Studies in Public Opinion
ATTITUDES, NONATTITUDES,
MEASUREMENT ERROR, AND CHANGE
Editors
Willem E. Saris
Paul M. Sniderman
PRINCETON UNIVERSITY PRESS
PRINCETON AND OXFORD
CHAPTER 12
A Consistency Theory of Public Opinion
and Political Choice: The Hypothesis
ofMenu Dependence
Paul M. Sniderman and John Bullock
IN THE END WE study the attitudes of citizens to understand the choices
they make as citizens-the candidates they choose to vote for, the public
policies they choose to support. Here we want to draw together some arguments
that have run through this book, together with some that have
been carried on outside it, to outline a general account of political choice.
The spine ofthis account is the concept ofconsistency. As with many social
science terms, the concept of consistency is inconsistently used. In the
context of research on public opinion and political choice, a trio of meanings
can be distinguished. Consistency can be a synonYln for constraint.
Construed as constraint, consistency indexes the predictability of citizens'
position on one issue given their positions on another.1
Then again, consistency
can be a synonym for stability. Construed as stability, consistency
indexes the predictability of citizens' positions on an issue at one point in
time given their positions on the same issue at an earlier point in time. Finally,
consistency can be a synonym for congruence. Construed as congruence,
consistency indexes the predictability of positions citizens take
on specific issues given their general political orientations.2
Empirically, this trio-constraint, stability, congruence-is broadly related.
The more tightly constrained citizens' positions across issues, the
more stable their positions are likely to be over time; and the more stable
and tightly constrained their positions, the more likely they are to be congruent
with underlying basic orientations. The premise of the theory we
present is thus that the first two senses of consistency are causally parasitic
on the third. Positions tend to be constrained across issues or stable over
time to the extent they are congruent with basic political orientations. And
just so far as citizens possess basic political orientations together with the
competence to call them into play, a consistency theory of public opinion
has a causal leg to stand on.
But a consistency theory ofpublic opinion that has just one leg has long
appeared too wobbly to stand, special circumstances aside.3
The dominant
themes of two generations of research have been that citizens tend to be
338 • Sniderman and Bullock
muddle-headed (the lack of constraint theme), empty-headed (the nonattitudes
theme), or both. True, a strong qualifying note also has been
sounded. Citizens can pull their ideas together conditional on political sophistication:
the more of the latter, the more ofthe former.4 But there has
seemed no way to get consistency of choice, defined as congruence, out of
the largest part of the public: they pay too little attention to politics, ~ow
too little about it, and invest too little in organizing their ideas about It.
We therefore want to point to a new conceptual path. It is necessary, we
will suggest, first to take account of the characteristics of choices that.c.itizens
face and then to attend to their characteristics as choosers. In polIncs,
citizens ;haracteristically are presented with an organized set, or menu, of
choices.s The choices they make are dependent on the organization ofthis
menu. Specifically, citizens are in a position to mal<e coherent choices just so
far as this menu is coherent. The distinctive feature ofthe consistency theory
that we present is thus that it stands on two causal legs. It posits that cons~stency,
understood as congruence, is jointly conditional on the ch~r~cterIstics
of citizens as choosers and the menu of options they face as Cinzens.
Admittedly, in proposing any version of a consistency theory, we a~e
malcing an uphill argument. A principal theme of half a century of PUb~IC
opinion research has been precisely the compa~ative absence.of conSIStency
in public opinion. Indeed, under the headIngs of nonatntudes and
ambivalence, the lability ofchoices is a surface motifofmany ofthe chapters
ofthis book. And yet, so far as the studies in this book seek to give a causal
analysis of the public's beliefs and choices that goes beyond a n~nattitudes
story, they are committed to a consistency accoun~ ofsome ~ariety. So, by
way of concluding comlnentary, it seems worthwhIle to consIder what the
premises of such an account most plausibly are. . ..
The most vital premise is this: in representative democracIes Cinzens do
not directly choose the alternatives. They only get to choose from amon~
the alternatives on the menu of choices presented to them. That menu IS
simplified, coordinated, and advocated above all thr.ough ele~~oral c?mpetition
between political parties. Accordingly, we claIm that cI~~ens In ~epresentative
democracies can coordinate their responses to polIncal chOIces
insofar as the choices themselves are coordinated by political parties.
To put our cards on the table at the start, the evidence backing our
claim is patchy at best. This is partly because the c!ain: has not ~et ~e~n the
object of systematic study and partly because this kin~ of claim I~ Inherently
difficult to study systematically.6 Our consideranon of preVIOUS research
is therefore frankly opportunistic. Still, iffuture work shows that we
were on the right track, we may be forgiven for giving our argument a
speculative shove here and there. . . .
We assume that people are motivated to be consistent In theIr behe~s
and sentiments about politics, just as they are in their beliefs and sennThe
Hypothesis ofMenu Dependence • 339
ments about other areas oftheir lives.7 The question is: how do they manage
to achieve consistency in politics, to the extent they actually achieve it?
B~ what .means, given a set of alternatives, do they select one congruent
WIth theIr general political orientations? How, for that matter, do they
for~. general po~tical orientations given the limited attention they pay to
polIncs and publIc affairs?
. We pro~eed in three steps. Our first step is to review critically the prinCIpal
conSIstency-generating mechanisms so far proposed. They are part of
an ans,:,er-~ut only as part of a broader perspective. Our second step is
to outlIne thIS broader perspective. A substantial part of the public is able
to achieve consistency in their political choices, we will suggest, just because
their choices are menu dependent. Finally, given the speculative
character of our account, our third step is to elaborate on some especially
obvious qualifications.
ESTABLISHED MECHANISMS
How, so far as citizens are consistent in their political choices, do they
manage to achieve consistency? There is no shortage of answers. By our
count, four are commonly given: on-line tallies, group affect, basic values
and political orientations, and judgmental heuristics.s Four mechanisms
may seem three too many. But each can do solid work. The difficulty, we
will suggest, is that even when taken together, they cannot do the necessary
explanatory job.
Taking the quartet of explanatory mechanisms in order of explanatory
reach, we start with the so-called Stony Brook account of motivated reasoning
and on-line impression formation.9 Lodge and his Stony Brook
colleagues have proceeded from a simple intuition. Voters form initial impressions
ofcandidates, stored in memory in the form offeelings (or affect).
