Public Disclosure Authorized. Public Disclosure Authorized. Public Disclosure Authorized. Public Disclosure Authorized

Transcription

1 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized

2 r 2009 Public Financial Publications, Inc. The Impact of Decentralization on Subnational Government Fiscal Slack in Indonesia BLANE D. LEWIS and ANDRE OOSTERMAN Since Indonesia began implementing its decentralization program in 2001, subnational unspent balances have grown rapidly and have reached levels that many officials find unreasonably high. But the extent to which subnational government reserves are excessive, in general, is not obvious. A not implausible decrease in the price of oil would reduce transfers to subnationals significantly and, if sustained, could possibly eliminate reserves in a relatively short time. Central government should not take any immediate action to reduce subnational slack resources directly but should instead focus on removing the underlying causes of such. INTRODUCTION A significant literature exists on subnational government slack revenues in more developed countries, especially the United States. 1 Most of this research focuses on the creation, use, and impact of budget stabilization funds and similar fiscal mechanisms. A major preoccupation of this work has been to examine the extent to which subnationals maintain sufficient reserves to stabilize expenditures during economic downturns. Issues related to the slack resources of subnational governments in developing countries have not yet been addressed in the academic literature. This is not surprising given the typical lack of accessible data on key variables. This paper focuses on the slack resources of Blane D. Lewis, Associate Professor, Lee Kuan Yew School of Public Policy, National University of Singapore, 469C Bukit Timah Road, Singapore , Singapore. Andre Oosterman, Consultant, World Bank, Jakarta Stock Exchange Building, Tower 2, 12th floor, Jl. Jenderal Sudirman, Kav , Jakarta 12180, Indonesia. 1. See Justin Marlowe, Fiscal Slack and Counter-Cyclical Expenditure Stabilization: A First Look at the Local Level, Public Budgeting & Finance 25, no. 3 (2005): and Rebecca Hendrick, The Role of Slack in Local Government Finances, Public Budgeting & Finance, 26, no : 14 46, for brief reviews and comprehensive citations. Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 27

3 subnational governments in Indonesia and would appear to be the first of its kind for a developing country. 2 Indonesian subnational governments comprise provinces and districts. Starting in 2001, these second- and third-tier governments began operating in a much more decentralized environment than had been the case before that date. Fiscal decentralization in Indonesia is characterized by a significant devolution of spending responsibility to subnational governments, especially at the district level. However, control over major sources of revenue remains highly centralized. By the end of 2007, subnationals accounted for about 38 percent of total public sector expenditure but only about 8 percent of total public revenues. As a result, transfers from the central government have been called upon to finance the bulk of expenditure decentralization to the subnational level. This is not an uncommon situation in developing countries that are decentralizing their public sectors. 3 Indonesian subnationals have accumulated substantial unspent balances since the government launched its decentralization program. 4 At their recent peak, subnational reserves reached just over Rp 110 trillion or about 3 percent of GDP. 5 The size of subnational unspent balances has caused significant concern among some central government officials, especially as the center has struggled to reduce its own fiscal deficit. 6 Policy discussions have focused on possible methods for encouraging subnationals to spend more of the resources available to them and, less feasibly given the current legal framework, on schemes for reducing transfers to subnationals if they don t increase spending. Three sets of questions about subnational reserves in Indonesia immediately present themselves. As indicated above, it is well known that it is sensible for subnationals to keep some funds in reserve in order to stabilize expenditures when economic activity slows and public revenues decline. This suggests a first question of significance: do Indonesian subnationals save too much from this point of view? A second set of interesting questions concerns the determinants of subnational savings. That is, do subnational governments save purposively or has the recent buildup of unspent revenues 2. Another reason for the relative dearth of research on subnational fiscal slack in developing countries may be that many subnationals, due to extremely limited revenues, have simply not saved much; that is, slack is nonexistent in many places. As transfers to and own-source revenues of subnationals grow, as they are in many developing countries, slack resources may become of greater concern. 3. Roy Bahl and Sally Wallace, Public Financing in Developing and Transition Countries, Public Budgeting & Finance, Silver Anniversary Edition (2005): 83 98, and Paul Smoke, Local Revenues under Fiscal Decentralization in Developing Countries: Linking Policy Reform, Governance and Capacity, in Fiscal Decentralization and Land Policies, eds. Gregory Ingram and Yu-Hung Hong (Cambridge, MA: Lincoln Institute of Land Policy Press, 2008). 4. We use the terms slack revenues, reserves, and unspent balances interchangeably in this paper. 5. Some government officials now refer to the increase in subnational reserves as the 100 trillion rupiah problem. 6. The cumulative central government budget deficit during is estimated to be just 4Rp 213 trillion. 28 Public Budgeting & Finance / Summer 2009

