25
7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 1/25 Explaining Turnout Decline in Britain, 1964-2005: Party Identification and the Political Context Author(s): Oliver Heath Source: Political Behavior, Vol. 29, No. 4 (Dec., 2007), pp. 493-516 Published by: Springer Stable URL: http://www.jstor.org/stable/4500258 . Accessed: 30/10/2013 16:42 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Springer is collaborating with JSTOR to digitize, preserve and extend access to Political Behavior. http://www.jstor.org This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PM All use subject to JSTOR Terms and Conditions

Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

Embed Size (px)

Citation preview

Page 1: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 1/25

Explaining Turnout Decline in Britain, 1964-2005: Party Identification and the PoliticalContextAuthor(s): Oliver HeathSource: Political Behavior, Vol. 29, No. 4 (Dec., 2007), pp. 493-516Published by: Springer

Stable URL: http://www.jstor.org/stable/4500258 .

Accessed: 30/10/2013 16:42

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of 

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Political Behavior.

http://www.jstor.org

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 2: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 2/25

Polit Behav (2007) 29:493-516

DOI 10.1007/s 11109-007-9039-4

ORIGINAL PAPER

Explaining Turnout Decline in Britain, 1964-2005:

Party Identification and the Political Context

Oliver Heath

Published online: 21 August2007

? SpringerScience+BusinessMedia, LLC 2007

Abstract Turnoutdecline in Britain s greater han it firstappearssince changesin

the social compositionof the electorate have had a positive impacton turnout.This

paper finds that whereas a weakening in the strength of party identification is

associated with the long-term decline, the political context influences short-term

variation.Partisandealignment s also changingthe dynamicsof the determinants f

turnout.Since non-identifiersare more strongly influencedby the political context

than strongidentifiers,and there are now more non-identifiers hanpreviously, the

politicalcontext is becominga moreimportant actor in determiningwhetherpeoplevote or not.

Keywords Turnout . Party identification . Political context

As in many other maturedemocracies, there has been growing concern in Britain

aboutdeclining levels of electoralparticipation.This concern came to a head in the

2001 General Election when turnout ell by an unprecedented12 percentagepoints

from 1997 to just 59.6%, and reached its lowest level since full adult franchise.Recent research, both in the UK and elsewhere, has devoted considerable

attention to this phenomenon,and three main argumentshave been put forward n

order to explain it. Broadly speaking these relate to socialisation effects and the

impact of generationalchange (Miller 1992; Wattenberg2002; Clarke et al. 2004;Franklin 2004), partisan dealignment and the weakening of party identification

(Wattenberg2000, 2002; Dalton et al. 2000), and (decreased) mobilisation by

political parties (Rosenstoneand Hansen 1993).

Although nterlinked, ach of these theoriesoffersverydifferent nterpretationsor

thecauses of turnout hangeand,by implication, he natureof therelationshipbetween

O. Heath(F?)

Departmentf Government,niversityf Essex,Colchester 04 3SQ,UKe-mail:[email protected]

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 3: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 3/25

494 Polit Behav (2007) 29:493-516

voters and parties. Accounts that emphasise the impact of generationalchangedominate the literature.Franklin's(2004) socialisation thesis posits thatvoting is a

habit whichis learned or not)in the firstfew elections thatanindividual s eligible to

vote in. Theelectoralcontext stimulatesnew votersto eitherparticipate rnot,andthis

leaves a 'footprint'on their futureparticipation.Turnoutdecline is thus primarilya

function of new cohortsentering he electorateand old cohorts eaving it.However, 'generation effects' are very difficult to empirically verify. Age,

cohort, and period are all co-linear which makes identificationimpossible unless

some simplifying assumptionsare made. For example, if age and cohort (year of

birth) are known, period (time of survey) is a given, or if cohort and period are

known, age is a given etc. Thus much previous researchon this subjecthas been

rathersuggestive. Analysis of the generationalbasis for turnoutdecline has tended

to rely on the constructionof 'psuedo panels' to illustratethe relativedifference in

levels of reportedturnout between different age cohorts over time. Clarke et al.

(2004) and Franklin(2004) adoptthis techniqueand arguethat modem day youngcohorts display much lower turnout han their olden day equivalents.The problemwith this approach s that it is not able to distinguishbetween the individual level

ageing processes that arerepeatedlyfound to have a positive impacton turnout,or

period effects, which may have short-term(ratherthan permanent)effects and

influence everyone (more or less) equally. But, perhapsmore fundamentally,the

main problem with accounts such as this is that they provide somewhat

deterministicexplanationsof political behaviour,which are unable to take account

of suddenchanges,

such as we observed in 2001 when there were over 5 million

fewer votes cast than in 1997.1Changesof this scale obviously cannot be the result

of generational replacement alone, so we must begin to look for additional

explanationsof turnoutchange elsewhere.

The second account of turnoutchange relates to the role of partyidentification

and partisandealignment. Since the work of Campbell et al. (1960, 1966) partyidentificationhas been a key featurein explanationsof why people vote, and since

its introduction o the British context by Butlerand Stokes (1969) it has repeatedlybeen found to be one of the most important predictorvariables for explaining

turnout(see Heath et al. 1985; Clarke et al. 2004). Althoughthere are conceptualambiguities surrounding he precise meaning of party identification,two broad

theories dominate the literature.Firstly,in its originalconception,it is firmlybased

in social psychology and can be regardedas a stable formof self-identification. t is

acquiredthrough early-life socialisationexperiences (Campbellet al. 1960; Butler

and Stokes 1969) and is strengthenedover time throughthe constantrepetitionof

voting andparticipating n the electoralprocess (Converse 1969, 1976). As such it

exerts a powerfullong-termforce on voting behaviour.By reducing he information

and decision making costs, partisanloyalties transform he act of voting from a

strictly instrumental ationalcalculationto a matterof habitborneout of a sense ofloyalty. This interpretationhas some similarities with the socialisation thesis of

generational replacement, and some scholars have suggested that partisan

1In 2001 there were 25,554,856 votes cast whereas in 1997 there30,496,924. PippaNorris TheBritish

ParliamentaryConstituencyDatabase 1992-2001.

1 Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 4: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 4/25

Polit Behav (2007) 29:493-516 495

dealignment contains a generational component (Clarke et al. 2004). However,

more recently this socialisationinterpretation as come underattack,and since the

1980s there has been a rise in the popularityof more rationalapproachesto the

studyof party dentification.Accordingto Fiorina(1981) party dentificationcan be

considereda weighted averageof past evaluationsof political parties,and follows

Downs' (1957) partydifferential erm,which states thatretrospectiveevaluationsofpolitical parties are a rationalway of calculating future expectations in decidingwhichpartyto vote for. It is continuallyupdated,andrepresentsa 'running ally' of

past evaluations. This line of thinking has been developed by, inter alia, Stokes

(1992), Achen (2002), Stewartand Clarke(1998) andClarkeet al. (2004). Although

differing n detail,theircore argument s thatpartyID is notjust the productof earlylife socialisation but responds to the macro political context and is sensitive to

political cues. It is updated bothin terms of directionandintensity)by information

relatingto evaluations of party political behaviour and variationsin the economy

(Clarkeet al. 2004: 28).

Lastly, the role of mobilisation has been emphasised by Rosenstoneand Hansen

(1993). This takesvarious orms,ranging rommobilisationby parties knockingup),civil rightsandcommunityorganisations o mobilisationby thepolitical contextand

the structureof electoralcompetition.Rosenstone andHansen(1993: 161) contend

that"people participaten electoralpolitics because someoneencouragesor inspiresthemto takepart",andso turnoutdecline is, accordingto them,mainlythe resultof

decreased mobilisation. This rather broad definition of mobilisation essentially

containstwo components.The firstrelatesprimarily o campaigneffects by politicalparties,specifically face to face campaigning.They arguethata move from labour

intensive door-to-doorcampaigning, o money intensive advertisingcampaignshas

had a negative impact on turnout(ibid 217-218). The second point relates to the

impactof the structure f electoralcompetitionand factorsthatmay stimulatepeopleto participate.Accordinglyturnout s likely to be higherin elections that areclosely

fought, since each vote counts (see also Denver and Hands 1974, 1985; Pattie and

Johnston2001; Clarkeet al. 2004; Franklin2004). However,since there hasbeen no

uniformdeclineinthecompetitivenessof elections inBritain,andthe electoralcontext

changesfrom election to election,thisfactorappearsunlikelyto explainthelong-termturnoutdecline observedover the last few decades,although t couldexplaina sudden

droplike, for example, in 2001.