Exposed to subsequent information, the affective tally ofpast impressions
comes to mind at the same time voters become aware ofthe affective "tag"
attached to the new information. In the process voters update their running
tally. Evaluations are thus constructed anew with each addition of
?ew information. But the impact ofeach new piece ofinformation is predIcated
on the tally ofprevious evaluations. So voters tend to tal<e onboard
information that confirms their previous view and to reject or discount information
at odds with it.
T~s hypothesis ofa running affective tally provides an attractively simple
solunon to the problem of mental bookkeeping. Citizens need remember
only one thing. And the one thing they need to remember is easy to remember.
To respond consistently to new information, voters need only know
how they feel about a candidate. They need not be aware of-indeed, they
340 • Sniderman and Bullock
probably do not remember-the specific reasons that led them to feel as
they do. .
The on-line model is welcome on another ground. It calls attentIon to
the pivotal role of affect in the maintenance of cognitive consistency. This
is a point ofsome importance. A distinction between cognition and affect,
though well-advised for understanding some problems, is ill-advised .for
understanding the organization of political belief systems. Many belIefs
cannot be genuinely held unless the sentiments appropriate to them are
also held. In politics, this entanglement of cognition and affect is inescapable
and useful. From one angle, what holds a structure of beliefs together,
what makes it cohere, is precisely its affective consistenc~.
Yet the on-line model is cramped. The fully worked-out verSIon of the
model to date gives an account only ofcongruence narrowly defined~the
accumulation ofevaluations ofpolitical objects, one by one. To give an account
of consistency of political choices over sets of choices (say, an array
of social welfare policies), a consistency-e;enerating mechanism of wider
scope is required. In their most recent work, Taber, Lodge and Glathar
, . h' 10 B thopropose "node-link structures' as a candIdate mec aniSIn. ut at IS
stage of their research it is not clear how much the invocation of structure
explains and how much it assumes. .
A second consistency-generating mechanism is predicated on the SImultaneous
role of social groups as objects of public opinion and public policy.
In his seminal essay, Converse picked out the politics of race as one of
two exceptions to the general rule ofminimal constraint in mass beliefsystems.
By way of argument, he first listed a series of public policies ta:geting
blacks-for example, to provide job training, t? increas.e educ,a~onal
opportunity, and to monitor discrimination in hOUSIng. Ordinary c~tI~e~s,
he then suggested, need not have detailed knowledge of these polICIes In
order to respond consistently to them. They need only know how they feel
about blacks.II The more they like and sympathize with black Americans,
the more likely they will be to support these policies; the more they dislike
and feel superior to black Americans, the more likely they will be to oppose
these policies. , .
On this view, citizens maximize congruence between polley alternatIves
and affect toward social groups that are the objects of these policies. They
thus can mount a coherent response to a set ofpolicies serially, responding
independently to each policy. Moreover, the politics of race are nO,t i~osyncratic,
but only an especially vivid illustration of the general prInCIple
that political choices are organized around social groups or around the
connections between social groups and political parties. As the authors of
The American Voter noted long ago, citizens can go a long way in putting
together a politically coherent set of beliefs by knowing th~t la~or un~ons,
for ex~mple, are specially tied to the Democratic Party, whIle bIg bUSIness
The Hypothesis ofMenu Dependence • 341
is specially tied t? the Republican Party. Indeed, knowing of alliances between
salIent.SOCIal groups and political parties, together with having prono~ncedfeellllgs
about them, can amount to "ideol<;>gy by proxy," in the
telling p~ase ?f Can:pbell and his colleagues (Campbell et al. 1960),
It ~er.ta.llllyIS ~osSIbl~ to extract a consistency story out of group likes
and dislikes th~t.IS conSIstent with the intermittent attention ordinary citizens
pay to politIcs. And perhaps this is the path to take. Still we have reservati?ns.
The issue ofrace aside, there is not a large amount ~fevidence that
chOICes over political issues are based on a group calculus. Moreover to
?e.t consistency ofchoice over sets ofpolicies dealing with diverse gro~ps,
It IS necessar.y to rely on a premise that feelings across groups are consistently
orgaruzed-and that premise is not plausible for large portions of
the public.
I2
I? any case, what is doing the real work in generating political
coherence IS know!edge ofpartisan coalitions. Absent knowing, for exam~le,
th~t labor uruon~ are allied with the Democratic Party and big
busllle~sWIth ~e RepublIcan Party, there is no basis for ideology by proxy.
A third conSIstency-generating mechanism, core values) is appealing on
~e gro~ndsof both explanatory scope and simplicity. On the one side, it
IS not chffic~lt to see how foundational values like liberty and equality can
ground chOices over large sets of specific issues. On the other side since
the n~beroffoundational values (as compared to opinions) is sm~l, it is
not dIfficult to see how ordinary citizens can organize them coherently. So
on both counts an account of consistency is conceivable, with citizens sele~ting
the alternative, from among those on offer, that is most congruent
WIth their core values.
Fel~an has, more than anyone, given empirical support to the hypotheSIS
ofcore ~alues coordinating policy choices. In pioneering work, he
develop~d.c~~dat~ measures of a triad of core values-egalitarianism,
econo~IC lllchvidualism, and support for free enterprise.I3 In subsequent
work WIth Steenbergen, Feldman has shifted focus to humanitarianism as
a core value, a shift we find appealing because ofits fit with recent normative
rethinking ofequality as an ideal ofhumanitarianism.
Intuitively, the appeal to core values as consistency generators is attractive.
It meshes smoothly with the language in which political thinkers conceive
political choices without requiring unreasonable assumptions about the
capacity ofordinary citizens :0be political thinkers. And empirically, it surely
must be true that Issue chOIces are grounded to a degree in basic values.
T~~ problem is that it appears to be true only to a limited degree. If the
empIrIcal benchmark of congruence is the power of measures of core values
to predict specific political choices, the conclusion to draw is that congruence
is modest at most. Of course, the problem may not lie with the
hypothesis ofcore values as consistency generators. As plausibly, the problem
may be the limitation ofcurrent measures to gauge adherence to core
342 • Sniderman and Bullock
values. Measurement of core values is still in an early stage. Methodologically,
it has concentrated almost entirely on the rating ofvalues.I4 This may
prove the best approach in the end, though Jacoby has recently introduced
an innovative approach involving the ranking of values.IS Future
development of ranking techniques, or indeed of rating ones, may show
that substantial numbers of ordinary citizens choose among political alternatives
by selecting the one most congruent with their core values. But if
that is the right lesson to draw, this is not the right time to draw it.
