The World Bank April PREMnotes 2010 NUMBEr 150 90184 ECONOMIC POLICY Agriculture Public Spending and Growth: The Example of Indonesia Enrique Blanco Armas, Camilo Gomez Osorio, and Blanca Moreno-Dodson This note1 analyzes the trends and evolution of public spending in the agriculture sector in Indonesia, as well as its impact on agriculture growth during the period 1976–2006. We demonstrate empirically that public spending on agriculture and irrigation had a positive impact on agriculture growth during that period, whereas public spending on fertilizer subsidies had the opposite effect. As Indonesia continues its efforts to revitalize the agriculture sector, public spending should be directed at improving the provision of public services rather than the subsidization of private inputs. Supplying Public Goods Versus the provision of public goods is often nega- Subsidizing Private Inputs tive (as will be discussed and shown in this According to the theory of public econom- “Economic Premise”). Subsidizing private in- ics, only the public sector can supply public put often represents only a transfer of re- goods efficiently (and at adequate amounts) sources, with no impact on the consumption because the market will always under-provide of that input; and, even if the subsidy increases those goods. When supplied in a cost-effec- its use, its impact on productivity is unclear. tive way, public goods will generate higher re- For example, there are diminishing returns on turns than will investments in private inputs fertilizer use beyond a point; additional use because they will create positive externalities may no (or a negative) effect on production if for the economy as a whole. Because govern- other production inputs are not available. ments have the capacity to collect individual contributions to provide public goods, they The Importance of Public Spending can also capture economies of scale, access for Agriculture Growth funding, and manage risk better than farm- The purposes of the public sector in the de- ers can manage it. As a result, they are better velopment of agriculture—primarily a private suited to supply public goods. sector activity—are to set an enabling environ- However, the impact on productivity of ment where private sector activities can flour- subsidizing private inputs is unclear. As shown ish; to correct circumstances under which the in the literature review conducted for the In- market fails to allocate resources efficiently; donesia Agriculture Public Expenditure Re- and to minimize the price distortions faced by view, the record of governments subsidizing both farmers and consumers—all while pro- private inputs in the agriculture sector is moting inclusive growth. In practice, those mixed, at best—although many governments tasks translate into interventions along several spend a considerable share of their budgets dimensions: on such subsidies. The productivity impact • correcting for externalities, which requires of subsidizing private inputs at the expense of making people pay (or be paid) for the FrOMTHEpOVErTYrEDUCTiONANDECONOMiCMANAGEMENTNETWOrK cost and benefits of their actions— such tivity stagnation from the 1990s onward is the as by discouraging the overuse of fertil- low levels of both private and public invest- izer that leads to pollution, or by reward- ments—with public investments in research ing with a patent advances in research and development, rural infrastructure, and and development; irrigation being necessary complements to • providing for public goods that are not effi- private investments in agriculture. ciently and sufficiently produced by the mar- We have reviewed areas through which ket—such as by building rural roads and government spending can have a positive im- irrigation systems, providing extension pact on agriculture growth—irrigation, agri- services and agriculture marketing, and culture research and development, extension funding more agriculture research and services, infrastructure/rural roads, and sub- development; sidies to private inputs. These do not neces- • addressing information asymmetries and sarily constitute an exhaustive list of factors eliminating information gaps so that farm- that determine agriculture growth; but our ers and consumers can make informed focus is primarily on the provision of public decisions on what to produce, with what goods and services with high fiscal implica- level of inputs, and at what price—for tions, and on other public spending cate- example, by certifying product input gories that are not necessarily public goods and output quality standards and ensur- (primarily, the provision of subsidies to pri- ing plant and animal health; and vate inputs such as seeds, fertilizer, or credit) • regulating against monopolistic behavior