When Can We Trust Population Thresholds in Regression Discontinuity Designs?

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. i Abstract: A recent literature has used variation just around deterministic legislative population thresholds to identify the causal effects of institutional changes. This paper reviews the use of regression discontinuity designs using such population thresholds. Our concern involves three arguments: (1) simultaneous exogenous (co-)treatment, (2) simultaneous endogenous choices and (3) manipulation and precise control over population measures. Terms of use: Documents inRevisiting the study by Egger and Koethenbuerger (2010), who analyse the relationship between council size and government spending, we present new evidence that these three concerns do matter for causal analysis. Our results suggest that empirical designs using population thresholds are only to be used with utmost care and confidence in the precise institutional setting.Acknowledgments: We would like to thank

Elected or appointed? How the nomination scheme of the city manager influences the effects of government fragmentation

“…For the RDD to measure the effect of the council size law, it is necessary that there are no other changes at the 3000-, 5000-or 10000-threshold of the assignment variable (see Lee andLemieux, 2011 andAde andFreier, 2011 for a further discussion on this topic).…”

Electoral Competition and Endogenous Political Institutions: Quasi-Experimental Evidence from Germany

“…The identifying assumptions in the Diff-in-Disc design are arguably less strict than in the RD design. Notably, we do not require that there is no co-treatment at the cutoffs (Ade and Freier (2011) show that co-treatment can be a major problem in standard RDDs with population thresholds). Instead, we only require that the effect of any co-treatments remains constant between the pre-and post-treatment periods.…”

The Effect of Direct Democracy on the Level and Structure of Local Taxes

“…Our institutional context violates this assumption. As discussed by Egger and Köthenbürger (2010) and Ade and Freier (2011), there exist a number of additional institutions that change at these thresholds, such as the size of the local council, wages of elected civil servants, status of the municipality (implying changes in expenditure and revenue duties), certain features of the horizontal fiscal equalization system, etc.…”

“…where the dependent variable y is the tax variable of interest as before; D is a dummy variable that is 1 if (normalized) municipal population size is above a given threshold; 10 T is a dummy variable that equals 1 starting from the year 1995 (the introduction of direct democracy); f (np) and g(np) are polynomials of (normalized) municipal population size Finally, an important requirement for a valid RDD is that municipalities should not have been able to selectively sort around the thresholds (Lee, 2008) by manipulating the population statistics (Ade and Freier, 2011). Population sizes are calculated by the state-level statistical office (outside of the control of municipal officials), however there might be some leeway through particular municipal (dis)incentives set for new residents to register or moving residents to unregister.…”