Is Central Government Intervention Bad for Local Outcomes? Mixed Messages from Indonesia

Abstract

Central government line agencies in Indonesia spend a significant amount of their budgets directly in the regions, much of it on functions that have already been decentralised to local governments. Such deconcentrated spending contravenes both international best practices and Indonesian decentralisation legislation. Empirical evidence on the question of actual impact of such spending in Indonesia is mixed. The share of central deconcentrated spending that is co-administered directly with local governments appears to have beneficial service delivery effects; but the portion organised through provinces – without significant input from districts – has a negative impact on service outcomes.

Acknowledgements

The author would like to thank staff at World Bank, Jakarta, Indonesia, for providing access to the data used in this article. The data can be downloaded from http://data.worldbank.org/data-catalog/indonesia-database-for-policy-and-economic-research.

Notes

1. Before decentralisation began, the Indonesian public sector operated through a parallel system of second- and third-tier central government deconcentrated agencies (Kantor Wilayah and Kantor Departemen respectively) and nominally autonomous provinces and districts. Deconcentrated agencies were abolished in 2001 and provinces and districts became fully autonomous. There are currently 34 provinces and 505 districts, including 412 rural districts (kabupaten) and 93 municipalities (kota).

2. TP may also be co-administered by villages, although the latter are not an official level of government.

3. Recent case studies have estimated that just over 95 per cent of Dekon and TP in the education sector have been used for functions that have already been decentralised to kabupaten/kota (Decentralization Support Facility, 2008).

4. Protected water, as defined by the Central Bureau of Statistics comprises: piped water; communal tap water; rainwater; spring water; and covered wells. These protected sources should be less than 10 metres from household and at least 10 metres from collected waste. Protected sanitation includes both individual and communal toilets.

5. These indicators have been selected based on the availability of data. Data problems appear to be significant for two other possible Susenas measures: average years of education and morbidity. As such, these indicators are not used. No other service delivery variables are readily available for all local governments over the time from of this study.

6. From a needs perspective, the influence of population on per capita Dekon or TP allocations is uncertain a priori. Population should have a positive effect on total allocations of course; but if local scale economies are presumed to exist, population’s impact on per capita allocations may well be negative.

7. Given the broad nature of this initial analysis interest focuses on the direction of influence of the various determinants and not the specific estimated quantitative impact.

8. The positive impact of transfers on own-source revenues in Indonesia contradicts most thinking on the issue and some empirical evidence from elsewhere in the world. The hypothesis is that rising intergovernmental transfers stimulate increases in local government own-source revenue, which in turn are used to fund increases in civil servant allowances, allowing the latter to keep pace with the increasing size of government. See Lewis and Smoke (in press) for some empirical evidence and a detailed explanation.

9. See Greene (2008a) for a description and discussion of stochastic frontier analysis (SFA) and links to an extensive literature.

10. For another stochastic frontier model using government spending as a factor of production (in the health sector), see Greene (2008b).

11. It might reasonably be asked why electricity access was not used as an explanatory variable or an instrument in the previous regressions explaining service access. Attempts were made to include the variable in both these ways. Electricity access effects on service delivery were insignificant and the variable’s use as an instrument proved problematic for several of the endogeneity related specification tests.