How have secularisation and educational expansion affected support for the mainstream right in Western Europe?

Abstract While the electoral decline of Social democracy has received considerable attention in the literature, much less is known about how the structural changes experienced by advanced capitalist societies in the past decades have affected support for the mainstream right. In order to fill this gap, this article examines the relationship between secularisation, educational expansion and support for Conservative and Christian democratic parties in 18 West European democracies since the 1960s. The analysis reveals that secularisation is negatively associated with support for the mainstream right, but the effect is only significant for Christian democratic parties. Moreover, the findings support the expectation that social conservatism tends to damage the electoral prospects of mainstream right parties in highly educated societies. This suggests that, although engaging in cultural wars might sometimes seem like a winning strategy in the short term, it is unlikely to produce long-term positive electoral returns for the centre-right in post-industrial contexts.

Christian Democratic parties were first affected by these trends, the more secular Conservative parties have not been immune to them either (Bale and Rovira Kaltwasser 2021).Despite cases of relative success (e.g.Finland and Sweden), and cases of parties that have (at least temporarily) overcome decades of slow electoral deterioration (Britain), electoral decline has affected the mainstream right in countries such as Belgium, Denmark and the Netherlands, and more recently also France, Italy, Germany and Spain, among others.
The scarce literature dealing with the electoral decline of the mainstream right has pointed to the impact of two parallel, but inter-related, processes of dealignment and realignment.On the one hand, secularisation has led to the weakening of the traditional religious cleavage which cemented support for many right-wing parties, making it more difficult for the latter to rely on social institutional links with certain Christian groups.On the other, post-industrialisation has increased tensions within the mainstream right between appealing to better-off, highly educated voters with right-wing economic preferences but liberal sociocultural values, and attracting voters who hold authoritarian values but do not favour economic liberalism (Bale and Rovira Kaltwasser 2021: 23).
While the literature has provided evidence of a weakening of the religious cleavage, the propositions regarding the effects of post-industrialisation, and in particular of educational expansion, on support for the mainstream right have rarely passed the status of untested hypotheses -especially when it comes to exploring the implications of these changes for the actual electoral results of mainstream right parties.Moreover, the strong focus on structural changes in the literature means that the role of political parties as agents, while never disregarded in theory, is often overlooked in practice, as if the mainstream right was doomed to suffer the consequences of social change without being able to do much about it.Parties' diverse strategic positions have, nevertheless, been shown to play an important role in explaining the Social Democrats' electoral success (or the lack thereof) in post-industrial democracies (Kitschelt 1994).The fact that the mainstream right's electoral decline has not followed the same trends everywhere in Western Europe, and that the fragmentation of the conservative political space has taken many a shape, suggests that the electoral strategies deployed by the mainstream right itself may be similarly important.
This article contributes to the literature on the electoral decline of mainstream right parties in two ways.First, it explores the impact of secularisation and educational expansion on the mainstream right's vote share over the long term.Second, it shows that mainstream right parties' strategic choices have either mitigated or exacerbated long-term patterns of electoral decline.Using aggregate data, I analyse electoral support for mainstream Conservative and Christian democratic parties in 18 countries since the 1960s.Results show that the prevalence of non-religious sectors in the population is negatively associated with support for the mainstream right, particularly for the Christian democrats.As for the expansion of education, its effect is found to depend on party strategies.Educational expansion is associated with positive electoral outcomes for parties with liberal positions on sociocultural issues, but not for their socially conservative counterparts.To triangulate these findings, I leverage individual-level data from nine waves of the European Social Survey. 1  Results are consistent with the findings based on aggregate data.First, religious identity increases the probability of supporting the mainstream right, and particularly so in countries where this adopts the form of a Christian democratic party.And second, education significantly increases (decreases) the chances of voting for right-wing parties with liberal (conservative) positions on sociocultural issues.
In the next section, I advance the main theoretical arguments and hypotheses of the article.This is then followed by a section on data and methods.The fourth section shows descriptive statistics on the evolution of support for the mainstream right across countries and over time.This is then followed by the results of the empirical analysis (aggregate analysis first, and individual-level analysis in the second place).The article then ends with a discussion of the findings and their main implications.

Theoretical considerations
Two main structural processes have been identified in the literature as potential factors upsetting the social basis of support for mainstream right-wing parties in the past few decades: the weakening of the religious cleavage, and the emergence of a new divide based on sociocultural issues that is strongly correlated with education.Although these two factors have often been analysed separately, many authors consider them to be part of the wider process of modernisation and (post)industrialisation of Western societies (Norris and Inglehart 2004).Both secularisation and educational expansion are associated (through different mechanisms) with a gradual shift towards more progressive sociocultural values, and are therefore expected to introduce instability in the potential electorate of centre-right parties.