4 been largely inadvertent? If the latter is the case, what then explains the unintended increase in reserves? A third important question is: What should be done about subnational government savings? This is the policy question on which many central government officials are fixated. The paper proceeds as follows. First, we use available aggregate central bank data on deposits to describe the growth, size, and geographic distribution of subnational reserves across provinces from 2001 through Second, using deposit and national budget data we simulate the impact of a decrease in the price of oil on subnational fiscal slack. This analysis is carried out in order to shed some light on the degree to which subnational reserves are excessive. Third, using the same data combined with information on the main intergovernmental transfers, among others, we specify and estimate a simple econometric model to investigate some of the main potential determinants of subnational government surpluses, during the period Fourth, we examine a significant number of audited subnational government budgets for with a view to further specifying the causes of surpluses. In addition, in this section of the paper we highlight the results of some limited field work recently undertaken on relevant issues as the analysis and discussion proceed. Finally, we conclude the paper with a summary of the main points and offer some conclusions of relevance for fiscal decentralization policymaking in Indonesia. GROWTH, SIZE, AND DISTRIBUTION OF SUBNATIONAL RESERVES This section of the paper uses subnational government bank deposits to indicate reserve funds. Subnational governments are not (yet) known to invest in central government certificates of deposit (SBI), treasuries, or other similar financial instruments to store their funds. 7 So most if not all slack funds are on deposit in commercial banks. In addition, subnationals have not borrowed much since decentralization to finance the development of infrastructure. 8 Instead, they mostly use own-source revenues and transfers for public capital expenditure. As such, funds on deposit do not include borrowed amounts. Finally, subnational governments are not allowed to maintain deposits in banks located outside the country. All things considered, the use of national deposits as a proxy for aggregate subnational reserves seems reasonable under the present circumstances. The data used in this analysis come from the Indonesian Central BankFBank Indonesia (BI). BI has information on all subnational government (demand, savings, and 7. The banks in which subnationals hold their reserves do, however, invest significantly in such instruments. 8. Christine R. Martell and George M. Guess, Development of Local Government Debt Financing Markets: Application of a Market Based Framework, Public Budgeting & Finance 26, no. 1 (2006): and Blane D. Lewis, On-Lending in Indonesia: Past Performance and Future Prospects, Bulletin of Indonesian Economic Studies 43, no. 1 (2007): Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 29

5 Rupiah (Blns) 120, , ,000 90,000 80,000 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 FIGURE 1 Subnational government bank deposits, Jan-01 Apr-01 Jul-01 Oct-01 Jan-02 Apr-02 Jul-02 Oct-02 Jan-03 Apr-03 Jul-03 Oct-03 Jan-04 Apr-04 Jul-04 Oct-04 Jan-05 Apr-05 Jul-05 Oct-05 Jan-06 Apr-06 Jul-06 Oct-06 Jan-07 Apr-07 Jul-07 Oct-07 Jan-08 Current Rupiah Constant Rupiah time) deposits in commercial banks in the country. The subnational government deposit data employed in this investigation have been aggregated at the provincial level. That is, for each province in the country, deposit data comprise those for the provincial government and all district governments within the province. The individual government level data have not been made available for use in this study due to BI s concerns about customer confidentiality. 9 Figure 1 details the rise of subnational reserves since Before the start of decentralization, subnational governments held approximately Rp 7 trillion in reserve funds. Between the start of 2001 and the end of 2007, subnational reserve funds expanded by an order of magnitude, reaching just under Rp 70 trillion. Reserves grew at an annual rate of 37 percent during the indicated period, in nominal terms and at a rate of just 425 percent per year in real terms. Observe the pattern of subnational government savings during any given fiscal year. Subnationals tend to increase their deposits during the first three quarters of the year and then draw down on reserves in the last quarter, the period during which most provincial and district government spending takes place. Among other things this shows that central government anxiety about the 100 trillion rupiah problem that subnational reserves supposedly represent is perhaps needlessly exaggerated, since the figure in question was just the peak in the year on year savings cycle. That is, what matters in this regard is 9. With a view to increasing transparency and monitoring subnational fiscal activity and performance, the Ministry of Finance formally requires subnational governments to report their reserve funds (and other assets, as well as liabilities). Unfortunately, most subnationals do not provide the relevant information as required and the Ministry of Finance has yet to employ available sanctions in an attempt to force them to do so. 30 Public Budgeting & Finance / Summer 2009

6 TABLE 1 Subnational Reserves, Nominal and as Percent of Expenditure Provincial (Rp Bln) 4,471 7,762 8,507 8,103 9,851 15,761 19,517 19,146 District (Rp Bln) 2,940 10,796 13,592 13,350 14,595 25,949 49,009 48,734 Total (Rp Bln) 7,411 18,558 22,099 21,453 24,446 41,710 68,526 67,880 Provincial (% Exp) District (% Exp) Total (% Exp) Source: Authors own calculations based on Bank of Indonesia and Ministry of Finance Data. the end of year balance, which in this case was approximately Rp 70 trillion, a considerably smaller amount. Table 1 shows provincial and district reserve funds at the aggregate level over the period in question. Note that at the end of 2000, provincial reserves exceeded those of districts. Provinces held 72 percent of total reserves compared with just 28 percent for districts. During the period, district reserves grew at a pace of almost 50 percent per year, whereas the reserve funds of provinces grew at only 23 percent per annum. At the end of the period districts maintained about 60 percent of total slack funds and provinces only 40 percent. It is common to examine total reserves as a percent of general revenues or expenditures or some similar standard. Since reserves are ostensibly to be used to smooth out revenues and spending, the normalization gives a better idea of the fiscal importance and relative magnitude of reserves, over time and across units of government. Table 1 also shows provincial, district, and total subnational reserves as a percentage of spending, from end 2000 to end Total reserves were just 15 percent of expenditure before decentralization, rose to 19 percent in 2002, declined to 14 percent in 2003, increased steadily to 29 percent in 2006, and finally, declined slightly to 25 percent at the end of Over the period, district reserves increased from just 8 percent of expenditures to around 25 percent. On the other hand, provincial reserve funds decreased from 36 to 26 percent of expenditures. Figure 2 provides a view of end 2007 reserves as a percentage of expenditures, across provinces. 10 Reserves and expenditures refer to those of the particular province in question along with those of all district governments within the province. (Within provinces, the data do not allow us to separate provincial from district reserves.) The distribution of funds is quite skewed. Subnational governments in Aceh, Kalimantan Timur, and Riau 10. At the time of this paper was written, disaggregated subnational government expenditures were not yet available for As such we divide 2007 reserves by 2006 expenditures. Since 2007 expenditures are likely to have exceeded those for 2006, Figure 1 probably overestimates the reserves to spending ratio. Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 31

