1 Introduction
Kai Arzheimer
Jocelyn Evans
Michael S. Lewis-Beck
How do we decide how to vote? Since the post-war behavioural revolution,
psephologists have proposed a set of relatively simple theories to account for
the social and psychological processes which result in the nature of the vote
cast. To test the simple theories, a larger set of varyingly complex and
technically sophisticated models have been operationalised to provide a more or
less rigorous empirical validation, or negation. The complexity of models has
derived both from a desire to combine theories, either to generate more complete
accounts of voting behaviour, or to test competing hypotheses, and from a need
to specify such models robustly, to avoid spuriousness and biased
estimations.
Party identification, the pivotal variable to any full model of voting in
post-war democracies, is a case in point. A baseline to any well-specified
voting model in countries where there has been demonstrated validity, the
underlying concept of an affective attachment to a party has remained relatively
untouched as a cornerstone of vote motivation. The principal source of its
development has been debated, the socialâpsychological bases to Michigan's take
on early political socialisation being challenged by the retrospective learning
process posited by Morris Fiorina and other rational choice adepts. This tension
between a socialâpsychological approach and a rational choice approach also
finds expression in the argument over the extent to which party identification
is exogenous or endogenous. Put another way, does party identification have the
characteristics of a stable, long-term enduring attachment, or the
characteristics of a changing, short-term, more unstable predisposition? This
question finds some resolution in the dynamic models of party identification
which enable both theories to consider how
identification evolves over time, information and experience recasting what
early socialisation has imbued.
But all these refinements, re-conceptualisations and re-interpretations
notwithstanding, the core of the argument made by Campbell et al. back in 1960
still stands. Taken together, the chapters in this Handbook demonstrate that the
act of voting grows out of some form of partisanship and sociological status
that anchors the voters in the long term, while they are buffeted about by
short-term forces, such as the pressing issues and leadership appeals of a
campaign. Of course, once in the voting booth, voters themselves must reconcile
these contradictions by making a unique party choice.
Sociological accounts of voting, though at first sight having experienced greater
changes over time, betray a similar stability at the conceptual level. The
binary simplism of the class divide, or the secularâreligious conflict which for
so long dominated many countriesâ LeftâRight content, has given way to more
variegated accounts of occupational class and gradations of religious practice.
At the same time, evidence for the growing influence of other socio-demographic
attributes as explanatory variables rather than the âstandard controlsâ they
were once relegated to â gender, age, ethnicity, amongst others â has been
clearly seen in countries even where class has been most widely (and somewhat
erroneously) written off. Underlying all such behavioural approaches to the
sociology of voting â as opposed to the structuralâfunctional determinism of
cleavage theory which in many ways relegated voters to ballot-wielding pawns to
be pushed along the electoral board â is an assumption that the environmental
context of individuals drives their behaviour and shapes information, attitudes
and preferences. An atomised account of value-neutral information gathering and
processing by a voter qua data miner is absent in any empirically founded
account of voting.
Given the debates which have taken place over the past 30 years regarding the
decline of socialâstructural and socio-demographic accounts of voting, some may
be surprised at the space we give to these individually. Undoubtedly, the role
of class, religion and other social identities has changed in accounts of
voting, and in some senses declined. Where religious identification, and more
importantly religious practice, has declined in absolute terms across most if
not all democratic societies, the amount of variance these factors can account
for in a voting model will likely decline. However, among a greater patchwork of
social identities and group belongings, rather than as one side of a binary
cleavage, religious affiliation can still be a powerful predictor of vote. The
patchwork metaphor works well for social class in the modern era too. The
Marxist dyad or the manual/non-manual split, which suited simple numerical
indices of class voting â but little else â have been replaced by more granular
and only quasi-hierarchical categories of occupational class and sector
belonging which are a far more satisfactory description of social structure by
employment than an a priori politicised frame with a decidedly Procrustean
approach to fit. Social structure remains important even where voting no longer lines up with its categories, because at
the very least it provides the researcher with a benchmark against which to see
both how voting has changed from the demand side and also how parties have
shifted their offer, consciously or otherwise, away from the social
structuration which previously oriented their programmes.