The decline of the religious cleavage
The decline of the religious cleavage has been the focus of scholarly attention since the 1990s, a decade that saw the electoral deterioration of the Belgian and Dutch Christian democrats alongside the disbandment of the Italian Democrazia Cristiana.As religion played a very important role in the emergence of Christian democracy, providing it with a loyal cross-class electoral base throughout the years, the gradual dwindling of practising Christians in West European societies was soon thought to be the culprit for the weakening of these parties' traditional electoral bedrock (Duncan 2003(Duncan , 2015)).As pointed out by Goldberg (2020), the mechanisms behind this are thought to be twofold.First, there has been a certain amount of structural decline due to the simple fact that the number of people who identify as Christian has decreased dramatically over the past decades, thereby reducing the proportion of religious voters.Second, there has also been behavioural dealignment, because the smaller role of religion in society has led many right-wing parties to de-emphasise religious issues in order to attract a growing number of secular voters.This, in turn, has further weakened the links between religious groups and the former.
The different trends of religious voting across countries have raised questions about whether secularisation has only affected mainstream right parties with a religious background, or whether the more secular Conservative parties have also suffered electorally as a result.There is some evidence that the growth of sectors in the population who are not members of any Christian church has negatively affected support for Christian democratic parties in countries such as Belgium, the Netherlands and Germany, whereas a similar relationship has not been found for secular mainstream right parties such as the British Conservatives (Goldberg 2020;Lachat 2007).However, it is worth noting that Christian democratic parties in countries such as Germany and Austria were very successful in a context of increasing secularisation, and have only recently experienced electoral decline.This casts some shadows on the actual electoral impact of secularisation.In fact, various scholars suggest that the effects of the latter on religious parties may have been overstated, as Christian democratic parties adapted very early on to a more secular context by stripping off all their 'explicitly and exclusively religious ideological baggage' in order to present under a new light a policy package that is now only broadly inspired by religious principles (Kalyvas and van Kersbergen 2010: 204).Moreover, while religious voters have been found to be over-represented in the electorate of Christian democratic parties, this is also the case for the more secular Conservative parties, albeit to a relatively lesser extent (Knutsen 2018).
One possibility is that secularisation may have only impacted parties that, regardless of their Conservative or Christian Democratic 'genetic' model (Panebianco 1988), have not been able to adapt their discourse on moral and lifestyle issues to an increasingly secular West European electorate.Indeed, individual-level research has found the magnitude of religious voting to be associated with parties' positions on moral and lifestyle issues (Graaf et al. 2013), even though some scholars caution that changes in parties' positions do not affect all age cohorts in the same way (Gomez 2022).However, it is important to bear in mind that religious identity and church integration have historically benefitted the mainstream right not only because of their impact on people's opinions on moral issues, but also because they contribute to the formation of a conservative identity and increase individuals' preferences for anti-redistributive economic policies, leading religious voters who would benefit from redistribution (i.e.those on a low income) to support right-wing parties (Stegmueller 2013).In fact, Langsaether (2019) finds that in Western Europe only around 10% of the effect of religion on party preferences is mediated by moral values, the rest being mostly explained by strong social institutional links between parties and religious voters.The role of religious networks, associations and organisations in creating and nurturing a political Christian identity has been such that, in some European countries, it even led to the emergence of a distinct and long-lasting subculture (Bale and Krouwel 2013).It is then easy to see how parties who relied on their links with religious groups may have struggled to keep the same level of loyalty in a secular environment.Adopting a moderate position on moral issues could make it easier to reach out to secular voters, but will not guarantee their loyalty; adopting a traditionalist position may, in turn, help parties to secure support among active religious voters, but these constitute an increasingly small group in the population.
As the evidence on the electoral impact of secularisation is mostly based on individual-level data, it is worth exploring how it has affected the actual vote share of different mainstream right parties.Thus, three alternative hypotheses will be tested here.The first and simpler hypothesis states that secularisation has had a negative electoral effect on mainstream right parties.Therefore, secularisation (measured as increases in the proportion of non-religious individuals) is expected to be associated with electorally weaker Conservative and Christian Democratic parties (H1).The second hypothesis states that the effect of secularisation depends on the parties' 'genetic model' .As historical links with religious groups were much stronger for confessional (Christian Democratic) parties than for Conservative parties, we should expect secularisation to be associated with electorally weaker Christian Democratic parties but not with weaker Conservative parties (H1a).Finally, a third hypothesis states that parties should have been able to avert the negative effects of secularisation by moderating their stances on moral issues.Thus, the negative effects of secularisation will be stronger for parties that adopt more traditional stances on moral issues (H1b).

