Introduction

In the last decade or so we have witnessed a proliferation of research on European Union (EU) public opinion. The initial Danish rejection of the Maastricht Treaty appeared to signal the awakening of public interest in the EU, the demise of the 'permissive consensus' and its replacement with what Hooghe and Marks (2005) term a 'constraining dissensus.' Most work to date has focused at the individual level on the identification of those factors that explain attitudes towards the EU (e.g. Gabel, 1998; Sanchez-Cuenca, 2000; Rohrschneider, 2002; Marks and Hooghe, 2004). What has gone largely unexamined, however, is the distribution of those attitudes. In particular, few scholars have sought to systematically investigate whether the extent of public disagreement over the European project has grown over time. Yet this is an issue of importance not only for the future of the EU but also for the domestic politics of the member states.

There is widespread recognition that in the post-Maastricht period Euro-scepticism has become a more significant phenomenon than in earlier decades.¹ This may simply reflect a decline in the average level of support for the EU. But, the rise in Euro-scepticism may also, or alternatively, be a function of a decline in consensus — that is, an increase in the dispersion of attitudes, along with a flattening or even polarization of opinion distribution. If we are to meaningfully talk of a shift from a 'permissive consensus' to a 'constraining dissensus,' we need to identify which of these two has taken place — that is, whether there has been a movement of the distribution in the issue space or a change in the shape of the distribution. In addition, given the centrality of Maastricht to the perception of a change in public attitudes, we need to establish whether the Treaty was indeed instrumental in moving or changing the shape of opinion distribution.

This question is doubly important because the shape of the distribution is also crucial to political contestation. While studies of public opinion typically focus on the central tendency of the distribution (e.g. Downs' (1957) 'median voter theorem'), the shape of the distribution can also tell us much about the likely nature of political competition. The more dispersed are attitudes, the more problematic it is to achieve consensus, not least because it creates electoral incentives for parties to locate at the extremes. Moreover, where public attitudes are widely dispersed, parties locating at the center of the distribution may find themselves confronting internal divisions and pressures to relocate in the issue space. A change in the shape of the distribution of opinion on the EU may thus have implications not only for the long-term trajectory of the Union but also for domestic politics.

In the following, we first outline why we believe attention to the extent of disagreement is warranted. We then briefly elaborate on why Maastricht and the post-Maastricht process of political integration may have led to the commonly postulated decline in consensus on the EU. To investigate the question, we employ a simple model that allows us to observe the average annual change in the mean, variance and kurtosis of support for the EU — the three moments of the distribution that define its shape. To briefly presage our findings, we do indeed find a significant decline in consensus as a consequence of Maastricht, this holds both across the EU as a whole and in almost every one of the member states that were members pre-Maastricht. Yet, in the post-Maastricht period, the degree of consensus is actually higher across the EU as a whole, and in a majority of member states, than was the case in the 1970s. There is also less difference between countries in the shape of their distributions today than 30 years ago. We then discuss the implications of the distributional shifts for national and EU politics.

Public Opinion

The proportion of the EU population that is opposed to further integration has increased in the period since the Masstricht Treaty and there now appear to be sizeable numbers of people either hostile to or ambivalent concerning the European project. The proportion of the EU population perceiving their country's membership to be a 'bad thing' increased from 9% in 1990 to 11% in 2002, while the proportion that perceived membership as 'neither good nor bad' or that simply did not know increased from 21% in 1990 to 36% in 2002 (Mannheim Eurobarometer Trend File, 1970–2002). Yet, a majority of the public, 53% in 2002, still viewed membership to be a 'good thing,' though this is markedly lower than the 70% holding the same view in 1990 (ibid.). These data would seem to suggest that in the 1990s public begun to diverge in assessments of the national value of EU membership — that is, public consensus on the EU declined.

The possibility of an emergent dissensus is also evinced at the level of political parties. As Taggart (1998), and, Taggart and Szczerbiak (2002) document, there are now an array of Euro-sceptical single-issue and protest parties, with Spain the sole member of the EU-15 not to have such a party. Moreover, while most of the mainstream parties, from the center-left to the center-right, remain supportive of the EU, many have witnessed the emergence of internal Euro-sceptical factions. Taggart (1998) lists six cases of such faction formation within mainstream parties in Finland, Ireland, Germany, Portugal, Sweden and the UK. Steenbergen et al. (2007, 20) would add to that list parties in Denmark and the Netherlands. And, as Gabel and Scheve (2007) note, drawing on expert surveys of EU political parties, while average levels of intra-party dissent are relatively low they nevertheless increased between the 1980s and 1990s, and, significant dissent is not uncommon, evidenced in more than 30% of the EU's political parties.

