Redistricting and the U.S. House of Representatives:
Illuminating Electoral Bias with the Brookes Method
by
Tony L. Hill
B.A., University of Minnesota, 1994
M.A., University of Minnesota, 2005
SUBMITTED TO THE DEPARTMENT OF POLITICAL SCIENCE IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
AT THE
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
AGHOVES
MA55ACHUSETT-7S INSiTUYE
OF TEC'HNOLOGY
JUNE 2010
JUN 2 9 2010
Copyright 0 2010 by Tony L. Hill. All rights reserved.
LIBRARIES
The author hereby grants to MIT permission to reproduce
and to distribute publicly paper and electronic
copies of this thesis document in whole or in part.
Signature of the Author:
Department of Political Science
March 31, 2010
Certified by:
Stephen D. Ansolabehere
Elting R. Morison Professor of Political Science
Thesis Supervisor
Accepted by:
Roger D. Petersen
Associate Professor of Political Science
Chair, Graduate Program Committee
ABSTRACT
This dissertation analyzes the effects of Congressional redistricting in the United States
using the Brookes Method, developed by R.H. Brookes, a New Zealand political scientist. The
Brookes Method disaggregates electoral bias into five separate components. My analysis shows
that the party winning control of the House benefits from the most prevalent component of bias
but that Democrats persistently benefit from the next most prevalent component. This means
that Republicans can never win the House as effectively as Democrats can.
The Brookes Method also informs electoral bias pertaining to racial gerrymandering.
Using the Brookes Method to evaluate three states with a history of using extremes in race-based
redistricting (Georgia, Louisiana, and North Carolina), my analysis reveals that these states have
disaffected Democrats in redistricting more sharply than has the country as a whole.
Using the Brookes Method to evaluate the small number of states using independent
commissions to carry out Congressional redistricting, I find that electoral bias in these states is
different from that found in states with legislative redistricting, and surprisingly, is often higher
in commission states. This suggests that commissions are in some cases not truly independent
and/or are merely fomenting a different kind of partisanship.
I propose a new formulation (the Hill Ratio) of a familiar compactness standard, the areaperimeter measure. Thousands of House districts across time are analyzed under the measure
and trends in compactness are noted. My analysis finds that districts in the U.S. have gotten
considerably less compact since the early 20th century, while districts in Canada are still more
compact than U.S. districts were even in the 1920s. Some of the states noted for their noncompact districts in the 2000s also had the least compact districts in the early 20th century.
Finally, compactness is used as a factor in voter knowledge. My analysis finds that
voters in non-compact districts are less likely to possess basic knowledge about their
representatives and districts than voters in compact districts, while knowledge about statewide
and national officeholders and party control is largely unaffected by the compactness of the
congressional district. This is true when analyzed both at the aggregate and the individual level.
These two measures of districting are harmonious with proportional representation ideals.
The Brookes Method is an explicit comparison of majoritarian seat outputs with a proportional
ideal. Compactness, in the words of Polsby & Popper, "tends to inhibit gerrymandering. By
inhibiting gerrymandering, in turn, one abets proportional representation ... by empirical
tendency."
TABLE OF CONTENTS
The achieving offair and effective representationfor all
citizens is concededly the basic aim of reapportionment.
(Reynolds v. Sims, 1964, 377 US 533 at 565-66)
1.
2.
3.
4.
5.
6.
7.
Introduction
An Overview of the Brookes Method
Partisanship and the Brookes Method
Racial Redistricting and the Brookes Method
The Use of Independent Boundary Commissions and the Brookes Method
Redistricting and Compactness: A New Formulation of a Measure
District Compactness and Voter Knowledge:
Information Heuristics Through Favorable Partitioning
Chapter 1 -- Introduction
The U.S. Constitution mandates that the seats of the House of Representatives be
reapportioned among the states following each decennial census. This is the main
purpose of the census. Some scholars use the term reapportionmentexclusively to refer
to this decennial reallocation of house seats and use redistrictingto refer to the process of
changing district lines within the individual states. However, most scholars use the terms
interchangeably. The decennial act has not involved Congress since the 1940 census, at
which time the formula for reapportionment was permanently changed. Before then,
Congress needed to pass a specific allocation for seats every 10 years. Sometimes they
increased the size of the House at this time, and the process of changing the allocation
from state to state was hard-fought and bitter. The formula for reapportionment has been
changed many times in history. One particular change happened after the 1880 census.
Alabama lost a seat under the new formula and protested that it should not lose seats even
as the House was increasing. This led to the formulation of a rule that no state should
lose seats when the House increases in size. This rule has been moot for almost 100
years, however, because from the time of and partly as a result of the 1910 census, the
House has remained fixed at 435. Following the 1920 census, Congress couldn't agree
on a formula for reapportionment, and as a result, no reapportionment was done from the
1920 census. Consequently, not very many states engaged in redistricting after the 1920
census. The size of the House became fixed at 435 by statute in 1929.
The formula for reapportionment works ever so slightly to the advantage of small
states. Rather than divide the total population of the states by 435, the first step in
reapportionment is to set aside one seat for each state, as required by the Constitution.
The next step might seem to be to divide the population of all states by the remaining
seats (385), but this is not so. Rather, an iterative method called the method of equal
proportions is used whereby for all 50 states, a priority score is calculated by dividing the
population of each state by the square root of the product of the number of seats it has
already been awarded times that number plus one. This obviously results in the most
populous state being highest priority. It is then given a second seat and its priority score
recalculated. Then the state with the new highest priority score (which might be the state
that was just awarded another seat) is given an additional seat. The process continues
through 385 iterations, whereupon all 435 seats have been awarded. This method does a
fairly close job of allocating the seats the same as if the population of the 50 states had
been divided by 435 and allocated proportionately to the states by population; however, it
differs crucially in who gets the last several seats made available by the method. In the
2000 census, using the divide-by-435 method allocated only 433 seats due to rounding.
The method of equal proportions allocated these last two seats to California and North
Carolina. Just missing an extra seat after the 2000 census was Utah. In essence, even
though there are 385 rounds of reapportionment with the method of equal proportions,
only the last 10 rounds contain any critical decisions. The algorithm for apportioning the
House has been unchanged since 1941 and has not been a significant source of
controversy in that time. Hayes points out, "This absence of discord is perhaps the one
bit of empirical evidence suggesting that algorithmic methods might really have
something to offer political science" (Hayes 1996).
With the House fixed at 435, there was no automatic change in the number of
states that prompted a large number of seats to redistrict. Obviously, when the House
was increasing in size at a time that most states were burgeoning in population, many
states saw a change in their seat count and therefore had a tremendous incentive to
redistrict. Since 1910, for most states, it has been a rarity for its seat count to change.
Thus, most of the impetus to redistrict from 1910 until the Reapportionment Revolution,
previously being a result of a need for the state to incorporate new House seats into its
polity, was gone. Many states felt no desire to redistrict after 1910, and with no
redistricting occurring in most places after the 1920 census, by the 1930 census,
redistricting was a thing of the past that had not happened in the time that the vast
majority of state legislators had been in office. Hanson notes that reapportionment
frequently did not happen "or was determined by practices more common to horse
trading than to political theory." (Hanson 1966, 38). Looking prospectively at the
question of redistricting, most of the incumbent state legislators saw that equal
redistricting would reduce the amount of representation their parts of the state had in
Congress. This is because state legislative district lines, similar to congressional district
lines, disproportionately favored rural areas over urban areas. This was not only because
of population standards that had been fixed since the 19th century, but it also resulted
from apportionment schemes in some states that were based on geography, such as the
requirement in Tennessee that every county have at least one senator and one
representative regardless of population. Since in most states a great many legislators then
represented rural areas - if not a majority, then at least a sizeable bloc in the legislature and the direction of the times (indeed, the entire period since the Civil War) saw rural
areas losing population relative to urban areas - it took no great amount of calculation for
them to realize that redistricting was bad for their interests and could lead to a vast
change in legislative (and congressional) control to urban voters. Malapportionment,
according to Hanson, "was preferred to a boundary adjustment which would reduce the
plurality of the incumbent, or even invite competition from the minority party in the
district." (Hanson 1966, 38).
As a result, many states ignored issues of malapportionment both on the state
legislative level and for congressional districts. In this period, the U.S. Supreme Court
rejected consideration of reapportionment cases on the grounds that they were patently
political decisions that should be mounted by the appropriate elected branches of
government. Some states got into the habit of decennial redistricting for both legislative
and congressional districts. Minnesota is one example of a state that began regular
redistricting in the early 20th century. However, there was no uniform standard for intrastate population deviation for districts even in the states that were fairly diligent about
carrying out regular decennial redistricting, and population inequality was thus a feature
even in states that engaged in regular redistricting.
After World War II, as suburbs grew all over the United States and inner cities
reached postwar peaks, the fundamental unfairness of the domination of the legislatures
of most states by rural majorities pre-ordained by redistricting decisions made in the late
19th or early 20th centuries began to demand a hearing. Lawsuits started appearing in
various courts around the country, but most courts followed the Colegrove precedent and
declined to get involved in what the Supreme Court had ruled was a political question.
The 1962 Supreme Court decisions in Baker v. Carr and its companions opened
the floodgates to challenges to districting in what were supposed to be equal-population
schemes all over the country. Left unanswered was whether these precedents required
both houses of a bicameral legislature to observe population-strict districting or whether
some parallel to the national bicameral legislature would be allowed. In the case of the
national government, the Great Compromise was to choose the U.S. House of
Representatives on the basis of population, while states would be equally represented in
the U.S. Senate. The answer to this came in 1966 with the decision in the case of
Reynolds v. Sims, in which the Supreme Court held that both houses of a bicameral
legislature had to be chosen on the basis of strict population standards. This led some to
question the necessity and even legitimacy of bicameral legislatures. Shortly after the
decision, a constitutional amendment was put forward by Sen. Everett McKinley Dirksen
(R-Ill.), the senate minority leader, which would have permitted states to use something
other than strict population to apportion one of its two legislative chambers. The Dirksen
Amendment failed, and efforts to pass it anew have attracted little support either among
the general public or in Congress since then. It ought to make sense that the Dirksen
Amendment would have no natural constituency after the 1972 elections, because the
state senators whose previous districting had been based on principles other than
population equality (guaranteed representation for counties, etc.) would have been
replaced by senators elected under population equality standards by then, at the latest.
Support for the Dirksen Amendment rested in large part upon its proponents being
elected based on non-population equality; it is no surprise that post-1966 state senators
elected under population equality standards are not eager to replace the system that
elected them with one that would surely not favor their interests, either political or
electoral. Since the 1970s, some of the same people who supported the Dirksen
Amendment in the 1960s have called for states to institute unicameral legislatures,
because there is no longer any difference between the representation standard for upper
houses and lower houses in the 49 states that have bicameral legislatures, making one of
the chambers, in the views of these advocates, redundant. In spite of this logic, there has
been no large movement and not a single successful one to abolish bicameralism in those
49 states. (Nebraska abolished its lower house in 1935, decades before rep-by-pop
became mandatory for both houses of state legislatures.) Unicameralism in the United
States is a difficult fight to wage, because state legislators have a vested interest in seeing
their offices retained. In most states, legislatures have much power to kill constitutional
amendments not favorable to their interests, if not directly then through the power to set
the ballot language or order or to couple the amendment with unpopular ones.
Reapportionment discussion for the remainder of the decade of the 1960s was
dominated by questions about malapportionment: Strictness of the standard for
population equality and questions about its implementations. In the 1970s and 1980s,
some states implemented districting schemes that relied upon extremes in population
equality. Of course, the only way to get congressional districts for an entire state within a
single-digit number of persons of each other was to haphazardly chop up districts at the
block level. (In 1992, 19 of 43 states had an intrastate variance of 10 persons or fewer;
the intrastate variance was over 1000 in only six states (Huckabee 2001, 4)). This had the
effect of making some voters politically impotent within their own districts outside of the
voting booth. Typically, political party groups are elected by enumeration units arrayed
using political boundaries. These might include cities, towns, counties, or election
precincts. When political parties assemble using districts formed using extreme methods
of redistricting that produce minute differences in population, it often happens that no one
is found to represent these corners of cities and counties from which most of the
population is in a different district. When someone is found to represent this tiny area,
they are often marginalized at political conventions. Typically convention delegates are
seated by county, so if at a particular convention, two counties each send 100 delegates
and a third county (of which only a small number of blocks is included) is allowed three
delegates, those three delegates generally find themselves sitting off on their own,
marginalized from the rest of the convention. For political precincting purposes, most
states require that everyone in a particular precinct have exactly the same ballot. This
means they have to be in the same legislative district, congressional district, city council
ward, etc. When courts and other redistricters break up existing electoral precincts to
allocate tiny areas to a different district, often these small groups of blocks have to form
their own election precinct. An election precinct, like any other body, needs a sufficient
population to function as a political unit. A precinct with 2000 residents is apt to form a
quite functioning unit within a political party, having a sufficient number of residents and
party activists to allow the recruitment of party officers, campaign volunteers, and the
like. A precinct with 50 residents is likely to have no one within its borders who wants to
volunteer for these things. This precinct is then apt to be ignored by the party or be
assigned as a colony to some neighboring precinct, whose officers are likely to deal with
the orphan precinct only when they are not occupied with the interests of their home
precinct. This tiny precinct is likely to have trouble finding even election judges within
its boundaries. Some such precincts might fall below the threshold of viability for having
an in-person election, and the small number of voters in these precincts might be told
they have to vote absentee or not at all. All of this suggests that taking liberties with
political subdivision boundaries to produce extreme equalities in districting has such an
effect on the voters shuffled across a subdivision boundary in order to accomplish it that
is akin to disfranchising them.
DISTRICT SIZE AND THE LACK OF THIRD PARTIES IN THE U.S.
Partisanship, participation, and district partitioning are all tied to demographics.
The United Kingdom and Canada both have viable third parties. Although these
parties are not viable in the sense of being able to form the government, they elect a
sufficient number of seats to have official party standing in the respective Houses of
Commons, and these elections occur with a degree of predictability with regard to
location that one can say that particular ridings (districts) in these countries are
dominated by those third parties. The population of U.S. house districts is a significant
cause of the lack of electoral competition for them. The U.K. has a population of some
61 million and 646 seats, or about 94,000 per seat. Canada has a population of about 33
million and 308 seats, or 107,000 per seat. The U.S. has 435 seats for a population of
about 300 million, or about 690,000 per seat. This means that U.S. House seats are
unfathomably large by British and Canadian standards, a factor around 7 times as great.
Having to build a district out of nearly 700,000 people means that districts are necessarily
heterogeneous on many dimensions; except in the very largest cities, having a 700,000
person district means that such a district cannot be overwhelmingly rich or poor, working
class or white collar, educated or uneducated. Cobbling together these large districts
means that a disaffected population group cannot easily dominate a congressional
election. U.S. districts take in vast swaths of a large portion of the state, necessarily
mixing relentlessly middle-class communities with whatever stratified districts lie in the
adjacency. Whereas in Canada or Britain, a riding might be centered upon a community
of 40-50,000 people, in the U.S. that same community is swallowed up with a dozen
other similar or dissimilar communities. This means that in Britain or Canada, districts
exist - created using perfectly natural boundaries with no need for gerrymandering - of
entirely poor communities (which find salience in left of center parties) or entirely upper
middle class communities (which do not burden left of center parties with the need to
campaign). In Britain and Canada, regional parties exist and thrive. Thus, the Scottish
National Party and the Bloc Qudbecois are able to grasp a foothold simply by virtue of
their numbers. When Quebec has 75 seats in the House of Commons, a regional party
capturing even a bare majority of them can produce a substantial caucus. (Indeed, the
Bloc Quebdcois formed the Official Opposition in Canada in 1993.) Were a U.S. state
with the equivalent population of Quebec or Scotland to elect a majority of members of a
similar regional party, they would send only four or five members to the House. This
small number of members would be marginalized by virtue of its innumeracy. Thus, no
one even bothers to organize a Virginia Party or a Colorado Party. The small Canadian
and British districts permit not only regions of the country to have their own parties, but
also allow the New Democratic Party and the Liberal Democrats, respectively, to thrive
in some very poor districts in inner cities and in some poorer rural districts and ridings
where organized labor still has a firm hand.
Third parties typically only have success when they are regionally based; the
Dixiecrats of the post-War American South are a cogent example. The Reform Party and
Bloc Quebecois were able to thrive by having nearly all of their support in particular
regions of Canada. Illustratively, the Progressive Conservative Party received nearly as
many votes in the 1993 Canadian general election as the Reform Party. But since the
Conservative Party was a national party whose support was spread across 295 ridings,
they came up victorious in only two ridings, while Reform won 52 seats. This is a classic
example of dissonance between seats and votes. That such outcomes fail to materialize
in the United States is mostly a function of so few voters casting their votes for third
parties; nevertheless, if we were to look at presidential voting instead of house seats, we
would find that several recent third-party presidential candidates (John Anderson in 1980;
Ross Perot in 1992; Ralph Nader in 2000) took a substantial enough share of the vote that
were their votes transposed to parliamentary elections in most of the world, they would
have amassed a fair number of seats. However, none of these candidates won even a
single electoral vote. In no small measure is this because they ran national campaigns;
their vote was spread (although not evenly) across the whole country. The last third party
candidate to actually win electoral votes was a politician whose appeal was calculatedly
regionally based: George C. Wallace, who won 46 electoral votes in 1968.
The size of the districts also affects the ability of challengers to appear. In a
94,000 population British riding, which might be won with 15 or 20,000 actual votes, and
which is centered upon a single suburb or small city, many people can fathom putting
together the electoral engine needed to win a party nomination or even capture the seat
outright. Such a campaign could have its genesis in the ordinary day to day activity
engaged in by many civic minded people.
Putting together the same campaign in a
district of 700,000 people is not merely the task multiplied by seven; no one has a social
network stretching across a wide swath of the state, encompassing many disparate
communities. In most cases, the only people who bother mounting such a campaign fall
into two groups: First, those who already have a leg up on the party machinery in some
way: state legislators or highly prominent local officials, who can make themselves
known to the party leaders of the congressional district through their legislative activity
or other political organizing. Second, those who are very wealthy and are willing to pay
to advertise in order to bring their profiles up to that of the first group. Both of those
groups consist of what are often called "rational actors" who are unwilling to mount
quixotic campaigns, given the sheer expense of campaigning in an electorate of over
300,000 voters. The size of American districts, then, keeps candidates and parties from
organizing what they see ex ante as hopeless causes; were the districts smaller and the
cost of campaigning much less, the evaluation of a particular district as a hopeless cause
would happen much less frequently. Duverger's Law says that a single-member plurality
system is apt to have a two party system. The small size of the electoral districts allow
British and Canadian third parties to exists as rumps in what are chiefly two-party
systems, or what some have taken to calling two-and-a-half-party systems. The large size
of U.S. congressional districts makes even the existence of third-party rumps unlikely.
Those British and Canadian third parties are also benefitted by the existence of different
two-party systems operating in different parts of the country. For example, in Quebec,
competition federally has primarily been between the Bloc Quebecois and the Liberals
during the recent period; provincially it is between the Parti Quebecois and the Quebec
Liberal Party, which is not affiliated with the federal Liberal Party. In fact, its leader
since 1998 was from 1993-1998 the leader of the national Progressive Conservative Party
(Hill 2002, xx-xxi). In Manitoba, the Liberal Party usually wins more of the province's
14 federal seats than it does of the 57 provincial ridings. Although winning consecutive
majorities in the provincial legislature, the New Democratic Party in recent years has not
won more than four of the 14 federal seats (Hill 2002, xvii). Were it not for the electoral
college, these kinds of patterns would be replicated all over the U.S., with different
parties taking up the state and national election mantles in the various states. According
to Stewart, the electoral college and the necessity it causes for having only two national
candidates in the presidential election prevents different two-party systems from forming
in the United States (Stewart, Analyzing Congress 2001, 244). Another way this
manifests itself is in the harmonization of the political party structures to mount
harmonized elections. In most of the U.S., state and national elections are held on the
same day. The regular schedule mandates an orderliness for the political parties to
organize in tandem for both elections, with the same delegates choosing the party's state
and federal candidates. In Canada, elections for national and provincial offices are held
separately, and the parties are organized in completely different fashion; it would not be
unusual for a person to be a member of the Conservative Party at the federal level and a
member of the Liberal Party at the provincial level. In most of the U.S., that is unheard
of. This harmonization of the parties works against diversity of political organizing and
the formation of third parties (Hill 2002, 9).
ROLE OF CONGRESSIONAL BUREACRACY IN FOSTERING REELECTION
Mayhew famously showed that the number of incumbents who win their seats by
only a small margin was in 1972 about half what it had been in 1956 (Mayhew 1974,
305). This was over a period where the Democratic Party maintained absolute
dominance of the House of Representatives. Campbell argues the persistence of the
Democratic House majority "may be the most important single feature of American
politics in the second half' of the 20th century (Campbell 1996, xviii). Fiorina has
alleged that the actual constituency service performed by incumbents leads to their high
re-election rates (Fiorina 1977). Born confirmed Fiorina's work by finding that House
members first elected between 1966 and 1978 were better able to fend off challenges
through their use of the advantages of office (Born 1979, 816). Incumbents are able to
build up a huge "personal vote," which is defined as people voting for the incumbent
specifically rather than merely voting for a party. Advantage of incumbency in the
United States is quite large, estimated generally between seven and thirteen percentage
points (Cover 1977, Payne 1980, Collie 1981, Krashinsky and Milne 1993, Ansolabehere,
Snyder and Stewart 2000), but if nothing else, it can make the difference between a
person winning or losing a very close contest (Erikson 1971, 404-405). King and
Gelman argue that every other measure of incumbency advantage is flawed, but
nevertheless show that their (allegedly) flawless measure yields approximately the same
results as the flawed ones (King and Gelman 1991, 1158). Garand and Gross find that
the advantage of incumbency began in the 1890s and not as recently as most scholars
show (Garand and Gross 1984, 21). Ansolabehere and Gerber point out that over the
period of large Democratic dominance of the House, Democrats tended to retire at lesser
rates than Republicans (Ansolabehere and Gerber 1997, 162). Prior makes the point that
the increase in incumbency advantage coincided with the development of television, and
through the power of this new medium, incumbents became better able to distinguish
themselves from challengers (Prior 2006). Incumbency advantage is much smaller in
Britain and Canada, where candidates campaign less in their own person and more as
agents for the national (or regional) party. This reinforces the tendency of partisans in
countries without strong incumbency to retire at times when they perceive bad fortunes
for their party and not take the chance that their personal electoral fortunes will help them
prevail (Ansolabehere and Gerber 1997, 174). Other scholars maintain that the huge
likelihood of an incumbent being reelected tends to dissuade quality challengers (i.e.,
persons with elective experience)- from running in the first place. According to Levitt and
Wolfram, "Virtually all of the growth in the incumbency advantage since the 1960s
appears to be attributable to a reduction in the relative quality of challengers." (Levitt and
Wolfram 1997, 56). The advantage of incumbency is well documented in the U.S. It is
theorized that some voters support candidates of the other party because of constituency
service they performed that benefitted the communities of interest to these voters, even if
the larger legislative stance of the candidates was antithetical to the voters' preferred
choices. The constituency service theory has its critics. Levitt and Wolfram find that
little of the incumbency advantage is attributable to incumbents' use of the benefits of
office. (Levitt and Wolfram 1997, 57). Cox & Katz maintain that much advantage of
incumbency stems not from the adroit use of campaign or official resources of Members
but from their prudence in knowing when to retire from Congress. (Cox and Katz 2002,
6). Ansolabehere & Snyder state it this way: "Periodic redistricting - more than any
other force in contemporary American politics - turns incumbents out of office and
brings in new people." (Ansolabehere and Snyder 2008, 264)
However, the constituency service theory has other flaws, because
parliamentarians perform constituency service in other countries as well, even those
where the advantage of incumbency is well below the U.S. level. Furthermore, it is well
understood that constituency service is a basic function of an elected official, and that any
elected official ought to be able to do it with competence. Thus, it is irrational for a voter
to support someone for their helpfulness at constituency service when their political
ideology and stance are adverse to the voter; to anyone for whom ideology and political
issues are important, it would be more rational to support someone who shared them,
even if that candidate initially might not be as proficient at performing constituent
service. On the other hand, the types of constituency service that are most efficacious at
generating constituency-level support are not of the casework variety, but those of
marshalling federal largesse for the district. This is obviously a type of constituent
service work that a member ought be expected to get better at with more experience, and
whether a member is part of the congressional majority or minority is pivotal in the
ability to get district projects funded. The question of whether voters explicitly consider
this when choosing to reject an incumbent in favor of a challenger has not been
specifically studied.
More salient to the ability of incumbents to rack up votes is their ability to
establish themselves as a brand. Nearly all members issue periodic newsletters, mailed at
government expense under the franking privilege to all postal patrons in the district.
These newsletters are helpfully color-coordinated to the member's own campaign
literature. Even voters who discard these iMailings without reading them are apt to notice
the name on them and take some sort of cue from the colors and logos. Then, on election
day, voters are more familiar with the incumbent's name than they are with any
challengers. The process has been likened to consumer purchases of everyday
commodities like canned vegetables. Although most consumers are not connoisseurs of
canned corn, they recognize brand names and logotypes from advertising and other
media. Therefore, Green Giant or Libby capture a higher market share in spite of having
a higher cost to consumers than other brands less familiar to them, even though the
product may be indistinguishable. Simply by being an incumbent, even in one of the few
remaining truly competitive house districts, the incumbent is able to establish himself or
herself as a brand, and this translates to a share of the vote not readily available to the
challenger.
Some argue that this is not a good thing for democratic government. As Cox &
Katz put it, "Whenever the resources of public office are used to insulate individual
politicians from electoral risk, their accountability to their constituents is weakened.
Whenever government resources are used to entrench a single party in government, its
accountability to the public at large is weakened" (Cox and Katz 2002, 7).
The question of whether incumbency is key is a controversial one. Baker alleges
that the advent of computers made it possible for partisans to create "equipopulous
districts for perpetual partisan advantage" (Baker, The Unfinished Reapportionment
Revolution 1990, 23). However, Ansolabehere and Snyder point out that incumbency
rates have increased for all kinds of offices since 1940, not merely those that are subject
to periodic redistricting (Ansolabehere and Snyder 2008, 267). This, they declare,
means that incumbency is a phenomenon all to itself and is not as closely tied to
redistricting as those who place weight in huge redistricting effects believe. However,
McDonald posits that a decline in district competitiveness occurred between 2000 and
2002 and that redistricting was apt to be a factor. (McDonald 2006, 100).
Chapter 2 - An Overview of the Brookes Method
The Brookes method of determining electoral bias is a standard in Westminster
democracies besides the United States. First put forward by R.H. Brookes in a 1959
treatment of New Zealand (Brookes 1959), the Brookes method calculates the difference
between the results of an election and what the outcome of that election would have been
using a pure proportional representation system on the same votes. Brookes's definition
of bias is the number of additional seats the second-place party would have received if it
had the same proportion of the two-party vote as the leading party, assuming a uniform
national swing (Rossiter, Johnston and Pattie 1997, 468). It has become a standard
approach to electoral bias in New Zealand, Australia, and the United Kingdom (Gudgin
and Taylor 1974, Gudgin and Taylor 1980, Johnston 1976, Johnston, Pattie and Dorling,
et al. 2001). Two Brookes treatments of Canada are known (Siaroff 2003, Hill 2004).
Only one brief treatment of the U.S. House has been published (Rossiter, Johnston and
Pattie 1997).
The Brookes method is generally compared favorably to the cube law approach.
Kendall & Stewart published their law of cubic proportions in 1950 (Kendall and Stewart
1950), but it has been a source of controversy in political science. Taagapera called it
"the only political science law that looks like a physics law," (Taagapera 1973) but Tufte
dismissed it as "British political folklore" and demonstrated that it only worked in one of
the six real-world instances he used to test it (Tufte 1973). March stated that its
significance was not due to its mathematical or aesthetic qualities but because it had been
successful in its explanations of British elections in the previous two decades (March
1957, 525). Gudgin and Taylor make the point that even if it does not qualify as a
physics law, it can possess explanatory power that can be refined into something more
useful (Gudgin and Taylor 1979, 79). Rossiter, Johnston, and Pattie criticize Gudgin &
Taylor's approach of evaluating principally the national share of the vote of the main
parties. They endorse the Brookes method because "it permits the components of bias to
be estimated in a straightforward and readily interpretable fashion" (Rossiter, Johnston
and Pattie 1997, 468). Brady and Grofman examined electoral bias in the period from
1850 to 1980 and found that the swing ratio was nearly a linear function of the percentage
of seats that were competitive (Brady and Grofman 1991, 261). Rae notes that the cube
law demonstrates the disproportionality in plurality elections as compared with majority
or proportional models (Rae 1971, 27). Campbell argues that swing ratios do not work
because they are grounded on a faulty premise, namely that the election system is neutral
(Campbell 1996, 54). This will be explored further herein.
Seats in legislative bodies do not always result from elections in proportion to the
popular vote for the parties. In fact, they rarely result proportionally to the party vote.
Political scientists often study this problem vis-a-vis the tendency of certain electoral
systems and certain party systems to produce a substantially disproportionate result.
Scholars have often looked with curiosity to the U.S. system for electing the U.S. House
of Representatives. Although the U.S. uses the same electoral system as the U.K. and
Canada and a number of West Indies republics (until rather recently Australia and New
Zealand also used this system), the U.S. has often served as an exceptional case in several
regards. First, the Democratic Party maintained a long dominance of the U.S. House of
Representatives through a period when the presidency was quite competitive. Second,
there have been no long-lived third parties in the United States. Third, most House
contests are characterized by a lack of competition. Most scholars believe that
competition is a good thing and that the political structure and processes should foster it.
"The uncompetitiveness of individual congressional elections undermines the electoral
process. ... If elections are uncompetitive, the electoral incentive for good representation
is missing, at least in the short run" (Campbell 1996, 5).
The tendency of single member plurality (SMP) voting in legislative elections is
for the majority party to receive an exaggerated share of seats relative to its popular vote
(Dixon 1982, 9). Backstrom, Robins & Eller term the pattern by which the majority seat
share is elevated relative to its vote share the balloon effect. They argue that the balloon
effect happens because a majority win by a party will have its greatest impact on
marginal districts. They caution against using a purely proportional standard for
evaluating statewide results for the reason that such an analysis deprives majorities of the
extra seats they win through the balloon effect (Backstrom, Robins and Eller 1990, 162).
Of course, one of the main reasons scholars use proportional standards for evaluating
polity-wide voting is simply because the balloon effect is one of the outcomes of
majoritarian electoral systems sought to be analyzed, and reformers see it as an excess to
be minimized through reform. Ansolabehere and Snyder reject the idea that districting is
a cause of manufactured majorities. They attribute the paucity of competitive seats for
the U.S. House to incumbency rather than redistricting:
Incumbents win by larger vote margins than they would
ever receive in open seat races, and open seat races are as
competitive as ever. When an incumbent runs, even if he
or she represents a district that leans toward the opposite
party, the advantages that come from the office itself
present a formidable hurdle to any challenger. Incumbents
win at very high rates for reasons other than districting,
such as campaign spending, constituent service, and simple
name recognition. (Ansolabehere and Snyder 2008, 271)
Browning and King posit a model that terms anything (excluding winner-take-all)
differing from a 1:1 relationship between seats and votes a "majoritarian" type (Browning
and King 1987, 312). This formalizes the notion of Backstrom and others that a majority
winner is apt to take a balloon effect in winning a higher proportion of seats than their
proportion of votes. As O'Rourke notes, "geographical, winner-take-all districting
virtually guarantees divergence between a party's overall proportion of the statewide vote
and its proportion of legislative seats" (O'Rourke 1980, 55).
The inputs of the Brookes method are mathematical and require minimal
assessments on the part of the researcher. The actual inputs and equations are given in
Appendix 2.
In function, one has to choose a party to serve as the object party. In a Brookes
analysis, positive numbers represent a bias in favor of this party and negative numbers
represent a bias against this party. In the U.S. case, it is most useful to treat the
Democratic Party as the object party because it was the majority party for most of our
recent history, and because there was a time in which Democrats were routinely
unopposed in large numbers of seats, primarily in the South. This was not true for the
Republicans in the same time period.
The Brookes method disaggregates the bias into five separate components. First,
there is the Gerrymandering effect (G). This is an inapt name, because it truly measures
maldistribution of a party's votes and not only those that are the result of intentional
gerrymanders. Lately, practitioners of the Brookes method have taken to calling the G
component Efficiency bias (Johnston 2006).