Althougheach of these theories has received some empirical support,there has

been little attemptto simultaneouslymodel theirrelative importance n relation to

each other in order to explain their unique impact on the dynamics of turnout

change. Previous research on turnout has thus tended to provide only partialaccounts of why people vote. This is partly because two different, yet not

incompatible,methodologicalperspectiveshave dominated he analysisof turnout.2

2 Therehavebeen many studiesthat examine the individual evel determinants f turnoutat one pointin

time (Clarkeet al. 2004: 237-260; Whiteleyet al. 2001) and from an over time perspective(Miller 1992;Abramsonand Aldrich1981).There have also beenmanystudies thatexaminethe aggregate evel impactof thepoliticalcontextat theconstituency evel (Whiteleyet al. 2001; Denver andHands1974, 1985)andat the national level over time (Clarkeet al. 2004: 261-268; Franklin2004). However, therehave been

very few attemptsto synthesise these two approaches ogetherto explain turnoutchange.

ISpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 5: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 5/25

496 PolitBehav 2007)29:493-516

The first uses cross-sectional survey data taken at one point in time to examine

individual level factors which explain voting in terms of social class location,sense

of political efficacy andparty dentificationetc. (see Heath et al. 2001; Verba et al.

1995). The second uses aggregatelevel data to examine the political context, such

as the expected closeness of the contest, the ideological difference between the

parties and the level of campaign spending and mobilisation (see Whiteley et al.2001; Pattie and Johnston 1998; Denver and Hands 1974, 1985). By considering

only time series aggregatedata there is the risk of ecological fallacy or spuriouscorrelation.However, by consideringonly cross-sectional data it is only possible to

discriminate between people within elections, not people across elections. This

meansit is not possible to examine factors thatmay influenceeverybody n the same

way, such as the expected closeness of the contest.

Notableexceptionsthat have attempted o integrate hese two approaches nclude

Rosenstone and Hansen (1993), Heath and Taylor (1999), Pattie and Johnston

(2001), and Franklin(2004). Rosenstone and Hansen (1993) comparethe relative

impact of individual-level factors and contextual-level factors on turnoutchange,but since theiranalysis is only based on two periods(the 1960s vs. the 1980s) their

results are unable to uncoverpreciselyhow thepoliticalcontextshapesand interacts

with individual characteristics.Similarly,PattieandJohnston(2001) combine over

time aggregatelevel analysis with individuallevel survey analysis from 1992 and

1997 to try andexplain turnoutdecline between the two elections. And, in perhapsthe most comprehensive attemptso far, Franklin(2004: Chapter6) examines the

impactof the electoral context on individual turnout

using pooledcross sectional

survey data from Germany,but he does not model the impact of this on turnout

change. Rather,his chapterson turnoutchange are all based on aggregatedata.By

building on this previous research, and extending the number of elections under

investigation, this paper integrates these two approaches by modelling both

individual evel factorsand contextual evel factorssimultaneously. n doing so it is

able to test the impact of both short-termcontextualfactors and long-term socio-

compositionalfactorson turnoutchange, and examine the interplaybetween them.

Data and Methods

In this paperI synthesisethese two methodologicaland theoreticalperspectivesand

analyse a nested model of electoral behaviour that examines individuals within

years.To do this I use pooled cross-sectionaldata fromthe BritishElection Studies

1964-2005 (see Appendix for details). This is the longest time series data on

political behaviour available in the UK and covers 12 elections and consists of

32,085 interviews. By using this approachI am able to pay particularattentionto

compositionalchanges thathave takenplace in the distributionof the sociologicaland attitudinalcharacteristicsof the electorate.Compositioneffects of this type are

rarely consideredin studies of turnout,which is a severe drawback(althoughsee

Heath and Taylor 1999). Concentrating olely on the effects of variables without

taking into account their compositional distribution in the electorate can be

particularlymisleading since in very high turnoutelections approaching100%,no

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 6: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 6/25

Polit Behav (2007) 29:493-516 497

variables will have significanteffects since everyone will vote. To make sense of

turnoutchange it is therefore necessary to model compositional changes in the

electorate,as employedin this paper,rather han ust concentratingon the effects of

variables to see if they have changed or not.

Of course it is possible that, as the composition of a factorchanges, so does its

effect.3 If this is the case then the assumptionof parametricuniformityover timemay not be valid, and modelling the effect of compositional changes may be

misleading. To test for this possibility I thereforespecify a number of interaction

terms between the independentvariables and year. Broadly speaking the results

show thatthere is no systematicparametricvariationover time, certainlynone that

is related to the composition of the variable.4 The assumption of parametric

uniformityis thereforefairly robust.

Turnout 1950-2005

Turnout levels in Britain have dramaticallyfallen over the last 50 years, from

around80%in the 1950s to around60%in the 2000s (Fig. 1). Fromthe equationof

the regression line we can see that on average turnout has fallen by about

3Forexample,the salience of class in determining urnoutmay weaken as society becomesmoremiddle

class. However, it is also possible that the effect of class may be stronger n some years thanin others

withoutany generaltendencyto change.If this is the case thenthe source of the variation s unlikelyto be

due to long-termforces and may be the result of particular hort-term actors, possibly relatingto the

politicalcontext. Underthese circumstances he assumptionof parametricuniformity s still broadlyvalid

but it may be necessaryto try and specify additional nteraction erms in orderto explain the differingeffects from year to year.

4 Including he interactionbetween class andyeardoes not resultin an improvementn the fit to the data

(Z2 change eq 7.4, df = 11, p = .767). This suggests that the working class have not become

disproportionatelyess likely to vote with Labour's shift to the right(Budgeet al. 2001). However,there

have been some modestchanges in the odds of voting over time by tenure(X2change eq 24.3; df = 10;

p = .007), and rathergreaterchanges by Franchise(X2changeeq 32.7; df = 8, p < .0005) partyID (X2

changeeq 42.5; df = 11,p < .0005) andage (Q2changeeq 34.4; df = 11;p < .0005). The changes bytenureare

relatively unimportantince it is

onlyin 2005 that

significantdeviationsareobserved

(whentheeffect was stronger han n otheryears).Thechangesby Franchisearealso modest(theeffect is somewhat

weaker in 1970 and 1975 andis somewhatstronger n 1987) suggestingthatthere is little structureo the

variation.In both instances, without sound theoretical ustificationfor including additional interaction

terms,the assumptionof parametrichomogeneityis thereforeupheld.The changes by age and partyID

aresomewhat more interesting.There are significantdifferencesin the effect of the age variable n 2001

and 2005 (the two elections with particularlyow turnout).In both these elections the effect of the ageterm was stronger han usual(meaningthatthe oddsof votingincreasemorerapidlywith eachyearof ageor thatyoung people were even less likely to vote in relation to old people thanusual). Thereare also

significantdifferencesin the effect of the partyID variable n 1983 andOctober1974, when the effect of

partyID was somewhat weaker than normal(meaning that strong identifiers were not so much more

likelyto vote than non identifiers than at other

times),and in 2001, when the effect of

partyID was

somewhatstronger han normal(meaningthat weak identifierswere even less likely to vote thanstrongidentifiers hanat othertimes). These short-term hangessuggestthatthe effect of age andpartyID maybe related to the political context (Franklin 2004). Since there has been no long-term change, and

moreover,since we can be surethatthechangesthathave takenplace are not relatedto the distribution f

the variable n the population, here is no reason to relaxthe assumptionof parametrichomogeneity(as a

general principle)at this point. However,we will return o these points in the second partof the paperwhen we considerthe political context.