Education-based realignment
The early focus on secularisation by studies on the mainstream right may have contributed to understating (at least until more recently) the influence of other inter-related social changes that have also taken place since the 1960s in advanced capitalist democracies.Among those is post-industrialisation, which the literature has identified as a key contributor to the erosion of West European party systems (see, e.g.Ignazi 1992).The term post-industrialisation traces back to Bell (2020Bell ( , 1976)), and can be defined as a process marked by significant skill upgrading, the increasing weight of the knowledge economy, and the corresponding growth of professionals and experts in advanced capitalist societies.While not everybody agrees with the term 'post-industrial' , because it can misleadingly evoke the idea that manufacturing jobs are no longer important, there is evidence that educational expansion, skill-biased technological change and globalisation are all associated with an increase in the proportion of occupations with higher complexity and skills requirements (Oesch and Piccitto 2019).Formal education, knowledge and cognitive skills have, thus, become primary sources of social stratification in modern societies (Kerckhoff 2001).As a social marker, education not only influences people's policy preferences and values, but its influence on the latter has also increased with post-industrialisation (Kalmijn and Kraaykamp 2007).This, in turn, has upset existing electoral equilibria, bringing about electoral challenges for the mainstream right as well as other mainstream parties (Bale and Rovira Kaltwasser 2021).
There are reasons to think the transition to a post-industrial economy may have had both positive and negative consequences for the mainstream right.On the positive side, educational expansion may have contributed to enlarging the pool of potential mainstream right voters.Education is associated with characteristics such as greater prospects of upward mobility, lower chances of unemployment and higher earned income, which are linked to preferences for lower taxation (Alesina and Giuliano 2011).Moreover, the occupational groups that have grown the most with educational expansion (professionals, managers and technical professionals, as well as skilled workers in export-oriented sectors) are expected to have relatively liberal economic policy attitudes (Kitschelt 1994).Finally, even though educational attainment is strongly impacted by social background (Souto-Otero 2010), the growing social role of formal education is likely to have reinforced the view that educational achievement and merit can grant access to better career opportunities, and therefore greater levels of wealth, power and prestige.This increased emphasis on self achievement, effort and individualism may have further contributed to the decline of class-based identity, thereby damaging support for left-wing parties.
Despite this, educational expansion may have also presented challenges for the mainstream right.Many scholars see the role played by education in post-industrial societies as one of the reasons behind political realignment, prompted by the emergence of a new issue dimension (some refer to it as a cleavage) concerning cultural diversity, traditionalism, nationalism, lifestyle autonomy and environmentalism (Hooghe et al. 2002;Inglehart 1977;Kitschelt and Rehm 2015;Kriesi et al. 2006).There is much evidence that education has become one of the main sources of preferences on sociocultural issues, with highly educated individuals being predisposed towards a more libertarian view of society than those with lower levels of education (Van De Werfhorst and De Graaf 2004).The reason why education is associated with sociocultural preferences is explained by at least three possible mechanisms (Hooghe et al. 2013;Meeusen et al. 2013;Stubager 2008).First, education is thought to have a 'liberalising' effect, which translates into individuals becoming more tolerant, open to change and socially liberal the longer they spend in formal education settings (although there is no consensus as to whether this is due to the socialising effect of formal education institutions, or whether there is also an intrinsic effect of developing higher-order thinking skills and acquiring knowledge that challenges previous assumptions about how the world works).Second, education has an indirect effect on people's values, because it correlates with variables such as job autonomy and economic security, which are empirically associated with social liberalism (Stubager 2008).Third, some scholars have also pointed out that education can become a social identity of its own, and play a central role in how individuals see themselves in relation to others (Zingher 2022).
Thus, while post-industrialisation is likely to have expanded the pool of voters with right-wing preferences on economic issues, it may also have increased the heterogeneity of the potential mainstream right electorate when it comes to sociocultural issues, because the educated 'post-industrial' middle class is expected to be more socially liberal than the 'old' middle class and other social groups that traditionally supported the mainstream right (Bale and Rovira Kaltwasser 2021).Indeed, there is mounting evidence that the proportion of cross-pressured voters (those who have conservative preferences on one issue dimension but not on the other) has multiplied in the past few decades (Gidron 2022).But the consequences that this has had on the actual election results of mainstream right parties are unclear.As Gidron (2022) shows, both welfare chauvinists (those who hold conservative positions on sociocultural issues but left-wing positions on economic issues) and market cosmopolitans (those who hold liberal positions on sociocultural issues but right-wing positions on economic issues) are more likely than other voters to identify with a right-wing ideology.However, despite placing themselves on the right of the political spectrum, these groups are not necessarily attracted to Conservative and Christian democratic parties.
On the contrary, it is often smaller right-wing parties, such as liberal parties and the radical right (but also centrist parties, agrarian parties, small religious parties, etc.), that are disproportionally favoured by them.So, while the growth of cross-pressured voters may have changed the balance between left and right, it may also have introduced increasing division (and potentially greater volatility) within the right-wing electorate.
This brings us to another important issue: the role of political agency.The problem with paying too much attention to structural social changes is that we are often left with a static picture that overlooks, whether intentionally or not, the behaviour of parties.Extant literature has found parties' strategic choices (and, in particular, their policy positions) to be essential for understanding the electoral mobilisation of different social groups (Evans and Graaf 2013).For example, parties with right-wing economic positions tend to be significantly more successful among the self-employed and the 'service class' than among other social classes, and those that promote lifestyle and moral traditionalism have been found to be more successful among churchgoers than among non-churchgoers (Elff 2009;Graaf et al. 2013).It therefore seems clear that the electoral success of mainstream right parties may also depend on their own strategic choices and how well they have responded to a changing social and political environment.
Conservatives and Christian democrats have always been able to navigate two-dimensional politics by exploiting both economic issues and issues concerning national identity, traditional morality and lifestyle choices.Social conservatism was the glue binding together a relatively broad coalition that included the 'old' middle and upper classes, conservative sectors of the working classes and other voters.For decades, this electoral formula appears to have allowed the mainstream right to develop cross-class appeals without confronting the electoral dilemma faced by reformist socialist parties, whose attempts to reach out to middle-class voters often ended up undermining their links with the working class (see Przeworski and Sprague 1986).However, educational expansion is likely to have upset the equilibrium of political preferences among potential conservative voters.In a scenario where the potential right-wing electorate is becoming more and more heterogeneous, and where parties can no longer rely to the same extent as they did in the past on social loyalties, the mainstream right now confronts a renewed electoral dilemma. 2 Mainstream right parties that maintain a conservative sociocultural agenda could potentially stop the emergence of radical right and other socially conservative competitors, but at the expense of creating opportunities for other parties to attract a growing sector of voters with liberal sociocultural positions.Conservative and Christian democratic parties with a socially liberal outlook will likely have the opposite problem; however, as the fastest growing sectors in post-industrial societies are precisely those who are expected to have liberal sociocultural values, these parties might be more successful than their more conservative counterparts in offsetting their net electoral losses.Therefore, we should expect the expansion of education to be associated with greater electoral success for Conservative and Christian Democratic parties that adopt more liberal stances on sociocultural issues compared to their socially conservative counterparts (H3).
At this point, it is important to clarify two issues.First of all, if adopting a relatively more liberal position on sociocultural issues is the least damaging strategy, should we not expect all Conservative and Christian democratic parties to have done so?The answer is 'no' , because we know that policy shifts are constrained by a number of variables, including uncertainty over the actual consequences of alternative policy positions, organisational and ideological constraints, credibility costs, the strength of party identification, the balance of office versus vote-seeking considerations, and the characteristics of the party system (Meyer 2013).Secondly, the considerations made above do not mean that Conservative and Christian democratic parties cannot adopt other successful strategies in the short term.As Kitschelt (1994) points out, political parties sometimes act like oligopolistic actors and make strategic moves that can be effective at preventing the growth or emergence of other competitors in the short run, even if these are counter-productive when sustained for a long time.And indeed, politicians often have stronger incentives to concern themselves with the short-term rather than the long-term.