7 FIGURE 2 Subnational reserves as percentage of expenditure, all hold reserves in amounts well over 50 percent of expenditures. 11 Subnational governments in these provinces are the principal beneficiaries of shared natural resource revenues, especially those from oil and gas. Subnationals in these three locations possess more than 30 percent of total subnational reserves. At the other end of the scale, subnationals in Nusa Tenggara Barat, Sulawesi Tenggara, Lampung, and Sulawesi Utara all have reserves comprising o10 percent of total expenditures. Subnationals in these locations are among the fiscally poorest in all of Indonesia. SUBNATIONAL FISCAL SLACK: SIZE RELATIVE TO NEED In order to discern the extent to which the level of subnational government reserves reported above represents excess savings it is necessary to define an objective standard against which such reserves can be compared. A rule of thumb in the United States suggests that state governments should keep in reserve an amount of funds equivalent to about 5 percent of current expenditures. Most analysts argue that municipalities in the United States may, in fact, need more reserves because their revenues are more volatile The overall average reserves to expenditure ratio decreases from around 25 percent to o20 percent when Aceh, Kalimantan Timur, and Riau are excluded from the calculation. 12. See Kenneth Kriz The Optimal Level of Local Government Fund Balance: A Simulation Approach, State Tax Notes, January, He estimates that municipalities in Minnesota should maintain reserve fund balances of between 16 and 91 percent of annual revenues in order to assure a 3-percent rate of growth of annual expenditures, given various scenarios regarding revenue volatility. 32 Public Budgeting & Finance / Summer 2009

8 Many observers in the United States claim that, in practice, states do not maintain reserve fund balances of sufficient size, even according to the somewhat crude 5 percent rule. 13 Others contend that they do. 14 Interestingly, it appears that most analysts are more comfortable with the magnitude of reserve holdings of U.S. municipalities. It s certainly clear that, in the event, municipalities do maintain reserves in relatively larger sums. 15 The above observations may not be particularly relevant for Indonesian subnationals, since the size and composition of revenues is significantly different from those of U.S. states and municipalities. Moreover, it is not productive to attempt to generalize about the reserve fund requirements of Indonesian subnational governments, as a whole. Ultimately the optimal size of any subnational s reserves depends on the potential (future) volatility of the public revenues of the particular government in question. In Indonesia since decentralization, the most important subnational revenues have consistently grown and have not yet experienced declines of any significant magnitude. 16 As such, it is difficult to estimate the future volatility (especially on the downside) of major subnational sources of public revenue, either in the aggregate or for individual provinces or districts. This is not to say, however, that subnational revenues will not necessarily experience a significant downturn at some point in the future. Perhaps the most significant risk for regional revenues relates to the price of oil. The pools of finance of three of the major sources of subnational government revenuefproperty tax receipts from the oil sector, nontax oil revenue sharing, and block grants (DAU)Fare all derived as a direct or indirect function of the price of oil. The following paragraphs discuss the manner in which oil helps determine these major sources of revenue and simulate the impact of decreases in the price of oil on revenues. 13. Alan Berube and Iris J. Lav When it Rains it Pours: A Look at the Adequacy of State Rainy Day Funds and Budget Reserves, State Tax Notes, May, 1999, and Phillip G. Joyce, What s So Magical about Five Percent? A Nationwide Look at Factors that Influence the Optimal Size of State Rainy Day Funds, Public Budgeting & Finance 21, no. 2 (2001): Gary C. Cornia and Ray D. Nelson Rainy Day Funds and Value at Risk, State Tax Notes, August, 2003 and Christian Gonzalez and Arik Levinson State Rainy Day Funds and the State Budget Crisis of 2002, State Tax Notes, August, Charlie B. Tyer, Local Government Reserve Funds: Policy Alternatives and Political Strategies, Public Budgeting & Finance 13, no. 2 (1993): and Michael Shelton and Charlie B. Tyer (with Holly Hembree). Local Government Reserve Funds and Fund Balance: Some Applications of Business Concepts, Municipal Finance Journal 21, no. 1 (2000): 1 18, have demonstrated, for example, that the majority of municipalities in North and South Carolina hold reserves in amounts greater than 50 percent of general expenditures. And Justin Marlowe, Balance, Net Assets, and Working Capital, in Handbook of Public Financial Management, ed. Howard Frank (New York: Taylor & Francis, 2006), provides some evidence to show that a recent sample of municipalities in Michigan and Minnesota also hold reserves in amounts at least 50 percent of general fund spending. 16. The most important own-source revenues for provinces are taxes on vehicle registration and change of title while those for districts are taxes on electricity consumption and hotel and restaurant sales. Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 33

