Resumé

Two models of opinion change in a representative democracy are subjected to an empirical test based on data from a series of surveys with voters and members of parliament in Sweden covering the period 1968-1994. An elite-driven opinion model proved to fit the data better than a mass-driven model. Voter opinions shifted more often in the direction of member opinions than vice versa. A run-from-above representation model is more valid for the Swedish case than an idealistic run-from-below model.

Introduction

Many studies of representative democracy, especially American studies, tend to be idealistic. They are pervaded by ideas about influence from below; that people and public opinion somehow have an independent influence on public policy, but they disregard political leadership and opinion formation from above. This tendency is clearly visible within the new area of research on the relationship between public opinion and policy making. The research questions in this area are usually cast in terms of how public opinion affects policy, not in terms of how policy affects public opinion or in terms of how elite opinion have influenced public opinion. Stimson, Mackuen & Erikson's introduction to their recent article on "Dynamic Representation" in The American Political Science Review is a good example: "How, if at all, does the force of public opinion enter into governance in America?", they ask and find that "policy responds dynamically to public opinion change." (Stimson, Mackuen & Erikson 1995, 543).

In a review article on opinion-policy research, Jacobs & Shapiro underscore that "most research has treated the opinion-policy relationship as unidirectional - generally focusing on public opinion's effect on policy makers" (Jacobs & Shapiro 1994, 12). Kuklinski & Segura find that "to most scholars, responsiveness is what representative democracy is all about"

---

* Sören Holmberg, Department of Political Science, Göteborg University, Sprängkullsgatan 19, S-411 23 Göteborg, Sweden.

and conclude that "empirical research on political representation consistently adopted the viewpoint that public opinion is exogenous . . .", i.e., not influenced by elite political opinion formation (Kuklinski & Segura 1995, 4, 17-18). Page & Shapiro, who have conducted the most comprehensive study on the opinion-policy relationship, typically use the responsiveness concept as their theoretical starting point: "The responsiveness of government policy to citizens' preferences is a central concern in normative democratic theory . . „" and they find that "public opinion is often a proximate cause of policy, affecting policy more than policy influences opinion," although they did not examine the extent to which political elites engaged in forming public opinion before the public opinion affected policy (Page & Shapiro 1983, 175; 1992).

Lyn Ragsdale contends that a historical change has taken place in America that is transforming the opinion relationship between the presidency and the public. She quotes Theodore Roosevelt: "I simply made up my mind what they (the people) ought to think, and then did my best to get them to think it," and claims that "unlike Roosevelt, presidents at the end of the century followed rather than created, public opinion." Following Ragsdale, "Reagan modified speeches according to what internal White House polls and focus groups showed" and "Bush received daily poll reports on American attitudes toward Iraq. . . . These polls helped ... to establish the ultimate direction of American involvement. ..." Furthermore, "Bill Clinton's specially prepared White House polls showed that the public felt the president should be more faithful to his centrist campaign promises. In the ensuing days, Clinton made announcements backing a middle class tax cut, denouncing the size of the federal government, and calling for a line-item veto." Ragsdale concludes that,

today's president must take into account the omnipresence of public opinion polls, which present an immediate and continuous plebiscite on their performance. The frequency and visibility of public opinion surveys put pressure on presidents to conform with the known preferences of the public as documented by the polls. Consequently, presidents are more likely to behave as delegates who react to public opinion than as trustees who shape it (Ragsdale 1997, 229-30).

The opinion-policy research approach studies the relationship between public opinion and the collective policy decisions of governmental bodies. This is in contrast to an older American tradition, initiated by Miller & Stokes in the late 19505, which examined representation on a more individual and district-by-district basis, looking for the extent to which constituency opinion affects congressional behavior (Miller & Stokes 1963, 45-63). Despite methodological differences, most studies in this older tradition, including Miller & Stokes', tend to lend support to the findings of the newer opinion-policy approach, i.e., there is, indeed, a relationship between constituency opinion and congressional decision making. However,

the Miller-Stokes model, like most models in the opinion-policy research tradition, is focused on the process of responsiveness and the emphasis is clearly on influence from below. The extent to which congressmen and the political parties are instrumental in forging constituency opinions is not studied within the framework of the Miller-Stokes model. Hence, the Miller model is idealistic and tends to see political decision making and representation as run from below.

