Cohort, Life-Cycle and Period Effects in a Period of Partisan Dealignment Hila Federer-Shtayer Department of Political Science Leiden University Wassenaarseweg 52 2333 AK Leiden The Netherlands hshtayer@fsw.leidenuniv.nl First draft, any comments or suggestions are welcome Paper prepared for presentation in the workshop: Electoral Change in the 21 st Century: De-Alignment or Realignment?, at the ECPR Joint Sessions of Workshops, Münster, Germany, 22-27 March, 2010 ABSTRACT: This paper analyses the possible effect of cohort differences on patterns of electoral behaviour in a period of partisan dealignment. It argues voters electoral behaviour in relation to partisan dealignment should be based on two types of behaviour: effective electoral participation that changing of electoral behaviour. The paper examines three different alternative explanations: cohort, life-cycle and period effects. Cohort effect refers to al difference (pre-dealignment s, postdealignment and children of dealignment ). Empirical analysis of two case studies the Netherlands and Israel--demonstrates that the last two effects period and life-cycle influence the electoral behaviour of people during a period of partisan dealignment. The period effects are mainly related to institutional and party system changes. In addition I will show that ageing has crucial role in explaining one s electoral behaviour as one becomes older he/she is less likely to be a dealigned voters.
1. INTRODUCTION Since the 1970s scholars have identified a change in the connection between voters and political parties in some of the Western democratic countries: the party-affiliated proportion of the electorate shrunk as the traditional party coalitions dissolve leading to dealignment (Inglehart & Hochstein, 1972:350) or partisan dealignment (Dalton, 2000; Dalton, Beck, & Flanagan, 1984a, 1984b) In the course of time, the evidence in favor of partisan dealignment has built up (as the longitudinal data for the period of partisan dealignment accumulate and partisan dealignment evolves in more and more countries). Several explanations for this political change are presented in the literature (Dalton et al., 1984a; Dalton, Beck, & Flanagan, 1984c), but the most prominent one is the functional model. According to this model partisan dealignment occurred due to changes at macro level (party system as a whole) and at micro-level (individuals) (Dalton et al., 1984c). While in the past voting behaviour was based on partisanship or socio-structural divide (Campbell, E.Converse, Miller, & Stokes, 1960; Lipset & Rokkan, 1967); this has changed since the voters need for political cues decline. This, in turn, is due to the spread of education together with increasing diversity of information (Beck, 1984a:237-8; Dalton et al., 1984c). Therefore, party ties are no longer necessary for guiding voters in making their political choice. This psychological change, which Dalton defined as cognitive mobilization, is argued to be concentrate among groups whose greater political skills and resources enable them to be more functionally independent: the better-educated, the middle class and the young. (Dalton et al., 1984c) The role of the last group the young people -- is prominent in the theoretical and empirical literature on partisan dealignment. It has been demonstrated that the decline in partisanship is most pronounced among the young (Abramson, 1979; Beck, 1984a:233) and therefore those who tend to be more volatile and also those who tend not to participate in election (Franklin, 2004; Topf, 1995). While early studies explain the loosening connection between voters and parties by pointing out to specific young who did not follow their parents, due to different socialization process (Alt, 1984; Inglehart & Hochstein, 1972:350). Later research shows that this is also found for young people in later s (Dalton, 2000:30). This paper deals with this group of people the young electorate with loosening connection with the political parties. More specifically it focuses on the difference between three s: those who were enfranchised before the partisan dealignment period begun, these are the pre- 2
dealignment ; and those who initiated this political change, these are the postdealignment ; and the current young people who are the offspring of the postdealignment and therefore will be defined as children of dealignment. Therefore my research question is do people of the three s behave differently in a period of partisan dealignment? The identification of different s is discussed in the literature of realignment and it is an important element for explaining the phenomena of realignment. Beck (1979:131) explains the realignment processes as related to al replacement. He identifies three distinct s: realignment (those who formed their party loyalties in the crucible of realignment dominant the electorate, children of realignment are the offspring of the realignment and children of normal politics who gain their political experience during a period of stable alignment. Carmines and Stimson s model (1984) of issue evolution also distinguishes between three different s: older voters who have strong partisan predispositions and are unaffected by the new issue cleavage, younger whose partisanship was formed during the heart of political conflict and young voters who have no direct experience with the issue (Carmines & A.Stimson, 1984:140). In the dealignment literature, however, this topic is not discussed and in this research I make an attempt to contribute to it. Besides examining the al differentiation, or the formative socialization effect, i.e. cohort effect, I will also examine two other possible explanations. The first is period effect the social and political events and circumstances that occurred over the period. The second is the lifecycle effect - the effects of passage through different life stages. This research will show that the last two effects life-cycle and period influence the electoral behaviour of people during a period of partisan dealignment. The period effects are mainly related to institutional and party system changes. In addition I will show that ageing has crucial role in explaining one s electoral behaviour as one becomes older he/she is less likely to be a dealigned voters. The research includes two cases the Netherlands and Israel. These two cases exemplify the most different cases research design. They are different from each regarding two important factors. Firstly, the main conflicts and the meaning of the Left-Right axis in these two polities are 3