The last set of consistency-generating mechanisms is heuristics, that is,
judgmental shortcuts. The intuition here is that even comparatively wellinformed
citizens have a limited amount of information to work with. If
they are to be able to make politically coherent judgments they need an
easy-to-operate calculus. Judgn1ental shortcuts would seem to fit the bill.
Consider the likability heuristic introduced by Brady and Sniderman.I6
The specific task is locating the positions of strategic actors in a political
landscape. Although some are concrete and immediate-men and women,
for example-others are more abstract and removed-liberals and conservatives,
for example. How, then, are citizens able to define correctly
what liberals and conservatives stand for-that is, accurately describe their
positions over an array of issues-even though they cannot accurately define
liberalism and conservatism?
By following a judgmental shortcut, Brady and Sniderman suggest. To
estimate accurately the issue commitments of any pair of competing
groups, it is necessary only for citizens to know their stand on the issue
and to take into account the difference in their feelings toward the two
groups. So, even without knowing what liberalism and conservatism are as
ideologies, citizens can know what liberals and conservatives stand for.
Notice, given our interest in consistency, the qualifying condition. The
heuristic only works to the degree that an individual likes one side and dislikes
the other. If they do not recognize that liking liberals entails disliking
conservatives, the likability heuristic fails.
The hypothesis of heuristics is frequently invoked.17 It seems to provide
a method of explaining how citizens can compensate for the limitedinformation
they have about political affairs. It is worth making plain why this
way of putting things is misleading, for it throws light on a neglected
problem.
It is true in one sense, but false in another, that citizens can compensate
for limited information by taking advantage of a heuristic. The sense in
which it is true is that even comparatively well-informed citizens are unlikely
to have all the information at hand to reason through an informed
decision. The sense in which it is false is that the likelihood of talcing advantage
of an effective judgmental shortcut itself depends on how wellinformed
citizens are. It takes smarts to make smart moves, if it is okay to
The Hypothesis ofMenu Dependence • 343
speak plainly. It accordingly is false to suggest that the public taken as a
whole can make judgments about a problem in public ·affairs by taking advantage
of heuristics that match the judgments they would make if they
were to be fully informed about it. It is only true that the better informed
they are, the less they are likely to be handicapped by their absolute lack of
knowledge.
The efficacy ofjudgmental shortcuts as a consistency mechanism is heavily
conditional on political sophistication. And so, a strealTI ofstudies make
plain, are the other consistency-generating mechanisms. One exception
aside, it is the politically more sophisticated who benefit more from each of
these mechanisms, that is, who choose more consistently in virtue ofthem.
That applies to judgmental shortcuts like the likability heuristic, core values,
and basic political orientations, and even to on-line processing, which
was initially commended precisely for its simplicity of operation.18
How, then, should the explanatory books be balanced? On the profit
side ofthe ledger, we have gotten better and better at giving an account of
how citizens make political choices the more politically sophisticated they
are. On the loss side of the ledger, however, we have gotten worse and
worse at giving an account of how citizens make political choices the less
sophisticated they are. Since there is not an excess of the former and no
shortage of the latter, this is explanatory progress of an ironic stripe. It
may be useful, therefore, to approach the problem of consistency from a
different perspective.
MENU DEPENDENCE: POLICY AGENDAS, ISSUE FRAMING,
AND ISSUE CENTRALITY
The capacity of citizens to malte consistent choices, we shall suggest, is
contingent on the organization ofthe menu ofchoices presented to them.
We shall explore three aspects of menu dependence for facilitating consistency
ofchoices: first, menu dependence over sets ofissues' second for issues
taken one at a time; third, for variation in consistency'across i;sues.
We start with policy agendas. To ask how citizens manage to achieve a
consistent response not merely to issues one by one but to whole sets of
them assumes that a substantial part ofthe public is in fact capable ofdoing
so. But in order to demonstrate that this assumption of consistency is in
fact warranted, it is necessary to know with respect to what they are striving
to be consistent.
What might an answer look like? To the extent citizens respond to issues
separately, evaluating each on the basis ofconsiderations unique to it, then
consistency understood either as constraint or congruence is ruled out. A
specific issue, just by virtue ofbeing specific, points to a particular matter-
344 • Sniderman and Bullock
whether the government should increase unemployment assistance, for
example, or whether it should ramp up job-training programs. And when
making choices about particular matters, citizens must be able to rely on a
more general view of the matter if they are to make consistent choices
across issues. But what might their general view of the matter consist in?
On the one side, it cannot be something as general as an ideological orientation,
since that is something so general as to be out of the reach of
most citizens. On the other side, it cannot be something so specific as their
feelings about particular social groups, since that is not general enough.
Something in between is needed, and that something, we suggest, is a policy
agenda.
To illustrate what we mean by a policy agenda, we draw on the research
of Carmines and Layman. Analyzing a series of National Election Surveys,
they pick out a three-dimensional structure of policy preferences. One dimension
is defined by issues like government support for jobs and standard
of living; a second, by issues like abortion and women's rights; a
third, by issues like government help for blacks and spending on programs
for blacks.19 Each attitudinal dilnension thus maps on to a policy agenda.
Carmines and Layman accordingly label the first the social welfare agenda;
the second, the cultural agenda; the third, the racial agenda. Our concern
is not whether there are three agendas or two-or four, for that matter.
For our purposes, Carmines and Layman's issue analysis makes two points
worth emphasizing, one negative, the other positive. The negative point is
that the ideas of ordinary citizens tend to be unrelated across policy agendas.
The positive point is that they tend to be consistent within them.
This distinction between consistency within and across policy agendas is pivotal.
The classical studies ofideology in mass publics proceeded on the premise
that the menu ofissue choices has no structure, that it is not divided into
distinct parts. So they took as a test of ideological thinking consistencydefined
as constraint-across the full spectrum ofcontemporary concerns,
ignoring demarcation of issues into distinct sets.20
In turn, they drew the
lesson that ordinary citizens cannot take in liberalism-conservatism as a
coherent whole.
But it does not follow that because the public as a whole cannot take in
liberalism-conservatism as a whole, it cannot take in their component parts
coherently. Each policy agenda-social welfare, cultural, and racial-eaptures
a component ofthe ideological divide in contemporary American politics.