European political scientists tend to be more cynical (or realistic), acknowledging the existence of party elites, political leadership and opinion molding from above. The Responsible Party Model, with its essential elitist character, is much more popular in multi-party Europe than in two-party America (Thomassen 1992, 260-75). According to the Responsible Party Model, parties (i.e., party leaders) formulate programs and policies and try to sell them to voters at election time. Such a model clearly involves opinion formation running from the top down. The view of how representative democracy works is as representation run from above (Esaiasson & Holmberg 1996, 3-4).

A rather typical elite model of the pull variant, i.e., a model where elites pull publics, is Deutsch's Cascade Model, which Wessels has found gives a fairly accurate picture of how attitudes toward European integration have developed (Wessels 1995, 161). Deutsch's model consists of different stratification and functional levels. At the top there are the social and political elites. Mass media and nets of opinion leaders are located below and function as mediators to the general public at the bottom level of the cascade. In a review of Deutsch's theory, Putnam concludes that "even in countries where levels of political sophistication are quite high, the 'bubble up' theory is less accurate than the one which Deutsch has termed the 'cascade' model" (quoted from Wessels 1995, 138; Putnam 1976, 138).

Granted that elitist models of representative democracy are more prevalent among European scholars than among their American counterparts, there are, of course, examples of American political scientists who are critical of the push model approach in examining political decision making, i.e., who questions whether publics push elites. A good case in point is Heinz Eulau, who in a critique of the Miller-Stokes model and Converse & Pierce's application of it in France, states that "in all modern representative democracies it is the electorate that responds in a more or less active manner to the elites' policy initiatives, thus indeed having some 'power' ... to hold the elites responsible within a policy framework set by the elites rather than by the citizenry" (Eulau 1987, 210-12; Converse & Pierce 1986). Eulau wants to jettison most of the ideas about influence from below that he thinks "pervade all of the current models of representative democracy." Another famous American critic of the ideas of democracy run from below is Schattschneider, who in the book The Semisovereign People (1960) strongly

argued that political leaders not only define the scope of conflicts but also
how people think about them.

Dichotomous either-or models are always appealing; their simplicity make them easy to understand and to apply. Elite pull or mass push, representation run from above or run from below, elite-driven or mass-driven opinion change, are all examples of such simple black and white models. Naturally, they all come out gray when confronted with empirical data. Representative democracy is never one hundred percent run from above or one hundred percent run from below. All opinion change is not totally elite driven or totally mass driven. Among all the elite pulls there are occasional mass pushes. Consequently, the really interesting scientific question is one of degrees. Given that reality is gray, are we talking about a black shade of gray or a white shade of gray? Are political changes more often elite driven than mass driven? Is representation run from below a more accurate model of democratic decision making than representation run from above?

Kuklinski & Segura, in their review article on political representation studies, end up in a nuanced compromise position of this sort. Their article is based on research on how individuals reach political decisions by among others Zaller (1992), Sniderman, Brody & Tetlock (1991), and Popkin (1991), and concludes that "elites and the information they provide truly do shape individual preferences and, in fact, often create them." But, on the other hand, they continue, "these preferences when aggregated into collective public opinion, become a force with which representatives must reckon." Kuklinski & Segura's final conclusion is Solomonic: "So both perspectives - Miller's and Stokes's and Schattschneider's - are right. At least half right. What contemporary students of public opinion have done is to demonstrate that both halves are needed to complete a picture of political representation." (Kuklinski & Segura 1995, 18).

Extenuating and conciliatory statements cannot hide the fact that what is most needed in representation research is not only more opinion-policy studies but also more thorough analyses of the interplay between elite and mass when it comes to opinion change. Presumably, opinion change on the part of political leaders most often precedes policy change; but not always, of course, since policy changes can be instigated out of opportunistic reasons or be based on anticipations of future mass opinions. However, an analysis of opinion change is an analysis of an early stage in the policy making process. The relationship between voters and elected officials in a representative democracy has never been intended to be static. Many institutional set-ups have been created to guarantee orderly changeability. General elections are the best example. The free formation of opinions and the elite recruiting nomination process are other examples. Hence, opinion change on the elite level as well as on the mass level can be achieved in many different ways. The most important question, however, is the democratic question whether a

dynamic interplay between political leaders and citizens in terms of opinion
formation actually exists, and if so, is the process most often elite driven or
most often mass driven?

