Over the last several decades many countries have devolved power to local governments (Grossman and Lewis 2014). Supporters of decentralisation point out that bringing governments closer to their constituencies leads to better public service delivery, allows policies to closer reflect local preferences, and facilitates monitoring (Tiebout 1956, Besley and Case 1995). Critics of decentralisation focus on the increased cost of government proliferation and vulnerability to capture by special interests (Boffa et al. 2016). The literature has mostly examined the implications of decentralisation for government financing, efficiency, and public services delivery. But surely the administrative and political upheaval caused by decentralisation must have implications for private business? These implications are likely to be greater in the context of developing countries, where administrative capacity is limited and thus the decentralisation process may not always be introduced in a smooth and efficient manner, as is the case in Indonesia (Nasution 2016).
Our study (Javorcik and Poelhekke 2024) takes a micro perspective and uses establishment-level data to examine the implications of decentralisation for the private sector. This contrasts sharply with the existing literature (Dahis and Szerman 2022, Cassidy and Velayudhan 2024), which assesses the economic implications of decentralisation using nighttime luminosity or region-level data. We study plant-level data from the Survei Manufaktur, the Indonesian Census of Manufacturing, covering all registered manufacturing plants with more than 20 employees during the period 1990-2009. Our main variable of interest is the investment rate, defined as the ratio of total investment relative to the value of fixed assets – both reported directly in the census. We also consider a plethora of other plant-level outcomes, including employment, taxes paid, the capital-labour ratio, and foreign direct investment.
Proliferation of districts in Indonesia
We consider a clean and plausibly exogenous policy shock, namely, the rapid process of government decentralisation in Indonesia that took place after the sudden fall of President Suharto in 1998 and the subsequent lifting of the presidential veto over district splits. As a result, the number of Indonesian districts increased from 284 to 511 between 1989 and 2014 (see Figure 1). The decentralisation process was associated with increased costs of running new administrations and a decline in central government transfers, leading to the proliferation of local taxes, the deterioration of the business climate, and the creation of uncertainty about future policies and the quality of governance in the newly created districts. However, it did not have a direct effect on economic output, thus offering a good setting to study implications of decentralization in isolation from other factors. Moreover, the staggered nature of the changes created an ideal laboratory for our analysis (as in Burgess et al. 2012) that allows us to trace the impact of decentralisation over time.
Figure 1 New districts created by year and political timeline
Decline in private investment
Under irreversibility of capital projects, investment is expected to react negatively to higher actual or expected taxation, deterioration in the business climate, as well as uncertainty. Irreversibility in combination with uncertainty leads to a positive option value of delaying investment until more information arrives. This mechanism relies on plants facing high costs of adjusting and reversing investment, resulting in periods with no investment followed by positive bursts.
An event study analysis shows that plants located in splitting districts do indeed invest less (relative to the size of their capital stock) after their districts split relative to plants operating in non-splitting districts (see Figure 2). The observed decline is substantial, with investment rates declining from a pre-shock mean of 35% by at least 8 percentage points for a duration of up to seven years. The analysis takes into account unobserved plant-level heterogeneity, sector-year heterogeneity, and democratisation at the local level. We find no evidence of differential pre-trends in splitting and non-splitting districts.
Figure 2 District splits event study graph: Investment rate (I/K)
State-owned enterprises are less affected
Several extensions boost our confidence that these findings capture the response of investment to decentralisation rather than to other factors. We posit that by the virtue of their government ownership, state-owned enterprises (SOEs) are not subject to the same kind of uncertainty that afflicts private establishments and that government transfers may compensate for proliferation of taxes. We test this hypothesis by allowing for a differential investment response of private and state-owned establishments in the post-split years. While in the post-split period, private establishments reduce their investment rate on average by 6-9 percentage points, we find that SOEs increase their investment rate by about 10-14 percentage points. As one would expect, this effect is driven primarily by SOEs owned by the central government.
SOEs belonging to regional governments reduce their investments as much as private establishments do. This is intuitive, given the decline in transfers regional governments obtained from the central government and the general decline in public investment in the splitting districts we document in our paper. We also find evidence of growing dispersion in plant performance in terms of total factor productivity. This pattern is present among private plants but not among SOEs, which is consistent with the latter group being sheltered from deterioration in the business climate, political uncertainty, and having the option of falling back on government support.
Increased tax burden, donations, and lower capital intensity
We show that establishments in the splitting districts see an increase in the tax burden. The ratio of indirect taxes paid to value added increases after a split for private establishments but not for SOEs. This also results in a drop in foreign direct investment. Further, we document an increase in ‘donations’ among establishment operating in the splitting districts, which is consistent with political donations or bribes being used to lobby or hedge against adverse actions of regional governments.
As the data show that output is not affected by district splits, we examine whether reductions in capital investment are accompanied by a shift towards more labour-intensive production methods. Given the likely high adjustment costs of fixed assets and irreversibility of physical investment, it may be less costly and less risky to adjust the size of the labour force.
As shown in Figure 3, we observe an increase in employment and a decline in the capital-labour ratio. This is suggestive of businesses dealing with increased taxation, deterioration in the business climate, and increased uncertainty by substituting labour for capital in order to avoid expenditures that are hard to reverse. The decline in the capital-labour ratio seems to persist over the full time horizon considered in the study. This is not surprising, given that the negative impact on investment is observed for seven years.
Figure 3 District splits event study graphs: Employment (top) and K/L-ratio (bottom)
Why did the unintended consequences take place?
The decentralisation process in Indonesia bestowed new powers on local governments, including giving them responsibility in the area of transport, agriculture, manufacturing industry, and trade. But as concluded by an Asian Development Bank study (Nasution 2016), “the decentralisation program was ill prepared” because “the capacity of subnational governments to produce public and private goods, increase productivity and employment, and promote economic growth in their jurisdictions, was not increased”. This capacity was lacking at the lower level due to the long tradition of centralisation and the fact that before the reform local governments mainly served as implementing agencies of national policies and programmes.
Conclusion
In sum, although decentralisation can potentially bring many benefits, we show that the decentralisation process itself is associated with economic costs by creating uncertainty, leading to a higher tax burden and a (at least temporary) deterioration in the business climate. All of the factors have a detrimental effect on investment, and thus a slowdown in structural transformation may be an unintended consequence of decentralisation.
References
Besley, T and A Case (1995), “Does electoral accountability affect economic policy choices?”, Quarterly Journal of Economics 110 (3): 769–798.
Boffa, F, A Piolatto, and G A Ponzetto (2016), “Political centralization and government accountability”, Quarterly Journal of Economics 131(2): 381–422.
Burgess, R, M Hansen, B A Olken, P Potapov, and S Sieber (2012), “The Political Economy of Deforestation in the Tropics”, The Quarterly Journal of Economics 127(4): 1707–1754.
Cassidy, T and T Velayudhan (2024), “Government Fragmentation and Economic Growth”, Review of Economics and Statistics.
Dahis, R and C Szerman (2022), “Development via Administrative Redistricting”, mimeo, PUC-Rio.
Grossman, G and J Lewis (2014), “Administrative unit proliferation”, American Political Science Review 108(01): 196–217.
Javorcik, B and S Poelhekke (2024), “Adjusting to Decentralization: Investment Dynamics and Beyond”, CEPR Discussion Paper No. 18081.
Nasution, A (2016), “Government centralization program in Indonesia”, Asian Development Bank Institute Working Paper 601.
Tiebout, C M (1956), “A pure theory of local expenditures”, American Political Science Review 64(5): 416–424.