different. While in the Netherlands the socio-economic cleavage is the most salient divide, in Israel the main political conflict runs along the peace-security issues and the religious divide. Secondly, the starting point of the dealignment period is different. The partisan dealignment period in the Netherlands began a long time ago. The first signs of change appeared already in 1967, when little difference in the party support between the younger and older voter group has been found (Irwin & Dittrich, 1984). Evidence of partisan dealignment became more prominent in the following election, 1971. Irwin and Dittrich report (1984:296) that [w]hereas only 23 percent of the older cohort report such a switch [i.e. switch of party support, H.F.-S.] close to 40 percent of the younger cohort switched. The continuity of the dealignment partisan period until recent years is confirmed by Dalton (2004:33). This long period of partisan dealignment will enable me to examine the influence of the three possible s on the partisan dealignment process. In Israel, the partisan dealignment period began a bit later, in the beginning of 1990s and for this Shamir et al s (2008:3) argue, can be no doubt. The Israeli case, however, can also give us some very important insights into the contribution of other explanations. First, I can examine the period effect. During this short period of partisan dealignment in Israel several institutional and political changes happened. These included change of electoral system and the introduction of direct election for the prime minister and later its abolishment and the change of the party system structure with the establishment of the center party Kadima. Second, Israel is an immigrants country: since 1989 a massive wave of immigrants changed the demographic composition of the Israeli society, as 620,000 people from Former Soviet Union immigrated to Israel between 1989 and 1996 (which was 12% of the population in Israel at that time) (Horowitz, 2003:56). I can use this large group of people to examine the life-cycle influence, as I can assume that the immigrants have gone through different political socialization process. 2. THE STUDY OF PARTISAN DEALIGNMENT The empirical research for identifying partisan dealignment is done by using diverse evidence to show partisan change, such as issue voting, candidate voting, (increase) in volatility, (increase in) split voting, (decrease of) party membership, (decrease of) turnout, etc. The most important aspect, however, is one s party affiliation or partisanship, while the others are seen as consequences of the partisan change (Dalton, McAllister, & Wattenberg, 2000). 4
The study of long-term party support is usually conducted based on levels of party identification (Dalton, 2000). In this research, however, I examine this subject based on an individual s past voting record. This is done because of the criticism raised against the application of the concept of party identification in European and other multi-party systems. The first objection is what Schmitt (2002:3-4) calls the stability assumption. Thomassen who studies the 1970 Dutch provincial elections and the 1971 and 1972 Dutch parliamentary elections, was the first one to identify this. He finds that party identification is less stable than voter preference and offers the opposed casual relation. According to him: party identification is not a psychological attachment, but simply a reflection of the vote preference (Thomassen, 1976:77) (see also (Eijk & Niemöller, 1983:319)). The second objection arose in another research of the Dutch electorate. This time it was against the assumption that voters identify with only one party (or what Schmitt (2002:3-4) calls the uniqueness assumption). Van der Eijk was the first scholar to finds that more than one third of the Dutch respondents admit to have multiple party identification (Eijk & Niemöller, 1983:338). Schmitt, in the most recent research on this subject, shows that this phenomena reveals more as he finds that for 14 countries between 1996 and 2000 10.2% of the responded identified with more than one party (Schmitt, 2002:10) This is even more intriguing as scholars found out that voters identify with groups of parties (Ventura, 2001) or have only left-right orientations (Percheron & Jennings, 1981). These strong arguments show the problematic reliability of party identification model for multiparty and European countries. As Thomassen argues the concept of party identification has no real meaning in the Netherlands (Thomassen, 1976:77) and van der Eijk concludes the application of the concept of party identification in relation to voting behaviour in the Netherlands is extremely doubtful (Eijk & Niemöller, 1983:339). Beck even argues that Michigan party identification measure does not seem as appropriate outside of the United States (Beck, 1984a:234). Therefore, I offer to examine the arguments of dealignment and more specifically the declines in partisanship by studying the patterns of electoral behaviour over the years. This is not the first time to study partisanship based on this kind of data. Already in the early American literature, before the publication of the American Voter book (1960) the study of (stable) partisanship was 5