But each represents a distinct component, with each having concerns
that mark it offfrom the others, in elite as well popular discourse. And accordingly
it is perfectly possible, indeed commonly the case as Carmines
and Layman's results indicate, that citizens can be liberal (or conservative)
with respect to issues like abortion or women's rights without being liberal
(or conservative) with respect to issues of government assistance for
The Hypothesis of Menu Dependence • 345
the poor or for those in search ofjobs. Carmines and Layman's results thus
underscore two points: first, that the three policy agendas are distinct in
the mind ofthe public; and second, that each of the agendas is sufficiently
bite-sized to be taken in as a coherent whole by the public as a whole. Ordinary
citizens thus tend to be consistently liberal (or conservative) agenda
by agenda in spite ofnot being consistently liberal (or conservative) across
agendas.
Now think ofthe overall view that ordinary citizens form about a policy
agenda as a latent trait-in our parlance, their general view of the matter.
How do citizens go about choosing among alternatives for an issue on a particular
agenda? By consulting their general view ofthat agenda.. Just so far as
their general view of the cultural agenda is conservative, they will consistently
choose the conservative alternative across issues on that agenda.
And just so far as their general view is liberal, they will consistently choose
the liberal alternative. In this way the public as a whole, and not just the
most sophisticated segment of it, can mal<:e consistently liberal and conservative
choices agenda by agenda even though they cannot make consistently
liberal and conservative choices across agendas.
But hO\\T do issues come to be bundled together as they are? Why do policy
agendas include some issues but not others? And, still more difficult,
how do some citizens achieve consistency not only within policy agendas
but also between them?
Part of the answer of how consistency is achieved across agendas is well
established. One characteristic of citizens as choosers, their level of political
sophistication, plays a crucial role. More exactly, a double role. Political
awareness facilitates consistency by facilitating a coherent set of core
beliefs and expectations, generalized priors ifyou like, on which to base a
specific choice. Additionally, political sophistication promotes consistency
by facilitating recognition ofthe relevance ofthese core beliefs and expectations
in Inaking specific issue choices.21 To this harmony of views we
would add just one discordant note. As we read the research literature, the
emphasis is on individuals as active information processors, imposing order
and coherence on what otherwise would be, in James's enduring phrase, a
blooming buzzing confusion.22 By contrast, consistent with our view on
the external organization of choice spaces, we propose that political sophistication
facilitates the more modest task of recognizing rather than
imposing coherence.
And that coherence is imposed, we suggest, through competition between
political parties and candidates for control of government. Partly
because of the commitments of their core members and partly because of
their strategic alliances with interest groups, parties compete against each
other agenda by agenda. And very largely because ofthe dynamics ofelectoral
competition they yoke agenda to agenda.
346 • Sniderman and Bullock
For evidence, indirec;t as it is, we rely on a recent study by Layman and
Carsey (2002)..They show that the link between the social welfare and cultural
agendas is conditional on strength ofparty identification and awareness
of partisan differences on the two agendas. For those identifying strongly
with a party and aware ofthe differences between them on the two agendas
there is a strong link; for those identifying with a party but not aware of
the differences between them on the two agendas, there is only a moderate
link; for those who do not identify with a party whether or not they are
aware of differences between the parties there is essentially no link.
Consider the implications of Layman and Carsey's findings. A consistency
theory of substantial scope has appeared a nonstarter because there
has appeared to be no way to get coherence ofchoice out ofcitizens themselves.
But the consistency-generating mechanisms, their findings suggest,
are institutional. Political parties provide the basis for the consistency ofindividuals;
indeed, twice over on Layman and Carsey's results: once through
the consequences of attachment to parties and once through the consequences
ofknowing parties' contrasting positions on issues.23 This double
role of parties testifies to the institutional organization ofpolitical choices.
By facilitating consistency across policy agendas, parties help the substantial
part ofthe public attached to them to mal<e choices across the spectrum of
contested issues on the basis ofa consistently liberal or conservative view of
the matter.
On our view, it is parties and candidates that do the heavy lifting necessary
for consistency in public opinion. They reduce the number of alternatives
open to choice to only a few-indeed, frequently to only two.
They portray those alternatives as competing courses of action. The implication
is that rejecting one means accepting the other. They stamp a partisan
and ideological brand on the arguments offered in their favor, signaling
that accepting one means rejecting the other. Political candidates, by assuming
the lead of a party, can do much to determine how issues are organizedinto
policy agendas or even how policy agendas are organized into
overarching ideological orientations.24 But at the end ofthe day, it is through
parties that the menu of choices on offer to citizens is organized.
And much to the advantage of citizens. By structuring political choice
spaces, parties facilitate citizens' reasoning consistently from basic principles.
It is not possible to derive a stable and coherent structure of choices
the other way around, from tne power ofordinary citizens to reason about
politics. Coherence at the level of individual citizens is conditional on coherence
in the menu of choices presented to them to tnake as citizens.
The role of political parties in bundling issues into bite-sized agendas
promotes consistency both within and across sets, or agendas, of issues.
But parties also promote consistency issue by issue. Consider the framing
and highlighting ofissues. Sniderman and Theriault (chapter 5 in this volThe
Hypothesis ofMenu Dependence • 347
ume) investigate efforts to frame or define "the essence of the problem"
for an issue, that is, to make a persuasive case as to "how it should be thought
about and ... [to] recommend what (if anything) should be done."25 They
show that when respondents are confronted with competing efforts to
frame an issue, rather than being confused, they are markedly more likely
to select the policy alternative congruent with their general view of the
matter. They embed this result in a larger story about the connection between
electoral competition and electoral choice. It is just so far as parties
and candidates compete over issues, they suggest, that voters are in a position
to make consistent choices.
Sniderman and Theriault present their story as one about the electorate
as a whole, not just the upper echelon of the politically sophisticated part
of it. Indeed, in their analysis, they show that their findings hold for both
less sophisticated and more sophisticated citizens. Both are more likely to
choose, from among the alternatives on offer, the one most congruent
with their general view ofthe matter when they are exposed to competing
efforts to frame an issue.
This result is important in its own right, illustrating as it does how citizens
can more accurately find their political bearings thanks to the clash of
competing arguments. But there is an aspect ofSniderman and Theriault's
results that they do not explore and that we want to call attention to here.