9 TABLE 2 Impact of Change in Oil Price on Regional Revenues (Rp Blns) Revenue source Rp (Blns) Rp (Blns) Rp (Blns) Oil price 5 $100 Oil price 5 $90 Rev loss Oil Sector Property Tax Shared with Regions 19,947 17,952 1,995 Nontax Oil Revenue Shared with Regions 37,728 32,914 4,814 DAU Dervied from Oil Revenue 50,675 44,518 6,157 Total Regional Oil-Based Revenue 108,350 95,384 12,966 Source: World Bank, Jakarta. Simulations assume 2007 cost structure and tax rates, 923 thousand barrels per day oil production, and an exchange rate of $1 5 Rp 9,150. Domestic nontax revenue from the oil sector is shared with regions. In 2008, the regional share of state oil revenues was 15 percent of total realized nontax revenues in the sector. The shared revenues are returned to the provincial point of origin and distributed among both producing and nonproducing districts within the province. Recent World Bank estimates put the elasticity of shared oil revenues with respect to changes in the dollar price of oil at The DAU pool of finance is calculated each budget year as 26 percent of total planned domestic revenues (net of amounts shared). Oil revenues are, of course, a major source of national revenues and thus have a major impact on the size of the DAU. World Bank staff estimate that the elasticity of oil sector-based DAU revenues with respect to changes in the budgeted price of oil is approximately Property tax in Indonesia derives from five sectors: urban, rural, estates, forestry, and mining (in particular, oil). Central government retains 9 percent of total property tax revenues as an administrative charge and distributes the remaining 91 percent to regions in various manners depending on the sector. Property taxes from the oil sector amount to about 75 percent of the total property taxes shared with subnational governments. 19 We have no specific information on the elasticity of subnational property taxes from oil with respect to changes in the price of oil and so we simply assume the elasticity to be one. While the price of oil has recently risen significantly and remained at historically high levels, it may not do so in perpetuity, of course. And if the price oil were to drop in world markets so too would subnational public revenues in Indonesia. Table 2 simulates the impact of changes to the price of oil on oil sector property taxes shared with regions, 17. Tim Bulman A Note on Oil Prices and Intergovernmental Transfers, mimeo, World Bank, Jakarta, Indonesia (2008). 18. Ibid. 19. The oil sector property tax operates somewhat like a sales tax, where property owners liabilities are derived as a percentage of total sales in the sector and not based on capital value of structures. See Blane D. Lewis, Property Taxation in Indonesia: Measuring and Explaining Administrative (Under-) Performance, Public Administration and Development 23, no. 3 (2003): Public Budgeting & Finance / Summer 2009

10 nontax oil revenue sharing, and the DAU. 20 We consider an oil price of US$100 per barrel as the base case. 21 The table shows the impact of a decrease in the price of oil from US$100 per barrel to US$90 per barrel. 22 As can be seen in the table, potential reductions in regional revenue due to a decrease in the oil price are not insignificant. A decrease in the price of oil from US$100 per barrel to US$90 per barrel results in a loss of revenue of Rp 1,995 billion, Rp 4,814 billion, and Rp 6,157 billion from property taxes, revenue sharing, and the DAU, respectively. The total decline in revenue under the simulated circumstances is just orp 13 trillion. This sum amounts to about 20 percent of end 2007 subnational reserves (Rp 68 trillion). Thus, in the aggregate, subnationals would appear to have more than sufficient funds on reserve to smooth out expenditures in the event of a one time 10 percent drop in the price of oil. Longer-term declines in the oil price would be more problematic for subnationals, however. The results here suggest, for example, that a 10 percent drop in the price of oil, sustained for 5 years, would eliminate total provincial and district reserves, if subnationals were to use the slack resources to maintain revenues at current levels, all other things remaining equal. 23 Larger decreases in the price of oil would eliminate fiscal reserves more quickly. The impact of an oil price decrease (limited and temporary, or otherwise) on the revenues of individual subnational governments would differ, of course. It is possible that even reasonably small and transitory decreases in the price of oil may reduce the revenues of some subnationals, those at or near the bottom of Figure 2, say, to a point at which they could not be covered by reserves. All things considered, the main problem with stock of subnational reserves in Indonesia would not appear to be one of excess level but to its uneven distribution across regions of the country. AN ECONOMETRIC EXPLANATION OF SUBNATIONAL SURPLUSES The approach used in this paper to econometrically explain subnational surpluses and reserves has a long history. 24 In broad terms, the (Gramlich-Galper) method posits that the subnational government surplus is a positive function of major (untied) fiscal trans- 20. A small amount of revenue from gas is also included in the simulations. The price of gas is pegged to the price of oil. As of 2008, the subnational share of nontax gas revenue was 30 percent. 21. The oil price assumption in the revised 2008 state budget was US$95 per barrel. The oil price for fiscal year 2009 is budgeted as US$100 per barrel (and may be subject to further revision). We use the latter figure for convenience. 22. The simulations assume that the 2007 cost structure of and tax rates in the oil sector prevail, a constant production level of 923,000 barrels per day, and an exchange rate of US$1 5 Rp 9, Among other things, this assumes no additional accumulation of subnational reserves in the interim. 24. E. Gramlich and H. Galper, State and Local Fiscal Behavior and Federal Grant Policy, Brookings Papers on Economic Activity (1973) 1: and E. Gramlich, The 1991 State and Local Fiscal Crisis, Brookings Papers on Economic Activity 2 (1991): Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 35