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 7: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 7/25

498 Polit Behav (2007) 29:493-516

90-

85

80-

0 75

70

. 65

60

55 - y = -0.2894x+ 646.31

R2 = 0.6186

50

1945 1955 1965 1975 1985 1995 2005 2015

Fig. 1 Turnout in Britain, 1950-2005 (official results). Source: 1945-1997 from Heath and Taylor

(1999: 165);FromElectoralcommission 2001 and 2005 reports http://www.electoralcommission.org.uk/

election-data/index.cfm)

0.29 percentagepointsa year, althoughfrom the dispersionof thepointsaround his

line we can see that the decline has not been uniform and in some years it dropped

considerably more than expected, whereas in other years it actually increased.

However,overall, it is evident that a trendof long-termdecline has takenplace. The

R2 of 62% shows that notwithstanding he short-termresidual variation,a linearmodel of decline providesa good fit to the data. This result is notjust dependenton

unusualstartor endpoints,and is robustto the exclusion of any particular lection.5

This holds even when as many as four elections are randomlyomitted at a time.

Indeed, even if the two highest turnout elections of 1950 and 1951 and the two

lowest turnoutelections of 2001 and 2006 arediscountedfromthe analysis,there is

still a significant long-term decline in levels of turnout.6Moreover, specifying a

linear trendover time providesa far betterfit to the data thana non-lineartrend.A

regressionusing the natural ogarithmof years passed since 1949 accountsfor only

47% of the variation,suggesting that this does not provide such a good fit to thedata. Although the trend of turnoutdecline is most pronouncedsince the 1950s,

there is also significantevidence of a longer term general downward trend since

1929 (the firstelection in which women were eligible to vote). On average, since

1929, turnouthas fallen by 0.14 points a year.7There is thus evidence of two distinctphenomena:a general long-termtrendof

decline, particularly since the 1950s, accompanied by short-term fluctuation

(residual variation).To understand he dynamics of turnoutchange it is therefore

5 Bootstrapping eveals that theseresults are robustto the exclusion of any particular lection, and arein

fact strongerwhen some elections are omitted.6 Linear regression of turnoutdecline 1955-1997: % turnout= 77.7 - 0.08 year (one sided t test

significantat 0.1 level).

7 Linear regression of turnout decline 1929-2006: % turnout= 79.7 - 0.14 year (one sided t test

significantat 0.01 level).

4Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 8: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 8/25

Polit Behav (2007) 29:493-516 499

necessaryto exploreboth these phenomenaandanalysethe factorswhich determine

each of them. Only by doing this is it possible to get a clear idea of how much

secular decline has takenplace over and above the short-termvariation rom which

turnout evels can easily bounce back. I arguethat long-termchanges in the socio-

psychological compositionof the populationareresponsiblefor the seculardecline,

and short-termchanges in the political context are responsible for the residualvariation. The firstpartof this papertherefore examines structural hanges in the

composition of the electorate,paying particularattentionto hypotheses of turnout

changederivedfrom generationaldifferences andpartisandealignment.The second

partof thispaperthenexamines the short-term actors which may cause the residual

variationaround he long-termdecline and considers the political context. In doingso I address he interplaybetween these factorsand theirimplicationsfor the future.

Social Change and Turnout

Demographic actors such as social class, housingtenure,age, genderandeducation

have been found to have modest associations with turnoutat the individual level in

Britain(Swaddleand Heath 1989; PattieandJohnston1998; Whiteley et al. 2001),

althougha very strongeffect elsewhere,particularlyn the UnitedStates (Verbaand

Nie 1972;Verba et al. 1995). In general,people with high socioeconomic status are

more likely to turn out thanthose with low status.Accordingto Verbaet al. (1995)socioeconomic status acts as a civic resource that facilitates

participation byreducingthe costs associatedwith understandingpolitics in general and the policy

positionsof the parties.Similarlyhousingtenurehas also been found to have a more

direct effect on voting by reducingthe costs of registration Brians 1997).As in manyother advanced ndustrialdemocracies, n the last 50 yearsthere have

been substantialchanges in the socio-economic composition of the electorate in

Britain which would appear to favour higher levels of turnout. The class

composition of the British population has gone througha major transformation,with the workingclass shrinkingand the Salariatexpanding.8The averagelevel of

home ownership s also much higherthanit used to be, andthe average age of theelectorate is older.9

Social changeof this type is often consideredto represent omethingof a paradoxin relationto levels of turnout.Since the middle class are more likely to vote than

the working class, and there are more middle class people in Britain today than

previously,it would seem logical all otherthingsbeing equal that turnout hould be

higher.The fact that it isn't means thatthe positive impactof these social changesmust be outweighed by the negative impact of othervariables.

This brings us to arguments about generational change. However, since, as

previously mentioned, there are identification problems when age, period and

8 If respondentdoes not have social class then social class of spouse is used as a proxy.

9 After 1992theaverageage of peopleon the electoralregisterwas substantiallyolderthan t was before.

However,rather hanreflectinga suddendemographic hangethis is probablymorelikely due to council

tax avoidance,which meant that in the 1990s many people, particularlyhe young,did not sign up to the

electoralregister.

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 9: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 9/25

500 Polit Behav (2007) 29:493-516

generationare considered simultaneously,it is thereforenecessary to make some

simplifying assumptions.To this end I draw on Franklin's (2004) socialisation

thesis which providesthe most plausibleaccountof why different cohortsmay turn

out at differentrates. Franklinsuggests that differentgenerationsor cohorts exhibit

different rates of turnoutdepending upon the characterof elections they were

exposed to when they were firstpolitically socialised. People coming of age in lowturnoutelections are likely to vote at lower rates throughout heir adult life than

people coming of age in high turnoutelections.

Therefore,ratherthan measuringcohort directlyI follow Franklinand measure

turnoutat the first election that each electoral cohort was eligible to vote in. For

example, people born between 1938 and 1943 were first eligible to vote in 1964,when turnoutwas 77.1%, so each person in this cohort scores 77.1 on the cohort

variable, people born in 1944 and 1945 were first eligible to vote in 1966, when

turnoutwas 75.8%,andso score75.8, andpeople bornbetween 1946 and 1952 were

firsteligible to vote in 1970 when turnoutwas 72.0% etc. This is repeatedfor eachelectoral cohort bornbetween 1904 and 1987. The measurethus tells us whetheror

not respondentswere socialised in relatively high turnoutelections.

Although this approachsolves the identificationproblem, since the electoral

cohort variable is not collinear with age andperiod, it possibly underestimates he

effect of generation.It is a theoreticalconstructrather hanan empiricalconstruct,

and as such theremaybe additionaldifferencesbetween generations hat it does not

take into account (for example in strengthof party identification).Moreover, the

measurementof the theoryis not perfecteither.In his originalformulationFranklin

suggests that socialisation takes place over a relatively long period-he cites three

elections-but the measure used here is based on exposure to just one election

(otherwise it would mean droppingthe first two and last two elections from the

series which wouldbe too heavy a price to pay). This meansthe expectedeffects of

the variable are likely to be somewhat weaker than Franklin would suppose.

However, despitethese limitations t still providesa potentially mportant ource of

turnoutchange, and at least allows us to consider the possible impacton turnoutof

generationalchanges, even if these are not captured n theirentirety.Franklinalso

arguesthat the extension of the franchise had

important long-term implicationssince people who reached the age of maturitybefore full franchisewas extended

were in a sense socialised into the act of not voting.10 The proportionof the

electorate who came of age before full franchisewas implementedhas decreased

from 23% in 1964 to less than 1%in 1997 and almost nothingthereafter.

Lastly, over the last few decades there has been a substantialdecline in the

strengthof party identification(see also Crewe et al. 1977; Dalton 2000; Clarke

et al. 2004). This phenomenon s notjust limited to Britainor the United States.In

an analysis of 19 advanced democracies, Dalton (2000) finds a similar trend in

practicallyevery country,and in the United States a similardealignmenthas beenfound to have a substantial mpacton turnoutdecline (Abramsonand Aldrich 1981;

10Therehavebeen various amendments o the franchise n Britainover the last 100 yearsor so. In 1918

women over 30 were given the vote for the firsttime, and all men over 21 were eligible to vote (priorto

that there were propertyrequirements or men). In 1928 full adult franchise as we know it today was

establishedwhen the age of maturity or women was reducedso that it was the same as for men.

i Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 10: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 10/25

Polit Behav (2007) 29:493-516 501

Rosenstone and Hansen 1993; Wattenberg2000; althoughsee Franklin2004 for a

counterargument).