that but that do affect the volume and composi- reduces social welfare—for example, by tion of the budget. Items such as trade policy having lower outputs sold at higher prices. or the business environment are also crucial There is renewed interest in improving for the development of agriculture; but be- our understanding of the impact of public cause public intervention in these areas usu- spending on agricultural growth.2 All these ally has lower spending implications, it will be studies emphasize the concept of opportunity excluded from this note. costs of subsidies. Although increased use of a particular input may have a positive effect on Trends of Agriculture Public production (such as the effect of fertilizer on Spending in Indonesia rice production), the impact of subsidizing The question that motivated the analysis was such inputs is often negative because it is whether the volume and composition of pub- done at the expense of providing public goods lic spending are having an impact on growth (such as funding for research on newer vari- in the agriculture sector. We first analyze re- eties or improvements to the irrigation net- cent trends for public agriculture spending work) that have a larger positive impact on in relation to growth, followed by a time-se- production. ries quantitative assessment of the impact of In Indonesia, Fuglie (2004) identifies public spending on per capita growth in In- the drivers of growth in agriculture between donesia’s agriculture sector during the peri- the 1960s and 2000. He argues that although od 1976–2006. agricultural productivity in the 1970s and Public spending on agriculture has in- 1980s was increasing, this trend has been flat creased recently in real terms and without a since the early 1990s, with most growth in corresponding increase in agricultural pro- agriculture being explained by increases in duction. During the years 2001–09, national production inputs (labor and land). Fuglie spending on agriculture3 increased from Rp also argues that the reason for the produc- 11.0 trillion to Rp 61.5 trillion, an annual av- 2 prEMNOTE MONTH 2010 Figure 1: Agriculture Spending, 2001–09 credit. As figure 2 shows, by end-2009 the allo- 40 cation for agriculture subsidies was four times 35 its 2001 level, while resources for irrigation have been flat since 2001. The budget of the 30 Ministry of Agriculture has increased signifi- 25 Share (%) cantly since 2001, but has grown at a slower 20 pace than agriculture subsidies. 15 Government investment in public goods 10 was largely behind Indonesia’s success in in- creasing agricultural productivity from the 5 1970s to the early 1990s. During the years of 0 the Green Revolution, Indonesia invested 01 02 03 04 05 06 07 08 09 heavily in its irrigation network, research and 20 20 20 20 20 20 20 20 20 Year development, extension services, and rural agriculture spending/agriculture GDP infrastructure; and it subsidized private agri- agriculture spending/national spending culture inputs (fertilizer, seeds, and credit). agriculture spending/GDP By the early 1990s, the country had achieved Source: World Bank staff calculations. high yields across several commodities, in- cluding rice, cereals, and potatoes (World Bank 1994). Unfortunately, in the 1990s the erage of 12 percent in real terms. This was upward trend in productivity flattened. Ex- the result of large budget increases and a big acerbated by declining levels of private and spending boost from decentralization across public investment, agricultural productivity all sectors, with even greater amounts for growth remains sluggish today. agriculture. As figure 1 illustrates, the agri- Spending as a share of GDP in agricul- culture share of the budget doubled from 3 ture averaged 10 percent and 8 percent in percent in 2001 to 6 percent by 2008; by that the 1970s and 1980s, respectively, compared year, it reached 1 percent of GDP because of with 40 percent today. As discussed earlier, increased spending on agriculture subsidies. we argue that because most of the increased This increase did not result in a correspon- ding rise in agricultural production, which increased an average of 3 percent between Figure 2: Index of Agriculture Spending, 2001–09 2001 and 2009. Low agriculture growth com- 5 bined with a constant share of labor force 4 Index (2001 = 1) participation in the sector has led to stagnant per-worker value-added. 