The decline in public perceptions of the benefits of EU membership coupled with the development of Euro-sceptical parties and party factions raises the question of whether public opinion is becoming more divided on the issue. If a decline in consensus is defined simply as a decline in the mean level of support then indeed this has occurred. If a decline in consensus is defined as an increase in disagreement, however, then change in the mean level of support tells us little, since it may equally be accompanied by an increase, a decrease or no change in the extent to which attitudes are congruent. In one scenario, it is possible that the distribution of public opinion has remained unchanged over the last two or three decades. In this case, the mean level of support may simply have declined in the 1990s as the distribution has moved in the issue space. In turn, the emergence of Euro-sceptical parties and party factions may simply reflect the growth in the proportion of the population with Euro-sceptic attitudes.² Yet, the emergence of Euro-scepticism need not reflect a decline in public consensus if the entire distribution has moved. In other words, if the entire population has become more Euro-sceptical the level of consensus may not have changed, just because positive evaluations of EU membership have declined. Alternatively, the public may have become increasingly divided in terms of evaluations of the EU. To the extent that this is the case, attitudes may have become more dispersed, the distribution of opinion may have become flatter, or even bimodal, and thus there may have been a decline in consensus. Moreover, as we elaborate below, the mean level of opinion may be identical across a range of different distributions, but different distributions can have quite different political implications.

As DiMaggio et al. (1996) note in their study of the polarization of social attitudes in the US, all else equal, as opinion becomes more dispersed it becomes more difficult to maintain a centrist political consensus. Thus, for example, an increase in the dispersion of public opinion increases the incentive for political parties not located at the center of the distribution to attempt to capitalize on that dispersion. In the case of the EU, to the extent that the dispersion of opinion increases, parties adopting stridently pro-EU or, as appears to be more pertinent, stridently anti-EU positions have a greater incentive to push the issue to the forefront of political debate. That is, they have an incentive to increase the salience of the EU issue for purposes of domestic political contestation. Indeed, as Steenbergen et al. (2007, 18) note, this appears to have been the case with Haider in Austria, Le Pen in France and Wilders in the Netherlands.

In addition, as opinion becomes more dispersed a mainstream party adopting a centrist position (i.e. centrist relative to the distribution) risks losing support as voters are increasingly located a greater distance from the party. Such parties may also experience growing internal party divisions as the distance between the party's position and that of some party members increases. In contrast, mainstream parties face less of a threat to their support if the distribution of opinion simply moves in the issue space in a more Euro-skeptical direction and the party follows that movement (i.e. becomes more Euro-sceptical, while maintaining a stance that is centrist relative to the distribution). Movement of the distribution in the issue space towards greater Euro-scepticism has clear implications for the future of the EU. An increase in the dispersion of public attitudes, however, has implications for both the future of the EU and for domestic political contestation.

Furthermore, if the distribution of opinion becomes flatter, or at the extreme, takes on a bimodal configuration — reflective of attitudinal polarization — the issue may constitute the basis for a domestic political cleavage. Thus, for example, not only may extreme parties locate at the tails of a distribution, mainstream parties may themselves be impelled to move away from the center and towards the tails if the distribution becomes bimodal. In turn, this implies the possibility of a widening of the distance between mainstream parties if they move towards opposite tails. Again, a change in the shape of the distribution has the potential to effect domestic political contestation in a way that the mere movement of the distribution does not.

A bimodal distribution might seem unlikely in the EU case given the generally high levels of support for the EU, and thus the likelihood that the distribution is single peaked and heavily skewed. Yet, as we illustrate below, this is not always and everywhere the case. In some countries and some years there is clear evidence of bimodality in the distribution of attitudes — that is, evidence of polarization on the EU. Moreover, even where the distribution of opinion is not clearly bimodal we cannot ignore the possibility that it has become flatter over time, and thus that the proportion of the population at the center of the distribution has declined while the proportions in the tails have increased. Identifying movement in this direction is important, even if the distribution does not clearly display a state of polarization.

The importance of the distribution of opinion for domestic and EU politics is obviously contingent on the extent to which the issue is salient, and to date there is no clear consensus on the topic. EU elections are commonly regarded as 'second-order national' elections since, even in what are clearly elections concerning the EU, the focus is predominantly on national issues rather than integration issues (Van der Eijk et al., 1996). Moreover, the EU issue dimension appears to be orthogonal to the left–right dimension, and the latter is the critical dimension structuring politics within the member states.

Yet, a number of recent studies indicate that the EU has indeed played a role in domestic electoral competition (e.g. Evans, 1999; Carruba, 2001; Tillman, 2004; Evans and Butt, 2007). And, as Steenbergen and Scott (2004, 189) stress, among parties in the EU-15 there are decreasingly few for whom integration is of no importance or only a minor issue — expert surveys indicate the number declined from 24 in 1984 to just eight in 1996, while the number of parties that considered integration to be one of the most important or the most important political issue increased fractionally from 28 in 1984 to 30 in 1996. Additionally, the increasing tendency for EU policy decisions to impinge on the lives of individuals, and, the tendency of governments to scapegoat the EU for unpopular policy decisions, suggests that that saliency is only likely to increase in the future.

