Decentralization and Political Institutions

Decentralization and Political Institutions Ruben Enikolopov (CEFIR and Harvard) and Ekaterina Zhuravskaya (CEFIR and CEPR) 7th Annual Global Development Conference, GDN St. Petersburg, January 19-21, 2006 1 Introduction  Does fiscal decentralization result in more efficient governance, higher economic growth, and better public goods?  We find strong empirical support to the Riker’s (1964) theory that the answer depends on country’s political institutions; particularly, on the strength of national political parties  Classic literature stresses both the costs and the benefits of decentralization  Benefits of decentralization come from  inter-jurisdictional competition (Tiebout 1956)  better information about voter preferences (Hayek 1948)  and higher preference homogeneity at the local level (Oates 1972)  Decentralization also has an important cost:  decentralized countries are characterized by lower internalization of externalities of local policies on neighboring jurisdictions (Musgrave 1969; Oates 1972)  Accountable local governments erect barriers to trade between regions in a single country as in China and Russia  or pursue unsound borrowing policies as in Argentina Introduction (cont.)  All benefits of decentralization are based on the premise that local politicians have political incentives to respond to the needs of their local constituency  The cost of decentralization also depends on this premise, because, as a consequence of accountability to local voters, politicians disregard preferences of population in other parts of the country  a trade-off b/w weights placed on national and local preferences in incentives of local politicians:  To minimize inter-jurisdictional externalities in a decentralized state, local politicians should place some weight on voter preferences in other jurisdictions of the country  To realize the benefits of decentralization, local politicians should have sufficiently high weight placed on preferences of their own local constituency  Since political incentives are shaped by political institutions, a decentralized country needs political institutions that strike a balance between interests of local and national population William Riker (1964)  Riker was the first to argue that political centralization (which aligns local politicians incentives with overall national objectives) may help outcomes of fiscal decentralization by making local politicians internalize interjurisdictional externalities to a larger extent  Riker considers two possible mechanisms of achieving political centralization:  Strong national political parties and Appointing local public officials  He argues that only one of them is effective  Our paper empirically tests for both these mechanisms the 1st Riker’s hypothesis  Riker argued that strong national political parties provide local politicians with efficient incentives because of 1. the need for political support at the local elections  which is more valuable for local politicians when parties are strong 2. career concerns  strong national parties have leverage over promotions of local politicians to the national level  Presence of strong national parties mitigates the interjurisdictional externalities because local politicians when choosing local policies want to please stronger national party to a larger extent  since it is able to offer better career prospects and higher support compared to a weak party  In presence of local elections, local officials are forced to be accountable to local constituency in order to be reelected when parties are strong just as when they are weak the 2nd Riker’s hypothesis  An alternative way to achieve political centralization is to let central authorities appoint local governments rather than have them being elected  In this case, central-level politicians should reappoint only “well-behaved” local officials who do not impose externalities  Riker’s view was that getting rid of local elections is inefficient because it undermines the benefits of federalism in the first place (unlike the strong national parties)  Appointed officials may stop caring for preferences of local population even though they know them better than the central politicians What this paper does…  We test the two Riker’s hypotheses:  how strong national parties and appointment vs. elections of local officials influence the effects of fiscal decentralization  Use x-section and panel data for 73 developing and transition countries and two case studies  Empirical evidence links outcomes of decentralization to strength of national political parties and presence of local elections  Case studies shed light on the channels of influence Case study: Argentina vs. Chile  Fiscal decentralization in the 80s and 90s:  Chile - improved provision of public health and education  Argentina - macroeconomic destabilization and a large-scale economic crisis  Chile - strong party system with national parties  strong career concerns about promotion to the national politics - a lot of upward vertical mobility, particularly from successful municipalities  Argentina - national political parties are weak; provincial parties dominate political arena at the national and provincial level  no career concerns about promotion to national politics – after serving in the central office politicians prefer to return to their provinces  while serving in the national office – politicians lobby for their provinces  Overall, in Argentina, national political parties do not serve as a mechanism for aligning incentives of local politicians with national objectives, while in Chile they do Case study: Russia vs. China (Blanchard and Shleifer 2001)  Decentralization is a major growth-promoting factor in China and an obstacle to growth in Russia  The reason is the difference in career concerns of local politicians  China - tight political control of provincial leaders by the communist party, which offers promotions to successful governors (e.g., President Hu Jintao was promoted this way)  Yeltsin’s Russia - large-scale political decentralization, national parties have plaid no role in shaping incentives of local politicians Variables of interest National party strength 1. Age of main parties  Intuition: career concerns  Control for: the age of democracy and time since independence 2. Fractionalization of governing parties  Intuition: career concerns  Control for:  majoritarian vs. proportional electoral rule  presidential vs. parliamentary regime  (Persson and Tabelini 2003)  Source: DPI, Beck et al., 2001 Elections 1. Are municipal executives appointed? 