3 Recent public spending trends in agri- 2 culture show that resources are being direct- 1 ed toward supporting private goods at the ex- 0 pense of providing public goods. In 2009, 01 02 03 04 05 06 07 08 09 the government of Indonesia directed 56 20 20 20 20 20 20 20 20 20 percent of agriculture resources (Rp 34.4 Year trillion) toward subsidizing private goods: agriculture spending/agriculture GDP fertilizer subsidies accounted for almost half agriculture spending/national spending agriculture spending/GDP (Rp 18.5 trillion), and the remainder was al- located to seeds, RASKIN,4 and agriculture Source: World Bank staff calculations. MONTH 2010 prEMNOTE 3 spending in agriculture is directed at private GDP; ATR is a ratio of 3 percent of total tax goods, it has not translated into a propor- revenues to agriculture GDP. Control vari- tional increase in growth. ables: l is the agriculture labor force (in thou- sands); la is the arable land, as an asset of the Empirical Application to Indonesia’s farmers (hectares); GD is an index of global Agriculture Sector demand for agriculture exports, total world In the empirical analysis, we look at the rela- demand for agriculture, crops, livestock, and tionship between agriculture public spending primary and processed exports (in billions of and the growth rate of agriculture GDP per U.S. dollars). Dummy98 is a dummy, 1 for the capita, using time-series data with both ordi- 1998 financial crisis; et is an error term; and nary least squares and general method of mo- b0, b1, b2, b3, b4, b5, b6, and b7 are the coeffi- ments econometric techniques. The model cients assigned to the independent variables. chosen for this policy note introduces specific Initially, the direct sensitivity of agricul- characteristics and innovations to fit the In- ture GDP per capita growth to public spend- donesian context as well as the broader analysis ing is tested using a functional specification objectives of the Public Expenditure Review.5 in which the rate of growth is the dependent Empirically, the connection between pub- variable. Two groups of independent vari- lic spending and growth has been difficult to ables are considered: fiscal variables and con- establish in the literature. The relationship trol variables. Regarding the former, the between agriculture and nonagriculture GDP spending effect initially is aggregated into to- is likely to be simultaneous rather than unidi- tal agriculture public spending (APE)6 and rectional—and that can introduce a simul- then is broken down into two components: taneity bias within the fiscal variables and the (1) development spending on agriculture dependent variable (growth rate of agricul- and irrigation and (2) fertilizer subsidies. All ture GDP). To minimize this risk, the fiscal fiscal variables include central government variables enter the specification lagged, and a and subnational spending and are consid- more rigorous use of instruments is consid- ered as ratios to agriculture GDP. The gov- ered by using the general method of mo- ernment budget constraint is considered in ments technique. Estimating the model with the specification function, for methodologi- general method of moments is a way to test cal reasons, by introducing ATR as a proxy the robustness of these empirical estimations for fiscal revenues.7 This assumes implicitly by exploring the relationship within a dynam- that, under an earmarking hypothesis, there ic setting. Also, although time-series analysis would be no fiscal deficit for the agriculture ideally would benefit from a longer time span, sector (period average). data availability limited the number of obser- The selection of control variables takes vations in the model to a 30-year period. into account the factors affecting supply of The basic model is as follows: and demand for agriculture output in In- donesia. Nonagriculture GDP is used to cap- (1) lnyt = b0 + b0lny2t + b2DAPEt–1 + ture the spillover synergies from growth in b3DATRt–1 + b4∆lnlt + b5∆lnlat + industry and services. On the supply side, the b6∆lnGDt + b7Dummy98t + et, labor and arable land variables are used as where t is the year, y is the per capita growth proxies for private inputs to capture possible rate of agriculture GDP, and y2 is the rate of complementarity effects between private and growth of nonagriculture GDP per capita. public assets. The agriculture demand vari- Fiscal variables: APE is the ratio of total agri- able GD is an index of global demand for culture public expenditures to agriculture agriculture exports whose fluctuations may 4 prEMNOTE MONTH 2010 have affected the production supply response of a two-pronged strategy that maximizes in Indonesia. Finally, a time dummy variable8 spending effectiveness, brings higher re- is introduced for 1998 to control for growth turns, and leads to growth for the agriculture effects related to events in that year—namely, sector; and pays attention to farmers’ welfare the Asian financial crisis affecting Indonesia. and people’s access to affordable food. The fiscal variables enter the functional The government may consider these in- specification with a one-year