A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty

Azis, Iwan J. Working Paper A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty ADBI Research Paper Series, No. 35 Provided in Cooperation with: Asian Development Bank Institute (ADBI), Tokyo Suggested Citation: Azis, Iwan J. (2002) : A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty, ADBI Research Paper Series, No. 35, Asian Development Bank Institute (ADBI), Tokyo, https://hdl.handle.net/11540/4141 This Version is available at: https://hdl.handle.net/10419/111127 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ADB INSTITUTE RESEARCH PAPER 35 A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty Iwan J. Azis March 2002 ADB INSTITUTE TOKYO ASIAN DEVELOPMENT BANK INSTITUTE ASIAN DEVELOPMENT BANK INSTITUTE Most studies attempting to link macroeconomic trends—particularly growth—and poverty have used aggregate cross-country data and unsophisticated regression models with limited usefulness for policy analyses. They do not really explain the mechanisms through which growth, let alone macroeconomic fluctuation, affects poverty. In the context of financial crisis, many studies compare poverty conditions before and after the crisis, as if everything that led to the rise of poverty was due to the financial shock. The current study is intended to fill the gap, by making use of an economy-wide model with a price endogenous feature, detailed financial sector, and explicit poverty module. Applied to the case of a specific country—Indonesia—the model is subsequently used to generate a set of counterfactual policy scenarios. It is shown that alternative policies during the crisis would have been more favorable in terms of employment, income distribution, and poverty, compared to the actual (benchmark) scenario. ADB Institute Research Paper Series No. 35 March 2002 A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty Iwan J. Azis II ADB INSTITUTE RESEARCH PAPER 35 Additional copies of the paper are available free from the Asian Development Bank Institute, 8t h Floor, Kasumigaseki Building, 3-2-5 Kasumigaseki, Chiyoda-ku, Tokyo 100-6008, Japan. Attention: Publications. Also online at www.adbi.org The Research Paper Series primarily disseminates selected work in progress to facilitate an exchange of ideas within the Institute's constituencies and the wider academic and policy communities. The findings, interpretations, and conclusions are the author's own and are not necessarily endorsed by the Asian D evelopment Bank Institute. They should not be attributed to the Asian Development Bank, its Boards, or any of its member countries. They are published under the responsibility of the Dean of the ADB Institute. The Institute does not guarantee the accuracy or reasonableness of the contents herein and accepts no responsibility whatsoever for any consequences of its use. The term "country", as used in the context of the ADB, refers to a member of the ADB and does not imply any view on the part of the Institute as to sovereignt y or independent status. Names of countries or economies mentioned in this series are chosen by the authors, in the exercise of their academic freedom, and the Institute is in no way responsible for such usage. Copyright © 2002 Asian Development Bank Institute & the author. All rights reserved. Produced by ADBI Publishing. ABOUT THE AUTHOR Prof. Iwan J. Azis of Cornell University and the University of Indonesia is a regular Visitin g Scholar at the ADB Institute. On the topic of the Asian Crisis, in early 1998 he spoke before the Joint Economic Committee (JEC) of the U.S. Congress, and was invited to present his views on the Indonesian case at the IMF meeting in Washington D.C. during the fall of 2000. H e has published on subjects such as ASEAN economies, spatial development, impacts o f economic reform, conflicts resolution, exchange rate and capital flows, reform sequencing, and financial crisis. He has authored or co-authored several books, and is currently working on another book on “Modeling Policy Analysis.” He received his BA from the University o f Indonesia and his MSc and PhD from Cornell University. During 1984-1993 he served as Chairman, Department of Economics, University of Indonesia, and Director of the World Bank-funded Inter-University Center. Prof. Azis is the author of the earlier ADBI paper to this study entitled Modeling Crisis Evolution and Counterfactual Simulations. III PREFACE The ADB Institute aims to explore the most appropriate development paradigms for Asia composed of well-balanced combinations of the roles of markets, institutions, and governments in the post-crisis period. Under this broad research project on development paradigms, the ADB Institute Research Paper Series will contribute to disseminating works-in-progress as a building block of the project and will invite comments and questions. I trust that this series will provoke constructive discussions among