In one of their experiments, the effect of being exposed to arguments
on both sides of an issue is conditional on the respondents' level ofpolitical
awareness. Finding that political sophistication makes a difference is a
standard result. But in one of their experiments, Sniderman and Theriault
find the opposite ofwhat is standardly found. Instead of finding that con.
gruence increases as political information increases, they find that the less
politically aware, not the more, benefit most from exposure to cOlnpeting
ways of thinking about a political issue.
As things stand, only two on-the-shelf explanations do better at accounting
for the choices ofthe less politically sophisticated than of the more sophisticated.
In the contemporary version of the nonattitudes model, the
decisive consideration is the extent to which people have an approximately
evenly balanced set ofreasons to support or oppose a policy.26 This certainly
counts as explanation, but there is no way to extract a consistency-based account
from a nonattitudes model, since the pivotal factor is precisely the
extent to which people's views about an issue are inconsistent.
The second on-the-shelfaccount centers on group affect. It can plausibly
be argued that the less politically sophisticated citizens are, the more likely
they are to lack a cognitive basis for malting political choices, and therefore
the more likely they are to rely on an affective one. There is some evidence
this is so, though less than one may think.27 Sniderman and Theriault point
to a third alternative for giving an account of the choices of the less polit-
348 • Sniderman and Bullock
ically aware portion of the public. It is a fact, and an important one, that
they are less adept at organizing their choices just by virtue of being less
aware. But so far as they can choose consistently, some external machinery
for organizing their responses is specially needed. Sniderman and Theriault
suggest that electoral competition supplies this machinery. Admittedly,
an inverse relationship between consistency of choice and level of
political sophistication holds for only one of their two experiments. So we
would therefore like to pitch their results at a slightly lower level. What
their results show, and not just in one of the experiments but in both, is
that the less sophisticated benefit at least as much as the more sophisticated
from being exposed to the clash of opposing arguments. And it is
electoral competition that generates the clash of arguments.
Issue-framing effects of all· varieties are illustrations of menu dependence.
But they are only a selection ofthe ways in which menu dependence
is driven by the dynamics ofelectoral competition. Consider differences in
the partisan centrality ofissues. There are issues that parties do not wish to
compete on; there are others that, out of a calculation of political advan:
tage or as a consequence of political conviction, they do wish to compete
on. Political issues can accordingly be located along a continuum ofpartisan
centrality, from those most vigorously contested to th?se least vigorously
contested.
In broad outline, the following hypothesis is worth examination. The
more central an issue is to electoral competition, the greater the effort that
political actors, including political parties, will make to call the electorate's
attention to them and to contra,st the alternatives open for choice. The
more peripheral an issue, the less attention directed to it, and the more
similar the alternatives open for choice will appear. Moreover, the more
central an issue is to electoral competition, the "stickier" will be the identification
of issues with parties over time; the more peripheral the issue, the
looser the linkage. The result: the more central the issue to partisan competition,
the stronger the tendency to congruence; the more peripheral the
issue, the weaker.
And what follows froln this? Very briefly, the relevance of alternative
public opinion models varies with the partisan centrality ofa political choice.
Consistency and contestation go hand in hand. The more central an issue
is to electoral competition, the more likely it is that causal accounts favoring
opinion consistency, whether in the form ofon-line processing or more
elaborately hierarchical models, will apply. Conversely, the more peripheral
an issue, the more likely it is that models accentuating inconsistencyConverse's
black-and-white model, Krosnick's satisficing model, and Zaller's
consideration sampling model with an important qualification-will apply.
Impressionistically, this hypothesis fits to a first approximation the standard
findings in public opinion research. Consistency (whether interThe
Hypothesis ofMenu Dependence • 349
preted in terms ofconsonance with other beliefs at one moment in time or
stability of the same belief over time) seems weal~est for issues that are remote
from established partisan battlegrounds (e.g., U.S. intervention in
Nicaragua) and strongest for those that are central to theln (e.g., racial
po~c.ies). Th~ hypothesis of a connection between partisan centrality and
opImon conSIstency seems to us promising, and offers an example of external
anchoring of belief, although of course further research is required
to test this hypothesis ex ante as opposed to trawling through previous research
ex post.
SOME ESPECIALLY OBVIOUS QUALIFICATIONS
Much ofour argument should be filed under the heading of "Suggestions
and Speculations." Apart from issuing this general caution, we want to
post some specific warning signs.
Themost conspicuous concerns the problem ofbeliefrevision. Any theory
ofchoice that is rational under any description ofrational must have a
provision for updating. External circumstances change, and there must be
some way of taking (some of) these changes into account. Yet consistency
theories take as their principal premise that what you believe and feel now
is conditional on what you believed and felt before; or still Inore strongly,
that in forming a beliefor feeling now, you aim at maximizing consistency
with what you believed and felt before. But if citizens have been consistency
maximizing for any extended period of time, how are they capable
ofsubstantially revising their beliefs?
One route to tal~e is to posit an accuracy motive operating in tandem
with a consistency motive. On this view, "[P]eople motivated to arrive at
a particular conclusion attempt to be rational and to construct a justification
of their desired conclusion that would persuade a dispassionate observer."28
Under some interpretation this surely is right. Still, it is less than
reassuring for revising beliefs about politics. For one thing, the notion of
a "dispassionate observer" is inherently difficult to pin down just so far as
political choices turn on conflicts between incommensurable values. For
another, an accuracy motive may work as a self-correcting device for fairly
specific motives, for example, maintaining a positive self-evaluation. But
once an internally consistent framework of beliefs, feelings, and expectations
has been established, how does one stand outside it? Just so far as a
framework is internally consistent, it carries with it a self-affirming warrant
that this is the way the world really is, as any dispassionate observer would
agree. In any case, in the absence of a theoretical basis, appealing to an
accuracy motive to override a motivation to consistency is unacceptably
ad hoc.
350 • Snidennan and Bullock
An alternative route to respond to the problem of updating is to take
advantage of Bayesian models of political learning.29
Gerber and Green
have put forward a case for unbiased Bayesian updating in the public as a
whole.30 On their interpretation, the case for unbiased learning is cemented
by the observation that in the aggregate, citizens identifying with opposing
parties revise their prior beliefs by approximately the same amount when
confronted by new experiences. This is a comforting outcome. But it is
not obvious that accepting Bayesian updating comes at the price of rejectii1g
consistency maxiinizing. Bartels argues in response that Gerber and
Green's central result illustrates partisan bias, not unbiased updating, provided
that Bayesianism is properly interpreted. Bartels's interpretation eases
the tension between consistency maximizing and updating in the face of
new experience (Bartels 2003).