Model Specification

To untanglethe impactof these differenttrendsandestablish theiruniqueimpacton

turnoutI specify a dynamicmodel which takes into accountcompositionalchangesin the electorate over time. To do this I examine a series of logistic regressionmodels. (Since turnoutis a binary variable, coded 1 for voted and 0 for did not

vote)." The logistic model examines individualsnested in years and whethertheyvoted or not. The firstmodel containsonly dummytermsfor eachelection year.The

magnitudesof these coefficients indicate the level of observed variation n reportedturnout n relationto the baseline figurein 1964, andcan be viewed as unexplained

year effects. The subsequent models each control for the composition of an

additionalsociological factor: model 2 also controls for age, sex, social class and

tenure;model 3 and 4 control for the generationalterms of full franchise (this is

interactedwith sex since priorto 1928 women did not have equal voting rights)and

electoral cohort(turnoutat firstelection thatpersonwas eligible to vote).12Finally,model 5 controlsfor partyidentification.13By controllingfor age, electoralcohort

and year in this way we are able to untangle the independent impact of age,

generationandperiodwhich have provedso problematic o differentiatebetween in

previousstudies on turnout.Thus the full

specificationfor the final model is:

Pr(Voting) = fn(Bo+ B1 1966 + B2 1970 + B3 1974F+ B4 19740 + B5 1979

+ B6 1983 + B7 1987 + B8 1992 + B9 1997 + B102001 + B112005

+ B12Age + B13Class+ B14Male+ B15Tenure+ B16Franchise

+ B17Franchise Male + B18Electoralcohort+ B19Party D) (1)

The extent to which controllingfor these factors influencesthe magnitudeof the

coefficients for the yearly dummyterms reveals the extent to which variation n, for

" Turnout s measuredby whetherrespondentsreportedhavingvoted in the election, or not.The base isall those on the electoralregister.Reported urnoutconsistentlyoverestimatesactualturnoutor validated

turnout. Swaddle and Heath (1989: 539) suggest that four main reasons account for this discrepancy:

misreportingby survey respondents,responsebias, failure to trace all movers, and redundancy n the

electoralregister.However,Heath andTaylor(1999) show a fairlyconstantrelationshipbetween official

turnoutandreported urnoutacross the years,with the latter racking he formeralbeit at a higher evel. If

the relationshipbetween the two measuresis constant, then this will affect the constant term in the

regressionsbut will not affect the parameter stimatesof interest.12 Electoralcohortsare definedby the first GeneralElection that an individualwas eligible to vote in and

the aggregateturnout n this election is computedfor each individual n that cohort. The rangeof valuesfor electoralcohort are centred on zero. People bornbefore 1904 (who were thus eligible to vote beforefull franchise was implemented)score zero. The dummyterm for Franchise hus tells us whetherpeoplebornbefore full franchise was implemented urnoutdifferentlyfrom those bornafter,and the interaction

with sex tells us whetherthis was different for women (who did not have the vote) and men.

13 Partyidentification s measuredusing the standard our point scale (0 = None; 1 = Not very strong;2 = Fairly strong;3 = Very strong).

ISpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 11: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 11/25

502 Polit Behav (2007) 29:493-516

example the age distributionof the electorate,influences turnoutchange. When we

control for age we hold the age distributionof the populationconstant over time. If

year effects then become largerwhen we control for age we can conclude thatthe

change in the distribution of the variable in question had a positive impact on

turnout evels, and similarly, if year effects become smaller we can conclude that

the variable had a negative effect. The addition of factors by blocs thus adds adynamic component to the model that shows how variation in the independentvariables over time influences variation n levels of turnout.

Results

Table 1 shows the parameter stimatesfor five models. Since the sample size is so

large it is appropriateo use robuststandard rrors n the analyses. The first model

contains only the dummy terms (or unexplainedperiod effects) for each election,with 1964 set as the reference category. The second model also includes age, sex

(1 = Male, 0 = Female), class (coded 1 for working class and 0 for other) and

housing tenure (code 1 for home owner and 0 for renter/other)as explanatory

variables,the third model also includes franchise(coded 1 for came of age before

full franchise,0 for post-fullfranchise)and franchise nteractedwith sex, the fourth

model also includes electoral cohort and the fifth model includes strengthof partyidentification.

From Model 5 we can see thatholding everythingelse constant,the log odds of

voting increase with each year of age (b = 0.02), are significantly lower for the

workingclass than the middle class (b = -0.32), and significantlyhigherfor home

owners than for renters (b = 0.49). There are also clear generation effects. The

franchise term and franchise interactedwith sex term are both highly significant,

althoughthe sex term by itself is not. This implies that women who came of agebefore full franchise was extended were much less likely to vote in subsequentelections than women who came of age with equal rights(b = -0.86), whereas men

were unaffectedby the extension of the franchise to women (b = -0.86 + 0.88 =

0.02).The second measure of

generational change,electoral

cohort,is also

significant.In line with Franklin'sthesis, people who came of age in high turnout

elections are also morelikely to vote in subsequentelections thanpeople socialised

in low turnoutelections, althoughthe magnitudeof the effect is small (b = 0.01).

Finally,we can see that the log odds of voting are muchhigherfor those with a very

strong partyidentification hanfor those with no identificationat all (b = 0.64).

However, in orderto understand he determinants f turnoutchangeit is of much

more interestto examine the impactthatcontrollingfor each of these variableshas

on the magnitudeof the coefficientsfor each of the election yeardummyterms. The

parameter stimates from Model 1 show thatreported urnoutwas significantlyandsubstantially ower in each of the subsequentelections after 1964. Forexample, the

log odds of someonevoting in the 2005 election were much lower than the log odds

of someone voting in the 1964 election (-0.75). However, when we control for the

compositionof age, class and tenure(that s whenwe hold the age distribution f the

electorate etc. constantover time) Model 2 shows that the predicted log odds of

SSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 12: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 12/25

Table 1 Reportedturnout over time and the impact of social change, 1964-2005, log odds (with robust standard

Model 1: year Model 2: +demogs Model 3: +franchise Model

b Robust SE b Robust SE b Robust SE b

1964

1966 -0.45*** 0.00 -0.45*** 0.01 -0.47*** 0.01 -0.47*

1970 -0.59*** 0.00 -0.57*** 0.00 -0.60*** 0.01 -0.59*

1974 -0.10*** 0.00 -0.11*** 0.00 -0.16*** 0.02 -0.16*

1975 -0.32*** 0.00 -0.32*** 0.00 -0.38*** 0.02 -0.37*

1979 -0.29*** 0.00 -0.30*** 0.01 -0.38*** 0.03 -0.37*

1983 -0.47*** 0.00 -0.54*** 0.01 -0.62*** 0.03 -0.61*

1987 -0.25*** 0.00 -0.36*** 0.01 -0.45*** 0.03 -0.44*

1992 -0.17*** 0.00 -0.26*** 0.01 -036*** 0.04 -0.35*

1997 -0.45*** 0.00 -0.63*** 0.02 -0.74*** 0.04 -0.72*2001 -0.81*** 0.00 -1.07*** 0.03 -1.19*** 0.04 -1.16*

2005 -0.75*** 0.00 -1.03*** 0.03 -1.15*** 0.04 -1.12*

Age 0.02*** 0.00 0.02*** 0.00 0.02**

Working class -0.29*** 0.02 -0.29*** 0.02 -0.29*

Male 0.01 0.03 -0.03 0.03 -0.03

Tenure 0.46*** 0.06 0.46*** 0.06 0.45**

Franchise -0.89*** 0.15 -0.87*

Franchiseby male 0.83*** 0.12 0.83**

Electoral cohort 0.01**

Party ID

Constant 2.08 0.00 1.14 0.10 1.15 0.10 1.17

aQ

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 13: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 13/25

Table 1 continued

Model 1: year Model 2: +demogs Model 3: +franchise Model 4

b Robust SE b Robust SE b Robust SE b

z2222

(df=

11)958

(df=

15) 1,032 (df = 17) 1,041Nagelkerke R2 .01 .05 .06 .06

(N = 31,188)

***Significant at p = .001

**Significant at p = .01

*Significant at p = .05

CD

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 14: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 14/25

Polit Behav (2007) 29:493-516 505

voting in 2005 are even lower thanthe log odds that we initially observed (-1.03

compared o -0.75). Controlling or franchise(Model 3) also has a similareffect; in

each case the magnitude of the coefficient for the dummy terms since 1964

increases.This shows thatchangesin the distributionof these social variableswould

have had a positive impact on levels of turnoutceteris paribus, and that if the

present day electorate had the same social composition as the electorate in 1964thenturnout n the 2000s would have been considerably ower than it actuallywas.

