National Survival and the Confederate Congress
Adam Ramey**
ABSTRACT:
I analyze the voting behavior of legislators in the Congresses of the Confederate States of
America during the American Civil War. I show that the occupation of Confederate
Congressional districts by Federal troops led legislators to abandon their previous voting
behavior and instead support the strengthening of the central government in Richmond. Specific
case evidence involving voting on a number of salient issues is provided to further demonstrate
the robustness of this result. Most important, the result leads to outcomes at odds with the logic
of secession as enunciated by Southern elites.

I would like to thank the Lanni-Wallis summer research fellowships for financial support during the course of this
project. I am also indebted to Keith Poole and Jeff Jenkins for advice and data, respectively. Information gathered on
legislators' characteristics would not have been possible without the diligent efforts of Alexander and Beringer. I
thank Larry Rothenberg for taking on the the task of advising me on this project. I would also like to thank
Dick Niemi, Lynda Powell, Gerald Gamm, Jeremy Kedziora, Shawn Ramirez, Michael Peress, Stu Jordan, David
Primo, and Curt Signorino for comments, suggestions, guidance, and general assistance. All remaining errors are my
own.
**
Assistant Professor of Politics, NYU Abu Dhabi, adam.ramey@nyu.edu
1
In Congress: The Electoral Connection, David Mayhew made the oft-cited observation
that members of Congress are “single-minded seekers of reelection” (1974, 17). Pork-barrel
spending, candidates’ declarations of independence from national parties during elections, and
legislators spending vast amounts of time on constituency casework all lend credence to this
observation. On the other hand, philosopher and politician Edmund Burke argued against this
view in 1774, insisting instead that a representative should be a trustee: “his unbiased opinion,
his mature judgment, his enlightened conscience, he ought not to sacrifice to you, to any man, or
to any set of men living…Your representative owes you, not his industry only, but his judgment;
and he betrays, instead of serving you, if he sacrifices it to your opinion” (Burke [1774] 1999).
Indeed, the debate between these two conflicting views of representation has endured for
centuries. But what happens when the debate is no longer one of philosophical opinion, that is,
when urgent circumstances mandate that national interest trump all else? Specifically, how can
matters of national survival, coupled with a direct impact on a legislator’s district, shape his
voting behavior in the legislature?
The American Civil War provides a rich historical case that can serve in better answering
these questions. It was at this unique point in the history of America where both (a) survival of
the nation was on the line and (b) members’ districts were directly impacted by the war. Despite
the wealth of information on that era and the substantive importance of the question posed, few
scholars have analyzed it thoroughly. Most works on the Civil War era are either wholly
concerned with the military side of the conflict or, if dealing with the political situation, are
solely descriptive. Rohde, Jenkins, Carson, and Souva (2001) attempt to bridge this gap by
looking at how various crisis-related variables (e.g. war deaths) affected elections in the Union.
2
While this piece provides valuable insight to the issue of how military failure can affect political
outcomes, it deals only with the Union and focuses on the electoral, and not legislative, context.
On the Confederate side, an even smaller subset of literature has emerged addressing the
political issues confronting the legislature. Wilfred Buck Yearns (1960) provided the first
comprehensive history of the Confederate legislature from the firing upon Fort Sumter to the
collapse at Appomattox Courthouse. However, Yearns’ work was descriptive and did not seek to
explain why Confederate legislators behaved the way they did. In The Anatomy of the
Confederate Congresses, Alexander and Beringer (1972) comprehensively analyze the roll-call
voting of Confederate legislators. Their work is impressive, documenting in great detail the
various roll calls and providing scholars with a biographical directory of the legislators, complete
with occupation, personal wealth, slaveholding data, and more. Yet they fail to provide a unified
model for the voting behavior of legislators. The closest they come is providing values for the
gamma measure of association between voting behavior and several demographic variables in
bivariate fashion (1972, 300-314).
In his analysis of Confederate legislative roll calls, historian Richard Bensel (1986)
comes perhaps the closest to identifying causal mechanisms behind voting behavior in that body.
Bensel makes the striking observation that “the consolidation of economic and social controls
within the central government of the Confederacy was in fact so extensive that it calls into
question the standard interpretations of southern opposition to expand federal power in both the
antebellum and post-Reconstruction periods” (1986, 68). Further, Bensel provides evidence
suggesting that legislators from occupied districts were more likely to support the strengthening
of the federal government. Nonetheless, he too fails to provide a unified model, taking into
account the cumulative effect of the various possible independent variables. Since neither he nor
3
Alexander and Beringer employ regression, they are unable to view Confederate voting in
anything more than a pairwise, bivariate manner.
To date, only one scholar has attempted to look at all of the variables associated with the
voting behavior of Confederate legislators in a systematic framework (Jenkins 2000). In this
analysis, Jenkins employs Poole and Rosenthal’s (1997) W-NOMINATE procedure for
estimating the legislators' ideal points and then uses these to run a series of regressions, treating
the six sessions of the Confederate House of Representatives as panel data. Using this method,
Jenkins finds no evidence to support the shock of district occupation as a significant explanatory
variable. However, by employing the static W-NOMINATE and not a dynamic estimation
routine, Jenkins’ results are called into question. Moreover, this issue highlights one of the
fundamental problems in estimating legislator preferences over time.