lag to capture terventions: the fact that the agriculture GDP per capita • Reallocate public spending from subsidiz- growth rate is dependent on the previous ing private inputs (fertilizer, seeds, and year’s public spending. Changes in taxation, grants to farmers and farmers’ groups) to for example, may have a particularly lagged providing agriculture and irrigation pub- effect because they affect the following fiscal lic goods and services. year. In contrast, the control variables are • Reorient government support to help considered contemporaneously, assuming small farmers and farmers’ groups meet that they are most likely to determine the the domestic demand for higher-value- agriculture growth rate within the same year.9 added products and gain access to the The overall results show that spending global value chain. This would entail a on agriculture has a statistically significant shift away from food or estate crops (the positive effect on the agriculture GDP per focus of the current agriculture policy). capita growth rate, after controlling for the • Continue the government’s income sup- effects of nonagriculture GDP per capita port to small and poor farmers, but pro- growth and for private inputs (labor and vide incentives to pursue productive in- arable land). We then split public spending vestments. Indonesia has extensive on agriculture into (1) spending for public experience in the area of cash transfers. goods (development spending for agricul- Providing incentives to put these cash ture and irrigation) and (2) spending on fer- transfers to productive use may result in tilizer subsidies. We find that spending on pub- increased investment and productivity in lic goods is a positive driver of the per capita the agriculture sector. Also, careful target- growth rate of agriculture GDP, whereas ing will be key to ensuring the efficiency spending on fertilizer subsidies appears to of the program and their fiscal viability. have a significant negative effect. The posi- • Put in place a comprehensive monitor- tive effect of public spending on agriculture ing and evaluation system that enables is associated only with the agriculture and ir- the government to assess the impact of rigation public spending component. Given its transfer programs. Such a system would the opportunity cost of further financing be instrumental in preventing and cor- subsidies at the expense of other agriculture recting mistakes in program design, max- spending and irrigation directly contributing imizing effectiveness in agricultural pro- to growth, the government should consider ductivity, and alleviating poverty in rural reallocating spending from fertilizer subsi- areas. dies to public goods (such as agriculture ex- tension services, research and development, References and irrigation) that could lead to faster sec- Allcott, Hunt, Daniel Lederman, and Ramón tor growth. López. 2006. “Political Institutions, Inequality, As Indonesia modernizes the agriculture and Agricultural Growth: The Public Expen- sector and income levels increase, it will be diture Connection.” Policy Research Working important to allocate resources on the basis Paper 3902. World Bank, Washington, DC. MONTH 2010 prEMNOTE 5 Bayraktar, Nihal, and Blanca Moreno-Dodson. sector, a large share of spending on subsidies to Forthcoming. “How Public Spending Can Help private inputs has a negative impact on agriculture You Grow. An Empirical Analysis for Develop- growth, given the corresponding lower spending ing Countries.” World Bank, Washington, DC. on the provision of public goods. López and Gali- Blanco Armas, Enrique, Blanca Moreno-Dodson, nato (2007) find similar results and argue that the and Camilo Gomez Osorio. Forthcoming. “Ag- positive effect of public spending on rural incomes riculture Public Spending and Growth: The is primarily dependent on the composition of Example of Indonesia.” PREMnote 149. spending. They estimate that a 10 percent reallo- Fuglie, Keith O. 2004. “Productivity Growth in In- cation from subsidizing private goods to providing donesian Agriculture, 1961–2000.” Bulletin of public goods can increase per capita income from Indonesian Economic Studies 40 (2): 209–25. agriculture by 5 percent. In a related piece of López, Ramón, and Gregmar I. Galinato. 2007. work, Santos and Ortega (2006) show how the “Should Governments Stop Subsidies to Private share of the budget allocated to subsidizing private Goods? Evidence from Rural Latin America.” inputs has a negative and significant effect on the Journal of Public Economics 91 (5–6): 1071–94. efficiency of public spending. Moreno-Dodson, Blanca. 