Are Marginals Different?*
Robert Hodgson and John Maloney
Abstract – We analyse the results of British general elections from 1950 to 2005, and
show that the national trend in an election is followed most strongly by seats which are
neither too safe nor too marginal. This is consistent with a voting model where
ideological factors are relatively dominant in instrumental voting, and valence factors are
relatively dominant in expressive voting. We also find that governments perform better
in their own marginals and opposition safe seats than in opposition marginals and their
own safe seats – and that it is this, not geographical polarization of party support, that
explains the declining number of marginals.
1.Introduction
Over the last sixty years there has been a trend – although not a steady or uninterrupted
one – for constituencies in a general election to show less uniform swings. How far these
are a matter of ‘noise’ and how far underpinned by shifting regional and socio-economic
electoral patterns has been intensively investigated. In this article we are going to look
specifically at differences in the behaviour of marginal, fairly safe and very safe seats.
The two uncontested propositions about marginal seats in Britain are that there are fewer
than there were (at least if you count them the in terms of two-party share of the vote) and
that turnout is higher in marginal seats than in safe ones. But does living in a marginal
constituency affect the way people vote as opposed to the likelihood that they vote at all?
Different types of seat might show different degrees of absolute volatility, or stronger and
weaker adherence to the overall national trend in a general election, or a different
propensity to swing away from and back to the government of the day. In this article we
put together the evidence of every general election between 1950 and 2005 to try to
answer these questions. We also ask whether being marginal or becoming marginal is the
more likely to produce distinctive behaviour, and whether it is right to focus in the first
place on seats with current small majorities, as opposed to those which would be close in
the event of a close national result – the so-called bellwether seats.
2. Literature
If people vote only when expected benefit (utility from deciding the government x
probability that your vote will do so) exceeds the cost of voting, no rational person will
vote. It is always open to political scientists to attribute bad statistical sense to voters, so
that they exaggerate their chances of personally ejecting or saving their MP, or even the
government itself. But ‘minimum rational choice’ and ‘general incentives’ have been the
more popular options over the years. According to ‘minimum rational choice theory’1
voters also define their political efficacy ‘in terms of the influence of groups of people
like themselves … Consequently they are motivated to take action because they feel they
can collectively make a difference.’2 If these groups include your political allies within
your own constituency, and collective efficacy is related to the likelihood that collective
action will be pivotal, we have a rationale for higher turnout in marginal seats. The
general incentives model3 encompasses the minimum rational choice approach but adds
other motives for voting, in particular the expressive motive first put forward by James
Buchanan4. And, in doing so, it provides not just a reason for higher turnout in marginals
but a rationale for a different pattern of voting. If voters have a mixture of instrumental
and expressive motives, the relative power of the latter will decline as the efficacy of the
vote increases5. Consequently, voters who would like to express a protest against the
government, but still prefer it to the main opposition party, might give it a grudging vote
in a marginal seat which they would withhold in a safe one.
2
This, then, is our first reason to expect marginal and safe seats to have different
patterns of swing. Price and Sanders extend the analysis by attributing two alternative
objectives to disenchanted government supporters
(a) punish the government as long as its majority is not reduced
(b) punish the government as long as it doesn’t lose office6
Either of these factors would reduce anti-government swings in marginal seats. But in
that case they must reduce any pro-government swings in marginal seats too. When the
trend is towards the government (i.e. it increases its majority), some voters in safe seats,
previously protesting against the government, will return to it. So will some in the
marginals, but not so many, because fewer deserted it in the first place. Whatever the
direction of the national swing, then, its beneficiary will gain fewer votes in marginal
seats than in safe ones.
The second reason why marginals are different is that parties campaign harder and
spend more in marginal seats. Again the most obvious – and well-documented – effect is
on turnout. Turnout in marginal constituencies has exceeded average national turnout in
every British election since 1950, but the size of this margin has fluctuated substantially.
Denver, Hands and McAllister trace the widening of the gap up to 1979 and its
subsequent narrowing during the Thatcher / Major years, only for it to open up again in
1997 and 2001.7 Controlling for socio-economic status (a significant determinant of
turnout) alters the story so far as the downturn after 1979 is now backdated to 1966. The
question raised is whether higher turnout in marginals comes from the more intensive
campaigning they undoubtedly endure, or whether the expected closeness of the result
would have called out more voters anyway. Denver, Hands and McAllister conclude
3
that both these factors matter, though their relative strength appears to vary significantly
between one election and another.
In addition, Curtice and Steed have found that the turnout effect of being a marginal
seat rises in elections that follow hard on their predecessors (1951, 1966, October 1974)
and then drops back at the subsequent election (1955, 1970, 1979).8 Presumably this is
because recent memories of the previous election include a memory of the constituency’s
marginal status. If so, this is itself evidence that higher turnout in marginals is not solely
due to higher spending and harder campaigning there. So, indirectly, is the finding by
Clarke et al that turnout is higher across the country when the national result is close,
given the lack of any inverse correlation between national winners’ margins and
combined campaign spending. 9
To assert that more people would vote in marginal constituencies even if no party
spent a penny on them is not, of course, to say that campaigning is ineffective.
Indeed recent evidence suggests not only that it is more effective than was previously
thought, but that the two main parties, even if they spend equal sums of money
nationally, do not just neutralize one another’s effects. Some of this is because parties
tend to spend more in their own constituencies.10 All of this means that we would expect
to see each party performing better in the marginals which it holds than in those it is
trying to win over, and Clarke et al. find exactly this in the 2001 election.11 In the present
article we find the pattern survives when we take the elections since 1950 in aggregate.
Thirdly, a generally positive personal vote for incumbent MP’s will give parties
particularly favourable swings in their own marginal seats taken as a whole. Unless a
party captured no seats at all from its opponents at the previous election, its marginals
will include a disproportionate number of candidates standing as the incumbent MP for
the first time, and thus enjoying an ‘incumbency swing’ less prevalent in safe seats.