To a degree the quality of democracy is at stake - there should be a dynamic interplay, a dialogue, between elite and mass opinions in a democracy, but whether the dialogue is run from above or from below is less decisive. In a society with freedom of expression, elite-driven as well as mass-driven opinion changes are legitimate democratic processes. The reason we are sometimes suspicious of opinions molded from above is that in many cases resources to exercise the freedom of expression are very unevenly distributed. In the information age, wealthy, eloquent people with access to the best PR people have an obvious advantage.

However, in pure theory, the role of political leaders and parties in the opinion forming process cannot be treated as suspect, as if it were somehow less autonomous than such other opinion forming factors as the media, friends, or personal experiences. Obviously, manipulation of opinions happens, but parties and candidates are not the only, and perhaps not the most frequent, perpetrators when it happens. Media, interest groups, and business interests are also prominent players on the modern opinion market. But, granted that outside the world of pure theories, politicians are no snow white virgins in the game of opinion manipulation and that elite groups almost always have more resources than the atomized masses, it is perhaps no surprise that mass-driven processes usually are perceived as somewhat more democratic than elite-driven processes. Representation run from below is closer to the original notion of direct democracy. Representation run from above may sound a little suspect to some, like a therapeutic democracy where people may air their grievances and elect representatives as long as the people do what the leaders tell them. A less sinister way of portraying a system where most representation is run from above is to acknowledge the fact that we are talking about a very familiar system, at least in Europe, where cohesive parties, ideological leadership, and accountability - not in advance, but afterwards - are the operative characteristics.

The Swedish Case

Dynamic representation is a new field of study, and as such in dire need of theoretical as well as empirical work. The purpose of this article is to alleviate some of that need when it comes to the empirical study of opinion change and the interplay between elite and mass. In contrast to most other examinations of dynamic representation, our setting is European, not American. Our data comes from an ongoing research program on political representation

in Sweden (Särlvik 1969; Holmberg 1974; Holmberg 1989, 1-36;
Holmberg 1996a; Esaiasson & Holmberg 1996).

As a Western European country, the Swedish political system differs from the American system. In contrast to the American experience, Sweden has a multiparty parliamentary system with extremely cohesive legislative party groups. Furthermore, Swedish electoral procedures, basedon multi-member districts and proportional representation, put the parties, not the individual candidates, at center stage. Consequently, the Swedish system, perhaps even more so than most other European systems, is a system where a kind of collectivist representation model centered on the political parties is more viable than various forms of individualistic representation models. Given the Swedish history of strong and cohesive parties and a tradition of ideological leadership, we might also expect a fair number of opinion change processes espousing a pattern of representation run from above, i.e., cases where opinion changes are more elite driven than driven by the public.

Our data originate from a series of Swedish representation studies performed in the years 1968/69, 1985, 1988, and 1994. In all instances, the data consist of a study including members of the Riksdag and interviews with a sample of the electorate. The Riksdag member studies were done in collaboration with the Swedish Election Studies Program. The 1969 Riksdag Study involved personal interviews with all the members of the Second Chamber; the response rate was 97 percent. The studies in 1985, 1988, and 1994 were done by mail questionnaires sent to all members of the now unicameral Riksdag; the response rate was still 97 percent in all three surveys. The election surveys of the electorate comprised approximately 3000 personal interviews each year with a response rate of about 80 percent (Holmberg 1994, 309-22).

A major advantage of our data sets is that they contain a large number of issue questions put in a similar, or as in most cases, in an identical fashion, to samples of voters as well as to members of the Riksdag. A fair number of those issue questions have been asked more than once to the public as well as to members, making it possible to study opinion change and the possible interplay between members and the electorate.

Issue Congruence

We start the analysis by looking at some baseline results concerning the degree of issue congruence between members and the electorate and the extent to which it has changed since our studies started in the 19605. Issue congruence between leaders and the public is of obvious importance in a representative democracy. As Pitkin said in the modern classic, The Concept of Representation: "The representative . . . must not be found persistently at

odds with the wishes of the represented . . ." (Pitkin 1967, 209-10). The results in Table 1 indicate that the opinions of Swedish MPs to an astonishing degree differ from those of the Swedish public, but that the degree of issue agreement has improved somewhat compared to the 19605.