Controllingfor electoral cohort (model 4) has practicallyno impacton the yeareffects: hardlyany of the dummytermschange in magnitude (althoughwhere theydo change it is in the expected direction).This suggests that Franklin'sfootprint

hypotheses,as operationalisedhere, is not well equippedto explain turnoutchange,

although it has merit in other respects. Somewhat contraryto received wisdom,

generationalchanges have on aggregatethereforehad a positive impacton turnout.The replacement n the electorateof women politically socialised before 1928, who

were much less likely to vote thanmen, with women socialised after 1928, who had

more or less the same likelihood of voting as men means that ratherthanlosing a

civic generationwe have in fact gained a fully franchisedgeneration.Since these social changes would have had a positive impact on turnout,to

explainthe generaldecline it is thusnecessaryto search for explanationselsewhere.

To do this I first consider the impact of partisan dealignment. Controlling for

partisanshipmodel 5) has a substantial mpacton the magnitudeof the coefficients

for each of the different election dummy terms. From Model 4 we can see that,without

controllingfor

partisanship, eported urnoutwas

significantlyower thanin

1964 in each of the subsequentelections. However, when partisanship s added to

the model, reportedturnout s no longer significantlylower in 1974 or 1992 and,moreimportantly, he magnitudeof the dummyterms aregreatlyreducedacrossthe

board.Forexample,the magnitudeof the coefficientfor 2005 decreases from -1.12

in model 4 to -0.72 in model 5, when partyID is added to the equation.It is possible to get a clearer idea of the magnitude of all these changes by

transforming he log odds into probabilities,which are slightly easier to interpret.14

14This is done using the following function:Pv= (elogxo+logxi)/(1 + elogxo+logxi), wherelog xo refersto

the constant(1964) term and log xi refers to the dummyterm for a particularelection year. Thus, the

probability of voting in 2005 (calculated from the log odds presented in Model 1 in Table 1) is

e2.08-0.75/1 e2.08-0.75 - 3.79/4.79 = 0.79. Or, put another way reported turnout among survey

respondentswas 79% (Table 2). This correspondsto our observed known value from the surveyestimates. We can then calculate the level of turnout n each year assumingthat variousdemographicvariableshadstayedthe same since 1964by comparing he magnitudeof the coefficient for each dummyterm afterwe control for age, sex, class and tenureetc. (Model 2) to the constant1964 term in Model 1.

This implicitlyholds demographicvariablesat the samedistributionas in 1964 and allows us to see what

turnoutwould have looked like in subsequentyears had therebeen no change. Thus, the probabilityof

voting in 2005, controlling for demographics(calculatedfrom the log odds presentedin Model 2 in

Table 1) is e2.08-1.03/ + e2.08-1.03= 2.83/3.83 = 0.74. Or, put anotherway, had the age, class and

tenureetc. distributionof thepopulationbeen the samein 2005 as it was in 1964 reported urnoutamongsurvey respondentswould have been 74%,or 5 percentagepoints lower than what we actuallyobserved.

These calculations are repeatedfor each block wise step.

6LSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 15: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 15/25

506 PolitBehav 2007)29:493-516

Table 2 presentsthe full results. We can see thatthe observedprobabilityof votingin 2005 is 0.79 (Model 1). Controllingfor demographics(Model 2) reduces the

predictedprobabilityof voting in 2005 to 0.74, controllingfor franchise(Model 3)reducesit even further o 0.72, controllingfor electoral cohort(Model4) leaves the

predicted probability of voting unchanged on 0.72. From this we can see that

holding demographic changes in the social and generational composition of theelectorate constantovertime, the predicted evel of reported urnout n 2005 (Model

4) is some 7 percentage points lower than the observed level of 79% (Model 1).Fromthis we can inferthat,otherthingsbeing equal, social changeboosted turnout

by about 7 percentagepoints,or that the 'real' decline in levels of turnout s in fact

7 percentagepoints greaterthanwhat it looks like on the surfacewhen we fail to

take into account social change.

Controlling for party ID (Model 5) greatly reduces the magnitude of this

decline. In real terms without controllingfor party identificationreportedturnout

fell from 89% in 1964 to 72% in 2005-a decline of 17 percentage points;however controlling for party ID reportedturnoutin 2005 was 80%-a decline

of just 9 percentage points from 1964. This finding contrasts with Franklin's

(2004: 164) claim that party ID does not generate high turnout.However, party

dealignmentis clearly not the whole story. Even controllingfor party ID turnout

was still substantiallylower in 2005 than it was in 1964, so there are clearlyadditional sources of turnoutchange that we need to consider. But the generaltrend of decline is greatly reduced: plotting predicted turnout from model 4

against yearreveals the linear trend of decline since 1964 in 'real terms' is

-0.30 percentage points per year (with an R2 of .61), but when we take partyidentificationinto account the average rate of decline falls to just -0.13 points

per year (with an R2 Ofjust .25). This suggests firstly, that partisan dealignmentaccounts fairly well for any secular decline, and secondly, that the remainingsources of variation probably do not follow a linear function since the R2 is

extremely low.

Short-term Factors: Political Context and Turnout

The above analysis suggests that,if party dentificationhad not decreased n Britain

over the last 40 yearsthenit is doubtfulwhetherthere would have been any generaldecline in levels of turnout.However, this does not mean that turnoutwould not

have variedfrom election to election. Previous researchhas shownthatthe political

context-particularly the expected closeness of the race-influences turnout evels

(Rosenstone and Hansen 1993; Pattie and Johnston 2001; Clarke et al. 2004;Franklin2004). Butbecause thepoliticalcontextchangesfrom electionto election it

cannot be expected to explain any general decline (cf. Rosenstone and Hansen1993). However, it can account for residual variationand explain why turnoutwas

particularlyhigh in some years and particularly ow in others.

Figure 2 shows the perceived ideological difference between the parties over

time (2 = A great deal of difference; 1 = Some difference; 0 = Not much

SSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 16: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 16/25

Polit Behav (2007) 29:493-516 507

Table 2 Predictedreportedturnoutover time andthe impactof social change, 1964-2005,

probabilities

Model Model 2 Model 3 Model 4 Model 5

1 year +demogs +franchise +electoral +partyID

cohort

1964 89 89 89 89 89

1966 84 84 83 83 83

1970 82 82 82 82 82

1974 88 88 87 87 89

1975 85 85 85 85 86

1979 86 86 85 85 88

1983 83 82 81 81 85

1987 86 85 84 84 88

1992 87 86 85 85 88

1997 84 81 79 80 84

2001 78 73 71 71 79

2005 79 74 72 72 80

difference).'5 The differences between the Conservative and Labour party were

perceived by the electorate to be at their greatest during the 1980s, and at their

narrowestduring he 1970s andin the late 1990s/early2000s.'6 Even thoughthereis

a considerableamount of individuallevel variation, his does not mean thatpeoplewho think that there is a great deal of difference between the parties in 2001, for

example are somehow wrong. Rather,responses to the question are likely to bedeterminedby what the respondentviews as the salient issue, which will varyfrom

person to person. In this respect the perceived ideological difference is a more

useful measurethana more 'objective' measure based on manifesto content. From

Fig. 2 we can see that the perceivedideological difference between the partieshas a

non-linear association with time. This means there is no evidence of either a

uniformincrease or decrease since the 1960s.