The second component is Constituency Size Variation effect (CSV). This
measure takes into account the aggregate difference between the average size of the
electorate in seats where a party wins versus seats where it doesn't. This is very
dependent on the actual vote in a district, not its population. In this sense, two districts
with the same population as of the most recent census might augur very differently in the
CSV measure. This is a real issue in the United States, where many inner-city seats have
highly non-competitive elections captured by Democrats. Some scholars note a
difference in analyzing elections where the voter rather than the district is the critical
unit: According to Campbell, "Counting each district equally essentially counts
individual voters quite unequally." (Campbell 1996, 84). In this chapter, the registered
electorate is simply that. In most Westminster democracies, voter registration is not as
voluntary as in the United States, and in practice the registered electorate is much larger.
This is because of the traditional practice of election officials going door to door to
register voters in advance of elections, something akin to a census. Only in recent years
have these canvasses been replaced by a U.S.-style registration system. (Canada went to
a permanent registry of voters in 2000.) Thus, in most of the countries where Brookes
analyses have been conventional, the registered electorate is closer to what is usually
called voting age population. Of course, using VAP in place of registered electorate
would also pose problems for an analysis, because political scientists have demonstrated
that there are many different conceptions of VAP which do not adequately cover who is
actually eligible to vote, excluding, among others, prisoners, felons, institutionalized
persons, and illegal aliens (McDonald and Popkin 2001). For our purposes, estimating
the voting eligible population, as McDonald and Popkin propose, is even more
problematical than using some of the fanciful statements of registered electorate, as
discussed below. Furthermore, even using VAP or VEP in place of registered electorate
does not totally assuage the problem of CSV, because constituencies are created on the
basis of total population, not VAP, so even if VAP were known with certainty, there
would still be CSV bias. Campbell makes the point that VAP or actual electorate is valid
for a determinant of population for redistricting in the U.S. notwithstanding the extreme
tendency to use actual population for the same; "What is politically important is whether
one voter's vote is worth as much as another's, not how many bystanders are nearby."
(Campbell 1996, 212).
The third component is Abstentions (A). It looks at the effect of potential voters
who do not participate and allocates a share of the seats on the basis of where the parties
stand relative to these pools of voters. The fourth component is Third Party Votes (TPV)
and is essentially the differential effect of third party votes upon the major party. That is,
a disproportionate share of seats is assessed when third party votes occur in one or the
other party's area of strength. It is assumed that the third party votes are
disproportionately affecting the dominant party. (That is, if a large number of third party
votes are cast in an otherwise Democratic area, that Democrats are being affected more
than Republicans.) Of course, TPV is of little consequence in the United States where
there are no viable third parties and where very few people cast votes for minor parties
and candidates.
Finally, the Third Party Wins effect (TPW) is the number of seats the object party
wins minus the number of seats in which it leads the other major party, minus the number
of seats the other major party wins minus the number of seats in which it leads the object
party. Thus, the TPW bias might be zero even if more than one seat is captured by
minor party candidates. That is because the bias is assigned based on which major party
leads. So in the instance where two minor candidates won, one in a district where the
Democrat leads, and one in a district where the Republican leads, the bias would be zero.
In recent history, the only third party candidate to regularly win is Socialist Bernard
Sanders of Vermont, now a senator. In function, he is a Democrat. He caucuses with the
Democrats and there is often no Democrat running against him. His wins therefore
translate to a bias against the Republicans, because they are otherwise leading in
Vermont. Simplified,
(n-n)-(n-(n+l))= 1
Thus, holding everything else equal, the Socialist win in Vermont translates to a
bias in favor of the Democrats. Actually, this is no different from simply construing
Sanders as a Democrat.
Problems of using the Brookes method in the United States
The lack of results for unopposed candidates. Five states, all in the South, by
law, did not report any vote totals for unopposed candidates during the period covered.
These states are Arkansas, Florida, Kentucky, Louisiana, and Oklahoma. This presents
problems comparing votes in these states directly with other states. Other states report
the number of votes for unopposed candidates, which are generally less than they would
be if that person had an opponent. Of course, whether the votes are reported for such a
candidate or not, they are generally not comparable to votes in contested races.
The problem of unopposed and underopposed candidates generally. As a
practical matter, in the other Westminster democracies, there are no unopposed
candidates. In Canada and the United Kingdom, the major parties as a matter of principle
contest every seat in the country, regardless of how poor their electoral fortunes may be
for a particular seat. This means that every voter in the country sees the names of all
major parties on the ballot. This is not the case in the United States. Some candidates
simply lack opposition. In many other districts, a semi-official candidate runs under the
banner of the losing party. Many political scientists lament this manifestation of the lack
of true competition in most congressional districts. "The voter ... has not been offered a
full and fair choice between the parties." (Campbell 1996, 47). These candidates have
fulfilled some formal requirement for serving as the party candidate, such as winning the
primary election or securing the party nomination at a convention (which might amount
to nothing but a legal formality in some instances) but have no official connection to the
party and receive little or no resources from the party in terms of money or organization.
Such candidates generally do poorly in the election. The results in these races exaggerate
the size of the winner's victory and reduce it as a meaningful estimator of partisanship in
the electorate vis-i-vis more accurate measures of partisanship in the district, such as
voting for president or for some state-level office such as secretary of state. This reveals
a defect in using the Brookes method for evaluating elections to the U.S. House. This
problem could be overcome by further reducing the scope of the analysis, using some
measure of "serious" opposition. However, this would be a most laborious and
information-intensive task which would thereby defeat the simplicity of the Brookes
approach in simply digesting aggregate electoral returns.
In yet other districts, the dominant party candidate is opposed only by strong (i.e.,
mildly competitive) minor party candidates (including independents) or only by noncompetitive minor party candidates. While these candidates are technically not
"unopposed," the race lacks the dynamic that would be present if there were even a
nominal major-party opponent. So these races are underopposed.
In considering U.S. House races under these circumstances, three separate
thresholds for analysis were considered: Excluding nothing (all 435 seats in every
election regardless of whether results were reported); excluding only those wholly
unopposed (which eliminates the problem of the states which don't report vote totals for
unopposed candidates); or excluding any race in which either major party is
unrepresented. The last was used in these analyses. Obviously, the first threshold has the
effect of diluting the results by including in the denominator some districts which
contribute nothing to the numerator. The second threshold eliminates only the most
egregious cases from that statistical distortion. The third threshold extends somewhat in
the direction of considering only seriously contested seats without employing the
subjective analytic techniques which a manipulation would require while still utilizing
the automatic data analysis that the Brookes method provides.
Relationship to the personal vote. The advantage of incumbency in the United
States is huge, generally estimated at 11 to 14 percentage points. Incumbent members of
the U.S. House have many resources at their disposal to help them build personal bases of
support in their districts, including their constituency offices, free franking privileges, and
access to the media. Ansolabehere, Snyder, and Stewart estimate the personal vote to be
approximately four percentage points in the time period analyzed herein (Ansolabehere,
Snyder and Stewart 2000, 11). This personal support gives them the ability to transcend
momentary electoral swings affecting their party generally or their presidential leader.
Ansolabehere, Brady, and Fiorina point out that incumbents are able to adjust the amount
of attention they pay to their districts based on their perceived level of electoral threat.
(Ansolabehere, Brady and Fiorina 1992, 27) The personal vote and the corresponding
advantage of incumbency are smaller in the other Westminster democracies and perhaps
negligible in one or two. The Brookes method is best at measuring pure partisanship,
exemplified by candidates who are complete ciphers in their districts and function only as
agents of the party and its leader. However, there is probably not such a thing as pure
partisanship even in the parliamentary democracies for which it was developed. With
electoral results distributed normally, some districts inevitably fall so close to the pivot
point that any factor can be alleged to have made the difference. Furthermore, the
personal vote and the advantage of incumbency pertain to both Democrats and
Republicans. Also, Ansolabehere, Snyder, and Stewart note that the personal vote is a
small part of the total advantage of incumbency and an even smaller part of the growth in
incumbency advantage in recent decades. They find that incumbents can expect to do 45 points worse in areas new to them through redistricting as compared with areas they
previously had represented (Ansolabehere, Snyder and Stewart 2000, 111).
Ansolabehere and Snyder demonstrate that redistricting lowers the personal vote of
incumbents and leads to their defeat in greater numbers (Ansolabehere and Snyder 2008,
266).
It is certainly a limitation of the Brookes method in studying the United States
that it does not take into account incumbency specifically; certainly the most salient
aspect of the Democratic hegemony from 1974 to 1994 is that the party and its
incumbents were able to denationalize elections and invest as much of their capital as
they could in building up incumbents as local brands rather than as agents of the party.
The 1994 election, however, underscores the usefulness of the Brookes approach insofar
as the Republican Party under Newt Gingrich did the maximum they could to nationalize
the election on their terms. The sudden movement of the bias measures in 1994
underscores the realigning nature of this election rather than showing it as a blip on a
panorama of continued Democratic dominance. Certainly the realignment could not have
been predicted solely from the bias numbers seen for 1990 and 1992 reported in table 1.
The New York multi-party system. New York has a unique system whereby
candidates are allowed to run on more than one party ticket. This poses a problem for the
Brookes analysis. For example, the Democratic candidate might also be the candidate of
the Liberal Party, the Right-to-Life Party or even the Republican Party. This confounds a
Brookes analysis which takes partisan voting as inputs. However, a vote for the
candidate of one of the major parties under the aegis of some minor party can still be
construed as a vote for the major-party candidate for our purposes. Thus, votes for major
party candidates on smaller party tickets are combined with the major parties. This
includes Republican votes being counted for Democratic incumbents when they carry
both labels. When minor parties run candidates other than the major party candidates,
these are included in Third Party Votes, as they would be in any other state.
Chapter 3 - Partisanship and the Brookes Method
It requires no special genius to recognize the political
consequences of drawing a district line along one street
rather than another. It is not only obvious, but absolutely
unavoidable, that the location and shape of districts may
well determine the political complexion of the area. District
lines are rarely neutral phenomena. They can well
determine what district will be predominantly Democratic
or predominantly Republican, or make a close race likely.
Redistricting may pit incumbents against one another or
make very difficult the election of the most experienced
legislator. The reality is that districting inevitably has and is
intended to have substantial political consequences.
(White, J., for the majority, Gaffney v. Cummings, 1973,
412 US 735 at 753)
Introduction
Redistricting first became a widespread and salient feature of the American
political system in U.S. House elections in the 1960s following a series of court decisions
beginning with Baker v. Carr(1962) holding that reapportionment meant not only
allocation of the seats among the states, as specified in the Constitution, but also equal
districting within the states. Political science has disagreed on the political salience of
redistricting, with a large body of literature holding that redistricting is largely irrelevant
to political outcomes. (Bicker 1971, Bullock 1975, Ferejohn 1977, O'Rourke 1980).
Other scholars have found that redistricting holds important effects. (Mayhew 1971,
Cain 1985).
This chapter seeks to demonstrate that the Brookes method can be used to show
that not only is redistricting not irrelevant to electoral outcomes, but that redistricting
affects different forms of electoral bias in sometimes contradictory ways.
Ideas about representative government being truly representative of the people
antedate the U.S. Constitution. One of the seminal acts of the Continental Congress was
the Northwest Ordinance of 1787, providing government for the vast acreage west of the
original 13 colonies which was acceded to American dominion under the Treaty of Paris
of 1783. The act provided that inhabitants, "shall always be entitled to ... proportionate
representation of the people in the legislature ... " (I. Stat. 50-2 (1787): An Act to Provide
for the Government of the Territory Northwest of the River Ohio, cited in (Baker, The
Unfinished Reapportionment Revolution 1990, 12))
With regard to the process in the several states, the only thing we can say with
certainly is that the states with only one seat have no decisions to make about districting.
In every other state, some decision or series of decisions has to be made about how to
establish the district boundaries. It would be less of an exaggeration to say no two states
do it the same way than to delineate a typology for how redistricting is done. In general,
however, we can divide states into two groups: those in which the legislature is the
primary agent of redistricting and those in which some other entity is the primary agent
of redistricting.
Within the states with legislative redistricting, the next demarcation is political
rather than structural: It is the division between states where a single party controls
redistricting versus those where more than one party has control of redistricting. The
degree of the second party's control exists on a continuum: It may be as little as enough
votes in one chamber to prevent override of a governor's veto, or it may be that the two
parties are equal in terms of each one controlling a chamber. North Carolina is the only
state where the governor does not have a veto over actions of the legislature.
In the states with non-legislative redistricting, it is quite true that no two states do
it the same way. The various methods of districting in the states with independent
redistricting are described at length in chapter 4.
In any state, it often happens that the default redistricter is the courts. Only rarely
does this happen by design; usually courts get involved because the designated
redistricter fails to act in time. Courts in some states have been very aggressive at
enforcing the demands of plaintiffs. It has happened that courts have issued injunctions
against the use of existing districts as soon as the day after the official census results are
released, providing the needed proof that disparity exists between districts. In any other
type of case, court action would be discouraged as unripe. However, the recent history of
obstruction by legislatures and the frequency with which courts have had to become
involved at the very last minute before irremediable decisions about elections were
imminent has led courts to become more aggressive players in redistricting at the outset.
The tendency of this is to make courts more active players in redistricting decisions rather
than to prod legislatures to do in a more timely fashion what the courts would have them
do. In the contemporary United States, it should be assumed that the courts will act if no
one else acts first. It is well known that courts are apt to be more aggressive in their
redistricting than legislatures are. This is because legislators often have as their first aim
in redistricting the protection of incumbents. By making "radical reapportionment" (as
some scholars have termed it) the default in case nothing happens, legislators have
incentive to get redistricting done on their own. As Ward described it, "radical
reapportionment overrides a legislature's instinct for leaving the districts alone yet leaves
unchecked the normal disposition of legislators to consult their own interests..." (Ward
1970).
This was not always so. Before the reapportionment revolution, when states came
to an impasse on redistricting, no changes were made to district lines. If the state gained
seats as a result of the apportionment of seats in the U.S. House, the additional seats were
elected at large by the entire electorate of the state. If the state lost seats through
reapportionment, then all of the seats in the state were elected at large until the legislature
could agree to a districting plan. This ended with revision of the law governing the
election of the House of Representatives in 1967. (It could be argued the policy was
already headed toward invalidation by courts before the law was changed.) Since 1968,
all representatives have been required to be elected in single-member districts. Public
Law 90-196, enacted December 14, 1967, governing election of Representatives in
Congress reads in part, "In each State entitled ... to more than one Representative ...
there shall be established by law a number of districts equal to the number of
Representatives to which such State is so entitled, and Representatives shall be elected
only from districts so established, no district to elect more than one Representative..."
(United States Code, Title 2, Chapter 1, Section 2(c)). This renders the term
"representative at large" obsolete, since the only states now allowed to elect members "at
large" are those having but a single district. Before the 1967 change, Hawaii was still
electing its two members at large, a practice New Mexico had only ceased as of the 1966
election. Thus, even though there have been 435 representatives in Congress since the
election of 1912, there have been 435 districts only since the 1968 election.
All but three states elect their House members through single-member plurality
(SMP) election. This means that the person who receives the largest number of votes in
the election wins. Campbell says, "Arguably the two most important features of the
House electoral system, single-member districts and the plurality-rule for deciding
election winners, are not even mentioned in the U.S. Constitution." (Campbell 1996, 18).
Three states, all in the South - Georgia, Louisiana, and South Carolina - use singlemember majority (SMM) for elections. Under SMM, if a candidate wins a plurality of
the vote but not a majority, they must face a runoff election with the second-leading votegetter in order to determine the SMM winner. Louisiana uses a completely different
election system than the other 49 states (and even different from the other two states that
use SMM). In Louisiana, all candidates compete in the same primary, regardless of
party. Any candidate who wins a majority in the primary is automatically elected, and
the winner runs unopposed in the general election. In essence, the general election (the
same day the other 49 states elect their representatives) serves as the runoff election in
Louisiana. Georgia has had two well known runoff elections for the U.S. Senate in the
recent period. In 1992, incumbent Senator Wyche Fowler, a Democrat, won a plurality
of the vote on election day but did not win a majority. He competed in the runoff against
Republican Paul Coverdell, who won. Voter turnout was much lower in the runoff
election than on general election day. In 2008, a similar matchup occurred when
incumbent Senator Saxby Chambliss, a Republican, fell just short of a majority on
general election day. He defeated Democrat Jim Martin handily in the runoff election.
(President-Elect Barack Obama chose not to campaign for Martin on the grounds that if
Martin lost, it would contribute to a perception of weakness by Obama out of proportion
to the significance of the senate seat loss.) Some political scientists have argued that
having a second election in which far fewer people vote is antithetical to democratic
values.
The alternative to a majoritarian system is proportional representation. The U.S.
has no tradition of proportional representation, and it is only used in scattered places in
the U.S., despite being the most common electoral system in Europe and parts of Asia.
Incorporating a proportional system necessarily means moving away from single-member
districts. As Dixon puts it, "the ideal of proportionate representation of parties does not
dovetail well with an election system based on the use of geographic legislative districts
and the plurality rule within each district." The two-party system, in turn, is dependent
upon the existence of single-member districts (Dixon 1982, 9). Some scholars,
concerned about the effects proportional representation is apt to have on the two-party
system have moved to pre-empt any discussion of particular standards of fairness in
redistricting, believing they lead inevitably to a proportional system. According to some
scholars, a test for partisan gerrymandering is a covert argument for proportional
representation (Schuck 1990, 240). Levinson says as much when he notes that by
putting emphasis on the fairest popular modes of election, advocates are inexorably (and
in most cases, unintentionally) pushing the courts toward proportional modes of
representation (Levinson 1995, passim). He characterizes the reapportionment cases of
the 1960s as "a radical intervention into long-established modes of apportioning
legislative seats" (Levinson 1995, 259) and suggests that in the future, an activist court is
apt to use the same type of activism to impose a proportional representation scheme on
the country. If these scholars are correct, the U.S. has been on track to a legal standard
prohibiting partisan gerrymandering for some time (but not yet arrived). Decades ago,
Justice Stevens argued, "Political gerrymandering is one form of 'vote dilution' that is
proscribed by the Equal Protection Clause." (Stevens, J., concurring, Karcher v. Daggett,
1983, 462 US 725).
In Justice O'Connor's worldview, excesses in redistricting, including
gerrymandering, are nothing more than perquisites of the majority party, one of the spoils
upon gaining office (Backstrom, Robins and Eller 1990, 148). Backstrom, Robins &
Eller caution, however, "Popular acquiescence in and support for laws of a democracy"
are dependent upon "the faith on the part of the losers in this legislative election that they
have a fair chance to be the victors in the next." (Ibid.)
According to Baker, the 1964 reapportionment cases suggest "at least three
interrelated components: political inequality of individual voters; majority rule rather
than oligarchy; representative institutions that can reflect significant shifts in public
opinion." (Baker, The Unfinished Reapportionment Revolution 1990, 11). Baker argues,
"representative institutions should not be static, but rather should be responsive to shifts
in public opinion." (Baker, The Unfinished Reapportionment Revolution 1990, 13).
Niemi & Deegan argue that since seats in a legislature should change as vote totals
change, "there is a need to incorporate the partisan division of the vote into the criteria
for fair districting." (Niemi and Deegan 1978, 1304). As Niemi puts it, "It hardly seems
fair or consistent with democratic principles to have a districting plan such that changes
in seats are heavily insulated from changes in votes." (Niemi 1982, 35).
Dixon argues that the most important function of the redistricting body is "to test
and discard unfair plans and not for the purpose of manufacturing artificial majorities in
the legislative assembly." (Dixon 1982, 11). Backstrom, Robins & Eller devise a test for
doing exactly that. They assert a standard based on a previous election for evaluating
redistricting plans. They note that persons involved in districting use exactly such a
partisan baseline for doing their work; therefore, scholars and courts evaluating plans
should do no less. Their process involves evaluating majority voting strength as the
primary indicator of fairness; if a plan exceeds this index, it constitutes a partisan
gerrymander (Backstrom, Robins and Eller 1990, 160). Schuck criticizes their method,
claiming, "a base race, to be a useful construct, must be on in which the effects of issues,
candidates' personalities, unusual party effort, and other contingent factors deemed
irrelevant to the parties' 'true' strength are minimal." (Schuck 1990, 241). But this is
exactly what Backstrom et al mean: They advocate use of election for secretary of state
or agriculture commissioner or some other fairly inert office for which partisanship is key
to the outcome. Where such a contest is lacking, Backstrom et al urge use of the next
most candidate-neutral contest, or where no such contest is present, an index constructed
from one or more other more prominent contests (Backstrom, Robins and Eller 1990,
160-61).
Backstrom points out that the only truly national base race is the vote for
president, which is inadvisable to be used because of the huge dynamics of the campaign,
incumbency, and particular characteristics of the candidates involved. (Backstrom, The
Practice and Effect of Redistricting 1982, 353). Backstrom et al were responding to a
line of argument that called for the establishment of standards in electoral matters. "In
the absence of any clear criterion as to what a fair result should be, it is not possible to
use the results as a criterion of unfairness in the establishment of electoral districts."
(Vickrey 1961, 106).
Many scholars have attempted to establish a set of criteria for fair districting.
Backstrom speculates that the reason the courts have over time been unwilling to engage
in litigation over reapportionment is because of "the lack of a precise definition of
partisan gerrymandering and the lack of recognized measures of it." (Backstrom,
Problems of Implementing Redistricting 1982, 45). Thus, definitions of the problem and
cogent solutions for it are important in a world where some people deny that a problem
even exists. One of the most basic sets of criteria is advocated by Niemi:
"Four criteria for fairness
1. Neutrality: A districting plan that treats all parties alike in allocating seats per
given vote totals is said to be neutral.
2. Range of responsiveness: The range of responsiveness of a districting plan is
defined as the percentage range of the total popular vote (for the entire state) over
which seats change from one party to the other. In other words, the low end of
the range is the minimum percentage of the total vote required to win at least one
seat, while the upper end is the minimum percentage of the total vote required to
win all seats.
3. Constant swing ratio: The swing ratio of a districting plan is defined as the rate at
which a party gains seats per unit increment in votes. When this rate is identical
for all vote percentage points over a specified range, the swing ratio is said to be
constant over that range.
4. Competitiveness: The competitiveness of a districting plan is defined as the
percentage of districts in which some normal or expected vote is within a fixed
difference of 50 percent." (Niemi 1982, 36)
As Niemi elaborated later, "The swing ratio is calculated for one party only, but in
a strictly two-party system (or if only the two-party vote is used), it will be identical for
each party" (Niemi 1990, 172).
Stern advocates creating politically competitive districts as an ideal of
redistricting. "Legislatures should be apportioned so as to minimize the extent to which
any representative is able to rely solely on a single interest group for his support." (Stern
1974, 401). He notes in a footnote, "such a process exerts a moderating influence on the
representative."
Dixon put forth three observations about the redistricting process:
1. "Whether non-population factors are expressly taken into account or not, they
influence all election outcomes in all sets of districts
2. "A large number sets of districts that observe equal population can be created, and
they will have different and non-neutral electoral impacts
3. "Gross electoral inequalities can still be produced with equal population" (Dixon
1982, 7-8).
The work of gerrymandering legislators is made easier by the tendency of voters
to segregate themselves. This is in a sense the process described by Kevin Phillips in The
Emerging Republican Majority (Phillips 1969). While not everyone in the suburbs is
now Republican, and many suburban areas remain competitive politically, the converse is
more true; most central cities in the U.S. are overwhelmingly Democratic to the point that
the Republican Party essentially does not exist in them (Campbell 1996). This chapter
will demonstrate that this fosters an asymmetry that moderates the impact of any growth
in Republican vote. Campbell argues that SMP generates bias in favor of Democrats
specifically because Democrats appeal to a lesser socioeconomic sphere than do
Republicans, and because people of a lesser socioeconomic sphere not only cluster
together in neighborhoods, but they also have less voter turnout than those of higher SES.
Stated more broadly, "single-member district electoral systems apportioned by
population are biased in favor of parties representing lower socioeconomic groups."
(Campbell 1996, 41-42). Campbell further goes on to say, "any party whose adherents
have characteristics associated with low turnout and are geographically concentrated by
these characteristics will benefit from a districting system that reflects these geographic
considerations." (Campbell 1996, 42).
The basic goal of redistricting is to equalize population in districts. Until the
1960s, there was gross malapportionment of districts at all levels in the United States, and
the most minimal goal of any redistricting program is to make the districts more equal.
This need not be pushed to an extreme, as some courts have done. (Michigan judges, for
example, districted all of the state's congressional districts within three persons in 1982.)
Backstrom has pointed out the folly of using census data which are known prospectively
to lack accuracy at the minute level to create districtings that supposedly are. The
Supreme Court was already speaking in these terms at the outset of the reapportionment
revolution: "It is a practical impossibility to arrange legislative districts so that each one
has an identical number of residents, or citizens, or voters. Mathematical exactness or
precision is hardly a workable constitutional requirement." (Reynolds v. Sims, 1964, 377
US 533 at 578). Some years later, Justice White made the point that some factions were
knowingly engaging in malfeasance by enacting redistrictings that they knew were
inaccurate: "Legislatures intent on minimizing the representation of selected political or
racial groups are invited to ignore political boundaries and compact districts so long as
they adhere to population equality among districts using standards which we know and
they know are sometimes quite incorrect." (White, J., dissenting, Wells v. Rockefeller,
1969, 394 US 542 at 544-545).
Some argue that redistricting is a political problem that ought to be left to the
elected branches of government to sort out. Indeed, until it wrought the reapportionment
revolution, this was the view maintained by the Supreme Court. But the notion that the
political system will somehow be able to work redistricting out without outside help is
challenged by many who have to deal with the excesses of legislators armed with the
perk of being able to draw their own districts. "Those who would leave the problem to
the give-and-take of the political process overlook the fact that the process itself often
resembles a monopoly more than a free market, with little 'give' but a lot of 'take."'
(Baker, The Unfinished Reapportionment Revolution 1990, 25). Rather than a
monopoly, Issacharoff calls the actions of the two parties relative to redistricting
"duopolistic gerrymandering," (Issacharoff 2002) in which both agree to protect each
other's safe seats and minimize the existence of competitive seats. Ansolabehere and
Snyder reject this argument, although they agree that politicians have tried and failed to
do on a large scale what Issacharoff alleges (Ansolabehere and Snyder 2008, 268).
Ansolabehere and Snyder posit that such a succession of duopolistic gerrymandering
would have resulted in a bimodal ordering of safe legislative seats in each chamber for
which it was implemented (Ansolabehere and Snyder 2008, 269). They tested the
duopolistic gerrymandering hypothesis for all fifty states for the decade of the 1980s and
found a bimodal distribution in only one chamber in one state: the New York state senate.
"The distribution of the vote across legislative districts is clearly unimodal."
(Ansolabehere and Snyder 2008, 270). They found the same pattern in the U.S. House,
with the popular vote approximating a normal curve. (Ibid.)
The Court in Bandemer decided that reapportionment is justiciable, but confined
itself to "boundary manipulation that goes beyond usual and expected partisanship"
(Baker, The Unfinished Reapportionment Revolution 1990, 21), where, as Justice White
put it in his majority opinion, "the electoral system substantially disadvantages certain
voters in their opportunity to influence the political process effectively..." (Davis v.
Bandemer, 1986, 106 S.Ct. 2797 at 2811). Grofman characterizes the decision as
prohibiting partisan gerrymandering which is intentional, severe, and "predictably nontransient in its effects." (Grofman, Toward a Coherent Theory of Gerrymandering:
Bandemer and Thornburg 1990, 30).
Data analysis
Regional breakdown. Results for the Brookes analysis are computed regionally.
For our purposes, the United States consists of four regions. The Midwest is least
controversial in terms of its composition, and includes these twelve states: Illinois,
Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio,
South Dakota, and Wisconsin. The South consists of the eleven states of the
Confederacy, plus Kentucky, Oklahoma, and West Virginia. The Northeast includes the
New England and Middle Atlantic states, including Delaware and Maryland. All other
states are included in the West. There is no perfect regional aggregation of states, but this
grouping encompasses that used by many other scholars. These are somewhat similar to
the U.S. census regions but not identical to them, because this analysis includes Maryland
and Delaware in the Northeast while the Census Bureau includes them in the South.
(United States Department of Commerce, Bureau of the Census 1992, A-4). Groupings
that use more regions typically disaggregate the Pacific states from the Rocky Mountain
states. Certainly there is reason to do so from a political analysis standpoint, with the
mountain states being roundly more conservative than the coast states; however, the
argument could then be made that the Great Plains states have more in common with the
mountain states than they do with the Great Lakes states. The neat pattern made possible
by the definition of four large regions then becomes muddied into a mere agglomeration
of states into geopolitically similar categories. While having less history behind it than
the other regions, "the West" is certainly salient as an ecopolitical area, much as the other
regions call to mind specific contexts in which political history has been situated.
The big picture in the bias analysis is one in which the Democrats typically suffer
from bias against them on measure G, but often more than make up for it in bias toward
them on measure CSV. The other components of bias, typically more prominent in
countries with more than two parties, are usually negligible. The most obvious historical
pattern is that the Democrats started with a system that was only nominally biased in their
favor, (excluding the South, it was mostly biased against them) and with their landslide in
the 1974 election, created a reservoir of advantage that they were able to sustain into the
1990s. This is the picture described in King & Gelman (King and Gelman 1991, 110).
Even in 1990 and 1992, the Democratic advantage exceeded 35 seats. Despite the
Republican victory in 1994, the Republicans have mostly held their own in House
elections and have not accrued victories as disproportionate as the Democratic margins of
the earlier period were. The confounding thing for the Republicans is that even though
the bias toward them on measure G was at the end of their 12-year ascendency higher
than what it was toward the Democrats at the peak of their hegemony in 1976, the CSV
factor has not similarly changed direction. This factor seems to be a persistent bias in
favor of Democrats no matter how the rest of the bias equation turns against them, which
wanes a little in redistricting years. Thus, an asymmetry has emerged in which CSV,
which used to undergird the Democrats' advantage on G, now counterbalances the
Republican advantage on G. Furthermore, this appears to be a systemic bias stemming
from persistent lower turnout (coupled with declining population) in strong Democratic
areas. The result is an apparent governor in the electoral system in which the Republican
advantage in House elections can never be as great as the Democratic advantage was in
the previous period.
Campbell argues, "District turnout disparities, in themselves, do not create
partisan bias." (Campbell 1996, 101). However, this contradicts exactly what the CSV
component of the Brookes model shows, namely that the disparity in turnout from one
district to another is producing disparities.
Campagna & Grofman make the suggestion, "there are some general effects of
redistricting that act relatively similar[ly] across all states and give rise to a slightly
higher level of swing immediately after redistricting has taken place." They point to a
typically higher level of swing in 1982 than 1980, something borne out by this table
(Campagna and Grofman 1990, 1254). But in general, these data prove their point
wrong. If anything, it would be easier to believe that swing is reduced immediately after
redistricting and that 1980, rather than 1982, was an anomalous year, being less
connected in terms of its bias patterns to the preceding years.
Table 1. National bias analysis, 1966-2008.
Year
Bias
G
CSV
A
TPV
TPW
1966
1968
1970
1972
1974
1976
1978
1980
1982
6.86
10.91
11.72
6.11
44.50
56.45
48.30
13.22
35.98
-13.83
-13.74
-15.33
-13.01
25.89
34.84
20.10
-11.74
29.30
18.62
18.98
18.05
8.87
2.21
24.21
23.94
24.51
0.41
1.71
4.25
9.27
10.63
16.44
-3.66
4.05
1.74
6.30
0.36
1.42
-0.27
0.62
-0.04
1.06
0.20
-0.28
-0.02
0
0
0
-1
0
0
0
-1
0
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
*2006
*2008
14.63
21.01
21.82
45.31
37.95
-3.05
-9.05
2.90
-6.32
-6.66
-12.33
15.52
31.03
-0.80
7.60
2.88
23.11
21.57
-21.94
-33.08
-12.41
-29.75
-32.01
-38.52
-4.71
17.82
20.78
16.08
16.89
17.18
15.04
16.98
23.82
17.16
22.95
20.86
24.25
19.32
12.41
-5.51
-2.70
2.24
4.81
0.90
1.95
-1.23
-2.24
0.79
2.61
1.74
0.89
1.20
0.16
0.03
-0.20
-0.79
-0.56
-0.04
0.43
0.39
-0.31
0.88
0.20
0.02
-0.41
0
0
0
1
1
0
1
0
0
1
0
0
0
Regionally, this augurs most forcefully in the South. The South has swung in our
time from being solid for the Democrats to being nearly as solid for the Republicans.
The paradigm shift, in terms of the Brookes analysis, occurred in 1996. From then on,
the system, which had heretofore been biased in favor of the Democrats in terms of
contested seats, came to be biased in favor of Republicans. Furthermore, the G bias has
grown to this time that the CSV bias does not compensate for even half of it. In 2004,
more than half of the G bias in favor of the Republicans was coming from the South.
(This analysis does not include unopposed seats and other seats in which one party does
not run, so it clearly understates what the Democratic advantage in seats was for decades.
Were every seat taken into account, the Democratic total bias in the South would be at
least 25 in every election from 1964 until 1978, and the Republican total bias would be
over 18 seats in 2004.)