Issue congruence has been measured in three different ways: with the help of the average difference between mean issue positions, with the help of the average summed percentage difference between dichotomized opinion distributions, and by examining the proportion of issues displaying different majority positions among members and the electorate. Compared across issues or parties, the three measures yield roughly identical results. If we

compare the degree of issue congruence in the late 1960s with the results for the 1980s and 1990s for the Riksdag as a whole and for the entire electorate, it is apparent that congruence has improved. The percentage difference measure indicates that the average percentage difference for dichotomized issues was 14 percent in 1994 compared to 20 percent in 1968/69; a substantial and statistically significant change for the better.

When we look at the results for the individual parties, it is evident that all parties, except one, also experienced an improved issue congruence between its MPs and its voters. The exception is the Conservative Party which exhibits worse issue agreement figures in the 1990s than in the 19605. Compared to the 19605, the Conservative Party of Sweden has clearly become a more ideological party, trying to impress voters with a neo-liberal market message. Many of the other parties, most noticeably the Social Democrats and the Left Party (former Communists), have become more pragmatic and moved toward the political center, and as a consequence their issue agreement figures have improved.

Given all the changes in Swedish politics and in the laws governing
elections since the 19605, it is not surprising that the degree of issue
agreement between leaders and the public has increased. Most of the changes

- the constitutional reform in the early 1970s which instituted more frequent elections and a stricter party proportionality in the distribution of parliamentary seats, the heightened importance of campaigns, including more media coverage, and the increased tendency among voters to switch between parties and to issue vote - would all be expected to enhance rather than diminish the possibilities of an improved opinion representation in Sweden (Gilljam & Holmberg 1995).

Granted that the level of issue congruence has increased compared to thirty years ago, it is nevertheless still true that the Swedish congruence level, in absolute terms, is not particularly impressive. For example, the extent of issue agreement between elite and mass is no higher in Sweden than in countries like USA, Germany, France or The Netherlands (Holmberg 1996b). It is about the same, which might come as a surprise given the common perception of Sweden as a more egalitarian society with shorter distances between leaders and followers than in many other non countries. Even if these perceptions are true, the realities behind them have not produced higher levels of issue congruence in Sweden than in other countries.

Furthermore, a Riksdag chosen by a totally random procedure, without resorting to the costly and time consuming processes of candidate nominations, election campaigns and popular voting, would almost always produce higher degrees of issue agreement than the ones we have found in our empirical studies. The probability of getting the kind of average summed percent differences between elite and mass opinions as we found in 1968/69,

1985, 1988, and 1994 (between 13 and 20 percentage points) would be less than one in a thousand if members were chosen randomly. Against this backdrop, it is easy to conclude that the level of issue congruence between Swedish MPs and the electorate is mediocre at best.

An Elite Polarized Model

A look at how the parliamentary party groups represent their voters' views on the all-important left-right issues further underscores our point. On average, issue congruence tends to be somewhat better on left-right issues than on most non-left-right issues. In a comparative perspective, Swedish politics is unusually unidimensional and dominated by the left-right dimension. Hence, it is not astonishing that issue agreement between leaders and the public is above average on the highly politicized and much debated left-right issues. The results in Figure 1 illustrate how opinion representation on the left-right dimension has changed since the 19605.

The analysis rests on four constructed left-right indices, one per study, based on concrete left-right issue questions. Members and voters were asked the same questions on each occasion, but the issue questions differed somewhat across time, which made exact comparisons of opinion averages difficult. Nevertheless, the results can be compared in a general way, since the findings are relatively robust, based as they are on a rather large number of issue questions; a replacement of one or two questions do not appreciatively change the results.

The results reveal that a rather profound change has taken place in terms of how the Riksdag parties represent the opinions of voters on the left-right dimension. A representation model characterized by elite left divergence in the 19605, where party MPs tend to be to the left of their voters, has been replaced by an elite conflict model in the 1980s and 19905, where opinion differences are greater among party elites than among party voters. In the new elite conflict model, Socialist and Green Party MPs are to the left of their voters, while non-Socialist members are to the right of theirs. The change is most profound for the Conservative Party which in the late 1960s used to have MPs who were very close to their voters on the left-right dimension, although somewhat to the left. In the 19905, the Conservatives in the Riksdag are clearly to the right of their voters, exhibiting a larger distance to their voters than any other parliamentary party group.