Data and method
In order to evaluate the hypotheses laid out in the previous section, I proceed in two ways.The main part of the article focuses on analysing the vote share of all major Conservative and Christian democratic parties in Western Europe from the 1960s (or since the country became a democracy in the cases of Greece, Portugal and Spain) until 2018.This period of around 50 years was chosen based not only on data availability, but also because post-industrialisation is often argued to have emerged in advanced capitalist economies during the 1960s and 1970s (Bell 1976).The analysis will focus on the mainstream right, defined as the main Conservative/ Christian democratic party in each country.It therefore excludes minor parties, where they exist, such as small religious and centrist parties.The list of parties and countries included in the analysis can be seen in Table 1.
The aggregate-level variables used in the analyses are as follows.The dependent variable is vote share, which was extracted from the European Journal of Political Research: Political Data Yearbook (2021) and ParlGov (Döring and Manow 2021).As for the main independent variables, the expansion of education and upskilling is measured through the Penn World Table's human capital index (Feenstra et al. 2015), which is a measure comprising average years of schooling and the rate of return to education.The percentage of the population who have completed at least secondary education (Barro andLee 2013, 2015) is also used as an alternative measure of educational expansion in additional models.Secularisation is measured by the percentage of population not affiliated with any religion or denomination, as estimated by the World Religion Project (Maoz and Henderson 2013).Party positions on sociocultural, economic and moral issues are constructed using party manifesto data from the MARPOR project (Volkens et al. 2019).Following Lowe et al. (2011), parties' positions are derived using a logit scale that contraposes the number of right-wing (conservative-authoritarian) quasi-sentences to the number of left-wing (liberal-libertarian) quasi-sentences in party manifestos.The items used to construct the sociocultural and economic scales are the ones identified by Prosser (2014), who employs an inductive content validity approach to ensure the inclusion of all empirically-relevant issues in each dimension. 3The resulting scales have been shown to score higher on reliability and internal validity than other well-known scales, including Laver and Budge's (1992) unidimensional scale and Bakker and Hobolt's (2013) two-dimensional scales.The moral scale was constructed using positive and negative mentions of a traditional way of life.Finally, the models include the following control variables: (logged) average district magnitude (Bormann and Golder 2013), trade union density (OECD and AIAS 2021), GDP per capita and GDP growth (OECD 2021), government party (Döring and Manow 2021), government spending (IMF 2020; The World Bank 2020) and turnout (International IDEA 2021).All of these are variables that could potentially impact support for the mainstream right, and many of them have also been used in research on the electoral success of other mainstream parties (Benedetto et al. 2020).
The second part of the article utilises data from nine waves of the European Social Survey (2020) covering the period 2002-2018, alongside party manifesto data from the MARPOR project (Volkens et al. 2019).The aim of this part is checking if the conclusions drawn from the analysis of aggregate data are consistent with evidence derived from individual-level data.Thus, the dependent variable is a dichotomous indicator of whether an individual voted for one of the parties in Table 1 at the last general election (1) or for a different party (0).At the individual level, the main independent variables are education (which is measured as the number of years spent in formal education to maximise comparability across different educational systems), and religious identity (a dichotomous variable indicating whether the respondent belongs to a Christian religion or denomination (1) or not (0)). 4At the aggregate level, the model includes parties' positions on the economic and sociocultural dimensions (see previous paragraph) and party genetic model (Christian Democrat or Conservative).The following controls are also included: age (measured in years), gender (which contains two categories: female (1) and male (0)), left-right self-position (an 11-point scale ranging from left (0) to right (10)), urbanisation of the place of residence (three dichotomous variables: village or rural, town, and city or suburbs), trade union membership (a dichotomous variable indicating whether the respondent is a union member (1) or not (0)), and occupational class, which follows Oesch's (2008) 8-category schema.
The aggregate data are analysed using OLS regression with party-level fixed effects, decade fixed effects and cluster-adjusted standard errors. 5 This method was chosen because it allows studying long-term change while controlling for potential time-invariant confounders.The individual-level data are analysed using a multilevel linear probability model with random effects by period (country × year) and country to account for the clustered nature of the data.

