A Service of
Leibniz-Informationszentrum econstor Wirtschaft Leibniz Information Centre Make Your Publications Visible. zbw for Economics
Edo, Anthony; Öztunc, Jonathan; Poutvaara, Panu
Article Immigration and Extreme Voting: Evidence from France
ifo DICE Report
Provided in Cooperation with: Ifo Institute – Leibniz Institute for Economic Research at the University of Munich
Suggested Citation: Edo, Anthony; Öztunc, Jonathan; Poutvaara, Panu (2017) : Immigration and Extreme Voting: Evidence from France, ifo DICE Report, ISSN 2511-7823, ifo Institut - Leibniz-Institut für Wirtschaftsforschung an der Universität München, München, Vol. 15, Iss. 4, pp. 28-33
This Version is available at: http://hdl.handle.net/10419/181258
Standard-Nutzungsbedingungen: Terms of use:
Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes.
Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence. www.econstor.eu FORUM FORUM
Anthony Edo, Jonathan Öztunc and PREVIOUS RESEARCH districts. Adding the share of electorate aged 45 and be ascribed to a presidential candidate). If votes for all Panu Poutvaara over boosts this figure to 85.7%. Adding population different presidential candidates are aggregated, they Previous research has already linked immigration and shares of EU15, EU accession and non-EU immigrants yield the number of valid votes. Immigration and Extreme extremist voting, as well as studied the mechanisms as additional explanatory variables does not change Since we are interested in the determinants of 1 Voting: Evidence from France behind opposition towards immigration. Otto and much, as the share accounted for is then 86%. To a large votes for far-left and far-right candidates, we identify Steinhardt (2014) find that larger immigrant shares extent, this can be expected to reflect endogenous presidential candidates that were classified as either increased support for the far-right in Hamburg, using migration responses. Migrants are more likely to far-left or far-right by the media in recent presidential data on city districts with fixed-effects. Halla et al. migrate to areas that are doing well economically, and elections. Jean Marie Le Pen, Marine Le Pen, Nicolas (2017) study the case of Austria and show that increas- where people are more likely to support European Dupont-Aignan, Philippe de Villiers and Bruno Mégret ing immigrant shares lead to higher vote shares for the Union membership, and have positive attitudes are included in the set of far-right presidential candi- INTRODUCTION far-right party. Card et al. (2012) use European Social towards immigration and globalisation more generally. dates. The set of far-left candidates consists of Jean- Survey (ESS) data to study the relative importance of In the present paper, we account for this potential iden- Luc Mélenchon, Nathalie Arthaud, Olivier Besancenot, In recent decades, immigration has become one of the labour market and cultural concerns in driving opposi- tification issue – i.e. the fact that migrants may prefer Philippe Poutou, Marie-George Buffet, Robert Georges most divisive issues in many Western countries. Oppos- tion to immigration. They conclude that compositional to settle in areas where the propensity to support far- August Hue, Pierre Juquin, André Francois Lajoinie, ing immigration has been a central pillar of the plat- amenities related to the utility that natives derive from right parties is low. Pierre Boussel, George Marchais and Arlette Laguiller. forms of extremist parties in many Western countries, their neighbourhoods, schools and workplaces are an Hainmueller and Hangartner (2013) studied dis- Finally, we aggregate the number of votes for all far- in the Leave campaign against British membership in important reason for negative attitudes towards crimination against immigrants in Switzerland, where right (far-left) presidential candidates to obtain an the European Union, and in Donald Trump’s electoral immigration. some municipalities used to decide on naturalisation aggregated number of votes for far-right (far-left) par- Anthony Edo campaign. Opposition to immigration and globalisa- When it comes to the psychological determinants of immigrants by referenda on individual applicants. ties in a presidential election. In order to calculate vote CEPII. tion were also central in Marine Le Pen’s campaign in of anti-immigration attitudes, Poutvaara and Stein- They find that the country of origin is a more important shares, we divide the aggregated votes by the total the French presidential election in 2017. The Front hardt (2015) show that bitter people who feel that they determinant of being naturalised than any other appli- number of votes cast (invalid and valid votes). National’s Marine Le Pen made it to the second round have not gotten what they deserve in life worry more cant characteristic, including and won 34% of votes. This was almost twice the 18% about immigration. Their analysis uses German