done in terms of an individual s past voting record (e.g. (Key, 1955, 1959)) and also in more recent research (Beck, 1984a). Beck also argues that increase in electoral instability is one of the two distinguishing characteristics of dealignment in most nations (Beck, 1984a:233). The phenomenon of electoral behaviour in relation to the partisan dealignment has three components: 1. casting a vote (did a person participate in the election?) 2. casting invalid vote (did the ballot is valid and counted?) 3. patterns of party support (does the voter cast his vote for the same party). These three components are important and each one of them is used as indicator for studying the phenomena of partisan dealignment (see for example (Gallagher, Laver, & Mair, 2006)). Therefore the study of one s electoral behaviour will be done based on two types of behavior: effective electoral participation (did a person participate in the election and cast a valid vote?) and changing of electoral behaviour. The second type of electoral behavior is equivalent to Ersson and Lane s (Ersson & Lane, 1998) Gross Volatility Index. This index measures not only a change in party support ( party switching ), but also takes into account all the eligible voters over the two elections and defining those changing between voting and non-voting as volatile voters (Ersson & Lane, 1998:25). These two dimensions of electoral behavior draw the line between voters who have strong party affiliation, aligned voters and those whose party affiliation is weak and therefore will be defined as dealigned voters. 3. HYPOTHESIS Recapitulating, this paper examines the possible influence of three s pre-dealignment, post-dealignment and the children of dealignment on individuals electoral behaviour in the Netherlands and Israel in their respective periods of partisan dealignment. My analysis is guided by two hypotheses and one question. First, I have to establish the argument about the presence of differences among the three s: pre- and post-dealignment and children of dealignment. 6
H1: There is a difference between the three s regarding their probability to be dealignment voters. The next hypothesis specifically deals with the expected proportion for each regarding two types of electoral behavior. I differentiate between aligned voters and dealigned voters. The first is the pre-dealignment who I expect to have a strong party affiliation, or in other words to be aligned voters. The post-dealignment is the first to have a weaker connection with parties and initiated the dealignment process, therefore I expect them to behave more as dealigned voters. The third is the children of the post-dealignment. Not only are they young people (and therefore we can assume them to be more dealigned voters), their parents are the post-dealignment and due to the combination of these two factors, I expect the proportion of dealigned voters for this to be higher than the proportion of the post-dealignment. H2: The proportion of dealigned voters for post-dealignment is higher than for the pre-dealignment, but lower than for children dealignment. To evaluate and establish the cohort effect the difference-- I have to control for two other effects: the life-cycle effect and period effect. While the first refers to long-term movements that are associated with progression through the life-cycle, the second are idiosyncratic events or circumstances occurring at particular time points (Markus, 1983:718). Therefore I will examine the following question: Q1: Can we identify life-cycle and period effects that can explain the electoral behaviour of the voters in a period of partisan dealignment? 4. THE EMPIRICAL RESEARCH Since each of my two cases the Netherlands and Israel has different circumstances I could not examine the hypotheses and the question in two cases in an identical way, therefore, the research design for each case is different. The long duration of partisan dealignment in the Netherlands (late 1960s), enables us to test the difference between the three s. In Israel, however, the dealignment period began only during the early 1990s and therefore only two different s can be found: pre-dealignment and post-dealignment. The Israeli case, however, 7
enables us to examine the voting behaviour of immigrants (which will function as control group, as will be explained below) and this aspect cannot be tested for the Dutch case, as the national election surveys do not include any question regarding this topic. All in all for each of the two cases the electoral behaviour of three different groups was examined. For the Dutch case these groups are the three dealignment s and for the Israeli case -- two dealignment s and the immigrants. I use data from the national election survey archive from the second election after the dealignment period begun until the last national election 1. For the Dutch case I examined eleven surveys (between 1972 and 2006) and for the Israeli case I examined five surveys. Some of the surveys are only pre-election surveys (1996, 1999, 2003) and some also include post-election wave (2006, 2009). The electoral behaviour of each respondent was differentiated based on his/her report on his voting behaviour in the last election (or his/her future voting behaviour in the coming election, when no data on last past election was available) and his/her voting behaviour in the preceding election. Appendix A specifies the variables per survey that were used. The study of electoral behaviour of the respondent is defined as aligned voter when the respondent reports to have participated in the two elections, cast valid votes in both of the election and supported the same party in the two consecutive elections. Respondent who did not vote in the last two elections, or cast an invalid vote, or changed his/her party support, or changed his/her electoral behaviour (for example, did not vote in one election and in the second election supports one of the parties or casts invalid ballot in one election and in the second did not vote, etc.) is defined as dealigned voter. Here I differentiate between forced and voluntarily change of party support. Forced change of party support happened as party merged and this has not been considered as a change. Regarding splits of parties, here I assume that when one moves to support the smaller fraction, this is a voluntarily change of party support and treat that accordingly. 1 Since I examined change of electoral behaviour between two consecutive elections, the research could be started only in the second election after the dealignment period has begun. 8