Trends in electoral support for the mainstream right
Before analysing the data, it is worth exploring how support for the mainstream right parties listed in Table 1 has changed since the 1960s.The vote share obtained by each party in parliamentary elections is shown in Figure 1 alongside a local polynomial regression line indicating the overall trend over time.Two pieces of information stand out in this figure.First, average support for mainstream right parties has been subject to a gradual decline in the period analysed, but particularly since the second half of the 1990s.Secondly, despite this decline in average support, there is wide variation between countries.
To help appraise country-specific trends, Figure 2 provides the same information for each individual country.As can be seen, the overall trend of gradual decline mentioned above can be appreciated in a majority of countries, but in most cases it is far from a straight linear downwards trend.This is a positive technical aspect of the data, because it ensures the stationarity of the dependent variable as a whole, but it is also interesting from a theoretical perspective.In those countries where there has been a decline in support, its timing varies greatly, with most parties showing signs of gradual deterioration from the 1990s onwards, but rarely earlier.Exceptions are Belgium and the Netherlands which do not only present the earliest declines, but also some of the sharpest drops.Among those cases where no clear trends can be appreciated are Luxembourg, Malta, Norway, Portugal, and, to an extent, Britain and Greece.On the other end are rare cases, such as Sweden, where support for the mainstream right has actually increased since the 1960s.All in all, there are interesting trends in the data that deserve an explanation.