The elite conflict model, in which much larger opinion contrasts are found among elected representatives than among their voters, is not a Swedish specialty, however. In fact, it is the model most often detected in Western representation studies (McClosky, Hoffman & O'Hara 1960; Miller & Jennings 1986; Herrera, Herrera & Smith 1992; Thomassen 1976; Herzog,

Rebensdorf & Wessels 1990; Converse & Pierce 1986; Barnes 1977; Dalton 1985; Huber & Powell 1994; Westerståhl & Johansson 1981). As a matter of fact, Australian theorist Peter Medding has claimed that there is what amounts to a functional necessity for party elites to make more differentiated

appeals in order to present policy options clearly to the public and to avoid
loss of support (Medding 1982, 393-412, quoted from McAllister 1991,
237-68).

The rather low degree of issue congruence between Swedish MPs and the electorate as well as the prevalence of a pronounced elite conflict model on left-right issues do not suggest that a mass-driven representation model would be the most accurate depiction of how Swedish democracy works. Quite the opposite, the results so far indicate that an elite-driven model with representation run from above would be more accurate. But, of course, the viability of the elite-driven versus the mass-driven models cannot be determined based on results solely derived from cross-section studies. Time series data and an examination of elite and mass interplay when it comes to opinion change are imperative.

Elite or Mass Driven Opinion Change?

Our data from the four Swedish representation studies are comprehensive enough that we can analyze different patterns of elite-mass opinion change in order to determine whether they primarily fit an elite-driven or a massdriven process. We have eighteen issues for which opinions of MPs and the electorate have been solicited on at least two occasions, for example 1968/69 and 1985 (two issues) or 1988 and 1994 (nine issues). To illustrate how the data look and how they can be analyzed, the results in Figure 2 show how opinions have changed across time on three issues among MPs as well as in the general public.

For the sake of simplicity, we measure the opinions of voters and representatives on the basis of the proportion with certain attitudes. A change in opinion implies that the proportion of voters or members with the indicated point of view increases or decreases. In the case where the aggregate opinion position of members shifts towards that of the public, at the same time that the opinion of the public is either stable or reinforced in the same direction, we speak of a mass-driven representation process. Elitedriven representation, on the other hand, involves opinion shifts in the public in the direction of member opinion. The causal process behind the changes is of no consequence for our classification. The only thing that is of consequence, and the only thing we can measure given our data, is whether the opinion changes are compatible with a potential elite-driven or a potential mass-driven opinion molding process.

The opinion change on the issue of the size of social benefits between the years 1968/69 and 1985 indicates a potentially mass-driven process. The attitudes among MPs have changed in the direction of where the voters were already in 1968. Like the electorate, members have become more negative

toward social benefit programs. In this case, the extent of issue congruence increased somewhat as a consequence of the opinion change, but that is not a requirement in mass or elite-driven representation processes. The only requirement is that the opinion changes among MPs and voters go in the same direction, i.e. that they are parallel, or stable at one level and closing in on the other level.

The results for the attitude changes between 1985 and 1994 in relation to the coming high tech society illustrate a potentially elite-driven opinion forming process; the public followed the elite and became more positive towards computers and industrial robots. Our third example, the short-term opinion changes on the nuclear power issue between 1985 and 1988, is a case where the attitude shifts in elite and mass were non-parallel. Member opinion became more negative towards nuclear power while, at the same time, the general public became more positive. The process could be called disconnecting dynamic representation, since not only did opinions move in different directions, the outcome was also a decreased degree of issue agreement. Elite and mass opinions moved apart and the attitude changes were divergent and non-parallel.

Table 2 summarizes the results of the examination of the opinion changes on all eighteen issues. The total number of cases, i.e., an issue measured at least at two time points among MPs as well as in the electorate, are thirtytwo for 1968/69-1985, eight for 1985-1988, nine for 1988-1994, and eleven for 1985-1994. Broken down by political party and summed across the parties, the total number of cases is one hundred fifty nine; five parties times thirty cases gives one hundred fifty cases plus nine cases for the Greens from 1988-1994. The Christian Democrats are not included in the analysis.

A sizable majority of our cases of opinion change among Swedish MPs
and voters have been parallel. Most often, elite and mass opinion winds have

blown in the same direction. Looking at the Riksdag and the electorate as wholes, not broken down by party, no less than 83 percent of the opinion change cases were parallel, and in most instances the process was potentially elite driven (50 percent). The proportion of cases compatible with a potentially mass-driven process was 33 percent.