Furthermore, as van der Eijk and Franklin (2004, 32–33) stress, due to its very orthogonality with the left-right dimension the EU issue may well constitute something of a 'sleeping giant.' The mainstream parties have failed to fold the issue into the left-right dimension, as is typically the case with new domestic political issues. What is more, the experience in the Netherlands and France of the referenda on the proposed Constitution suggests that parties find it far more difficult to marshal their supporters in EU referenda than at national elections. Consequently, the EU issue has the potential capacity to undercut the bases for contemporary party mobilization in many member states (van der Eijk and Franklin, 2004, 33). If the distribution of public opinion affects the incentive of political elites to activate the EU issue, then it may well play a part in determining whether, and if so when, that 'sleeping giant' is roused.

Central to both the perception of a decline in consensus on the EU and the increasing salience of the EU, are the Maastricht Treaty and the post-Maastricht emphasis on political integration. Economic and Monetary Union, coupled with moves to integrate policy in the areas of foreign relations, security, defence, immigration, asylum and policing, all constituted integration of an unequivocally political nature. Simultaneously, the adoption of qualified majority voting in the Council, and, the expansion of Parliament's powers via the cooperation and co-decision procedures underpinned an effective increase in supranational authority. Maastricht also 'endowed the EU with political objectives and constitutional characteristics,' such as EU citizenship, and, the charge to uphold democracy, human rights and national identities (Dinan, 2004, 291).

The rancorous and protracted debates over the Treaty and, in varying degrees, over ratification at referenda in Denmark, Ireland and France, raised the profile of both the political content of integration and the increase in supranational authority; concerns that were further echoed in the subsequent resurgence in debates over the 'democratic deficit' in the years immediately after Maastricht. By taking on a more explicitly political character, the integration process raised in a sharper and more acute fashion the relationship between national sovereignty and national identity, on the one hand, and European integration, on the other. Accordingly, while the process may have remained somewhat obscure, distant and not well understood, there can be little dispute that debate in relation to the process has become both more important and more accessible to the European public. Moreover, subsequent developments, in particular the growth in Euro-scepticism and the recent failure to ratify the Constitution in France and the Netherlands indicate public concern has not abated.

If the Maastricht Treaty was instrumental in fostering greater public concern over the integration process and the development of a 'constraining dissensus,' it is at this point that we should expect to see the beginning of a decline in support for the EU. Yet, whether there has been a decline in consensus, conceived as a greater dispersion in attitudes and/or a polarization of opinion, is less clear-cut, since this is not an automatic consequence of a decline in the average level of support. In the following analyses, we therefore examine change in the shape of opinion distribution over time and also, critically, the impact of Maastricht on the shape of the distribution.

Measurement, Data, and Models

The shape of a distribution can be characterized by three moments, its mean, variance and kurtosis. The mean of the distribution addresses the possibility that the distribution of opinions has shifted such that overall levels of support for the EU are more negative or positive. In this case, growing Euro-scepticism is a function of a general disapproval of integration rather than a growing divide in the electorate. The variance and kurtosis of the distribution provide measures of the degree of consensus in the electorate. The two measures tap, respectively, change in the dispersion of attitudes and movement towards or away from bimodality. The higher the variance the greater the dispersion of attitudes and thus the greater the likelihood that any two randomly selected individuals will have different views. In contrast, the kurtosis of a distribution measures the degree of peakedness of a distribution.³ If kurtosis has a positive value, then the higher the value, the more peaked the distribution.⁴ If kurtosis has a value below zero, the distribution is flatter than that of a normal distribution, and below –1.2 (i.e. a uniform distribution) the center of the distribution empties out until –2 is reached, at which point the distribution is perfectly bimodal. Two different distributions may have the same variance but different values for kurtosis. The same variance score may be generated by a few extreme deviations or a greater number of smaller deviations. Consequently, variance and kurtosis each tap a distinct dimension of consensus and thus are not substitutes for one another.

For each of these moments, we are interested in observing the trends in distributional change over time and seek to observe the effects of the Maastricht Treaty on the distribution of opinion. To this end, we employ Eurobarometer data from the Mannheim Trend file 1970–2002. We use the question that addresses perceptions of country benefits from EU membership (i.e. 'Generally speaking, do you think that (your country's) membership of the European Community is .... a good thing, a bad thing, neither good nor bad'). This choice was driven in large measure by data availability, since the series is available from 1973 to 2002.⁵ We are thus tapping changes in the distribution of instrumental perceptions of the E U. This is important to note because public support for EU level decision-making can vary contingent on the particular issue under investigation (see e.g., Hooghe, 2003). Moreover, Lubbers and Scheepers (2005) have shown that while instrumental and political Euro-scepticism (i.e. scepticism concerning EU level decision-making on specific policies) are somewhat correlated they are nevertheless distinct dimensions of scepticism. However, the choices presented to the public at elections come in the form of policy packages presented by political parties. Parties themselves, either explicitly or as a consequence of the particular policy packages they adopt, have positions on the EU that can in turn be arrayed in terms of the extent to which any party is more or less supportive of the EU, and it is typically these positions on which voters express preferences at elections. Therefore, since our concern with public disagreement is motivated by a concern for the EU as a domestic political issue, an instrumental focus appears more relevant than a policy-oriented focus. It should be recognized, however, that change in public disagreement over which policies should be located at the EU level is also an underexplored issue warranting further attention, not least because it may plausibly affect elite decision-making concerning the policy packages they present to voters.