Insofar as redistricting was most lacking in the South before 1972, for the years
considered here, there is little support for King & Gelman's contention that redistricting
reduces electoral bias per se (King and Gelman 1991, 541). Indeed, with regard to the
Brookes analysis, bias overall in the South got worse before it got better, and in terms of
the G component, bias is higher in absolute numbers now than at any other time in this
period, even taking into account the many seats removed from competitiveness at the
beginning of the period by the absence of the Republican Party.
Table 2. Electoral bias in the South, 1966-2008.
Year
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
Bias
10.97
14.68
8.48
8.71
13.28
19.67
17.62
9.55
18.85
5.47
10.16
11.08
19.03
17.42
5.54
-10.35
-6.46
-10.60
-8.03
-12.54
-37.96
-24.12
G
4.42
11.97
1.84
5.06
5.89
15.06
9.86
3.60
13.42
0.53
8.37
5.67
14.11
9.60
-0.69
-17.20
-11.68
-14.92
-14.52
-20.38
-46.55
-25.48
CSV
7.56
2.49
1.31
1.97
3.83
6.93
5.37
5.46
-1.58
9.71
9.92
4.82
4.68
5.88
6.70
7.96
1.83
6.37
4.37
8.81
7.34
6.25
A
-1.02
0.04
5.38
1.99
3.86
-2.65
2.37
0.96
7.09
-4.82
-8.18
0.52
0.22
1.55
-0.51
-1.23
3.31
-0.92
2.10
-0.88
1.15
-4.66
TPV
0.01
0.18
-0.04
-0.31
-0.30
0.33
0.02
-0.47
-0.08
0.04
0.05
0.07
0.02
0.38
0.05
0.12
0.08
-0.12
0.02
-0.08
0.10
-0.23
TPW
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-1
0
0
0
0
The Midwest is the region of the country that has historically favored
Republicans. The party was born there, and quickly gave it a Northern foothold, which
remained persistent into the New Deal era. In some ways, the Midwest best illustrates
the ascent of the Democrats in 1974 and their rapid descent in 1994. The overall bias
number was negative through 1972, as was the G component. This suddenly turned
around in 1974, with the bias being a near mirror image of what it was in 1972. With the
exception of 1980, the Democrats had a bias in their favor until the Republican landslide
of 1994. The salience of this period of Democratic hegemony is well characterized by its
mere presence in this region which had for so long been unwinnable by Democrats. It
was here that the redistricting revolution first made waves by eliminating Republican bias
in the Midwest at a time the South was still subject to Democratic bias. In a sense, this is
why the Democrats were so effective at keeping a congressional majority even when the
presidency was alternating between the two major parties. According to Cox & Katz,
"the eradication of pro-Republican bias in the translation of Congressional votes into
seats resulted in an abrupt decline in the Republicans' probability of attaining a majority
of seats in the House of Representatives" and made recruitment more difficult for them
and resulted in the incumbency advantage being improbably more significant for them
than for Democrats (Cox and Katz 2002, 7).
Table 3. Electoral bias in the Midwest, 1966-2008.
Year
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
Bias
-17.53
-16.52
-7.18
-8.39
8.34
5.59
6.56
-3.43
4.58
2.07
5.50
7.07
11.03
9.75
-4.49
-3.10
2.14
G
-22.78
-18.82
-12.51
-11.95
4.42
2.33
-0.37
-10.37
8.59
1.88
3.84
3.64
6.95
7.68
-5.52
-6.57
0.49
CSV
4.59
1.92
3.84
0.31
-3.32
1.98
5.71
5.94
0.76
-0.53
0.36
3.00
2.82
2.33
1.20
3.79
3.13
A
0.68
0.40
1.65
3.14
7.36
1.18
1.16
0.91
-4.86
0.65
1.30
0.36
1.24
-0.51
-0.18
-0.71
-1.62
TPV
-0.02
-0.02
-0.17
0.11
-0.11
0.09
0.05
0.08
0.09
0.07
0.00
0.07
0.02
0.25
0.01
0.40
0.14
TPW
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2000
2002
2004
2006
2008
-6.19
-9.39
-11.32
-4.93
6.81
-9.48
-11.18
-14.13
-8.67
4.91
3.51
3.37
3.85
3.52
2.95
-0.53
-1.66
-1.20
0.15
-0.87
0.30
0.08
0.16
0.08
-0.19
0
0
0
0
0
The Northeast has been a strong area for Democrats since the rise of Franklin D.
Roosevelt, and they maintained their positive numbers on the bias scores during the
Republican ascendancy. Nevertheless, this is due in large part to their persistent positive
CSV score. In terms of G bias, the Democrats have occasionally seen an aggregate bias
against them in the Northeast, even in good Democratic years. The abstention factor
typically works to their advantage, almost as heavily as CSV bias, in some years.
Table 4. Electoral bias in the Northeast, 1966-2008.
Year
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
Bias
9.66
10.14
5.51
2.50
12.43
18.94
15.58
5.60
10.45
4.45
2.56
3.25
8.19
4.87
2.96
9.05
8.60
6.37
7.72
G
4.99
0.68
-1.52
-0.95
7.31
11.90
9.35
-1.48
11.35
-1.25
-2.20
-0.35
2.45
-0.26
-1.78
2.76
4.27
-1.04
1.67
CSV
2.99
7.53
4.32
1.79
1.52
5.15
7.06
6.08
-1.54
7.85
3.05
4.32
4.65
4.22
4.99
3.23
4.70
3.51
3.81
A
1.52
1.30
2.71
2.16
3.19
1.49
-0.94
1.91
0.68
-2.11
1.65
-0.47
0.41
0.05
-0.22
2.53
0.03
3.04
1.03
TPV
0.15
0.62
0.00
0.50
0.41
0.39
0.11
0.09
-0.04
-0.04
0.06
-0.25
-0.32
-0.13
-0.04
-0.47
-0.40
-0.14
0.21
TPW
0
0
0
-1
0
0
0
-1
0
0
0
0
1
1
0
1
0
1
1
2004
2006
2008
6.94
27.98
45.93
0.20
23.98
41.56
3.63
2.45
1.81
3.20
1.61
2.69
-0.09
-0.06
-0.14
0
0
0
The West is a politically amorphous region. It is dominated by California, which
is by this time one of the most reliable Democratic states. It has not always been so, and
were it not for California, the West would be the strongest Republican region. However,
despite the West's Republican bias for several election cycles beginning in 1994, it has
reverted beginning in 2000 to its pre-1994 pattern of mild Democratic bias. Also, the
West was not as biased against the Democrats before the 1974 Watergate landslide as the
Midwest, turning in positive numbers in the pre-1974 elections and only small biases
against the Democrats on the G measure in those years. Despite the near elimination of
G bias in California in 2002 and 2004 (see below), the Democrats suffer from persistent
G bias in the West, highlighting that the situation is worse for them outside California.
Table 5. Electoral bias in the West, 1966-2008.
Year
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
Bias
3.46
2.21
4.84
3.24
10.45
12.06
8.47
1.51
2.04
2.58
2.45
0.51
7.32
6.69
-7.15
-4.67
G
1.29
-1.41
-1.29
-1.72
8.79
7.67
3.81
-1.77
-0.44
-1.80
-1.83
-5.65
-0.24
5.41
-12.17
-13.53
CSV
2.99
3.72
6.45
2.67
-0.30
7.65
4.94
5.83
1.16
3.91
3.89
5.61
6.37
2.53
3.68
10.08
A
-0.79
-0.30
-0.16
2.05
2.01
-3.51
-0.25
-2.56
1.28
0.38
0.44
0.54
1.39
-0.95
1.47
-1.49
TPV
-0.03
0.21
-0.16
0.24
-0.04
0.25
-0.02
0.01
0.04
0.08
-0.05
0.01
-0.20
-0.30
-0.12
0.26
TPW
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1998
2000
2002
2004
2006
2008
-1.94
3.84
2.45
4.39
8.81
20.29
-7.16
-3.95
-7.23
-1.06
2.66
13.24
6.60
7.52
7.55
6.24
5.41
5.20
-1.41
0.93
2.16
-0.68
0.93
1.59
0.03
-0.66
-0.04
-0.11
-0.20
0.26
0
0
0
0
0
0
An examination that focuses on the impact of redistricting on electoral bias is
most salient in the CSV category. The elimination of disparities in population between
districts is the main objective of redistricting. Indeed, when judges have been the ones to
implement redistricting, they have been known to draw districts absurdly similar in
population to avoid any charge that they did not fix disparities. However, the Brookes
method as it is applied here looks at disparities not in population but in registered voters.
This emulates the function of proportional representation of equalizing actual voters
rather than population. Even so, one would expect the CSV bias to be smallest in
redistricting years, when the districts newly reflect the population as it was 2 1/2 years
earlier; and higher in the last year of a district map, when the census that created the lines
is more than 10 1/2 years old.
The pattern generally visible in the data is that the CSV bias is reduced with each
redistricting and then grows over the life of that map. While it is not a perfect pattern,
the paradigm of CSV bias favoring Democrats is overwhelmingly true. (Only five of 85
data points are negative.) This suggests, ceteris paribus, that Democrats' advantage in
House elections ought to increase through the life of a redistricting map. However, this
might be only be relevant in a statistical analysis and convey nothing meaningful
politically, as it might mean only that districts which are already overwhelmingly
uncompetitive on the Democratic side have become more so.
Table 6. CSV bias nationally and by region, 1972-2008.
USA
MW
NE
1.79
0.31
8.87
1972
1.52
-3.32
2.21
1974
5.15
1.98
24.21
1976
7.06
5.71
23.94
1978
6.08
5.94
24.51
1980
-1.54
0.76
0.41
1982
7.85
-0.53
20.78
1984
3.05
0.36
16.08
1986
4.32
3.00
16.89
1988
4.65
2.82
17.18
1990
4.22
2.33
15.04
1992
4.99
1.20
16.98
1994
3.23
3.79
23.82
1996
4.70
3.13
17.16
1998
3.51
3.51
22.95
2000
3.81
3.37
20.86
2002
3.20
3.85
24.25
2004
1.61
3.52
17.88
2006
2.69
2.95
17.10
2008
Horizontal lines reflect redistricting years.
S
W
1.97
3.83
6.93
5.37
5.46
-1.58
9.71
9.92
4.82
4.68
5.88
6.70
7.96
1.83
6.37
4.37
8.81
7.34
6.25
2.67
-0.30
7.65
4.94
5.83
1.16
3.91
3.89
5.61
6.37
2.53
3.68
10.08
6.60
7.52
7.55
6.24
5.41
5.20
A bivariate regression using the number of years since redistribution as the
independent variable (year ending in 2= 0, ... , year ending in 0 = 8) and the CSV bias as
the dependent variable, with the 15 elections from 1972 to 2004, inclusive, for the U.S. as
a whole and each of the four regions (N=85) yields a coefficient of 0.51 (SE 0.25), with
significance at the .95 level. While this is of only modest salience, it points in the
direction in which a more casual analysis of the data also points.
An object lesson can be taken by applying the Brookes method to California by
itself. In 2002, redistricters in California created 53 safe districts. There is not a
competitive House district left in the state. The efficacy of this redistricting as a
bipartisan compromise shows in the Brookes analysis, because the G bias factor
diminishes almost to zero, comprising -0.246 in 2002 and 0.558 in 2004. This is a drop
from a G bias in favor of the Democrats of 1.79, or almost two seats, in 2000, under the
previous map. However, as is often the case in redistricting chess moves, amelioration
of this aspect of bias only makes another one worse. As the map is supposedly made
more fair to Republicans in suburban areas by reducing the G bias, the CSV bias expands
to take up the slack. Thus, the overall bias of the redistricting is approximately the same
as before, and perhaps even more biased in favor of the Democrats. While neither party
has to fight for its seats in California anymore, the Republicans are still fighting harder to
achieve parity.
Table 7. Bias analysis of U.S. House elections in California, 2000-2004.
2004
2002
2000
0.558388
1.79025 -0.24576
G
3.885317 5.310184 4.446103
CSV
0.333325'1.473966 1.021839
A
-0.06005 0.019658 0.116509
TPV
0
0
0
TPW
5.948839 6.558045 6.142839
Bias
* excludes races where either D or R not running
The method also sheds light on the infamous Philip Burton gerrymander of 1982.
This plan, one of the few non-racial gerrymanders to be classified as truly extreme by
geographers and political scientists, cut up neighborhoods and ran district lines every
which way in order to give the Democrats an advantage in House elections. Burton was
not the least apologetic about using gerrymandering. He declared that the first priority
for a legislator should be to "get yourself in a position [to] draw lines for [your own]
district. Then, you draw them for all your friends before you draw anyone else's."
(California Journal, August 1983, quoted in (Baker, The 'Totality of Circumstances'
Approach 1990, 208). Owen & Grofman explained "aberrant features of the plan could
almost always be traced to manipulations designed to achieve probably partisan
advantage." (Owen and Grofman 1988, 19).
However, a Brookes analysis shows that in terms of G, the 1982 election was
much less biased in the direction of the Democrats than the 1980 was biased in the
direction of the Republicans. Essentially, the redistricting reversed the direction (if not
the magnitude) of the G bias in California and even cut the sizeable bias the Democrats
had in CSV going into the election, which was more than enough to wipe out the G bias
under the previous map. This is consistent with the assertion of Ostdiek (1995) and
others that in order to properly dilute its opponents' districts, a party must reduce its own
advantage in its safe seats. Desposato and Petrocik succinctly state the paradox of the
partisan gerrymander this way:
At the margin, ensuring electoral security and maximizing
the party's number of seats are conflicting goals. But a
balance is struck when the dominant party provides the
'assurance' of victory to the nth candidate of the largest
possible majority for their party, while providing a
'guarantee' of victory to the smallest possible number from
the other party by packing minority party supporters
together in a minimum number of districts. (Desposato and
Petrocik 2003, 18).
This paradox and its implications are quite visible in this analysis.
Table 8. Bias analysis of U.S. House elections in California, 1980 and 1982.
1982
1980
-5.35571 1.171245
G
5.503137 2.949975
CSV
0.342799 0.895863
A
0.243822 -0.00324
TPV
0
0
TPW
0.73405 5.013841
Bias
*excludes races where either D or R not running
Conclusion
Perhaps the key finding in this analysis is illustrated in Chart 1. Even as the bias
in G drifted deeper into the Republican zone between their victory in 1994 and the
Democratic recapture of the House in 2006, the CSV bias has remained largely constant
for the Democrats and has moderated the total bias against that party. As mentioned,
CSV bias is less easily corrected than G bias. It would disappear under most forms of
proportional representation, but it would also be greatly reduced under certain
applications of multi-member districts. No state has used multi-member districts since
1966, when Hawaii was still electing its two members at large. But multi-member
districts would not necessarily help Republicans in the larger scheme of things, because
they would also reduce its more sizeable advantage in G bias. Thus, the current state of
American politics can be described through a Brookes analysis as having an asymmetry
in which the Democrats can more easily convert their votes into seats than the
Republicans can. This meshes with the redistricting hypothesis in that this chapter finds
that while redistricting has some impact on the CSV bias in favor of the Democrats, it
does not eliminate it completely.
That CSV tends to drop in redistricting years and then increase for the rest of the
decade is consistent with the core ideal of redistricting: to equalize population.
In many cases, the effects attributable to districting are small. As Ansolabehere
and Snyder put it, "Political cartography today results from struggles among the
legislators themselves over many different objectives and goals. The new lines have
become increasingly twisted as parties and incumbents, constrained by the courts and the
governors, fight for any little gain they can make. But the gains are just that - little."
(Ansolabehere and Snyder 2008, 271).
Nevertheless, Campbell argues that turnout disparities between districts - made
more obvious through the Brookes analysis - ought to be addressed somehow: "It would
be merely a constitutional nicety rather than a guarantee of popular sovereignty to insist
that districts be equally populated, if the numbers of actual voters in districts varied
widely, with very few casting ballots in some districts while huge numbers voted in other
districts in one election after another." (Campbell 1996, 96).
Presidential voting in House districts and party polarization
Analyzing presidential voting in house districts gives us the ability to look beyond
the effects of congressional incumbency and redistricting. Although there are incumbent
presidents too, the visibility of the presidential contest makes the incumbency advantage
less important. Although there are also state effects in presidential voting, there are no
district effects because voters in the individual states are mobilized to vote by campaigns
and targeted by media regardless of the congressional district they live in. Looking at
presidential voting by house district therefore lets us see things about districts that would
not be visible if only house contests, with their severe incumbency effects and the sharp
self-elimination of challengers, were analyzed. When looking at presidential voting in
house districts, we often see the kind of lopsided elections occasionally experienced in
Canada and Britain and other parliamentary democracies. For example, in 1972,
Republican Richard Nixon won 376 congressional districts and Democrat George
McGovern won only 59. On that same day, by comparison, Democrats won 242 house
seats while Republicans won only 192 (and one other seat went independent). From
those House results, one wouldn't know that a presidential landslide was happening
elsewhere on the ballot. Similarly, in 1984, Republican Ronald Reagan won 369
congressional districts and Democrat Walter Mondale won only 66. Meanwhile, House
Democrats won 252 seats to 183 for House Republicans. In the American example,
House contests have been marked with greater consistency from election to election and
presidential contests with greater fluctuation. Presidential voting is therefore more
similar than congressional voting to the parliamentary model in Britain and Canada.
Conservatives won 211 of 282 seats in the Canadian election of 1984 but were reduced to
only two in 1993 (and took 169 in the intervening election in 1988). Liberals won only
40 seats in that election in 1984, increased to a not much better 83 in 1988, and then
boomed to 177 in 1993. These landslide elections where old governments are voted out
as much as new governments are voted in are much more reminiscent of presidential
elections where incumbents are defeated than they are of the rare U.S. House election in
which the majority party changes hands. Therefore, looking at presidential voting in the
context of House districts can give us clues about the effects of those districts (including
the districting that created them) that is not available to those who only examine voting
for Members of Congress.
The degree of polarization in the country has also changed in the period being
studied. Abramowitz et al note that the 1976 and 2004 presidential elections were
decided by the same margin in the electorate but that 1976 was mostly about many close
states whereas by 2004, the country was more polarized (Abramowitz, Alexander and
Gunning, Don't Blame Redistricting for Uncompetitive Elections 2006, 88). Pildes &
Niemi state the problem this way: "Democratic theory might accommodate either
proportional representation or territorial districting," but they hasten to add while citing
Polsby & Popper, "trying to force the kinds of concerns a proportional-representation
system addresses into a territorial system eventually stretches the latter to a breaking
point." (Pildes and Niemi 1994, 502). Polsby & Popper decry gerrymandered SMP
systems as possessing "the worst aspects of both Madisonian democracy and proportional
representation." (Polsby and Popper 1991, 306). They argue that gerrymandered
majorities are less in need of forming coalitions and thus more ideologically polarized.
(Polsby and Popper 1991, 307).
Brookes analysis of U.S. presidential voting by Congressional district
1984
66
369
66
369
196889.1
219834.8
286299.5
280021.8
89410.43
60186.96
1988
135
300
135
300
193818.2
214375.9
306956.2
304670.8
113138
90294.88
1996
1992
257
279
178
156
257
279
178
156
192257.1 206536.8
193299.3 194270.6
328919
303466.6
315225.7 362380.6
64346.8 116606.7
54732.33 136306.8
42800.63 18041.65
48926.21 19536.99
-323.11 -27.7799 -232.353 -327.946
-77.9973
24.90803 12.51027 7.759074 2.197369 -4.78541
-2.63797 -9.59473
22.9687 52.88677
30.2127
0
0 -4.39835
0
3.259583
0
0
-18
0
0
2.947858
-181.667
-302.519
-70.4677 -320.195
-187.383
-1.62316
24.08191
0
0
-164.924
77.82472
8.242214
10.31116
-3.62386
0
92.75423
1968
159
229
191
243
133921.1
150796.4
191561.2
214641.5
34969.66
41304.53
22670.47
22540.58
G
CsV
A
TPV
TPW
Total
1972
59
376
59
376
170097.6
174894
218452.5
231111.6
50064.12
60775.41
1976
219
216
219
216
170148.1
195935.6
238435.3
247094
68287.18
51158.35
1980
129
306
129
306
152711.7
194205
260864.1
263507.5
108152.4
69302.45
9425.612
14626.26'
2000
207
228
207
228
216952.8
247129.4
347199.6
402982.8
120896
148039.6
9350.768
7813.768
136.4404 -49.4948
32.643
20.81629
-7.44963 -6.48448
0.21582 1.679647
0
0
150.0229 -21.6567
2004
180
255
180
255
261497.2
289293.3
372889.6
414902.6
109229
123590.4
2613.497
2439.924
2008
242
193
242
193
286135.6
309584.9
380602.4
418728.2
90590.62
104952.9
3876.202
4190.394
31.9266
-97.2002
23.47866 20.68385
-1.52779 -3.63827
0.246967
0.03888
0
0
-75.0024 49.01105
Chart 1. Electoral bias and G and CSV components, for districts with both D and R candidates
BLAS INHOUSE SEATS- TWOG-PARTY CONTETONLY
80.00(10.00
40.00
..
-I-
0.00
-20.00
-40.00
-60.00
Bias
-G
""***
Appendix 1. Notes on methodology.
There is no single source for the size of the registered electorate by Congressional
district. Where such numbers were included in the Almanac of American Politics, they
were used. For 1970, where none were available, they were copied from 1968. For
1972, the statewide electorate from Leip for other states except Iowa, Missouri, North
Dakota, and Wisconsin was used and interpolated by proportion of the 1972 presidential
vote allocated straight-line. For Iowa, Missouri, North Dakota, and Wisconsin, the
electorate was fixed by dividing the presidential vote by .7 by comparison with other
midwestern states. For 1974, the electorate was copied from 1972. For 1976, the
statewide electorate for Georgia, Massachusetts, New Jersey, New Mexico, Ohio,
Oregon, and West Virginia were taken from Leip. For other states, except North Dakota
and Wisconsin, it was obtained from the Almanac of American Politics 1978. The
electorate by district was established by allocating the proportion of presidential vote
straight-line. For North Dakota and Wisconsin, the electorate was fixed by dividing the
presidential vote by .72 by comparison with Illinois, Minnesota, and South Dakota. For
1978, the electorate was copied from 1980. For 1980, the electorate was apportioned
straight-line from the Almanac of American Politics 1982 by proportion of presidential
vote. For North Dakota and Wisconsin, the electorate was fixed by dividing the
presidential vote by .72 by comparison with Minnesota and South Dakota. For 1982, the
electorate was copied from 1980. For 1984, the electorate was calculated from statewide
electorate figures from the Almanac of American Politics 1986 by apportioning the
presidential vote by district straight-line, for all states except Alabama, North Dakota,
and Wisconsin. For North Dakota and Wisconsin, the electorate was fixed by dividing
the presidential vote by .715 by comparison with Minnesota and South Dakota. The
numbers didn't make sense for Alabama, so a proportion of .7 was assumed. For 1986,
the electorate was copied from 1984. For 1988, the electorate was calculated from
statewide electorate figures from the Almanac of American Politics 1990 by apportioning
the presidential vote by district straight-line, for all states except Wisconsin. For
Wisconsin, the electorate was fixed by dividing the presidential vote by .71 by
comparison with Illinois and Minnesota. For 1990, the electorate was copied from 1988.
For 1992, the electorate by Congressional district was available for Maryland on the
state's web site. For all other states except North Dakota and Wisconsin, the electorate
was calculated from statewide electorate figures from the Almanac of American Politics
1994 by apportioning via straight-line the presidential vote by district. For North Dakota
and Wisconsin, the electorate was fixed by dividing the presidential vote by .73 by
comparison with Illinois, Minnesota, and South Dakota. For 1994, the electorate by
Congressional district for Florida and Maryland was obtained from those states' web
sites. For all other states, it was copied from 1992.
In applying presidential election voting using congressional districts, most
presidential election results by congressional district were taken from various editions of
the Almanac of American Politics and Congressional Quarterly's Guide to U.S.
Elections. Some recent results were obtained from online sources. The size of the
registered electorate was taken from the Almanac of American Politics where available.
The figures from 1976-1988, inclusive, were first obtained from Congressional Districts
in the 1970s (Congressional Quarterly 1974) and Congressional Districts in the 1980s
(Congressional Quarterly 1983), but these figures were deemed to be too inflated to be
directly comparable with data from other sources for the years 1968 and 1972, and 1992
to 2008. Thus, data for 1976-1988, inclusive, as contained in the aggregate analysis,
were interpolated from the years 1968, 1972, and 1992, using a quadrennial inflator of 8
percent. This was apportioned in proportion to the figures obtained from Congressional
Districts in the 1970s and Congressional Districts in the 1980s. Figures for abstentions in
this period were also deemed to be inflated, and so the aggregate was calculated from
straight subtraction of the election results from the electorate as estimated above. These
results at the congressional district level were not posted backward to the database, so the
database continues to contain the suspect figures from Congressional Districts in the
1970s and Congressional Districts in the 1980s.
Appendix 2. Equations of bias components
x = number of seats won by object party (Democrats)
y = number of seats won by other major party (Republicans)
b = number of seats where object party leads other major party
f= number of seats where other major party leads object party
P = average number of combined votes for two major parties where object party leads
other party
Q = average number of combined votes for two major parties where other party leads
object party
R = average registered electorate in seats where object party leads
S = average registered electorate in seats where other party leads
C = average number of abstentions in seats where object party leads
D = average number of abstentions in seats where other party leads
U = average number of minor party votes in seats where object party leads
V = average number of minor party votes in seats where other party leads
G = gerrymander effect
CSV = constituency size variations (malapportionment effect)
A = abstentions effect
TPV = third party votes effect
TPW = third party wins effect
G = {[f(Pb/Qf-1)] - [b(Q/Pb-1)1}} / 2
CSV = {[f(S/R-1)] - [b(R/S-1)]} / 2
A
=
TPV
{f*[(R/(R-C))*[(C/R)-(D/S)}]-b*[(S/(S-D))*[(D/S)-(C/R)]I} / 2
=
{*[(R/(R-U))* [(U/R)-(V/S)}]-b*[(S/(S-V))*[(V/S)-(U/R)]]}
/2
TPW = (x-b)-(y-J)
Source: Alan Siaroff, "Electoral Bias in Quebec Since 1936," paper presented at the 17th biennial
conference of the Association for Canadian Studies in the United States (ACSUS),
Portland, Ore., November 19-23, 2003, p. 17, adapted from Johnston, et. al., From
Votes to Seats: The United Kingdom's ElectoralSystem in OperationSince 1945.
Manchester, U.K.: Manchester University Press, 2001, pp. 229-230.
Chapter 4 - Racial Redistricting and the Brookes Method
Racial redistricting has evolved from a form of segregation derided by minorities
and advocates for minorities within the larger population to a form of segregation praised
by minorities and these advocates. A view expressed early in the reapportionment
revolution held, "Racial gerrymandering is simply a particular kind of political
gerrymandering" (Dixon 1971). Actions of minorities and their white sympathizers over
the next decade suggested that they did not see the act of racial gerrymandering as Dixon
did. However, when partitioning of states to segregate minorities was carried to excess in
the 1990s, Dixon's words were vindicated and support for racial gerrymandering from
courts, the general public, and even minorities themselves began to wane. Backstrom
notes that a fair amount of demand for highly racially-conscious redistricting came from
blacks themselves: "Black people wanted to be represented as blacks and not merely as
people." (Backstrom 1982, 45). Polsby & Popper argue that many minority voters see
themselves as members of a minority group first and as a member of a political party only
secondarily (Polsby and Popper 1991, 337).
Long before anyone started consciously packing black voters into particular
congressional districts, scholars were aware that majority-minority districts were on their
way through a more natural process. Jewell notes that because black residents tend to
live together in certain residential areas, a districting plan is apt to create some blackmajority districts (Jewell 1969, 15). Campbell argues that the single-member plurality
electoral system, "which permits the existence of cheap seats and the overrepresentation
of its voters, significantly augments the representation of African-Americans." (Campbell
1996, 208). Pildes & Niemi argue that having safe districts created to benefit minorities
might be essential to "avoid their submergence in a hostile majority" (Pildes and Niemi
1994, 526).
In 1982, Congress renewed the Voting Rights Act. Contained in the act was a
provision to use outcomes rather than merely the intent of election officials in evaluating
electoral rules, including district boundaries. The act states explicitly that it creates no
right to proportional representation on racial lines (Issacharoff 1995). Striving to
increase minority representation in the House, states moved to redesign districts to give
minorities greater opportunities to win seats. By the late 1980s, it was clear that doing
this was tantamount to using the gerrymandering strategy of packing. This is because
blacks are overwhelmingly a Democratic constituency and Hispanics are only somewhat
less so. In the round of redistricting following the 1990 census, Republicans moved
aggressively to create majority-minority districts and thereby pack Democrats. In three
states - Georgia, Louisiana, and North Carolina - Republicans drew districts that formed
tentacles reaching into black
neighborhoods, linking disparate
Atlanta
Augusta
communities over a large swath of the
states into a single district. The Georgia
district was based in east-central Georgia
and had tentacles reaching into Augusta
Savannah
Sources of silhouette maps: Election Data Services,
MarkMonmonier.com
and
Savannah and toward Atlanta.
The Louisiana district ran from the Baton
ret
Rouge area along the Mississippi River and
Monroe
thence westerly along the Arkansas border, with
tentacles reaching into black neighborhoods
along the way and as far west as Shreveport.
The state border was clearly recognizable in the
silhouette of the district.
The North Carolina district stretched
from Gastonia to Durham, running along Interstate 85 for much of its length and in some
places no wider than the freeway itself. The joke was that someone could drive down I85 with both doors open and kill most of the people in the district. It has also been
likened to a large intestine unfurled, complete with vermiform appendix reaching to
include part of Gastonia. Where previously black opportunity districts had been 55-60
percent black, the three districts in these instances were supermajority black - over 90
percent. As a result, Republicans were able to be more competitive in the non-minority
districts of these states. This was
Greensboro
a factor in Republicans winning
Winston-Salem
Durham
High Point
control of Congress in 1994.
Bullock describes the
process of packing minorities into
certain districts as "bleaching"
Charlotte
adjacent districts and thereby
making them more Republican (Bullock 1995). Petrocik and Desposato minimize the
impact of redistricting, calling it a "friendly" redistricting for Democrats (although they
admit many Democrats had lost many black constituents) that happened to coincide with
a rising tide for Republicans and an aggressive push by Republican leadership to recruit
quality challengers (Petrocik and Desposato 1998, 630).
The creation of these districts provoked a backlash, and the complaints of some
voters who claimed their rights were compromised by this aggressive redistricting went
to the Supreme Court in Shaw v. Reno. The Court urged North Carolina to come up with
a greater justification for the highly irregular districts. According to Justice White, the
central holding of Shaw is "race-conscious redistricting that 'segregates' by drawing
oddly shaped lines is qualitatively different from race-conscious redistricting that affects
groups in some other way." (Pildes and Niemi 1994, 499). Pildes & Niemi note, "at a
certain point, the use of race can amount to value reductionism that creates the social
impression that one legitimate value has come to dominate all others." (Pildes and Niemi
1994, 501). Upon remand, the trial court found the North Carolina districts to be
justified. The case, by then renamed Shaw v. Hunt, went back to the Supreme Court,
which ruled the new 12th district unconstitutional. North Carolina then redid its
districting again. There were still two majority-black districts, but they were much less
supermajority black than had been the case in the districting for the 1992 election.
Arguing that North Carolina's egregious 12th district of 1992 resulted from racial
gerrymandering being an afterthought once incumbent protection had been attended to,
Polsby & Popper state, "The construction of nonugly districts might have been easier if
the districtmakers were not trying to do so many things at once." (Polsby and Popper
1993, 653). Georgia successfully redrew its districts, but Louisiana and North Carolina
were each sent back to the drawing board with
Winston Salem
their revised majority-minority districts.
'-xoton
Stat..vilm.
Thus, in the decade of the 1990s, each of
those states had three very different sets of
Kannapolis
Charlotte
districts, with North Carolina finally coming
up with a plan for the 12th that was not only
not extreme in its racial gerrymandering, but
also no longer extreme in its deviation from the compactness ideal too (at least by 1990s
standards in the Deep South).
Scholars disagree about the impact of these majority-minority districts on partisan
bias. There can be no question that the largely unfettered zeal of the 1992 round of
redistricting produced severe biases against Democrats. However, scholars who focus on
the post-Shaw rounds of redistricting naturally find less empirical support for the idea
that the creation of majority-minority districts packs Democrats. It has often been noted
that minorities can usually be used as a proxy for Democrats in redistricting due to the
high affinity with particular minorities for the Democratic Party. "The higher the
percentage minority, the greater the probability of electing a Democrat to office."