Non-parallel opinion changes were more common within the parties, i.e.,
when we compare the attitude shifts among MPs and voters of the same party

- 37 percent compared to only 17 percent for the Riksdag and the electorate as wholes. However, the intra-party results are statistically less reliable than the results for all MPs and the entire electorate, since some of the parties are quite small and comprise rather few interviewed persons. Consequently, a conclusion that intra-party opinion forming interplay between elite and mass tends to be less orderly and more chaotic has to be drawn with some caution. On the other hand, all the parties, including the Social Democrats and Conservatives, the two big ones, exhibit a larger proportion of non-parallel opinion changes than the Riksdag and the entire electorate taken as wholes. So, maybe we have found the first indication that the representation system as a whole functions more efficiently in creating parallel elite and mass opinion changes than its most important subparts, the political parties. Some misrepresentations on the part of individual parties could even out when aggregated to the level of the entire parliament.

An example of when this evening out process was working effectively was on ideological representation in the 1980s and 1990s on the left-right dimension. The elite polarized congruence pattern on the left-right continuum meant that most parliamentary party groups represented their voters' leftright positions rather poorly, while the Riksdag as a whole did a better job of reflecting the center of gravity of the electorate's position. The left-right mean position of the elite polarized parliament ends up close to the mean position of the less polarized left-right mass distribution. However, in the 19605, there was no evening out-process in left-leaning parliamentary party groups.

Going back to our opinion change data for the parties in Table 2, it is evident that the elite-driven process occurred somewhat more frequently for the parties which, on most issues, hold flank positions, i.e., for the Left Party, the Greens, and for the Conservatives. Presumably, these parties have a larger stake in molding opinions than the center parties on most issue dimensions. Consequently, we would expect the middle of the road parties to exhibit more instances of mass-driven opinion change patterns, and that is precisely what we find. The Social Democrats, the Center Party, and the Liberals reveal a larger proportion of potentially mass-driven opinion changes than the flank parties. The differences are not overwhelming, but they are there and they fit the hypothesis.

Another difference which is small, but which also follows a plausible hypothesis, is the difference between results for short-term and long-term opinion changes. At least normatively, one would expect mass-driven processes to play a more prominent role over the long term. Sustainable opinion molding by political leaders over long periods of time is somewhat difficult to reconcile with traditional notions of democracy. If the political leaders lead all the time and the followers follow patiently all the time, our democratic ideals get strained.

Strained ideals or not, our data show the perhaps comforting result that mass-driven opinion changes were somewhat more frequent when we studied opinion forming processes over longer periods (between nine and seventeen years) than when we looked at attitude shifts over shorter periods (between three and six years).

For the Riksdag as a whole, a mass-driven model fit 39 percent of the cases involving a long-term opinion change, and 29 percent when the time period was short term. The comparable summed results for the political parties are 37 percent for long-term opinion change and 29 percent for shortterm. Admittedly, the differences are small and merely suggestive, but they are in accordance with the hypothesis.

Hannah Pitkin established the rather obvious democratic claim that "representatives . . . must not be found persistently at odds with the wishes of the represented . . ." (Pitkin 1967, 209-10). In the long run, congruence between elite and mass on politicized issues should eventually occur. A representative democracy where leaders and voters most of the time change opinions according to some kind of disconnecting dynamic model is not what is usually meant by an ideal democracy. In our Swedish case, with elite and mass attitudes measured across time on eighteen issues, congruence increased over time in a majority of cases (53 percent) for the Riksdag and the electorate taken as wholes. If we concentrate on the cases covering the longest time periods (nine years and up), the proportion of cases exhibiting an increased degree of issue agreement becomes an even more impressive 69 percent. Evidently, issue congruence is produced over time, especially over longer time periods.

The other side of the coin is that issue congruence is less effectively achieved over shorter time periods; in our case only 41 percent of the shortterm opinion interplays between all leaders and the electorate resulted in increased levels of issue agreement. The comparable intra-party result was about the same, 43 percent. Parties were somewhat more successful in creating intra-party issue congruence over longer time periods, but only to a very limited degree. Among our cases of intra-party, long-term opinion changes, only 45 percent led to an increased level of issue agreement. Like we found previously, compared to the representative system as a whole, individual parties, as subparts of the system, are less effective at producing issue congruence. The whole is more efficient than the sum of its parts.