Secularisation, educational expansion and support for the mainstream right
The models in Table 2 examine the relationship between secularisation and educational expansion and the vote share of mainstream right parties in Western Europe.The first model (Model 1), which includes all the independent variables of interest but no interactions, can be used to test H1, which states that support for the mainstream right will decrease with secularisation.As can be seen, the coefficient for secularisation (measured as the percentage of population who do not belong to any religion or denomination) is negative and statistically significant, suggesting that the decline of organised religion is associated with lower electoral support for the mainstream right, as predicted by H1.
Other findings in this first model are also worthy of mention.First, as expected, educational expansion is not, on its own, associated with the electoral success of mainstream right parties.The coefficient for the human capital index, which measures the average amount of time spent in formal education, is statistically non-significant.Moreover, the positive sign of the coefficient indicates that, if anything, increases in human capital are associated with stronger mainstream right parties.This is consistent with Kitschelt's (1994) argument that educational expansion can potentially increase the mainstream right's electoral base, although the lack of significance of this coefficient indicates the likely absence of a systematic positive effect for most parties.Using the percentage of the population who has completed at least secondary population as an alternative proxy for educational expansion produces the same result (see online appendix).When it comes to the mainstream right's policy positions, there is no evidence of a direct effect of sociocultural liberalism/conservatism.This contrasts with the effect of economic left-right.Here, the coefficient is negative and statistically significant.This suggests that, as one would expect based on standard spatial models of voting, mainstream right parties are more successful when they adopt moderate policy stances on economic policy.As for the control variables, none of them have a significant effect in this model.This does not mean that these variables are not associated with the electoral prospects of the mainstream right.The model investigates long-term changes, and some of the controls that change more decisively over time might perhaps only make a difference in the shorter term.
The aim of Models 2 and 3 is testing H1a and H1b, which respectively state that the negative effect of secularisation may have been stronger for mainstream right parties with a religious background (Christian democrats) or for those parties that adopt a more traditionalist position on moral issues.Model 2 introduces an interaction between the percentage of non-religious people in the population and a dummy variable accounting for whether a party is Conservative (0) or Christian democratic (1).As can be seen, the interaction term is negative and statistically significant, suggesting that Christian democratic parties may have been, on average, more negatively affected by secularisation than their secular counterparts.As there is no constituent term for the dummy variable (the effect of being a Christian democratic or Conservative party is already absorbed by the fixed effects), we can interpret the coefficient for 'non-religious (%)' as the effect of secularisation on the base category (Conservative).Interestingly, although the coefficient is also negative, it is statistically non-significant, which suggests that secularisation is not negatively associated with support for Conservative parties, or at least not in a systematic manner. 6Model 3 then introduces an interaction between secularisation and parties' positions on the moral traditionalism scale.In this case, the findings do not support H1b, as the interaction coefficient is not only positive (rather than negative, as predicted by H1b) but it is also statistically non-significant.This suggests that the aggregate effect of secularisation may have been related to the weakening of the traditional party-voter links that Christian democratic parties used to rely on, rather than the role played by parties' positions on moral issues.This is an important finding, as it speaks to the limits of partisan strategies in counteracting the effect of certain structural social changes.
In order to test the hypothesis concerning educational expansion, Model 4 (Table 1) includes an interaction between the human capital index and the mainstream right's position on the sociocultural liberal (libertarian)/conservative (authoritarian) dimension.Educational expansion is expected to have a negative effect for mainstream right parties that adopt conservative positions on the sociocultural dimension, but to benefit those with relatively liberal policy stances (H2).As can be seen in the table, the interaction between the human capital index and the mainstream right's sociocultural position is both negative and statistically significant, which is consistent with H2.
To ease interpretation, the total effect of this interaction is shown in graphical form in Figure 3.The graph displays the average marginal effect of adopting a more conservative position across different levels of educational expansion.The slope of the effect is negative, indicating that social conservatism is associated with worse electoral outcomes when levels of education are greater in a country.The evidence is consistent with the hypothesis that in societies where people spend more time in formal education, mainstream right parties that adopt a conservative position on sociocultural issues tend to have worse electoral returns than those that adopt a more liberal policy stance.
This finding remains when the interaction between secularisation and the dummy variable accounting for the parties' genetic origin (Christian democratic or Conservative) is included (Model 5).