language skills, education, vote share that her father Jean-Marie Le Pen won in Socio-Economic Panel (SOEP) data and holds both in and socioeconomic status. Figure 2 2002, the only previous presidential election in which cross sections and a panel approach, when changes in The applicants from ex-Yugo- evelopment of Unemplo ment ate over ime Front National made it to the second round. The Front bitterness are used to explain changes in worries about slavia and Turkey are rejected National’s platform is anti-EU, anti-immigration and immigration. The link between bitterness and worries considerably more often than anti-globalisation. about immigration holds among different skill catego- applicants with similar age, 10 Jonathan Öztunc In this article, we summarise our ongoing research ries, men and women, those living in former West and education and labour market Master’s student at with Yvonne Giesing on extreme voting in France. Given former East Germany, and young and old. Furthermore, status from northern and west- Barcelona Graduate School of Economics. the central role that France plays in the European the link cannot be explained away by labour market ern Europe. Union, together with Germany, understanding French competition, as it holds among civil servants who have politics is important in its own right. Furthermore, permanent contracts and are not affected by labour DATA 4 French politics is an ideal setting to test the role of market competition as a result. immigration and economic concerns in the rise of far- Nikolka and Poutvaara (2016) analysed voting in We investigate the determi- 2 left and far-right voting more generally. The Front the Brexit referendum in 326 local authority districts in nants of voting outcomes for National has run, and won more than 10% of votes in all England. They show that the share of the electorate the first round of the presiden- 0 French presidential elections since 1988. Far-left candi- with some tertiary education alone can explain 80% of tial elections that occurred in 1 1 5 2002 2007 2012 dates have also won over 10% of votes in all presiden- variation in the Leave vote share across local authority 1988, 1995, 2002, 2007, 2012 and ource re c ce u t re c tio l I titute for t ti tic tial elections since 1988, apart 2017. The data on voting out- Eco omic tu ie I EE ifo I titute Panu Poutvaara from in 2007. Ours is the first comes are available for around ifo Institute, Ludwig- paper that separately analyses Figure 1 36,000 French municipalities. Figure 3 Maximilians-University Munich, CESifo, CReAM the effects of immigration on They record the aggregated and IZA. voting in terms of political sup- Vote hare for ar eft an ar ight an i ates in rance number of registered voters, evelopment of the Immigration hare over ime port for the far-left and far-right. r left p rtie abstentions, cast votes, valid r ri t p rtie Importantly, our analysis con- 0 and invalid votes and the votes 10 trols for various economic and for each presidential candidate 25 demographic factors that could in each municipality. Regis- also explain extreme voting, tered voters refer to all people 20 and accounts for the fact that who are eligible to cast a vote immigrants may prefer to reside 15 at the ballot box. Registered in areas where the propensity voters are split into abstentions 4 to vote for extreme parties dif- 10 (people who refrain from vot- fers from other places. ing) and cast votes (people who 2 5 1 This article is based on the research paper fill out a ballot paper at the bal- “Immigration and Electoral Support for the Far 0 lot box). Cast votes are split into 0 Left and the Far Right” by A. Edo, Y. Giesing, J. Öztunc and P. Poutvaara, presented in the 1 1 5 2002 2007 2012 2017 invalid votes (blank and errone- 1 1 5 2002 2007 2012 OECD-CEPII conference “Immigration in OECD ous votes on the ballot paper) Countries” in Paris in December 2017, and avai- ource re i e ti l electio t 2017 m e il ble b t e re c o er me t ource re c ce u t re c tio l I titute for t ti tic lable as ifo Working Paper No. 244, 2017. D t o ote re for t e e r 1 come from t e ce tre e o e ocio politi ue ifo I titute and valid votes (votes that can Eco omic tu ie I EE ifo I titute
28 ifo DICE Report 4 / 2017 December Volume 15 ifo DICE Report 4 / 2017 December Volume 15 29 FORUM FORUM