11 TABLE 3 Subnational Data, Totals, (Constant Rupiah) Revenue Sharing (Rp Blns) 33,367 37,726 45,251 49,862 57,210 69,184 68,461 Block Grants (Rp Blns) 100, , , , , , ,787 Reserves (Rp Blns) 30,949 33,504 30,946 32,957 48,020 73,049 67,872 GDP (Rp Trns) 2,306 2,479 2,588 2,861 3,197 3,338 3,782 GDP deflator Population (Mlns) Source: Authors own calculations based on Ministry of Finance data. fers to subnational units and a negative function of subnational expenditure needs. In this scheme, the subnational surplus is but one part of a larger fiscal model comprising subnational spending and taxes, as well, both of which are also a function of transfers and expenditure needs. 25 We use the methods here to determine subnational surplus during the period Table 3 supplies aggregate data on variables of interest in this analysis. It shows the amount of the major transfers to subnational governments, that is, revenue sharing and the general purpose block grant, along with data on subnational reserves, GDP, the GDP deflator, and population for the period The data are shown for background purposes only and are not discussed in much depth. Three points are worth highlighting, however. First, it should be emphasized that both revenue sharing and the general purpose block grants are untied. That is, the expenditure of those funds is completely at the discretion of subnational governments. While subnationals do receive some tied funds in the form of a special purpose capital grant (DAK), those transfers are very limited in amount and have not been included in this analysis. Second, notice the sizable real increase in shared revenues in 2005 and then again in This growth was largely driven by the significant rise in the oil price during those 2 years, which forces increases in subnational oil-based revenues, as described above. Third, note the substantial increase in block grants in This is also a function of the rising price of oil, as well as the fact that government adopted a reasonable estimate of that price in the state budget for the first time in many years. Previously, the government purposefully underestimated the price of oil in the national budget in order to 25. Blane D. Lewis, Indonesian Local Government Spending, Taxing, and Saving: An Explanation of Pre- and Post-Decentralization Fiscal Outcomes, Asian Economic Journal 19, no. 3 (2005): , employed the general Gramlich-Galper framework to study Indonesian local government spending, taxing, and savings just before and after decentralization. 36 Public Budgeting & Finance / Summer 2009

12 reduce its transfer obligations to the regions. 26 All other things being equal, a higher budgeted price of oil drives up planned domestic revenues, which serve as the basis for determining the pool of finance for the general purpose block grants. In any case, the considerable increase in revenue sharing and block grants in 2005 and 2006 corresponds closely with the significant rise in subnational surplus during those 2 years, as shown in Figure 1. This suggests at least one possible factor that might help determine the annual subnational surplus and thus the accumulation of reservesfincreases in untied intergovernmental transfers, as the Gramlich-Galper model indeed posits. This is investigated in more detail below. We specify a simple two-way fixed effects model to operationalize the methods described above. More particularly, it is assumed that: y it ¼ a 0 þ a i þ g t þ b 0 x it þ e it ð1þ where y is per capita subnational government surplus, defined as the difference between per capita reserves in the present and just previous periods. In addition, x is a vector of explanatory variables, a, g, and b are parameters to be estimated, e is the usual error term (with mean and variance equal to zero and s 2, respectively), and subscripts i and t refer to the province in which subnational governments are located and the year, respectively. The data set comprises 176 observations on 30 provinces over the period Explanatory variables comprise, among others, tax and nontax shared revenues and the DAU. 28 Transfer variables are measured in per capita terms. We expect per capita surplus to rise along with increases in intergovernmental transfers, all other things being equal. Prices faced by subnational governments are perhaps the broadest indicator of subnationals differential expenditure needs and we use the government s geographic price index as a proxy for those prices-cum-expenditure needs. 29 We hypothesize that per capita surplus would decline as the price index increases due to increasing cost of inputs and higher spending. All variables used in the econometric analysis are defined more precisely in Table As noted, the pool of finance for the DAU is based on planned domestic revenues. However, if realized domestic revenues exceed 130 percent of planned revenues at the time of midyear budget amendments, then amounts in excess of that 130 percent are transferred to the regions. Revenues shared with subnational governments are determined as a function of actual national revenues. 27. Data are available for 29 provinces for all 6 years and for one recently recreated province (Sulawesi Barat) for 2 years. Jakarta is excluded from the analysis due to lack of information on transfers. 28. Tax and nontax shared revenues are combined in the empirical analysis here. Estimation of the model with the two types of transfers treated separately resulted in severe problems of multicollinearity. Specific-purpose grants (DAK) are not considered as a possible explanatory factor of surpluses. As noted, such grants are very small in comparison with others and are of less interest in the current context. 29. The government geographic price index is based on consumer price information gathered in 50 cities across Indonesia, where each of the 30 provinces is represented by at least one city. In cases where a province is represented by only one city, that city s index value is used for the province; where a province is represented by more than one city, the average of the price indices is used. The base year of the index is 2002 and the cross provincial average in that year is equal to 100. Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 37

