اقتصاد سیاسی امور مهاجرت در یک دموکراسی مستقیم :مورد سوئیس

In this paper, we analyze the determination of immigration policy in a direct democracy setting. We formulate a model of voting and participation behavior integrating instrumental and expressive motivations. The model is estimated using data drawn from a survey carried out after a vote in Switzerland in 2000 on a popular initiative proposing to implement immigration restrictions. The model enables us to recover estimates of participation costs and preferences towards immigration and analyze how these preferences are translated into actual voting outcomes. The results reveal a substantial gap (“participation bias”) between attitudes towards immigration in the general population (43% favorable to restrictions) and the outcome of the vote (26%).

In many countries of the Northern Hemisphere, opinion polls show that a majority of residents would prefer to reduce the number of immigrants to their country. For example, in the 1995 survey of the International Social Survey Programme (ISSP), more than 50% of respondents in 20 countries say that the number of immigrants should be reduced a little or a lot. Even in the four countries of the sample where this is not the case (Ireland, Spain, Japan and Canada) more respondents are in favor of reducing immigration than in favor of increasing immigration. These responses reveal a discrepancy between the immigration policies of these countries and popular demands for tighter immigration control. Here we have a puzzle for the analysis of the political economy of migration: Why is public opinion not accurately reflected in actual policies? According to Chiswick and Hatton (2003) this puzzle should be addressed by answering two questions: (1) what drives public opinion and (2) why is it not reflected in policy? According to Rodrik (1995), an adequate description of individual preferences should indeed be the first element of a political economy model. However, understanding how preferences on immigration are formed is not enough: the model must also “contain a description of how these individual preferences are aggregated and channeled (…) into political demands for a particular policy or another”. Finally, the policymakers' preferences and the institutional setting should be specified. It is in these latter elements that an explanation for the immigration policy puzzle should be sought. In this paper, we address Chiswick and Hatton's puzzle in a direct democracy setting. We formulate a model of voting and participation behavior in order to analyze the political economy of immigration policy in the Swiss context. The model is estimated using data collected after a vote in September 2000 on a popular initiative proposing to restrict immigration. The model enables us, on the one hand, to recover estimates of preferences towards immigration and to analyze, on the other hand, how these preferences are translated into actual political outcomes in the context of a direct democracy. Let us consider these two aspects in turn. First, we find that there is a substantial gap between general attitudes towards immigration and the outcome of the vote. The survey provides information on individual attitudes only for those individuals who participated in the vote. However, our structural model enables us to identify the underlying preferences of the entire population. According to our estimations, there is a substantial “participation bias”: 43% of Swiss citizens are in favor of immigration restrictions but, among those who participated in the vote, only 28% voted in favor of the popular initiative. This difference between the outcome of the vote and underlying attitudes can mainly be attributed to unobserved factors: citizens in favor of immigration restrictions tend to have higher participation costs. This can be intuitively understood in the following way: economically literate citizens who fear that the initiative would result in a large aggregate economic loss also have low participation costs because of their informational advantage over other citizens. Second, the use of a structural model that accounts explicitly for the participation decision is crucial if one wants to analyze the determinants of immigration preferences. Simple descriptive evidence for voters (see Table 1) seems to indicate that individuals with high earnings are more inclined to accept immigration restrictions than those with low earnings (0.309 vs. 0.250), the opposite of our findings in the estimated structural model. Moreover, if preferences towards immigration are estimated by a simple probit using only the subsample of voters (see Table 2), human capital variables do not seem to have a significant influence on individual attitudes. By contrast, the structural estimates indicate that attitudes towards immigration restrictions depend negatively and significantly on human capital. Table 1. Summary statistics.a Variable Mean Standard deviation Vote 0.288 Politics Left 0.098 Center/indifferent 0.270 Right 0.508 Earnings Low 0.250 High 0.309 Participation 0.532 Politics Left 0.616 Center/indifferent 0.451 Right 0.696 Earnings Low 0.377 High 0.687 Education (years) 12.488 2.128 Potential earnings 5.848 1.570 Political: right 0.190 Political: left 0.208 Facilitated postal vote 0.778 Vote at the canton level 0.780 Share foreigners in agglomeration 0.196 0.066 French part 0.228 Age 47.805 17.522 Female 0.498 a There are 953 observations for all variables except the vote (507 observations). Standard deviations are only given for continuous variables. Earnings are classified as “low” if they are below the median. Table options Table 2. Simple probit estimation results for the popular initiative in 2000a (dependent variable: voting preferencesb). Voting preferences Model (1) (2) (3) (4) (5) (6) Education (years) −0.040 −0.039 −0.029 (0.031) (0.031) (0.029) Potential earnings −0.086 −0.086 −0.063 (0.054) (0.054) (0.052) Share foreigners in agglom. −0.975 −0.949 −0.710 −0.652 −0.497 −0.454 (1.026) (1.026) (1.109) (1.112) (1.071) (1.074) Political: right 0.620⁎⁎⁎0.620⁎⁎⁎ 0.624⁎⁎⁎0.624⁎⁎⁎ 0.616⁎⁎⁎0.616⁎⁎⁎ 0.620⁎⁎⁎0.620⁎⁎⁎ (0.143) (0.143) (0.143) (0.143) Political: left −0.652⁎⁎⁎−0.652⁎⁎⁎ −0.654⁎⁎⁎−0.654⁎⁎⁎ −0.646⁎⁎⁎−0.646⁎⁎⁎ −0.648⁎⁎⁎−0.648⁎⁎⁎ (0.183) (0.183) (0.183) (0.183) French part −0.113 −0.125 −0.204 −0.212 (0.176) (0.176) (0.172) (0.171) Age −0.063⁎⁎⁎−0.063⁎⁎⁎ −0.042⁎−0.042⁎ −0.063⁎⁎⁎−0.063⁎⁎⁎ −0.042⁎−0.042⁎ −0.053⁎⁎−0.053⁎⁎ −0.038 (0.022) (0.025) (0.022) (0.025) (0.021) (0.024) Age2/1000 0.651⁎⁎⁎0.651⁎⁎⁎ 0.496⁎⁎0.496⁎⁎ 0.652⁎⁎⁎0.652⁎⁎⁎ 0.496⁎⁎0.496⁎⁎ 0.595⁎⁎⁎0.595⁎⁎⁎ 0.479⁎⁎0.479⁎⁎ (0.213) (0.231) (0.213) (0.232) (0.205) (0.224) Female −0.304⁎⁎−0.304⁎⁎ −0.317⁎⁎−0.317⁎⁎ −0.297⁎⁎−0.297⁎⁎ −0.310⁎⁎−0.310⁎⁎ −0.362⁎⁎⁎−0.362⁎⁎⁎ −0.372⁎⁎⁎−0.372⁎⁎⁎ (0.133) (0.134) (0.134) (0.134) (0.127) (0.128) Constant 1.541⁎⁎1.541⁎⁎ 0.952 1.501⁎⁎1.501⁎⁎ 0.920 1.097 0.672 (0.726) (0.591) (0.729) (0.593) (0.684) (0.557) Log_likelihood −268.883 −268.471 −268.674 −268.218 −291.879 −291.595 a Standard errors in parentheses, ⁎⁎ significant at 10%; ⁎⁎⁎⁎ significant at 5%; ⁎⁎⁎⁎⁎⁎ significant at 1%, N = 507. b The binary variable of the probit equation is coded as follows: yes for immigration restrictions =1, no =0. Table options Our empirically oriented model combines instrumental and expressive motivations for voting and is consistent with group-based motivations which figure prominently among recent explanations of voter turnout (Dhillon and Peralta, 2002 and Feddersen, 2004). Instrumental theories assume that the individual's contribution to the outcome of the vote is the main motivation for participation. Individuals participate in the vote if the expected benefit from voting is higher than the cost of voting. Although the probability of casting the decisive vote is extremely small in large electorates (“voting paradox”), empirical studies suggest that instrumental motivations matter at the margin. Group-based models are able to account for these empirical results by postulating a model structure where a reduced number of agents (i.e. groups) interact strategically. Two versions of group-based models can be distinguished. The first one, due to Shachar and Nalebuff (1999), assumes that each group has a leader who expends effort to motivate her followers to vote. The leaders' effort level (measured by political advertising and grassroots campaigning in a later paper, Shachar, 2009) is determined by strategic interaction and depends in equilibrium on the pivotalness of the circumscription. In the “ethical” version of group-based models (Coate and Conlin, 2004 and Coate et al., 2008), individuals are seen as rule-utilitarians who participate in the vote if the ethical benefit from voting is greater than the cost of voting. As decisions are based on the outcome for the entire group, these models predict realistic participation rates. Our model can be reinterpreted as a simplified version of an ethical group-based model. In addition, we account for expressive motivations by introducing social identity or self-image (Akerlof and Kranton, 2000). In our model, the voting decision is endogenously determined and voting preferences influence the participation decision. First, an individual decides whether she is in favor or opposed to the initiative proposing immigration restrictions. Second, she decides to participate in the vote if the intensity of her preferences (i.e. absolute value) is greater than the participation costs. As voting decisions are only observed for the individuals who participate in the vote, our econometric model resembles a Heckman selection model with two major differences: (i) both the outcome and selection variables are qualitative and (ii) there is a strong non-linearity in our model (the selection equation depends on the absolute value of the latent preference variable). Identification conditions require that there is at least one variable in the preference equation that is excluded from the participation cost. In our model, identification is ensured by variables of political identity and is based on the assumption that expressive motivations are relevant for the participation decision only to the extent that they influence the intensity of preferences, in favor or against the popular initiative. This identifying assumption reflects the strong polarization along the left–right spectrum which occurred in Switzerland in the 1990s when an established right-wing party (SVP/UDC) turned to systematic opposition against immigration. This situation is