In this paper, I seek, primarily, to fill the gap in the literature on the Confederate
Congresses and, secondarily, to address the methodological problems in dynamic ideal point
estimation in an applied setting. Specifically, I explain how the war-induced crisis affected the
roll-call voting behavior of legislators in the First and Second Confederate House of
Representatives. First, I describe the First and Second Congresses, providing information of their
context and of legislators’ individual characteristics. Therein, I examine the previous ideal point
estimation of Jenkins, surveys extant approaches to dynamic ideal point estimation, and propose
a methodological approach that produces ideal point estimates that are comparable over time.
The next section presents a theory that explains the factors that affect the shift in ideology of
legislators. I argue that the occupation of a congressional district by Union troops, given the
legislature’s survival crisis, led members to support a strong central government in Richmond.
The following section uses the estimated ideal points to evaluate the central hypothesis of this
4
paper. I also look at some specific and significant roll calls to further substantiate the result. In
Section 6, since these results are not what one would expect from a legislature that nominally
based itself upon the Jeffersonian anti-nationalist mentality, I discuss the substantive
implications of them.
The Confederate Congress
Historical and Organizational Basics
With the election of 1860 handing victory to Republican Abraham Lincoln, many leading
White southerners felt that remaining in the Union was no longer a viable option for maintaining
the Southern way of life. Seven southern states seceded soon following the election: South
Carolina, Georgia, Florida, Alabama, Louisiana, Mississippi, and Texas. In April of 1861, when
Confederate forces fired upon Fort Sumter, South Carolina, Virginia, North Carolina, Tennessee,
and Arkansas joined suit. Meanwhile, delegates from these states plus Missouri and Kentucky
convened in Montgomery, Alabama, on February 4, 1861 to begin the first session of the
Provisional Congress of the Confederate States of America (Martis 1994, 9). A constitution was
drafted and elections were scheduled. This Provisional Congress came to a close on February
17, 1862 (Martis 1994, 2). During its short existence, the Confederacy had two regular
Congresses. The First Congress began February 18, 1862 and closed February 17, 1864; the
Second began on May 2, 1864 and ended March 18, 1865 (Martis 1994, 2). The short tenure of
the Second Congress can be attributed to the Confederacy’s military surrender and the
subsequent fall of the government in Richmond.
The structure of the legislature was very similar to that of the United States. Members of
Congress (MCs) were elected in single-member districts. These districts were allocated
proportional to the state’s population. It is important to add here that, although they remained
5
part of the Union, Southern-sympathizing citizens in Kentucky and Missouri established rival
Confederate state governments and thus seated members in the Confederate Congresses (Martis
1994, 117).1 The MCs dealt with some of the very same issues as their Union counterparts.
From 1862-1865, the legislature handled hundreds of roll-calls dealing with matters as diverse as
trade and foreign affairs, central-government powers, appropriations for public works, and the
pork-barrel minutiae that have become a staple in the modern U.S. Congress (Alexander and
Beringer 1972). Thus, while this legislature was in a certain sense unique, it was in many ways
similar to its Union counterpart and, more broadly, much like contemporary Congresses.
Indeed, one of the greatest advantages in studying the Confederate Congresses or the
Confederacy, more broadly, is their remarkable similarity to the United States Congress. The
Confederate Constitution was essentially identical to that of the U.S., with a few notable
exceptions. First, slavery and states’ rights were specifically enumerated and, hence, were
“closed cases,” whereas the U.S. Constitution was vague on these points (Thomas 1979, 37).
Second, Section IX of the Confederate Constitution forbade export tariffs and captitation/direct
taxes, matters that had been the cause of quarrels in the U.S. Congress. The important
implication of these clauses was the de facto prohibition of a competitive, meaningful party
system (Alexander and Beringer 1972; Beringer 1967; Martis 1994; Bensel 1986). When the old
Democrat-Whig divisions broke apart in the aftermath of the Compromise of 1850, Southern
Whigs and Democrats came together and formed a single-issue coalition around slavery or, more
broadly, states’ rights (Jenkins 1999). Once the South seceded, the coalition had no more
meaning, as slavery and other matters of contention were taken care of in the Constitution. Thus,
1
Since both of these states were occupied from the outset, the behavior of the legislators from these districts could
potentially be quite different from their unoccupied counterparts. However, upon examination of the subsequent
data, it appears that Kentucky and Missouri legislators behaved no differently than legislators from other occupied
districts.
6
with the absence of political party as a factor, there was not immediate rationale for parties.
Practically speaking, what this means is that one has to look to other possible variables affecting
legislators’ voting behavior.
Variables affecting voting behavior in the Confederate Congresses
One particularly attractive variable is district occupation. As the war progressed, more
and more districts were occupied by Federal troops. In fact, by the midpoint of the Second
Congress, almost half of the Confederate states, mostly those along the borderlands of the
Confederacy (e.g. Arkansas, Kentucky, etc.), were occupied.2 The occupation forced legislators
to remain in Richmond and make decisions without communication with their constituents.
Intuitively, it seems reasonable that severing the electoral connection could impact the behavior
of legislators.3
In order to analyze the legislative behavior of members of Congress (MCs) given the lack
of party organization and the above-mentioned structural circumstances, it is necessary to
develop an alternative measure to categorize Confederate roll call voting. More specifically,
since party is, in the modern sense, the way in which legislative voting tends to divide members
of the body, the absence of party in the Confederate Congress must imply that some other
construct governed voting behavior. Alexander and Beringer develop such a measure and call it
the Confederate support score. They derive it by performing content analyses of all roll calls in
the Congresses (1972, 307-313). The score identifies how intensely a legislator supports a strong
central government. The range is from zero to nine, with the former representing weak support of
a powerful central government and the latter representing strong support of a powerful central
2
The data will be presented in a later section.