The (two or three) different s were examined by tracking the age groups as they move through time. This is the conventional approach to resolving the conflict between al and life-cycle explanations (Beck, 1984b:523). The dealignment period in the Netherlands began in 1967, but according to Irwin and Dittrich (1984), the difference between the two s (pre and post-dealignment) began to be higher only in 1971 election. Therefore I followed Irwin and Dittrich (1984:295) and refer to those who could vote in 1967 election as pre-dealignment and those who joined the party system for the first time in 1971 onwards, as post-dealignment. In addition, I also defined voters who were enfranchised since the 1989 election as offspring of dealignment, since 18 years passed since the first voters of the post-dealignment could vote. For the Israeli case I divide the electorate into three groups: pre-dealignment (those who were enfranchised before the 1992 election), post-dealignment (those who were enfranchised after the 1992 election) and the immigrants (those who immigrated to Israel in 1989 or after). The immigrants group can help me control for the life-cycle effect as I am able to track the same life-cycle influence for people whose socialization process occurred in different political environments. Therefore I include only those voters who are in the same age range as the predealignment in this group. The lowest age for this group is 26 years old, and therefore we can assume that the immigrants experienced their political socialization process in a different country, as research shows that the crucial age is until the age of 25 (Kent & Niemi, 1981:8).Younger immigrant voters were omitted from the study as for them I cannot control for their socialization process. Appendix B specifies the age range of each one of the s, over the years, for each empirical case. The more important variable for controlling life-cycle effect is the age variable. This is continuous variable which was recoded from 0 (the youngest voter) to 1 (the oldest voter). The period effect is tested by examining the probability of people being dealigned voters in each election year. Therefore dummy variables were created for each election year, where the first year 1972 (for the Netherlands) and 1992 (for Israel) was the reference time-point. In addition, I control for the effect of cognitive mobilization. As Dalton s explanation for partisan dealignment argues (Dalton, 1996) that dealignment seems to be concentrated not only among the 9
youngest, but also among the better educated and post-materialists voters. These Apartisan voters are cognitively mobilized, which implies high levels of political involvement and sophistication, though these citizens remain unattached to any political party, since they have the political resources to follow the complexities of politics, and they are free of affective party ties (Dalton, 2006:195; Dalton et al., 1984c:462). Therefore, I have created a cognitive mobilization variable which is a combination of the indicators: level of education and political interest. 2 5. THE EMPIRICAL ANALYSIS To test my hypotheses I use a binary logit model. Since the research designed for each case was a bit different, I will examine each case separately. 1. The Dutch case I start by looking for evidence to support my arguments about the differences among the three s in terms of their likelihood to be dealigned voters. For this I first ran two models of binary logistic regression. The dependent variable of the two models dealigned voters was coded 0=alignment voters and 1=dealigned voters. The independent variable of the two models was different. In the first model (model no. 1.1) it was one dummy variable for the two postdealignment s (post-dealignment and children of dealignment, and the pre-dealignment was the reference group). In the second model (model no. 1.2) I differentiated between the two last s and created a dummy variable for each post-dealignment (again the reference group was the pre-dealignment ). The logit model shows that the Chi-square for the first model (model 1.1) was 336.640 (p=0) and for the second model (model 1.2) it was 382.667 (p=0). Put differently, the two models show that there are better predictors than the null model (which includes only the constant), as the Chi- 2 Regarding the political interest variable for the Dutch case, I have used the political interest index (and recoded the same index for 1971 and 1972 survey). For Israel I have used the survey question on political discussion. This question is changed a bit over the years. In 1996 and 1999 election surveys the question was: How frequently do you discuss politics with friends or family? (the answers were: often, sometimes, little or never). In 2003-2009 election surveys the question was: To what extent can you and your friends influence government policy? (the answers were: great extent, certain extent, not very much, not at all). All in al, this variable in the two cases has range of 1(lowest) -4 (highest)). Regarding the education variable while for the Israeli case I used the variable years of schooling, for the Dutch case, I recode the question on education level according to 1989 election survey codebook (as this survey has the lowest number of categories) (in total there are four categories: 1- elementary school; 2 elementary school plus 3- secondary level education; 4- post secondary level education (undergraduate and graduate level). 10
square of the two models were statistically significant. Moreover, model no. 1.2 which examines the difference between the three s is better model, as the value of the Chi-square was higher. After establishing that there is statistical support for dividing the s into three groups, I turned to examine the trends of electoral behaviour of the three s over the years, based on aggregate data. Table 1 percentages of dealigned voters, for each in the Netherlands, between 1972-2006 Election year Predealignment Difference Postdealignment difference Children of difference dealignment 1972 35.8 52.9 1977 31.6-4.2 52.7-0.2 1981 28.2-3.4 47.7-5 1982 29 0.8 43.7-4 1986 26.1-2.9 42.2-1.5 1989 23.6-2.5 34.6-7.6 1994 34.9 11.3 46.7 12.1 56.8 1998 35 0.1 43.8-2.9 57.7 0.9 2002 40.8 5.8 48.7 4.9 54-3.7 2003 27.6-13.2 36.2-12.5 35.3-18.7 2006 33.6 6 45.1 8.9 54.6 19.3 Table 1 presents the proportion of dealigned voters, for the three s between 1972 and 2006. The percentages in the table give an indication that the trends are in the expected directions; the proportion of dealigned voters for the post-dealignment is higher than the proportion for the pre-dealignment over the whole period. In addition the percentages of dealigned voters for the post-dealignment were lower than those for the children dealignment, except for the 2003 election, but the difference was only 0.9%. Examining the change in the percentages of dealigned voters for each gives us some interesting indications. From the 1972 election to 1989 the proportion of dealigned voters between two consecutive elections decreased, with the exception of the 1982 election where the percentages increased with 0.8%. Since 1994 onwards we see a fluctuation in the level of dealigned voters, but the type of trends decrease or increase is almost identical for the three s. There are two exceptions to that. The first is 1998 election, where the proportion of dealigned voters for the pre-dealignment stayed almost stable. For the post- 11