similar to most European countries but differs from the US, where positions on immigration policy cut across party lines. Most studies on attitudes towards migration rely on opinion polls (Scheve and Slaughter, 2001, Mayda, 2006, O'Rourke and Sinnott, 2006, Hanson et al., 2007 and Facchini and Mayda, 2009) that are likely to suffer from the so-called “hypothetical bias” since individuals have little incentive to reveal their true preferences knowing that their answer will have no real consequences. A vote on a popular initiative, such as ours, provides a context which differs fundamentally from opinion polls. First, the result of the vote is binding: the acceptance of a popular initiative by vote implies a change in the Swiss Constitution.1 Second, the political discussion preceding a vote enables individuals to take a more informed decision on the issue up for vote than is the case when answering an opinion poll. In representative democracies, the existence of lobbying groups influencing immigration policies provides an additional explanation of the Chiswick–Hatton paradox, see e.g. Freeman (1992), Joppke (1998) and Facchini et al. (2007). Further, it can also be argued that the government takes into account both social welfare and contributions offered by domestic lobbies when deciding over immigration policy (Facchini and Willmann, 2005). The remainder of the paper is organized as follows. The next section presents a theoretical framework for our analysis. This is followed by a formulation of the econometric model and a discussion of the estimation method in Section 3. Section 4 describes the data used and empirical results are analyzed in Section 5. Simulations of a fall in voting costs are described in Section 6 and the paper ends with some concluding remarks in Section 7.

In this paper, we analyze the determination of immigration policy in a direct democracy by formulating and estimating a model of participation and voting behavior. The model is used to uncover determinants of attitudes towards immigration, on the one hand, and to analyze the link between attitudes and political outcome, on the other. We account for instrumental and expressive motivations for voting by introducing political identity into the model. The direct democracy in Switzerland provides an institutional context that avoids the hypothetical bias which hampers the analysis of opinion polls. According to our estimations, human capital is an important determinant of attitudes towards immigration. This result is confirmed by the fact that education is only significant in the subsample of individuals in the labor force, which tends to exclude the possibility that our education variable captures other influences such as openness and tolerance towards other cultures. Our results show that Chiswick and Hatton's (2003) observation (as to the discrepancy between immigration policies and popular demands for tighter immigration control) also applies in the direct democracy case where the link between individual attitudes and political outcome is much more direct than in a representative democracy. In the particular vote that we analyze in this paper, there seems to have been a weak mobilization of citizens in favor of immigration restrictions and a relatively strong mobilization of individuals opposed to such restrictions. Our model, and the available data, do not allow to identify the factors that were responsible for this participation bias since it is mainly explained by the positive correlation between the error terms in the voting and participation cost equations. Simulation experiments generalizing postal voting (and introducing e-voting) show that the resulting increase in participation would lead to a slight increase in the percentage of yes-votes without changing the outcome of the vote. Before broader policy conclusions can be drawn from our results, future work should address the question whether the participation bias occurs in other votes on migration issues in Switzerland. One should also be cautious before generalizing our findings to other countries since Switzerland differs from other European countries in several respects.19 First, most immigrants to Switzerland have European origins (there was a large immigration flow from ex-Yougoslavia during the 1990s preceding our vote) and therefore cultural and religious differences were less of an issue than in other European destination countries. Second, the Swiss naturalization law was (and is) among the most restrictive in Europe (an immigrant must have lived for at least 12 years in Switzerland and be well integrated into Swiss society) resulting in comparatively low naturalization rates in the 1990s. Swiss voters might therefore be less concerned about immigrants becoming future citizens than voters in other countries where individual preferences about immigration should take into account the consequences of current decisions on the composition of the future electorate (Ortega, 2005). Third, the direct democratic process analyzed in our paper differs from the political process in representative democracies. It would be interesting to address the question whether the link between individual attitudes and the political outcome differs in the two political systems.