An interesting aside is that, despite the lack of contact with constituents and stringent circumstances on the
electorate brought about by the war, election turnout rates were relatively high (Martis 1994, 58). Further, even in
the occupied districts, those who could vote did vote, and even soldiers on the battlefield voted when they were
found.
3
7
government. Unfortunately, the measure is discrete (i.e., integer values) and does not offer very
much comparative leverage. For example, two legislators may have a score of 3 but might
actually have different, albeit nuanced, voting behaviors. Fortunately, there is a procedure
available that does provide us with a more precise interval-level tool: W-NOMINATE.
Applying W-NOMINATE to voting in the Confederate Congresses
The use of roll-call scaling has become commonplace in analyses of legislative voting
behavior. Jenkins (1999) applies Poole and Rosenthal’s (1997) W-NOMINATE (weighted,
nominal three-step estimation) scaling procedure to the Confederate roll call votes. He argues
that, upon an inspection of the substantive content of the Confederate roll-calls, the first
dimension W-NOMINATE score represents a “central government” support dimension. Scores
range from negative one to positive one, continuously. Those close to negative one represent
hostility toward a strong central government, whereas scores near positive one indicate support
of a strong central government. Further, through an analysis of the optimal classification, only
slightly more leverage is gained by the addition of higher dimensions. Thus, it is wholly
unnecessary to interpret the substantive content of the second (or any higher) dimension(s), as
they are for our purposes, theoretically irrelevant.
Jenkins (1999) does not provide evidence validating his interpretation of the content of
the first dimension’s issue space. While this may not be wholly necessary, it is fruitful to gain as
much substantial leverage as possible in describing the content of the measure. One way of
doing this is checking whether the Alexander and Beringer measure described above correlates
highly with the W-NOMINATE first dimension score. If the answer is affirmative, then we may
proceed with confidence. The correlations for both the First and Second Congresses are .82 and
.87, respectively.
8
[Insert Figure 1 about here]
Ideological and Demographic Make-up of the Congresses
The W-NOMINATE scores will now assist in painting a better ideological picture of the
Confederate Congresses. As is seen in Figure 1, the distribution of legislators’ first-dimension
[Insert Figure 2 about here]
W-NOMINATE scores is roughly normal, with the exception of the slight depression near the
center. The reader will recall that negative scores indicate dislike of strong central authority
whereas positive scores indicate support of a strong government in Richmond. The mean score
among all members (N=111) is -.007. Thus, in the First Congress, the body was essentially split
on the matter. The Second Congress was a whole other matter. In Figure 2, one notices a large
skew to the right in the distribution of the W-NOMINATE scores. The mean score for the
Second Congress was .26, indicating a large change from the First to the Second Congress.
[Insert Figure 3 about here]
The large change in the W-NOMINATE is not simply a result of a large turnover in
membership from the First to the Second Congress. Figure 3(a) shows the First Congress WNOMINATE distribution for legislators who served in both Congresses. The shape bears a great
resemblance to Figure 1. Figure 3(b) shows the distribution for incumbents in the Second
Congress. Notice that this too bears a striking similarity to its counterpart, Figure 2. Note that
this is not what one would expect when considering incumbency. Incumbents should have had
(nearly) the same distribution from Congress to Congress and non-incumbents should have been
driving the shift. However, it appears as if incumbents had a fundamental change of heart. It is
clear that something happened that affected both incumbents and non-incumbents that led to the
observed shift.
9
[Insert Table 1 about here]
The best way to tackle this question is to compare known information about these two
Congresses. Alexander and Beringer provide detailed information about district wealth,
legislators’ personal wealth, district slaveholding, partisanship, secession stance, and district
occupation status (1972, 354-389). Table 1 provides the mean of all legislators on each of these
dimensions, by Congress. Former political party takes a value of 1 if the legislator was a
Democrat and 0 if he was a Whig. Secession stance is set to 1 if the legislator was pro-secession
and 0 if he was unionist. District occupation is set to 1 if the district was occupied by U.S.
troops for at least half of the Congress’ duration and 0 otherwise.4 Personal wealth is the 1860
estate value of the legislator in 100,000s of U.S. dollars. District wealth is the average per acre
district wealth in 1860 U.S. dollars. This variable could potentially be significant if legislators
from the wealthy cotton and tobacco districts voted differently from those representing poorer
districts. Last, the district slaveholding index is an ordinal measure of the extent of slaveholding
in the district; it ranges from values of 0 (low levels) to 9 (high levels).
District wealth, personal wealth, and slaveholding stayed fairly stable from the First to
the Second Congress, although the average personal wealth did decline slightly.5 Three
observations can be made about the remaining three variables. First, the average member was
far more likely to be a unionist in the Second Congress than in the First. Second, nearly half of
all districts were under Federal occupation for at least half of the duration of the Second
Congress. Third, many more former Whigs were elected to the Second Congress than to the
This is a rather crude measure and the choice of “greater than half” is arbitrary. However, if I am incorrect in my
measurement of occupation then one would expect that I underestimate the effect of district occupation. This is
because many occupied districts were occupied for the full duration of the Congress. Thus, choosing “greater than
half” the time as my cutoff, if anything, only a small number of districts would change. As will be shown in the
results section, the variable is so significant that any further precision would make little difference.
