Government Constraints and Accountability: Economic Voting in Greece Before and During the IMF Intervention

Abstract

Incumbent parties in Southern Europe experienced losses in their electoral support that came along with a series of economic reforms imposed by the EU and the IMF. However, recent theories of accountability would predict lower levels of economic voting given the limited room left for national governments to manoeuvre the economy. To resolve this puzzle, the paper presents and models quarterly vote intention time series data from Greece (2000–2012) and links it with the state of the economy. The empirical results show that after the bailout loan agreement Greek voters significantly shifted their assignment of responsibility for (economic) policy outcomes from the EU to the national government, which in turn heightened the impact of objective economic conditions on governing party support. The findings have implications for theories linking international structures, government constraints and democratic accountability.

Acknowledgements

Part of this work was completed in the Hellenic Observatory of the London School of Economics where the author held the ‘National Bank of Greece Fellowship in Contemporary Greek Studies’ research fellowship in 2012. An earlier version of this work appeared in the GreeSE discussion papers of the Hellenic Observatory. I would like to thank Kevin Featherstone, Vassilis Monastiriotis, Penny Apostolopoulou, Athanasia Chalari, Georgios Xezonakis, Guido Tiemann, Sara Hobolt and participants of the panels in the 2012 Midwest Political Science Association (Chicago) and the 2013 European Political Science Association (Barcelona). All errors in this manuscript remain my responsibility.

Notes

1. Given that the data is quarterly, some reservations should be held about the role of these events on

public opinion. A rigorous analysis of this time series is not provided in this paper.

2. The prominent polling company Metron Analysis, which is a member of the Hellenic and European Market Research Association and of the Worldwide Association of Public Opinion Research, gathered the survey data. Metron Analysis does not release any individual-level survey data and thus only macro expectations can be tested.

3. The second time in four years that a Greek government admitted that the Greek statistics are trustworthy.

4. The reasoning behind these protests were (1) the austerity measures and its severe consequences and (2) the popular belief that national sovereignty was on hold due to the two rescue packages and the IMF impositions. In reality, the actual room to manoeuvre the economy was indeed limited because of the mutual agreement between the Greek government and its international lenders (IMF/ECB/EU).

5. However, note here that the elected Greek parliament voted in favour of the two bailout agreements with 172 MPs in May 2010 and 216 in February 2012.

6. Notice that an interrupted time series design would also be appropriate here. To be sure, the specification is very close to that setup and the upcoming analyses offer a similar test. I thank an anonymous reviewer for pointing this out.

7. Most of them are depicted in Figure 1. Additional events include the quarter the Greek police arrested the 17 N terrorist group, the summer of the Olympic games, the Grigoropoulos events, Vatopedi, the phone-hacking incident, the Zahopoulos events and the junkbonds scandal.

8. Additional models (Prais–Winsten/Newey–West) that take autocorrelation into account show that the main results of the study are robust. The Newey–West regression estimates, moreover, account for the fact that some of the missing data was linearly interpolated.

9. What is interesting is that the period after the election of New Democracy in 2004 the uncertainty around the estimate increased substantially. This is presumably due to the fact that the previous government had stayed in office for almost 11 consecutive years and the assignment of responsibility for economic outcomes was ambiguous.

10. However, some reservations should be held with rolling coefficients, as they are not stable across different specifications. The rolling coefficient presented in Figure 3 is from a simple specification that controls for elections and political events. However, alternative specifications more or less corroborate the main findings of the study.

11. The variable is worked out as follows: Greece/(Greece + EU).

12. Note that the coefficient for the lagged levels of the unemployment measure is positive and significant in Table 2. When the responsibility measure is purged from its economic causes, the coefficient on the lagged levels of unemployment is in line with the theoretical prediction (negative and significant Unemployment_t–1 = –1.08; p < 0.00). The full regression table corresponding to Figure 5 is available upon request.

13. The x-axis measures the innovations from the conditional variable. The 0 point on the x-axis of Figure 5 is now the equivalent of the 0.5 middle point in Figure 4.