2008. “Assessing the Im- 3. National spending on agriculture includes pact of Public Spending on Growth. An Em- central government spending on irrigation by the pirical Analysis for Seven Fast Growing Coun- Ministry of Public Works and on agriculture by tries.” Policy Research Working Paper 4663. the Ministry of Agriculture. Subnational govern- World Bank, Washington, DC. ment spending on agriculture and irrigation is Santos Rocha, Józimo, and Jorge Ortega. 2006. done by district and provincial governments and “Crecimiento, inversión privada y eficiencia agriculture subsidies. del gasto público en las áreas rurales de Améri- 4. RASKIN (Beras Miskin) is not a subsidy to agri- ca Latina y el Caribe.” In Políticas Públicas y De- culture inputs, but primarily an instrument to sub- sarrollo Rural en América Latina y el Caribe—El sidize rice consumption for the poor. To the extent papel del gasto público, ed. Fernando Soto Ba- that it increases domestic demand for rice and it is quero, Józimo Santos Rocha, and Jorge Orte- partly used to stabilize prices and therefore provide ga, 127–63. Santiago de Chile: Food and Agri- an incentive for increased rice production, it will culture Organization of the United Nations. also have an impact on rice production. In any case, World Bank. 1994. “Indonesia, Stability, Growth we include it as an agriculture subsidy because the and Equity in Repelita IV.” Washington, DC. Ministry of Finance includes it; note, however, that it is not entirely a subsidy to agricultural production. Endnotes 5. Different function specifications were con- 1. This note is based on the results of the In- sidered, taking into account previous analyses of donesia Agriculture Public Expenditure Review. the impact of public spending in the agriculture The work is being carried out within the Initiative sector. See, for example, López and Galinato for Public Expenditure Analysis framework, a (2007); Moreno-Dodson (2008); and Bayraktar joint initiative by the government of Indonesia, and Moreno-Dodson (forthcoming). donors (the Dutch government, the European 6. Agriculture public spending is defined as the Commission), and the World Bank. Dwi Endah sum of development spending on irrigation and Abriningrum provided research assistance. agriculture plus spending on fertilizer subsidies. All 2. Evidence provided by a U.N. Food and Agri- data come from the Indonesian budget publica- culture Organization research project conducted tion Indonesia Statistical Yearbook. in 20 countries in Latin America shows that public 7. Because there is no variable capturing the spending in rural areas has a positive impact on fiscal revenue burden on the agriculture sector, agriculture growth (Allcott, Lederman, and López we use 3 percent of total tax revenues (based on 2006). The study also shows that both the volume the fact that average agriculture public spending and the composition of spending matter. Assum- has represented about 3 percent of total public ing a fixed amount of spending in the agriculture spending during the period of analysis). 6 prEMNOTE MONTH 2010 8. Inclusion of the dummy fortifies the results, This is an important precondition to meet when but it does not change them. It corrects the noise conducting time-series analysis, because the pos- introduced by the economy’s contraction follow- sibility of including a nonstationary series can re- ing the Asian financial crisis. GDP declined over sult in picking up a spurious effect. 13.0 percent in 1998, and agriculture declined by 1.3 percent. Without the dummy, the relationship About the Authors between the fiscal variables and per capita agri- Enrique Blanco Armas is a senior economist and culture GDP growth is similar; but significance Camilo Gomez Osorio is an economist of the Poverty levels range between 5 percent and 10 percent. Reduction and Economic Management (PREM) Net- 9. After testing for the stationarity of the series, work, East Asia; and Blanca Moreno-Dodson is a the variables are estimated in first difference, senior economist of PREM, World Bank, Washington, which ensures the stability of the model over time DC. To learn more about PREM, please visit http:// (and corrects for the possibility of unit roots and www.worldbank.org/prem. the spurious regression problem in the results). This note series is intended to summarize good practices and key policy findings on PREM- related topics. The views expressed in the notes are those of the authors and do not neces- sarily reflect those of the World Bank. PREMnotes are widely distributed to Bank staff and are also available on the PREM Web site (http://www.worldbank.org/prem). If you are in- terested in writing a PREMnote, email your idea to Madjiguene Seck at mseck@worldbank.org. For additional copies of this PREMnote please contact the PREM Advisory Service at x87736. PREMnotes are edited and laid out by UpperCase Publication Services, Ltd. Prepared for World Bank staff