dealignment it decreased by 2.9% and for the children it increased by 0.9%. The second exception is 2002 election in which the percentages of dealigned voters increased substantially for the pre- and post-dealignment, but decreased substantially for the dealignment. These almost identical trends might indicate that there are events or other changes which influenced the electorate, in other words there might be a period effect. Therefore, I turn to examining the life-cycle and period effects. For this I run binary logit model with two dummy variables for s pre-dealignment and children of dealignment (therefore the reference group is the post-dealignment ), age (which is recoded between 0 and 1), and dummy variable for each election year (the 1972 election year was the reference group). I initially included a variable for cognitive mobilization, but when running this model, I found out that the coefficient of the cognitive mobilization variable (as interaction variable between education and political interests) was very low and not significant. Therefore I omitted this variable (and its constitutive terms) from the model model number 1.3. Table 2 shows the raw scores for the binary logistic regression coefficients of dealigned voters for all the independent variables in model number 1.3. Model 1.3 gives a statistically significant improvement in fit over the model with evaluate only cohort effect (model no. 1.2), as the Chi-squared more than doubled (!) (681.95, p=0.00). This model gives us an interesting picture of the three different effects cohort, life-cycle and period. This model estimates that the s are not differ regarding their probabilities for being dealigned voters. The odds coefficients were not statistically significant, but also the odds show the opposite direction, the odds for being a dealigned voter for the pre-dealignment were higher than those for the post-dealignment and those for the children of dealignment were lower than those for the post-dealignment. The confidence intervals which ranged between the directions of increase and decrease and the fact that the odds were not statistically significant indicate that this estimation does not have any statistical basis. I received very similar results when I changed the group of reference of the two dummies variables for the effect to the pre-dealignment. 3 Therefore we can 3 The Exp(B) for post-dealignment was 0.986, c.i. 0.887-1.095 (p=0.79); and for the children dealignment it was 0.936, c.i. 0.789-1.109 (p=0.44). 12
say that with the introduction of the other effects life-cycle and period, the cohort effect disappeared. Table 2: Model No. 1.3 - Binary Logistic Regression Analysis: prediction of dealigned voters in the Netherlands, 1972-2006 Predictor Variable B (S.E.) Exp(B) 95% Confidence constant.045 (.057) 1.046 Interval for Exp(B) pre-dealignment.014 (.054) 1.014.913 1.127 children of dealignment -.052 (.066).949.834 1.081 age -1.932 (.131)***.145.112.187 1977 election -.045 (.075).956.826 1.107 1981 election -.134 (.071).875.760 1.006 1982 election -.154 (.072)*.857.744.987 1986 election -.172 (0.77)*.842.723.980 1994 election.285 (.076)*** 1.330 1.146 1.544 1998 election.334 (.076)*** 1.396 1.204 1.619 2002 election.586 (.079)*** 1.798 1.539 2.099 2003 election.070 (0.84) 1.073.909 1.266 2006 election.573 (.077)*** 1.774 1.525 2.064 Chi-squared 681.953 Nagelkerke R Square.057 Log Liklihood 20668.503 N 15993 Note: The dependent variable is coded 0 if respondent voted for the same party in the two consecutive elections and 1 if the respondent changed its electoral behaviour pattern or his/her electoral behavior is not effective. *p<=0.05, ** p<=0.01, *** p<=0.001 Regarding the other two effects, the binary model confirms these two. The Exp(B) of the age variable indicates us that when one ages, the odds of being a dealigned voter decreases. Regarding the period effects, the model renders that between 1977 and 1986 the odds of being a dealigned voter were lower than those for 1972 (but this is not statistically significant for the first two elections -1977 and 1982). From 1994 election onwards the odds of being a dealigned voter are higher than those for 1972 and all of them are statistically significant (with the exception of the 2003 election). 13
2. The Israeli case For the Israeli case, I also ran a binary logistic regression to establish the difference between the three groups pre- and post-dealignment and the control group, the immigrants-- about their likelihood to be dealigned voters (model no. 2.1). The dependent variable dealigned voters was coded same way as for the Dutch case (i.e. 0=alignment voters and 1=dealignment voters). There were two dummy variables as independent variable. The first one for the post-dealignment and the second for the immigrants (therefore the group of reference is the predealignment ). The chi square value of model 2.1 was 85.856 (p=0.00) and allows rejection of the null model accordingly all coefficients in the model expect the constant are zero. Therefore I assume that the three groups: pre- and post-dealignment s and the group of immigrants are different in their odds of being dealigned voters. Table 4 percentages of dealigned voters, for each group in Israel, between 1996-2009 Pre-dealignment Difference Postdealignment Difference Immigrants difference 1996 28.2 35.9 58.8 1999 30.8 +2.6 37.9 +2 28-30.8 2003 35.7 +4.9 51 +13.1 57.9 +29.9 2006 66.3 +30.6 66.9 +15.9 89.5 +31.6 2009 49.3-17 64.6-2.3 49.5-40 Table 3 presents the differences between the three groups over time in percentages based on aggregate data. The data indicate that the percentages of dealigned voters for the postdealignment s are higher than those for the pre-dealignment over the entire period, but in 2006 the difference between the two groups is only 0.6% which is according to my expectations. Not less interesting are the numbers for the immigrants group. In three out of five elections (1996, 2003, 2006) the proportion of dealigned voters for immigrants were high not only for the pre-dealignment, but also for the post-dealignment. In the 1999 elections the proportion was the lowest compared to the other two groups and in the 2009 elections the percentages of dealigned voters for immigrants were almost identical to the predealignment (the difference was only 0.2%!) and lower than those for the post- 14