To identify any systematic change over time in the distribution of attitudes toward integration, we calculate the mean, variance and kurtosis in each year, for each country, and the EU as a whole for the membership question. We then estimate an OLS regression for the EU as a whole and each country over time that allows us to observe the average annual change in each of the distributional moments. As outlined above, although we are interested in tracing cross time change in the degree of opinion change on the EU, we are also interested in the effects of the apparent increase in public engagement with the European project in the early 1990s that coincided with a shift in the substantive content of European integration signaled by the Maastricht Treaty. Accordingly, the model allows us to compare the immediate change in the distribution as well as the average annual change in distribution before and after the Maastricht Treaty.

The full equation below is estimated for each country using OLS regression and each of the mean, variance and kurtosis as dependent variables.

The variable 'Year' is a time trend variable coded as the year of analysis minus 1970. The Maastricht variable is a dummy variable coded '0' for each year prior to 1992 and '1' for all years after 1991. The term 'Maastricht Year' is an interaction between the dummy variable and the year variable. Substantively, the year coefficient is the average annual change in the dependent variable prior to Maastricht. The coefficient for the Maastricht and year interaction provides the difference between the slopes before and after Maastricht. Thus, the post-Maastricht slope is the sum of ₁ and ₃. The interpretation of the coefficient on the Maastricht dummy variable is less straightforward. Technically ₂ is the difference between the constant in the model (the expected value of the dependent variable in 1970 based on the pre-Maastricht slope) and the expected value of the dependent variable in 1970 based on the post-Maastricht slope. Combining this term with the others in the model allows us to compare the estimated value of the dependent variable in 1991 with the estimated value in 1992 to determine the immediate expected change in the mean, variance and kurtosis in response to the Maastricht Treaty. Thus, with this model, we are able to determine whether there was a substantial change in the mean, variance, and kurtosis between 1991 and 1992 and whether the trends in the distributions before and after Maastricht are significantly different.

Analysis

We begin our analysis with an examination of the means of the distribution. We estimate the regression equation for distribution means across the EU as a whole and within individual countries and present the results in Table 1. Figure 1 graphically displays the predicted means.

Figure 1.

Mean of support for EU membership.

Full figure and legend (145K)

Table 1 - Analysis of the mean support for EU membership.

Full table

In addition to the regression coefficients and standard errors for each country, Table 1 contains two columns that estimate the possible effect of the Maastricht treaty. The column titled '91–92 Difference' shows the predicted change in the dependent variable between 1991 and 1992. The P-values indicate the probability that the difference between 1991 and 1992 are truly zero.⁶ The entries in the column titled 'Post-92 Slope' are the calculated slopes for the post-Maastricht period described in the section above. The P-values indicate the probability that the slope in the post-Maastricht period is zero.⁷

Prior to Maastricht, the average level of support increased annually in six countries — Denmark, Great Britain, Ireland, Netherlands, Northern Ireland and Portugal, but elsewhere there was no significant change. Between 1991 and 1992, we observe a significant decline in support in Belgium, France, Germany, Greece, Italy, Luxembourg, Portugal and Spain. Moreover, in France, Italy and the Netherlands there was also a statistically significant decline in support in the years after the Maastricht Treaty. Only in Spain was there a 'rebound' in support after 1992. In all remaining countries, there was no significant change in the post-Maastricht period, meaning that for most countries support for the EU in 2002 was still lower than in 1991. The combined analysis for the whole of Europe shows a general trend of increasing support for the EU prior to Maastricht, a significant decline in support between 1991 and 1992, followed by no significant change (a flat slope) in the post-Maastricht period. For the countries entering the Union after Maastricht (Austria, Finland and Sweden), only in Austria does the average level of support increase after 1992.⁸

The results here are broadly consistent with similar analyses (e.g. Eichenberg and Dalton, 2007), indicating that the aggregate level of support was lower in the 1990s than in the preceding decades. Yet, as the graphs in Figure 1 illustrate, in most countries support was higher in the 1990s than in the early 1970s, and only in countries with particularly high levels of support in the 1970s was it lower in the 1990s. The key point is that the critical effect of the Maastricht Treaty has been to break the previous steady upward trajectory in support in those member states not already displaying high levels of support in the 1970s. The Netherlands aside, the five countries displaying a marked inter-temporal increase in the average level of support, prior to Maastricht, all joined in the second and third waves of enlargement, and critically, joined with levels of support well below those of the founding six.⁹ This doubtless reflects the fact that following the accession, public debate on the merits of membership tended to dissipate as the domestic political significance of membership receded. Post-accession decision-making concerned technical economic matters, determined by elites with little public input, and, national public began to develop a familiarity with the community. That familiarity is important because the relatively minimal impact the EEC/EC had on the daily lives of citizens in the member states doubtless served to assuage pre-accession concerns over decision-making moving from national parliaments to Brussels.