(Handley, Grofman and Arden 1998, 13). Bullock notes that challenger quality is tied to
racial redistricting. "Some whitened southern districts attracted more formidable
Republicans than would have emerged had the racial proportions remained unchanged"
(Bullock 1995).
Lublin and Voss conceive as the 1994 election as having eliminated the moderate
Democrats who previously dominated the South and replacing them with either very
conservative Republicans or with black Democrats (Lublin and Voss 2003, 234). They
attribute this change to redistricting after 1990 that packed Democrats and particularly
blacks into majority-minority districts. Bullock argues that due to the unpopularity of
President Bill Clinton among southern whites, Republicans would have gained seats
anyway in the South, "but affirmative action gerrymandering greatly contributed to these
advances" (Bullock 1995).
Abramowitz and his co-authors make the point that the change in the number of
safe and competitive districts over the past 30 years has not been due to redistricting and
demonstrate that these numbers changed little between 1990 and 1992 and between 2000
and 2002, which is where the change should have been noted if redistricting (which
comes into place before the year ending in 2) were the cause (Abramowitz, Alexander
and Gunning, Don't Blame Redistricting for Uncompetitive Elections 2006, 88). Rather,
they blame the eradication of marginal districts on the tendency of Americans to selfsegregate into homogenous areas. However, in another article published by the same
authors that year, they attribute the net entire change in party standings in the U.S. House
elections in 2004 to the gerrymander in Texas (Abramowitz, Alexander and Gunning,
Incumbency, Redistricting, and the Decline of Competition in U.S. House Elections
2006, 75). Contrary to this view, Bullock argues that the results of the first two elections
after a remap ought to be considered jointly, because "the full impact of redistricting is
not usually felt in the first election" (Bullock 1995).
Although the United States has many minority groups, discussion on redistricting
questions (and other issues involving race) has often dealt with the country as if it were a
binary black-white polity. However, as of 2010, this is changing rapidly. More and
more places which formerly had a Hispanic community too small to be given parity with
blacks on questions of redistricting now have a Hispanic community of considerable size;
and in a growing number of places, blacks are no longer the largest minority group in
numeric terms. The question of how to deal with competing minority groups is a tricky
one. Some of the most contentious districts, particularly the "earmuffs" district in Illinois
and the New York City district that jumps the East River twice, were created not to favor
blacks but to engender a Hispanic-majority district in the context of the Hispanic
community living in neighborhoods highly proximate to blacks. In both of the
aforementioned cases, the unusual shape of the district comes from the challenge of
creating a Hispanic-majority district without breaking up adjacent black-majority
districts. In the case of Chicago, this posed a particularly difficult geometry problem. As
of 2000, Chicago had two large majority-Hispanic communities, one on the North Side
and one on the South Side, which were separated by a nearly all-black area. The black
West Side of Chicago needed to be contiguous with largest black area of Chicago, on the
South Side, in order that a black-majority congressional district could exist. This means
the two Hispanic communities could not
be connected either directly or by way of
downtown Chicago. The solution was to
extend the district encompassing the
Hispanic neighborhoods westward via a
spaghetti string at some points only one
block wide, to the western county line.
This resulted in the Hispanic district surrounding the western end of the black district and
also gave the district its unusual shape which nearly everyone who comments on it likens
to a pair of earmuffs. Luis Gutierrez won the district. The goal of electing a Hispanic to
the House of Representatives was achieved, but this came at the cost of making everyone
aware of how blatantly redistricting was used to make it happen. As a result of the mere
existence of such a district, the legitimacy of redistricting for the entire House of
Representatives is called into question.
By no means is race the only factor causing extreme compromises of compactness
to happen in Illinois. The eight Downstate districts (11-12, 14-19) are all very noncompact, as is discussed in the chapter on compactness and voter knowledge. Race is a
fairly small consideration in Downstate congressional redistricting since there is no
opportunity to create a majority-minority district there. Rather, these compromises of
compactness are caused by political gerrymandering, informed somewhat by the
likelihood of minorities to also be Democrats.
At the same time, racial segregation combined with a compactness standard
serves to guarantee the existence of at least one majority-minority district, if the minority
population is large enough (Barabas and Jerit 2004, 424). This is harmonious with a
simulation of ward districting in Mobile, Alabama, which found that the most compact
district solution was very similar to a scenario that assured one majority-black district out
of three. In this instance, the majority-black district (90 percent black) was identical in
the two scenarios (O'Loughlin and Taylor 1982, 329).
Race is in many ways the national obsession in the United States, much the way
language is in Canada and questions of national identity are in some European countries.
Race is so important to redistricting that the very first census data made available are
population breakdowns by racial categories. This is known as PL 94-171 data, after the
law passed in 1974 that had no small role in putting race near the front of redistricting
discussions.
At the same time as race is at the forefront of redistricting, in some ways,
questions about race are put off limits in redistricting discussions. Because the Voting
Rights Act prohibits redistricting aimed at curtailing the electoral success of minorities,
and because many states (covered by section 5 of the Voting Rights Act) are subject to
having their redistricting plans (and other electoral measures) pre-cleared by the U.S.
Department of Justice, discussion about race in redistricting is about how minorities can
be advantaged through redistricting. In many places there are "minority-opportunity"
districts. These typically have a large share of minority voters concentrated in them so
that a minority candidate can have a chance of being elected. Supposedly, a minority
group would fare better at electing a member if all members of the group were placed in a
single district than if they were divided among multiple districts. If this were done in a
partisan sense, it would probably be called packing, but minorities, by definition,
constitute a minority in the country and special efforts are often taken to help them boost
their chances of succeeding in the electoral sphere. In some states, this creation of
minority-opportunity seats has gone further and has resulted in the creation of majorityminority districts. Obviously, some minority-opportunity districts will naturally be
majority-minority districts in places where there is a large population of a single
minority, such as Chicago or Los Angeles. Majority-minority districts have existed in
some of the larger cities since the early 20th century. Controversially, in the 1990s, some
states began creating majority-minority districts that reached across large distances in the
state to take in widely separated minority communities in different metropolitan areas of
the state.
One of the main debates over majority-minority districting is whether the race of
the representative makes a difference. Cameron et al speak of increasing the presence of
blacks in Congress as descriptive representation, while enacting legislation of benefit to
minorities is substantive representation (Cameron, Epstein and O'Halloran 1996, 794).
Bullock argues that the process taking place in the South is bad for blacks. "The
replacement of moderate white Democrats with conservative Republicans, even with the
addition of a few African American legislative seats, bodes ill for the ability of African
American legislators to find the allies they need to achieve their policy goals." (Bullock
1995). Anecdotal evidence has been mixed on the question, and this is not a recent
phenomenon. Jewell reported that two black legislators in Ohio did not believe that
redistricting benefited the black population. One thought his legislative effectiveness was
diminished "because he was perceived as a representative of a single district and a single
interest rather than as a representative of the whole county." (Jewell 1969, 17). Another
black Indiana legislator interviewed by Jewell thought that districting hindered his ability
to represent black interests effectively (Jewell 1969, 17). This very debate about the
merits of whether minority representatives should only try to represent their own group
has reached the level of the Supreme Court. As Justice O'Connor put it in her majority
opinion in Shaw, "When a district obviously is created solely to effectuate the perceived
common interests of one racial group, elected officials are more likely to believe that
their primary obligation is to represent only the members of that group, rather than their
constituency as a whole." (Shaw v. Reno, 1993, 509 US 630, 648).
Pildes & Niemi argue that the primary reason for the odd shape of North Carolina
12 (the controversial 1-85 district of the 1990s) is not race but rather partisan political
considerations. They argue that once the justice department ordered a second majorityminority district in the state, the legislature could have created a compact black-majority
district. That they chose to craft a bizarre district was an exercise of their political
judgment (Pildes and Niemi 1994, 517). They further argue that VRA gives politicians a
false excuse to use to justify extremes in redistricting, or "Machiavellian lengths to
protect their seats and pursue their partisan agendas," which Pildes & Niemi characterize
as "self-interest masquerading as race consciousness." (Pildes and Niemi 1994, 518).
However, in the same article, they claim that black-dominated districts are apt to exist
solely to enhance the representation of minorities (Pildes and Niemi 1994, 526).
Lublin argues that rather than majority-minority districts, blacks would fare better
with minority-opportunity districts where they would constitute the swing vote rather
than be packed into a single district (Lublin 1997, 121). Cameron, Epstein & O'Halloran
argue that the creation of majority-minority districts "dilute[s} minority representation in
surrounding areas which may then elect representatives unsympathetic to minority
concerns." (Cameron, Epstein and O'Halloran 1996, 794). They posit that a trade-off
between substantive and descriptive representation is necessary and conclude that
majority-minority districts do not necessarily maximize legislative outcomes benefitting
minorities. (Cameron, Epstein and O'Halloran 1996, 808). They advise that in the South,
blacks would do best if districts were created with "slightly less than a majority of black
voters." (Cameron, Epstein and O'Halloran 1996, 809).
They also accuse the courts and
justice department for using a rule of thumb for majority-minority districts - 65 percent -
that has the effect of diluting minority strength. (Cameron, Epstein and O'Halloran 1996,
809-810). In the non-southern states, they advocate "distributing black voters equally
among all districts." (Cameron, Epstein and O'Halloran 1996, 809). Bullock points out,
"the impact of redistricting is apparent when the number of black voters removed from a
district exceeds the GOP victory margin. ... All districts held by Democrats in 1991 in
which redistricting reduced the black percentage by more than 10 points" fell to
Republicans in either 1992 or 1994 (Bullock 1995). According to Cain et al, "Political
geography and the [Voting Rights Act] give the Democrats a big edge in safe seats over
the Republicans. No plan, no matter who draws it, can change that." (Cain, MacDonald
and Hui 2006, 4). Obviously they are correct about the Democratic edge in safe seats,
but what about the much more important question of who has the edge in total seats, or at
least marginal seats?
Using the Brookes analysis produces some interesting results in looking at the
impact of majority-minority districts. For purposes of analysis, these states are classified
as having created majority-minority districts in the 1990s: Florida, Georgia, Illinois,
Louisiana, Maryland, North Carolina, New York, and Texas. An analysis for the decade
following the 1992 redistricting that incorporated the entire country proved inconclusive.
Results from some of the majority-minority states in which racial packing was confined
to small areas of the state (e.g., New York, Florida, Illinois) tended to water the results
for the majority-minority states. A more thorough analysis would break down these
states into areas with majority-minority districts and areas without (e.g., disaggregate the
New York metropolitan area from Upstate; disaggregate the Chicago area from
Downstate). A more concentrated study was made by considering only three medium-
sized states that had engaged most aggressively in racial packing in the 1990s: Georgia,
Louisiana, and North Carolina. Even though these states created only two or three
racially packed districts each, these districts had impacts over most of the congressional
districts of these states. Whereas the drawing of the earmuffs district had little impact
outside the Chicago city limits (indeed, it appeared to make no difference at all even in
the suburban district it bordered on its shoestring portion), the creation of the
gerrymandered 4th district in Louisiana had an impact on all of the districts outside the
New Orleans area. The Brookes method lets us look at the partisan impact of majorityminority districts in the states where we would expect them to have the greatest impact.
Although this can be construed as a biased way of looking at majority-minority districts,
it can also be seen as putting these three states under the magnifying glass.
For each of the five elections beginning with 1992, the magnitude of the bias on a
per-district basis is greater in the three states considered than in the non-majorityminority states (that is, the states not listed above as using majority-majority districts
from 1992 through 2000). The magnitude of the bias is also greater on a per-district
basis in the sum of the three states than in the other states that used majority-minority
districts in the period. In the decade, the three states of Georgia, Louisiana, and North
Carolina had a combined 30 House seats. For 1992, the combined bias in the three states
was 8.06 seats in favor of the Democrats. For the middle group of states (majorityminority districts other than the three considered separately), which had a total of 112
seats, the total bias was 20.59 seats in favor of Democrats. The rest of the states had a
total of 293 seats and had a total bias of 53.98 seats toward Democrats. Expressed on a
per-seat basis, this means bias in the three states was +0.26 per seat, in the middle group
was +0.18 per seat, and in the rest was +0.18 per seat. This means bias in the three states
was 1.46 times what it was in the non-majority-minority states and also 1.46 times what it
was in the other majority-minority states.
In 1994, with the Republican capture of Congress, total bias in the three states
turned against Democrats, with a total bias of 6.04 in favor of Republicans, or -0.2 per
seat. Meanwhile, the other majority-minority states turned in total bias of 6.81 seats in
favor of Democrats, or +0.06 per seat. The rest of the states had a total bias of 24.69
seats against the Democrats, or -0.08 per seat. The ratio of the three states to the nonmajority-minority states is 2.39. The ratio of the three states to the other majorityminority seats (in terms of absolute value of magnitude only, since the signs are different)
is 3.31. The patterns are similar in the final three elections of the decade. In each case,
the bias against the Democrats on a per-seat basis is higher in the three states than in the
states that weren't using packing tactics to create majority-minority districts. For all
years except 1992, a huge portion of the total bias in the country against Democrats in
House elections comes from just these three states. In absolute value, the bias against
Democrats in the three states was at least a fifth of a seat each and was an average of 0.28 for the four years. The average absolute value also remains 0.28 if 1992 is included.
From this limited Brookes analysis focusing on the three states most associated
with using racial packing to create majority-minority districts in the 1990s, the
conclusion is that the level of racial packing conducted in the three states led to electoral
bias against Democrats. This continued to be true even after the states were forced to
moderate their racial gerrymandering by the Supreme Court and lower courts over the
course of the decade.
Brookes analysis of Congressional voting in opposed races, by MM type
x
y
b
f
P
Q
R
S
C
D
U
V
G
CsV
A
TPV
1992 MM 1992 MID 1992 NON 1994 MM 1994 MID 1994 NON 1996 MM 1996 MID 1996 NON 1998 MM 1998 MID 1998 NON 2000 MM 2000 MID 2000 NON
57
139
10
57
144
10
59
143
19
66
173
12
59
134
11
148
53
20
148
55
20
152
55
19
158
52
18
119
46
11
19
66
173
12
60
134
11
57
140
10
57
144
10
59
144
11
46
120
18
52
159
19
55
153
20
55
149
20
53
149
190826
197391
219679
113279
142609
166276
195362
167017
200880
159619
130541
163861
216346
176624
213840
231292
229752
230477
148650
171705
176893
198396
208161
218035
168460
159601
174008
216360
236660
233013
425348
422612
428548
426017
420894
426380
423868
417532
425361
425633
420882
427312
426196
422612
427312
432857
433037
420552
430446
432527
419873
427827
430981
419182
426615
430981
419761
426615
429744
419065
230068
228989
198203
304253
282752
258152
225520
260295
219707
262993
296771
266811
206266
251924
13669
201565
203285
183619
281796
260822
242223
229431
222820
196228
258155
271380
248387
210255
193084
17262
2011
16084
9588
3
7367
3802
4111
5851
5157
3632
11696
4585
8025
8618
10171
627
12031
9851
0
8886
4530
5031
10083
7225
2153
7908
6195
9833
12840
11008
5.171601
0.41888
2.421384
0.050075
11.58118
0.866398
7.564621
0.57494
45.99821 -10.2379 -2.38829
-1.15143 0.458373 0.935051
8.250929 3.742168 9.449877
-0.12053 0.000108 -0.18411
-10.4077
0.478746
11.91956
-0.55369
-25.0406
-2.43026
13.71844
-0.76352
-10.8163 -9.31108 -13.8119
0.141636 0.478746 -1.93688
0.672384 10.82471 13.37241
0.053609 0.517035 -0.59044
-10.001 -10.4911
0.141636 0.156057
-0.14022 16.33088
-0.06216 -0.56122
0
0
1
Total
Per seat
8.06194
0.268731
20.58714
0.183814
53.97717
0.184222
Ratio
1.458733
2.387852
5.402845
49.40146
4.516354
Abs ratio
1.461976
3.308477
20.86332
14.801
6.912001
TPW
x
NUM SEATS WON BY PARTY A
y
NUM SEATS WON BY PARTY B
b
NUM OF SEATS WHERE A LEADS B
0
-34.0975 -8.23165
-1.99499 0.246043
10.67386 -0.01433
-0.27465 -0.03019
-1
1
0
0
0
-6.03726 6.812534
-0.20124 0.060826
-24.6933
-0.08428
-8.03013
-0.26767
1.436931
0.01283
-14.516
-0.04954
f
NUM OF SEATS WHERE B LEADS A
P
AVG NO OF MPVOTES WHERE A LEADS B
Q
AVG NO OF MPVOTES WHERE B LEADS A
R
AVG REG ELEC IN SEATS WHERE A>B
S
AVG REG ELEC IN SEATS WHERE B>A
C
AVG NO OF ABSTENTIONS WHERE A>B
D
AVG NO OF ABSTENTIONS WHERE B>A
U
AVG NO MINOR PARTY VOTES WHERE A>B
V
AVG NO MINOR PARTY VOTES WHERE B>A
0
1
-9.94866 2.509402
-0.33162 0.022405
0
-1.96685
-0.00671
0
-17.6041
-2.41017
-1.3847
-0.35964
0
0
-10.0617
5.43457
-0.33539 0.048523
-21.7586
-0.07426
Chapter 5 - The Brookes Method and the use of independent redistricting bodies
Politicians, if left to their own devices, do not act benignly
in the public's interest. (Ansolabehere and Snyder 2008,
271).
Redistricters ought to go for as much partisan advantage as
they can (while staying within federal guidelines, if
applicable) because courts are apt to use compromises in
formulating a plan anyway. (Babcock 1998, 121-22).
Introduction
Redistricting has been a mandatory part of the political process in the United
States since a series of court decisions beginning with Baker v. Carr(1962) holding that
reapportionment meant not only allocation of the seats among the states, as specified in
the Constitution, but also equal districting within the states. This has come to mean that
every unit in the United States that uses districts has fallen under a higher degree of
scrutiny than was previously the case. From the Progressive Era of the early 20th century
until the Baker era, there was little redistricting at all; some states had used the same
political maps since the 19th century. Most states made no changes to their
Congressional districts in the 1920s and then got out of whatever redistricting habit they
had. (The same pattern holds in most state legislative redistricting as well.) Only a few
states performed redistricting decennially, as has now become standard, and even
compulsory. Political science has disagreed on the political salience of redistricting, with
a large body of literature holding that redistricting is largely irrelevant to political
outcomes. (Bicker 1971, Bullock 1975,-Ferejohn 1977, O'Rourke 1980, Saffel 1982).
Other scholars have found that redistricting holds important effects. (Mayhew 1971,
Cain 1985).
The first sessions of redistricting after Baker were contentious. The
implementation of Baker in the 1960s was done as a matter of urgency; most of this was
reflected in districts used for the 1966 election, although a few states redrew their
districts in time for the 1964 election. Redistricting for 1972 was also contentious given
the air of upheaval prevailing in the national spirit at the time, and redistricting for 1982
represented a large shift of Congressional voting power from the Snowbelt to the Sunbelt.
Almost from the time of the post-Baker redistricting, a sense was felt in certain quarters
that politicians ought not be drawing their own district boundaries, and that the matter
should be instead left to a disinterested (if not truly non-partisan) panel. In the 1970s,
several states set up independent redistricting commissions, and a movement to establish
them was put forward in many others. In the 1980s, other states joined the movement.
Although it never spread in the manner of wildfire, by the 1990s, nine states - Arizona,
Connecticut, Hawaii, Idaho, Indiana, Iowa, Montana, New Jersey, and Washington - had
implemented some form of taking boundary making decisions out of the hands of elected
politicians and making an independent body responsible for line drawing. (In the case of
Iowa, the responsibility devolved to a committee of bureaucrats, which, although
technically subservient to the legislature, had achieved practical control over the state's
Congressional district boundaries.) Montana was reduced to a single Congressional
district for the 1992 election, and so it is excluded from this analysis. Karch et al found
that the imperative to establish independent redistricting bodies typically came from the
good-government movement and was often dependent on their ability to launch
initiatives, bypassing the legislature. In other states, voter desire for a commission was
fueled by the excesses of partisanship in redistricting. (Karch, McConnaughy and
Theriault 2007). Some scholars dredge up dark images of redistricting: "Changing only
the lines on a map and not a single vote, the People's voice can be dramatically altered.
Those who draw those lines can become master ventriloquists of the People's voice."
(Polsby and Popper 1991, 351).
The case for independent redistricting bodies usually centers on the inadvisability
of elected politicians drawing their own districts. Kang alleges, "the political motivations
of self-interested elected officials, those in charge of redistricting, are the problem. ...
Self-dealing incumbents can and do substitute their political interests as the overriding
priority for redistricting in place of any broader sense of the public good" (Kang 2006,
682-683). The public is aware of the propensity for redistricting mischief. Ansolabehere
and Snyder state it bleakly: "redistricting is widely viewed not as a corrective but as a
malignancy of American politics." (Ansolabehere and Snyder 2008, 265). Polsby &
Popper allege, "Gerrymandering violates the American constitutional tradition by
conceding to legislatures a power of self-selection." (Polsby and Popper 1991, 304).
This attitude was present at the start of the reapportionment revolution.
"Whenever the drawing up of the boundaries is left even slightly to the discretion of an
interested body, considerable latitude is left for the exercise of the art." (Vickrey 1961,
105).
In what was surely one of the most erroneous statements of the decade, Vickrey
speculated about the response to his proposed solution to gerrymandering: "Politicians
are going to be reluctant in most cases to desist from discreet gerrymandering, but it is
just barely possible that presentation of a scheme [to produce more equitable districts]
would develop sufficient support for its inherent fairness to overcome this reluctance."
(Vickrey 1961, 108-109). We can safely say that not only did politicians not desist
"discreet gerrymandering," but they took off onto frolics of the most egregious and
blatant kinds of gerrymandering the world has ever known.
Stem makes this observation about the practice of gerrymandering: "The goal of
gerrymandering is to create a districting plan that facilitates the retention of seats by
incumbents or allows the political group in power to enlarge the number of seats it
holds." (Stem 1974, 404). He notes that incumbent-protection gerrymanders are the kind
most likely to occur when the redistricting process is not controlled by a single party or
faction (Stem 1974, 404). "In almost every case ... gerrymandering creates safe districts.
This tactic diminishes effective representation by decreasing the number of politically
competitive districts and by reducing the effectiveness of the franchise through a
reduction in the number of voters whose vote can affect the outcome." (Stem 1974, 405).
The question addressed here is whether these independent boundary commissions
do redistricting differently from the politicians that proponents of commissions argue are
supposedly too close to the process to make such decisions in a neutral and effective
manner. Researchers have commented that non-legislative actors in redistricting are
given to avoid partisan political considerations and be more concerned with more neutral
principles including contiguity, compactness, and the ability of representatives to be
responsive to a district (Carson and Crespin 2004, 457). Morrill argues, "a non-partisan
commission is not and really cannot be politically neutral in effect, and in the attempt to
be 'mathematical' may well produce results which ignore the needs and interests of
people." (Morrill 1982, 367). McDonald argues that there are partisan and bipartisan
commissions and that partisan commissions act in predictable partisan ways, while
bipartisan commissions broker compromises in redistricting (McDonald 2004).
The partisanship chapter demonstrated that the Brookes method can be used to
show that not only is redistricting not irrelevant to electoral outcomes, but that
redistricting affects different forms of electoral bias in sometimes contradictory ways.
This chapter will use the same method to show the different impacts that commissions
have on electoral bias and district compactness in redistricting vis-i-vis state legislatures.
Data analysis
This analysis looks at each House race in 1996, 2000, and 2004 from states with
more than one House member (excluding thereby Alaska, Delaware, Montana, North
Dakota, South Dakota, Vermont, and Wyoming) in which there was at least a Democrat
and a Republican on the ballot. For 1996, 25 contests are excluded. For 2000, the
number of excluded contests rises to 67. For 2004, 70 contests are excluded. The higher
number of excluded cases reflects a conscious decision by both the Democratic and
Republican parties to no longer bother with losing House races in an expanded number of
states, including Massachusetts, California, Illinois, and North Carolina. In 1996, only
five of the 18 contests excluded because of underopposition were outside the South, but
by 2000 this figure had risen to 21 (of 60, meaning that underopposition had risen in the
South too).
An examination that focuses on the impact of redistricting on electoral bias is
most salient in the CSV category. The elimination of disparities in population between
districts is the main objective of redistricting. Indeed, when judges have been the ones to
implement redistricting, they have been known to draw districts as equal in population as
2-20 people to avoid any charge that they did not fix disparities. However, the Brookes
method as it is applied here looks at disparities not in population but in registered voters.
This emulates the function of proportional representation of equalizing actual voters
rather than population. Even so, one would expect the CSV bias to be smallest in
redistricting years, when the districts newly reflect the population as it was 2 1/2 years
earlier; and higher in the last year of a district map, when the census that created the lines
is more than 10 1/2 years old. (Jacobson & Kernell note that one confounding factor in
redistricting years is that when plans are late in being assembled, challengers are more
disadvantaged than incumbents because of the shortened window for parties and
challengers to decide where to focus their efforts (Jacobson and Kernell 1983, 101)).
The Brookes analysis of commission vs. legislative redistricting of Congressional
districts is a curious one. One would approach the question expecting that commissions
would foment less bias than legislatures in creating districts (being supposedly
disinterested persons) but the result is that on a per-district basis, commission-carved
districts created more bias than legislative-created districts in all three of the years under
review, and commission districts also created more bias on a raw numbers basis in two of
the three elections. This is a truly curious finding, because the N of non-commission
districts is a factor of 6-7 of commission districts in these three years.
In 1996, the sum of 53 commission districts was a bias of -5.492 against the
Democrats, while the sum of 357 non-commission districts was a bias of -1.664 against
the Democrats. In 2000, 52 commission districts produced an aggregate of -3.09 seats
against the Democrats, and 316 non-commission districts produced an aggregate of 1.633 seats against the Democrats. In 2004, 48 commission districts summed to a bias of
-3.015 against the Democrats, and 317 non-commission districts produced a total of -
8.233 seats against the Democrats. Some of these non-commission biases are a
negligible number of seats. The biases in 1996 and 2000 amount to only 0.5 percent of
non-commission seats, while the aggregate bias from commission seats in those years
were 10.4 and 5.9 percent, respectively. In 2004, while the raw magnitude of the bias
was higher for non-commission seats, expressed as a percentage, it is still considerably
less, accounting for 2.6 percent of non-commission seats as opposed to the bias from
commission seats that year of 6.3 percent.
Part of the answer may be deduced from looking at the breakdowns by bias
component. As is typical for Brookes analyses of the U.S. House of Representatives, the
G bias is large in favor of the party that won control of the House in the election (the
Republicans, in each of these cases), while the CSV bias is large in favor of the
Democrats. This pattern is preserved in each of these three elections for the legislativecreated seats. For the commission seats, however, the G bias in each of these three
elections maintains its pattern of going in the direction of the (Republican) winner of the
elections, but the CSV bias is not as strong to help the Democrats compensate. This
suggests the possibility that commissions are doing something to mitigate CSV bias; for
example, perhaps they are not packing inner city Democrats as aggressively as their
legislative counterparts do. The small number of states having commissions removes a
great deal of energy from this assertion; it happens that the states with commissions tend
to be more suburban dominated (e.g., Arizona, Connecticut, New Jersey) and the states
with inner cities large enough to have districts to themselves (e.g., New York, California,
Illinois) are still using legislative redistricting.
The notion that commissions reduce CSV bias is worth further study, insofar as
was noted in the partisanship chapter that CSV bias is less easily mitigated than G bias.
The late redistricting scholar Robert G. Dixon recommended a particular use of
the commission method of districting. He recommended instituting a bipartisan
commission model with a tie-breaker device because it "allows combining the population
equality principle with political realities and a better informed public scrutiny." (Dixon
1982, 10). He argued that the role of the commission should be to produce fair or same
treatment of all parties, or "neutrality in this special sense." As he put it, "The vice to be
avoided is differential advantage, one party over another, in the cause of manufacturing a
congressional delegation majority out of a reasonably predictable statewide minority of
the popular vote." (Dixon 1982, 11). However, his prescription for type of commission
might have been too specific for it to have any traction since he commented, "a
straitjacketed commission may be worse than no commission at all." (Dixon 1982, 18).
Commissions and Compactness
Compactness is described in detail in the subsequent chapter. Relating
compactness to the theme of the work of independent electoral boundary commissions,
the average Hill ratio of seven states using commissions as of 2006 is 1.793. (Hawaii is
excluded from this, because its archipelagic nature makes any measure of compactness
useless.) The average of states having more than one district and not using commissions
is 2.027. This suggests that commissions are more cautious about creating non-compact
districts than are legislatures. However, a note of caution can be added by analyzing the
same groups of states from 1960, before the reapportionment revolution. The seven
states currently using commissions had an average Hill ratio of 1.427 while the rest of the
states had an average of 1.56 1. This can be viewed one of two ways (at least). One is
that these states have a history of fashioning districts more compact than average; the
other is that the commissions have exerted a moderating effect in these states. (Looked at
mathematically, the commission states are 26 percent less compact than in 1960, whereas
the non-commission states are 30 percent less compact, which does not constitute a
severe moderating effect.)
In any case, the fact stands naked that in this age of extremes in redistricting,
commissions are producing more compact electoral districts than state legislatures.
Conclusion
If the key finding of the earlier chapter that G bias goes to the winner whereas
CSV bias is persistently in favor of the Democrats is correct, then it appears that
something is happening in commission states that reduces the natural advantage that
Democrats have from CSV while at the same time allowing the kind of large G bias that
has traditionally favored the winner of the election. Part of the answer could lie in the
approach that commissions take to their work. If they are, in fact, acting "disinterested,"
they may be unwittingly be bending over backward to accommodate political parties that
do not deserve as large a share as the commission is handing them. To contrast a recent
example, for 2002, the California legislature created 53 safe seats, 35 for Democrats and
18 for Republicans. This was the best deal Republicans could get under the political
situation at that time in California. Presumably an independent commission charged with
the same task would swing closer to a 50-50 distribution of safe seats. By doing so, the
mere existence of the commission would fundamentally alter the distribution of political
spoils within the state, of which the partisan bias of a Congressional seat is certainly one.
Commissions are also creating more compact districts than legislatures are. This
is no doubt another manifestation of them being or wanting to appear to be disinterested.
By being less willing to fashion non-compact districts (which might seem to be a political
perquisite), they are producing districts that are more likely to meet judicial standards of
objective measures of this particular district ideal.
The movement toward independent redistricting commissions seems to have
stalled. That makes this a good time in history to take a closer look at the partisan impact
commissions are having. The cursory examination of three recent elections herein seems
to suggest that commissions do not reduce electoral bias vis-A-vis the electoral bias that
the legislatures they replace in the redistricting process would themselves create.
Chapter 6 - Redistricting and compactness in the United States and Canada
Introduction
Compactness is the degree to which the spatial area of an object is related to its
center. Compactness is a criterion frequently considered by courts, perhaps because of its
simplicity: Any layperson can appreciate that a square or circle is more compact than an
elongated or sinuous shape. Ideals of compactness can be tied to the ideal of proportional
representation, which is the benchmark used in the Brookes method. "By purely
mechanical operation," Polsby & Popper demonstrate, "a compactness requirement tends
to inhibit gerrymandering. By inhibiting gerrymandering, in turn, one abets proportional
representation, not by fiat, but by empirical tendency." Schwartzberg alleges that the
heart of compactness is nothing other than a plain and simple notion of fairness
(Schwartzberg 1966, 444). He notes that a legislative committee early in the
reapportionment revolution defined compactness as the absence of any attempt:
1. To divide (a territorial unit) into election districts in an unnatural and unfair
way with the purpose of giving one political party an electoral majority in a
large number of districts while concentrating the voting strength of the
opposition in as few districts as possible.
2. To divide (an area) into political units in an unnatural and unfair way with the
purpose of giving special advantages to one group. (U.S. House of
Representatives 1965, 2).
This corresponds closely to what most people mean by gerrymandering,but it bears no
relationship to the mathematical measures for compactness that scientists have devised in
the past 200 years. At the beginning of the reapportionment revolution, Hacker defined
gerrymandering as "the art of political cartography." (Hacker 1964, 54). But this is
obviously an inadequate definition. Backstrom, Robins & Eller state that gerrymandering
has come to be known as "the excessive manipulation of the shape of legislative
districts." (Backstrom, Robins and Eller 1978, 1122). Like a lot of questions in the area
of redistricting, there is no consensus as to what constitutes excessive manipulation.
Much has centered on the idea of compactness, but even this much is not clearly agreed
upon by scholars and certainly not by activists in the field of redistricting. According to
Polsby & Popper, "The diagnostic mark of the gerrymander is the non-compact district."
(Polsby and Popper 1991, 302). Pildes & Niemi view the Shaw court as maintaining,
"oddly shaped race-conscious districts compromise the values of political integrity and
legitimacy." (Pildes and Niemi 1994, 502).