Nevertheless, it remains a difficult argument to make that, even where the direct
causal link of social structure on vote is broken, the more proximate
explanations of vote, such as attitudes towards policies and leaders, derive
from something other than social context. While the disconcertingly random voter
discovered through Converse's belief system paradigm has since been shown to be
somewhat more consistent and ordered than previously assumed, homo
economicus is certainly not a reasonable label for modern voters any
more than it was for their forebears. A greater quantity of political
information is available through traditional, electronic and social media, but
this information is processed through biased filters and prejudiced cognitive
systems, if it is received at all. Modern voters no more sit down and neutrally
construct a political ideology based upon beliefs and context than they apply
this strategy to any other habitual and affective behaviour in their lives.
Context preconditions behaviour through attitudes. If not the context of social
position, gender, ethnicity and identity, then what?
Modern studies of voting have come up with a complementary answer to social
context that previous work had only hinted at more or less anecdotally, by
beginning to operationalise the workings of the human cognitive and affective
system(s) and, even beyond that, the very building blocks of human life â DNA
and genetics â which associate with political choice. Where Wallas described
impulse and instinct, and the non-rational side of politics, biology, psychology
and neuroscience now show us how such unconscious drivers activate in political
behaviour as much as in other aspects of life. For a discipline which spends
much of its time asserting its broad relevance and interest to an external
audience increasingly turned off by the topic's content, political science has
generally proved remarkably resistant to studies that lodge âpolitical
behaviourâ in the âbehaviourâ box. Methodological concerns over causality, the
extrapolation from lying in an MRI machine to standing in front of a voting
machine, and processes which remain largely âblack boxâ can be well taken, if
they acknowledge that similar concerns need to apply as strongly to more
traditional approaches with exactly the same methodological issues. Clinical
science variables may be latecomers to the party, but they are not
gatecrashers1.
Again, however, the challenge comes not to the âheavy variablesâ (les
variables lourdes as French psephologists call them) of social context
and party identification, or the more proximate issue causes, but rather to the
process of interaction between the individual and their context, and how
cognitive and affective processes within that context arrive at a political
choice. The history of voting studies has been at its most productive not from
trying to turn traditional voting accounts on their heads, but rather by better
explaining how those accounts work under
the bonnet. Party identification is measurable as a phenomenon in the United
States. Refinements in understanding the conditions of its appearance, its
mutability, and its manner of influence do not undermine the concept, but rather
reinforce its core role in votersâ decision making. True, the extent to which
strong identifiers have been increasingly offset by âleanersâ demonstrates that
the factors responsible for party identification have changed in their relative
strength and composition. But party remains the key shortcut to vote choice. The
findings of Columbia, indirectly, and Michigan, directly, remain valid seven
decades on.
Continuity of a similar kind will strike readers who cover the majority or all of
the chapters. A handful of key texts return no matter what the content. The
American Voter by the Michigan team and VO Key's The Responsible
Electorate, along with Anthony Downs's Economic Theory of
Democracy and Fiorina's Retrospective Voting in American National
Elections, are four texts whose concepts and approaches have influenced
a substantial proportion of all research into voting. Those readers who are
struck by the rich collection of sociological explanations may well be those
equally struck by the absence of a dedicated rational choice chapter. As a field
which in the past has been characterised as cartelising schools of thought and
journals in a way that many physicists have more recently bemoaned string theory
doing2, why have we apparently bypassed
this theoretical canon? The answer is to be found in the abundance of chapters
where rational choice's influence lies implicit in the assumptions about voter
behaviour. What is implied may not be perfectly informed homo economicus
(not that Downs ever assumed perfect information) with a cognitively processed
cost-benefit analysis resulting in utility maximisation. Rational choice's
eventual influence may well have come instead from the modifications imposed by
boundedness, by psychological insights into low-cost habitual, affective action
and by all manner of shortcuts, heuristics and thought-saving devices that
voters employ. Rational choice's first-principles derived theory was required to
step away from the empiricist induction characteristic of many âexplanationsâ â
pattern spotting and anecdotal in many cases â of vote, to spur more realistic
accounts of the decision calculus employed by the electorate. As the chapters in
this Volume attest, party identification; economic voting; strategic voting;
core ideology; valence; geolocation â all partly or totally rely upon a level of
cognitive evaluation and choice for the theoretical framework to function.