4
Several of the election studies by Steed and/or Curtice notice an incumbent government
holding rather successfully onto its marginals when it is losing ground in general. A
particularly strong and persistent effect in their studies is how well a party generally does
when its incumbent MP fights a seat he or she gained from an incumbent opponent at the
previous election. This ‘double incumbency’ effect seems to have been particularly
striking in the elections of 2005, 1992, 1987 and 1970.12 In the 1970 case Steed finds
that the double incumbency effect entirely explains why the outgoing Labour government
suffered a lower hostile swing in its 100 most marginal seats than across the country as a
whole. (In 1979, by contrast, when a Labour government was again defeated but again
did less badly in its own marginals, Curtice and Steed find the double incumbency effect
explaining only a part of the phenomenon.) 13
Curtice and Steed are also among those who find a sudden drop in the number of
marginals after 1979, noting that this is reflected in an increase in the standard deviation
of Conservative/Labour shares across the constituencies, partly caused by a negative
kurtosis flattening the peak of the distribution away from normality.14 Norris and Crewe
agreed with the repeated demonstrations by Curtice and Steed that there were now fewer
two-party marginals -- seats where the Conservatives polled between 45 and 55% of the
combined Conservative / Labour vote – but argued that, if the point of identifying a
marginal seat was to pick out those seats most liable to change hands, then a better
criterion would be the number of seats where the margin between the winner and the
runner-up (of whatever party) was less than 10%, or alternatively 20%, of all the votes
cast. 15 Table 1 (columns 2 and 3), however, shows that, even on this classification, and
extending the data to 2005, there is a clear long-run downward trend to the number of
marginals, though at nothing like the rate given by the Curtice/Steed criterion (column
1).16
5
TABLE 1
1950
1951
1955
1959
1964
1966
1970
1974 (Feb)
1974 (Oct)
1979
1983
1987
1992
1997
2001
2005
(1)
27.9
29.1
27.9
28.0
27.0
28.6
25.3
19.3
18.3
17.2
15.2
14.1
17.5
13.2
10.0
15.8
(2)
31.3
30.6
28.5
30.0
31.2
31.2
27.9
27.0
28.6
22.6
23.0
21.3
24.9
22.1
17.4
24.7
(3)
56.4
55.5
53.4
52.9
58.8
57.2
56.6
65.1
59.5
48.0
47.9
46.1
46.4
45.4
45.2
52.9
(4)
15.1
15.2
15.0
14.4
14.3
16.0
13.8
12.1
13.9
13.5
11.1
13.8
13.9
17.3
14.4
12.4
(5)
1.55
0.39
-0.01
-0.09
0.87
0.59
0.13
1.21
0.88
-0.14
-0.53
-0.06
-0.09
-0.57
-0.33
-0.03
Column 1: seats where Conservatives polled 45-55% of combined Conservative /Labour
vote, as percentage of all seats
Column 2: Seats where Conservative or Labour won with a majority of less than 10% of
the total vote, as a percentage of all seats
Column 3: Seats where Conservative or Labour won with a majority of less than 20% of
the total vote, as a percentage of all seats
Column 4: standard deviation of absolute Conservative or Labour percentage majority
from zero
Column 5: kurtosis
The main reason put forward over the years for the decline in the number of
marginals has been increasing geographical polarization. If Southern seats become more
intensively Conservative while Northern ones swing still further towards Labour, this will
turn Conservative southern and Labour northern marginals into safer seats. It will have
the opposite effect on safe Conservative seats in the north and safe Labour seats in the
south, but, because there were relatively few of these, not all the vanishing marginals will
be replaced. That polarization between North and South has occurred can be seen from
the first column in table 2. Regional polarisation here is measured as follows:
[% of seats in South that are more Conservative than national average minus % of seats in
6
South that are more Labour than national, average] + [% of seats in North that are more
Labour than national average minus % of seats in North that are more Conservative than
national average]
where South includes the South-Eastern and South-Western regions of England as
currently defined by the British government, plus any other county which adjoins
London. North comprises Scotland, plus the North-East and North-West regions, and
Yorkshire and Humberside, as defined by the government.
TABLE 2
Year
Regional Polarisation
Lab/Con 1st
1950
1951
1955
1959
1964
1966
1970
1974 (Feb.)
1974 (Oct.)
1979
1983
1987
1992
1997
2001
2005
Consistency index
8.04
8.20
8.88
12.97
14.37
15.00
15.92
16.04
17.83
26.09
31.14
40.52
36.02
30.56
29.20
31.00
n/a
n/a
n/a
n/a
76
83
78
50
51
42
45
43
46
36
37
n/a
The high point of polarization was reached in the general election of 1987, when Labour
won only three seats south of a line from the Wash to the Bristol Channel (London
excepted) and the Conservatives, despite a twelve-point national lead in the vote, only
won ten out of a possible seventy-two seats north of Hadrian's wall. Since then
polarization has been partially reversed, but even so the index for 2005 was nearly twice
anything seen before 1979.
7
However as geographical political identities have become more polarized, class ones
have become less so. A simple index of the decoupling of class and voting is the
consistency index shown in column 2 of table 2 17:
[% of middle-class voters who vote Conservative minus % of working class voters who
vote Conservative] plus [% of working-class voters who vote Labour minus % of middleclass voters who vote Labour]
Clearly the maximum score – absolute correlation between class and party – is 200. The
index has always taken on a value well below half of this, but nonetheless declined from
76 in 1964 to 37 in 2001.
Because the two indices are comparable (they use the same method to measure,
respectively, the correlation between being Southern and Conservative and the
correlation between being middle-class and Conservative) we can take the much larger
absolute change in the consistency index than in the regional one to mean that class has
weakened as a voting indicator more rapidly than region has strengthened. If regional
polarization had been the only effect at work, we would indeed have an explanation for
the falling number of marginals. The trouble is that all the while there has been an even
stronger opposing effect – the continuing detachment of voting from social class. And
even if social de-polarisation had been no stronger than regional polarization, the net
effect of the two forces would still have been to increase the number of marginals.18 To
explain why marginals have become fewer, we need a different approach altogether, and
we pursue this below in section 7.
What all the above implies is that there are good reasons to think that nonuniformity of constituency swings is likely to be systematic as well as random, and that
8
there is unlikely to be a simple relation whereby marginal seats feature lower or higher
swings than safe ones. And certainly the successive analyses of election swings by
Michael Steed and John Curtice do not add up to any such pattern. Aggregating their
surveys, however, is not easy. The definition of a marginal changes between surveys; in
1987, for example, they looked only at government marginals; in elections which follow
boundary changes they do not try to identify marginals at all. This is not a criticism.
Different approaches are appropriate to different electoral situations. But it does mean
that an aggregative study of very safe, moderately safe, marginal and bellwether seats
reaching down from 1950 has yet to be conducted. That is what we are going to do.
3. Model
FIGURE 1
Suppose voters live along a left-right axis, and that there are two parties, L and R, located
at ! L and ! R on this axis.
Voters in constituency x are normally distributed with variance
9
! 2 (common to all constituencies) around a mean of x . Let the ideological position of
voters in seat x who are indifferent between party L and party R be ! (x) .
The number of voters at ! (x) will thus be q(! (x)) =
$( x $ ! ( x ))2
1
2"#
2
e
2# 2
(1)
Now suppose that expressive voting (voting to demonstrate support for a party) and
instrumental voting (voting to try and affect the outcome of the election) both depend on
(i) the parties’ respective valences (ii) the voter’s ideological distance from the parties.
Let voter i derive the following expressive utilities from voting for L and R respectively:
EXP
U i,EXP
L = ! vL " # | $ i " $ L | U i, R = ! vR " # | $ i " $ R | where v is valence and ! is ideological location.