In brief, where widespread antipathy to the community did exist in every country it tended to fade away following accession — that is, until Maastricht. The significant and in some cases precipitous drop in support that most countries displayed at the time of Maastricht, the clear post-Maastricht continuation of that decline in France, Italy and the Netherlands, and the cessation, at Maastricht, of the tendency for the mean level of support in the newer member states to converge upwards on the levels displayed by the original six, are all indicative of the 'system transforming' nature of the Treaty. Yet, as noted earlier, in the absence of an analysis of the variance and kurtosis of the distribution we simply do not know whether Maastricht inspired a simple decline in support or whether the public became more divided on the EU. Accordingly, it is to these two moments of the distribution we now turn.

As with the mean, we estimated the regression model against the variance for the EU as a whole and within individual countries. Table 2 below presents the results of those estimations and the predicted variances from the regressions are displayed graphically in Figure 2.

Figure 2.

Variance in support for EU membership.

Full figure and legend (131K)

Table 2 - Analysis of the variance in support for EU membership.

Full table

The first point of note is that in no country prior to 1992 is there any evidence of increasing variance in attitudes. In six of the countries, almost exclusively the founding member states, there is simply no statistically significant time trend in the data. In the remaining countries (i.e. Denmark, Great Britain, Greece, Ireland, the Netherlands, Northern Ireland and Portugal) there is a statistically significant decline in the variance of attitudes in the pre-Maastricht period. This result is also reflected in the significant negative coefficient on year for the EU as a whole. In other words, between 1973 and 1991, consensus was increasing over time in almost all the states that joined the EU in the 1970s and 1980s, and across the EU as a whole. This trend clearly parallels the trend for the mean — in those countries in which the predicted level of support for the EU increases in the pre-Maastricht period the variance also declines.

Interpreting the magnitude of the variances and their changes is not straightforward, but some conclusions can be drawn about the substantive effect of the change. The scale ranges from 0 (all respondents are contained in a single category) to 1 (exactly half of the distribution is in each of the most extreme categories). Further, it can be shown that for each increase of 0.10 in the variance, 10% of the electorate will shift away from the middle category dividing themselves equally betwixt the extreme categories. The converse is also true, for each decline of 0.10 in the variance, 5% of the electorate will move from each of the extreme categories to the center. So, in the case of Portugal, the decline in variance from 0.45 in 1984 to 0.23 in 1991 means that approximately 11% of respondents moved from the extreme categories into the center (holding the mean constant), or that the Portuguese population in agreement on the EU increased by 22% in the 7 years prior to Maastricht.

By 1991, the predicted variance ranged from 0.20 in the Netherlands to 0.67 in Denmark. In 1992, there was a dramatic increase in the variance in nine countries, relative to 1991, and in no country was there a significant decline in variance. The magnitude of these changes was substantially larger than any of the average annual changes in the variance prior to 1992. Thus, the largest (in absolute value) average annual change in the pre-Maastricht period was -0.03, in Portugal, producing a 22% increase in agreement across the span of the 7 years prior to Maastricht (Ceteris paribus). In the course of a single year, however, that increase had all but dissipated with a decline in consensus of more than 21%.

After 1992, in most countries, there was no significant average annual change in variance (i.e. it was not statistically different from zero). Only in the Netherlands was there a significant post-Maastricht increase in variance. Of the countries for which a Maastricht effect was not pertinent (i.e. Austria, Finland, East Germany¹⁰ and Sweden), only in Austria and East Germany does the variance display a significant time trend, increasing in East Germany and decreasing in Austria. Thus, in general the trend is one of increasing consensus prior to Maastricht, a sharp one-time decline in consensus between 1991 and 1992, followed by no significant change in the post-Maastricht period.

The critical point is that the decline in support precipitated by Maastricht did not simply constitute a movement of the distribution in the issue space. Public disagreement over integration increased, and did so quite markedly. The transformation of the European Community into the EU clearly constituted more than just another step in an ongoing integrative process. Yet in turn, with the sole possible exception of the Netherlands, Maastricht did not initiate a new trend of increasing dissensus in the member states. Maastricht may have been system transforming, both in the sense that the prior trend towards increased consensus ended and disagreement sharply increased, but it did not establish a new trend. In almost all member states, there was no significant difference in the dispersion of attitudes in 2002 than in 1992.