In order to show that our results are not sensitive DESCRIPTIVE STATISTICS ON FAR-RIGHT AND far-left candidates. The increase in the vote share for for omitted variables that could affect immigration and to the geographical unit of analysis used in the empiri- FAR-LEFT VOTING these candidates was particularly concentrated in the political outcomes within an area. In particular, we cal section, we aggregate the data on votes at three north western and eastern French region (ranging include the share of unemployment in the population different regional levels: the canton, department and Figure 1 shows how the first-round vote share of far- between about three and five percentage points). and control for the age, education and employment region level. While there are around 2,000 cantons, right and far-left candidates has changed between 1988 structure of each area. there are 96 departments and 22 regions. Using larger and 2017. Both vote shares have increased dramatically EMPIRICAL METHOD: SPATIAL CORRELATION However, it is still possible that immigrants settle geographical areas allows us to show that our results since 2007, following the financial crisis, the Eurozone APPROACH in places with a small share of votes for anti-immigra- are not contaminated by the fact that French citizens crisis and, most recently, the refugee crisis. Figures 2 tion parties. In order to tackle this endogeneity issue, may respond to the arrival of immigrants in a given area and 3 show how unemployment and the population Our empirical strategy exploits the fact that immi- we use an instrumental variable (IV) strategy. The idea by moving away. share of immigrants have changed between 1988 and grants tend to cluster in a limited number of geograph- is to use variations in immigration that are due to We use the French censuses from 1990, 2007 and 2012 (2017 data is not yet available). Both unemploy- ical areas. We use the spatial distribution of immigrants another variable whose changes are plausibly exoge- 2012 to infer the number of immigrants for the presi- ment and the population share of immigrants have in order to estimate their impact on far-right and far- nous (unrelated) to the outcome. In this respect, litera- dential elections of 1988, 2007 and 2012. No census was increased steadily since 2002. left-voting. The idea of this spatial correlation approach ture on the subject generally uses past migrant net- implemented for the years 1995 and 2002. Instead, we Figure 4 displays the regional distribution of the is to compare the changes in votes for far-right and far- works (past settlements) as predictors for future use the pooled 1994-1995 labour force survey (LFS) and first-round vote share for far-right and far-left candi- left voting of high-immigration places with those of migration flows. In our study, we use the historical dis- 2001-2002 LFS to ensure a high level of precision in esti- dates in 1988. Far-right candidates were initially very low-immigration places. tribution of immigrants across French departments mating our variables for these two election years. For strong in the southeast of France, while far-left candi- This approach is subject to the main limitation that from the 1968 French census as a predictor for their the year 2017, we use the most recent wave of available dates were popular in the north, centre and south of immigrants are not randomly distributed across areas. subsequent flows. Our instrument is based on the idea data, which is the 2015 LFS to infer the number of immi- France. They may prefer to settle in areas experiencing positive that the stock of previous immigrants has an impact on grants for each French region. We define an immigrant Figure 5 respectively displays the corresponding economic shocks and where the share of far-right vot- subsequent flows through network effects, while as a person born abroad without the French citizen- change in the vote share from 1988 until 2012 for far- ers is relatively low. This behaviour among migrants assuming that past immigrant concentrations are ship. This definition allows us to exclude the migrants right and far-left parties across departments. In con- will create a spurious negative correlation between uncorrelated with current unobserved economic with French nationality who can vote and avoid any trast to the initial vote share, the increase in the vote immigration and far-right voting, contaminating the shocks (for details, see Edo et al. (2017)). composition effect due to their inclusion in the share for far-right candidates was concentrated in measured effects of immigration on political outcomes. The use of the 1968 census allows us to predict sub- sample. north eastern departments, departments in the center In order to limit this potential bias, we estimate the sequent inflows based on immigration patterns that We can therefore investigate the impact of immi- and to some extent in the southwest of France, as well impact of changes in local immigrant shares on changes took place at least 20 years earlier. Moreover, the Front gration on far-right and far-left voting at the canton as Corsica. In these departments, the vote share for far- in vote shares. This estimation in first differences National, which is the first post-1945 extreme right level between 2002 and 2012, at the department level right candidates increased by between 7 and 16 per- accounts for all time-invariant differences between party, was created in 1972 and participated in the pres- between 1988 and 2012 and at the regional level centage points. The right-hand side of Figure 5 presents areas that may affect immigrant inflows and votes. We idential election for the first time in 1988. As a result, between 1988 and 2017. the change in the vote share from 1988 until 2012 for also introduce a large set of control variables to account the spatial distribution of immigrants in 1968 was not