2. Are province-level executives appointed?  Source: DPI, Beck et al., 2001 Fiscal Decentralization 1. Subnational revenue share 2. Subnational expenditure share  Source: GFS, IMF Methodology  Standard methodology for growth and the quality of governance regressions  Treisman, 2000; La Porta et al., 1999; Barro and Sala-i-Martin, 1995; Sala-iMartin,1997  Add decentralization and political institutions measures and their interaction  Sample of developing and transition countries  Measures of political institutions vary too little and do not reflect career concerns in developed countries as much as in developing countries  whether the main political party is 150 years old or 200 years old may not have a direct effect on career concerns  Cross-section and panel regressions  In X-section – use geographical area of countries is an instrument for decentralization  In panels – use lags as instruments for decentralization and political institutions Measures of outcomes Quality of government (x-section)      Corruption indices Control over corruption Government effectiveness index Regulatory quality index Rule of law index LR Growth (x-section)  GDP per capita growth rate, PPP Public goods provision (panel data available)     Pupil to teacher ratio DPT Immunization Infant mortality Illiteracy rate Standard control variables (in logs): X-section  Fixed investments (WDI)  GDP per capita (WDI)  Population (WDI)  Openness (WDI)  Fertility (WDI)  Current level of democracy (Polity IV)  Democratic tradition (Polity IV)  Ethno-linguistic fractionalization (Roeder, 2001)  Protestants share (La Porta et al., 1999)  Latitude (La Porta et al., 1999)  Legal origin (La Porta et al., 1999) Panels  Country and year fixed effects  Fertility (WDI)  GDP per capita (WDI) Results – party strength Very solid support Riker’s theory:  Strong parties (low government fractionalization and high party age) make decentralization more efficient: improve quality of government, growth, public goods  In x-section, the results are very strong and consistent across outcomes  Panel results are consistent with x-section results!  the former are subject to possible omitted country characteristics that  Effects for Immunization and Pupil-to-teacher ratio - significant; and no results for Infant mortality and Illiteracy; this is to be expected since the first two measure chance in the short run Immunizatoin Neg_Inf_Mort Reg_Quality Neg_P_to_T GDP_grwth Gov_Effect Neg_Illiter influence outcomes, political institutions, and decentralization party_age* rev_dec gov_frac* rev_dec x-section panel x-section panel 0 -** +*** +** +* + +*** + +*** +*** 0 0 -*** -*** -*** -*** -*** -*** -*** -* -** -* 0 0 +** +** +*** Rule_Law Cont_Cor TI - Cor Residual partial plots: 1 HRV DOM MNG CHL ARM EST CHN CRI IDN TTO MEX AZE ARG BGR THA HUN NIC MUS ISR POL IND SVK MDA ALB SVN CZE MY S ROM PAK PER Gov. fractionalization measure BOL LVA GEO BLR KAZ BRA UKR -1 0 ZAF -2 3 RUS Party age measure Residuals of Rule of Law index -3 0 10 20 government parties fractionalization * Subnational revenues share 30 IND MNGPAK MEX BRA POL MY S EST KAZ HRV THA HUN BOL LVA AZE ISR TJK GEO CZE DOM MUS SVN UKR RUS IDN ARM NIC MDA CRI PER BLR ROM SVK BGR TTO ZAF ALB ARG CHN -1 0 0 1 2 50 100 150 Logarithm of age of the main parties * Subnational revenues share 200 Again we find support for Riker’s hypothesis that administrative is not a good way to ensure political centralization  In x-section – no results for Municipal elections; and very weak occasional negative coefficients for State elections. In panel regressions, the effect is positive  The weak minuses for state elections, thus, should be attributed to omitted variable bias which is eliminated with the use of country fixed effects Results – elect / appoint  Overall, there is no evidence that appointing municipal or state officials helps the outcomes of fiscal decentralization (if anything – it hinders) Immunizatoin Neg_Inf_Mort Reg_Quality Neg_P_to_T GDP_grwth Gov_Effect muni_elect* rev_dec x-section panel state_elect* rev_dec x-section panel 0 0 0 - 0 -* 0 0 0 - 0 -* 0 0 0 0 0 +*** +* -** +*** 0 Neg_Illiter 0 + 0 Rule_Law Cont_Cor TI - Cor  Influential observations Sensitivity analysis  China is most influential in x-section; without China results go through; we checked for other influential observations as well  Controlled for other possible driving forces for results  Transition  Global trends in decentralization and its efficiency  Additional controls  Elimination of controls used in the base line  Results are stable!  Initial GDP per capita squared, federation dummy, regional dummies, colonial dummies, etc. Conclusion   The key contribution of this paper is the finding that political institutions play an important role in determining the results of fiscal decentralization Previous empirical literature was inconclusive about the effects of fiscal decentralization We show that this is because it overlooked the effects of political institutions   We find strong empirical support for the two predictions of the Riker’s theory of federalism: 1. Strong national political parties are an effective way of providing right political incentives:  On the one hand, they help to create incentives for local  2. politicians to internalize externalities of their policies On the other hand, they do not hinder local accountability which is vital for efficient decentralization In contrast, appointing rather than electing local public officials does not help the outcomes of decentralization because it undermines local political accountability