dealignment. The fluctuation in percentages for the immigrant group in comparison to the other two groups does not support any explanation for cohort or life-cycle effect. The trends over the years of the groups are identical (!), the proportion of dealigned voters between 1996 and 2006 increased from year to year and decreased in 2009; the only exception is the decrease for the immigrants group in 1996 election. This strong indication for a period effect caused me to run another binary logit model. First, I ran the full model for examining the life-cycle and period effects, controlling for cognitive mobilization. Similar to the Dutch case, the coefficients for the cognitive mobilization variable were not statistically significant, so I omit this variable, together with its constitutive terms (i.e. education and political interest) from the model. The model (model 2.2) includes the following independent variables: two dummies variable of post-dealignment and children of dealignment (thus the reference group is the pre-dealignment ); four dummy variables for year election since 1999 onwards (thus 1996 election function as reference group); and the variable of age. Table 4 shows the raw score binary logistic regression coefficients of dealigned voters for all the independent variables (model 2.2). The model gives a substantial statistical improvement in fit over model 2.1 (which only controls for the effect), as the Chi-squared is 611.01 (p=0.00). The model rejects our hypothesis of cohort effect as the odds were in the wrong direction as the odds for the post-dealignment were lower than those for the pre-dealignment (but this was not statistically significant). I computed very similar findings when I changed the group of reference to the pre-dealignment. 4 This together with the confidence intervals (which ranged between the two directions of increase and decrease) leads me to reject the cohort effect hypothesis. Similar to the Dutch case, the cohort effect disappeared when introducing the other effects life cycle and period. Regarding the immigrants group I got an interesting picture. The odds of members of this group being dealigned voters are not only higher compared to the pre-dealignment but also 4 The Exp(B) for pre-dealignment was 1.073, c.i. 0.886-1.300 (p=0.47). 15
compared to the post-dealignment. 5 Therefore we cannot reject the possible life-cycle effect. The high negative and statistically significant odds of the age variable point to a life-cycle effect; when one becomes older, the odds of being a dealigned voter decreases. For the period effect, I find evidence for strong and statistically effect as since the 2003 election the odds of being a dealigned voter are higher compared to the 1996 election. Table 5: Model No. 2.2 Binary Logistic Regression Analysis: prediction of dealigned voters in Israel, 1996-2009 Predictor Variable B (S.E.) Exp(B) 95% Confidence Constant -.430 (.095)***.651 Interval for Exp(B) Post-dealignment -.070 (.098).932.770 1.129 Immigrants.751 (.100)*** 2.119 1.742 2.577 Age -1.486 (.212)***.226.149.343 1999 election -.016 (.101).984.807 1.200 2003 election.572 (.097)*** 1.771 1.466 2.140 2006 election 1.813 (.098)*** 6.132 5.060 7.431 2009 election 1.093 (.110)*** 2.983 2.406 3.699 Chi-squared 611.005 Nagelkerke R Square.162 Log Liklihood 5900.447 N 4719 Note: The dependent variable is coded 0 if respondent voted for the same party in the two consecutive elections and 1 if the respondent changed its electoral behaviour pattern, or his/her electoral behavior is not effective. *p<=0.05, ** p<=0.01, *** p<=0.001 6. DISCUSSION AND CONCLUSIONS This paper deals with the influence of different s on party affiliation in a period of partisan dealignment. It proposes to differentiate between three different possible s: pre-dealignment, post-dealignment and children of dealignment. The party affiliation of an individual is examined by aid of two components of electoral behavior: effective electoral participation and the change in patterns of electoral behavior. 5 In a similar model in which the reference group was the post-dealignment ; the Exp(B) for the immigrant group was 2.273, c.i. 1.737-2.974. 16
Besides the cohort effect (i.e. al), the paper also examines the influence of life-cycle explanations (i.e. age) and period effects (social or political events or change of circumstances). This is done in two different case studies the Netherlands and Israel. The aggregate data for the two cases gave us an indication of the cohort effect. This is expressed by the fact that the proportions of dealigned voters for post-dealignment were higher than those of the pre-dealignment and lower than those for the children dealignment. The logit models include life-cycle effect (i.e. age) and period effect (i.e. election year) as well and reject the cohort explanation as the direction of the odds in the two cases were opposite and not statistically significant. The life-cycle theory finds support here. In the two cases I find that when one s become older the odds for being dealigned voter decreases. This evidence fits the life-cycle partisanship model accordingly young adults will have weaker partisan loyalties than their elders. They are likely to strengthen their party identification through party support, therefore their partisanship strengthens as they become older (Abramson, 1992). Not less interesting are the period effects which were found in the two cases. The period effect for the Dutch political system can be explained mainly by party system change. The decrease in the proportions of dealigned voters between 