Figure 4 Figure 5 Initial Vote Share for Far-Right and Far-Left Parties in France in 1988 Increase in Vote Share for Far-Right and Far-Left Parties in France in 1988
r Ri t rtie Far-Left Parties r Ri t rtie Far-Left Parties
15 25 to 25 perc poi t 1 02 to 21 5 perc poi t 10 05 to 15 57 perc poi t 5 to 4 1 perc poi t 1 5 to 15 25 perc poi t 10 47 to 1 02 perc poi t 7 17 to 10 05 perc poi t 2 1 to 5 perc poi t 10 4 to 1 5 perc poi t 1 to 10 47 perc poi t 4 to 7 17 perc poi t 0 4 to 2 1 perc poi t 5 to 10 4 perc poi t 4 15 to 1 perc poi t 12 to 4 perc poi t 5 5 to 0 4 perc poi t
ote e e tm p re m e il ble from ttp // t ou fr/fr/ t et /co tour e ep rteme t fr c i i u ope treetm p/ come from ote e e tm p re m e il ble from ttp // t ou fr/fr/ t et /co tour e ep rteme t fr c i i u ope treetm p/ come from t e co tributor of pe treet p e u e t e 2017 er io e t i il ble u er t e pe D t b e ice e t e c rto r p i lice e CC t e co tributor of pe treet p e u e t e 2017 er io e t i il ble u er t e pe D t b e ice e t e c rto r p i lice e CC e cop ri t of t e m p lie it pe tree p ttp // ope treetm p or /cop ri t/e Co tributor of pe treet p u er Db lice ce e cop ri t of t e m p lie it pe tree p ttp // ope treetm p or /cop ri t/e Co tributor of pe treet p u er Db lice ce ource ut or illu tr tio u i t from t e pe l tform of re c ublic D t 2017 ifo I titute ource ut or illu tr tio u i t from t e pe l tform of re c ublic D t 2017 ifo I titute
30 ifo DICE Report 4 / 2017 December Volume 15 ifo DICE Report 4 / 2017 December Volume 15 31 FORUM FORUM
caused by extreme right voting Figure 6 increases the share of votes for far-right parties by 2.02 Understanding the grievances of those voting for and is very likely to be unre- percentage point at the department level. extreme parties is also important to those who do not lated to the share of votes for elationship et een ar ight Voting an Immigration agree with them. Ignoring these grievances puts gains far-right parties in subsequent from globalisation and migration at jeopardy at the bal- C e i oter for f r ri t p rtie i THE IMPACT OF IMMIGRATION ON FAR-RIGHT years. 0 1 VOTING ACROSS EDUCATION-NATIONALITY lot box. In the French context, it is notable that the total GROUPS first-round vote share for far left and far right candi- 0 12 THE AVERAGE IMPACT OF dates increased from 34.3% in 2012 to 47.3% in 2017. If IMMIGRATION ON VOTES 0 0 In Edo et al. (2017) we go beyond the average impact of labour market reforms do not succeed in boosting eco- FOR FAR-RIGHT AND FAR- immigration on far-right voting by decomposing its nomic growth and reducing unemployment, an extrem- LEFT PARTIES 0 04 effect across education-nationality groups. We break ist candidate is likely to win in the next French presiden- 0 00 down the immigrant population into six education-na- tial election. Figure 6 provides a preliminary tionality groups and use three education groups: low, look at the correlation between 0 04 medium and high education groups. The low education REFERENCES the change in the vote share for group is composed of people who have an elementary 0 0 Card, D., C. Dustmann and I. Preston (2017), “Immigration, Wages, and far-right parties within a given school diploma or no diploma, the medium education Compositional Amenities,” Journal of the European Economic Associa- department and the change in 0 12 group is composed of people who have a high school tion 10 (2012), 78–119. the share of immigrants for 0 10 0 05 0 00 0 05 0 10 degree and a French diploma giving access to high Edo, A., Y. Giesing, J. Öztunc, and P. Poutvaara (2017), “Immigration and C e i immi r t re i Electoral Support for the Far Left and the Far Right”, ifo Working Paper that department.2 The figure school, the high education group is composed of peo- ote E tim te coefficie t 0 2 tu e t t te t 4 00 No. 244. indicates a positive and signifi- ource re i e ti l electio t re c ce u e l bour force ur e 1 2012 ple who have a college degree, some college or a French Hainmueller, J. and D. Hangartner (2013), “Who Gets a Swiss Passport? cant relationship: the esti- ut or c lcul tio ifo I titute diploma giving access to the university. For each educa- A Natural Experiment in Immigrant Discrimination,” American Political mated coefficient (and T-statis- tion group, we compute the change in the share of Science Review 107, 159-187. Halla, M., A. F. Wagner and J. Zweimüller (2017), “Immigration and Vot- tic) is -0.32 (4.00). This non-European and European immigrants. ing for the Far Right,” Journal of the European Economic Association 15, preliminary