5
Though it is tangential to this study, note that the average legislator personal wealth was $74,800 (First Congress)
and $68,000 (Second Congress). Factoring in inflation, this amounts to $1,536,590.53 and $1,396,900.48 in 2005
U.S. dollars, respectively.
4
10
First. Considering these three large shifts, some combination of these three factors could be
behind the large shift in the W-NOMINATE scores discussed above.
Unfortunately, assessing this possibility is not straightforward. Specifically, Jenkins’
(1999; 2000) dynamic analyses rely on W-NOMINATE, an inherently static method that
precludes comparisons over time. An obvious correction would be to use Poole and Rosenthal’s
dynamic version of W-NOMINATE, DW-NOMINATE. However, this not appropriate, since
there were only a small number of time periods in the Confederate Congress.6 Alternatively,
Martin and Quinn (2002) propose a dynamic Bayesian technique that is better suited to the
Confederate roll call environment. Due to computational burden, it remains a desirable yet
impractical solution.
To remedy the static ideal point estimates, I propose a simple adjustment to WNOMINATE. Following Groseclose, Snyder, and Levitt (1999; hereafter, GLS), I connect
period-by-period ideal point estimates by assuming there is potential stretch and scale in the
policy space over time. These authors apply the equivalent of a “temperature conversion” to
interest group ratings of legislators, estimating distortion parameters across years and then using
them to connect the yearly ideological ratings.
In the context of this paper, I organize the roll calls of the Confederate Congresses in
seven separate data sets – one for the Provisional Congress, four for the different sessions of the
First Confederate House, and two for the sessions of the Second Confederate House. I estimate
the ideal points using W-NOMINATE for the seven unique sessions; denote these by 𝜃𝑖𝑡 , where 𝑖
indexes individual legislators and 𝑡 indexes time. Following GLS, we assume that 𝜃𝑖𝑡 = 𝛼𝑡 +
𝛽𝑡 𝑥𝑖 + 𝜀𝑖𝑡 , where (𝛼𝑡 , 𝛽𝑡 ) are the annual distortion parameters and 𝑥𝑖 is a mean, latent preference
6
DW-NOMINATE requires, at the minimum, five periods. It is better suited to data that extends over many more
timer periods than we have with the Confederate Congress.
11
parameter for each individual. For identification, we must set a certain time period as a baseline
– i.e., 𝛼𝑡 = 0 and 𝛽𝑡 = 1; I chose the middle time period, the third session of the First House. If
we assume the error term is normally distributed, then this model may be estimated
straightforwardly. Upon running the estimation procedure, the raw ideal points may be adjusted
according to a simply formula: 𝜃̂𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 = (𝜃𝑖𝑡 − 𝛼𝑡 )/𝑏𝑡 . Henceforth, when referring to WNOMINATE scores, I refer to scores adjusted in this fashion.
[Table 2 about here]
These results in hand, we can examine how unionism, partisan divides, and district
occupation. I argue that neither party nor secession stance are at the root of this phenomenon. If
one disaggregates the mean W-NOMINATE scores by party and secession stance, my point will
be clearer. Table 2 shows the mean W-NOMINATE scores of the party and secession stance
variables, by Congress. Note that both former Democrats and secessionists (many of whom were
one and the same) went from slightly opposing a stronger central government in the First
Congress to overwhelmingly supporting the central government in the Second Congress.
Further, as was noted above, the Second Congress saw a change in partisan and secession-stance
makeup, such that the legislature was roughly evenly divided between unionists and
secessionists, former Whigs and former Democrats. It is clear that neither former party nor
secession stance is at work in this change. Consider first former political party. Notice that the
mean adjusted W-NOMINATE score of former Democrats increased dramatically from the First
to the Second Congress and their membership declined during the same time. However, the
body’s mean increased dramatically as well. If the former Democrat’s numbers were diminished,
but the body’s mean adjusted W-NOMINATE score went up, it is hard to see how party could
have been the cause. Similarly, with respect to secession stance, the secessionist’s mean W-
12
NOMINATE score increased and their membership declined, yet the body’s mean still
increased.7 Thus, it too is inadequate as an explanation. In the next section, I construct a
theoretical model demonstrating that it is district occupation status this is at the root of the
change.
Theory
When a district was occupied by Federal troops, a number of things could have happened.
One intuitive possibility is that a legislator “voted his conscience.” That is to say, a legislator
was no longer constrained by district opinion and, if he disagreed with his district on a matter, he
had the ability to use his own judgment, as opposed to adopting a “single-minded seeker of
reelection” mentality. This makes sense, in that the occupation of his district made a legislator
acutely aware of the stakes at hand. Failure to provide the government in Richmond with the
resources necessary to conduct the war, even if in opposition to district opinion, would result in
the downfall of the Confederate nation and, consequently, the loss of his job. Another possibility
is that he became a “delegate” in the fashion of Edmund Burke, focusing his efforts on the
Confederacy’s best interests and setting aside the wants and desires of his district.8 Of course,
there is the possibility that there was no change, that a legislator just stayed on the path he was
on before the change in situation.
Which of these best applies to the Confederate Congresses? The answer to this question
lies in the nature of the circumstances. Not only were Confederate districts being occupied, but
the very survival of the Confederacy was on the line. As was seen in the previous section, the
drift in support of a strong central government in the Second Congress is indicative of a “survival
7
I should add that this line of argument is especially convincing when looking the unionists and Whigs. The
average unionist and Whig scores were negative in the Second Congress.