Since the same change in variance can be brought about by either a few extreme deviations or a larger number of small deviations, the peakedness of the distribution must be examined to identify the full extent of public division on the EU. Accordingly, the same model was run against the kurtosis of the distribution across each country and across the EU as a whole. The results are presented in Table 3 below and displayed graphically in Figure 3.

Figure 3.

Kurtosis of support for EU membership.

Full figure and legend (138K)

Table 3 - Analysis of the kurtosis of support for EU membership.

Full table

The results very closely mirror those presented in Table 2, in that a significant increase (decrease) in variance is matched with a significant decrease (increase) in kurtosis. This is not wholly surprising, given that we are working with three categories; it is axiomatic that an increase in the dispersion of opinion (if the mean is held constant) equates to increasing bimodality. Across the EU as a whole and in eight member states, the coefficient is significant and positive pre-Maastricht, indicating increasing 'peakedness' of the distribution (i.e. Denmark, Great Britain, Greece, Ireland, Netherlands, Northern Ireland, Portugal and Spain). However, between 1991 and 1992 the distribution becomes significantly flatter in most countries and in the EU as a whole. Only in the Netherlands does the flattening continue in the post-Maastricht period, while in Ireland and Spain, the distribution becomes more unimodal. For all other countries, there is no reversal of the marked hollowing of the center of the distributions that occurred between 1991 and 1992. Of the countries for which the Maastricht intervention is not estimated, only two countries display a significant time trend, Austria towards greater peakedness and East Germany towards a flatter distribution.

It is also worth noting that in some country-years there is clear evidence of polarization on the EU — that is, some country-years register a kurtosis value of less than –1.2 (Austria, Denmark, Finland, Great Britain, Greece, Ireland, Northern Ireland and Sweden, comprising some 90 of the 398 total country-years). Yet, the proportion of country-years in which some degree of bimodality is evident declines over time, from 31% in the 1970s, to 25% in the 1980s, to 20% from 1990 to 2002. And, after 1990, only Austria, Denmark, Finland, Great Britain and Sweden display a kurtosis value of less than –1.2 in some years. Thus, the degree of polarization on the EU has actually declined in the last 30 years, rather than increased.

Again the central story is the termination of a steady convergence of public opinion prior to the 1990s. Maastricht served to flatten most national distributions and that effect persisted beyond the treaty, but it did not escalate. That is, we observe no new tendency of increasing polarization after Maastricht, the sole exceptions being East Germany and the Netherlands. Yet in the case of East Germany there was no prior trend to reverse, and thus it would be problematic to attribute to Maastricht the tendency toward bimodality evident in the 1990s. In the Netherlands, while there is an unequivocal tendency towards a flattening of the distribution in the 1990s, and it is a sharp reversal of the pre-Maastricht trend, the Netherlands was nevertheless a marked outlier immediately preceding Maastricht. The Dutch kurtosis value was both the highest of all countries and almost double that of the next nearest country (i.e. Luxembourg) in 1991. Moreover, despite the dramatic decline in the kurtosis value in the 1990s, the Dutch distribution was still more peaked than those in any other country in 2002, excepting Ireland and Luxembourg. In other words, while the center of the Dutch distribution hollowed out substantially after Maastricht that process simply brought the shape of the distribution more in line with those of the other member states.

In all three of the analyses of opinion distribution, we see the clear enduring effect of Maastricht. Between 1991 and 1992, the distribution of public opinion changed dramatically. Not only did increasing support for the EU come to a halt, but also opinion became more negative in most countries. This shift in the central tendency of the distributions coincided with a change in the shape of the distributions, which became flatter with greater dispersion post-Maastricht than in the 1980s. In other words, public became both more divided and less supportive of the integration process. Thus, that Maastricht constituted a 'system transforming' moment in the process of European integration is not limited to the institutional changes wrought by the Treaty. Maastricht also clearly served to transform public opinion and in so doing altered the terrain on which the future of the integration process would be played out.

Conclusion

That attitudes toward the EU have changed is not a particularly novel finding. However, that the level of consensus surrounding the EU has changed is an important and under studied phenomenon. When Lindberg and Scheingold (1970) first coined the term 'permissive consensus' in 1970 they conceived of 'consensus' as receptivity to the future growth of the European Community — that is, they meant by consensus support for the European Community. To the extent that such a consensus can be said to have prevailed in 1970 it certainly has not disappeared, since the level of support is little different today than that prevailed in the early 1970s. Rather, it was the secular trend in most countries of increasing levels of support for the community through the 1970s and 1980s that came to a halt in the early 1990s. Yet, despite the drop in support at the time of Maastricht, in only four countries is support lower in 2002 than in 1973, and only in France and West Germany is the difference non-negligible. Moreover, the cross-national gap in the mean level of support is substantially lower in 2002 than it was in 1973 — that is, national means have converged over the last 30 years. The growth in Euro-sceptical parties and party factions, the apparent willingness of the public to reject party cues at referenda and the reticence of governments to put EU issues to referenda all appear to reflect greater public concern and engagement with the process of integration. When viewed over the long run, however, it is far less clear that these phenomena are a product of a decline in support for the EU. Accordingly, if in 1970 it were pertinent to refer to support for integration as constituting a 'permissive consensus,' that consensus has not evaporated — it is simply lacking the permissive quality it once possessed (a likelihood acknowledged from the earliest days of the process should integration move onto the terrain of 'high politics').