result suggests that the votes for far-right the 1988 and 1995 presidential elections when using Our break-down firstly shows that the average 1341–1385. parties grew fastest in the departments that experi- the department as an alternative unit of analysis. positive effect of immigration on extreme right voting is Nikolka, T. and P. Poutvaara (2016), “Brexit - Theory and Empirics,” enced the highest increase in immigration. The econo- Regressions across French regions even allow us to asymmetric across education groups. This effect is fully CESifo Forum 17(4), 68–75. metric results reported in Table 1 shows the robustness account for the first-round results during the 2017 pres- driven by the share of poorly educated immigrants. Otto, A. H. and M. F. Steinhardt (2014), “Immigration and Election Out- comes: Evidence from City Districts in Hamburg,” Regional Science and of this correlation. idential election. This result is consistent with the fact that low-skilled Urban Economics 45, 67–79. Table 1 reports the OLS and IV estimated effects of The results from Table 1 indicate that immigration immigration may have detrimental labour market Poutvaara, P. and M. Steinhardt (2015), “Bitterness in Life and Attitudes immigration on the change in votes for far-right and far- has a positive impact on votes for far-right parties and effects, as compared to highly-skilled immigration. Towards Immigration”, CESifo Working Paper Series 5611, CESifo Group Munich. left parties across French areas. We use past immigrant a modest negative impact on those for far-left parties. Secondly, the positive impact of poorly educated immi- settlement patterns as an instrument. In addition to The fact that the OLS estimates are weaker than the IV grants on extreme right voting is only driven by those using the 96 French departments as our baseline geo- estimates is consistent with the fact that immigrants migrants who have a non-European nationality. By con- graphical unit of analysis, we also use cantons (1,989) are more likely to migrate to regions where the vote trast, medium and high educated non-European immi- and regions (22) as alternative units. Cantons are share for far-right parties is low; or to regions with thriv- gration have insignificant or negative effects on sup- smaller than departments, while regions are larger. The ing economies that may be less inclined to support far- port for far-right parties. Taken together, our results regressions at the Canton level are performed over the right parties. In particular, our IV estimates implies that suggest that the educational composition of immi- 2002-2012 period. We extend this period of analysis to a one percentage point increase in the immigrant share grants, as well as their origin, matter in determining their impact on votes for far-right parties. 2 More specifically, the points in the scatter diagram are the residuals from a regression of the change in votes for far-right parties and the change in immigrant share on a set of year fixed effects. The year fixed effects remove CONCLUSION any year-specific effects that are common to all geographical areas. We estimated the impact of immigration on voting for Table 1 far-left and far-right parties in France, using panel data Impact of Immigration on Extreme Voting Across Alternative Geographical Units on presidential elections from 1988 to 2012 (and in
Geographical unit of analysis some analyses until 2017). We found that immigration Canton Department Region Region increases support for far-right candidates, in all ana- OLS IV OLS IV OLS IV OLS IV lysed geographical units of observations. This result was especially strong when using instrumented immi- Far-right 0.42*** 2.55*** 0.38*** 2.02*** 0.97** 2.52** 0.99** 2.29** gration flows, but it was also present in ordinary least (8.61) (3.39) (3.57) (3.46) (2.52) (2.26) (2.63) (2.17) squared regressions. There is no robust pattern on far- left voting. Far-left 0.02 -0.27 -0.10*** -0.16 -0.38* -0.64*** -0.32 -0.70*** Our additional analyses suggest that an increase in the electoral support for the far-right is driven primarily (0.71) (-0.50) (-2.82) (-0.81) (-1.90) (-2.75) (-1.60) (-2.88) by low-skilled immigrants from non-Western countries. Time Period 2002–2012 1988–2012 1988–2012 1988-2017 Furthermore, our results indicate that both economic concerns (related to the educational level of migrants) Observations 3,895 384 88 110 and cultural concerns (related to migrants’ region of Note: ***, **, * mean different from 0 at the 1%, 5%, 10% significance level. T-statistics are indicated in parentheses below the point estimate. origin, most saliently whether they come from Western Source: Presidential election data, French censuses and labour force survey (LFS) (1988–2012); authors‘ calculations. or non-Western countries) play an important role.
32 ifo DICE Report 4 / 2017 December Volume 15 ifo DICE Report 4 / 2017 December Volume 15 33