1977 and 1989 compared to 1974 election can be explained as the effect of the electoral alliance between the Anti-Revolutionary Party (ARP); the Christian Historic Union (CHU) and the Catholic People s Party (KVP), which later unified into a single party, the Christian Democratic Appeal (CDA). In 1989 the same effect occurred as the Evangelical People s Party (EVP), The Radical Political Party (PRP), the Pacifist Socialist Party (PSP) and the Communist Party of the Netherlands (CPN) merged to form the Green-Left (GL). The 1994 election is the turning point, since from this election onwards the proportion of dealigned voters increased, compared to the 1972 election. This is the only election year which cannot be explained by changes in the party system. In this election the two government parties CDA and the Labour party (PvdA) lost votes severely. While the electoral support for the PvdA decreased in almost 8% points compared to the preceding election, the decrease for the CDA was more than 13% points. There are several explanations for this major change: faultiness election campaign of the biggest party, the CDA and its leadership crises and public opposition to changes 17
in pensions (AOV), which was discussed by the government (Bosmans, 1999 ). As a result of the 1994 election, the CDA for the first time did not take part in the government coalition, and the support for the Purple coalition (with the PvdA and the Liberal Party (VVD)) is seen in the following election, in 1998. A very high proportion of dealigned voters is measured in 2002 and this is the effect of the new extreme-right party List Pim Fortuyn (LPF) (which received 17% of the votes). The coefficients for the 2003 elections were not statistically significant and the aggregate data indicates that proportion of dealiged voters decreased (compared to the 2002 election), and this can also be a aftershock effect of the high electoral support for LPF in 2002 election. Change in the Dutch party system structure happened also in 2006 as two new parties which represent new topics run to the parliament for the first time; Freedom Party (PVV) (Anti Islam), and Animals Party (PvdD). The period effects for the Israeli period can also be explained by political system change but also institutional change. The 1996 election which was the reference group in the statistical model was the first national election held accordingly to the new electoral system in which each voter was allocated two votes; one ballot for the Israeli parliament and another for the prime-minister (Kenig, Rahat, & Hazan, 2005). The second one was the 1999 election. This probably explains the downward trend that was identified in the analysis of the aggregate data and also by the nonsignificant coefficient in the logit model. The higher proportions of dealigned voters in the 2003 election are probably due to the second reform in the Israeli electoral system as the direct election for prime-minister was canceled. This finding supports Kenig et al s arguments (2005) regarding the influence of the electoral system reform on voting behaviour (such as the creation of patterns of split voting) (Kenig et al., 2005) (for the opposite argument see (Andersen & Yaish, 2003)). Not less interesting are the results for the 2006 election. The higher proportions of dealigned voters are probably due to the emergence of the hybrid-center party Kadima. The same results were found for the 2009 election. This is probably due to the change in the party system structure from two blocs (Right-Left) to three blocs (Left-Center-Right) (as was already named by Samoha the big bang (Smuha, 2006) (see also (Shamir et al., 2008)). The statistical evidence in favour of a higher proportion of dealinged voters in the two last elections measured against the 1996 election, implies that this party system change has much more influence than the first institutional change (i.e., the introduction of new electoral system). This supports Shamir et al s observation the 1992 change in the Law of Direct Election for the Prime Minister may have been a catalyst in the acceleration of this trend, but not its major source. (Shamir et al., 2008:4) 18
All in all we can say that electoral behaviour of voters over period of dealignment cannot be explained by the cohort effect, but by life-cycle and period effects. While the first is sociologicalpsychological explanations - ageing, the second is mainly explained by institutional and party system change. Since partisan dealignment as Dalton (2006:191) state appears to be a continuing feature of contemporary politics, it is important to understand its dynamic and transitions over time. My suggestions for future research include the development of a more comprehensive model that explains the dealignment processes as a function of life-cycle effects (age) and institutional and party system change. In addition, there is room for more extensive research on the cohort effect which includes a larger number of cases. 19
Appendix A: variables of electoral behaviour, per survey for the Netherlands (NL) and Israel (IL) Did you vote (election before) Party support (election before) Did you vote (latest election) Party support (latest election) NL 1972 election V154 V155 V143 V150 NL 1977 election V94 V95 V315 V323 NL 1981 election V016 V017 V512 V513 NL 1982 election V1171 V1172 V1045 V1046 NL 1986 election V203 V204 V180 V181 NL 1989 election V055 V056 V146 V147 NL 1994 election V055 V056 V280 V281 NL 1998 election V0165 V0166 V0610 V0611 NL 2002 election V0235 V0236 V0646 V0647 NL 2003 election V0646 V0647 X0195 X0196 NL 2006 election V220 V 221 V510 V512 IL 1996 election CCC37 CCC23 IL 1999 election C17 C12 IL 2003 election B70 B63 IL 2006 election C40 D5 D6 IL 2009 election V162 Q1 Q3 Appendix B: The different groups in each election survey, per case study The Netherlands Group no. 1: Pre-dealignment Group no. 2: Post-dealignment Group no. 3: Children of dealignment 1972 election 23 - older 18-22 1977 28 - older 18-27 1981 32 - older 18-31 1982 33 - older 18-32 1986 37 - older 18-36 1989 40 - older 21-39 18-20 1994 45 - older 26-44 18-25 1998 49 - older 31-48 18-30 2002 53 - older 35-52 18-34 2003 54 - older 36-53 18-35 2006 57 - older 39-56 18-38 Israel Group no. 1: Pre-dealignment 1996 26-older 18-25 1999 29-older 18-28 2003 33-older 18-32 2006 36-older 18-35 2009 39-older 18-38 Group no. 2: Post-dealignment Group no. 3: new immigrants who are in the same age range of the predealignment 20