A look at which process, the elite-driven or the mass-driven, is most effective at producing increased levels of issue agreement, yields a result that could be disturbing to people who subscribe to an idealistic view of how democracy works. As it turns out, the elite-driven process proves to be most effective, not only for the system as a whole, but for the intra-party processes as well. The results in Table 3 show that 65 percent of all potentially elitedriven, intra-party opinion changes led to increased levels of issue congruence. The comparable result for potentially mass-driven, intra-party attitude shifts was only 35 percent.

A similar, but less pronounced outcome is obtained for the entire Riksdag and the electorate. An increased degree of issue congruence is produced in a majority of cases where the opinion change was potentially elite driven (53 percent). When the opinion change was potentially mass driven, the proportion of cases with increased levels of issue agreement was lower, about 40 percent. Hence, Swedish elite-driven opinion formation has been successful in one very important aspect: instigated opinion shifts have tended to lead to improved levels of issue congruence between leaders and voters, while, in a majority of cases, mass-driven opinion changes have resulted in diminishing degrees of issue agreement.

Political Representation Run from Above

Top down opinion formation is a legitimate form of politics in a democracy characterized by free and fair elections and freedom of speech. Representation run from above may sound a bit sinister, but it exists and we should not shy away from it even if the phenomenon is not espoused in standard text

books on democracy. Furthermore, we are not talking about a pure model, a totally black model of one hundred percent elite-driven attitude shifts in contrast to a white idealistic model of only mass-driven opinion changes. As the Swedish results proved, the empirical reality is more gray and less clearcut. Opinion changes in Sweden were not totally elite driven. We encountered many instances of opinion shifts that were potentially mass driven, especially on issues where we could measure attitude changes over longer time periods.

However, and that is our main conclusion, our Swedish study indicates that run-from-above models must be taken seriously. If we had to choose, in the Swedish case, a run-from-above representation model would be more valid than a mass-driven model. Formulated more harshly, Swedish politicians, to the extent that they seek public support at all, primarily obtain the citizens' consent to carry out policies which they themselves have defined and sold on the opinion market.

Consequently, we must conclude that Swedish politics, and probably European politics more generally, are more elitist than US politics; at least if we can believe the results from most studies on American representation, which tend to emphasize the relevance of the run-from-below model in the US. Perhaps we have stumbled upon a genuine difference between parliamentary systems and those based on complex separation of powers procedures. It is conceivable that US politicians are more sensitive to changes in public opinion than their European counterparts; the avalanche of opinion polls in US politics may be a case in point. In parliamentary systems with cohesive and ideological parties, as a contrast, it is equally legitimate to justify policies with reference to decisions taken by party congresses as with references to public opinion. Indeed, Swedish politicians trying to justify a controversial decision by referring to public opinion are often accused of superficial populism (Esaiasson & Holmberg 1996, 318-19).

Representative democracy is a delicate system fundamentally built on trust and a fine-tuned balance between political leadership and responsiveness (Holmberg 1996b, 13). Too much leadership leads to elitism, too much responsiveness leads to populism. Our results from Sweden, a European parliamentary democracy, indicate that the scale is somewhat tilted. Political leadership and representation run from above carry more weight than responsiveness and representation run from below.

REFERENCES

Barnes, S. H. 1977. Representation in Italy: Institutional Tradition and Electoral Choice.
Chicago: University of Chicago Press.

Converse, P. E. & Pierce, R. 1986. Political Representation in France. Cambridge, Mass.: The
Belknap Press of Harvard University Press.

Dalton, R. J. 1985. "Political Parties and Political Representation. Party Supporters and Party
Elites in Nine Nations," Comparative Political Studies 18, 267-99.

Esaiasson, P. & Holmberg, S. 1996. Representation from Above. Members of Parliament and
Representative Democracy in Sweden. Aldershot: Dartmouth.

Eulau, H. 1987. "The Congruence Model Revisited," Legislative Studies Quarterly 12, 171—
214.

Gilljam, M. & Holmberg, S. 1995. Väljarnas val. Stockholm: Norstedts Juridik.

Herzog, D., Rebenstorf, H., Werner, C. & Wessels, B. 1990. Abgeordnete und Burger.
Opladen: Westdeutscher Verlag.