Even so, Bayesian updating and consistency maximizing make an illmatched
couple. The whole thrust of a consistency account is to emphasize
the recalcitrance of people, to revising their beliefs even in the face of
clear evidence that their preconceptions have failed thein. And what is worse,
this appears to be at least as true for those who have given more thought
to a problem, at any rate when it comes to politics. Consider Tetlock's
(1999) studies of theory-driven reasoning. He shows that experts cope
with the interpretive cOillplexities and ambiguities of world politics by resorting
to theory-driven strategies ofthinking that allow them to: (1) mal<:e
confident counterfactual inferences about what would have happened had
different policy paths been pursued (plausible pasts); (2) generate predictions
about what might yet happen (probable futures); and (3) defend
both counterfactual beliefs and conditional forecasts from potentially disconfirming
data. For exalnple, experts who were convinced in the late
1980s that the Communist Party of the Soviet Union would continue to
control the levers of power in the Kremlin deep into the 1990s argue that
this outcome nearly did happen (the attempted coup of August 1991),
whereas experts who expected the EU currency convergence project to fail
could argue that the project nearly did come undone at several junctures
(again, the close-call counterfactual defense) and eventually will still collapse
(the just-off-on-timing defense).
The weight ofTetlock's findings is that experts do not come close to living
up to the terms of the Bayesian reputational bets they make about the
relative likelihood ofevents. On his view, there are two basic reasons-one
rooted in the human mind and the other in the structure of the political
environment-why experts find it relatively easy to resist changing their
minds when the unexpected occurs. Resistance is easy, in part, because experts
come cognitively equipped with a complex of cause-effect schemata
that allow them to portray events that were not viewed as likely ex ante as
The Hypothesis of Menu Dependence • 351
close to inevitable ex post. Resistance is also easy, in part, because historical
data lend themselves to a variety of alternative causal interpretations
and it is rarely possible to achieve ideological consensus on how history
would have unfolded in counterfactual worlds in which alternative policy
paths were pursued. When reality constraints are weal<:, strongly held preconceptions
fill the void.
There is another problem. When it comes to political Illatters at the center
of electoral competition, political sophistication promotes divergence, not
convergence, of political choices along partisan and ideological lines. It is
not obvious tl1at divergence can be accommodated on Bayesian principles.31
Still, consistency n1aximizing without a provision for updating is a formula
for irresponsible voters in a world in which things do change. How
can citizens change in response to changing circumstances if they are indeed
consistency maximizers?
Framing the problem of change in terms of belief revision (Bayesian or
otherwise) presupposes that the question that needs an answer is how citizens
on their own hook revise their beliefs in the face of new experience.
But the central argument of our project is that there is much that citizens
don't do on their own hook. Throughout our claim has been that political
institutions do the heavy lifting. That is just the point of emphasizing the
menu dependence of choice.
And what falls out ofthis focus on menu dependence is updating via another
route. Partisan elites have incentives to keep the menu of basic
choices fixed for extended periods of time. But/to the extent their continuing
hold on political power is put at risk by changing circumstances, they
are under pressure to respond. They must introduce new policies or revise
old ones. Sometimes they do the one, sometimes the other, sometimes under
old labels, sometimes under new ones. The question of updating takes a
quite different form so viewed. It is not necessarily the beliefs of citizens
that get revised. It is instead the courses of action they get to choose that
are revised. And so far as this is true, updating is built into the menu of alternatives
on offer to citizens.
Again we think it is useful to distinguish between the characteristics of
citizens as choosers and the characteristics of the choices presented to
them. Yet drawing distinctions by twos .risks supposing that the two are
separate, as though some portion ofthe variance in consistency is to be attributed
to characteristics of citizens and a different portion to characteristics
of choices. The temptation to pronounce one or the other Illore
fundamental appears nearly irresistible-irresistible and ironic.32
Thus,
students of legislative institutions devote themselves to analysis of the
strategic choices ofpolitical elites, yet they take electoral preferences to be
the most fundamental causal force. Generously reciprocating, students of
352 • Sniderman and Bullock
mass behavior devote themselves to the investigation of public opinion,
yet they take the initiatives ofelites to be the most fundamental causal force.33
It is all the more worthwhile to emphasize that public opinion and elite
strategic choice are part of a single account.
Elsewhere we have written about political choice spaces.34 A choice
space is, simply enough, the menu ofalternatives on offer. Alternatives are
defined by political elites partly by anticipation of electoral preferences,
partly by consequences of past commitments. Alternatives are chosen by
mass publics partly on the basis of preferences, partly on the basis of aptitudes
in connecting preferences to alternatives. Plus, there is a complicating
factor on both sides ofthe equation. Working with just two terms, elite
actions and electoral preferences, is working with two terms too few. From
time to time, shocks intrude-a slump in the economy, a threat overseas.
Elites in office have to respond just because their eyes are on the next election.
So they draw on their respective ideologies, broadly liberal or conservative
depending on which party they belong to. And apart from the
demands of interest group politics, they draw on their ideologies in an effort
to take advantage of their best theories of how the world actually
works. The whole idea is thus to tie together elite initiatives and citizen responses
through the vehicle of a choice space.
Another especially obvious point of qualification: the explanatory limits
of our account. We have worked to reduce explanation of choice to the
maximization of consistency, where consistency is defined as congruence
between selection of a policy alternative and direction of an underlying
orientation. Analyzing judgments of the political commitments of strategic
political actors, Brady and Sniderman have formalized a consistency
model of judgment that derives accuracy of cognitive judgments from
basic likes and dislikes, coherently organized.35 This reduction, focusing
on consistency, offers gains. It brings the benefit ofparsimony. It facilitates
formulation of a formal model cast in maximizing form. And it generates
non-obvious predictions.
Still, our account has notable gaps. One illustration will have to do for
many. A key question, we remarked, was "With respect to what, exactly,
are most citizens trying to be consistent?" The answer, we suggested, was
their general orientation toward policy agendas, one at a time. The tone of
our account certainly suggests that the organization of policy agendas is
the work ofpolitical elites. Even to our eyes, this suggestion is muzzy. Political
institutions do the immediate work ofbundling issues. But the opinions
that elites have about the shape of public opinion surely must play
some role. If so, how does the process work? To answer that question it
will be necessary to detail how particular policy agendas come to be organized
by political activists.