Individual-level analysis
The macro-level analysis performed in the previous section suggests that secularisation is negatively associated with support for Christian democratic parties and that formal education is associated with poorer election results for mainstream right parties that hold conservative positions on sociocultural issues.This section now tests whether these findings are consistent with evidence using individual-level (survey) data.Rather than evaluating over-time changes as in the previous section, this section tests whether: (a) religious identity increases the chances of voting for Christian democratic parties more than it does for Conservative parties, and (b) the effect of education on the probability of voting for the mainstream right depends on the sociocultural position adopted by the party.As the analysis is based on relatively recent data (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018), results should be taken as a snapshot of the variables associated with individual support for mainstream right parties in the post-industrial West European societies of the 21 st century.The data are not suitable to properly test for long-term structural effects, because it only covers a limited number of elections within each country, but it can provide evidence of how voter profiles vary for different types of mainstream right parties.
The results of a multilevel model explaining vote for the mainstream right are shown in Table 3 (Model 6).Among other variables, the model contains an interaction between religious identity and party genetic model (Christian democrat versus Conservative), as well as an interaction between the length of formal education (measured in years) and parties' positions on sociocultural issues.
Starting with the impact of religion, the constituent term for religious identity is positive and statistically significant, as is the interaction between religious identity and the dichotomous variable measuring party genetic model.This suggests that, although individuals who belong to a Christian religion or denomination are more likely than other individuals to vote for Conservative parties, the former are even more likely to vote for the mainstream right in countries where this takes the form of a Christian democratic party.This is indeed consistent with the findings in the previous section, which showed the negative association between secularisation and the electoral results of mainstream parties to be considerably stronger for Christian democratic parties.
Regarding the effect of education, as can be seen in Model 6 the interaction between the mainstream right's position on sociocultural issues and education is negative and statistically significant.In other words, unlike their culturally conservative counterparts, the probability of voting for mainstream right parties with liberal positions increases significantly with education.The total effect of the interaction, which can be found in the online appendix (Figure 1a), clearly shows how the total effect of education depends on the mainstream right's position on sociocultural issues.Every additional year of formal education increases the probability of voting for a socially liberal mainstream right party (1 standard deviation below the mean on the sociocultural conservatism scale) by about 0.44 percentage points.This may seem little, but it amounts to 6.6 points for an individual with 15 years of formal education.For a party with an average position on sociocultural issues, the estimated average effect of education amounts to 0.26 percentage points per year, and to a non-significant 0.08 for a socially conservative party (1 standard deviation above the mean).
All in all, the survey data analysed in this section offer evidence that in post-industrial West European societies, the electoral mobilisation of education is associated with parties' positions on the sociocultural dimension.While these results do not tell us how different types of mainstream right parties ended up with a particular voter profile, they are consistent with the finding in the previous section that educational expansion is associated with better electoral outcomes for liberal mainstream right parties.It is important to bear in mind that the model also controls for occupational class, which is, of course, related to the levels of skills an individual possesses.So, the estimates for education are likely to be conservative.
Regarding the effects of control variables, the probability of voting for mainstream right parties is greater among white-collar employees (with the exception of sociocultural professionals) and self-employed individuals than among workers.In addition, both age and living in rural areas increase the probability of voting for the mainstream right, while gender has no significant effect once we control for whether the mainstream right party is Conservative or Christian democrat.This is consistent with other findings in the literature, because religion has been shown to be one of the factors explaining why women, and particularly older generations of women, used to be relatively more likely than men to vote for Christian democratic parties (Duncan 2017;Shorrocks 2018).

Robustness
Robustness tests can be found in the online appendix.Individual-level models were re-estimated using church attendance rather than religious identity, and conclusions remained the same.Aggregate models were re-estimated using the percentage of the population who have completed at least secondary education as an alternative proxy for educational expansion (a measure that is available for a slightly more limited number of cases); using random effects with a lagged dependent variable rather than fixed effects; controlling for the position of the centre-left; using alternative policy position scales; and excluding the Irish case.Overall, conclusions do not change.The results presented in this article hold when using the empirically-validated party manifesto scales based on Prosser (2014) as well as Bakker and Hobolt (2013).The use of non-validated deductive social scales containing a more limited number of manifesto items weakens the statistical significance of findings in a considerable manner, but does not affect the sign of coefficients.