If, however, consensus is conceived as the extent of disagreement, then post-Maastricht we have indeed moved into a world in which there is more significant public disagreement than that prevailed pre-Maastricht, and this is not simply a function of a decline in support for the EU. In almost every member state the shape of the distribution of opinion changed, with attitudes more dispersed and the distribution flatter. Thus, the potential for electoral gain by parties adopting a non-centrist stance on the issue does appear to have increased, at least to the extent that the issue is salient to voters. Yet, viewed over the long run, the change that has occurred is quite limited. In eight of the EU-13 (i.e., the EU-12 plus Northern Ireland), only five countries have a greater dispersion of attitudes in 2002 than at the earliest point for which we have data, and in only France and West Germany is the change non-negligible. Similarly, while eight member states had distributions that reflected some degree of polarization (i.e. hollow in the center, as evinced by a kurtosis value less than –1.2) at the earliest point for which we have data, only in Great Britain (-1.3) and Sweden (-1.4) did this still hold true in 2002.

The significant effect of Maastricht was to bring about a halt to the prior trend towards increasing consensus. Despite the subsequent moves to push political integration further (albeit with small steps) at Amsterdam and Nice, the completion of Monetary Union and ongoing attempts to develop defense and foreign policy coordination, only in the Netherlands do we observe a clear secular trend of decreasing consensus in the post-Maastricht period. Thus, while Maastricht may have improved the terrain for Euro-skeptic parties, both in the sense that consensus decreased and the public has become more engaged, the near total absence of further significant increases in the variance or decreases in the kurtosis of opinion distribution suggests that that terrain is little more propitious for these parties today than it was more than a decade ago. Accordingly, to the extent that public attitudes are a central determinant of electoral success, there is little evidence to suggest Euro-sceptical parties are better placed to make electoral breakthroughs today than they were 15 years ago, at least on the basis of the EU issue alone.¹¹

Lastly, it is worth noting that the overall picture emerging from our analyses is one of long run convergence between countries in each of the three moments of the opinion distributions (a convergence noted by Eichenberg and Dalton (2007) in relation to net support for the EU). There was less disparity between countries in terms of the mean, variance and kurtosis of the distributions in 2002 than in 1973. What is more, the cross-national range of each of these moments was lower in 2002, despite an increase in the number of member states, than in 1973. Those countries with relatively low means, high variance and low kurtosis all experienced some increase in their means, decrease in variance and increase in kurtosis over the 30-year period. In contrast, those countries with relatively high means, low variance and high kurtosis all to some extent experienced the reverse. In short, the national distributions shared a greater resemblance to one another in 2002 than they did 30 years ago. Thus, while public opinion may well be more constraining today than it ever was in the past, whether it is 'consensual' or 'dissensual' hinges on the meanings we attribute to the terms.

Notes

¹ Indeed, recent issues of both Acta Politica (Volume 42, Number 3, September 2007) and European Union Politics (Volume 8, Number 1, March 2007) have been entirely devoted to the subject.

² The growth of Euro-sceptical parties and party factions may have driven the movement of the distribution, may be a response to that movement, or, as Steenbergen et al. (2007) argue, reflect reciprocal causation.

³ Kurtosis is defined as: K=(X-)⁴/N⁴-3, where is the mean and is the standard deviation.

⁴ This assumes the standard '- 3' correction included in the above formula. That is, a correction to allow a normal distribution to evince a 0 kurtosis.

⁵ A better indicator of attitudes towards the EU would be the question addressing whether respondents are for or against efforts to unify Western Europe. This question, however, ceased to be asked after 1995. In consequence, we lack significant post-Maastricht data. This is important because we seek to identify whether the pre- and post-Maastricht periods display similar trajectories in terms of changes in the distribution of attitudes.

⁶ The P-values are derived from an F-test of the hypothesis: [+₁1991]-[+₁1992+₂+₃1992]=0.

⁷ The p-values are derived from an F-test of the hypothesis: [₁+₃]=0.

⁸ East Germany (sampled separately in our data) became a member upon unification, but there are not enough data prior to 1992 to justify including the full model with interaction terms. In East Germany, from 1990 to 2002, support for integration significantly declined.

⁹ As can be seen from Table 1, the slope coefficient on the Netherlands is significantly smaller than that of the other countries displaying a systematic pre-Maastricht increase in average support for the EU.