The Hypothesis of Menu Dependence • 353
There is a more general point. Only after the consistency mainspring is
wound do the explanatory clocks ofconsistency theories start ticking. But
of course this involves some fairly heavy question begging. Why do some
p~ts.of the pu~lic identify with one political party while others identify
WIth Its competItor(s), and why are some disposed to a liberal or broadly
l~ft poli~cal O~tlook while others are disposed to a conservative or broadly
rIght orlentauon? How exactly do the major pieces of citizens' political
furniture, their basic beliefs and loyalties, come to be acquired? The answers
must come from studies of the acquisition of basic political beliefs
and allegiances. In this sense our account ofpolitical choices is parasitic on
accounts ofpolitical socialization and learning.
A FINAL WORD
Perhap.s quixotically, we want to suggest that the weakest point ofour hypotheSIS
of menu dependence may prove its strongest recommendation.
When in need ofan unseen (explanatory) hand, we point to political institutions'
political parties and electoral competition in particular. But invoking
institutions presents a special problem bf proof. So far as political
institutions are responsible for the menu of political choices (as opposed,
for example, to exogenous shocks), the basic choices tend to be fixed for
extended intervals, often a generation or more. Certainly that seems the
case for a political system like the American one. But to say that the menu
of basic choices is fixed over time is to say it is a constant. And so far as it
is a constant, then how can it be shown to constrain political choice?
We have worried about this problem for some time. Historical accounts
aside, we believe the best way forward is through randomized experiments
in public opinion surveys. Many features ofmenu dependence lend themselves
to experimental manipulation-the effect of taking polar positions,
the party branding of policy alternatives, and the provision of explicit alternatives,
among them. This is the road we have started down. We invite
others to join us.
NOTES
We gratefully acknowledge the assistance of the National Science Foundation
through an Achievement-Based Award (SES-0111715) in preparation ofthis book.
On the individual front, we are deeply grateful to James H. Kuklinski, Philip E. Tetlock,
and Michael Tomz for criticism laced with encouragement.
1. We interpret constraint broadly, referring not only to the connectedness of
specific beliefs but also to the connectedness ofany set ofidea elements at the saIne
354 • Sniderman and Bullock
level ofspecificity. Our usage thus conforms to Peffley and Hurwitz's (1985) concept
ofhorizontal linkage. Accordingly, we often speak of coherence in place of
constraint.
2. The notion ofcongruence corresponds to Peffley and Hurwitz's vertical linkages
in belief systems.
3. The two classic exceptions are group-centered politics, where affect-for example,
toward blacks-can assure constraint across issues dealing with blacks, and
issue publics, where UnCOlnmon motivation and focused attention on a specific set
of concerns can also do the trick. See Converse 1964.
4. See, for example, Sniderman, Brody, and Tetlock 1991; Zaller 1992; and
Jackman and Sniderman 2001. It deserves to be emphasized that Converse first
sounded this note in his seminal essay, although he qualified this by positing that
only a thin slice of the public is sophisticated enough about politics to form politically
coherent belief systems.
5. For the concept ofmenu dependence in economic analysis, see, e.g., Sen 2002.
6. For the principal difficulty, in our view, see our concluding comment on
menus as explanatory constants.
7. Festinger 1957.
8. For all the obvious reasons, we olnit consideration ofideology, understood as
an abstract and integrated outlook on politics.
9. Hastie and Park 1986; Lodge, Steenbergen, and Brau 1985; Lodge, McGraw,
and Stroh 1989.
10. Taber, Lodge, and Glathar 2001 :216, italics theirs.
11. Converse 1964.
12. See the discussion below on the connection between affective coherence
and levels of political sophistication.
13. Feldman 1988. Feldman has carried this on in an interesting way with humanitarianism-see
Feldman and Steenbergen 2001. See also Frankfurt 1988.
14. Robinson, Shaver, and Wrightsman 1999.
15. Specifically, the procedure scores forced choices over all possible pairs of a
lilnited set of basic values. See Jacoby 2002.
16. Brady and Snidennan 1985.
17. E.g., Popkin 1991 and Lupia 1994.
18. For the conditionality of on-line processing on political awareness, see McGraw,
Lodge, and Stroh 1990.
19. Carmines and Laylnan 1997; Shafer and Claggett 1995.
20. Converse 1964, especially Table VII.
21. Converse 1964; Sniderman, Brody, and Tetlock 1991; Zaller 1992.
22. E.g., Taber, Lodge, and Glathar 2001:198.
23. See, specifically, Layman and Carsey 2002:798.
24. For a key argument on the contingent character of issue bundling, see
Cannines and Stimson 1989.
25. For a review of previous efforts, see Nelson and Kinder 1996:1057. See
also Entman 1993.
26. Zaller 1992.
27. Sniderman, Brody, and IZuklinski 1984.
The Hypothesis of Menu Dependence • 355
28. Kunda 1990:482-83, emphasis ours. Contrary to the impression that this
point of disagreement may leave, we benefited greatly from Kunda's work. Her
analysis on motivated reasoning introduces a very useful distinction between motives
for accuracy and for directional goals.
29. See, for example, Achen 1989, 1992; Bartels 2003; Gerber and Green
1998,1999; Zechman 1979; and Husted, Kenny, and Morton 1995.
30. Green, Palmquist, and Schickler 2002: chap. 5; Gerber and Green 1999,
1998.
31. For the connection between convergence and Bayesian updating, see Bartels
2003; for evidence on the connection between polarization and political sophistication,
see Sniderman, Brody, and Tetlock 1991.
32. We can attest to the temptation, having succumbed ourselves in earlier efforts.
We are especially indebted to James Kuklinski for pointing out to us the error
of our ways.
33. Zaller 1992 is the preeminent example.
34. Sniderman 2000; Jackman and Sniderman 2001.
35. It may be worth remarking that the Brady-Sniderman model has the nonintuitive
property of generating accurate cognitive judgments about the political
commitments of strategic actors (e.g., political parties, liberals and conservatives)
from affective orientations to those actors. See Brady and Sniderman 1985.
REFERENCES
Achen, Christopher. 1989. "Prospective Voting and the Theory of Party Identification."
Paper presented at the annual meeting of the Atnerican Political Science
Association, Atlanta, August 30-September 3.
---.1992. "Social Psychology, Demographic Variables, and Linear Regression:
Breaking the Iron Triangle in Voting Research." Political Behavior 14:195-211.