13 TABLE 4 Variable Names and Definitions Name Dependent variable SPLSPC Explanatory variables SHRVPC BLKTRPC GPI Definition Total subnational government surplus, within province, per capita Total revenue sharing transfers to subnational governments, within province, per capita Total block grants to subnational governments, within province, per capita Geographic price index Table 5 provides the regression results. In addition to the output for the two-way fixed effects model in equation (1), for comparison purposes, the table also shows the results for the one-way (i.e., provincial) fixed effects, and the simple OLS (i.e., with no fixed effects) versions of the model. For each of the models, the table provides the estimated coefficients of the independent variables, the relevant t statistics, and an indication of the statistical significance of the estimated coefficients. In addition, the number of observations, log-likelihood function, restricted log-likelihood function, w 2, and adjusted R 2 values are shown at the bottom of the table. The usual likelihood ratio and F tests indicate that the two-way fixed effects model performs best among the three. 30 We restrict discussion of the output to the two-way model. As the table shows, the marginal effects of an increase in per capita shared revenues and per capita block grants are 0.62 and 0.14, respectively; both coefficients have the expected signs and are statistically significant. At the margin, therefore, an extra rupiah of revenue sharing leads to nearly four and a half times as much savings as does an additional rupiah of block grants. At the point of means of relevant variables, these marginal effects lead to estimates of elasticities of per capita surplus with respect to per capita revenue sharing and per capita block transfers of 0.98 and 1.17, respectively. In other words, a percentage increase in block grants would lead to higher subnational savings than same percentage increase in revenue sharing would. This is not a trivial matter from a policy perspective. Increases in the pools of finance for transfers are typically implemented in percentage terms. In 2008, for example, the government increased the subnational pool of finance for the DAU from 25.5 to 26 percent of planned domestic revenues. In 2009, it will increase the region s share of national oil revenue from 15 to 15.5 percent. The estimation results here 30. More specifically, the tests indicate that the one-way fixed effects model does not improve on the fit of the simple OLS model but that the two-way fixed effects model performs better than either the one-way or the OLS model. Detailed results are available from the authors upon request. 38 Public Budgeting & Finance / Summer 2009

14 TABLE 5 Surplus Regression Output a OLS One-way fixed effects Two-way fixed effects Variable Coefficient t stat Coefficient t stat Coefficient t stat Constant 2,279, * F F 4,355, * SHRVPC * * * BLKTRPC * * GPI 42, * 44, * 79, * No. Observations Log likelihood 2, , ,172.3 Restr log likelihood 2, , ,393.8 Chi square Adjusted R a Dependent variable is SPLSPC. * Coefficient is statistically significant at the 0.01 level. imply that the percentage increase in block grants will have led to larger subnational savings than the percentage increase in oil revenue sharing, all other things being equal. Note also that the coefficient of GPI is negative and statistically significant, as hypothesized. The outputs implies that a one point increase in the price index results in nearly Rp 80,000 decline in the per capita surplus, all else remaining the same. It is instructive also to consider the two sets of fixed effects from the estimation of equation (1) in more detail. Table 6 provides the estimated fixed group and time effects. The group effects are ranked from lowest to highest. Note that the provinces with the some of the smallest fixed group effectsfriau and Kalimantan Timur, especiallyfare also those with the largest reserves. This implies that subnationals in those provinces actually tend to spend more and save less than would typically be the case, holding transfers (and prices) constant. On the other hand, the fixed effects for some provinces on JavaFJava Tengah, Banten, and Java TimurFare among the largest. The interpretation here is that subnationals in those provinces actually spend less and save more than is usual, given transfers (and prices). More generally, we might conjecture that those subnationals with large positive fixed group effects could be suffering from some kind of capacity constraints regarding the spending of available funds. The results for the fixed time effects are also illuminating. As the table shows, these fixed effects continuously increase over time until 2006 at which point they decline precipitously. This may indicate that subnational governments are perhaps finally beginning to learn how to spend their resources to a fuller extent. It s worth noting that the slight decline in reserves in 2007, as seen in Figure 1 and Table 1, is at least consistent with this hypothesis. Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 39