On the eve of the reapportionment revolution, Reock advised that a notion of
compactness is central to equal districting. "Without some requirement of compactness,
the boundaries of a district may twist and wind their way across the map in fantastic
fashion in order to absorb scattered pockets of partisan support." (Reock 1961, 71).
Compactness is centrally about shape. As Taylor wrote fairly early in the
reapportionment revolution, "Politicians, political commentators, and political scientists
have become fascinated by the shapes of electoral districts." (Taylor 1973, 947). Baker
makes the point that ideals in compactness will never be ideal for everyone: "A district
pattern of symmetrical squares, although conceivable, well can operate to submerge a
significant element of the electorate," although he doesn't say how voters are submerged
by square districts. "Furthermore, a benign gerrymander," he continues, "in the sense of
some asymmetrical districts, may well be required to assure representation of submerged
elements within a larger area. Shape requirements focus on form rather than the
substance of effective political representation." (Dixon 1982, 16) Dixon is correct about
this last point, of which there can be no doubt, but this is attributable in great part to the
fact that focusing on form can be made wholly objective while focusing on substance
requires at least some subjectivity and most likely a great deal of subjectivity.
The role of compactness as a criterion for fairness has been debated since the start
of the reapportionment revolution. According to Sickels, "Dragons, bacon strips,
dumbbells, and other strained shapes are not always reliable signs that partisan (or racial
or ethnic or factional) interests are being served, while the most regularly drawn district
may turn out to have been skillfully constructed with an intent to aid one party." (Sickels
1966) However, Morrill takes a nearly opposite view, arguing, "except in isolated
instances, it is quite difficult to gerrymander compactly." (Morrill 1981, 21). As Polsby
& Popper put it, "An ugly map implies that a human ambition of some kind, with
politically strategic ulterior motives, has been hard at work." (Polsby and Popper 1993,
652). Grofman argues, "violations of natural communites of interest, ill-compact shapes,
or excessive crossings of local jurisdictional boundaries can be seen as prima facie
indicators of gerrymandering." (Grofman, Toward a Coherent Theory of
Gerrymandering: Bandemer and Thornburg 1990, 40)
In an earlier article, Grofman
posited 12 markers of gerrymandering, of which the aforementioned constitute three.
(Grofman 1985, 117-18). Morrill defines nine characteristics of "poor quality
districting," of which eight constitute gerrymandering (Morrill 1990, 213).
However,
Dixon argues against the promulgation of a lengthy set of criteria for redistricting:
"Districting method is more important than districting standards." (Dixon 1982, 18).
Dixon alleges that population exactness and preserving municipal boundaries are at
loggerheads: "The extent to which population subdivision boundaries may be honored is
... an inverse corollary of the degree of population stringency required." (Dixon 1982,
17). However, Backstrom, Robins & Eller agree with the minority viewpoint of Justices
Powell and Stevens in Bandemer that excessive division of local boundaries is evidence
of gerrymandering, stating, "the needless splitting of subdivision lines is usually done to
achieve partisan advantage." (Backstrom, Robins and Eller 1990, 153). Baker predicted
in 1990 - with fairly good prescience - "dramatically contorted districts will likely be
prominent in most gerrymandering suits." (Baker, The 'Totality of Circumstances'
Approach 1990, 205). According to Morrill, "Compactness is not an end in itself, but
rather an operational aid in avoiding discriminatory gerrymandering against parties or
territories. Compactness is inherently preferable to irregularity, simply because compact
territories tend to have easier communication and greater internal cohesion .... It is easier
for the representatives and the represented to develop a mutual identity with the district."
(Morrill 1990, 214). Reock declines to prescribe a standard for compactness, insofar as
some districts must inevitably lack compactness due to the existence of panhandles,
indented coastlines, and islands (Reock 1961, 73-74). Even so, he says that these
exceptions should be rare rather than common, and "the mathematical degree of
compactness may be used to test the reasonableness of any districting act." He says that
districts falling in the bottom quarter by his measure (which involves circumscribing a
circle around the district) bear close examination and those rating least compact of all
"should be considered suspect until proven valid." (Reock 1961, 74). According to Stern,
a compactness standard is "no more likely to result in reduced effectiveness of
representation than any other objective standard for drawing district lines. Such a
standard is more administratively workable than the alternatives and also promotes
competitive elections by facilitating transportation and media access within the district."
(Stem 1974, 415). Compactness is also an element in VRA litigation, because one of the
prongs of liability requires proof that a reasonably compact minority district could have
been created. (Pildes and Niemi 1994, 528).
The earliest known measure of compactness was proposed in 1822 by German
geographer Karl Ritter for evaluating the shape of grains of sand. German mathematician
Christian Heinrich von Nagel posited a measure of compactness using only the perimeter
and area of the objects as inputs. This has been adopted by scholars in such diverse fields
as geography, mineral engineering, and psychology. Cox wrote one of the earliest
treatments of the problem in English. (E. Cox 1927). Schwartzberg introduced Nagel's
index of compactness to political science by proposing that it be used for purposes of
evaluating electoral districts. He stated that the value of a compactness measure is to
"restrict the latitude for manipulation of district boundaries toward [gerrymandering] and
reduce the number and magnitude of abuses." (Schwartzberg 1966, 448). Polsby &
Popper colorfully praised this measure by noting, "The Schwartzberg criterion measures
a gerrymanderer's self-indulgence as surely as a breathalyser measures a drunkard's."
(Polsby and Popper 1991, 349).
Scholars have often debated the merits of various measures of compactness.
Other methods of determining compactness that have been developed include the ratio of
the area to the area of the smallest circumscribing circle (Ehrenberg 1892); the ratio of
the area to the largest inscribing circle; the diameter of the largest inscribing circle
divided by the diameter of the smallest circumscribing circle (Haggett, Cliff and Frey
1966); the diameter of a circle of equal area divided by the diameter of the
circumscribing circle (Schumm 1956); the area of intersection of the object and circle of
equal area divided by the area of union of the object and a circle of equal area (Lee and
Sallee 1970); the ratio of the longest axis to the shortest axis; the variance in the length of
radials extending outward from the object's center (Boyce and Clark 1964); the
dispersion of unit of area around the center (Blair and Biss 1967). None commands a
consensus of users.
Some measures which take into account only the widest points of a district fail to
account for variation in the intervening points. Such a measure is demonstrated in Figure
1, where the measure is based on the longest line that can be constructed inside the
district, line AB. Line XY is the longest line that can be formed perpendicular to AB.
However, this measure does not take into account the line forms other than those; the
lines forming the perimeter of the district may be as straight as AY or they may be
indented and meandering, so long as they don't affect the two lines constructed or allow
them to be exceeded in magnitude.
A
Y
Figure
I
Source of Figures 1-3: Polsby & Popper
Another group of measures construct a box around the district and then analyze
the district with regard to the size of the box. However, Figure 2 shows three
hypothetical shapes that would all be evaluated the same using the boxlike method,
because they fit the same box. Obviously, these three hypothetical examples are not
equally compact. Therefore, the boxlike measure is not a workable one.
Figure 2
Another group of measures operates by circumscribing a circle around the district
and then inscribing another circle inside the district such that it crosses no line of the
district's perimeter. The ratio of these two circles is examined. If the ratio is the same
(i.e., the district is a circle), the ratio is 1:1. As the district becomes less circular, the ratio
increases. The circle method would properly detect the non-compactness of the three
hypothetical districts of Figure 2. Since the districts fit the same box, they would also fit
the same circle. But the circle inscribed within them would be tiny compared to the
circumscribing circle, and the district would be revealed to be non-compact. The
challenge for the circle method is shown in Figure 3. These three districts are very
compact except for the projection extending upward from them. Although the
projections are equal in size and thus in their actual impact on the compactness of the
district, the differing angles of the projection impact the size of the circumscribing circle.
The circumscribing circle measure is thus impacted disproportionately by this minor
deviation from compactness. Obviously, then, it misses something when it comes to
evaluating the compactness of an electoral district.
Figure 3
Perimeter and area measures thus emerge as the most sensitive measure of
compactness. A measure which takes into account the total perimeter of the district and
the total area of the district is affected by everything that affects the shape of the district.
The perimeter-area method also has the advantage of being easily computable using only
the perimeter and area of the district, two things which in the 21st century are easily
generated from mapping software as soon as the district is proposed or created. Thus,
the perimeter-area method is not only the most efficacious, it is the most easily computed.
Most perimeter-area measures compare the district to a circle. These perimeterarea ratios were not necessarily devised with political science objectives in mind; most
were devised for use in the natural sciences, for evaluating the shapes of rocks and grains
of sand or insect bodies. Circle-based measures have the disadvantage of finding no
analogue in the real world of electoral districts. There are no districts based on circles
(although it is a useful measure for rocks and grains of sand and the shapes of lakes), and
it would be impossible to create a set of circular districts, because having one circular
district would necessarily mean that the ones adjacent to it could not be circular. This
chapter proposes using the square as the ideal for district compactness, because while no
real-world districts are shaped like circles, some are nearly squares. Using the circle as
an ideal means that no district will ever attain a perfect score, but with the square as a
measure, some districts come very close. It is much more useful to know how square a
district is as opposed to how round it is. For example, the nearly perfect square of the
5th congressional district of Texas from 1932 into the 1960s gets a Cox ratio of 0.783; an
Attneave and Arnoult ratio of 0.466; and a Nagel ratio of 1.13 (these are all circle-based
measures). None of these measures immediately communicates to the user that this is a
square district.
The formula for the Hill Ratio is
P/4
A perfectly square district would have a Hill Ratio of 1, and as districts became
less compact, their ratios increase proportionately. The Texas district mentioned above
has a Hill Ratio of 1.001, which does more to tell the user it is almost a square than ratios
of 0.466, 1.13 or 0.783 do. Other examples of Hill Ratios are shown in the appendix.
Another method conceived of for measuring entire districting plans, rather than
individual districts, is the total perimeter length in the plan. Obviously, as districts
become more complicated in their shape, the total length of perimeter increases. This
method has the obvious disadvantage that it can only be used for evaluating plans in their
totality and not of evaluating individual districts. Also, it is disproportionately affected
by line lengths in large districts; a plan might be rated less compact because one or two
rural districts make use of a complicated boundary whereas a plan that is really less
compact makes a straight line between those rural districts while many irregularities are
constructed using the much shorter lines separating urban districts.
The area-perimeter method overcomes weaknesses in the other methods. For
example, those based on the ratio of the largest and smallest lines that can be constructed
inside the district have the disadvantage that they are unaffected by changes to the district
not involving those two lines. Similarly, the districts based on the size of circles that
circumscribe the district or can be inscribed in it are greatly affected by the creation of
tiny angles that extend into the inner core of the district or extend outward from the
perimeter. These tiny angles might be close to meaningless in the construction of the
district but greatly move its place in the measure. The Blair & Biss and Boyce-Clark
measures are affected by the size of the district and not merely its shape and so cannot be
used to compare districts of unlike size. Polsby & Popper say of the area-perimeter
measure, "it is so sensitive to any deviation that it is impossible to comfortably
gerrymander" (Polsby and Popper 1991, 350).
Also, the area-perimeter method has simplicity as a sizeable advantage. It uses
only the area and perimeter of the districts. These measurements are available as soon as
a districting scheme is proposed and geographic data (e.g., map boundary files) are made
available in digital form, or they can be easily calculated by digitizing a paper map or by
aggregating existing units (towns, census tracts, precincts, etc.) into the desired district
form.
Political scientists are often called upon to pass judgment on redistricting
proposals or even to create those proposals themselves. Compactness, in general terms,
is readily distinguishable to the human eye. The scientist's role is to add quantification in
order to establish that what is believed to be compact or non-compact is really so. While
various scientists can argue for one or other measure of compactness to be most reliable,
100
none can dispute the speed at which compactness ratios can be produced from a large
number of districts by the area-perimeter method. Thus, the area-perimeter method by
its ability to contribute to a speedy analysis of a pending districting scheme more than
makes up for any deficiencies alleged against it.
According to MacEachren, "compactness can probably be considered the single
most important aspect of geographic shapes." (MacEachren 1985, 65). Pounds declared
compactness second only to size in significance when it comes to evaluating countries.
He notes that compactness affects ease of travel, communication, and the homogeneity of
the population (Pounds 1972, 54-55).
As noted above, over the decades since redistricting plans came to be evaluated
by courts, political scientists and geographers have proposed multiple measures of
compactness. Some of the most readily accessible take into account only the area and
perimeter of the district. These measures relate the object to a circle, maintaining that a
circle is the most compact figure. Therefore, an object that has the same area as a circle
whose circumference is the same as the object's perimeter (i.e., it is a circle) would
receive a perfect score on such a measure, whereas a polygon that has more than a circle
of equal perimeter would receive a lower score. (Some of these measures are inverted, so
that a lower score is higher, or vice versa.) One such measure is the Nagel ratio, which
is two times pi times the district's perimeter divided by the square root of the district's
area divided by pi. Simply stated, it is the ratio of the circle formed from the district's
perimeter to the circle formed from the district's area.
The area-perimeter measures can be divided into two groups: Those that are
consistent with Cox and those that are consistent with the square root of Cox (Attneave
and Arnoult 1956). All of the Cox-consistent area-perimeter measures work the same in
terms of how they rank the same set of districts, and all of the square root based measures
do the same among themselves. Many of the variants employ an inverted form whereby
a low ratio means greater compactness.
The chief failing of the existing measures of the area-perimeter ratio is that they
are based on circles. There is no electoral district in the world based on a circle.
Therefore, all districts will fail to meet the ideal represented by the circle. Harris notes
that circles are the most compact geometric figure, but it would not be possible to
redistrict an entire political unit using only circles. "If one or several districts were
formed into circles, the remaining districts would be odd-shaped ... A few circular
districts would mean a sacrifice of compactness in other districts." (Harris 1964, 220).
Meanwhile, actual square districts exist but are underrecognized for their innate
compactness due to the use of pi to create existing compactness measures. The
arbitrariness of the circle ideal is reflected in the scales used for the area-perimeter
measure. The example of the nearly square 5th district of Texas from the 1930s into the
1960s given above illustrates this.
The Hill ratio is the district's perimeter divided by four, divided by the square
root of the district's area. Simply stated, it is the ratio of the square formed from the
district's perimeter to the square formed from the district's area. Since squares and
circles based on the same measures are definitively proportional, the Hill ratio can be
expressed as the Nagel ratio times 0.886227. The 5th district of Texas mentioned above
has a Hill ratio of 1.001. A scale that incorporates 1 as a perfect square and larger
numbers being proportionately less compact is much more useful to a consumer than a
102
scale that is purely relative, especially if the scale is inverted so that a higher number
indicates greater compactness.
In testing the Hill ratio against other measures of compactness, the author
digitized congressional district maps from the period before 1992 by aggregating
counties. Counties were the principal building blocks of congressional districts until the
1960s when states, facing the necessity of adhering to strict standards for
malapportionment, began to use other areas as atomic units, including precincts and even
blocks. For 1962 and earlier, it is possible to reproduce most districts digitally merely by
aggregating counties (or in the case of New England, towns). For this purpose, the maps
of the post-1960 redistricting and the pre-existing redistricting contained in a volume
issued by Congressional Quarterly were used. (Congressional Quarterly 1962). For the
mid-i 960s redistricting necessitated by Baker and its progeny and for the 1970s and
1980s, many map sources are available. For the period after 1990, complete digital
district maps of the country are available.
The most compact district in the U.S. as of 2009 is the state of Wyoming.
Compared to a circle, its Nagel ratio is 1.137. Compared to a square, its Hill ratio is
1.008. It is much
more useful to know
how square
Wyoming is as
opposed to how
round it is.
The
most uncompact
103
district in U.S. congressional history, the 29th district of Texas created for the 1992
election, has a Hill ratio of 10.498.
It has been described as Pegasus the flying horse
dragging a dead lion. The degree to which Texas legislators used city blocks as atomic
units of the district is obvious even in the silhouette. This is a far cry from the early 20th
century when counties were the atomic unit of congressional districts not only in Texas
but in most of the country. The reapportionment revolution dictated that a smaller atomic
unit be used - initially it was the municipality and then the election precinct - but as
gerrymanderers have gotten more bold, the city block is now the basic unit. One can only
speculate if individual parcels of property will soon become the atomic unit, with
politicians choosing not only which neighborhoods and blocks they will represent, but
also which households.
Morrill says that districts with Nagel ratios of 2 or greater ought to be considered
suspect, and justification for them should be demanded (Morrill 1981, 22). He makes the
point that extremes in non-compactness should be ipso facto suspect because
gerrymandering is the main reason someone would go to the trouble of fashioning an
extreme district (Morrill 1987, 249). Pildes & Niemi note that an extremely uncompact
district calls attention to itself and invites stricter scrutiny of itself (Pildes and Niemi
1994, 575). Indeed, such a district attracts stricter scrutiny not only for itself but for the
entire plan.
A Nagel ratio of 2 corresponds to a Hill ratio of 1.77. As of the 2006 election,
192 Congressional districts have Hill ratios of 1.77 or less, and 244 exceed this threshold.
In other words, more districts fail Morrill's suspect test than pass it. Schwartzberg's
standard of Hill 1.48 (Nagel 1.67) would cause even fewer districts to pass scrutiny.
104
Dixon argues, "A rule of compactness and contiguity, if used merely to force an
explanation for odd-shaped districts, can have much merit." (Dixon 1968, 460). Niemi
et al state that a compactness measure should only be used to compare districts, and that
there should be no particular score that constitutes an unacceptable level of noncompactness. (Niemi, Grofman, et al. 1990, 1176).
Pildes & Niemi allege that majority-minority districts would disproportionately
suffer from any rule that mandated a particular compactness measure (Pildes and Niemi
1994, 567). However, their noting of this tendency suffers from a failure to test these
districts against any objective compactness standard; their evaluation stemmed from only
looking at current outliers.
It was not always the case that so many states ignored notions of compact
districts. The author has determined compactness ratios for a set of 1899 districts in
existence back to 1922 (some of which were created in the 19th century). For the 1990s
and 2000s, digital boundary files are available for the whole country. For the earlier
periods, digitized congressional district boundaries were created by aggregating counties
into districts. This was non-problematical in most instances because counties were in fact
the units of aggregation for congressional districts until the reapportionment revolution.
In a fairly small number of instances, the digitized districts are somewhat different from
the real districts in terms of those districts that were essentially groups of counties with a
piece of an urban county attached to them. In these cases, the size of the urban
component is small relative to the total area of the district, and this variation in the
composition of the district makes little or no difference in the total area and perimeter of
the district. For some states, the deviations these variations would have made were
105
unacceptably large, and so for some states in some periods, only a partial set of digitized
districts exist. A table of these districts is included as an appendix to this chapter. In an
inversion of the usual pattern of data availability, the gap in the data is not in the early
part of the 20th century but the mid-I 960s to the 1980s. This is because when the
Supreme Court mandated an equipopulation standard, states began fragmenting counties
as units of aggregation for congressional districts. From then until the 1990s, when
systematic, nationwide digital boundary files became available, the ability of the author's
method to generate digital boundary files is small. As a result, the dataset contains a
more complete set of districts from the 1930s than from the 1980s.
As of the 1960 election, the average Hill ratio of all districts for which data are
available was 1.526. As of 1982, it was up to 1.676; then to 2.146 in 1992, and since
then the average has declined slightly to 2.103. However, the decline from the 1990s to
the 2000s was largely due to a reduction of the excesses of 1990s racial and partisan
gerrymandering in four states: North Carolina, Georgia, Louisiana, and Texas. If these
states and Maryland are excluded, the change from 1992 to 2006 is negligible. The trend
is for less compactness with each redistricting cycle, rather than more. Carson et al
argue, "redistricting plans, for the most part, create districts that are more extreme
relative to previously drawn seats." (Carson, Crespin and Finocchaiaro, et al. 2007, 884).
Pildes & Niemi state, "there is no denying that the present congressional districts
are less compact than those they replaced." (Pildes and Niemi 1994, 573). They identify
three factors in causing the sharp increase in non-compactness of congressional districts:
1) Post-Daggett decisions to strictly equalize congressional district populations
intrastate;
106
2) New computer technologies that let actors be more sophisticated with their
district manipulation;
3) Post-Gingles decisions to create more districts favorable to minorities, or as
they also suggest, to use creation of majority-minority districts as an excuse to
ignore traditional geographic principles in redistricting (Pildes and Niemi 1994,
574-575).
There is considerable variation among states in the compactness ratio. The eight
districts of Minnesota have an average compactness ratio of 1.594. The eight districts of
Maryland have an alarming average of 4.659. (Even excluding three complicated
districts in eastern Maryland for which redistricters should possibly be given some
latitude, the average is still above 3.) Necessarily, the districts including the Eastern
Shore of Maryland and Cape Cod will always be uncompact, but this does not justify
making all of the other districts in Maryland and Massachusetts equally irregular.
The relative inequalities in compactness between states is also not a function of
the composite units of the states with low compactness averages being relatively square.
After the Northwest Ordinance, Congress mandated that the new states Ohio and west
would be surveyed using a regular square measure, a notable improvement over the
metes-and-bounds system that was used to establish everything from neighborhoods to
counties in the thirteen original colonies. The averages of square-based Indiana (1.774)
and Oklahoma (1.779) are practically indistinguishable from that of metes-and-boundsbased Connecticut (1.781) and haphazard Kentucky (1.759).
The aggressiveness of
redistricters in various states is aptly reflected in the compactness ratios. Oddly enough,
many states that were outliers as early as 1922 (North Carolina, Maryland, California)
107
continue to pursue extremes in redistricting today. Nevertheless, not everyone believes
that states reaching new extremes in non-compactness means automatically that
democratic virtue is being compromised. "The contours of district maps today more
closely represent the average voter's preferences than they did fifty years ago."
(Ansolabehere and Snyder 2008, 272). This may simply be a veiled way of saying that
modem districts are partitions of partisanship.
Canada has not yet seen the proliferation of partisan and racial redistricting that
the United States has experienced. The national average Hill ratio for all ridings except
those classified as "excluded" (The territories, Labrador, the Sable Island portion of
Halifax riding, and discontiguous portions of two Quebec ridings) was 1.399 in 1997 and
1.400 in 2004.
This is lower than the national average for the United States has been in
the entire period studied herein, 1922 to 2006. At its low point before the
reapportionment revolution began, the U.S. national average was 1.526.
In the current 2004 remap of Canada, New Brunswick has the highest ratio, 1.547.
Quebec is second at 1.501. Thus, the two Canadian provinces with the least compact
federal ridings are approximately on a par with the U.S. state with the fifth most compact
congressional districts. Newfoundland and Labrador (although Labrador riding is
excluded from this analysis) has the most compact ridings in Canada, with a ratio of
1.238. Disaggregating the ridings further shows that the 18 ridings of Montreal Island
have an average ratio of 1.315; the 23 ridings of the city of Toronto have an average of
1.260, illustrating that urban ridings in Canada tend to be more compact than rural
ridings. The averages for Winnipeg (1.400), Calgary (1.236), Edmonton (1.347), and the
108
British Columbia Lower Mainland (Greater Vancouver, 1.222) confirm this. The nine
principally rural ridings of northern Ontario, by contrast, have an average ratio of 1.552.
One spot in Canada where redistricting has been contentious in the last two
rounds is New Brunswick. As stated, the province has the least compact federal ridings
in Canada, on average. The province's ratio increased from 1997 to 2004 by 0.126, the
largest increase in Canada. Another province where redistricting was controversial in
1997 and 2004 is Saskatchewan, but the change in this time is negligible, and in fact the
province's ridings overall are more compact than either of its provincial neighbors,
Manitoba and Alberta, where federal redistricting has been reasonably non-controversial.
The issue in New Brunswick prior to the 2004 remap was that districters attempted to
pack all of the Canadian Indian
reserves in the province into the
Miramichi riding. In Saskatchewan,
the issue before the 1997 remap was
that districters had done a textbook job
of diluting the urban voters of the
province's two major cities, Regina
and Saskatoon, by putting them each
into four urban-rural fringe districts
that ran from the inner cities of each
Source: Mondopoliticocom from Elections Canada base
map
out for a considerable distance (as far
as 120 miles) into agrarian territory.
(Indeed, the remap practically quartered the two cities and their surrounding countryside.
109
This act illustrates that it is possible to gerrymander compactly.) Community-ofinterests theory in redistricting holds that Regina and Saskatoon should each have as
many districts to themselves as would fit, and that at most one district would be extended
beyond the municipal or metropolitan boundary.
The small change in averages from 1997 to 2004 (0.001 nationally) suggests that
redistricting is not viewed by Canadian politicians as nearly as great a political
opportunity as is the case in the United States. Canadians are increasingly paying
attention to both intraprovincial and interprovincial malapportionment. In Ontario, the
federal riding of Kenora has about half of the population of the largest riding in that
province. Canadian provinces are protected from losing seats in the House of Commons
by virtue of the 1982 constitution that provides that no province can ever have fewer
House seats than it had at the time the constitution was promulgated. Because of this, in
reapportionment years, provinces that are increasing in population gain seats, but
provinces that lose population do not lose seats. Although these compactness data only
cover two maps and one round of redistricting, the low ratios point to no past tradition of
gerrymandering in Canada.
110
The same cannot be said for the United States, where a district created under the
leadership of Gov. Elbridge Gerry of Massachusetts led to such notoriety that the entire
practice of manipulating districts has been named for him. Most have never seen an
accurate representation of
the district. A cartoon
image of the district drawn
in the likeness of a
salamander is familiar to
many students of
redistricting. However, an
examination of the map
Map by author
reveals that much of the
salamanderlike shape of the district came not from gerrymandering but from natural land
forms, such as the state boundary and the coastline.
The southern border of the
panhandle is the Merrimack River, and the northern border is the line three miles north of
the Merrimack fixed as the boundary when New Hampshire was created as a separate
colony (Stein 2008, 137). Using a major river as a boundary was certainly a defensible
decision for the time. The district is essentially a backward seven, and would be largely
unremarkable if fomented on a modern election map. Although extreme for its time, the
gerrymander district has a lower Hill ratio than 32 congressional districts currently in use
in the United States, in more than half a dozen states. (Elbridge Gerry's famous
salamander was not the first reapportionment controversy in American history. The very
first presidential veto in 1792 was over a reapportionment bill. (Hayes 1996)).
In summary, compactness has declined precipitously in most of the United States
over the period of time in which redistricting has been a regular happening. The same
has not happened in Canada, with the Canadian average being below what the U.S.
average was even in 1922. Canadian districts are also more uniform in their compactness
ratios from sea to sea than American districts. American redistricters continue to create
non-compact districts, often to extremes, apparently subject only to the limits that courts
will accept. The tendency to aggressively seek to increase vote share through
redistricting has not come to Canada except in rare instances.
The clash over compactness is part of a larger issue. Geography has to always
play an important part in redistricting, and compactness is one of the most fundamental
and (depending on the measure one uses) objective criteria for geographical standards in
districting. Pildes & Niemi argue that interest can never completely trump geography in
districting decisions, lest the process be "reduced to a single-dimensional process in
which interest appears to dominate overwhelmingly." (Pildes and Niemi 1994, 504-505).
While electoral districting will probably never be based solely on geographic
considerations alone, attempts to quash the compactness standard demonstrate what is apt
to happen when politicians decide to leave geographic factors completely out of the
redistricting process.
112
National Compactness Average
2.400
2.200
2.000
1.800
1.600
1.400
1.200
1.000
1922 1932 1960 1962 1972 1982 1992 2006
113
--
USA
---
Canada
COMPACTNESS OF CONGRESSIONAL DISTRICTS 1922-2006
Change Change Change
CD
CD
CD
CD
CD
CD
CD
CD
1992-
1960-
1922-
State
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
1922
1.676
1932
1960
1.694
1.414
1.648
1.569
1.765
1.378
1962
1972
1.630
1982
1.545
1.758
1.735
2006
2.221
1.743
1.775
2.288
1.756
1.781
2006
0.085
-0.136
0.007
0.601
-0.154
0.124
2006
0.527
0.328
0.127
0.720
-0.010
0.403
2006
0.544
1.414
1.604
1.501
1992
2.136
1.878
1.768
1.687
1.910
1.657
Florida
1.711
1.549
1.596
2.596
2.419
-0.177
0.870
0.708
Georgia
1.504
1.517
2.349
1.779
-0.570
0.262
0.275
1.500
1.564
1.403
1.776
2.011
1.711
2.195
-0.065
0.184
0.147
0.793
0.695
1.817
1.396
1.442
1.890
3.636
1.774
1.497
1.495
1.759
2.102
-0.043
0.102
0.053
-0.131
-1.533
0.336
0.215
0.251
-0.006
0.427
0.390
0.161
0.194
0.043
2.164
2.212
2.110
4.659
-0.055
2.447
0.260
2.561
0.126
2.660
2.488
1.530
2.213
1.651
-0.276
0.121
-0.077
0.226
0.144
Idaho
Illinois
1.877
1.605
Indiana
Iowa
Kansas
Kentucky
Louisiana
1.384
1.336
1.301
1.716
Maine
Maryland
1.984
1.999
Massachusetts
Michigan
1.507
Minnesota
1.442
Mississippi
1.615
Missouri
1.467
1.516
1.449
1.299
1.250
1.643
1.524
1.717
1.366
1.438
1.282
1.244
1.766
1.676
1.288
1.279
1.811
1.850
2.099
1.958
2.038
2.290
1.425
2:240
1.310
1.424
1.545
1.594
0.049
0.284
0.152
1.598
1.668
1.580
2.015
1.896
-0.119
0.228
0.281
1.512
1.498
1.655
1.440
1.616
1.748
0.133
0.250
0.282
1.384
1.506
0.086
0.144
0.876
0.764
1.425
1.392
2.200
1.459
1.236
1.754
2.340
1.373
Montana
1.453
1.464
0.011
Nevada
1.549
1.797
0.247
New Hampshire
New Jersey
1.991
2.255
1.959
2.322
-0.032
0.067
Nebraska
1.320
1.352
1.378
1.557
1.446
1.412
New York
1.323
1.528
1.505
North Carolina
1.807
1.848
1.828
North Dakota
1.351
Ohio
1.293
1.355
1.862
Oklahoma
Oregon
1.477
1.423
1.640
1.424
1.930
1.775
1.822
1.979
1.654
1.971
1.292
New Mexico
South Dakota
1.825
1.628
1.535
1.797
1.531
1.481
-0.050
2.407
2.010
-0.397
0.482
0.687
3.814
2.721
-1.092
0.873
0.914
1.927
0.065
0.572
0.634
1.779
1.774
-0.150
-0.001
0.140
0.350
0.303
0.351
2.390
1.949
0.568
-0.030
0.995
2.406
1.904
-0.502
0.408
2.202
2.215
0.013
0.773
3.490
2.143
-1.348
0.680
1.172
1.413
Pennsylvania
Rhode Island
South Carolina
-0.101
0.683
1.395
1.373
1.283
1.581
1.496
1.605
1.250
1.288
1.399
1.442
1.675
Tennessee
0.581
1.561
1.463
1.493
1.491
-0.002
0.075
Virginia
1.871
1.898
1.811
2.407
2.219
-0.188
0.408
0.347
Washington
West Virginia
Wisconsin
1.432
1.998
1.429
1.464
2.016
1.464
2.036
1.377
2.180
2.035
2.018
1.787
2.352
1.649
1.721
2.303
1.726
-0.066
-0.049
0.076
0.258
0.267
0.348
0.289
0.305
0.297
Grand
1.535
1.527
1.526
1.551
1.575
1.676
2.146
2.103
0.577
0.568
Texas
1.462
0.323
1.760
1.415
Utah
_
Source: Tony L. Hill, Redistricting and Compactness
in Canada and the United States, presented at CPSA 2009 tlh@alum.mit.edu
Average Compactness by Province and Region of Canada
diff notes
-0.076 excl Labrador
Province
Newf
1997
1.314
2004
1.238
PEI
1.247
1.266
0.019
NS
1.317
1.438
0.121 excl Sable Island
NB
1.421
1.547
0.126
Que
1.499
1.501
1.327
1.315
-0.012
1.335
1.333
-0.002
Montreal Isl
Ont
0.002 excl Iles-de-la-Madeleine, non-contig parts of Manic
N. Ontario
1.400
1.552
0.152
Rest of Ont.
1.286
1.313
0.027
Toronto
1.315
1.260
-0.055
Subn Toronto
1.251
1.231
-0.020
1.415
1.367
-0.048
1.418
1.400
-0.018
Sask
1.258
1.264
0.006
Alta
1.357
1.385
0.028
1.214
1.236
0.022
1.330
1.347
0.017
1.501
1.447
-0.054,
1.285
1.222
-0.063
1.399
1.400
Man
Winnipeg
Calgary
Edmonton
BC
Lower Main
Natl
0.001 excl excluded
Source: Redistricting and Compactness in Canada and the United States
by Tony L. Hill, tlh@alum.mit.edu
INVENTORY OF DISTRICT COMPACTNESS DATABASE
P 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90
State
PpPPPPPPPPPPPPPPPPPPPP
xxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Alabama
Arizona
I
I
XXXXXXXXXXXXXXXXXXXXXXXXXX
I
....-o"'XXXXXXXXX)
XXXXXXXI'0 ..