policymakers as well as researchers about where Asian economies should go from the last crisis and recovery. The conference version of this paper was presented on 7 December 2001 at the ADB Institute’s Fourth Anniversary on “Poverty Reduction: Quality of Growth, Governance, and Social Development”. (www.adbi.org/povred/pov2001.htm) Masaru Yoshitomi Dean ADB Institute IV ABSTRACT This studya sequel to ADBI paper no.23attempts to establish a link between macroeconomic (financial) shocks and poverty by modeling the detailed and complex mechanisms of how household incomes and prices are determined. The model is of a general equilibrium type with an explicit and detailed financial sector. One of the novel features is that the poverty measures are derived endogenously. Indonesia during 1997-1999 is used as a case. The strong co-existence of economic and political crises not only makes the country most interesting to study, but also forces the model to include a parameter reflecting the political risk, the fluctuation of which is commensurate with the country’s risk premium. There are two major components to the model: (1) the macroeconomic and financial sector; and (2) social indicators such as household incomes, prices, and poverty measures. The macroeconomic part details the relations among macroeconomic variables, e.g., outputs, inputs, general prices including exchange rate, exports, imports, capital flows, interest rates, government budget, and labor market. The general social indicators include unemployment, income distribution, and income poverty. The latter is measured particularly by the headcount ratio, poverty gap, and poverty severity. The main thrust of the study is how to link (1) and (2). Major sources of household incomes are factor incomes, transfers, and returns on assets. The latter is specified according to Tobin’s portfolio model, in which there is no perfect substitutability in the allocation of narrow money, domestic time deposits, foreign time deposits, and equity. The specific allocation is determined by households’ preferences and/or tastes. The benchmark simulations shows that the generated income distribution tends to fluctuate, i.e., worsening towards Stage 6 (May 1998) and Stage 7 (December 1998), and improving towards the end of the simulation period (Stage 8, March 1999). It is revealing that there is a close correlation between worsening (improving) income distribution and the trend of increased (decreased) interest rates. Surely, asset (interest) incomes and windfalls from foreign assets holdings in an environment of super-high interest rates and exchange rate collapse during the crisis have produced a not insignificant effect. As far as poverty impacts are concerned, the main channel of transmission is through endogenous price changes (affecting the poverty line) and household incomes (affecting the level and patterns of consumption). Under the benchmark simulation, the poverty incidence increases faster in urban than in rural areas. In some rural households (i.e., agricultural workers), the head-count ratio actually drops, since many of them are employed in the plantation export sector, which benefited from currency depreciation. Similar trends are also observed for poverty gap and poverty severity. Two counterfactual experiments are conducted, i.e., preventing interest rates from rising persistently, and a combination of such a policy with a partial debt resolution. V The results show that these two alternative policies would have produced lower poverty lines. But the per capita household incomes would have been also lower under the policy mix of less tight and partial debt resolution; they are higher only under the less tight interest rates policy. When the poverty line is lower and the per capita incomes decline, the poverty incidence can change in different directions (indeterminate). In the Indonesian case, however, the results clearly indicate that both counterfactual policies produce lower poverty incidence than in the benchmark case, suggesting that the impacts of prices on poverty are far more significant than the impacts of income changes during the crisis. From this standpoint, the actual policies of removing subsidies at once in order to tighten the budget, and injecting liquidity funds to the banking sector that made the base money surge should have been avoided. The model is also capable of endogenizing poverty gap (P1) and poverty severity (P2). The latter is particularly important since a person that can afford to consume only food that is 1,000 calories short of daily requirements might be 16 times more vulnerable to diseases than a person with a 250 calorie shortfall, not four times as would be the case if the poverty gap measure had been adopted. Should policymakers be concerned with such distributional issues, they must pay more attention to the measure of poverty severity. During the crisis, the increase of poverty severity was higher