Now, over and above any expressive utility, let voter i’s welfare gain from having L’s
policies, not R’s, implemented be U i,INST
! U i,INST
= " (vL ! vR ) ! # (| $ i ! $ L | ! | $ i ! $ R |) L
R
Hence i’s instrumental utility from voting for L is:
p(x)(U i,INST
! U i,INST
L
R ) = p(x)[" (vL ! vR ) ! # (| $ i ! $ L | ! | $ i ! $ R |)] where p(x) is their perceived probability that their vote will be decisive if they live in
constituency x.
Hence, putting expressive and instrumental utilities together, i’s total excess utility
from L rather than R is:
U i.L ! U i, R = " (vL ! vR ) ! # (| $ i ! $ L | ! | $ i ! $ R |) + p(x)[% (vL ! vR ) ! & (| $ i ! $ L | ! | $ i ! $ R |)]
10
We now assume in addition that
vL = !vR , and that the valences are sufficiently small to
ensure that voters in seat x who are indifferent between L and R (i.e. voters at ! (x) )
come between L and R ideologically. Making these assumptions, and writing vL as v
from now on, means that the utility function of voters at ! (x) , if living in constituency x,
is:
U! ( x ), L " U! ( x ), R (= 0) = 2# v " $ (2! (x) " ! L " ! R ) + p(x).(2% v " & (2! (x) " ! L " ! R )) (2)
i.e. ! (x) =
" + # p(x)
! + !R
v+ L
(3)
$ + % p(x)
2
so that
d! (x) " + # p(x)
=
dv
$ + % p(x)
(4)
This is the shift in the ideological boundary between L and R voters, in constituency x, in
response to a unit change in v.
The swing between the parties in constituency x will thus be:
number of voters at ! (x) times
shift in ! (x) Graphically:
11
FIGURE 2
Or algebraically, using equations (1) and (4)
d! (x)
s = q(! (x)).
=
dv
$( x $ ! ( x ))2
1
2"# 2
2# 2
e
% + & p(x)
' + ( p(x)
(5)
where, as can be seen from Fig.2, x ! " (x) measures the safety of the seat. (It is not the
actual majority, but is a monotonically increasing function of the majority.) We
henceforth write this term as n (shorthand for n(x)) and further assume that voters’
estimate of their chances of being pivotal when they live in seat x (p(x)) is inversely
related to the absolute majority in x and hence inversely related to n. We can thus
rewrite (5) as
d! (x)
s = q(! (x)).
=
dv
$ n2
1
2"#
2
e 2#
2
% + & p(n)
where p’(n)<0.
' + ( p(n)
(6)
Before continuing, note that if voting were either purely expressive or purely
instrumental, marginal seats would unambiguously have the highest swings. Purely
instrumental voting is the case where ! and ! are both zero: (6) becomes:
12
d! (x)
s = q(! (x)).
=
dv
$ n2
2 ' % *
e 2 # ) , where ds/dn<0
(&+
2"# 2
1
If voting is purely expressive, ! = " = 0 , so that now (6) becomes:
d! (x)
s = q(! (x)).
=
dv
$ n2
1
2"# 2
e
2# 2
'%*
)( & ,+ Again ds/dn<0.
Returning to the case of composite voting motives, we now differentiate (6) with respect
to n:
ds / dn =
1
2!" 2
e
#
n2
2" 2
( #n($ + % p(n)) (&% # $' )p '(n) +
* " 2 (& + ' p(n)) + (& + ' p(n))2 - (7)
)
,
so that (ds / dn = 0) ! n =
" 2 (#$ % &' )p '(n)
(& + $ p(n))(# + ' p(n))
(8)
Consider equation (7) in the case where !" > #$ . Since n represents absolute distance
between x and ! (x) and cannot be negative, while p '(n) must be negative
(or voters
would
rank seats with small majorities as safer than those with large majorities), it
follows that ds/dn is negative throughout: the safer the seat, the smaller the swing.
The logic goes as follows. The component terms of !" > #$ are:
! =
responsiveness of expressive voting to valence
! =
responsiveness of instrumental voting to valence
! =
responsiveness of expressive voting to
ideological distance between voter and party
! = responsiveness of instrumental voting to
ideological distance between voter and
party.
13
But if
!" > #$ ,
then
! / " > # / $ . This means that (compared with their respective
sensitivities to valence) expressive voting is relatively sensitive to, and instrumental
voting relatively insensitive to, ideological differences between the voter and the party.
But in a safe seat, the expressive motive is relatively strong and the instrumental one
relatively weak. It follows that voting in these seats is ideology- rather than valencedriven, so that it requires a big change in valence to get a given shift in ! (x) ,
the
ideological position of the swing voter. In this case, then, a given change in valence will
shift ! (x) by less in safe seats than in marginal ones, and this reinforces their existing
tendency towards smaller swings, given by the fact that there are fewer voters at ! (x) in
the first place. The situation is as in Fig.3:
FIGURE 3
Now take the case where !" > #$ . We can see from (8) that a positive value of n exists
such that ds/dn=0. But is s (swing) maximized or minimized at this point? When the
first-order condition holds, the term in the square brackets in equation (7) is zero, so that
d s / dn =
2
2
1
2!" 2
e
#
n2
2" 2
.
d ( #n($ + % p(n)) (&% # $' )p '(n) +
+
dn *) " 2 (& + ' p(n))
(& + ' p(n))2 -,
14
=
1
2!" 2
e
#
n2
2" 2
( #1 $ + % p(n) n(&% # $' )p '(n) (&% # $' )p ''(n) 2(&% # $' )' [ p '(n)]2 +
* " 2 . & + ' p(n) # " 2 (& + ' p(n))2 + (& + ' p(n))2 #
(& + ' p(n))3
)
,
Assuming p ''(n) >0 (or some seats would have a negative probability of changing
hands), the first three terms in the bracket are negative but the final one is positive. Thus
d 2 s / dn 2 could be either positive or negative. It could even change sign depending on the
value of n. If it is consistently negative (positive), then the swing will have a maximum
(minimum) point as n increases. The function s(n) is hump-shaped (U-shaped):
intermediate seats will feature higher (lower) swings than marginal or safe ones.
Why? This time the intuition is as follows. When !" > #$ (i.e.
! / " > # / $ ) it is
instrumental voting that is relatively sensitive to, and expressive voting relatively
insensitive to, ideological differences between the voter and the party. It is thus in the
marginal seats (where instrumental voting is most important) where the vote is most
strongly driven by ideology, and where it thus requires the largest change in valence to
get a given shift in the position of the swing voter ( ! (x) ) along the ideology axis. In this
case, then, a given change in valence shifts ! (x) further in safe seats than in marginal
ones, and this counteracts their tendency towards smaller swings, given by the fact that
there are fewer voters at ! (x) in the first place. With these two effects now working in
opposite directions, the situation is in Fig.4 and there is no longer a monotonic relation
between existing majority and swing.