In similarfashion,the expected closeness of the contest varies from one election

to thenext, and theexpectedgap atT1 is a very poorpredictorof the expectedgapat

T2. For example the contest in 1992 was expected to be very close (the meandifference in the polls was in fact 0.4 percentagepoints) whereasin 1997 the polls

put Labour 16 percentage points ahead (Fig. 3).Of course, not everyone may react to these stimuli in the same way. From

analysis in the firstpartof the papertherewas some evidence thatthe age variable

15 In the 2001 and 2005 surveythe questionon the perceived ideological difference was asked in the

mailbackpartof the survey.Non responsesare set to the mean for the year in question.16 This fits well with conventionalaccountsof the difference between the parties.In the 1970s andlate

1990sHaroldWilson andTony Blairrespectively,steered he LabourPartyawayfromthe Left andcloser

to the Centreground, n bothcases reducing he gap between the two mainparties.By contrast, he 1980s

was characterisedby sharpconflictbetweenMrs.Thatcher'sbrandof neo-conservative iberalismon the

one hand,and Labour'smilitant eft on theother.This, combinedwith the narrowwidth of theconfidence

intervals,suggeststhat on the whole the electorateunderstood airlywell the ideological platformsof thetwo majorpartiesand how they differed from each other, and that there was not much divergence in

opinion.

IL Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 17: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 17/25

508 Polit Behav (2007) 29:493-516

1.80 -

00 1.60 -

C 1.40 t

O 1.20

I I1970 1980 1990 2000

Fig. 2 Partydifferential:perceived deological differencebetween ConservativeandLabour,1964-2001

(0 = not much difference, 1 = some difference,2 = a lot of difference),N = 32,085

and party ID behave differently in different electoral contexts. Since party

identification structures ndividual behaviour and makes it relatively stable andpredictable,it is possible that short-termfactors may have a stronger impact on

those with weaker identificationsince their decision to vote is less ingrained.We

would thereforeexpect the impact of partyID to be greaterin electoral contests

where there is less at stake. To test thishypothesisit is necessaryto fit aninteraction

term between partyidentificationand the perceiveddifference between the parties.

Similarly,Franklin uggests thatyoung people are also morelikely to be influenced

by the politicalcontextthanold people who have become stuck in theirways, so this

interaction erms is also included as a control.There areproblemsof identification f

the model contains dummy terms for all the election years and aggregate levelelection variables since period effects are largely the result of the characterof the

election variables.The finalmodel thereforeonly includesthe dummytermsfor the

years that are significantafter all the otherindependentvariables arecontrolled for

(these are 1966, 1970, 2001 and2005). The equationfor the finalregressionmodel

can be expressedas:

Pr(Voting)=fn(Bo + B1Age + B2Class+ B3Male+ B4Tenure

+ B5Electoral ohort+ B6Franchise+ B7Franchise Male

+ B8Party D+ B9Ideologicaldifference+ B10Expectedpolldifference

+ BI Previouspolldifference+ B12 deologicaldifference Party D

+ B13 deologicaldifference* NewVoter)

(2)

l Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 18: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 18/25

Polit Behav (2007) 29:493-516 509

20.00 -

a) * *

S15.00-

()

0

V)a) *t

S10.00

1,)

"00

t'. 5.00 -

xw, *

O*

O*

0.00-

1970 1980 1990 2000

Fig. 3 Expectedcloseness of the race: meangapbetweenConservativeandLabourparty n finalopinion

polls 1964-2005. Source: 1965-1997 from Heath andTaylor(1999: 178);2001, 2005 fromall published

polls (http://www.mori.com/polls/trends/voting-allpub-trends.shtml)

whereBo

is theconstant,

B1-B8

are the individual evelparameters

romequation1,and B9-B13are the parametersassociatedwith the political context.

Table 3 presents the results from the final model. Overall the model performs

quite well-particularly for the purpose of explaining change over time. The

dummy terms for the elections in 1966, 1970, 2001 and 2005 are still significant,

meaning that there are still some unexplained period effects that the model is not

able to account for. Althoughthis should be treatedwith caution,it is also possiblethat these additionalsources of variationmay be somewhatparticularisticn nature,

meaningthatthey cannoteasily be modelled. Forexample, in 1970 the election was

held duringthe summerholidays when many people were away. However, the factthat the final model managesto reduce the magnitudeof these coefficients,and none

of the otherdummytermsare significantmeans that the model is fairly well able to

capture the dynamics of turnoutchange in Britain over the last 40 years. For

example, in 2001 the coefficient for the 'real' decline (when only demographicsare

included in the model) is -0.725. This is reduced is to -0.584 when partyidentification is included, and -0.300 when the political context variables are

included.

From Table 3 the firstthingto notice is thatthe political context variableshave a

significantimpact on turnout.The log odds of voting increase when there is a bigideological difference between theparties(b = 0.33) anddecreasewhen the election

is expected to be a landslide (b = -0.02). The closeness of the race from the

previous election is also significant, suggesting that the decision to vote is

influencedby bothretrospectiveandprospectiveperceptionsaboutthe closeness of

the race. The interactionterm between party ID and the perceived ideological

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 19: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 19/25

510 Polit Behav (2007) 29:493-516

Table 3 Reported urnoutandthe impactof the politicalcontext, log odds (withrobust standard rrors)

b RobustSE

1966 -0.23*** 0.02

1970 -0.41*** 0.03

2001 -0.30*** 0.052005 -0.46*** 0.05

Age 0.01*** 0.00

Workingclass -0.30*** 0.03

Male -0.05 0.03

Tenure 0.47*** 0.07

Electoral cohort 0.01*** 0.00

Franchise -0.79*** 0.14

Franchiseby male 0.88*** 0.14

PartyID 0.68*** 0.05

Partydifferential 0.33*** 0.07

Expectedcloseness of race -0.02*** 0.00

Previouscloseness of race -0.01* 0.00

Partydifferentialby partyID -0.06* 0.03

Partydifferentialby new voter -0.10** 0.03

Constant 0.02 0.13

z22,500 (df = 17)

NagelkerkeR2 .132N= 31,188

***Significantat p = .001

**Significantatp = .01

*Significantatp = .05

distance betweenthe partiesis significantandnegative,meaningthatthe difference

between the parties is a more important predictorof reportedturnoutfor non-

identifiers hanit is forstrong

identifiers.17imilarly

the interactioneffect between

new voters andthe difference betweenthe parties s significant,meaningthatyoung

people are more influenced by the political context than old people (see also

Franklin2004). The other individuallevel variablesare much the same as before

and so need no furtherdiscussion here.

We can get a clearer idea of how these factorsinfluence turnout evels by usingthe regression equationto predict turnout n each year (Table 4).18Replacing the

sample means with the corresponding1964 means for demographicvariables,we

can see that,for example, predicted urnout n 2005 would have been 6.0 percentage

17 The interactionbetweenparty dentificationandtheexpectedcloseness of the contest is not significantbut this result is not surprisinggiven the data constraintsof having only 12 level 2 datapoints.

18 On the whole the predictionsslightly over estimatethe observedlevel of reported urnout by about

2 percentage points), but otherwise they capture the dynamics of turnoutchange over time fairly

accurately.From Table 4 we can see thatpredictedturnout n 1964 was 90.3%andpredictedturnout n

2005 was 81.7%,(a decline of 8.6 percentagepointswhich is not far off the 10 point observeddecline).

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 20: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 20/25

PolitBehav 2007)29:493-516 511

Table4 Summarympact f theindependentariablesnturnouthangeoverime,percentages

Year Observed Predicted Impactof social Impactof party Impactof political

reported reported change on turnout ID on turnout context on turnout

turnout turnout (% points) (% points) (% points)