8
The difference between “voting his conscience” and voting as a Burkean delegate are not mutually exclusive. In
the former case, conscience could imply either strengthening or weakening the central government, whereas the
Burkean delegate idea pertains exclusively to the latter.
13
response.” Taking both of these into consideration, I argue that, even with the mean WNOMINATE score shifting from the First to the Second Congress, the occupation of their
districts gave legislators a “shell shock.” Not only was national survival on the line, but for these
legislators their own districts were captured. Thus, they most certainly chose to strengthen the
central government, as occupation of their districts amplified the crisis of survival.
Further, it is also clear that the legislators were not “voting their consciences.” This was
a government that argued for its very independence on the issue of states’ rights—of the
submission of the federal government to the whims of the states. If legislators’ voting behavior
was moving toward their “true” ideologies, the drift in the Second Congress would have been in
the other direction. Table 2 demonstrated how former Democrats, members of the party of
states’ rights and central supporters of secession, were slightly against a strong central
government in the First Congress. However, during the Second Congress, they largely
abandoned this ideological plank in favor of a strong central government. This means that they
were certainly not voting the overt political philosophy of most southern elites. Thus, I argue
that the occupation status of a district is the most significant variable in explaining the roll call
behavior of a legislator.
I contend that occupation may affect behavior in two ways. First, it may be the case that
district occupation has permanent and lasting effect on legislator behavior. This would imply that
a legislator from an occupied district would always be more likely to support strengthening the
central government. Second, it may be that district occupation causes a short-term shock that
bumps the legislator’s ideal point to right. Note that these two possibilities are not mutually
exclusive: one could be “shocked” and continue to shift behavior after the initial event.
14
In order to test these two possibilities, I consider two regression models. In the first
model, I disaggregate the data by time period and regress the adjusted W-NOMINATE score on
party, secession stance, personal wealth, district wealth, and slaveholding. This attempts to
analyze differences between occupied and unoccupied districts over time. In the second model, I
analyze the impact of occupation’s shock on the difference in preferences from year-to-year.
Specifically, let Δ𝑊𝑁𝑂𝑀 = 𝑊𝑁𝑂𝑀𝑡 − 𝑊𝑁𝑂𝑀𝑡−1 and let OCC denote a change in district
occupation from 𝑡 − 1 to 𝑡. I regress Δ𝑊𝑁𝑂𝑀 on OCC and the same controls listed above. If
occupation’s shock is as I theorize, then the coefficient on OCC should be positive and
significant. Results from both sets of models are found below.
Results
General Regression Results
[Insert Table 3a about here]
The results are shown in Table 3a. The seven columns are seven separate models, each
run for a unique session of the Confederate Congress (Provisional Congress, four sessions of the
First House, and two sessions of the Second House). Most striking in this table is occupation
status’ negligible effect on W-NOMINATE scores, save for the last two time periods. In these
cases, the coefficients are negative and significant; this suggests that those from unoccupied
districts were actually more supportive of a strengthened central government in Richmond than
were those who were occupied. Somewhat predictably, the two economic variables –
slaveholding and district wealth – have large, statistically-significant effects. They are
increasingly positive towards the end of the war, suggesting that legislators from cotton-heavy
districts were willing to do whatever it takes to win the war. International relations scholars
frequently dub this “gambling for resurrection
15
[Insert Table 3b about here]
Though the results from Table 3a suggest that occupation may not have a period-byperiod effect on W-NOMINATE scores, the results in Table 3b tell a different story. Herein, I
regress the difference in the ideology scores from year-to-year on the difference in occupation
status and other relevant controls. Like the previous model, district wealth and slaveholding are
positive and statistically significant, indicating that legislators with the most to lose economically
were increasingly supportive of the centralized war effort. Most critical to my argument in this
paper, the coefficient on the shift in district occupation is positive and statistically significant.
This suggests that there is a discernible shift in legislator preferences, controlling for other
relevant variables. Thus, while there may not be a period-by-period difference across legislators,
there is indeed a shock present when we look within legislators.
While this demonstrates support for the occupation hypothesis, there is still further to
consider. Since the W-NOMINATE scores aggregate over a wide range of issues (not just central
state authority matters), one might argue that the conclusions that have been drawn are too broad.
Fortunately, there is one more matter that will help solidify the argument. During the course of
the Confederacy’s short existence, there were four issues that fundamentally challenged the
whole basis for secession (i.e., states’ rights): suspension of the writ of habeas corpus,
impressments, central control of state militias, and conscription. By analyzing each of these,
disaggregating by occupation status, the substantive significance of this variable will further
underscore its statistical significance.
[Insert Table 4 about here]
Specific Case I: Suspension of Habeas Corpus
16
The writ of habeas corpus, a Latin phrase meaning “ye should have the body,” is
considered one of the corner-stones of Anglo-Saxon law (Yearns 1960, 150; Martis 1994, 90). It
expressly denies the government the right to unlawfully detain persons in captivity. As is often
taught in American history courses, President Abraham Lincoln suspended the writ of habeas
corpus in the North during the war in the face of desertion, draft riots, and other miscellaneous
domestic problems (Yearns 1960, 150-152). Less well known is that the same measure was
invoked by President Jefferson Davis in the Confederacy. It is perhaps more interesting that the
Confederate government suspended the writ, in that it this action is expressly opposed to the
entire purpose for secession in the first place—states’ and individuals’ rights. A number of the
South’s outspoken statesmen made clear statements on these matters. One John Murray of
Tennessee remarked that he did not understand “that political doctrine that teaches that in order
to get liberty you must first lose it” (Alexander and Beringer 1972, 172-173). Another legislator,
Reuben Davis, thought that the suspension of this civil liberty would lead Davis to suspend
Congress (Yearns 1972, 152). Thus, the debate on the suspension of habeas corpus was intense
and it is clear that the suspension was at odds with the Confederate nation’s ideals.