15
FIGURE 4
All this casts a new light on the proposition that a government falling out of favour
may lose more support in safe seats (where voters feel free to express their feelings of
protest) than in marginal ones (where some voters would like to protest but do not want
the government actually ejected from office.) The implicit assumption behind this story
is that there are voters who are expressively anti-government but instrumentally progovernment. This could be the case – indeed it corresponds to the case where
! / " > # / $ in our model (as the government’s valence declines, expressive sentiments
change faster than instrumental calculations.) But it doesn’t have to be the case – it is
equally possible that the declining valence has more impact on the instrumental motive
than on the expressive one. In that case relatively few voters will be deserting the
government as a protest, and many more doing so as a strategy. As the marginal seats are
the ones where strategy is most preferred to protest, the government will do worst in the
seats where it can least afford to lose ground.
16
But so far we have been assuming that national swings are caused by changes in the
valences of the parties. What if voters are switching because one of the parties has shifted
its ideological position? Equation (3) was:
! (x) =
" + # p(x)
! + !R
v+ L
$ + % p(x)
2
so that
d! (x)
= 0.5 , and equation (6) becomes
d! L, R
$n (
% 1
d! (x)
2
s = q(! (x)).
= 0.5 '
e 2 # * which is declining in n. Swing is once again
2
d! L, R
& 2"#
)
2
inverse to constituency majority. ( ! (x) has shifted the same distance in all seats, so the
ranking of swings is simply down to how many voters were at ! (x) in each seat.) If it is
the ideology of the voters that has shifted, that is equivalent to the ideology of both the
parties shifting in the other direction. The swing will double, but will remain lowest in
the safest seats.
The picture is getting monotonous. If voting is purely expressive or purely
instrumental, the highest swings will be in the most marginal seats. If voting is partexpressive and part-instrumental, with ideology more dominant in the expressive
component than in the instrumental one, then a fortiori the highest swings are in the most
marginal seats. If parties are gaining or losing ground because their own ideology, or that
of the electorate, has shifted, again the swing will be higher the more marginal the seat.
The only possibility that stands out against the consensus is the case where swings are
driven by valence changes, voting is part-expressive and part-instrumental, and valence is
more dominant in the expressive component of the vote than in the instrumental one.
Then and only then will the ideological position of the swing voter shift furthest in safe
seats, compensating for the fact that they have fewer swing voters to start with, and
17
possibly producing a hump-shaped or U-shaped relation between existing majority and
swing. Should we find either relationship, we will have narrowed down decisively the
possibilities of what is going on.
4. Method and Data
To analyse the impact of a seat’s marginality on swing we make use of all constituency
results (Northern Ireland excepted) where either Labour or the Conservatives won,
between 1950 and 2005. However the elections of February 1974, 1983 and 1997, all of
which followed major boundary changes, are used only to calculate swings at the
succeeding election (and to establish where it was being contested by incumbent MP’s).
Our basic regression model is run using OLS and is as follows:
Swingjt = !1M jt + ! 2 X jt + ! 3D t
Here j refers to the constituency and t to the general election. Swing is swing between
Labour and Conservative19, with positive swing representing either swing to the gainer
(party which gains seats compared with the previous election) or swing to the
Government (party that is the currently in Government). M is a matrix of variables
measuring the marginality of the seat, X is a matrix of conditioning variables, and D is a
set of year dummies, i.e. fixed effects for each election. Fixed effects for each
constituency would not be appropriate: the majority of constituencies were consistently
safe or marginal between each set of boundary changes. Controlling for constituency
characteristics would thus largely control for just the variable we need to isolate for
analysis.
18
For the purposes of this article we use traditional or ‘Butler’ swing. Traditional or
‘Butler’ swing looks at parties’ share of the entire vote and averages one party’s gain and
another one’s loss. Two-party swing replaces shares of the poll by shares of the
combined vote of the two parties concerned. In some contexts two-party swing captures
the dynamics of the political process better than traditional swing. (For example, it does
not find a spurious ‘swing’ between the two main parties as a result of voters switching
between third parties and abstention.) Nonetheless, we use the traditional version of
swing in this article. This is because much of our analysis is to do with the possible
behavioural effects of voters’ realization that they are, or are not, in a marginal
constituency. Public discussion of the swing needed to topple an MP or a government is
entirely couched in the language of traditional swing. ‘The Tories’ 96th most winnable
seat’ is not the one where the Conservative : Labour vote ratio is the 96th nearest to 50%
but the one where Labour’s percentage margin over the Conservatives, as a proportion of
the entire vote, is the 96th smallest. Commentators on election night do not present twoparty swings, nor do they appear to have them in mind when they say ‘that’s one Labour
should have won if it’s seriously hoping to form a government.’ Consequently we use
Butler swing as the one most likely to shape electors’ perceptions of the consequences of
their vote.
An alternative to either Butler or two-party swing would be to adopt the three-party
method pioneered by Dorling, Johnston and Pattie20. This draws an electoral triangle with
100% Labour, Conservative and Liberal shares of the vote at each vertex, places each
constituency where its results dictate, and traces its movements between elections to find
any regularities that exist. (In a later paper Cornford and Dorling also use this method to
examine the likelihood that seats at various points in the map will change hands next time
round.)21 Given that the focus of the current article is on Conservative- and Labour-held
19
marginals (albeit with ‘other parties vote’ as an important control variable), we did not
employ the electoral triangle on this occasion.
A further important methodological point is how to model the kind of incumbency
effects, postulated and confirmed by Steed and Curtice, whereby a sitting incumbent
might specifically brake or accelerate the national electoral trend. If there is a positive
incumbency effect, the swing to the national gainer will be amplified in seats where the
gainer is fielding an incumbent this time, but was not doing so last time. Let us give this
situation an ‘incumbency score’ of one. If it was the party which is losing ground
(nationally) this time which fielded an incumbent last time, but is not doing so this time,
again the national gainer gets an incumbency score of one for that constituency – they are
losing a previously hostile incumbency vote. Where the current incumbent belongs to the
gaining party and is replacing an incumbent of the opposite party, there is a double
incumbency effect and the incumbency score is two. Clearly, for each of these positive
scores there is a negative counterpart – if, for instance, there was no incumbent at the last
election but a current incumbent, who belongs to the party which is currently losing
ground nationally, the party gaining ground gets an incumbency score of minus one.
Formally:
TABLE 3
INCUMBENCY SCORES
Incumbent at current election
National gainer at
current election
National loser at
current election
Other, or no
incumbent
0
-2
-1
2
0
1
1
-1
0
Incumbent at
previous election
National gainer at
current election
National loser at
current election
Other, or no
incumbent
20
5. Results
We now test the prediction of our model: that the party on the favourable end of the
swing will either (1) achieve higher swings in more marginal seats (2) enjoy a humpshaped or U-shaped relationship where swings are at their extreme in seats which are
neither too safe or too marginal.