1964 88.9 90.3 0.0 0.0 0.0

1966 83.6 85.7 0.1 -0.2 -2.4

1970 81.5 84.4 0.3 -0.5 -1.5

1974 87.9 88.9 0.7 -1.4 -0.8

1975 85.3 87.8 0.8 -1.5 -1.8

1979 85.6 87.9 1.3 -2.7 -1.0

1983 83.3 85.3 2.0 -3.7 -3.4

1987 86.1 88.0 2.3 -3.4 -1.2

1992 87.1 89.9 2.1 -2.9 0.4

1997 83.6 86.6 3.6 -4.0 -3.4

2001 78.0 80.6 5.9 -6.9 -5.8

2005 79.1 81.7 6.0 -6.7 -3.2

points lower thanit actuallywas (81.7 - 75.6 = 6.0), meaningthat,all otherthings

being equal, social change has had a positive impact of 6.0 percentagepoints on

turnout levels over this time period. By contrastpartisan dealignmenthas had a

negativeimpactof 6.7 percentagepointsandthepoliticalcontexthas hada negative

impactof 3.2 percentagepoints. Whereasthe impactof social change and partisandealignmenthas been more or less cumulative over the last 40 years, the impactof

the political context varies much more from election to election. This suggests that

partisandealignmentis responsiblefor the seculardecline in turnoutover the last

40 years and the political context is responsible for the short-termvariation. The

secular decline has therefore been somewhat less than the overall decline. For

example, in 2001 when turnoutfell off an apparentcliff, we can attributeabout

6.9 percentagepoints of the drop to long-term secular decline and 5.8 percentage

points to short-term variation. Bearing in mind that the baseline 1964 election

witnessed a fairly tightrace with clear ideological waterbetweenthe parties,we cansee that the elections that deviated most from the long-term patternof generaldecline since then (1966, 1979, 1983, 1997, 2001, 2005) are the ones that on

average had the least competitive electoral contests with the least difference

between the parties(as illustratedby the high impact of the political context).Variation in levels of turnout over time can thus be attributed o two distinct

processes.The firstis compositionaland refers to the generaldecline in the strengthof party identificationin the electorate since 1964. The second is contextual and

refers to the natureof the political contest.This varies fromyear to yearand is able

to accountfor muchof the residualvariation n levels of turnout rom the long-termtrend of generaldecline. Whereas to a certain extent partyidentificationstructures

individual behaviour and makes it relatively stable and predictable,the politicalcontext is muchmore volatile. In some sense then,these two determinants f turnout

SSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 21: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 21/25

512 PolitBehav 2007)29:493-516

pull at each otherin oppositedirections,and the decline of the firstsuggests thatthe

latter is thereforebecoming more important.This is supportedby the significanceof the interaction ermbetweenpartyID and

ideological difference. Whereas for strong identifiers the difference between the

parties hardlymakes any difference to their probabilityof voting (0.94 if a good

deal of difference;0.92 if not muchdifference),for non-identifiers he probabilityofvoting is greaterwhen there is a good deal of difference between the partiesthan

when there is not muchdifference(0.76 compared o 0.60). Since non-identifiersare

more influenced than strong identifiersby short-term actors, and since there are

now more non-identifiers hanpreviously,de facto short-termactors areassuminga

greaterrole in determining urnout evels thanwas the case in the 1960s.This canbe

illustratedby imputingthe sample means for partyID from each election into the

regression equationand varying the level of perceived partydifference between a

hypotheticallow of 0 and a hypothetical high of 2, while holding all other values

constant (at their 1964 levels). Varying the ideological difference between the

parties has only a slight impact on the predictedlevel of turnout n 1964, with a

predictedreported urnoutof 88.3%when there is no differencebetween the partiesand 91.5% when there is a substantialdifferencebetween the parties,giving a rangeof 3.2 percentage points (Table 5). In 2005 this range is somewhat larger at

7.1 percentagepoints.

Conclusion

Previous researchhas tended to examine turnoutchange as either a function of

individuals or as a functionof elections. Both these approachesappearto be miss-

specified. Although Franklin(2004) makes an importantcontributionby drawingattentionaway from individualisticaccountsof turnout o highlightthe role played

by the electoral context, he goes too far in the opposite direction, stating that

"turnout s an aggregate-levelconcept" (ibid 167). Althoughturnout s influenced

by the political context, its impact is mediatedby how individualsrespond to it.

Different individualsrespond

in differentways,

so thecomposition

of different

groupsis important or understandinghe impactof the election. By modelling the

joint impact on turnout of individual factors and macro factors relating to the

structureof the nationalpolitical contest, the paperexamines how the dynamicsof

voting are changing over time. The political context has a significant impact on

individual's willingness to turnout to vote, even when we control for individual

level attributes.However, individualsrespondto politicalcues at the macro level in

different ways, and to understandwhy people vote (or not) we therefore need to

understandhow theirbehaviourrelatesto the widerpolitical context. In this respect

turnoutcannotbe fully understoodpurely with reference to individuals'attributes,but neither can it be understoodpurely with referenceto the political context.

To explain turnoutchange it is necessaryto firstknow what one is attempting o

explain. This paper arguesthat there are two distinctphenomenaat work. The first

relates to partisan alignment (or the strengthof party identification).In countries

where this is high (such as in Malta) turnout s likely to be high and will not vary

SSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 22: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 22/25

Polit Behav (2007) 29:493-516 513

Table 5 Predictedreported

turnoutand the changing impact

of partydifferential,1964-2005,

percentages

Year When party When party Rangedifferential= 0 differential= 2

1964 88.3 91.5 3.2

1966 85.4 89.4 3.9

1970 82.7 87.3 4.61974 86.6 90.5 3.8

1975 86.6 90.4 3.9

1979 84.9 89.4 4.5

1983 84.2 88.9 4.8

1987 83.9 88.8 4.9

1992 84.2 89.0 4.8

1997 84.2 89.0 4.8

2001 78.1 84.6 6.5

2005 75.4 82.5 7.1

much from election to election. In countries where it is lower but stable (such as in

Denmark)turnout s likely to be correspondingly ower and will vary accordingto

the political context (both up and down but no general decline). However, in

countries where party identification is declining (such as in Britain, the United

States and many others) turnout will decline over the long-term, and become

increasingly nfluencedby the structureof short-term orces relatingto the politicalcontext.

There is no doubt that this process has significant mplicationsfor representative

democracy. Party-basedrepresentativedemocracy implies that political represen-tatives articulateand express the political interests and preferencesof the peoplewho have elected them (Key 1964; Dalton 2000). In this light, the strongerthe

alignmentbetweenthe public andparties,the greaterclaim partieshave to speakon

the behalf of theirelectors. However, as Dalton and Wattenberg 2000: 278) state,whereas "partiesused to ensure that elections were judgements on broad policychoices thatwould guide governmentaction-this function has become less clear".

There are not only fewer people voting than previously, but those people who do

vote are now more likely to belong to the growingband of non-identifiers or weak

identifiers) hanstrongidentifiers.For these people the act of voting is not so much

of a judgmenton broadpolicy platforms,but a response to more narrow concerns

related to short-term actors associatedwith the political context. Thus the politicalcontext is becoming more important.This also helps to provide a context for

understandingwhy short-termfactors such as the 'personalisationof politics',

leadershipeffects andcampaigneffects appear o be gaining importance n terms ofexplaining vote choice and the decision to vote. When partyID is high turnout s

high and so individual level variables will not differentiateeffectively between

those who vote and those who don't (since almost everyone will vote). However,when partyID is low turnout s correspondingly ower and there is more room for

other factors to come into play.

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 23: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 23/25

514 Polit Behav (2007) 29:493-516

The paperthus sheds light on why people participate,how electoralparticipationrelates to the political processandin some sense, whatparticipation ctuallymeans,

andwhatcan be inferred andwhatcannot)from bothhigh levels of turnoutand low

levels of turnout.I show that individuals need both resources and motivation to

participate n politics. People have the resourcesto participatemore so now than

ever before, but this in itself is not enough to ensure high levels of electoralparticipation.Turnouthas not declined because it has become more difficult to vote,

but because the attachments between political parties and the electorate have

become much weaker over time. If this is not re-energisedthen baseline levels of

turnoutwill continue to decline, no matter how easy it is to vote or how close an

election is expected to be.

Acknowledgements I would like to thank David Sanders, Paul Whiteley, Anthony Heath, David

Denver,JohnBartle,MichaelLewis-Beck, JouniKuha and threeanonymousreviewers for theirhelpfulcommentsandsuggestions.I am also gratefulto the ESRC(PTA-026-27-0486)and the BritishAcademyfor researchfunding.

Appendix

BritishGeneralElection Study, 2005

The 2005 BritishGeneralElection Study was fundedby the ESRC. The studywas

carried outby University

of Essex andUniversity

of Dallas at Texas. Thesamplewas drawn from the Postcode Address file. The fieldwork was carried out by

National Centrefor Social Research. The surveyconsisted of two waves. The first

wave took place before the election. It had a response rate of 61% and achieved

3,589 interviews. The second wave took place immediatelyafter the election. It had

a response rate of 88% (of Wave 1) and achieved 2,959 interviews. In addition a

top-upsurveywas also carriedout which had a responserate of 54% and achieved

1,202 interviews. The data used in the papercomes from the post election panel,which has a valid samplesize of 3,431 respondentson the electoralregister.Further

details are available at the BES website (http://www.essex.ac.uk/bes).