Why would a legislator support such a thing? John Murray and Reuben Davis were not
alone in their opposition to this apparent usurpation of individuals’ rights vis-à-vis the central
government. According to my hypothesis, one would expect legislators from occupied districts
to be much more in favor of the suspension of the writ than their unoccupied counterparts.
Although the measure was unpopular, those from occupied districts could support this
“necessary” ceding of power to the executive without fears of electoral retribution. Further,
legislators may have supported suspending the writ because the occupation of their districts
made vivid the implications of an un-centralized government. As one sees in Table 4, 82% of
17
legislators from occupied districts supported the suspension of the writ. Contrast this with a
mere 22% support from legislators in unoccupied districts. Given the intense and divisive nature
of this issue, the evidence provided in Table 4 (in light of previous findings) most certainly
affirms my hypothesis.
[Insert Table 5 about here]
Specific Case II: Central Control of State Militias
It is widely known that military units, both North and South, were organized at the state
level during the Civil War. In the South especially, the notion of the state militia was quite
important—a as it was the states that seceded on their own, not as a unified bloc.9 Thus,
centralizing the state militias was a very divisive matter. To be sure, Wilfred Buck Yearns
observed that, “The adoption of conscription made…important changes in the method of raising
an army: to a large extent it dispensed with the instrumentality of the states…” (1960, 66).
Fierce debate surrounded the issue, culminating with a vote on March 16, 1865. Though late in
the war, the results of this roll call are telling (Table 6). As predicted by my hypothesis, the vast
majority (73%) of legislators from occupied districts supported the relinquishing of state militia
power to the central government. Only 17% of legislators from unoccupied districts supported
the measure.
[Insert Table 6 about here]
Specific Case III: Impressment
Impressment was yet another hotly contested issue in the Confederate Congresses. The
basic question was: should the government be able to seize private property? While the general
intention on the part of the Confederate government was to use impressed property to support the
9
This is a crucial point. The states seceded independently. Their congregation in Montgomery, Alabama in 1861
was to create a confederation and not a federal republic like the nation they seceded from.
18
army and the war effort, it was still a case of the central government usurping the rights of
individuals (Martis 1994, 90). Further, when property was impressed, mounting military defeats
during the Second Congress had a depressing effect on Confederate currency, and this led to
property owners receiving an unjust compensation for their stolen land (Alexander and Beringer
1972, 154). Due to contentiousness of this issue, we should once again expect a sharp divide
between occupied and unoccupied districts. In Table 5, one sees that a majority of legislators
from both occupied and unoccupied districts supported impressment in the First Congress.
However, by the time of the Second Congress, those from occupied districts more often
supported impressments and those from unoccupied districts significantly decreased in their level
of support. Once again, my hypothesis is confirmed with respect to this issue; legislators from
occupied districts were much more willing to cede authority to the government in Richmond.
Specific Case IV: Conscription
The last specific case is conscription. Intimately tied to central control of state militias,
conscription was a divisive matter. Yearns (1960) comments that, “[f]ew people were neutral
about conscription. The patriotic objected to its implications, while would-be civilians resented
compulsory army service” (66). In the creation of a confederacy, the framers of the Confederate
Constitution had in mind a free and unbinding association of independent states. In this sense,
there was nothing compulsory about joining the Confederacy. Imposing these sorts of central
conscriptions is a coercive act on the part of the government, most certainly outside the scope of
what was intended by the intellectual fathers of the South.
Table 6 shows the average support for conscription by Congress, where average support
is calculated as the percentage of the time a legislator supported increasing conscription. The
levels of support are essentially static from Congress to Congress. While this matter was less
19
decisive than the suspension of habeas corpus, impressment, or central control of state militias,
one still notices a roughly twenty percentage-point gap between members from occupied and
unoccupied districts. Further, if one bears in mind that these percentages are based on average
levels of support, the substantive impact is broader. For example, as one sees in Table 6, roughly
76% of legislators from occupied districts expressed high levels of support for conscription.
This means that over three-fourths of legislators in this subgroup consistently voted to support
conscription. Moreover, this result is in some sense more telling than certain single-vote results
presented above, for it is a consistently large percentage.
Conclusion
In this paper, I argue that the national crisis brought about by the Civil War, exacerbated
by the Federal occupation of Congressional districts in the Confederacy, led Southern legislators
to alter their voting behavior. To demonstrate this, I introduce empirical evidence for the shifting
of ideal points of legislators. More importantly, this shift is shown to have a close relationship
with the occupation of the legislator’s district. Though not a primary goal of the paper, I
necessarily address two methodological issues. First, I demonstrate that treating static ideal
points as cross-temporally comparable can lead to faulty inferences in subsequent analyses.
Second, and related to the first, I have brought to light some of the challenges presented by
estimating preferences of decision-makers over time.