References List Abramson, P. R. (1979). Developing Party Identification: A further Examination of Life-Cycle, Generational, and Period Effects. American Journal of Political Science, 23(1), 78-96. Alt, J. E. (1984). Dealignment and the Dynamic of Partisanship in Britain. In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 298-329). Princeton: Princeton University Press. Andersen, R., & Yaish, M. (2003). Social Cleavages, electoral reform and party choice: Israel s natural experiment Electoral Studies 22, 399-423. Beck, P. A. (1979). The electoral Cycle and Patterns of American Politics. British Journal of Political Science 9, 129-156. Beck, P. A. (1984a). Patterns of Dealignment In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 233-239). Princeton: Princeton University Press. Beck, P. A. (1984b). Young vs. Old in 1984: Generations and life stage in Presidential Nomination Politics. PS, 17(3), 515-524. Bosmans, J. (1999 ). Staatkundige Vormgeving In Nederlands- Dell 2 De Tijd Na 1940 (12 ed.). Assen-Mastrict Van Gorcum. Campbell, B. A., E.Converse, P., Miller, W. E., & Stokes, D. E. (1960). The American Voter. New York: John Wiley and Sons, INC. Carmines, E. G., & A.Stimson, J. (1984). The Dynamics of Issue Evolution: The United States. In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 134-153). Princeton: Princeton University Press. Dalton, R. J. (1996). Citizen Politics - Public opinion and Political Parties in Advanced Industrial Democracies (Second ed.). Chthaman, New Jersey: Chatham House Publishers, INC. Dalton, R. J. (2000). The Decline of Party Identification. In R. J. Dalton & M. P. Wattenberg (Eds.), Parties without Partisans - Political Change in Advanced Industrial Democracies (pp. 19-36). New York: Oxford University Press. Dalton, R. J. (2004). Democratic Challenges, Democratic Choices - The erosion of political support in advanced industrial democracies Oxford: Oxford University Press Dalton, R. J. (2006). Citizen Politics - Public Opinion And Political Parties In Advanced Industrial Democracies (Fourth ed.). Washington, DC: CQ Press. Dalton, R. J., Beck, P. A., & Flanagan, S. C. (1984a). Electoral Change in Advanced Industrial Democracies. In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 3-22). Princeton: Princeton University Press. Dalton, R. J., Beck, P. A., & Flanagan, S. C. (1984b). Electoral Change in Advanced Industrial Democracies - Realignment or Dealignment? Princeton: Princeton University Press. Dalton, R. J., Beck, P. A., & Flanagan, S. C. (1984c). Political Forces and Partisan Change In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 451-476). Princeton: Princeton University Press. Dalton, R. J., McAllister, I., & Wattenberg, M. P. (2000). The Consequences of Partisan Dealignment. In R. J. Dalton & M. P. Wattenberg (Eds.), Parties without Partisans - 21
Political Change in Advanced Industrial Democracies (pp. 37-61). New York: Oxford University Press. Eijk, C. V. d., & Niemöller, B. (1983). Electoral Change in the Netherlands - empirical results and methods of measurement. Amsterdam: CT Press. Ersson, S., & Lane, J.-E. (1998). Electoral Instability and Party System Change in Western Europe. In P. Pennings & L. Jan-Erik (Eds.), Comparing Party System Change (pp. 23-39). London and New York: Routledge. Franklin, M. N. (2004). Voter Turnout and The Dynamic of Electoral Competition in Established Democracies Since 1945. Cambridge: Cambridge University Press. Gallagher, M., Laver, M., & Mair, P. (2006). Representative Government In Modern Europe (Fourth ed.). New York (etc.): Mc Graw Hill. Horowitz, T. (2003). The Increasing Political Power of Immigrants from the Former Soviet Union in Israel: From Passive Citizenship to Active Citizenship. International Immigration, 41(1), 47-71. Inglehart, R., & Hochstein, A. (1972). Alignment and Dealignment of the Electorate in France and the United States. Comparative Political Studies, 5(3), 343-373. Irwin, G., & Dittrich, K. (1984). And the Walls Came Tumbling Down: Party Dealignment in The Netherlands. In R. J. Dalton, P. A. Beck & S. C. Flanagan (Eds.), Electoral Change in Advanced Industrial Democracies- Realignment or Dealignment? (pp. 267-297). Princeton: Princeton University Press. Kenig, O., Rahat, G., & Hazan, R. Y. (2005). The political consequences of the introduction and the repeal of direct elections for the prime minister. In A. Arian & M. Shamir (Eds.), The elections in Israel-2003 (pp. 33-62). New Jersey: Transcation Publishhers. Kent, J. M., & Niemi, R. G. (1981). Generations and Politics: a panel study of young adults and their parents. New Jersey Princeton University Press. Key, V. O. (1955). A Theory of Critical Elections. Journal of Politics, XVII, 3-18. Key, V. O. (1959). Secular Realignment and the Party System. Journal of Politics 21, 198-210. Lipset, S. M., & Rokkan, S. (1967). Cleavage Structures, Party Systems and Voter Alignments. In S. M. Lipset & S. Rokkan (Eds.), Party systems and voter alignments: Cross-national perspectives (pp. 1-64). New York: The Free Press. Markus, G. B. (1983). Dynamic Modeling of Cohort Change: The case of Political Partisanship. American Journal of Political Science 27, 717-739. Percheron, A., & Jennings, M. K. (1981). Political Continuities in French Families: A New Perspective on an Old Controversy. Comparative Politics, 13(4), 421-436. Schmitt, H. (2002). Multiple Party Identifications, the Comparative Study of Electoral Systems the WZB in Berlin: (http://www.mzes.unimannheim.de/projekte/manifestos/schmitt/berlin.pdf) Retrieved 30/1 30/10/2009. Shamir, M., Ventura, R., Arian, A., & Keidar, O. (2008). Kadima - Forward in a Dealigned Party System. (http://web.mit.edu/okedar/www/materials/sham-ven-ar-ked.pdf) Retrieved 02/02, 2010. Smuha, S. (2006). The Big Bang in Israeli Politics. Ofakim (www.ofakim.org.il) Retrieved 18-06 2009 (in Hebrew). Thomassen, J. (1976). Party Identification as a Cross-National Concept: Its meaning in the Netherlands In I. Budge, I. Crewe & D. Farlie (Eds.), Party Identification and Beyond - representations of voting and party competition (pp. 63-79). London (etc.) John Wiley & Sons 22