TABLE 4
(1)
Majority
Majority
squared
Conservative
majority
Cons. Maj.
squared
Labour
majority
Lab. Maj.
squared
Labour seat
(2)
Swing to
gainer
0.0041
(0.118)
---
(3)
---
Swing to
gainer
0.0340***
(0.000)
-0.0006***
(0.000)
---
---
---
---
---
---
---
---
---
Swing to
gainer
-----
Other parties 0.0014
(0.810)
Incumbency 0.7091***
score
(0.000)
1951
1.070
1955
1.364
1959
0.895
1964
3.671
1966
2.890
1970
4.555
1974 (Oct)
1.937
1979
4.510
1987
2.563
1992
3.024
2001
1.409
2005
3.436
-0.0015
(0.787)
0.7230***
(0.000)
0.829
1.125
0.660
3.445
2.662
4.345
1.730
4.296
2.353
2.805
1.222
3.215
0.0724***
(0.000)
-0.0019***
(0.000)
0.0392***
(0.000)
-0.0005***
(0.002)
-0.2066
(0.255)
-0.0050
(0.413)
0.7137***
(0.000)
0.819
1.137
0.699
3.479
2.647
4.327
1.746
4.294
2.387
2.856
1.242
3.283
Adjusted R2
0.529
0.536
0.540
No. of obs.
5887
5887
5887
21
Column (1) regresses constituency swing to national gainer on percentage majority, other
parties’ (i.e. not Labour or Conservative) share of the vote and incumbency score as just
defined (p-values in brackets). The coefficient on majority is insignificant, thus giving
no reason to believe that marginal seats in general reflect the national trend at an election
any more or less strongly than safe ones. It is possible, though, that we are imposing a
linear trend on a non-linear relationship. Consequently column (2) tries out a quadratic
version, with swing to gainer regressed on both percentage majority and its square. Both
coefficients are now significant at the 1% level; their respective positive and negative
signs give us a humped curve. The downturn of the hump, however, comes quite late: it
takes a majority of 30% before the swing starts to turn down again. The implication,
then, is that the safer the seat, the more strongly it follows the national trend, except for a
slight reversal of the picture when seats are very safe indeed.
Before trying to interpret this, we make a distinction between Labour and
Conservative-held seats. According to column (2), both the linear and the quadratic
relationship between majority and swing are stronger when the seat is Conservative, but
the ratio of the linear to the quadratic coefficient is smaller, implying that Conservative
seats reach their ‘hump’ at a much lower majority than do Labour ones. Table 5
illustrates:
22
TABLE 5
Excess swing to gainer over swing in knife-edge seat
Seat held by
Majority
Labour
Conservative
10%
0.341
0.535
20%
0.580
0.690
30%
0.719
0.468
40%
0.753
-0.136
Majority which produces
strongest swing to gainer
38.1%
18.1%
So why might the national trend be strongest in relatively safe seats, as opposed to
marginal seats or extremely safe seats? This was seen to be consistent with our model if,
comparatively speaking, valence considerations do more to drive expressive voting, while
ideology does more to drive instrumental voting. It is perfectly plausible that things
might be this way round. Expressive voting might well be driven by a change in the
voter’s opinion of a party rather than by their long-term views. You might particularly
want to express a protest against a party you used to admire but which you think has let
you down: if you have always disliked that party, you may vote against it instrumentally
but expressive sentiments might have worn themselves out long ago. But it is a change in
valence which is most likely to provoke a mass expressive reaction: changes in ideology
are not only relatively slow-moving compared with the ups and downs of perception of a
party’s competence in government (or even in opposition); they will also attract some
voters even as they repel others. So the idea that expressive voting is comparatively more
about valence – and hence instrumental voting comparatively (if not necessarily
23
absolutely) more about ideology – does, perhaps, resonate more than the opposite version
does.
6. Some extensions
We now consider some factors which might affect voters’ perceptions of their ability to
affect the result. So far as these change p(x), p’(x) or p’’(x), they will alter the relative
weights of expressive and instrumental voting and hence the relation between existing
majority and swing. In the first place, voting may become more expressive and less
instrumental when there is little doubt about who is going to win nationally. Swinging
the result in a marginal seat may seem less important to the voter when the marginal seat
has no apparent chance of swinging the national outcome. The difference in swing
between marginal and safe seats would then be less pronounced. Thus we create the
interactive variables MAJ * NATMAJ and MAJ 2 * NATMAJ , where NATMAJ is the
national percentage lead (over the runner-up) of the election winner. Should the
coefficient on MAJ * NATMAJ be significant and opposite in sign to that on MAJ , that
would be consistent with our conjecture -- that the more decisive the national result, the
less distinctive the voters’ behaviour in marginal seats.
24
TABLE 6
MAJ*POLLLEAD
swing to gainer
0.0363
(0.010)***
-0.0009
(0.000)***
-0.0012
(0.643)
0.0001
(0.048)**
---
MAJ2*POLLLEAD
---
Other parties
1951
1955
1959
1964
1966
1970
1974 (Oct)
1979
1987
1992
2001
2005
0.0014
(0.808)
0.7213
(0.000)***
0.9536
1.2371
0.6827
3.5246
2.5838
4.4289
1.6919
4.2066
2.1947
2.7156
0.9834
3.2162
0.0003
(0.881)
0.0001
(0.110)
0.0018
(0.759)
0.7270
(0.000)***
0.8791
1.2379
0.7554
3.6008
2.4260
4.3860
1.5630
4.2159
2.2133
3.0100
0.7185
3.1865
Adj R2
0.5381
0.5402
0.4872
No. of obs.
5869
5869
1608
MAJ
MAJ2
MAJ*NATMAJ
MAJ2*NATMAJ
Incumbency score
Swing to gainer
0.0317
(0.022)**
-0.0009
(0.000)***
---
Swing to gainer
0.0504
(0.000)***
-0.0012
(0.000)***
---
---
-------0.0327
(0.002)***
0.3953
(0.001)***
0.9450
------2.9537
--2.1938
-----------
Column (1) of table 6 shows little evidence for the conjecture. MAJ*NATMAJ is
statistically insignificant. In contrast MAJ2*NATMAJ is significant at the 5% level but
the coefficient is tiny and, in any case, all this means is that the swing in very safe seats
does not decline so fast when the national result is more decisive – which does not follow
from our conjecture. The above does assume, however, that it is acceptable to proxy
voters’ expectations by the actual result. A better measure might be the average of
opinion poll forecasts on polling day. The variable POLLLEAD is the average percentage
25
lead, on polling day and across all polls, of the party forecast to win (from Butler and
Butler, 2000 and 2006).22 But, as column (2) shows, there is now, if anything, even less
support for the conjecture that marginal seats behave more like safe ones when the
national result is not expected to be marginal.