All of these objectives, either primary or secondary, are certainly interrelated, at least to
the degree that they influence one another. The use of statistical methodology to evaluate
historical events and institutions is surely a positive development. Indeed, modern
methodological tools afford scholars new and innovative ways to investigate old claims and
hypotheses. At the same time, however, it is imperative that the methods employed be
20
appropriate to the historical application. In the case of preferences, using dynamic ideal points
allows scholars to evaluate historically-based claims surrounding decision-makers’ preferences
in a coherent and correct fashion. Nonetheless, the methods themselves are not infallible---all of
them rely on a set of more-or-less restrictive assumptions. Depending on the application, some
assumptions are more reasonable than others. Thus, it befalls on researchers to evaluate which
approach’s assumptions are appropriate to the situation at hand.
For the case of the Confederate Congress discussed at length in this paper, the methods
employed have provided solid evidence that legislators’ preferences shifted as a result of their
Congressional district’s occupation. Even when controlling for all core factors that might
otherwise have affected preferences (e.g., former party, secession stance), district occupation still
has a substantial impact on ideal point shifts. Though one cannot have a complete sense of the
psychological factors that affected legislators’ decisions to change, it seems clear that the reality
of military loss and subsequent political change caused enough of a stir in legislators’ minds to
induce ideal point shifting. However, one must not ignore the intense normative implications of
these legislators’ choices. The calls for secession among Southern intellectuals were almost
wholly theoretical. They based the legitimacy of a Southern Confederacy on the principles of
limited, accountable government and individual liberty. To the contrary, the evidence presented
in this paper has shown quite clearly that the occupation of legislators’ districts caused them to
vote in way contrary to these principles. Consequently, a new question begs to be answered: in
the aftermath of these anti-states’ rights votes, what was the new basis for an independent
Southern nation? If the South did in fact win its independence, would legislators go back to their
“old selves”? How would they answer to their newly un-occupied constituencies?
21
Future research should explore these issues further, as similar historical circumstances
have reared their heads in the nearly 150 years since the Civil War. By exploring legislators’
motives, political scientists will become one step closer to explaining the relationship between
historical circumstances and individual behavior. In turn, this will shed light on our
understanding of legislative institutions, decision-making, and, more broadly, human behavior
when under fire.
22
References
Alexander, Thomas and Richard Beringer. 1972. The Anatomy of the Confederate
Congresses. Nashville: Vanderbilt University Press.
Bensel, Richard. 1986. “Southern Leviathan: The Development of Central State
Authority in the Confederate States of America.” In Studies in American Political
Development, vol. 2. New Haven: Yale University Press.
Beringer, Richard. 1967. “A Profile of Members of the Confederate Congress.” The
Journal of Southern History: 518-541.
Burke, Edmund. [1774] 1999. “Speech to the electors of Bristol.” In Select Works of Edmund
Burke. A New Imprint of the Payne Edition. Indianapolis: Liberty Fund.
Groseclose, Timothy, Levitt, Steven, and James M. Snyder, Jr. 1999. "Comparing Interest Group
Scores across Time and Chambers: Adjusted Ada Scores for the U S
Congress." American Political Science Review, 93(1), 33-50.
Jenkins, Jeffery. 1999. “Examining the Bonding Effects of Party: A Comparative
Analysis of Roll-Call Voting in the U.S. and Confederate Houses.” American Journal of
Political Science 43: 1144-65.
Jenkins, Jeffery. 2000. “Examining the Robustness of Ideological Voting: Evidence from
the Confederate House of Representatives.” American Journal of Political Science 44:
811-22.
Martin, Andrew and Kevin Quinn. 2002. “Dynamic Ideal Point Estimation via Markov Chain
Monte Carlo for the US Supreme Court, 1953-1999. Political Analysis 10: 134-153.
Martis, Kenneth. 1994. The Historical Atlas of the Congresses of the Confederate States
of America: 1862-1865. New York: Simon & Schuster.
23
Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale
University Press.
Poole, Keith and Howard Rosenthal. 1997. Congress: A Political-Economic History of
Roll Call Voting. New York: Oxford University Press.
Rohde, David, Jamie Carson, Jeffery Jenkins, and Mark Souva. 2001. “The Impact of
National Tides and District-Level Effects on Electoral Outcomes: The U.S.
Congressional Elections of 1862-63.” American Journal of Political Science
45:887-98.
Rable, George. 1994. The Confederate Republic: A Revolution against Politics. Chapel
Hill: University of North Carolina Press.
Thomas, Emory M. 1979. The Confederate Nation: 1861-1865. New York: Harper & Row.
Wooster, Ralph. 1962. The Secession Conventions of the South. Princeton: Princeton
University Press.
Yearns, Wilfred Buck. 1960. The Confederate Congress. Athens, Ga.: University of
Georgia Press.
