Revisiting the link between growth and federalism: A Bayesian model averaging approach

Abstract

Die Dis cus si on Pape rs die nen einer mög lichst schnel len Ver brei tung von neue ren For schungs arbei ten des ZEW. Die Bei trä ge lie gen in allei ni ger Ver ant wor tung der Auto ren und stel len nicht not wen di ger wei se die Mei nung des ZEW dar.Dis cus si on Papers are inten ded to make results of ZEW research prompt ly avai la ble to other eco no mists in order to encou ra ge dis cus si on and sug gesti ons for revi si ons. The aut hors are sole ly respon si ble for the con tents which do not neces sa ri ly repre sent the opi ni on of the ZEW. Non-technical SummaryIn the classic case fiscal federalism leads to economic growth by a more productive and possibly smaller public sector through higher preference homogeneity and enhanced intergovernmental competition. On the other hand, decentralized political decisions create inter-jurisdictional spillovers which may negatively affect growth by distorting local tax and fiscal incentives. In recent years, this discussion has led to a growing body of literature aimed at understanding the empirical link between the two phenomena. The results on the existence, direction or sign of such a link, however, are as ambiguous as ever.The present paper takes a fresh look at the empirical relationship between fiscal federalism and output growth using a Bayesian model averaging approach. This approach is increasingly becoming a standard econometric tool in the empirical growth literature, because it offers a coherent procedure to deal with both model and parameter uncertainty in a context of weak theoretical guidance which has previously led researchers to choose their empirical specifications (explanatory variables and functional forms) on an arbitrary basis, or at best using some unknown rule-of-thumb rules. Surprisingly, however, the earlier literature on the link between growth and federalism -by definition being a part of this much larger literature on the determinants of economic growth and, thus, sharing the same methodological limitations -has never adopted these advancements.Obviously, endogeneity and causality are the main concerns of these studies, but the methodological superiority of the present analysis allows for claiming that the results of previous research might have been additionally biased due to over restrictive model specification. In contrast, the results from a sample of 23 OECD countries over indicate that after controlling for unobserved country heterogeneity, there is no robust link, neither positive, nor negative, between output growth and fiscal federalism. Das Wichtigste in Kürze

Revisiting the growth effects of fiscal policy: A Bayesian model averaging approach

“…With regard to the control variables, most standard growth regressors that the literature identifies as robust in global and cross country settings are not robust in the panel of OECD countries. This is in line with the findings of Eicher et al (2007), who look at growth determinants in a cross country setting for the OECD, and Asatryan and Feld (2015), who examine growth determinants in OECD countries in a panel data setup. Both find that in the OECD sample only a few variables are robust determinants of economic growth -but none of these two references examines the effects of fiscal policy.…”

Partial fiscal decentralization and sub-national government fiscal discipline: empirical evidence from OECD countries

“…Given the arguments in favor of the former measure of revenue autonomy, which are discussed throughout this paper and supported by what can be called a general consensus of the relevant literature (e.g., Ebel and Yilmaz 2002;Stegarescu 2005;Baskaran 2010;Asatryan and Feld 2014), it can be argued that the analyses using more disaggregated measures of revenue autonomy can indeed be more reliable.…”

“…Although these measures are available for a relatively large sample from the IMF's Goverment Finance Statistics (GFS), they are criticized widely for not distinguishing between SNGs' real functions and those imposed and regulated by the center (eg., Ebel and Yilmaz 2002;Asatryan and Feld 2014).…”

Nonlinear predictive model selection and model averaging using information criteria

“…Model averaging is a widely applied method to deal with model uncertainty and reduce or eliminate the risk of using only a single model. Model averaging approaches such as AIC-and BIC-based averaging methods have been used in many applications (Asatryan & Feld, 2015;Cade, 2015;Kontis et al, 2017;Moral-Benito, 2015). The model averaging approach with AIC involves the computation of the delta AIC and the Akaike weights.…”