Holmberg, S. 1974. 'Riksdagen representerar svenska folket.' Empiriska studier i
representativ demokrati. Lund: Studentlitteratur.

Holmberg, S. 1989. "Political Representation in Sweden," Scandinavian Political Studies 12,
1-36.

Holmberg, S. 1994. "Election Studies the Swedish Way," European Journal of Political
Research 25, 309-22.

Holmberg, S. 1996a. "Svensk asiktsöverensstämmelse." In Rothstein, B. & Särlvik, B. (eds.),
Vetenskap om politik. Festskrift till professor emeritus Jörgen Westerståhl. Göteborg:
Statsvetenskapliga institutionen.

Holmberg, S. 1996b. Policy Congruence Compared. Göteborg: Statsvetenskapliga institutionen.

Holmberg, S. & Esaiasson, P. 1988. De folkvalda. Stockholm: Bonniers.

Huber, J. & Powell, B. 1994. "Congruence Between Citizens and Policymakers in Two
Visions of Liberal Democracy," World Politics 46, 291-326.

Jacobs, L. R. & Shapiro, R. Y. 1994. "Studying Substantive Democracy," PS: Political
Science & Politics 27, 9-17.

Kuklinsky, J. H. & Segura, G. M. 1995. "Endogeneity, Exogeneity, Time, and Space in
Political Representation," Legislative Studies Quarterly 20, 3-21.

MacAllister, I. 1991. "Party Elites, Voters and Political Attitudes: Testing Three Explanations
for Mass-Elite Differences," Canadian Journal of Political Science 24, 237-68.

McClosky, H., Hoffman, P. J. & O'Hara, R. 1960. "Issue Conflict and Consensus among Party
Leaders and Followers," American Political Science Review 54, 406-27.

Medding, P. Y. 1982. "Ruling Elite Models: A Critique and an Alternative," Political Studies
33, 393^52.

Miller, W. E. & Stokes, D. E. 1963. "Constituency Influence in Congress," American Political
Science Review 57, 45-56.

Miller, W. E. & Jennings, K. M. 1986. Parties in Transition. New York: Rüssel Sage
Foundation.

Page, B. I. & Shapiro, R. Y. 1983. "Effects of Public Opinion on Policy," American Political
Science Review 77, 175-90.

Page, B. I. & Shapiro, R. Y. 1992. The Rational Public: Fifty Years of Trends in Americans'
Policy Preferences. Chicago: University of Chicago Press.

Pitkin, H. 1967. The Concept of Representation. Berkeley: University of California Press.

Popkin, S. L. 1991. The Reasoning Voter. Communication and Persuasion in Presidential
Campaigns. Chicago: University of Chicago Press.

Putnam, R. D. 1976. The Comparative Study of Political Elites. Englewood Cliffs, N.J.:
Prentice Hall.

Ragsdale, L. 1997. "Disconnected Politics: Public Opinion and Presidents." In Norrander, B.
& Wilcox, C. (edsj, Understanding Public Opinion. Washington, D.C.: CQ Press.

Schattschneider, E. E. 1960. The Semi-Sovereign People. Hinsdale, 111.: The Dryden Press.

Stimson, J. A., Mackuen, M. B. & Erikson, R. S. 1995. "Dynamic Representation," American
Political Science Review 89, 543-65.

Särlvik, B. 1969. Representationsundersökningens forskningsprogram. Relationer mellan
väljare och valda på rikspolitisk nivå. Göteborg: Statsvetenskapliga institutionen.

Thomassen, J. 1976. Kiezens en gekozen in een representatieve demokratie. Alphen aan den
Rijn: Samson.

Thomassen, J. 1992. "Empirical Research into Political Representation. A Critical
Reappraisal." In Klingemann, H-D., Stöss, R. & Wessels, B. (eds.), Politische Klasse und
Politische Institutionen. Opladen: Westdeutscher Verlag.

Wessels, B. 1995. "Support for Integration: Elite or Mass Driven?" In Niedermayer, O. &
Sinnott, R. (eds.), Public Opinion and Internationalized Governance. Oxford: Oxford
University Press.

Westerståhl, J. & Johansson, F. 1981. Medborgarna och kommunen. Studier av medborgerlig
aktivitet och representativ folkstyrelse. Stockholm: DSKn 1981:12.

Zaller, J. R. 1992. 77te Nature and Origins of Mass Opinions. New York: Cambridge
University Press.