With an emphasis on theoretical refinement and improvements in analysis, rather
than paradigm shift, the technical sophistication of explanatory models has
inevitably increased over the years. We would be mistaken to think that anything
pre-1990 and the advent of PC-based statistical software could only be
rudimentary. Nevertheless, the realms of maximum likelihood, simulations and
Bayes were computationally, if not theoretically, off-limits. As the respective
chapters on cross-national data sources, multi-level modelling and technology
show, psephology has always been at the forefront of technological innovation in
political science and has in turn benefitted hugely from novel ways of
collecting, sharing and analysing data.
The best models employ the technical sophistication to reflect as accurately as
possible the functional form of an outcome, to map involved causal pathways, to
anticipate the role that time plays in variation of effects, to correct bias â
in short, to try to depict a simplified reality as accurately as possible in
quantitative terms. But where technical sophistication begins to outweigh
simplicity as an end, rather than a means, the balance has shifted, and for the
worse.
This may seem like a truism, but where simple models of electoral behaviour still
add immense value is in exploring new inflections to the vote calculus which
better data now provide. For, while we would underline that the principal
explanations of vote remain constant in their relevance, if not in the hierarchy
of their importance, there is still work to be done in exploring minor
influences on voter choice. The personal vote, candidate effects, campaign
effects more generally and even electoral institutions may not always be
game-changers, but rather condition baseline voting propensities, yet their
influence adds to our understanding of what matters to voters beyond the main
causal drivers. This knowledge has helped tremendously in developing a new
branch of electoral studies, that of election forecasting. Of course, the causal
factors enabling successful forecasts of this sort have always been present in
voting research, at least on the margins and often at the heart, for example in
The American Voter. Yet, the dominance of large-n national sample
survey work has in the past diverted the study of candidate-specific, or
campaign-driven, investigations of aggregate-level electoral effects to a
separate stream of research. Advances both in choice modelling and data linking
mean that this stream has now become a river, whose major forks (e.g. political
markets or econometric forecasting) thankfully branch back to the main channel
eventually.
Such refinements have occurred largely in the countries where voting models have
been most prevalent â from the US origins through to mainly European
counterparts. In these countries, commitment to collection of large-scale
election data, and to funding supplementary studies with specific analytical
foci, has ensured that researchers can test across time and space. However, it
is evident that an increasing number of newer democracies provide auspicious
setting for refining and retesting models. In Latin America, Asia and Africa,
rigorous election studies are increasingly present for researchers wishing to
find different contexts in which to test their assumptions. European studies
have a growing set of cases with the inclusion of Central and Eastern European
countries. Whilst comparative datasets across large numbers of countries have
not always provided the highest level of rigour or comparability in the past,
national election studies predicated upon the full model of voting are now
increasingly the norm. Even where such data are still not directly accessible,
or do not exist, developments in datascraping of online sources mean almost no
country is now beyond the scope of some degree of electoral analysis, and from
multiple sub-disciplinary perspectives.
Indeed, one might argue that, roughly a century after its beginnings, psephology
has come full circle: while political science is at its core, psephology is
forever borrowing methods, concepts and data from other disciplines â geography,
statistics, sociology, economics, psychology, and now even neuroscience,
genetics and computer science. Looking back at the highlights of this remarkable
story would have made for an interesting (if slightly nostalgic) book, too. But
instead of a sentimental journey, the authors of this book have aimed for
something much more rewarding: each of the following chapters presents the
current state of knowledge on the respective topic whilst highlighting several
points of departure for new research. Inevitably, the state of the art moves
continually forward and, by the time of publication, steps will have been taken
to refine voting models further, and to apply them to new contexts. Like the
survey data which provide the raw material for many electoral studies, this
collection is a set of snapshots of the status quo, which provide a guide to the
community of electoral research. We hope you enjoy the tour.
Notes
1. One suspects no small part of the problem also comes
from a certain envy at the âinterlopersâ having better toys at their
disposal.
2. It is presumably coincidental that both have been
criticised in their respective fields as being long on theory, short on
empirical testing.
Part I Institutional Approaches
2 Institutions and Voter Choice: Who Chooses, What Do They Choose Over, and How Do They Choose
Shaun Bowler
One of the important sources of variation across democratic systems is the variation in institutional context. That variation in institutional context is important in helping to understand the workings of electoral democracy in a comparative sense. Indeed, in one sense what is being compared in comparative studies of voter beha...