Another possibility is that voters’ awareness that the seat is marginal, and hence the
way some of them vote, depends on how long ago the last election was. As we have
already seen (p.4), Curtice and Steed have found that turnout in marginals is greater than
in non-marginals and that turnout in marginals is greater still in elections that come
quickly after their predecessors (1951, 1966, October 1974). But is there a parallel effect
on the way the inhabitants of marginals vote, as opposed to the likelihood that they vote
at all? When we run swing to gainer on majority and majority squared for the elections
of 1951, 1966 and October 1974 only, the coefficients on both variables go up (column
3). Three elections is rather a meagre data set, but our finding, so far as it goes, implies
that voters in marginal seats do behave more distinctively when there has been a recent
election to remind them that the seat is marginal.
A fundamental objection to the proposition that ‘marginals should be different’ is
that it is not the existing marginals that are going to swing the election, but rather those
seats that would be marginal in the event of a close national result (the ‘bellwether’
seats). All the above assumes that voters will regard as crucial those seats which currently
have small majorities. What if they realise that the seats crucial to the outcome are the
bellwether seats? If Labour needs to gain 100 seats to win the election, then, even with
the certainty of non-uniform swing, the 100th most marginal Conservative seat is far more
likely to be decisive than the most marginal one. Might it therefore be in the bellwether
seats that we see the most distinctive voting behaviour? To test this possibility we re-run
column (2) of table (4), except that our main independent variable is no longer the
26
existing majority in each seat, but the absolute difference between its percentage
government majority (positive or negative) and the percentage government majority in
the government’s nth most marginal seat, where 2n is its parliamentary majority over the
main opposition party—i.e. the bellwether seat.
TABLE 7
Swing to
gainer
BELL
0.0387
(0.000)***
2
BELL
-0.0008
(0.000)***
Other
-0.0002
parties
(0.973)
Incumbency 0.7157
score
(0.000)*
1951
0.8737
1955
1.1819
1959
0.7129
1964
3.4980
1966
2.6933
1970
4.3549
1974 (Oct)
1.6938
1979
4.3100
1987
2.3795
1992
2.9132
2001
1.2113
2005
3.2531
Adjusted R2
0.5380
No. of obs.
5869
The coefficient on BELL and BELL2 (0.039 and -0.0008) are slightly higher than they
were on MAJ and MAJ2 (0.034 and -0.0006). There is not a great deal in it, but if
anything the hump-shaped curve is more pronounced when we take bellwether, not
marginal, status as our predictor. However a t-test shows there to be no significant
difference in the magnitude of the coefficients between either MAJ and BELL or MAJ2
and BELL2.
27
7. Swing to government
We now go on to compare marginal and safe seats not in terms of ‘swing to gainer’ but
with respect to ‘swing to government.’ Does the government of the day typically fare
best in its safe, marginal or intermediate seats? And how does the result differ when these
seats are held by the opposition party?
TABLE 8
(1)
(2)
Swing to
govt.
-0.0095***
(0.000)
--
Swing to
govt.
--
--
Incumbency
score
1951
1955
1959
1964
1966
1970
1974 (Oct)
1979
1987
1992
2001
2005
-0.1424**
(0.013)
0.6926***
(0.000)
-0.938
1.680
1.255
-3.439
3.271
-4.345
2.458
-4.225
-2.223
-2.701
-1.048
-3.035
0.0267***
(0.000)
-0.0307***
(0.000)
0.9812***
(0.000)
-0.1284**
(0.024)
0.6940***
(0.000)
-1.506
0.990
0.605
-4.088
2.718
-4.817
1.908
-4.715
--2.904
-3.390
-1.480
-3.582
Adj R2
0.536
0.544
No. of obs.
5887
5887
Majority
Lagged swing
Opposition
Majority
Government
majority
Government
seat
Other parties
---
--
28
Column (1) of table 8 indicates that governments on average fare best (in terms of swing)
in marginal seats. But column (2) which differentiates between government- and
opposition-held seats (not just government and opposition incumbent MP’s), sheds a
different light on the whole picture. Our finding is illustrated in Fig.5
FIGURE 5
Swing to govt.
 opposition majority
0
govt. majority 
i.e.
(1) Parties do better in their own marginal seats than in those held by their opponents
(2) Subject to this exception, the swing in favour of a party is inverse to its existing share
of the vote in that constituency. You lose the most where you have been doing best.
We now search for explanations for each of these phenomena separately -- starting with
the second one -- before putting them back together to see what they portend about the
long-term fates of marginal and safe seats respectively.
29
7.1 Swing is inverse to existing share of the vote
If we take this effect by itself, it seems to predict that there is some kind of long-term
entropy occurring, and that a long enough wait would turn every safe seat into a
marginal. However, as will shortly be seen, any such trend will be checked, if not
reversed, when we combine it with parties’ tendency to do better in their own marginals.
And quite apart from this, a tendency for large majorities to get smaller may not be
entropy at all. Another possibility is that constituencies suffer shocks but then revert to
their own particular mean. In this case, a low vote at the last election and a favourable
swing at this one would be correlated because both would result from an unfavourable
shock at the last election.
We might call these competing explanations of our results the ‘entropy’ explanation
and the ‘mean-reversion’ explanation. We can distinguish between them empirically by
regressing constituency swings on both majorities and swings at the previous election.
The entropy theory implies that the coefficient on the former will remain significant and
negative; the mean reversion theory says it will lose its significance, transferring it to the
previously omitted variable of lagged swing.
Column (1) of table 9 suggests that there is some truth in both hypotheses. A 1%
constituency-specific swing at the previous election will, on average lead to a 0.15%
swing back at the current election (the year dummies are still in the regression but have
not been reported). Some weak mean-reversion is taking place. But controlling for this
effect does not destroy the significance of the coefficients on government majority and
opposition majority. The entropy hypothesis has survived the challenge.
30
TABLE 9 (year dummies omitted)
Opposition
majority
! Opposition
majority
Government
majority
! Government
majority
Government seat
First-timer
Other parties
Incumbency score
Lagged swing
Swing to govt.
Swing to govt.
Swing to govt.
0.0283***
(0.000)
---
0.0329***
(0.000)
---
-0.0289***
(0.000)
---
-0.0339***
(0.000)
---
0.931***
(0.000)
---
1.009***
(0.000)
0.081
(0.726)
-0.0095
(0.150)
0.8271***
(0.000)
---
0.0390***
(0.000)
-0.166*
(0.051)
-0.0350***
(0.000)
-0.004
(0.581)
1.062***
(0.000)
---
-0.0118*
(0.069)
0.8731***
(0.000)
-0.1549***
(0.000)
-0.001
(0.848)
0.701***
(0.000)
---
Adj R2
0.650
0.623
0.601
No. of obs.