British General Election Study, 2001

The 2001 British GeneralElection Studywas fundedby the ESRC. The studywas

carried out by Universityof Essex and Universityof Dallas at Texas. The samplewas drawnfrom the Postcode Addressfile. The fieldworkwas carriedout by NOP.

The surveyconsisted of two waves. The firstwave took place before the election. It

had a responserate of 53% and achieved 3,220 interviews. The second wave tookplace immediatelyafter the election. It had a responserate of 74% (of Wave 1) and

achieved 2,359 interviews. In additiona top-upsurvey was also carriedout which

had a responserate of 47% andachieved 681 interviews.The data used in the paper

comes from the post election panel, which has a valid sample size of 2,081

SSpringer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 24: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 24/25

Polit Behav (2007) 29:493-516 515

respondentson the electoral register. Furtherdetails are available in Clarke et al.

(2004: 329-336).

BritishGeneralElection Study, 1997

The 1997 British GeneralElection Studywas fundedby the ESRC and the GatsbyCharitableFoundation,and the Commission for Racial Equality. The study was

carriedout by CREST. The samplewas drawn from the Postcode Addressfile. The

fieldwork was carried out by Social and CommunityPlanningResearchin May to

June 1997. The issued sample was 6,540, of which 5,814 were eligible addresses.

There were 3,615 interviews giving a responserate of 63%. Furtherdetails of the

survey are given in Heathet al. (2001: 170).

BritishElections, 1963-1992

The combinedBritishGeneral Election studydataset,1963-1992 was compiled bythe Data Archive, University of Essex. The original studies were carriedout byNuffield College, 1963-1970, University of Essex, 1974-1979, and Nuffield

College, 1983-1992. The samples were drawn from the Electoral Register. The

achievedsamplesizes were 2,009 (1963); 1,769 (1964); 1,874 (1966); 1,843 (1970);

2,462 (1974F); 2,365 (19740); 1,893 (1979); 3,955 (1983);3826

(1987); 3534(1992). Furtherdetails are available in Heath et al. (1994: 302-308).

References

Abramson,Paul,& Aldrich,John(1981). The decline of electoralparticipationn America. TheAmerican

Political Science Review, 76, 502-521.

Achen,Christopher2002). Parental ocialization andrationalparty dentification.Political Behavior, 24,151-170.

Brians, Craig (1997). Residential mobility, voter registration,and electoral participation n Canada.Political Research Quarterly,50, 215-227.

Budge, Ian, Klingemann,Hans-Dieter, Volkens, Andrea, Bara, Judith, & Tanenbaum,Eric (2001).

Mapping policy preferences: Estimatesfor parties, electors, and governments.Oxford: Oxford

UniversityPress.

Butler,David, & Stokes, Donald (1969). Political change in Britain.London: Macmillan.

Campbell,Angus,Converse,Philip,Miller,Warren,& Stokes,Donald(1960). TheAmericanvoter.New

York: JohnWiley & Sons.

Campbell,Angus,Converse,Philip,Miller, Warren,& Stokes,Donald(1966). Elections and thepoliticalorder. New York: JohnWiley & Sons.

Clarke, Harold,Sanders,David, Stewart,Marianne,& Whiteley,Paul(2004). Political choice in Britain.

Oxford: OxfordUniversity

Press.

Converse,Philip (1969). Of time and PartisanStability.ComparativePolitical Studies,2, 139-172.

Converse, Philip (1976). The Dynamics of Party Support: Cohort-AnalyzingParty Identification.

Beverley Hills, CA: Sage Publications.

Crewe, Ivor,Sarlvik, Bo, & Alt, James(1977). Partisandealignment n Britain 1964-74. British Journal

of Political Science, 7, 129-190.

L Springer

This content downloaded from 158.109.174.204 on Wed, 30 Oct 2013 16:42:27 PMAll use subject to JSTOR Terms and Conditions

Page 25: Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

7/27/2019 Explaining Turnout Decline in Britain, 1964-2005; Party Identification and the Political Context.pdf

http://slidepdf.com/reader/full/explaining-turnout-decline-in-britain-1964-2005-party-identification-and 25/25

516 Polit Behav (2007) 29:493-516

Dalton, Russell J. (2000). The decline of party identification.In Russell J. Dalton, & Martin P.

Wattenberg Eds.),Parties withoutpartisans:Political change in advanced ndustrialdemocracies.

Oxford:OxfordUniversityPress.

Dalton,RussellJ., & Wattenberg,MartinP. (Eds.). (2000). Parties withoutpartisans:Political change in

advancedindustrialdemocracies. Oxford: OxfordUniversityPress.

Denver, David, & Hands, Gordon(1974). Marginalityand turnout n Britishgeneral elections. British

Journalof Political Science, 4, 17-37.

Denver, David, & Hands, Gordon(1985). Marginalityand turnout n British general elections in the

1970s. British Journalof Political Science, 15, 381-388.

Downs, Anthony(1957). An economic theory of democracy.New York:Harperand Row.

Fiorina,Morris(1981). Retrospectivevotingin Americannational elections. Newhaven: Yale UniversityPress.

Franklin,Mark N. (2004). Voter turnout and the dynamics of electoral competition in established

democracies since 1945. Cambridge:CambridgeUniversityPress.

Heath,Anthony,Jowell, Roger, & Curtice,John(1985). How Britain votes. Oxford:PergamonPress.

Heath, Anthony, Jowell, Roger, & Curtice, John (2001). The rise of new labour. Oxford: Oxford

UniversityPress.

Heath,Anthony,Jowell,Roger,Curtice,John,& Taylor,Bridget(1994). Labour's ast chance? The1992

election and beyond.Aldershot:Dartmouth.

Heath, Anthony, & Taylor, Bridget (1999). New sources of abstention?In Geoffrey Evans, & PippaNorris (Eds.), Critical elections Britishparties and voters in long-termperspective.London:Sage.

Key, Valdimer Orlando(1964). Politics, Parties and PressureGroups.New York:Crowell

Miller,Warren 1992). The puzzle transformed:Explaining decliningturnout.Political Behavior,14, 1-

43.

Pattie, Charles,& Johnston,Ron (1998). Voter turnoutand constituencymarginality:Geographyand

rational choice. Area, 30, 38-48.

Pattie, Charles, & Johnston,Ron (2001). A low turnout andslide: Abstentionand the British generalelection of 1997. Political Studies,49, 286-305.

Rosenstone, Steven J., & Hansen, John Mark (1993). Mobilization,participation,

anddemocracy

in

America.New York: Macmillan.

Stewart,Marianne,& Clarke,Harold 1998). Thedynamicsof party dentificationn federalsystems:The

Canadiancase. AmericanJournalof Political Science, 57, 368-377.

Stokes,Donald(1992). Valencepolitics. In DennisKavanagh Ed.),Electoralpolitics. Oxford:Clarendon

Press.

Swaddle, Kevin,& Heath,Anthony(1989). Officialandreported urnoutn theBritishgeneralelection of

1987. British Journalof Political Science, 19, 537-551.

Verba,Sidney, & Nie, Norman H. (1972). Participationin America.New York:HarperandRow.

Verba, Sidney, Schlozman, Kay, & Brady, Henry (1995). Voice and equality: Civic voluntarismin

Americanpolitics. Cambridge,Mass: HarvardUniversityPress.

Wattenberg,MartinP. (2000). The decline of partyidentification. n Russell J. Daltonand,& Martin P.

Wattenberg Eds.),Parties withoutpartisans:Political change in advanced industrialdemocracies.

Oxford:OxfordUniversityPress.

Wattenberg,MartinP. (2002). Wherehave all the votersgone? Cambridge,Mass: HarvardUniversityPress.

Whiteley, Paul, Clarke, Harold,Sanders,David, Stewart,Marianne(2001). Turnout.In Pippa Norris

(Ed.), Britain votes 2001. Oxford: OxfordUniversityPress.