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxx
Arkansas
PPPPPPPPPPPPPPPPPPPPPP jPVPVPVPVPVPV
California
IPPPPPPPPPPPPPPPPP
I
lxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Colorado
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Connecticut
jXXXXXXXXXXXXXXPPPP
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX I
Florida
xxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Georgia
IAAAAAAAXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
Idaho
IPPPPPPPPPPPPPPP
pppppppppppppppppppppppppppppppppppppppppp
Illinois
Indiana
FxxxxxxxxxxxxxxxxxxxR7 ]fPPPPPPPPPP xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxXXXXXXXXXXXXXXXX)XXXXXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxXXXXXXXXXXXXXXXXAAA
Iowa
ppppppppppp
xxxxxxxxxxxxxxxxxxxxxXXXXXXXXXXXXXXXX)XXXXXxxxxxxxxxxxxxxxxxxxxxxxpppp
Kansas
XXXXXXXXXXXXXXXXXXXXXXXIAAXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXX
Kentucky
I
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Louisiana
Maine
lxxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvVVVVVV)PPPPPPPPPpppppp
Maryland
I
I
I
Massachusetts
-XXXXXXXXXXXXXXXXXXXX,
XXXXXXXXXXX4 XXXXXXXX,
1
1
PPPPPPPPPPPPPPPPPPP PVPVPVPVPVPVPVPVPVPVPVP PPPPPPPPPPPPP
Michigan
VVVVVVVVVVVVVVVVVVVVV AA PPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP
Minnesota
XXXXXXXXXXXXXX
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Mississippi
1PPPPPPPPPPP
PPPPPPPPPPPPPPPPPPPPPPPP JAA PPPPPPPPPPPPPPPPPPPP jPPPPPPPPPPPPPPP
Missouri
Montana
JAAAAAA XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Nebraska
AAAAAAA
115
jjjj
1111,!,
111111!1111111
411112
1AAAAAA 11,41
Nevada
New Hampshire
I
I
i
I
PPPPPPPPFHFFFFFPPPPPPPIPPPPPPFFFHFFFPPPPPPPFFFFVFPPPPPPPPPPP
New Jersey
I
AAAAAAAAAAAAAAAAAAAAAA NAAAAA
XXXXXXXXXXXXXXXXXX)XXXXXXXXXXX)
New Mexico
jPPPPPPPPPPPPPPP
PVPVPVPVPVPVPVPVPVPVPVPVPVPVPVPVPVPVP
New York
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
North Carolina
IXXXXXX", , , "","XPPPPPPFFIFF
I
I
XX
XXXXXXTXXXXXXXXXXXXXXXXI
North Dakota
pppppppppppppppppppppppppppppppppppppppplpp IPPPPPPPPPPPPP
Ohio
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxlxxxxxxxxxxxxxxxxxxxxxx
Oklahoma
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Oregon
PPPPPPPPPP
Pennsylvania
PPPPPPPPPPPPP
PPPPPPPPPPP,
Rhode Island
I
I
I
I
I
I
I
1PPPPPPPPPPP PPPPPPPPPP
X = complete
P = partial
V = varied
PV = partial and varied
A = all reps at large
objel le sdai Ile = V
P91JBA
pue leiped = Ad
P,91JeA = A
lei:ped = d
9191dwoo = X
06 88 98 V9 Z9 09 9L 9L VL ZL OL 99 99 V9 Z9 09 99 99 V9 Z9 09 9V 9V tV Zt OV 9C 9C VC ZC OC 9Z 9Z VZ ZZ OZ 81, 91, ti, ZI, d
IdddddddddddddddddddddddddddddddddddI dddddddddddddddddddddd
xxxxxxxxxxxxxxxxxxxxxx kxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxx
dddddd
ddddddddddddddddddddddddddddddddddddddddddddddddddd
xxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxx 'XXXXXXXXXXXXXXXXXXXXxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
ddddddddddd XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
xxxxxxxxxxxxxxxxxi
I
ddddddddd7dd
dddd
I
rvvvvvvvvvxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx:xxxxxxxxxxxxxxxxxxxxxxxI
ldddddddddxxxxxxxxxxxxxxl xxxxxxxxxxxxxxxxxxxxxxxxxxxXXXXXXXXXXXXXXXXXX<XXXxxxxxxxxxxxxxxxxxxxxxx
Glels
uisuooslm
eiu!bJIA IS9M
U016ulqsem
MOM
win
SeXGJL
qqssquuqjL
ejoNeC] qjnoS
euilojeo LjjnoS
3SVE3ViVG SS3NiOVdNOO IMUSIG =10 MAOiNEIANI
Chapter 7 - District Compactness and Voter Knowledge:
Information Heuristics Through Favorable Partitioning
Having developed a workable and meaningful measure for compactness, the
question arises as to the utility of compact districts. Scholars have often argued that
compactness has no real impact on electoral districts. Polsby & Popper note that the
harm gerrymandering causes "is easier to characterize than to prove" (Polsby and Popper
1991, 304). Backstrom, Robins & Eller claim, "It is, in truth, hard to develop a powerful
case for the intrinsic value of having compact districts..." They argue that the only one
who benefits from compactness is the representative herself, and then only if she lives at
the center of the district (Backstrom, Robins and Eller 1990, 152). They concede,
however, that compactness has "a symbolic virtue" insofar as the esthetic value of
compact districts serves as a cue to the public. "Crooked districts," they argue, "lead the
public, often correctly, to suspect crookedness by someone manipulating the districting
process in order to gain unfair advantage." (Ibid.)
Fenno relates a telling anecdote from Rep. Gary Ackerman (D-N.Y.). "I don't
know half the time whether I'm in or out of the district. Neither do my constituents. They
argue among themselves." Ackerman and his constituency aides have been known to
argue the question too: "'Are we in the district now?' 'I think so.' 'No, I don't think so.'
'This must be the dividing line. I don't know, but it must be along here somewhere. It's
crazy."' (Fenno 2000, 169, Yoshinaka and Murphy 2009, 451). If the Member of
Congress and staff members based in the district and charged with dealing with
constituent matters don't know where the district lines are, how can voters be expected to
118
do any better? It is tautologous that districts with clearly defined boundaries have clearly
defined boundaries; these can be obvious to everyone involved.
Past research on this question has not been particularly detailed with regard to the
issue of district compactness. Niemi, Powell & Bicknell analyzed 1978 NES data with
regard to the number of representatives overlapping in a community (i.e., metropolitan
area) and found that voters who lived in communities with only one congressional district
were more likely to know information about the incumbents and candidates than those
from communities with more districts (Niemi, Powell and Bicknell 1986, 192). They
found that congruence between the district and media market made it much more likely
that voters would know about challengers in the congressional election, but the difference
was much more scant when it came to knowing who the incumbent was (Niemi, Powell
and Bicknell 1986, 196).
Kramer analyzed a particular New York state assembly district that leaped from
Staten Island to Lower Manhattan and found a large friends-and-neighbors effect in terms
of voters from one island or the other siding with political candidates from the same
island (Kramer 1990). If this is true in a state district (much smaller than a congressional
district), the expectation would be that it is even more true in a large U.S. House district.
Stem argues that a successful representative must be integrated fully with his or
her districts. "To represent fully the concerns of his constituency, a representative must
not be merely an agent whose opinions match those of some ideological majority in his
district, he must be attuned to a diversity of interests." (Stem 1974, 400). On the other
hand, rather than merely serving the district access problems of the Member, compact
districts facilitate greater voter knowledge of the district. A poor district functions only
119
as an electoral district. A good district that is formed compactly from regular boundaries
has purposes other than as an electoral district. A well-formed district is apt to function
as an economic area, tourism region, or as a quadrant of a particular state. A compact
district, such as the current 6th congressional district of Michigan, can
be described as "the southwest corner of Michigan"; or "the Kalamazoo
*
and Benton Harbor area." There is no significant quarrel with this
designation. People in the district get most of their information from
media in those cities. A non-compact district, however, such as the current 4th district of
Massachusetts (or nearly any of that state's others) can only be described as "the 4th
district." The people of the 4th really have nothing in common except that they live in
the district. It does
not conform to any
c Zfeconomic or
transportation
function. The district
awkwardly overlaps
disparate media
markets. The district
00
includes two
I I
prominent Boston suburbs, Brookline and Newton, and then snakes westward, then
southward, and then eastward to take in the disparate communities of Taunton and New
Bedford. The district includes a minority of the city of Fall River. This is not a unified
district by any measure. A large share of the residents of Brookline and Newton have
120
never been to New Bedford and perhaps have only a vague idea where it is. The daily
concerns of these largely white-collar suburbs of Boston have little in common with the
maritime New Bedford and old-era industrial cities of Taunton and Fall River. People in
the 4th do not read the same newspapers. The newspaper editor in New Bedford, of
course, focuses largely on the 4th because that is where nearly the entire newspaper
readership resides. The media in Boston have many other districts to cover - some of
which lie entirely within the Boston area - and cannot devote much coverage to the 4th.
The newspaper editor in Fall River must contend with allocating most congressional
coverage to the 3rd district, which includes much more of the Fall River area than does
the 4th. The 3rd district is only slightly easier to comprehend than the 4th. It reaches
upward from Fall River to contain Massachusetts's 3rd largest city, Worcester, although
it does not include all of Worcester's suburbs. Some of them are included in the 2nd
district, a congressional district dominated by Springfield, the second largest city in
Massachusetts, although the 2nd does not totally dominate Springfield; a good share of
metro Springfield is in the 1st congressional district.
121
In Illinois, the eight Downstate districts (11-12, 14-19) are all very non-compact.
Race is a fairly small consideration in Downstate congressional redistricting since there is
no opportunity to create a majority-minority district there. Rather, these compromises of
compactness are caused by political gerrymandering. Downstate Illinois is a fairly
Republican area; only the 12th district, including the heavily Democratic East St. Louis
area, contains a compact Democratic area. Essentially, the Democratic legislature
devised several districts with long tentacles
reaching into Democratic areas of small and
medium-sized cities (Springfield, Decatur,
Bloomington, even Parma and Sterling) in
order to advantage Democrats running in the
11th and 17th districts. This has been a
successful strategy for them, as the party has
been able to hold the 17th and won the 11th
when it opened up in 2006. (The Democrats
also won the 14th district in a special election
in 2008 after Speaker Dennis Hastert retired.)
These districts are much more suspicious-looking than those in use at the time it was
routinely alleged that Illinois politics was controlled by machines, Chicago by the
Democrats and the state government by the Republicans. Interestingly, Kaiser simulated
a redistricting of Illinois early in the reapportionment revolution that not only produced
districts that were much, much more compact than the present ones, but they were also
more compact than the actual Downstate districts in use at the time. (Kaiser 1966, 210).
122
The case of the 17th is an interesting one. The Republican candidate in 2006 was
a television news anchor from the Quad Cities, Andrea Zinga. Television news anchors
have a long history of being able to win local and district-level elections. Although she
lost to Democrat Phil Hare decisively (57-43%), Zinga surely would have had a better
chance of winning had the boundaries of the district more closely corresponded to
television market boundaries, as is often the case when districts are shaped using natural
boundaries or those resulting from natural human transportation and communication
patterns. As it was, the 17th contained only part of the Illinois side of the Quad Cities
television market and then swept southward along the Mississippi River to take in Quincy
and some territory stretching into Springfield and Decatur. Of course, ideals in
redistricting should not favor the election of a television news anchor or any other person,
but politicians of all stripes understand that campaign costs and efforts are reduced when
the district is compact and can be reached through a minimum of media buys (Campbell,
Alford and Henry 1984); in a district like the 17th, parties and candidates aiming to reach
the entire district can only make inefficient use of broadcast media and newspapers: They
only desire to reach a small percentage of viewers in each of many markets, because most
of the viewers in those markets live in other congressional districts. The issue of multiple
states in a media market is also a challenging one. Snyder and Str6mberg report,
"newspapers exhibit an in-state bias, covering representatives from their home state more
heavily than out-of-state representatives," even when their market area overlaps into an
adjoining state (Snyder and Strdmberg 2010, 11).
The 46th district of California presents another illustration of the role of
compactness in the conception of a district and the decisions that follow from this
123
conception. The 46th contains most of its population in some coastal suburbs in northern
Orange County. In an effort to create a safe Republican district as a part of the bipartisan
gerrymander of California after the 2000 census, the district also includes the Palos
Verdes Peninsula, at the southwest corner of Los Angeles County. In order to form a
E
Torrance
S
.
mC
desEE es
47
RaW
est-Aa
P
Es
Ver
Rea
1
W/
Source: State of California
contiguous district, the 46th has a spaghetti strip running along the heavily Democratic
city of Long Beach and the Los Angeles section known as San Pedro, also a heavily
Democratic center. (A small functional part of Long Beach near California State
University is included in the 46th.) The spaghetti strip includes some harbor islands,
which are heavily industrial. (In the earlier version of this district created after the 1980
census, the spaghetti strip through Long Beach and San Pedro was so narrow that it was
said to be contiguous only at low tide (Butler and Cain 1992, 61)). This district has no
center as that term is used by ordinary people. There is no
/
I'
X
way to drive continuously across the district. The Orange
County residents have little contact with the voters of the
124
Palos Verdes Peninsula, and vice versa. The ordinary driving distance between
Huntington Beach, in the Orange County portion of the district, where Congressman
Dana Rohrabacher maintains his district office, and Rancho Palos Verdes is 32 miles.
According to Google Maps, this is a 47 minute trip under favorable traffic conditions.
Nearly all of this trip takes place outside the borders of the district.
Source: Google MTaps
If the incumbent congressman and any challengers are strategic about what they
do, they spend most of their time in Orange County and hold a few events on the Palos
Verdes Peninsula. They have little or no incentive to do anything with the spaghetti strip
running across Long Beach and San Pedro. (The district also includes Santa Catalina
Island with a population of about 4000 and unpopulated San Clemente Island.)
In terms of media coverage, the distrct is at the center only for the residents of
the Orange County portion, and even so, the main media serving the district (the Orange
County Register and Los Angeles television stations) have many other districts to cover.
For the Palos Verdes portion of the district the main newspaper is the South Bay Daily
Breeze. However, most of the residents of the South Bay area live in the much more
compact 35th and 36th districts. So from the perspective of the Daily Breeze (and similar
media), the 35th and 36th districts are their main market and the 46th is a periphery. The
125
Long Beach Press-Telegram (and other media in the country's largest suburb) have little
incentive to do much with the 46th district. The vast majority of the population of Long
Beach is in the 37th district and few people to the east or west pay much attention to the
Long Beach paper. In a large media market, it is difficult for a new MC to get covered
by the press when they are so accustomed to quoting and interviewing the more senior
members of the local delegation. "It does not appear that newspapers are mainly
interested in providing necessary information to help voters," Snyder and Strdmberg
allege, noting that newspapers are more likely to run articles about retiring members than
those who are new and relatively unknown (Snyder and Str6mberg 2010, 11).
Voters in the spaghetti strip are at a tremendous disadvantage when it comes to
learning about their Congressman and their district. Voters in the Palos Verdes
Peninsula, essentially a panhandle of the district, also face a disadvantage. And even the
voters in the Orange County portion of the district are at a disadvantage relative to the
position they would be in if their district included only Orange County territory; by
sharing part of what should be an Orange County district with people in a non-compact
area, they are losing the focus of the Congressman, campaigns and media that would be
enhanced in a compact district. When Congressman Rohrabacher holds a town hall
meeting, it behooves him to concentrate on Orange County. Holding a town meeting in
downtown Long Beach, even if one could be held in the block or two along the ocean that
is in the spaghetti strip and part of the 46th, would likely be unsuccessful. The Long
Beach media would not do much to publicize it, and the small population of the spaghetti
strip would mean that there is no great pool of voters from which attendance at the
meeting can be drawn. Even holding such an event in the eastern part of Long Beach
126
where the district is more than two blocks wide would not do much to engage the citizens
of the spaghetti strip. The district is not much better descriptively than the
Massachusetts 4th; it can be weakly described as a district "on the
harbor," which might be true but doesn't convey the essence of the
represented territory as neatly as the appellation "Northeastern
Minnesota" does for that state's 8th congressional district.
In essence, the incumbent and contenders are forced to
consciously choose to spend time in a particular area of the district. In a district that is
both compact and small, this is not a choice that has to be made; merely being anywhere
in the district is proximate to all other points in the district. A congressman representing,
for instance, Milwaukee and a few suburbs need not consciously spend time in each
neighborhood of the district; it will suffice for most to know that the congressman is
present in the city. The same is not true for a member from a panhandled district like
California 46. In essence, the extreme deviation from a compactness standard forces the
politician to make a conscious choice to spend a certain amount of time in or effort on a
particular part of the district. This is similar to the choices that have always confronted
those who represent rural districts, who have to make constant decisions about how much
in resources to allocate to particular communities within their districts. In this sense,
then, redistricting with non-compactness complicates the lives of the politicians who
serve them (or run to) relative to the effort that would be necessary were the districts
more compact.
Obviously, adding a few more square miles of Orange County to
California 46 would eliminate the expense of resources involved in representational tasks
for covering the elongated and unwieldy district in its present form.
127
In a world heavily dependent on information, when district boundaries impose
restrictions on voter abilities to learn about politicians and on politicians' abilities to
communicate with voters, democracy suffers. Voters are hampered in learning about
politicians, and, it is assumed, politicians in learning about them. In a polity or media
construction that consists of two large portions from two districts and a small section
from a third district, the district contributing a small portion is apt to be marginalized. If
the polity in question is a school district, government teachers are apt to invite
representatives from the two main districts to speak at the school and participate in
forums. The school district is likely to keep in close contact with those representatives
and keep an eye on what legislative activities occupy them. The legislators from the third
district are apt to recoil in horror at their exclusion from the activities of this school
district, much as they strive to keep up with its activities and policies. These legislators
are apt to be caught off-guard while campaigning or doing constituent service in the
relatively small portion of their district that overlaps with this school territory.
Geography is working against them. People who make decisions and are aware of the
overlapping geography are likely to place a premium on inclusion of the politicians with
the large overlap and minimize those who have only a small overlap; they might even
assume they're bothering those legislators by attempting to make them aware of the
activities of their polity. Politicians, of course, blanch at the notion of being excluded
from anything.
The example of the school district overlapping with several legislative districts
resonates with real-world examples of media markets and congressional districts,
including those mentioned above. Many congressional districts overlap neatly with
128
media markets, and for politicians who serve these districts (or have ambitions to),
dealing with the media may be smooth sailing; they are apt to find that media in their
districts are eager to hear from them and even seek them out. For politicians who have to
deal with getting coverage in many media markets where the overlap with the territory is
very small, dealing with the media can be nothing but an exercise in frustration.
Politicians might tire of sending out press releases that go unreleased, and their press
secretaries no doubt tire of making cold calls to newsrooms where the feeling is that this
congressman has too few constituents in the newspaper's coverage area to warrant the
amount of coverage the congressman feels he deserves. Since news coverage is an
important aspect of voter information, these decisions that politicians and editors make
are critical in the process of voters obtaining information.
These media market-district overlaps are also crucial in congressional campaigns.
A campaign can be made much more expensive when a party or candidate has to buy
advertising time in a peripheral media market in order to reach a particular district. On
the state level, this has many ramifications across the U.S.
-
New Jersey, for example, is situated in two of the five largest
media markets, New York and Philadelphia. It is extremely
expensive and very inefficient for New Jersey parties and
-
candidates to pay to advertise on the radio and television
stations of these cities. It is inefficient because the New York
Source for TV market
maps: Dishuser.org from
U.S. Census Bureau base
maps
media market includes many more people in New York and
Connecticut and even a few in Pennsylvania that the advertiser
has to pay for when using the station (advertising rates take into
129
account the entire audience, not merely the ones the advertiser wants to reach); the
Philadelphia market is mostly in Pennsylvania and also includes viewers in Delaware and
Maryland.
By contrast, when advertising in the media markets in California, advertisers pay
for only viewers in California due to the remoteness of California from other states.
Campaigns seeking to reach New Hampshire - not only
a pivotal state in general elections but the location of the
first and most prominent presidential primary - must pay
for the expensive Boston media market. In his first bid
for the U.S. Senate from West Virginia in 1982, Jay
-Rockefeller
-
spent a considerable sum of money on the
costly Washington media market, which he needed to
reach only two small counties in the West Virginia eastern panhandle. Examples abound.
Chicago and Cincinnati are needed to reach all of Indiana; reaching sparsely populated
western Wisconsin requires paying for Minneapolis-St. Paul. Indeed, the only states that
are coverable from a single media market that includes no significant population in
adjoining states are Utah and Hawaii.
Stewart and Reynolds point out that
Paducah, KY Cape Girardeau, MO-'
Harrisburg Mt. Vernon, I
linois
the television news boss in Bangor, Maine
-
**"/
faces no decisions to make about which
senate race to cover (the market is entirely
within Maine), but the counterpart in
WO
Paducah, Kentucky has to consider multiple
130
senate races in the local coverage area, which includes parts of not only that state but also
Indiana, Illinois, Missouri, and Tennessee. (Stewart and Reynolds 1990, 500). As a
result, Maine voters hear only about their own state, whereas a Kentucky voter in the
Paducah market hears about several states and doesn't get as much about the home state
fed directly. (Ibid.) Stewart & Reynolds conclude that incumbents benefit from a
fragmentary structure of television markets, because challengers are more dependent on
television advertising and less likely to start with name recognition. Incumbents also get
more free coverage by television (i.e., news coverage) than challengers (Stewart and
Reynolds 1990, 512).
At the congressional district level, relatively few districts are contained within a
single media market or can be reached without incurring the expense of vast unrelated
consumers. Some overlap a string of media markets that mostly expand into adjoining
states. One object case is the 17th district of Illinois, stretching from the Davenport-Rock
Island-Moline media market (which includes a fair chunk of Iowa) on the north end,
through the Quincy-Hannibal media market (which is about half in Missouri), to the
periphery of the St. Louis media market (a major market, the overwhelming majority of
whose population is outside the district), then reaching through the Springfield area with
a narrow tentacle and ending in part of Decatur. Even with the relatively inexpensive
medium of radio, this district is a chore to cover. Not only are campaigns paying for
listeners they don't want to reach, those listeners are fatigued by hearing constant
advertising for campaigns they don't vote in. It complicates the matter of learning about
one's own congressman and district when one hears so much unrelated and confusing
input.
Data from the Cooperative Congressional Election Study (CCES) for 2008 can
illuminate the impact of compactness on voter knowledge. CCES (hereafter CCES 2008)
is a large-n Internet survey which includes close to 34,000 voters in its common content
(Ansolabehere 2009). The survey asks several general knowledge questions of voters
pertaining to the race, partisanship, and identity of their representatives. The theses
articulated here are (1) compactness will impact voter knowledge at the congressional
district level; (2) compactness will make little or no difference on voter knowledge at the
statewide level; (3) extremes in redistricting disproportionately affect Democrats. The
first two theses require little explanation. If the compactness thesis is correct, voters in
non-compact districts are disadvantaged in learning about their representatives and
districts more than those in compact districts, but the ability to learn about senators,
governors, and the composition of Congress as a whole and of state legislatures is largely
unaffected. The third thesis closely relates to the earlier chapters in this dissertation
which elaborated on extremes of redistricting primarily being used for purposes of racial
gerrymandering, which most heavily impact black voters, who are disproportionately a
Democratic electorate, and also disadvantage white Democratic voters in the states that
pack black Democratic votes; and that partisan gerrymandering, in the 21st century, has
been most acerbic in Texas, where the Republican state government has successfully
packed Democrats, employing in the process some of the least compact districts ever
devised. The bipartisan gerrymander in California in 2002, for the perverse reason that
because it created twice as many safe Democratic districts as safe Republican districts,
and used extremes in non-compactness to accomplish this end, did ipso facto affect
Democrats more than Republicans. The impact of these two gerrymanders, just by
132
themselves, is quite significant; California and Texas, the two largest states, have
between them almost 20 percent of all congressional districts, and more than 20 percent if
the states having only one or two districts are excluded. Thus, in the main, non-compact
districts affect Democrats more than Republicans. McDonald points out, "bias and
responsiveness in national congressional elections can be accomplished by controlling the
redistricting process in key states with large, heterogeneous populations" (McDonald
2006, 3-4).
CCES 2008 asked respondents to give the race of their House member (CCES
2008, question 319). Overall, 82.5 percent gave the correct answer. However, this was
not uniform across the country. Those who got the question correct were more likely to
live in a compact district. The average Hill ratio of the district of respondents who
answered correctly is 2.04; for those who answered incorrectly, it is 2.15. This difference
is statistically significant at the p<.0001 level. The 99.99 percent confidence level for
those answering correctly (n=26973) is 2.02 to 2.06 and for those answering incorrectly
(n=5758) is 2.12 to 2.19.
The survey also asked respondents for the party of their member (CCES 2008,
question 309d). Those who answered correctly have an average Hill ratio of 2.05
(n=22805; 99.9% CI 2.03-2.06) while those who answered incorrectly have an average
Hill ratio of 2.08 (n=9718; 99.9% CI 2.069-2.121). This is statistically significant at the
p<.001 level. In a bivariate regression operationalizing knowledge of member's party as
the dependent variable and compactness as the sole independent variable, the coefficient
is -0.768 (p=0.03265). Party identification is apt to have some role in this process.
133
Snyder and Str6mberg find that respondents whose party currently occupies their seat are
better at identifying the candidates (Snyder and Str6mberg 2010, 17).
A third useful item from CCES about the district does not pertain to an
information-related question directly, but tells about voters' use of that information.
Respondents were asked if they had contributed money for various electoral contests,
including the U.S. House race in their own districts (CCES 2008, question cc4l6a_4).
Among respondents who contributed money (at any level) in 2008, the average Hill ratio
of those who contributed to the contest in their own U.S. House district (N=1008; 99.9%
CI 1.919-2.04) is 1.979 and for those who didn't is 2.077 (n=8636; 99.9% CI 2.056-2.1).
This is statistically significant at the p<.001 level. There is no statistical significance to
the difference between means of the Hill ratios of those who did or did not contribute to
House candidates in other districts or those who gave to presidential candidates.
These results suggest that lack of compactness affects the information voters have
about their own districts. As discussed above, media messages are apt to be more
fragmented when voters are in a non-compact district, especially if the voters themselves
are situated in the part of the district responsible for giving the district its non-compact
shape (i.e., a tentacle or panhandle of the district). Voters are also less likely to be graced
by the presence of the congressman in a non-compact district. It is more difficult for
campaigns to recruit volunteers to deal with voters in tentacles and appendages of the
district, and as a result, there is apt to be less campaigning dealing specifically with the
congressional district contest in these areas of the district (contributing to an overall
effect for the district as a whole) than in compact districts. Advantages that Members
build up as individuals - which is key to the incumbency effect, according to Fiorina and
134
others (Fiorina 1977, Ansolabehere, Brady and Fiorina 1992) - then become diluted in a
non-compact district when many members of the polity are unable to ascribe particular
things that incumbents do or particular benefits that incumbents have delivered to the
district to that particular officeholder.
This gels nicely with the finding of Snyder and Str6mberg: Voters whose districts
overlap closely with media markets (their term is "congruence") have the opportunity to
read many more articles about their member than other voters (Snyder and Str6mberg
2010, 11). They argue that this affects information that voters have. The difference
between being low on congruence and high on congruence is quite large in terms of voter
information, "about as large as the effect of changing a respondent's education from
grade school to some college" (Snyder and Str6mberg 2010, 16). Using ANES data, they
find that voters are more likely to know a candidate's name the more closely their district
overlaps with a newspaper market area (Snyder and Str6mberg 2010, 17).
The second hypothesis is that compactness will make little or no difference in
voter information for offices other than the U.S. House district. Respondents were asked
which party controls the U.S. House of Representatives (CCES 2008, question cc308a).
The correct answer was the Democratic Party. Overall, 22,212 respondents (68.1
percent) gave the correct answer. Only 9 percent of respondents said Republicans, but
405 said neither party controlled the House, and 21.6 percent weren't sure. All of these
can be construed as the wrong answer and are operationalized that way for purposes of
this analysis. The average Hill ratio of those who gave the correct answer is 2.075 (95%
CI 2.04-2.1). The average Hill ratio of those who have one of the various wrong answers
is 2.058 (95% CI 2.04-2.07). This difference lacks significance at the p<.05 level. A
135
parallel question asked which party controls the U.S. Senate (CCES 2008, question
cc308b). In the total survey, 65.9 percent gave the correct answer and said the
Democratic Party controls the Senate; 8.8 percent said the Republicans do, 5.1 percent
said neither, and 20.3 percent weren't sure. The average Hill ratio of those with the
correct answer is 2.063 (95% CI 2.05-2.07). The average Hill ratio of those with a wrong
answer is 2.058 (95% CI 2.04-2.07). This difference lacks significance at the p<.05 level.
Respondents were also asked questions about their knowledge of state level
politics. Survey respondents were most familiar with the partisanship of their state
governor, on average, of all political offices referenced in the survey (CCES 2008,
question cc309a). In total, 81.8 percent got this question correct. The average Hill ratio
of those with the correct answer is 2.044 (95% CI 2.06-2.07). The average Hill ratio of
those who answered incorrectly is 2.065 (95% CI 2.03-2.06). This difference lacks
significance at the p<.05 level. Respondents were much more challenged when it came
to the political control of their state legislative chambers. Only 48 percent knew which
party controlled their state's upper chamber (CCES 2008, question cc308c) and only 46.5
percent knew which party controlled their state's lower chamber (CCES 2008, question
cc308d). (CCES operationalizes Nebraska's unicameral senate as a lower chamber.)
Relatively few of these gave an actual wrong answer; the bulk of respondents who didn't
know the correct answer admitted that they didn't know. The average Hill ratio of those
who gave a correct answer for control of their state's upper chambers is 2.082 (95% CI
2.068-2.095). For those who answered that the wrong party controls the upper chamber,
it is 2.053 (95% CI 2.033-2.072).
This difference lacks statistical significance at the
p<.05 level. The average Hill ratio of those who gave a correct answer for party control
136
of their state's lower house is 2.086 (99.9% CI 2.063-2.11). For those who gave an
incorrect answer, it is 2.04 (99.9% CI 2.022-2.058).
This difference is statistically
significant at the p<.001 level. This result is confounding because not only does it
represent a statistically significant difference but also because the result suggests that
voters in non-compact districts are more likely to know the correct answer. Partisanship
is part of the answer. In states where the Republican Party controls the lower chamber,
the difference is statistically significant at the p<.05 level, but the difference is highly
significant in states where the Democratic Party controls the lower chamber. In terms of
the upper chamber, it is nearly the opposite. In states where the Democratic Party
controls the upper chamber, the difference is statistically significant at the p<.05 level, if
barely. In states where the Republican Party controls the upper chamber, the difference is
statistically significant at the p<.O1 level. In general, however, knowledge of which party
controls the upper and lower house of the legislature is low, and there is wild fluctuation
from state to state. This is therefore not a very good measure of voter knowledge at all,
let alone a good one to use to evaluate the impact of congressional redistricting.
In the case of U.S. Senators, the results are mixed. For the senior senator (CCES
2009, question cc309b), the average Hill ratio of those getting the correct answer is 2.064
(95% CI 2.054-2.074) while for those who gave the wrong answer, it is 2.052 (95% CI
2.036-2.069). This is not statistically significant at the p<.05 level. For the junior
senator (CCES 2009, question cc309c), the average Hill ratio of those with the correct
answer is 2.049 (99.9% CI 2.036-2.0649) and for those who gave the wrong answer, it is
2.095 (99.9% CI 2.065-2.125). This is statistically significant at the p<.001 level. This is
another result that confounds the theory.
137
However, when analyzed on a state by state basis, knowledge of the senior
senator relative to the junior senator is greatest in these states: North Carolina, Delaware,
Kentucky, Arizona, Georgia, Louisiana, Utah, South Carolina. Some of these states are
notable for their excesses in non-compactness. Many of them had new senators going
into the 2008 election who were little-known in the state. At the other end, knowledge of
the junior senator relative to the senior senator is greatest in these states: New Hampshire,
Illinois, New York, Colorado, Mississippi. Two of those five had presidential candidates
running in the 2008 election cycle who were obviously better known than the senior
senator. Indeed, the top three senators in terms of home-state voter knowledge of
partisanship were the junior senators from Illinois and New York and the senior senator
from Arizona. Of course, three of those top five states are known for their excesses in
non-compactness and their aggressive use of majority-minority districting. These factors
contributed to distorting the significance of knowledge of the junior senator vis-i-vis
compactness. Overall, the senior senator is only slightly better known than the junior
senator, with a 3.3 percentage point difference. This suggests that the normal state of
affairs is for knowledge of the senior senator to be slightly higher than the junior senator.