Discussion and conclusions
Despite their historically dominant position on the right of the political spectrum, in recent decades West European Conservative and Christian democratic parties have faced increasing levels of electoral volatility and seen new and existing right-wing competitors bite into their traditional electoral base, sometimes even outranking them as the main party of the right.The findings in this article suggest that those trends are, at least in part, associated with two long-term structural processes: the de-aligning effect of secularisation and the re-aligning effect of educational expansion.
The evidence shown in this article is consistent with the idea that secularisation has contributed to the erosion of long-established links between voters and the mainstream right, making it more difficult for parties to use religious identity to retain voters.However, the negative effects of secularisation seem to be mostly circumscribed to Christian democratic parties.Indeed, religion was more central for Christian democratic parties than it ever was for their Conservative counterparts, and has played a key role in their greater ability to appeal to a wide cross-class electorate (Duncan 2015;Kalyvas and van Kersbergen 2010).Interestingly, though, there is little evidence that the negative impact of secularisation can been alleviated by the adoption of relatively more liberal stances on moral issues by the mainstream right.This is consistent with findings in the literature indicating that, although religious voters tend to hold more traditional values on moral issues, the latter accounts for a small percentage of the effect of religion on party choice (Langsaether 2019).Therefore, the negative association between secularisation and the vote share of Christian democratic parties might be related to the gradual weakening of party-voter links in societies where an increasing proportion of voters do not consider themselves religious.This highlights the limits of partisan strategies to quickly counteract the effect of social developments, especially when these erode the longstanding structural foundations of parties' electoral support.
Alongside secularisation, the other structural change that has affected support for the mainstream right is educational expansion.Upskilling and the expansion of formal education are linked to the development of the so-called information revolution and post-industrial knowledge societies, and have brought about new opportunities for the mainstream right while also disrupting previous electoral equilibria.Findings in the literature have shown education to be associated with preferences for lower levels of redistribution, but also with sociocultural liberalism.Thus, while the expansion of education is expected to increase the demand for liberal economic policies, it has introduced further heterogeneity in the potential electorate of centre-right parties by fostering the growth of sectors with liberal sociocultural preferences.Although we need to be careful not to make strong causal assumptions based on correlational evidence, the findings presented in this article suggest that mainstream right parties might be able to offset some of the negative consequences of growing heterogeneity in the electorate by adopting a relatively liberal position on sociocultural issues.Such a strategy would enable the mainstream right to secure the support of those groups that have experienced the fastest relative expansion in post-industrial societies.While the risk of losing voters to the radical right should not be underestimated, neither should the growth of educated sectors of the population, who can easily be tempted by parties promoting policies that better match their political values.
Overall, results show that, far from being a threat, educational expansion could represent an opportunity for the mainstream right.Structural changes, including secularisation, have opened the door to increased electoral volatility and facilitated the emergence of new competitors.While this is likely to increase pressures on the mainstream right coming from all sides, parties can minimise the electoral losses caused by such changes by adopting a strategy aimed at building renewed and sustainable social electoral coalitions.Such coalitions may comprise many different social groups, but what the empirical evidence suggests is that, in the context of modern societies with knowledge-based economies, strategies based on engaging in culture wars by taking on very conservative stances on social issues are unlikely to provide the mainstream right with as much long-term gain as would a more moderate strategy.Having said that, politicians do not always plan their strategies based on long-term gains.Sometimes, the need to prevent the growth or the emergence of new competitors, including the radical right, may lead parties to pursue a different path, in what Kitschelt (1994) calls 'oligopolistic' party strategies.When sustained over time, however, taking on very conservative stances on sociocultural issues is likely to gradually erode support among the growing liberal sectors of the population.
One caveat of this research is that, although it triangulates findings using both aggregate-and individual-level data, it does not look at the micro-dynamics of electoral change.For example, it might be that new generations of more educated voters have been less likely to vote for the mainstream right from the start while older generations, who have a well-established identification with those parties, are less affected by the emergence of sociocultural issues.Therefore, future research should not only look at how exactly education interacts with policy shifts on both economic and sociocultural issues, but also at whether younger voters with higher levels of education are more likely than other voters to switch to liberal and centrist parties when these emerge.Moreover, the lack of suitable aggregate-level data has prevented this article from looking at interactions between party positions and the size of particular occupational groups.As mentioned in the article, the literature has argued that preferences for redistributive policies vary among those with higher levels of education, with sociocultural professionals having more left-wing preferences than managers, self-employed professionals and technicians.The availability of this sort of data in the future should make it possible to explore with more precision the mechanisms laid out here.Finally, this piece only focussed on Christians and non-affiliated individuals, but the growth of non-Christian migrant communities in Western Europe raises questions about the ability of centre-right parties to attract those new groups of voters.While the West European centre-right tends to lack institutional links with religious organisations from other faiths, social and moral conservatism could nevertheless prove to be a useful tool to appeal to non-Christian religious individuals.This is, therefore, an important point that deserves further exploration in future research.

Notes
1. Research ethnics details can be found here https://www.europeansocialsurvey.org/about/ethics.html 2. As others have pointed out, Social democratic parties currently face a similar kind of dilemma, as the policies associated with sociocultural liberalism tend to be more appealing to educated professionals than to the more economically left-wing working class (Benedetto et al. 2020;Kitschelt 1994).This is, nevertheless, different from the earlier dilemma described by Przeworski and Sprague (1986), whose analysis does not capture the changes occurred in the past 50 years and the need for Social democratic parties to more decisively embrace multi-dimensional competition strategies.Positive; 701 Labour Groups: Positive.4. Those affiliated with a non-Christian religion have been excluded from the analysis, as there is no clear expectation about the direction of their vote in Western Europe once other variables are controlled for.5.The reason why fixed effects are introduced by party, rather than country, is because the dependent variable is measured at the party level to deal with a small handful of cases where a party was dissolved and its place was taken over by a qualitatively different political actor.Those cases are Belgium, where the mainstream right split into linguistic lines in 1968; Italy, where Christian Democracy was dissolved with the end of the first republic; and the Netherlands, where the three mainstream Christian democratic parties merged together in 1973.6.This is confirmed when the model is tested separately on the subsamples of Conservative and Christian democratic parties.

Figure 2 .
Figure 2. the electoral support of the mainstream right since 1960 across countries.

Figure 3 .
Figure 3. the effect of the mainstream right's social conservatism by level of educational expansion.note: the histogram shows the distribution of the human capital index in the sample.

Table 1 .
list of cases.

Table 3 .
individual-level model of voting for a mainstream right party (Mr).