¹⁰ As noted earlier, given the relative lack of data for East Germany in the pre-Maastricht period (i.e. two data points) comparing this period with the post-Maastricht period and addressing the issue of a Maastricht effect seems neither sensible nor warranted. The sign and the significance of the coefficient for year are unaffected by whether or not the interaction terms are included.

¹¹ Needless to say, the nature of inter-party competition and institutional features of the political system constitute critical intermediate variables that may serve to either advantage or disadvantage Euro-sceptical parties, quite apart from the distribution of public opinion.

References

1. Carruba, C.J. (2001) 'The electoral connection in European Union politics', Journal of Politics 63: 141–158. | Article |
2. DiMaggio, P., Evans, J. and Bryson, B. (1996) ''Have Americans' social attitudes become more polarized', American Journal of Sociology 102(3): 690–755. | Article |
3. Dinan, D. (2004) Europe Recast: A History of European Union, Boulder, CO: Lynne Rienner Publishers.
4. Downs, A. (1957) An Economic Theory of Democracy, New York, NY: Harper & Row.
5. Eichenberg, R. and Dalton, R. (2007) 'Post-Maastricht blues: The transformation of citizen support for European integration, 1973–2004', Acta Politica 42(3): 128–152. | Article |
6. Evans, G. (1999) 'Europe: A New Electoral Cleavage?', in G. Evans and P. Norris (eds.) Critical Elections: British Political Parties and Voters in Long-term Perspective, London: Sage, pp. 207–222.
7. Evans, G. and Butt, S. (2007) 'Explaining change in British public opinion on the European Union: Top down or bottom up?', Acta Politica 42(3): 173–190. | Article |
8. Gabel, M. (1998) 'Public support for European integration: An empirical test of five theories', Journal of Politics 60: 333–354. | Article |
9. Gabel, M. and Scheve, K. (2007) 'Mixed messages: Party dissent and mass opinion on European integration', European Union Politics 8(1): 37–59. | Article |
10. Hooghe, L. (2003) 'Europe divided? Elite vs public opinion on European integration', European Union Politics 4(3): 281–305. | Article |
11. Hooghe, L. and Marks, G. (2005) 'Calculation, community and cues', European Union Politics 6(4): 419–443. | Article |
12. Lindberg, L.N. and Schiengold, S.A. (1970) Europe's Would-Be Polity, Englewood Cliffs, NJ: Prentice Hall.
13. Lubbers, M. and Scheepers, P. (2005) 'Political versus instrumental Euro-scepticism. Mapping scepticism in European countries and regions', European Union Politics 6(2): 223–242. | Article |
14. Marks, G. and Hooghe, L. (2004) 'Does identity or economic rationality drive public opinion on European integration?', PS: Political Science and Politics 37(3): 415–420.
15. Marsh, M., van der Eijk, C. and Franklin, M.N. (1996) 'What voters teach us about Europe-wide elections; what Europe-wide elections teach us about voters', Electoral Studies 15: 149–166. | Article |
16. Rohrschnieder, R. (2002) 'The democracy deficit and mass support for an EU-wide government', American Journal of Political Science 46: 463–475. | Article |
17. Sanchez-Cuenca, I. (2000) 'The political bases of support for European integration', European Union Politics 1(2): 147–171. | Article |
18. Steenbergen, M. and Scott, D.J. (2004) 'Contesting Europe? The Salience of European Integration as a Party Issue', in G. Marks and M. Steenbergen (eds.) European Integration and Political Conflict, Cambridge: Cambridge University Press, pp. 165–192.
19. Steenbergen, M., Edwards, E. and de Vries, C. (2007) 'Who's cuing whom? Mass-elite linkages and the future of European integration', European Union Politics 8(1): 13–35. | Article |
20. Taggart, P. (1998) 'A Touchstone of dissent: Euroscepticism in contemporary Western European party systems', European Journal of Political Research 33: 363–388.
21. Taggart, P. and Szczerbiak, A. (2002) 'The party politics of euroscepticism in EU member and candidate states', SEI Working Paper No. 51 (Opposing Europe Research Network Working Paper No. 6), Sussex European Institute, Sussex.
22. Tillman, E. (2004) 'The European Union at the ballot box? European integration and voting behavior in the new member states', Comparative Political Studies 37: 590–610. | Article |
23. van der Eijk, C. and Franklin, M.N. (2004) 'Potential for Contestation on European Matters at National Elections in Europe', in G. Marks and M. Steenbergen (eds.) European Integration and Political Conflict, Cambridge: Cambridge University Press, pp. 165–192.

Acknowledgements

We thank Liesbet Hooghe, Gary Marks, and the anonymous reviewers for their helpful comments.

About the authors

Ian Down is Assistant Professor of Political Science at the University of Tennessee, Knoxville.

Carole J. Wilson is Assistant Professor of Political Science at the University of Texas at Dallas.