By excluding Illinois, New York, Arizona, New Hampshire, and North Carolina
from the dataset, the difference between the Hill ratios for those who know and don't
know can be brought down below the level of statistical significance at the p<.05 level.
For those who answered correctly, the Hill ratio is 2.042 (95% CI 2.03-2.0524) and for
those who answered incorrectly, it is 2.07 (95% CI 2.0521-2.089).
Of course, there is
some ad hoc calculation involved in this, but it illustrates the point that overall,
significance is low when comparing knowledge of senatorial party relative to
138
compactness and that an unusual situation occurred in 2008 that produced two
presidential candidates who were both junior senators, one of whom was already one of
the best known figures in the country and the other of whom became one of the best
known figures in the country as a result of the presidential campaign.
The third hypothesis is that Democrats are more greatly affected by
discompactness than Republicans. Using 3-point party identification (CCES 2008,
question cc307), the average Hill ratio of Democrats is 2.09 (99.9% CI 2.0626-2.115), of
Republicans is 2.03(99.9% CI 2-2.0622), and of independents is 2.04 (95% CI 2.02-2.06).
This is statistically significant at the p<.001 level. This means that Democrats tend to
systematically live in less compact districts than do Republicans or independents. Thus,
if the effects argued in this chapter are real, then they would tend to have greater impact
on Democrats than either Republicans or independents. Mechanistically, this would also
extend to the parties and campaigns themselves, making campaigning more difficult for
Democrats than for Republicans. Using 7-point party identification (CCES 2008,
question cc307a), we find a significant difference between strong Democrats and strong
Republicans at the p<.O1 level and between not very strong Democrats and strong
Republicans at the p<.05 level. The average Hill ratio of strong Democrats is 2.085
(99.9% CI 2.062-2.11) and of strong Republicans is 2.025 (99.9% CI 2-2.052). The
average Hill ratio of not very strong Democrats is 2.09 (95% CI 2.06-2.13). This shows
that the most staunch Democrats are apt to live in discompact districts while the most
staunch Republicans are more likely to live in compact districts. Indeed, the next most
dedicated group of Democrats live in even less compact districts than the strong
139
Democrats. Meanwhile, the independents and the weak partisans are much more similar
in their compactness averages to Republicans than they are to Democrats.
In order to get around the confounding nature of the results for some offices
trending differently from what the theory predicts, the author developed a different
method for evaluating the impact of compactness on voter knowledge. The zip code of
each respondent was among the pieces of information collected in CCES 2008. Of the
32,800 respondents, 361 had a faulty zip code that did not match the state the respondent
claimed. These were excluded from this analysis. No effort was made to determine if the
state-matching zip codes were correct for the congressional district the respondent
claimed to live in. From these zip codes, latitude and longitude were coded from thr
Boutell zip code database representing a centroid point in the zip code. An additional
211 zip codes were not codable, mostly because the zip code was newly created. This
left 32,228 respondents for this analysis, although the 67 coded responses from the
District of Columbia were not used in the analyses that follow. The average latitude and
longitude of all respondents in each congressional district was computed to determine a
centroid point for each congressional district. (The term "centroid point" is used in this
analysis to indicate that the point created herein does not necessarily correspond to the
geographic center of the district or to the population center created using a different
method.) Using an algorithm, the distance between each respondent's zip code and the
centroid point for each congressional district in the same state was computed.
Of course, some respondents are closer to the centroid point of congressional
districts in other states than to other points in their own state and even their own district.
However, there was no point in computing these distances because proximity to
140
congressional districts in other states is not information used by redistricters in creating
congressional districts.
The minimum distance to a congressional district centroid point within the state
was computed, and respondents were coded whether this corresponded to their own
district or not. Overall, 22,839, or 70.9 percent, of the respondents' zip code centroid
points are closer to their own congressional district centroid points than to others in the
same state.
The significance of this measure is that the analysis of compactness ratios and
voter knowledge above does not take into consideration that individual respondents may
not reside in the panhandled parts of the districts; thus, from their standpoint, the district
is seemingly a compact unit even though others at the periphery of the district may see
things differently. If all districts were square, everyone would be closer to the center of
their own district than to the center of any other district. Thus, the fact that individual
respondents are closer to the centers of other districts than to the center of their own
district means that they live in a panhandle, projection, extrusion, or some other kind of
sinuosity, rather than in the center of the district. Computing the actual population
centers of each of the 435 congressional districts is beyond the scope of this analysis (and
not particularly useful since the finest level geography available in the CCES dataset is
the zip code anyway), and using the zip code geography to generate a centroid point will
suffice for this analysis.
A chi-square analysis yielding odds ratios was computed for each of the voter
information measures of CCES 2008. In each instance, the two groups of the chi-square
are own-district closest (OC) and own-district not closest (NOC). The condition tested
141
for is knowledge of the voter information item (knows information = condition present;
does not know information = condition absent).
The key voter information items hypothesized to be linked to compactness are
found to be significant in this analysis. First, knowledge of one's own member was had
by 16,247 persons whose own district centroid point is closest to them and lacking in
6,328, or 71.97 percent who knew this. Among those whose own district centroid point
is not the closest to them, 6,141 knew the member and 3,189 did not, meaning 65.82
percent did. By a factor of 1.0934, those in central parts of districts (as defined by the
centroid of one's own district being closer than any other in-state district) have greater
awareness of their own member's partisanship. This corresponds to an odds ratio of
1.333, with a Pearson chi-square of 119.25. This is significant at p<.0001.
Second, the
race of one's own member was given correctly by 19,186 in the OC group as opposed to
3,535 who gave an incorrect answer (84.44 percent correct). This compares to only
77.93 percent in the NOC group who got the answer right (7,317 correct; 2,072
incorrect). That is a factor of 1.0835 for those in the central part of the district, a rate
very similar to the rate by which the knowledge of the OC group exceeded the NOC
group on the partisanship of the member. This generates an odds ratio of 1.5369 and a
chi-square of 195.36, also significant at p<.0001.
Also corresponding to the hypothesis, voter information about other offices senior senator, junior senator, control of the U.S. Senate, control of the U.S. House,
governor - is not related to whether one is closer to the center of one's own district than
to the center of other districts in the same state. The relative factors (also known as risk
ratios) and odds ratios are very close to one for each of the other offices, and the Pearson
142
chi-square coefficients are all less than one, with the p values all being too high to
achieve significance. A complete table of these chi-square and odds ratio statistics
comprise Table 2 at the end of this chapter.
The only part of this data analysis that continues to confound is knowledge of
party control of the lower house of the state legislature. While the pattern of by OC or
NOC of knowledge of party control of the upper house of the state legislature is similar to
the other statewide level knowledge items above, the chi-square coefficient pertaining to
the lower house is 12.98, with a p value of a highly significant 0.000315. No explanation
for these deviating results is obvious. While it is tempting to consider that perhaps states
that gerrymander their congressional districts also gerrymander their state legislative
districts, this does not explain why the result is significant for the lower house and not the
upper house. Nor is it a satisfactory hypothesis for the information gap; the thesis for this
chapter contends that non-compactness impacts the knowledge of voters of their own
representatives, but it would not prevent them from knowing the partisanship of the
leadership of their state's legislative chambers. A more likely explanation is that since
overall, knowledge of control of state legislative chambers is low (when "don't know"
responses are considered, less than half of CCES respondents are able to correctly
identify the party that controls either house of their state legislature; this is considerably
less than for the houses of Congress), eccentricities in the data and the patterns of
information in particular states are apt to have a greater impact on this information than
on other information which is more widely held.
The finding that information about larger political offices tend not to be affected
by compactness aligns well with research by Snyder and Str6mberg. They posit that their
143
key variable, congruence, while highly influencing the amount of coverage a newspaper
gives a member should not influence the amount of coverage the newspaper gives to
other topics, such as the majority party or U.S. senators. (This is not entirely correct,
since some media markets overlap state lines, and there are apt to be state effects on
coverage of senators much as the paper shows there are for house members. The
newspapers and other media in the Boston media market assuredly give more coverage to
Massachusetts senators than they do to those from New Hampshire.) They find that
congruence is not correlated with political knowledge that is not tied to specific
congressional districts (Snyder and Str6mberg 2010, 20).
This ties in with the larger theories about redistricting and competitiveness
posited by Campbell and others. One scholar noted, "Compared to gerrymandered
districts, compact districts more effectively preserve political homogeneity, because they
frustrate attempts to construct safe districts for incumbents or isolate opposition votes in
certain districts." (Stem 1974, 414-415).
The circulation areas of newspapers (which reflect age-old communication and
transportation patterns) are key to how voters and MCs perceive each other. Snyder and
StrSmberg characterize newspapers as "key provider[s] of information about
congressmen." (Snyder and Str6mberg 2010, 23). Snyder and Str6mberg find overall that
congruence affects voter information and even participation in many ways. Furthermore,
they argue that this influences congressional voting behavior as congruence gives them
greater incentive to align themselves with the interests of their district. (Snyder and
Str6mberg 2010, 44). The impact of these findings on the compactness and voter
knowledge thesis is that as compactness decreases, congruence is apt to decrease too and
this therefore constitutes additional evidence that non-compact districting has a negative
impact on voter information.
Some scholars have noted the increasing tendency of voters to assort themselves
geographically in particular ways and allege that compactness exacerbates such a sorting
process. Erikson calls these "accidental gerrymanders" (Erikson 1972, 1237). Chen and
Rodden put it this way: "Since the realignment of the party system, Democrats have
tended to live in dense, homogenous neighborhoods that aggregate into landslide
Democratic districts, while Republicans live in more sparsely populated neighborhoods
that aggregate into geographically larger and more politically heterogeneous districts."
(Chen and Rodden 2009, 27). However, they fail to find that the partisan bias they
observe in recent elections in Florida are the result of intentional partisan
gerrymandering. (Chen and Rodden 2009, 28) Instead, they caution against the
assumption that electoral bias results from partisan gerrymandering. In this instance, they
demonstrate that electoral bias results merely from the ordinary mobility of the electorate.
Furthermore, they concede, "the best hope for Democrats to obtain a seat share that
approximates their vote share in Florida would be to strategically draw long, narrow
districts shaped like pie slices emanating from downtown Miami and Tampa into the
suburban and rural periphery." This is contrary to the evaluation of Campbell, who
contends, "So long as district boundaries are drawn with respect for established
communities, with concern for geographic compactness and contiguity, and without
regard to normal turnout levels in the proposed districts, the single-member-district
electoral system should favor the party with low-turnout characteristics that tend to be
also concentrated geographically." (Campbell 1996, 42). But Campbell argues that if,
145
inter alia, "districts were not drawn in such a way that there were no aggregate
socioeconomic differences among districts, then the single-member-district system would
not generate electoral bias based on turnout differences." (Campbell 1996, 42-43). Of
course, the solution Chen & Rodden propose is an absurd one and one that can't be done
in concert with establishing majority-minority districts under the aegis of the Voting
Rights Act. It also can't be done in the context of most states having statutory or
constitutional provisions mandating compact, contiguous districts.
Furthermore, this idea contradicts the trend of courts considering districting plans
under "communities of interest" theory dating back to the 1980s and probably earlier.
The "pie-shaped" congressional districts of Minnesota in use in the 1970s (all but one
outstate district reached into the Minneapolis-St. Paul area) were thrown out by the court
in favor of a "four-four"
plan that gave four
cCity of fackson
o
Hinds County
...
districts to the
miles
I%
Jackson
/
15%
2ek*
10 Miles
Minneapolis-St. Paul
area and four to outstate
Concentrations of
black population
1%Percent of citys b,
Minnesota. (Schwab
population. by district
1988, 46). Chen and
Source: Monmonier 2001, 29, compiled from Parker 1990, 155-156.
Rodden advocate an even
more egregious plan that would extend districts like this not only into the metropolitan
areas but right into the inner city. Doing so would probably not only reduce minority
representation, it might even result in no representatives coming from the inner city at all.
Consider, for example, the textbook case of racial gerrymandering to exclude minorities,
146
the redistricting of Hinds County, Mississippi in 1973 that divided Jackson's population,
including its sizeable black population, among all five districts. (Monmonier 2001, 29).
Similarly, Archer and Shelley explore redistricting scenarios using pie-shaped districts
and conclude that inner-city issues would suffer under the implementation of such a
scheme (Archer and Shelley 1986, 80). The solution, then, if there is one in the context
of an SMP electoral system, is for Democrats not to segregate themselves in the core of
urban areas. Thus, Chen and Rodden are implicitly calling for the SMP electoral system
to be replaced by a system of proportional representation.
Rather, the thesis of Chen and Rodden is harmonious with this dissertation and
with other work cited herein; many scholars have noted the tendency for winners to take
a larger share of seats than their vote share would seem to call for. Chen and Rodden
over-demonize compactness as an ideal rather than focus on the advantages compactness
brings to representation. Instead, they miss the larger picture (argued particularly in the
compactness chapter) that Democrats accrued larger seat counts than presidential voting
suggested they ought to have been getting in the 1966 to 1992 period. If the 2000
election showed that Republicans had an unfair bias in Florida, as Chen and Rodden
allege, then that must mean Democrats had an unfair bias for a much longer period of
time in many other places and in the U.S. as a whole.
Chen and Rodden are not the first to allege that compactness can work against
Democrats. In a highly polemical article, two legal advocates for Democrats in
redistricting cases declare that compactness is not a neutral criterion for districting but
rather one that systematically advances Republican interests due to the tendency of
Democrats to cluster in urban areas (Lowenstein and Steinberg 1985, 23-24). Polsby and
147
Popper respond that the authors don't cite evidence that shows the ratio of Democrats to
Republicans in Democratic districts is higher than the ratio of Republicans to Democrats
in Republican districts (Polsby and Popper 1991, 334-335); Grofman characterizes their
data as "sketchy to the point of nonexistence." (Grofinan 1985, 92). Tongue firmly in
cheek, Shapiro points out, "proving that a government policy favors Republicans is no
longer the knock-out blow it once was" - and with fairly good prescience, that a Supreme
Court that attacks gerrymandering will strike down both Democratic and Republican
gerrymanders and the public will not have much idea whose side was favored in the total
score (Shapiro 1985, 237-238).
Ultimately, what keeps gerrymandering from getting out of control is the
tendency of the system to punish greed. Gerrymandering is a zero-sum game. The only
way partisans can make their opponent's districts weaker is by making their own districts
weaker. Legislators simply don't want to cede their "surplus votes" to make an adjacent
district easier for the party to win. Polsby & Popper add that the most intricate
gerrymanders are based on assumptions as to the partisanship of certain marginal
districts, for which miscalculation will cause the gerrymander to backfire on the
gerrymanderer. "A few unintentional marginal losses can eviscerate a gerrymander,
because to be effective, a gerrymander must produce wins, not just in a majority of
marginal districts, but in a supermajority of those districts." (Polsby and Popper 1991,
335).
They further argue that in the context of a hypothesized gerrymander fomented by
the process of compactness that Lowenstein & Steinberg deride, there is no reason to
believe Republicans could carry a supermajority of the marginal districts and that such a
mapping "may even redound to the Democratic Party's advantage where their core
148
districts were more irrefragably partisan than were the strongholds of the opposition."
(Ibid.) No one denies that Democratic core districts are now more partisan than they
were when that article was written.
Moreover, even if those who question the neutrality of compactness and naturalboundary districts were correct, it would not be a legitimate reason to abandon the
compactness ideal. If a party can only achieve a majority through gerrymandering, then
perhaps that party ought not exist. Polsby & Popper point out that no such partisan
argument saved the malapportioned districts of the pre-Baker era from the equipopulation
standard. (Polsby and Popper 1991, 335-336). Shapiro adds that the problem of
compactness favoring Republicans, if any, is not a function of districting but the
distribution of Democratic and Republican voters. He contends that the only thing that
can prevent such a scenario is "an affirmative gerrymander in favor of the Democratic
Party." (Shapiro 1985, 238). He cites work by Grofman (Grofrnan 1985) that even after
courts determine that gerrymandering is justiciable, the natural distribution of voters
through residential clustering will continue to exist (Shapiro 1985, 240). He then argues
himself,
Neither party chose to represent whom they did because of
their geographic stacking or dispersion or with an eye to
how their choice would affect their electoral fortunes if the
world were suddenly to come ungerrymandered. If
geography favors the Republicans in an ungerrymandered
world, that is a purely fortuitous result, unforeseeable by
either party when it chose its ideologies and clienteles.
Such stacking ought to be treated as extraneous to the goal
of constraining the self-serving actions of legislatures.
(Shapiro 1985, 240).
149
Compact districts are more functional than merely being "the 4th district of
Massachusetts"; a compact district is part of and integrated with a community, a polity, a
media market. In a functional district, people live and work there, attend school there,
raise their children there, interact with others and share opinions and information there.
They do not only vote there. Justice Powell wrote in Karcher, "A legislator cannot
represent his constituents properly - nor can voters from a fragmented district exercise
the ballot intelligently - when a voting district is nothing more than an artificial unit
divorced from, and indeed often in conflict with, the various communities established in
the State." (Powell, J., dissenting, Karcher v. Daggett, 1983, 462 US 725 at 787).
A question to be answered in the future is whether the public at large cares about
compactness in redistricting. Cain, among others, has argued that public knowledge of
redistricting is so low that there is unlikely to be any wide knowledge of compactness
effects. (Pildes and Niemi 1994, 538). (However, Pildes & Niemi themselves note that
post-Shaw, courts are apt to reject Cain's argument. (Ibid.))
In the end, compactness produces not only procedural fairness but substantive
fairness as well, when the end result is to produce districts that are more fair and
functional for a greater number of voters.
Schattschneider wrote nearly seven decades
ago, "These are the forces engaged in a war of extermination for supremacy in Congress,
the no man's land of American politics. It is a war of extermination because it is real, no
mere game played for exercise." (Schattschneider 1942, 207).
150
Table 1. Compactness ratios (Hill) of voter congressional districts by voter knowledge,
PID, and campaign contributions
Knowledge item
Party of House member
Race of House member
Party of senior senator
Party of junior senator
Party of governor
Party of U.S. Senate
Party of U.S. House
Party of upper chamber
Party of lower chamber
No
N (Yes)
N (No)
2.095
2.154
22805
2.039
26973
24794
24083
26673
21405
22166
15445
15068
9718
5758
7789
8498
5948
11326
10565
16744
17324
1.979
2.077
1008
8636
**
2.135
2.071
2.062
2.041
602
9402
NS
8308
1336
NS
Yes
2.048
2.041
2.064
2.049
2.065
2.063
2.063
2.082
2.086
2.052
2.095
2.044
2.058
2.058
2.049
NS
NS
NS
NS
**
Participation item
Gave to House candidate
in own district
Gave to House candidate
in other district
Gave to presidential cand
3 point PID
Dem
Rep
Ind
Other
Not Sure
7 point PID
Strong Dem
Not very strong Dem
Lean Dem
Independent
Lean Rep
Not very strong Rep
Strong Rep
not sure
*p<.05 **p<.01 ***p<.001
N
Diff
Dem
**
2.089
2.033
9864
7752
**
2.042
6786
*
2.047
904
2.018
971
2.085
2.093
2.046
2.043
2.031
2.049
2.025
2.042
NS
NS
N
7174
Diff SD
2660
NS
NS
NS
2857
2600
2625
2333
Diff
Rep
NS
NS
NS
Diff
NVSD
NS
NS
NS
NS
NS
NS
NS
5399
546
****p<.0001
Source: Cooperative Congressional Election Study, 2008
151
Diff
LID
NS
NS
Diff
Ind
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Dif LR
Diff
NVSR
Diff
SR
*
NS
**
*
NS
NS
NS
NS
NS
--
NS
NS
--
NS
NS
NS
NS
NS
NS
NS
NS
NS
*
TABLE 2. Chi-square analysis of proximity of repondents to the centroid of their own district
Knows information__________
Party of own member
Own dist closer (OC)
Own dist not closer (NOC)
Race of own member
Own dist closer (OC)
Own dist not closer (NOC)
Yes
No
Rate
Factor
Odds
OR
xA2
16247
6328
72.0%
1.093
2.568
1.833
119.25
<.0001
6141
3189
65.8%
19186
3535
84.4%
1.537
195.36
<.0001
7317
2072
77.9%
0.972
0.81
0.3681
Own dist not closer (NOC)
0.990
0.12
0.7290
0.973
0.93
0.3349
0.984
0.38
0.5376
1.008
0.09
0.7642
1.926
___
Party of governor____
Own dist closer (OC)
p
__
____
_
_
3.531
4147
81.7%
7682
1673
82.1%
17213
5412
76.1%
7124
2218
76.3%
____
_____
_____
18502
5.427
1.084
4.462
0.995
4.592
Party of senior U.S. senator
Own dist closer (OC)
Own dist not closer (NOC)
3.212
Party of junior U.S. senator
Own dist closer (OC)
Own dist not closer (NOC)
3.181
0.998
________
____
16703
73.8%
5917
0.993
2.823
2.901
6947
2395
74.4%
14819
7724
65.7%
6160
3160
66.1%
15387
7216
68.1%
6338
2996
67.9%
Party control of state upper house
10626
Own dist closer (OC)
4524
Own dist not closer (NOC)
3261
1316
76.5%
77.5%
0.988
3.258
3.438
0.948
2.07
0.1502
Party control of state lower house
10340
Own dist closer (OC)
2829
78.5%
0.971
3.655
0.865
12.98
0.0003
Party control of U.S. Senate
Own dist closer (OC)
Own dist not closer (NOC)
Party control of U.S. House
Own dist closer (OC)
Own dist not closer (NOC)
Own dist not closer (NOC)
0.995
1.919
1.949
____
4450
1053
80.9%
1.003
2.132
2.116
4.226
BIBLIOGRAPHY
Abramowitz, Alan I., Brad Alexander, and Matthew Gunning. "Don't Blame Redistricting for
Uncompetitive Elections." PS: PoliticalScience and Politics 39 (2006): 87-90.
Abramowitz, Alan I., Brad Alexander, and Matthew Gunning. "Incumbency, Redistricting, and the Decline
of Competition in U.S. House Elections." JournalofPolitics68 (2006): 75-88.
Altman, Micah. "A Bayesian approach to detecting electoral manipulation." PoliticalGeography 21
(2002): 39-48.
-.
"Districting Principles and Democratic Representation." Pasadena, Calif.: Ph.D. thesis, California
Institute of Technology, 1998.
Ansolabehere, Stephen D. Cooperative CongressionalElection Study, 2008. Release 1. Cambridge, Mass.:
Massachusetts Institute of Technology, February 2, 2009.
Ansolabehere, Stephen D., and Alan Gerber. "Incumbency Advantage and the Persistence of Legislative
Majorities." Legislative Studies Quarterly22 (1997): 161-178.
Ansolabehere, Stephen D., and James M. Snyder. The End ofInequality: One Person,One Vote and the
Transformation ofAmerican Politics.New York: Norton, 2008.
Ansolabehere, Stephen D., David Brady, and Morris Fiorina. "The Vanishing Marginals and Electoral
Responsiveness." British JournalofPoliticalScience 22 (1992): 21-38.
Ansolabehere, Stephen D., James M. Snyder, and Charles H., III Stewart. "Old Voters, New Voters and the
Personal Vote: Using Redistricting to Measure the Incumbency Advantage." American Journal of
PoliticalScience 44 (2000): 17-34.
Ansolabehere, Stephen D., John Mark Hansen, Shigeo Hirano, and James M. Snyder Jr. "The Decline of
Competition in U.S. Primary Election, 1908-2004." In The Marketplace ofDemocracy: Electoral
Competition and American Politics,edited by Michael P. McDonald and John Samples, 74-101.
Washington: The Brookings Institution, 2006.
Archer, J. Clark, and Fred M. Shelley. American ElectoralMosaics. Washington: Association of American
Geographers, 1986.
153
Attneave, Fred, and Malcolm D. Arnoult. "The Quantitative Study of Shape and Pattern Perception."
PsychologicalBulletin 53 (1956): 452-47 1.
Babcock, Tuckerman. "Predictably Unpredictible: The Alaskan Supreme Court and Reapportionment." In
Race and Redistrictingin the 1990s, edited by Bernard Grofman, 121-133. New York: Agathon
Press, 1998.
Backstrom, Charles H. "Problems of Implementing Redistricting." In RepresentationandRedistricting
Issues, edited by Bernard Grofman, Arend Lijphart, Robert B. McKay and Howard A. Scarrow,
43-53. Lexington, Mass.: D.C. Heath, 1982.
Backstrom, Charles H. "The Practice and Effect of Redistricting." PoliticalGeography Quarterly 1 (1982):
351-359.
Backstrom, Charles H., Leonard Robins, and Scott Eller. "Establishing a Statewide Electoral Effects
Baseline." In PoliticalGerrymanderingand the Courts, edited by Bernard Grofman, 145-170.
New York: Agathon, 1990.
Backstrom, Charles H., Leonard Robins, and Scott Eller. "Issues in Gerrymandering: An Explanatory
Measure of Partisan Gerrymandering Applied to Minnesota." Minnesota Law Review 62 (1978):
1121-1159.
Baker, Gordon E. "Gerrymandering: Privileged Sanctuary or Next Judicial Target?" In Reapportionment in
the 1970s, edited by Nelson Polsby, 121-142. Berkeley, Calif.: University of California Press,
1971.
Baker, Gordon E. "Lessons from the 1973 California Masters' Plan." In PoliticalGerrymanderingand the
Courts, edited by Bernard Grofman, 296-317. New York: Agathon Press, 1990.
Baker, Gordon E. "The 'Totality of Circumstances' Approach." In PoliticalGerrymanderingand the
Courts, edited by Bernard Grofman, 203-211. New York: Agathon Press, 1990.
Baker, Gordon E. "The Unfinished Reapportionment Revolution." In PoliticalGerrymanderingand the
Courts, edited by Bernard Grofman, 11-25. New York: Agathon Press, 1990.
Balinski, M.L., and H. Peyton Young. "A New Method for Congressional Apportionment." Proceedings of
the NationalAcademy of Science 71 (1974): 4602-4606.
154
Barabas, Jason, and Jennifer Jerit. "Redistricting Principles and Racial Representation." Politics and Policy
Quarterly4 (2004): 415-435.
Bicker, W. "The effects of malapportionment in the states: A mistrial." In Reapportionmentin the
Seventies, edited by Nelson Polsby. Berkeley, Calif.: University of California Press, 1971.
Bickerstaff, Steve. Lines in the Sand: CongressionalRedistrictingin Texas andthe Fallof Tom DeLay.
Austin, Tex.: University of Texas Press, 2007.
Blair, T.J., and T.H. Biss. "The Measurement of Shape in Geography: An Appraisal of Methods and
Techniques." Bulletin of QuantitativeDatafor Geographers11 (1967).
Born, Richard. "Generational Replacement and the Growth of Incumbent Reelection Marginals in the U.S.
House." American PoliticalScience Review 73 (1979): 811-817.
Bositis, David A. "Political Parties, Redistricting, and Minority Representation: The Southern States, 19922002." In Voting Rights and Minority Representation:Redistricting, 1992-2002, edited by David
A. Bositis, 33-50. Lanham, Md.: University Press of America, 2006.
Boyce, Ronald R., and W.A.V. Clark. "The Concept of Shape in Geography." GeographicalReview 54
(1964): 561-572.
Boyd, William J.D. "Reapportionment: Problems, Prospects, Probabilities." In Representation and
Misrepresentation:Legislative Reapportionment in Theory and Practice,edited by Robert A.
Goldwin, 115-140. Chicago: Rand McNally, 1968.
Brady, David W., and Bernard Grofman. "Modelling the Determinants of Swing Ratio and Bias in US
House Elections 1850-1980." PoliticalGeographyQuarterly 10 (1991): 254-262.
Brady, David W., and Bernard Grofman. "Sectional Differences in Partisan Bias and Electoral
Responsiveness in U.S. House Elections, 1850-1980." British JournalofPoliticalScience 21
(1991): 247-256.
Brookes, R.H. "Electoral Distortion in New Zealand." AustralianJournalofPolitics andHistory 5 (1959):
218-223.
Brookes, R.H. "The Analysis of Distorted Representation in Two Party, Single Member Elections."
PoliticalScience 12 (1960): 158-167.
155
Browning, Robert X., and Gary King. "Seats, Votes, and Gerrymandering: Estimating Representation and
Bias in State Legislative Redistricting." Law & Policy 9 (1987): 305-322.
Brunell, Thomas L. Redistrictingand Representation: Why Competitive Elections areBadfor America.
New York: Routledge, 2008.
Bullock, Charles S. "Affirmative Action Districts: In Whose Faces Will They Blow Up?" Campaigns &
Elections, April 1995.
Bullock, Charles S. "Redistricting and Congressional Stability." JournalofPolitics 37 (1975): 569-575.
Burton, Orville Burton. "Legislative and Congressional Districting in South Carolina." In Race and
Redistricting in the 1990s, edited by Bernard Grofian, 290-314. New York: Agathon Press, 1998.
Butler, David, and Bruce E. Cain. CongressionalRedistricting: Comparativeand TheoreticalPerspectives.
New York: Macmillan, 1992.
Cain, Bruce E. "Assessing the Partisan Effects of Redistricting." American PoliticalScience Review 79
(1985): 320-333.
-.
The ReapportionmentPuzzle. Berkeley, Calif.: University of California Press, 1983.
Cain, Bruce B., John A. Ferejohn, and Morris P. Fiorina. The Personal Vote: Constituency Service and
ElectoralIndependence. Cambridge, Mass.: Harvard University Press, 1987.
Cain, Bruce E., Karin MacDonald, and Iris Hui. Competition and Redistrictingin California:Lessons for
Reform. Berkeley, Calif.: Institute of Governmental Studies, University of California at Berkeley,
2006.
Cameron, Charles, David Epstein, and Sharyn O'Halloran. "Do Majority-Minority Districts Maximize
Substantive Black Representation in Congress?" American PoliticalScience Review 90 (1996):
794-812.
Campagna, Janet, and Bernard Grofman. "Party Control and Partisan Bias in 1980s Congressional
Redistricting." JournalofPolitics 79 (1990): 1242-1257.
Campbell, James E. Cheap Seats: The DemocraticParty'sAdvantage in U.S. House Elections. Columbus,
Ohio: Ohio State University Press, 1996.
Campbell, James E., John R. Alford, and Keith Henry. "Television Markets and Congressional Elections."
Legislative Studies Quarterly9 (1984): 665-678.
156
Canon, David T. Race, Redistricting, and Representation: The Unintended Consequencesof Black Majority
Districts. Chicago: University of Chicago Press, 1999.
Carson, Jamie L., and Michael H. Crespin. "The Effect of State Redistricting Methods on Electoral
Competition in United States House of Representatives Races." Politics andPolicy Quarterly4
(2004): 455-469.
Carson, Jamie L., Michael H. Crespin, Charles J. Finocchaiaro, and David W. Rohde. "Redistricting and
Party Polarization in the U.S. House of Representatives." American Politics Research 35 (2007):
878-904.
Chen, Jowei, and Jonathan Rodden. "Tobler's Law, Urbanization, and Electoral Bias: Why Compact,
Contiguous Districts are Bad for the Democrats." Working paper, 2009.
Cho, Yong Hyo, and H. George Frederickson. Measuring the Effects ofReapportionment in the American
States. New York: National Municipal League, 1976.
Clayton, Dewey M. African Americans and the Politicsof CongressionalDistricting.New York: Garland,
2000.
Collie, Melissa P. "Incumbency, Electoral Safety, and Turnover in the House of Representatives, 19521976." American PoliticalScience Review 75 (1981): 119-131.
Congressional Quarterly. CongressionalDistricts in the 1970s. 2nd Edition. Washington: Congressional
Quarterly, 1974.
-.
-.
CongressionalDistricts in the 1980s. Washington: Congressional Quarterly, 1983.
CongressionalRedistricting:Impact of the 1960 Census Reapportionmentof House Seats. Washington:
Congressional Quarterly, 1962.
"Congressional Reapportionment and Partisan Deadlock." In The PoliticsofReapportionment, edited by
Malcolm E. Jewell, 131-133. New York: Atherton Press, 1962.
Cover, Albert D. "One Good Term Deserves Another: The Advantage of Incumbency in Congressional
Elections." American JournalofPoliticalScience 21 (1977): 523-541.
Cox, E.P. "A Method of Assigning Numerical and Percentage Values to the Degree of Roundness of Sand
Grains." JournalofPaleontology 1 (1927): 179-183.
157
Cox, Gary W., and Jonathan N. Katz. Elbridge Gerry's Salamander: The Electoral Consequences of the
ReapportionmentRevolution. Cambridge, U.K.: Cambridge University Press, 2002.
Cummings Jr., Milton C. "Reapportionment in the 1970s: Its Effect on Congress." In Reapportionmentin
the 1970s, edited by Nelson Polsby, 209-241. Berkeley, Calif.: University of California Press,
1971.