15 TABLE 6 Fixed Group and Time Effects Group/province Fixed effects Time/year Fixed effects Riau 438, ,270 Sumatra Barat 376, ,186 Kalimantan Timur 300, ,900 Papua 202, ,051 Sumatra Selatan 110, ,968 Aceh 47, ,039 Sulawesi Tengah 35,017 Malluku Utara 27,228 Sulawesi Tenggara 22,125 Maluku 19,627 Kalimantan Tengah 15,049 Sumatra Utara 9,226 Sulawesi Barat 4,993 Gorontalo 5,321 Sulawesi Utara 13,104 Bengkulu 21,410 Kalimantan Selatan 28,283 Lampung 35,198 Sulawesi Selatan 36,530 NTT 36,603 Yogyakarta 37,186 Kalimantan Barat 39,590 Bali 45,797 Java Barat 46,864 NTB 51,449 Java Tengah 53,731 Banten 65,518 Bangka Belitung 82,750 Java Timur 86,364 Jambi 198,051 EXAMINATION OF AUDITED SUBNATIONAL GOVERNMENT BUDGETS The regulatory framework for subnational government budgeting is found in Law 32/04, Government Regulation 58/05, and Ministry of Home Affairs Decree 59/07. The regulations do not treat the development, use, and management of formal reserve funds (dana cadangan) in much depth. Subnational governments are simply informed that they may create such funds and that to do so they must issue a local regulation, noting the 40 Public Budgeting & Finance / Summer 2009

16 general intent of the dana cadangan, along with its size, source of funds, 31 and the kinds of activities that will be financed from the formal reserves. Subnational governments must create special bank accounts for reserve funds. In the event that reserve funds are not put to immediate use they may be invested in financial instruments with steady return and low risk. 32 Any earnings from such investments are to be channeled back into dana cadangan. Finally, subnational governments are required to report their reserves on their balance sheets, which must be sent to the Ministry of Finance for review annually. In the event, it appears that very few subnational governments create formal reserve funds of the kind outlined in regulation and described above. Our review of subnational audited budgets revealed that o5 percent of total subnational government funds on deposit in commercial banks were held in formal reserve fund accounts. This, by itself, suggests that the accumulation of subnational reserves may be adhoc and inadvertent. If subnational governments do not plan for budget surpluses, then they generate surpluses unintentionally, either by underestimating revenue, by overestimating expenditure, or by some combination of the two. 33 We explore the extent to which subnational governments misestimate their revenues and expenditures just below. We base the analysis on the audited budgets and budget out-turns of 27 (out of 31) provinces and 220 (out of 440) districts over the period. The provincial and district budgets used for this analysis were not chosen randomly but selected based exclusively on the availability of information. Overall, the selected provinces and districts cover 89 and 63 percent of the total population, respectively. In addition, we highlight key results from some limited field work carried out in conjunction with this study. The field work comprised structured interviews with subnational government finance officials from three provinces and 10 districts. 34 Table 7 presents information on planned and realized revenues, by major source, and expenditures, by main type, for provinces and districts in our sample of subnational governments for (The results for 2004 and 2005 are similar and are not shown here.) The initial point to be gleaned from the table concerns the subnational deficit/ surplus. As the table shows, while both provinces and districts planned for fiscal deficits, in the event they ran quite significant fiscal surpluses. As can be readily be determined from information in the table, provinces planned for a deficit in the amount of just over 11 percent of (planned) expenditures but in the end ran a surplus of 7 percent of (re- 31. Regulations assert that all regional revenues may be used to fund such reserves, except the special purpose grant. Furthermore, subnationals may not use borrowed funds to create reserves. 32. As examples of such financial instruments the regulations cite bank deposits, government treasuries, and other government secured investments. 33. Tyer (1993). 34. Andre Oosterman and Bambang Tata Samiadji Increases in Surpluses of Regional Governments in Indonesia: An Empirical Analysis, mimeo, Final Report Ministry of Finance. Decentralization Support Facility, Jakarta, Indonesia, Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 41

17 TABLE 7 Planned and Realized Subnational Revenues and Expenditures, 2006 (Blns Rp) Provinces Districts Planned Realized Difference % Planned Realized Difference % Revenues Own-Source 27,308 28, ,294 9,322 1, Shared 16,758 18,412 1, ,436 25,488 5, DAU 11,959 11, ,358 71, DAK ,758 5, Other 4,676 5, ,979 7, Total 60,711 63,770 3, , ,542 6, Expenditures Current 52,027 46,491 5, ,572 79,532 9, Capital 16,337 13,145 3, ,323 26,170 8, Total 68,364 59,636 8, , ,702 17, Deficit/surplus 7,653 4,134 11, ,069 13,840 23, Source: Authors own calculations based on audited provincial and district budgets for alized) expenditures; and districts planned for a deficit in the amount of 8 percent of expenditures but ran a surplus of 13 percent of spending. Table 7 also shows the difference between planned and realized revenues and expenditures. These differences are shown in nominal terms and as a percentage of the difference between the planned deficit and realized surplus (in absolute value). We term the latter the total forecasting error. The percentage figures show the contribution of the revenue and expenditure forecasting errors to the total forecasting error. The revenue forecasting error determines about 26 and 28 percent of the total forecasting error of provinces and districts, respectively. In both cases, underestimated shared revenue makes up the bulk of the problem; it is important to note that most of the difficulty in this regard relates to revenue sharing in the natural resources sectors, particularly oil. Forecasting errors related to own-source revenues and other revenues are considerably less problematic. And such errors associated with DAU and DAK are trivial. Inaccurate subnational planning associated with shared revenues in the oil sector is a function of two main underlying problems. The first concerns the general difficulty the central government has in correctly forecasting the price of oil and the impact that such inexact forecasts have on subnational revenue planning. The Ministry of Finance provides estimated revenue sharing distributions to subnational governments for a given year at the end of the before fiscal year, based on the forecasted price of oil at the time. Recently the price of oil used by government at the end of one fiscal year for such purposes has been significantly less than the actual average price in the following fiscal 42 Public Budgeting & Finance / Summer 2009