Bartels, Larry M. 2003. "Beyond the Running Tally: Partisan Bias in Political Perceptions."
Politi£al Behavior 24:117-50.
Brady, Henry E., and Paul M. Sniderman. 1985. "Attitude Attribution: A Group
Basis for Political Reasoning." American Political Science Review 79:1061-78.
Campbell, Angus, Philip Converse, Warren Miller, and Donald E. Stokes. 1960.
The American VOter. New York: Wiley.
Carmines, Edward G., and Geoffrey C. Layman. 1997. "Value Priorities, Partisanship,
and Electoral Choice: The Neglected Case ofthe United States." Political
Behavior 19:283-316.
Carmines, Edward G., and James A. Stimson. 1989. Issue Evolution: Race and the
Transformation ofAmerican Politics. Princeton: Princeton University Press.
Converse, Philip E. 1964. "The Nature ofBeliefSystems in Mass Publics." In Ideology
and Discontent) ed. David E. Apter. New York: Free Press.
Entman, Robert M. 1993. "Framing: Toward Clarification of a Fractured Paradigm."
Journal ofCommunication 43:51-58.
Feldman, Stanley. 1988. "Structure and Consistency in Public Opinion: The Role
of Core Beliefs and Values." American Journal ofPolitical Science 32:416-40.
356 • Sniderman and Bullock
Feldman, Stanley, and Marco R. Steenbergen. 2001. "The Humanitarian Foundation
ofPublic Support for Social Welfare." AmericanJournal ofPolitical Science
45:658-77.
Festinger, Leon. 1957. A Theory ofCognitive Dissonance. Stanford: Stanford University
Press.
Frankfurt, Harry G. 1988. The Importance of What We Care About. New York:
Cambridge University Press.
Gerber, Alan, and Donald P. Green. 1998. "Rational Learning and Partisan Attitudes."
American Journal ofPolitical Science 42:794-818.
---. 1999. "Misperceptions about Perceptual Bias." Annual Review ofPolitical
Science 2:189-210.
Green, Donald, Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts and
Minds: Political Parties and the Social Identities ofVoters. New Haven: Yale University
Press.
Hastie, Reid, and Bernadette' Park. 1986. "The Relationship between Memory
and Judgment Depends on Whether the Task is Memory-Based or On-Line."
Psychological Review 93:258-68. "
Husted, Thomas A., Lawrence W. Kenny, and Rebecca B. Morton. 1995. "Constituent
Errors in Assessing Their Senators." Public Choice 83:251-71.
Jackman, Simon, and Paul M.Sniderman. 2001. "The Institutional Organization
of Choice Spaces: A Political Conception of Political Psychology." In Political
Psychology) ed. Kristen Renwick Monroe. Malwah, NJ: Lawrence Erlbaum.
Jacoby, William. 2002. "Core Values and Political Attitudes." In Understanding
Opinion) ed. Barbara Norrander and Clyde Wilcox. 2nd ed. Washington, DC:
CQ Press.
Kunda, Ziva. 1990. "The Case for Motivated Reasoning." Psychological Bulletin
108:480-98.
Layman, Geoffrey C., and Thomas M. Carsey. 2002. "Party Polarization and Conflict
Extension in the American Electorate." American Journal ofPolitical Science
46:786-802.
Lodge, Milton, Kathleen McGraw, and Pat Stroh. 1989. "An Impression-Driven
Model of Candidate Formation." American Political Science Review 83:
399-420.
Lodge, Milton, Marco R. Steenbergen, and Shawn Brau. 1995. "The Responsive
Voter: Campaign Information and the Dynamics of Candidate Evaluation."
American Political Science Review 89:309-26.
Lupia, Arthur. 1994. "Shortcuts versus Encyclopedias: Infonnation and Voting
Behavior in California Insurance Reform Elections." American Political Science
Review 88:63-76.
McGraw, Kathleen M., Milton Lodge, and Patrick Stroh. 1990. "On-Line Processing
in Candidate Evaluation: The Effects of Issue Order, Issue Importance,
and Sophistication." Political Behavior 12:41-58.
Nelson, Thomas E., and Donald R. Kinder. 1996. "Issue Frames and GroupCentrism
in American Public Opinion." Journal ofPolitics 58:1055-78.
Peffley, Mark A., and Jon Hurwitz. 1985. "A Hierarchical Model ofAttitude Constraint."
American Journal ofPolitical Science 29:871-90.
The Hypothesis of Menu Dependence • 357
Popkin, Samuel L. 1991. The Reasoning Voter: Communication and Persuasion in
Presidential Campaigns. Chicago: University of Chicago Press.
Robinson, John P., Phillip R. Shaver, and Lawrence S. Wrightsman, eds. 1999.
Measures ofPolitical Attitudes. New York: Academic Press.
Sen, Amartya. 2002. Rationality and Freedom. Cambridge, MA: Belknap Press.
Shafer, Byron E., and William]. M. Claggett. 1995. The Two Majorities: The Issue
Context ofModern American Politics. Baltimore: Johns Hopkins University Press.
Simon, HerbertA. 1985. "Human Nature in Politics: The Dialogue ofPsychology
with Political Science." American Political Science Review 79:293-304.
Sniderman, Paul M. 2000. "Taking Sides: A Fixed Choice Theory ofPolitical Reasoning."
In Elements ofReason) ed. Arthur Lupia, Mathew D. McCubbins, and
Samuel L. Popkin. New York: Cambridge University Press.
Sniderman, Paul M., Richard A. Brody, and James H. Kuklinski. 1984. "Policy
Reasoning in Political Issues: The Problem of Racial Equality." American Journal
ofPolitical Science 28:75-94.
Sniderman, Paul M., Richard A. Brody, and Philip E. Tetlock. 1991. Reasoning
and Choice: Explorations in Political Psychology. New York: Cambridge University
Press.
Taber, Charles S., Milton Lodge, and Jill Glathar. 2001. "The Motivated Construction
of Political Judgments." In Thinking about Political Psychology) ed.
James H. Kuklinski. New York: Cambridge University Press.
Tetlock, Philip E., 1999. "Theory-Driven Reasoning about Plausible Pasts and
Probable Futures in World Politics: Are We Prisoners of Our Preconceptions?"
American Journal ofPolitical Science 43:335-66.
Zaller, John. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge
University Press.
Zechman, Martin. 1979. "Dynamic Models of the Voter's Decision Calculus."
Public Choice 34:297-315.