3688
3754
4063
One extremely salient fact that we would expect to produce a degree of entropy is
the progressive decoupling of social class and party allegiance since the 1970’s. If class
were the only systematic determinant of the vote, it would all be extremely clear: if, over
time, more ABC’s and fewer DE’s voted Labour, Labour would get the most (least)
favourable swings in the seats with the most ABC’s (DE’s) i.e. the seats where it was
previously doing worst (best.) The most obvious objection to this simple story in the
British case is that we have seen geographical polarisation of the vote progressing
alongside its social decoupling. However, as argued above, the weakening of class as a
predictor of the vote appears to dominate the rising influence of geography – and, as has
31
been shown elsewhere, the regional effect would need to increase by more than the class
effect decreases in order for their consequences to balance out.23
7.2 Parties do best in their own marginals (partisan effect)
We cannot explain this by MPs’ personal incumbency effects because we are already
controlling for them with our variable ‘incumbency score.’ But a second possibility is
that parties put more time, money and effort into holding their own marginal seats than
into capturing marginals from their opponents. As we have seen, Clarke et al. find that
this is indeed the case.24 But in this case, we would see particularly favourable swings in
seats which a party captured from the other side at the previous election (over and above
any personal incumbency effect for the MP.) As column (2) of table 9 shows, we have
found no such effect. ‘First-timer’ is a dummy with a value of one for all seats that the
current government gained at the previous election. Although its coefficient is positive, it
is very small and statistically insignificant. After controlling for personal incumbency,
parties seem to get as good a swing in long-standing marginals as in ones which they
have just won over.
A third possibility is that the ‘chameleon effect’ is at work. It is well established
that voters take some of their political colouring from their surroundings. 25 It might be
down to the influence of friends and acquaintances26, an unconscious imbibing of the
dominant political culture or even a conscious decision to conform. It goes without
saying that any such effect will be much weaker in a marginal constituency which may
have recently belonged to the other side and where the vote by definition is fairly evenly
split. And again we have to ask whether any favourable swing from this source can live
on after the first election where the party fights the seat as sitting tenant. After that, any
chameleon effect, while sustaining the share of the vote, will not produce any further
32
swings. It could, however, be that the chameleon effect is slow-acting on some voters, so
that a change in the ownership of a seat, if not reversed, is still exerting some effect on
the voting two or more elections later.
To sum up, then, the zigzag curve which we found can be explained by a
combination of incumbent party (not just personal) effects in constituencies and a
general entropy effect produced by the decreasingly clearcut relation between social class
and party preference. The partisan effect is the more problematic of the two: can we
believe that seats which are fairly evenly balanced and also probably change hands from
time to time can have an incumbent party effect which is not only fairly strong but which
goes on producing favourable swings (not just holding on to an existing advantage)
election after election? Clearly more work needs to be done here.
But, whatever the explanation of the zigzag curve, we can go on to elucidate its
implications. Take a seat currently held by the party which is currently in government
(party A). Let the current election be denoted by 1 and the previous one by 0. Omitting
control variables, we can see from col.2 of table 8 that the swing in favour of party A in
seat x will be
s(A)x1 = NS(A)1 ! 0.0307MAJ(A)x 0 + 0.9812
where NS is the national swing before the effects of seat x’s majority and government
status are added.
i.e. MAJ(A)x1 = MAJ(A)x 0 + 2s(A)x1 = 0.938MAJ(A)x 0 + 2NS(A)1 + 1.962
33
Now suppose the swing at election 1 was against the government ( NS(A)1 < 0 ), and is
sufficient to put the opposition (party B) in power. x however stays with party A and
thus becomes an opposition-held seat. Consequently at election 2, the swing in favour of
party A in seat x will be
s(A)x 2 = NS(A)2 ! 0.0267MAJ(A)x1
i.e.
MAJ(A)x 2 = MAJ(A)x1 + 2s(A)x 2 = 0.946MAJ(A)x1 + 2NS(A)2
= 0.887MAJ(A)x 0 + 2[NS(A)2 + NS(A)1 ] ! 0.106NS(A)1 + 1.856
Now assume that the national swings in the two elections are equal and
opposite (NS(A)2 + NS(A)1 = 0) . Hence MAJ(A)x 2 > MAJ(A)x 0 if
0.113MAJ(A)x 0 < !0.106NS(A)1 + 1.856 . But since NS(A)1 < 0 , this condition will be
more than fulfilled provided 0.113MAJ(A)x 0 < 1.856 , i.e. MAJ(A)x 0 < 16.42 . All seats
with a majority of less than 16.42%, provided they are not so marginal that they
change
hands at election 1, will be safer at election 2 than they were at election 0.
Clearly this is all a gross simplification. Elections do not feature uniform alternating
swings, and constituencies can move into and out of the marginal category for all kinds of
idiosyncratic reasons. However, the simulation above is perhaps not entirely worthless in
demonstrating that incumbency effects have probably been strong enough relative to
entropy effects to explain the declining number of marginals.27
34
8. Being marginal or becoming marginal?
Our ‘swing to gainer’ model in section 3 implied that the key relationship was between
swing and current majority. There was no place for history – whether a marginal was of
long standing or had only become marginal at the previous election would make no
difference. But the putative factors behind the swing to government being highest in
government-held marginals (and lowest in opposition-held ones) all relate change in
government vote to the length of a seat’s marginal status. Incumbency effects produce a
swing when they first take effect, after which they hold the incumbent’s vote up but
increase it no more. If parties try hardest in their own marginals, this could only produce
cumulative swings if they make cumulative efforts. Similarly the chameleon effect will
stop showing up in election swings once the chameleon’s change of colour is complete.
None of these considerations means that effects on swing will blow themselves out after
one election. It may very well take longer than that for incumbency to pull its full weight,
or for parties to put seats on their priority list (a seat which is always marginal surely
deserves more care than one which is occasionally marginal), or for the chameleon’s new
colour to be noticed by one and all. Nonetheless, all three effects give some ground for
thinking that becoming a marginal may be at least as important as being a marginal when
we look at the fortunes of the party holding the seat. If parties get better results in their
own marginals, they will do best in terms of swing in those which have just become
marginal and worst in those that have just ceased to be marginal. Accordingly column 3
of table 9 regresses swings on both the majority and the change in the majority at the last
election.
It looks as if being a marginal, not becoming a marginal, is what matters. The
coefficient on ‘change in government majority’ is very small and insignificantly different
from zero, while that on ‘change in opposition majority’ is almost significant at the 5%
35
level, but with the wrong sign—if the opposition majority went up last time that should
be good for the swing to the government this time because the seat is now less of an
opposition marginal. The failure of ! MAJ to capture the significance held by MAJ
reinforces the implications of our finding (table 9, column 2) that the party incumbency
effect on swing does not exhaust itself at the election after the one in which that party
won that seat.
9. Conclusions
We have found two basic relationships between a constituency’s majority and its likely
swing at the next election. The first is that the strongest directional swings are to be
found in seats which are neither too safe nor too marginal. This might be because very
safe seats have fewer floating voters, while marginal seats are dominated by instrumental
voting which might be less responsive to changes in parties’ valences than its expressive
counterpart. Ranking seats not by their current majority but by their distance from the
‘bellwether’ seat (the one which would be most marginal if the election as a whole were
on a knife-edge) did not produce significantly different results. The second finding was
that parties do better in their own marginal seats and their opponents’ safe seats than in
their own safe seats and their opponents’ marginal seats. This could be due to the
combination of two forces. On the one hand, small majorities tend to firm up as parties
switch time and effort to their retention, and as incumbent MP’s dig themselves in. On
the other, there could be a long-term entropy at work causing parties, other things equal,
to do worst in their strongest seats and best in their weakest. The progressive weakening
of the link between social class and voting would achieve this result – and we argue that
any geographical polarization of voting behaviour has been insufficient to offset it.