18 year. In 2005, for example, the end of year forecast for oil revenue sharing with regions was about Rp 16.5 trillion; in the event, realized revenue sharing in the sector was closer to Rp 19 trillion in 2006 (representing a 15 percent forecasting error). It is the former figures that work their way into subnational revenue budgets and the latter that account for realized revenues of subnationals. As such, it is easy to see how provinces and districts underestimated their revenues from this source during The second problem relates to the timing of natural resource revenue payments to regions. The payment of revenue shares has recently been made to subnationals rather late in the fiscal year. In 2006, the central government did not make any natural resource revenue transfers to regions until the third quarter. It seems that this problem may be related in the first instance to cash flow difficulties of the central government. That is, the central government itself typically does not get access to much natural resource revenue itself until somewhat late in the year and it simply chooses to wait until a convenient time to transfer onward to subnational governments their respective shares of the revenue. In any case, subnational revenue received late in the fiscal year is difficult to spend in full. Expenditure forecasting errors explain most of the total forecasting error of both provinces and districts. As Table 7 shows overestimated expenditure makes up 74 and 72 percent of the total error of provinces and districts, respectively. Provinces apparently have more difficulty in forecasting current expenditures while districts seem to underforecast current and capital spending to about the same extent. Three main problems drive the overestimation of spending: inflexible budget rules, the national anticorruption program, and insufficient subnational capacity in project development and implementation. Subnational governments are required to prepare very detailed expenditure budgets. Each unit (dinas) of the province or district develops a budget, according to exhaustive spending categories, including those by type (or object) and (economic) function. The budget is subsequently approved (by the relevant higher level government and the subnational parliament) and expenditures are then authorized. However, no provincial or district government dinas is allowed to spend more than the budget ceiling, or even reallocate planned spending from one budget category to another, unless it undertakes to formally revise its budget, something that requires subnational parliamentary approval. Thus, all subnational government unit requests for additional spending must be coordinated and presented to the subnational parliament for approval. This is a time consuming process that some but not all subnational governments manage to carry out a maximum of one time each year. The end result is, invariably, less spending than originally envisioned. The recent and ongoing national program against government corruption has been a positive force. At the same time, however, it also appears to have played a role in constraining investment spending by subnationals. The basic problem seems to be in the formation of local tender committees, which are required by government regulation to review proposals and award construction contracts. Subnationals appear to have found it increasingly difficult to find officials willing to take part in such committees. In a number of cases, officials have been visited by police immediately after their names have Lewis and Oosterman / Impact of Decentralization on Subnational Government Fiscal Slack 43

19 been proposed as tender board members but even before meetings have take place. The implicit suggestion is that the police are acting in an overzealous manner in seeking out potential corruptors. Of course they may also be searching for opportunities to participate in fraudulent activities. In any case, this has made many subnational government officials wary of participating in tender committees. The resulting delays in forming committees (if they are formed at all) have adversely affected public investment in infrastructure. Finally, perhaps the most important cause of underspending concerns the lack of subnational government capacity to plan and implement development projects. It is difficult to provide objective measures of this lack of this rather intangible phenomenon. However, capacity constraints came out clearly in the interviews conducted as part of this research. It seems that many subnational government officials have a limited understanding of their roles as providers of public infrastructure services. Officials that do have some appreciation that their main function is to spend on public capital and deliver services often have in mind infrastructure of rather limited scale, believing that major public expenditures should be made by the central government instead. In this case, they appear to see their role as one relegated to lobbying appropriate central officials to make needed public investments rather than arranging to undertake these investments themselves. 35 SUMMARY AND POLICY IMPLICATIONS Policymakers in Indonesia have become increasingly concerned at the rapid and significant buildup of unspent balances at the subnational level since decentralization. Nominal subnational reserves grew at an annual rate of 37 percent between end 2000 and end As of 2007, reserves amounted to just orp 70 trillion, an order of magnitude larger than in During the period in question, district reserves grew even more swiftly, almost 50 percent per year. District slack resources now comprise about 60 percent of the total, up from 30 percent just before decentralization. It is difficult, however, to determine the extent to which subnational reserves are, in fact, excessive. While Indonesian provinces appear to maintain reserves in larger proportions of total expenditure than their state counterparts in the United States, for example, district governments in Indonesia do not seem to hold slack funds in amounts exceeding those of many U.S. municipalities. In the end, the optimal amount of reserves that Indonesian subnationals should maintain is a function of potential revenue volatility, not what transpires in other countries. That is, subnationals need to have sufficient funds on reserve that they can use to keep expenditures at constant levels, in the event of a downturn in public revenues. In Indonesia, subnational revenues from all sources have 35. Blane D. Lewis and Jasmin Chakeri, Central Development Spending in the Regions Post Decentralization, Bulletin of Indonesian Economic Studies 40, no. 3 (2004): Public Budgeting & Finance / Summer 2009