Cunningham, Maurice T. Maximization, Whatever the Cost: Race, Redistricting,and the Department of
Justice. Westport, Conn.: Praeger, 2001.
Dauer, Manning J., and Robert G. Kelsay. "Unrepresentative States." NationalMunicipalReview 44
(1955): 571-575; 587.
de Grazia, Alfred. "The Applied Science of Equality: The Case for Apportionment, with Special Attention
to the Idea of Equi-Populous Districts." In Representation and Misrepresentation:Legislative
Reapportionment in Theory and Practice,edited by Robert A. Goldwin, 169-189. Chicago: Rand
McNally, 1968.
Desposato, Scott W., and John R. Petrocik. "The Variable Incumbency Advantage: New Voters,
Redistricting, and the Personal Vote." American JournalofPoliticalScience 47 (2003): 18-32.
Dixon, Robert G. DemocraticRepresentation:Reapportionment in Law andPolitics.New York: Oxford
University Press, 1968.
Dixon, Robert G. "Fair Criteria and Procedures for Establishing Legislative Districts." In Representation
and RedistrictingIssues, edited by Bernard Grofman, 7-19. Lexington, Mass.: D.C. Heath, 1982.
Dixon, Robert G. "The Court, The People, and One Man, One Vote." In Reapportionment in the 1970s,
edited by Nelson W. Polsby, 7-45. Berkeley, Calif.: University of California Press, 1971.
Engstrom, Richard L., and Jason E. Kirksey. "Race and Representational Districting in Louisiana." In Race
andRedistrictingin the 1990s, edited by Bernard Grofinan, 229-168. New York: Agathon Press,
1998.
Erikson, Robert S. "Malapportionment, Gerrymandering, and Party Fortunes in Congressional Elections."
American PoliticalScience Review 66 (1972): 1234-1245.
Erikson, Robert S. "Sources of partisan bias in US congressional elections: an update stimulated by Ron
Johnston's essay." PoliticalGeography21 (2002): 49-54.
158
Erikson, Robert S. "The Advantage of Incumbency in Congressional Elections." Polity 3 (1971): 395-405.
Fenno, Richard F. Congress at the Grassroots:RepresentationalChange in the South, 1970-1998. Chapel
Hill, N.C.: University of North Carolina Press, 2000.
Ferejohn, John. "The Decline of Competition in Congressional Elections." American PoliticalScience
Review 71 (1977): 166-176.
Fiorina, Morris. "The Case of the Vanishing Marginals: The Bureaucracy Did It." American Political
Science Review 71 (1977): 177-181.
Flaherty, Mark S., and William M. Crumplin. "Compactness and Electoral Boundary Adjustment: An
Assessment of Alternative Measures." CanadianGeographer36 (1992): 159-171.
Forgette, Richard, and Glenn Platt. "Redistricting principles and incumbency protection in the U.S.
Congress." PoliticalGeography24 (2005): 934-951.
Garand, James C., and Donald A. Gross. "Changes in the Vote Margins for Congressional Candidates: A
Specification of Historical Trends." American PoliticalScience Review 78 (1984): 17-30.
Gay, Claudine. "Legislating Without Constraints: The Effect of Minority Districting on Legislators'
Responsiveness to Constituency Preferences." JournalofPolitics 69 (2007): 442-456.
Gelman, Andrew, and Gary King. "A Unified Method of Evaluating Electoral Systems and Redistricting
Plans." American JournalofPoliticalScience 38 (1994): 514-554.
Gelman, Andrew, and Gary King. "Enhancing Democracy Through Legislative Redistricting." American
PoliticalScience Review 88 (1994): 541-559.
Gelman, Andrew, and Gary King. "Estimating Incumbency Advantage Without Bias." American Journalof
PoliticalScience 34 (1990): 1142-1164.
Gimpel, James G., and Frances E. Lee. "The Geography of Electioneering: Campaigning for Votes and
Campaigning for Money." In The Marketplace ofDemocracy: Electoral Competition and
American Politics,edited by Michael P. McDonald and Frances E. Lee, 125-148. Washington:
The Brookings Institution, 2006.
Glazer, Amihai, Bernard Grofnan, and Marc Robbins. "Partisan and Incumbency Effects of 1970s
Congressional Redistricting." American JournalofPoliticalScience 31 (1987): 680-707.
159
Grofman, Bernard. "Criteria for Districting: A Social Science Perspective." UCLA Law Review 33 (1985):
77-184.
Grofman, Bernard. "For Single Members Districts, Random Is Not Equal." In Representationand
RedistrictingIssues, edited by Bernard Grofinan, Arend Lijphart, Robert B. McKay and Howard
A. Scarrow, 55-58. Lexington, Mass.: D.C. Heath, 1982.
Grofman, Bernard. "Toward a Coherent Theory of Gerrymandering: Bandemer and Thornburg." In
PoliticalGerrymanderingand the Courts, edited by Bernard Grofman, 29-63. New York:
Agathon, 1990.
-.
Voting Rights, Voting Wrongs: The Legacy of Baker v. Carr.New York: Priority Press Publications,
1990.
Grofman, Bernard, and Lisa Handley. "Estimating the Impact of Voting-Rights-Related Districting on
Democratic Strength in the U.S. House of Representatives." In Race andRedistricting in the
1990s, edited by Bernard Grofman, 51-65. New York: Agathon Press, 1998.
Gudgin, Graham, and Peter J. Taylor. "Electoral Bias and the Distribution of Party Voters." Transactionsof
the Institute of British Geographers63 (1974): 53-74.
-.
Seats, Votes, and the Spatial Organisationof Elections. London: Pion, 1979.
Gudgin, Graham, and Peter J. Taylor. "The Decomposition of Electoral Bias in a Plurality Election." British
JournalofPoliticalScience 10 (1980): 515-521.
Hacker, Andrew. CongressionalDistricting: The Issue of Equal Representation.Revised Edition.
Washington: The Brookings Institution, 1964.
Hagens, Winnett W. "The Politics of Race: The Virginia Redistricting Experience, 1991-1997." In Race
andRedistrictingin the 1990s, edited by Bernard Grofman, 315-342. New York: Agathon Press,
1998.
Haggett, Peter, Andrew D. Cliff, and Allan E. Frey. LocationalAnalysis in Human Geography. London:
Edward Arnold, 1966.
Handley, Lisa. "Drawing Effective Minority Districts: A Conceptual Model." In Voting Rights and
Minority Representation:Redistricting, 1992-2002, edited by David A. Bositis, 51-72. Lanham,
Md.: University Press of America, 2006.
160
Handley, Lisa, Bernard Grofman, and Wayne Arden. "Electing Minority-Preferred Candidates to
Legislative Office: The Relationship Between Minority Percentages in Districts and the Election
of Minority-Preferred Candidates." In Race and Redistricting in the 1990s, edited by Bernard
Grofman, 13-38. New York: Agathon Press, 1998.
Hanson, Royce. The PoliticalThicket: Reapportionmentand ConstitutionalDemocracy. Englewood Cliffs,
N.J.: Prentice-Hall, 1966.
Harris, Curtis C. "A Scientific Method of Districting." Behavioral Science 9 (1964): 219-225.
Hayes, Brian. "Machine Politics." American Scientist, November-December 1996.
Hill, Tony L. CanadianPolitics,Riding by Riding: An In-Depth Analysis of Canada's301 Federal
ElectoralDistricts.Minneapolis: Prospect Park Press, 2002.
-.
"District Compactness and Electoral Volatility in Canada and the United States." Presented at the
national conference of the Midwest Political Science Association, Chicago, April 2-5, 2009.
-.
"Electoral Bias and the New Democratic Party: Provincial Competition in Manitoba, Saskatchewan,
and British Columbia." Paper presented at the annual convention of the Society for Socialist
Studies, Winnipeg, June 3-6, 2004.
-.
"Electoral Bias and the Partisan Impact of Independent Redistricting Bodies." Presented at the national
conference of the Midwest Political Science Association, Chicago, April 3-6, 2008.
Hofeller, Thomas, and Bernard Grofman. "Comparing the Compactness of California Congressional
Districts Under Three Different Plans: 1980, 1982, and 1984." In PoliticalGerrymanderingand
the Courts, edited by Bernard Grofman, 281-288. New York: Agathon Press, 1990.
Holmes, Bob. "Reapportionment/Redistricting Politics in Georgia in the 1990s and 2001-2002: Reflections
of a Participant-Observer." In Voting Rights and Minority Representation:Redistricting,19922002, edited by David A. Bositis, 87-112. Lanham, Md.: University Press of America, 2006.
Holmes, Richard A. "Reapportionment Strategies in the 1990s: The Case of Georgia." In Race and
Redistrictingin the 1990s, edited by Bernard Grofman, 191-228. New York: Agathon Press, 1998.
Horn, David L., Charles R. Hampton, and Anthony J. Vandenberg. "Practical Application of District
Compactness." PoliticalGeography 12 (1993): 103-120.
161
Huckabee, David C. The House of RepresentativesApportionment Formula:An Analysis of Proposalsfor
Change and Their Impact on States. CRS Report for Congress, Washington: Congressional
Research Service, Library of Congress, 2001.
Issacharoff, Samuel. "Gerrymandering and Political Cartels." HarvardLaw Review 116 (2002): 593-648.
Issacharoff, Samuel. "Has the Supreme Court Destabilized Single-Member Districts?" In Voting and
Democracy Report 1995. Takoma Park, Md.: Center for Voting and Democracy, 1995.
Jacobson, Gary C. "Competition in U.S. Congressional Elections." In The Marketplace ofDemocracy:
Electoral Competition andAmerican Politics,edited by Michael P. McDonald and John Samples,
27-52. Washington: The Brookings Institution, 2006.
Jacobson, Gary C., and Samuel Kernell. Strategy and Choice in CongressionalElections. Second Edition.
New Haven, Conn.: Yale University Press, 1983.
Jewell, Malcolm E. Metropolitan Representation:State LegislativeDistrictingin Urban Counties. New
York: National Municipal League, 1969.
Jewell, Malcolm E. "Political Patterns in Apportionment." In The PoliticsofReapportionment, edited by
Malcolm E. Jewell, 1-48. New York: Atherton Press, 1962.
Jewell, Malcolm E. "The Consequences of Single- and Multimember Districting." In Representation and
RedistrictingIssues, 129-135. Lexington, Mass.: D.C. Heath, 1982.
Johnston, Ron J., interview by Tony L. Hill. (2006).
Johnston, Ron J. "Ifit isn't a gerrymander, what is it?" PoliticalGeography 21 (2002): 55-65.
Johnston, Ron J. "Manipulating maps and winning elections: measuring the impact of malapportionment
and gerrymandering." PoliticalGeography 21 (2002): 1-31.
Johnston, Ron J. "Spatial Structure, Plurality Systems and Electoral Bias." CanadianGeographer20
(1976): 310-328.
Johnston, Ron J., and Charles Pattie. "Money and votes: a New Zealand example." PoliticalGeography 27
(2008): 113-133.
Johnston, Ron J., Charles Pattie, Danny F.L. Dorling, and David J. Rossiter. From Votes to Seats: The
UnitedKingdom's ElectoralSystem in Operation Since 1945. Manchester, U.K.: Manchester
University Press, 2001.
162
Johnston, Ron J., David Rossiter, and Charles Pattie. "Changing the scale and changing the result:
Evaluating the impact of an electoral reform on the 2000 and 2004 US Presidential elections."
PoliticalGeography 25 (2006): 557-569.
Johnston, Ron J., David Rossiter, and Charles Pattie. "Disproportionality and bias in US Presidential
Elections: How geography helped Bush defeat Gore but couldn't help Kerry beat Bush." Political
Geography 24 (2005): 952-968.
Kaiser, Henry F. "An Objective Method for Establishing Legislative Districts." Midwest Journalof
PoliticalScience 10 (1966): 200-213.
Kang, Michael S. "De-Rigging Elections: Direct Democracy and the Future of Redistricting Reform."
Washington University Law Review 84 (2006): 667-716.
Karch, Andrew, Corrine M. McConnaughy, and Sean M. Theriault. "The Legislative Politics of
Congressional Redistricting Commission Proposals." American PoliticsResearch 35 (2007): 808825.
Karlan, Pamela S. "Maps and Misreadings: The Role of Geographic Compactness in Racial Vote Dilution
Litigation." HarvardCivil Rights-Civil Liberties Law Review 24 (1989): 173-248.
Kendall, M.G., and A. Stewart. "The Law of Cubic Proportions in Election Results." British Journalof
Sociology 1 (1950): 183-197.
King, Gary, and Andrew Gelman. "Systemic Consequences of Incumbency Advantage in U.S. House
Elections." American JournalofPoliticalScience 35 (1991): 110-138.
Kousser, Morgan. "Reapportionment Wars: Party, Race, and Redistricting in California, 1971-1992." In
Race and Redistrictingin the 1990s, edited by Bernard Grofman, 134-190. New York: Agathon
Press, 1998.
Kramer, Daniel C. "'Those people across the water just don't know our problems': An Analysis of Friends
and Neighbours Voting in a Geographically Split Legislative District." PoliticalGeography
Quarterly9 (1990): 189-196.
Krashinsky, Michael, and William J. Milne. "The Effects of Incumbency in U.S. Congressional Elections,
1950-1988." Legislative Studies Quarterly 18 (1993): 321-344.
163
Lee, David R., and G. Thomas Sallee. "A Method of Measuring Shape." GeographicalReview 60 (1970):
555-563.
Levinson, Sanford. "Gerrymandering and the Brooding Omnipresence of Proportional Representation: Why
Won't It Go Away?" UCLA Law Review 33 (1995): 257-281.
Levitt, Stephen D., and Catherine D. Wolfram. "Decomposing the Sources of Incumbency Advantage in
the U.S. House." Legislative Studies Quarterly22 (1997): 45-60.
Lijphart, Arend. "Comparative Perspectives on Fair Representation: The Plurality-Majority Rule,
Geographical Districting, and Alternative Electoral Arrangements." In Representation and
RedistrictingIssues, edited by Bernard Grofinan, Arend Lijphart, Robert B. McKay and Howard
A. Scarrow, 143-159. Lexington, Mass.: D.C. Heath, 1982.
Lowenstein, Daniel H., and Jonathan Steinberg. "The Quest for Legislative Districting in the Public
Interest: Elusive or Illusory?" UCLA Law Review 33 (1985): 1-75.
Lublin, David. "Racial Redistricting and Southern Republican Congressional Gains in the 1990s." In Voting
Rights and Minority Representation:Redistricting, 1992-2002, edited by David A. Bositis, 113129. Lanham, Md.: University Press of America, 2006.
-.
The Paradoxof Representation:Racial Gerrymanderingand Minority Interests in Congress. Princeton,
N.J.: Princeton University Press, 1997.
Lublin, David, and D. Stephen Voss. "The Missing Middle: Why Median-Voter Theory Can't Save
Democrats from Singing the Boll-Weevil Blues." JournalofPolitics 65 (2003): 227-237.
Lyons, Michael, and Peter F. Galderisi. "Incumbency, Reapportionment, and U.S. House Redistricting."
PoliticalResearch Quarterly48 (1995): 857-871.
MacEachren, Alan M. "Compactness of Geographic Shape: Comparison and Evaluation of Measures."
GeografiskaAnnaler 67 (1985): 53-67.
March, James G. "Party Legislative Representation as a Function of Election Results." Public Opinion
Quarterly 21 (1957): 521-542.
Mayhew, David. "Congressional Elections: The Case of the Vanishing Marginals." Polity 6 (1974): 295317.
164
Mayhew, David. "Congressional Representation: Theory and Practice in Drawing the Districts." In
Reapportionmentin the Seventies, edited by Nelson Polsby, 249-285. Berkeley, Calif.: University
of California Press, 1971.
McDonald, Michael P. "A Comparative Analysis of Redistricting Institutions in the United States, 200102." Politicsand Policy Quarterly 4 (2004): 371-395.
McDonald, Michael P. "Redistricting and Competitive Districts." In The Marketplace ofDemocracy:
ElectoralCompetition and American Politics, edited by Michael P. McDonald and Frances E. Lee,
222-244. Washington: The Brookings Institution, 2006.
McDonald, Michael P. "Re-Drawing the Line on District Competition." PS: PoliticalScience andPolitics
39 (2006): 99-101.
McDonald, Michael P. "Regulating Redistricting." PS: PoliticalScience and Politics 40 (2007): 675-679.
McDonald, Michael P. "Seats to Votes Ratios in the United States." Working paper, 2006.
McDonald, Michael P., and Richard L. Engstrom. "Detecting Gerrymandering." In Political
Gerrymanderingand the Courts, edited by Bernard Grofman, 178-202. New York: Agathon Press,
1990.
McDonald, Michael P., and Samuel L. Popkin. "The Myth of the Vanishing Voter." American Political
Science Review 95 (2001): 963-974.
McKay, Robert B. "Affirmative Gerrymandering." In Representation and RedistrictingIssues, edited by
Bernard Grofman, Arend Lijphart, Robert B. McKay and Howard A. Scarrow, 91-94. Lexington,
Mass.: D.C. Heath, 1982.
McKee, Seth C., and Jeremy M. Teigen. "Probing the reds and blues: Sectionalism and voter location in the
2000 and 2004 U.S. presidential elections." PoliticalGeography28 (2009): 484-495.
Mellow, Nicole, and Peter Trubowitz. "Red versus blue: American electoral geography and congressional
bipartisanship, 1898-2002." PoliticalGeography 24 (2005): 659-677.
Milnor, A.J. Elections andPoliticalStability: An Analytic Study. Boston: Little, Brown, 1969.
Minar, David W. "Equilibrium in Illinois: Frustration and Accomodation of the Parties." In The Politicsof
Reapportionment, edited by Malcolm E. Jewell, 134-148. New York: Atherton Press, 1962.
165
Monmonier, Mark E. Bushmanders and Bullwinkles: How PoliticiansManipulateElectronic Maps and
Census Data to Win Elections. Chicago: University of Chicago Press, 2001.
Monomonier, Mark E. Drawing the Line: Tales of Maps and Cartocontroversy.New York: Henry Holt,
1995.
Moore, Toby. "Comments on Ron Johnston's 'Manipulating maps and winning elections: measuring the
impact of malapportionment and gerrymandering'." PoliticalGeography 21 (2002): 33-38.
Morrill, Richard L. "A Geographer's Perspective." In PoliticalGerrymanderingand the Courts, edited by
Bernard Grofnan, 212-239. New York: Agathon Press, 1990.
Morrill, Richard L. "Electoral Geography and Gerrymandering: Space and Politics." In Reordering the
World: GeopoliticalPerspectives on the Twenty-First Century, edited by George J. Demko and
William B. Wood, 117-138. Boulder, Colo.: Westview Press, 1999.
-.
PoliticalRedistrictingand GeographicTheory. Washington: Association of American Geographers,
1981.
Morrill, Richard L. "Redistricting Standards and Strategies after 20 Years." PoliticalGeographyQuarterly
1 (1982): 361-369.
Morrill, Richard L. "Redistricting, Region and Representation." PoliticalGeography Quarterly 6 (1987):
241-260.
Morrill, Richard L., Larry Knopp, and Michael Brown. "Anomalies in red and blue: Exceptionalism in
American electoral geography." PoliticalGeography 26 (2007): 525-553.
Musgrove, Philip. The General Theory of Gerrymandering.Beverly Hills, Calif.: Sage Publications, 1977.
Niemi, Richard G. "The Effects of Districting on Trade-offs among Party Competition, Electoral
Responsiveness, and Seats-Votes Relationships." In Representation and RedistrictingIssues,
edited by Bernard Grofman, Arend Lijphart, Robert B. McKay and Howard A. Scarrow, 35-42.
Lexington, Mass.: D.C. Heath, 1982.
Niemi, Richard G. "The Swing Ratio as a Measure of Partisan Gerrymandering." In Political
Gerrymanderingand the Courts, edited by Bernard Grofman, 171-177. New York: Agathon Press,
1990.
166
Niemi, Richard G., and John Deegan. "A Theory of Political Districting." American PoliticalScience
Review 72 (1978): 1304-1323.
Niemi, Richard G., and John Wilkerson. "Compactness and the 1980s Districts in the Indiana State House:
Evidence of Political Gerrymandering?" In PoliticalGerrymanderingand the Courts, edited by
Bernard Grofman, 255-265. New York: Agathon Press, 1990.
Niemi, Richard G., and Stephen G. Wright. "Majority-Win Percentages: An Approach to the Votes-Seats
Relationship in Light of Davis v. Bandemer." In PoliticalGerrymanderingand the Courts, edited
by Bernard Grofman, 266-278. New York: Agathon Press, 1990.
Niemi, Richard G., Bernard Grofman, Carl Carlucci, and Thomas Hofeller. "Measuring Compactness and
the Role of a Compactness Standard in a Test for Partisan and Racial Gerrymandering." Journalof
Politics 52 (1990): 1155-1181..
Niemi, Richard G., Lynda W. Powell, and Patricia L. Bicknell. "The Effects of Congruity Between
Community and District on Salience of U.S. House Candidates." Legislative Studies Quarterly 11
(1986): 187-201.
Niemi, Richard G., Lynda W. Powell, William D. Berry, Thomas M. Carsey, and James M. Snyder Jr.
"Competition in State Legislative Elections, 1992-2002." In The Marketplace ofDemocracy:
ElectoralCompetition andAmerican Politics,edited by Michael P. McDonald and John Samples,
53-73. Washington: The Brookings Institution, 2006.
O'Loughlin, John, and Anne-Marie Taylor. "Choices in Redistricting and Electoral Outcomes: The Case of
Mobile, Alabama." PoliticalGeography Quarterly 1 (1982): 317-339.
O'Rourke, Timothy G. The Impact ofReapportionment.New Brunswick, N.J.: Transaction Books, 1980.
Ostdiek, Donald. "Congressional Redistricting and District Typologies." JournalofPolitics 57 (1995): 533543.
Owen, Guillermo, and Bernard Grofman. "Optimal Partisan Gerrymandering." PoliticalGeography
Quarterly7 (1988): 5-22.
Papayanopoulos, Lee. "Compromise Districting." In RepresentationandRedistrictingIssues, edited by
Bernard Grofinan, Arend Lijphart, Robert B. McKay and Howard A. Scarrow, 59-63. Lexington,
Mass.: D.C. Heath, 1982.
167
Payne, James L. "The Personal Electoral Advantage of House Incumbents, 1936-1976." American Politics
Quarterly 8 (1980): 465-482.
Petrocik, John R., and Scott W. Desposato. "The Partisan Consquences of Majority-Minority Redistricting
in the South, 1992 and 1994." JournalofPolitics 60 (1998): 613-633.
Phillips, Kevin P. The Emerging Republican Majority. New York: Arlington House, 1969.
Pildes, Richard H., and Richard G. Niemi. "Expressive Harms, "Bizarre Districts," and Voting Rights:
Evaluating Election-District Appearances after Shaw v. Reno." Michigan Law Review 92 (1994):
483-587.
Polsby, Daniel D., and Robert D. Popper. "The Third Criterion: Compactness as a Procedural Safeguard
Against Partisan Gerrymandering." Yale Law and Policy Review 9 (1991): 301-353.
Polsby, Daniel D., and Robert D. Popper. "Ugly: An Inquiry into the Problem of Racial Gerrymandering
Under the Voting Rights Act." MichiganLaw Review 92 (1993): 652-682.
Pounds, Norman J.G. PoliticalGeography. 2nd Edition. New York: McGraw-Hill, 1972.
Prior, Markus. "The Incumbent in the Living Room: The Rise of Television and the Incumbency
Advantage in U.S. House Election." JournalofPolitics 68 (2006): 657-673.
Pritchett, C. Herman. "Representation and the Rule of Equality." In Representationand Misrepresentation:
Legislative Reapportionmentin Theory and Practice,edited by Robert A. Goldwin, 1-20.
Chicago: Rand McNally, 1968.
Rae, Douglas W. "Reapportionment and Political Democracy." In Reapportionment in the 1970s, edited by
Nelson Polsby, 91-112. Berkeley, Calif.: University of California Press, 1971.
-.
The PoliticalConsequences ofElectoralLaws. Revised Edition. New Haven, Conn.: Yale University
Press, 1971.
Rasky, Susan. "The Bottom Line -- The Age of Do-It-Yourself Redistricting." CaliforniaJournal,February
1,2000.
Reock, Ernest C. "Measuring Compactness as a Requirement of Legislative Reapportionment." Midwest
JournalofPoliticalScience 5 (1961): 70-74.
Rossiter, David, Ron J. Johnston, and Charles J. Pattie. "Redistricting and Electoral Bias in Great Britain."
British Journalof PoliticalScience 27 (1997): 466-472.
168
Rush, Mark E. Does RedistrictingMake a Difference?: PartisanRepresentation and ElectoralBehavior.
Baltimore: Johns Hopkins University Press, 1993.
Rush, Mark E. "Redistricting and Partisan Fluidity: Do We Really Know a Gerrymander When We See
One?" PoliticalGeography 19 (2000): 249-260.
Saffel, David. "The Policy Consequences of Reapportionment: State Legislators' Perspectives." In
Representation and RedistrictingIssues, edited by Bernard Grofman, Arend Lijphart, Robert B.
McKay and Howard A. Scarrow, 203-219. Lexington, Mass.: D.C. Heath, 1982.
Scarrow, Howard A. "The Impact of Reapportionment on Party Representation in the State of New York."
In Representation and RedistrictingIssues, edited by Bernard Grofman, Arend Lijphart, Robert B.
McKay and Howard A. Scarrow, 223-236. Lexington, Mass.: D.C. Heath, 1982.
Schaffner, Brian F. "The political geography of campaign advertising in U.S. House elections." Political
Geography25 (2006): 775-788.
Schattschneider, E.E. Party Government: American Government in Action. New York: Rinehart, 1942.
Schousen, Matthew M., David T. Canon, and Patrick J. Sellers. "Representation and Ambition in the New
African-American Congressional Districts: The Supply Side Effects." In Race and Redistrictingin
the 1990s, edited by Bernard Grofman, 39-50. New York: Agathon Press, 1998.
Schubert, Glendon. Reapportionment.New York: Charles Scribner's Sons, 1965.
Schuck, Peter. "Partisan Gerrymandering: A Political Problem Without Judicial Solution." In Political
Gerrymanderingand the Courts, edited by Bernard Grofman, 240-251. New York: Agathon Press,
1990.
Schumm, Stanley A. "Evolution of Drainage Systems and Slopes in Badlands at Perth Amboy, New
Jersey." Bulletin of the GeologicalSociety ofAmerica 67 (1956): 597-646.
Schwab, Larry M. The Impact of CongressionalReapportionmentand Redistricting. Lanham, Md.:
University Press of America, 1988.
Schwartzberg, Joseph E. "Reapportionment, Gerrymanders, and the Notion of 'Compactness'." Minnesota
Law Review 50 (1966): 443-452.
169
Sellers, Patrick J., David T. Canon, and Matthew M. Schousen. "Congressional Redistricting in North
Carolina." In Race and Redistrictingin the 1990s, edited by Bernard Grofman, 269-289. New
York: Agathon Press, 1998.
Shapiro, Martin. "Gerrymandering, Unfairness, and the Supreme Court." UCLA Law Review 33 (1985):
227-256.
Siaroff, Alan. "Electoral Bias in Quebec Since 1936." Paper presented at the 17th biennial conference of
the Association for Canadian Studies in the United States (ACSUS), Portland, Ore., November 1923, 2003.
Sickels, Robert J. "Dragons, Bacon Strips, and Dumbbells: Who's Afraid of Reapportionment?" Yale Law
Journal75 (1966): 1300-1308.
Snyder, James M., and David.Stramberg. "Press Coverage and Political Accountability." Working paper,
2010.
Stein, Mark. How the States Got Their Shapes. New York: Collins, 2008.
Stern, Robert S. "Political Gerrymandering: A Statutory Compactness Standard as an Antidote for Judicial
Impotence." University of Chicago Law Review 41 (1974): 398-416.
Stewart, Charles H. Analyzing Congress.New York: W.W. Norton, 2001.
Stewart, Charles H., and Mark Reynolds. "Television Markets and U.S. Senate Elections." Legislative
Studies Quarterly 15 (1990): 495-523.
Stokes, Donald. "Is There a Better Way to Redistrict?" In Race and Redistrictingin the 1990s, edited by
Bernard Grofman, 345-366. New York: Agathon Press, 1998.
Swain, John W., Stephen A. Borrelli, and Brian C. Reed. "Partisan Consequences of the Post-1990
Redistricting for the U.S. House of Representatives." PoliticalResearch Quarterly 51 (1998): 945967.
Taagapera, Rein. "Seats and Votes: A Generalisation of the Cube Law of Elections." Social Science
Research 2 (1973): 257-275.
Taylor, Peter J. "A New Shape Measure for Evaluating Electoral District Patterns." American Political
Science Review 67 (1973): 947-950.
Thernstrom, Abigail. "Redistricting, Race, and the Voting Rights Act." NationalAffairs 3 (2010): 52-67.
170
Tufte, Edward R. "The Relationship Between Seats and Votes in Two-Party Systems." American Political
Science Review 67 (1973): 540-554.
U.S. House of Representatives. H.R. Report No. 140, 89th Congress, 1st Session. Washington: U.S. House
of Representatives, 1965.
United States Department of Commerce, Bureau of the Census. Census ofPopulationandHousing (1990):
Summary Social, Economic, andHousing Characteristics.Washington: Government Printing
Office, 1992.
Vickrey, William. "On the Prevention of Gerrymandering." PoliticalScience Quarterly 76 (1961): 105110.
Wahlke, John C. "Logic and Politics in Electoral Engineering." In Representationand RedistrictingIssues,
edited by Bernard Grofman, Arend Lijphart, Robert B. McKay and Howard A. Scarrow, 161-168.
Lexington, Mass.: D.C. Heath, 1982.
Ward, Elliott. "Prometheus, Proteus, Pandora, and Procrustes Unbound: The Political Consequences of
Reapportionment." University of Chicago Law Review 37 (1970): 474-492.
Warf, Barney. "Voting technologies and residual ballots in the 2000 and 2004 presidential elections."
PoliticalGeography 25 (2006): 530-556.
Weaver, James B., and Sidney W. Hess. "A Procedure for Nonpartisan Districting: Development of
Computer Techniques." Yale Law Journal73 (1963): 288-308.
Wells, David. "Against Affirmative Gerrymandering." In Representation andRedistrictingIssues, edited
by Bernard Grofman, Arend Lijphart, Robert B. McKay and Howard A. Scarrow, 91-94.
Lexington, Mass.: D.C. Heath, 1982.
Yarbrough, Tinsley E. Race and Redistricting: The Shaw-CromartieCases. Lawrence, Kans.:: University
Press of Kansas, 2002.
Yoshinaka, Antoine, and Chad Murphy. "Partisan gerrymandering and population instability: Completing
the redistricting puzzle." PoliticalGeography28 (2009): 451-462.
Young, H. Peyton. "Measuring the Compactness of Legislative Districts." Legislative Studies Quarterly 13
(1988): 105-115.
NEIGHBORING
CONGRESSIONAL DISTRICTS IN TEXAS
1932-1964
Compactness
12th
5th
Cox
0.779
0.783
A&A
Nagel
Hill
0.470
1.133
1.004
0.466
1.130
1.001
TWO CURRENT TEXAS DISTRICTS
Measure
Cox
A&A
Nagel
Hill
Ranking
19th
District
12th District
0.155
0.919
2.539
2.250
1581/1899
0.444
0.745
1.500
1.329
329/1899
FIVE CURRENT HOUSTON-AREA DISTRICTS
Measure
Cox
A&A
Nagel
Hill
Ranking
CD 18
CD 7
0.111
0.175
0.943
0.908
2.996
2.392
2.655
2.120
1503/1899 1736/1899
CD 29
CD 9
0.116
0.145
0.940
0.924
2.938
2.623
2.604
2.324
1626/1899 1726/1899
18
CD 22
0.102
0.948
3.133
2.776
1763/1899
I
Arizona CD 4 (Hill 1.16)
MOST COMPACT DISTRICTS IN UNITED STATES AND CANADA
Langley ridinq, B.C. (Hill 1.03)
Ohio CD 3 (Hill 1.91)
MEDIAN COMPACT DISTRICTS IN UNITED STATES AND CANADA
Mount Royal riding, Que. (Hill 1.36)
Florida CD 4 (Hill 5.40)
LEAST COMPACT DISTRICTS IN UNITED STATES AND CANADA
Okanagan-Coquihalla riding, B.C. (Hill 2.09)
THEBETTMANN
ARCHIVE
GERRYMANDER, a fictional creature based on the shape of an electoral district of Massachusetts, as set up for political reasons.
Measure
Cox
A&A
Nagel
Hill
Ranking
Gerry
0.084
0.957
3.444
3.052
1807/1899