24
Figure 1: W-NOMINATE Distribution, First Congress
Figure 2: W-NOMINATE Distribution, Second Congress
25
Figure 3: W-NOMINATE Distribution for Members Who Served in Both Congresses
(a)
(b)
26
Table 1: Descriptive Information for the Congresses of the Confederate States
Variables
First Congress
Second Congress
Percent formerly
Democratic
58%
52%
Percent pro-secession
65%
55%
Percent of districts occupied
at least half of the session of
Congress
36%
48%
Average personal wealth
(x10,000)*
8.2074
6.8176
Average per acre district
land value*
11.28
11.43
Slaveholding index
4.50
4.35
* In 1860 U.S. Dollars
Table 2: Average W-NOMINATE Score by Former Party and Secession Stance
Disaggregated Party and
Secession Stance
Former Democrats
Former Whigs
Secessionists
Unionists
First Congress
Second Congress
-.232
.260
-.195
-.116
.187
.336
.226
.305
27
Table 3a: Multivariate model predicting W-NOMINATE scores+
(1)
Model
(2)
Model
(3)
Model
(4)
Model
(5)
Model
(6)
Model
(7)
Model
0.000943
(0.00175)
0.0448
(0.0413)
0.000810
(0.000948)
0.0495
(0.141)
-0.00308
(0.0488)
-0.353***
(0.0503)
-0.347***
(0.0494)
Party
0.00430
(0.00272)
0.104
(0.0630)
0.00222
(0.00137)
0.252
(0.160)
-0.0760
(0.0483)
-0.105
(0.0695)
-0.103
(0.0683)
Secession
-0.00166
(0.00186)
-0.0346
(0.0438)
-0.000832
(0.000960)
-0.102
(0.111)
0.0316
(0.0333)
0.0425
(0.0483)
0.0418
(0.0474)
Party*Secession
-0.00356
(0.00369)
-0.0869
(0.0861)
-0.00182
(0.00188)
-0.203
(0.217)
0.0629
(0.0658)
0.0661
(0.0913)
0.0649
(0.0896)
Personal Wealth
1.15e-08
(8.45e-09)
2.62e-07
(1.98e-07)
5.84e-09
(4.41e-09)
6.86e-07
(5.20e-07)
-2.04e-07
(1.57e-07)
-3.61e-07
(2.23e-07)
-3.54e-07
(2.19e-07)
Slaveholding
-0.00739**
(0.00370)
0.00323***
(0.00107)
-0.167**
(0.0830)
-0.00383**
(0.00186)
-0.456**
(0.222)
0.141**
(0.0686)
0.248**
(0.0967)
0.244**
(0.0950)
-0.0762***
(0.0242)
-0.00170***
(0.000540)
-0.194***
(0.0638)
0.0586***
(0.0196)
0.0958***
(0.0278)
0.0941***
(0.0273)
0.0636***
(0.00485)
0.337***
(0.111)
-0.0747***
(0.00244)
0.907***
(0.297)
-0.303***
(0.0917)
0.183**
(0.0805)
0.286***
(0.0790)
109
0.18
0.13
109
0.19
0.13
109
0.19
0.13
109
0.18
0.12
109
0.18
0.12
109
0.21
0.16
109
0.21
0.16
Variables
Occupation
Status
District Wealth
Constant
Observations
R-squared
Adj. R-squared
Note: Cell values are Ordinary Least Squares regression coefficients. Robust standard errors are in parentheses.
Dependent Variable:
Adjusted W-NOMINATE first dimension score.
Independent Variables:
Occupation Status = 1 if occupied for greater than half of the session’s duration; 0 otherwise
District Wealth = District per acre land value in 1860 U.S. Dollars
Personal Wealth = Legislator estate value in 1860 U.S. Dollars
Slaveholding = District slaveholding index; 0 = “low”, 9 = “high”
Party = 1 if former Democrat, 0 if former Whig
Secession Stance = 1 if secessionist, 0 if unionist
*p < .05, **, p < .01
28
Table 3b: Multivariate model predicting shift in W-NOMINATE scores+
Variables
Occupation Shift
0.116*
(0.0726)
Party
-0.00741
(0.0124)
Secession
0.0285***
(0.00979)
Party*Secession
-0.00769
(0.0202)
Personal Wealth
-1.57e-07***
(4.64e-08)
Slaveholding
0.0372*
(0.0200)
District Wealth
0.0189***
(0.00627)
Constant
-0.0945***
(0.0263)
Observations
R-squared
Adj. R-squared
762
0.01
0.00
Note: Cell values are Ordinary Least Squares regression coefficients. Robust standard errors (clustered by
individual) are in parentheses.
Dependent Variable:
Difference in adjusted W-NOMINATE first dimension score from last time period.
Independent Variables:
Occupation Shift = 1 if change in occupation status occurred since last period; 0 otherwise
District Wealth = District per acre land value in 1860 U.S. Dollars
Personal Wealth = Legislator estate value in 1860 U.S. Dollars
Slaveholding = District slaveholding index; 0 = “low”, 9 = “high”
Party = 1 if former Democrat, 0 if former Whig
Secession Stance = 1 if secessionist, 0 if unionist
*p < .05, **, p < .01
29
Table 4: Support for Suspension of the Writ of Habeas Corpus (December 8, 1864)
Occupation Status
% Voting Yea
N
Occupied
82
37
Unoccupied
22
13
Note: Cell values represent the percent of legislators who voted in support of suspending the writ of habeas corpus.
Source: Martis 1994, 107.
Table 5: Support for Central Control of State Militias (March 16, 1865)
Occupation Status
% Voting Nay
N
Occupied
73
36
Unoccupied
17
9
Note: Cell values represent the percent of legislators who voted nay, that is, in support of allowing the central
government to control state militias. These values are disaggregated by district occupation status. The N is the
number of legislators within that group (occupied or unoccupied). Source: Martis 1994, 108.
Table 6: Average support for Impressment and Conscription by Congress
Issue
Occupation
Status
% High Support
(First Congress)
N
% High Support
(Second Congress)
N
Impressment
Occupied
Unoccupied
66
54
25
37
78
30
39
17
Conscription
Occupied
Unoccupied
76
53
29
37
73
55
36
31
Note: Cell values represent the percent of legislators who voted yea, that is, in support of impressments or
conscription. The N is the number of legislators within that group (occupied or unoccupied). Since there were a
number of roll calls on the matter, the percents are average level of support (see Martis 1994, 110-111, 113).
30