Migrant mix and election outcomes: a district-level analysis of greater Istanbul metropolitan area^†

† An earlier version of this paper was presented at the fourth Biennial EconAnadolu Conference held in Eskişehir, Turkey, 10–12 June 2015, and benefited greatly from the comments received from the audience there.

Abstract

About 40% of the Turkish population resides in a province other than the one in which they or their fathers were born. Forty-three percent of such people live in the greater Istanbul metropolitan area (Istanbul, Kocaeli and Yalova provinces) and make up 82% of its population. Whether party preferences differ between migrants and natives, and between migrants from different parts of Turkey are studied with the help of robust regressions. It is shown that 2011 vote shares of the three major political parties at various districts of the metropolitan area in question can be explained by the proportions of immigrants from different regions living at these districts. This conclusion is not altered even after socioeconomic and demographic differences between the districts are controlled. Thus, a strong region-of-origin effect is found on the party choices of internal migrants.

Keywords:

• internal migration
• elections
• voter behavior
• regional analysis
• Turkey
• Istanbul

JEL Classification:

• D72
• R23

Acknowledgements

I am grateful to Dr Lyn Squire, Managing Editor of the Middle East Development Journal, and an anonymous reviewer of the journal for their suggestions which improved the paper substantially

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

† An earlier version of this paper was presented at the fourth Biennial EconAnadolu Conference held in Eskişehir, Turkey, 10–12 June 2015, and benefited greatly from the comments received from the audience there.

1. The figures given are for place of birth registry. In Turkey, children are registered at the location where their father was born. Thus, the migrant definition in this paper includes second-generation migrants as well, which is more appropriate, for the purpose at hand. The birthplace data are not available for 2011 anyway. For that reason, another study analyzing 2011 election results on the same subject, Başlevent (2013), also used place of birth registry in place of birthplace. However, it should be noted that the birth registry of women is transferred to the birthplaces of their husbands after they marry. Also, until 2006, people could transfer their birth registry from one province to another. Nevertheless, as partners in most marriages are from the same province, and registry transfers occurred only on rare occasions, any distortions these may have caused should be negligible. That province of registry and province of birth data can proxy each other quite well can be seen from the fact that the correlation between 81 pairings of the two variables for Istanbul province (the former variable for 2011 and the variable latter for 2014, both measured as proportions of respective total populations) is 0.82. The corresponding correlations for the provinces of Kocaeli and Yalova, are, respectively, 0.96 and 0.86. Had the two series being related been for the same year, no doubt the correlations obtained would have been even higher.

2. The literature exploring why that is the case, beyond what is summarized here, is surveyed in Akarca and Başlevent (2010).

3. Altınova district of Yalova, which also has a large rural population, is left in the sample because its northern part is urban and shows contiguity with the rest of the metropolitan area.

4. ED² is also tried but it turned out to be statistically insignificant.

5. Kocaeli and Yalova natives are lumped together because there are only three observations on the latter province.

6. The breakdown value is a measure of the proportion of contamination that an estimation method can withstand and still maintain its robustness.

7. The least trimmed squares (LTS) procedure developed by Rousseeuw (1984) is another high breakdown value method. The S-method is preferred here because, given the same breakdown value, S estimation has a higher statistical efficiency than LTS estimation.

8. To give an idea about earlier years, it will be useful to mention that in 1965, no province had more than a fourth of its birth population living elsewhere. Only for four provinces, this ratio was between one-fifth and one-fourth. Two of those were in the Black Sea region and one in the Central-East region. The latter province was one of those with heavy Alevi concentration.