36
FOOTNOTES
* This project was funded by the Leverhulme Trust, award number SC-04151.
Constituency results between 1950 and 1997 are from Caramani,
D.,
Elections
in
western
Europe
since
1815:
electoral
results
by
constituencies
(London:
Macmillan, 2000),
while Constituency results for 2001 and 2005 are taken from Norris P., The British
Parliamentary Constituency Database, 1992-2005 (Release 1.3). The incumbency status
of candidates is compiled from Whitaker’s Almanack. Sources of all other figures are
specified in the relevant footnotes below.
1 see W. Riker, and P.C.Ordeshook, An Introduction to Positive Political Theory
(Englewood Cliffs, Prentice-Hall, 1973)
2 H.D.Clarke, D.Sanders, M.C.Stewart and P.Whiteley Political Choice in Britain,
(Oxford, OUP, 2004), p.248
3 see P.Whiteley, P.Seyd, J.Richardson and P.Bissell ‘Explaining party Membership: The
Case of the British Conservative Party’, British Journal of Political Science 24 (1994).
79-94; P.Whiteley, P. and P.Seyd , High-Intensity Participation – The Dynamics of Party
Activism in Britain (Ann Arbor, University of Michigan Press 2002)
4 J.M.Buchanan, ‘Individual choice in voting and the market’, Journal of Political
Economy, 62 (1954), 334-343.
5 G.Tullock, ‘The charity of the uncharitable’. Western Economic Journal, 9 (1971),
379-392.
6 S.Price and D.Sanders ‘By-elections, Changing Fortunes, Uncertainty and the
Mid-Term Blues’, Public Choice, 95 (1998), 131-48
37
7 D.Denver, D., G.Hands and I.McAllister, ‘Constituency Marginality and Turnout in
Britain Revisited’, Journal of Elections, Public Opinion and Parties, 13 (1998), 174-94
8 J.Curtice, and M.Steed, ‘An analysis of the voting’, Appendix 2 of D.Butler and
D.Kavanagh, The British General Election of 1979 (London, Macmillan, 1980) 390-431,
p.422
9 Clarke et. al., Political Choice in Britain
10 Clarke et. al., Political Choice in Britain
11 Clarke et. al., Political Choice in Britain
12 J.Curtice, J., S.Fisher and M.Steed, ‘The Results Analysed’, Appendix 2 of
D.Kavanagh & D.Butler, The British General Election of 2005 (Basingstoke, Palgrave
Macmillan, 2005), 235-59, p.248; J. Curtice, J. and M.Steed ‘The Results Analysed’,
Appendix 2 of D.Butler and D.Kavanagh, The British General Election of 1992 (London,
St Martin’s Press, 1992), 322-362, p.340; J.Curtice, and M.Steed (1988), ‘Analysis’,
Appendix 2 of D.Butler and D.Kavanagh, The British General Election of 1987 (London,
Macmillan, 1988), 316-362, pp.333-4; M.Steed, ‘An Analysis of the Results’, Appendix
2 of D.Butler and M.Pinto-Ducshinsky, The British General Election of 1970 (London,
Macmillan, 1971) 386-415, p.404
13 Curtice & Steed ‘An analysis of the voting’ (1980)
14 J. Curtice, and M.Steed (1984), ‘An analysis of the voting’, Appendix 2 of D.Butler
and D.Kavanagh, The British General Election of 1983 (London, Macmillan, 1984), 333373; Curtice and Steed, ‘Analysis’ (1988)
15 Pippa Norris and Ivor Crewe ‘Did the British Marginals Vanish? Proportionality and
Exaggeration in the British Electoral System Revisited’, Electoral Studies, 13 (1994), 201–
22
38
16 The explanation is straightforward. 1979-83 saw a rise in the third-party vote that
proved permanent. Clearly the larger the third-party vote, the more a constituency
majority as a proportion of the two-party vote exceeds its counterpart as a proportion of
total vote. Hence any trend towards fewer marginals will be amplified by the use of twoparty swing.
17 Taken from Clarke et. al., Political Choice in Britain, p.43
18 see R.Hodgson and J.Maloney ‘Why Has Britain Fewer Marginal Seats Than It Used
To?’ University of Exeter Department of Economics Discussion Paper 11/05 (2011)
19 This includes swings between Conservative and Labour in seats where one of these
parties won but the other was not the runner-up. We declined to exclude such seats: they
are not the only ones in which swings may have been distorted by tactical voting, and in
any case there is a control variable in the form of ‘other parties’ vote.’
20 D.Dorling, R.J.Johnston. and C.J.Pattie, ‘Measuring Electoral Change in three party
systems: an alternative to swing’, Political Science and Politics 26 (1993), 737-741.
21 J.Cornford and D.Dorling, ‘Crooked Margins and Marginal Seats’, Journal of
Elections, Public Opinion and Parties, 7 (1997), 174-90
22 figures from D.Butler and G.Butler, British Political Facts, 1900-2000 (Basingstoke,
Palgrave Macmillan, 2000) and D.Butler and G.Butler , British Political Facts since
1979, (Basingstoke, Palgrave, Macmillan, 2006)
23 Hodgson and Maloney, ‘Why Has Britain Fewer Marginal Seats Than It Used To?’
24 Clarke et. al., Political Choice in Britain
25 D.Butler and D.Stokes, Political Change in Britain (London, Macmillan, 1974),
chapter 6, and W.L. Miller, Electoral Dynamics, (London, Macmillan, 1977) are among
those who provide a variety of evidence for this proposition.
39
26 S.D.McClurg, ‘The Electoral Relevance of Political Talk: Examining Disagreement
and Expertise Effects in Social Networks on Political Participation’, American Journal of
Political Science, 50 (2006), 737-54; S. Richey,‘The Autoregressive Influence of Social
Network Political Knowledge on Voting Behaviour’ British Journal of Political Science,
38 (2008), 527-42
27 With the entropy effect causing parties to lose a proportion of their majority at each
election (but being counteracted by incumbency effects) it might sound as if we are
signing up to something like the Butler-Stokes hypothesis in which the party on the
wrong end of the national swing loses a proportion of its vote, but this is reconciled with
allegedly uniform swings by counteracting chameleon effects which, mysteriously, are
claimed to switch on only when a party is losing ground (Butler and Stokes, Political
Change in Britain). In fact we are saying something rather different. In our version the
tendency of the parties’ majorities to shrink is not caused by the national swing but is the
background against which national swings take place.
40