in urban than in rural areas. This trend is fairly robust, valid for the benchmark as well as the two counterfactual experiments. VI TABLE OF CONTENTS About the Author II Preface III Abstract IV Table of Contents VI 1. Introduction 1 2. Modeling Household Income, Price Determination, and Poverty Module 2 3. The Evolution of Poverty During the Crisis 12 4. Policy Environment 18 5. Model Simulations 22 6. Concluding Remarks 33 References 36 Tables and Figures (in body of text) Table 1. Number of Households and Populations, 1995-1999 11 Table 2. Morbidity by Consumption Quintile 19 Table 3. Impacts of High Interest Rate on GDP, Prices, Employment, Income Distribution , and Poverty 28 Table 4. Endogenous Poverty Measures: Benchmark and Counterfactual Simulations 33 Figure 1. Household Portfolio Allocation Decision 3 Figure 2. Potential Negative Impacts of Exchange Rate Depreciation on Income Distribution 6 Figure 3. Impacts of Higher Interest Rates and Debt Resolution (Counterfactual Policy Scenarios) 9 Figure 4. From Pre-to Post-Crisis Povety: Indonesia 13 Figure 5. Fluctuating Monthly Inflation Rate in 1998 14 Figure 6. Annual Growth of Poverty Line: Urban and Rural 14 Figure 7. Poverty Line by Regions: 1996-1999 15 Figure 8. Annual Growth of Nominal and Real Wages 16 Figure 9. Cumulative Density Function, 1996 and 1999 (at constant 1996 prices) 17 VII Figure 10. Parametric Income Distribution for Indonesian Household Groups 1996 versus 1999 (at constant 1996 prices based on GDP Deflator) 18 Figure 11. Gross Enrollment in Urban and Rural Areas 20 Figure 12. Labor Real Income 23 Figure 13. Household Real Income 24 Figure 14. Income Distribution 25 Figure 15. Macroeconomic and Social Indicators: Simulation Results 29 Figure 16. Income Distribution: Benchmark & Counterfactuals 30 Figure 17. Prices: Benchmark & Counterfactuals 31 Figure 18. Prices for Poverty Line: Benchmark & Counterfactuals 31 Appendix Figure 1a.Agricultural Workers 39 Figure 1b.Farmers with Land 39 Figure 1c.Rural Low 39 Figure 1d.Rural Non-Labor Force 39 Figure 1e.Rural High 40 Figure 1f. Urban Low 40 Figure 1g.Urban Non- Labor Force 40 Figure 1h.Urban High 40 Figure 1i. Rural Groups 41 Figure 1j. Urban Groups 41 Figure 2a.Agricultural Workers 42 Figure 2b.Farmers with Land 42 Figure 2c.Rural Low 42 Figure 2d.Rural Non-Labor Force 42 Figure 2e.Rural High 43 Figure 2f. Urban Low 43 Figure 2g.Urban Non-Labor Force 43 Figure 2h.Urban High 43 Figure 2i. Rural Groups 44 Figure 2j. Urban Groups 44 7 equation. Since in most emerging markets a considerable portion of intermediate inputs are usually imported, the composite intermediate inputs INTM are necessarily modeled as a CES function of domestic and imported inputs (DOMINTM and FORINTM). When necessary, one can alter the elasticity of substitution of some of these inputs. In the second stage, domestic output is specified as a CES function of value-added VA and composite intermediate inputs. The resulting price of value-added PV is: p pppp pVA INTMPINTMXPX PV ×−× = (8) where PINTM is the price of intermediate inputs. The unit price of imported and domestically produced intermediate inputs (PDINTM and PFINTM) are, respectively, ∑×= pp pppppp PDaadPDINTM }{ , (9) ∑×= pp pppppp PMaamPFINTM }{ , (10) where aad and aam are the share parameters, and subscripts p and pp refer to the production sector. Given (9) and (10), the following equation for price of composite intermediate inputs is derived: p pppp pINTM FORINTMPFINTMDOMINTMPDINTM PINTM ×+× = (11) More relevant for poverty measures is the Consumer Price Index (CPI)-related price (PINDEX), which is the aggregate prices of Armington goods, ∑×= ppp PQwtqPINDEX (12) where wtq is the share parameter. To arrive at the prices of basic needs (prices presumably paid by the poor), the trend of any price index to be used should meet the following conditions: (1) differentiated between urban and rural, and (2) linked to the fluctuation of PINDEX. For example, if one uses the average domestic price PD (denoted by PDAVG), the fluctuation of such prices must be adjusted by PINDEX fluctuation. In order to distinguish the rural poverty line prices from the corresponding prices in urban areas, consumption patterns in the two areas have to be taken into account, such that the resulting poverty line prices reflect those actually paid by poor households in urban and rural areas. The different consumption patterns are reflected through the sectoral consumption parameter α pr,u. Hence, the poverty line PL for both areas can be written: ∑××       =p ur p ur PD PDAVG PINDEX PL ,, α (13) Note that all variables in the above prices are derived endogenously, except for the consumption parameter α pr,u. Once the incomes of different household groups and 8 the poverty line prices in urban and rural areas are determined, various poverty measures can be applied. The starting point is to select a basket of Basic Needs (BN) reflecting the consumption pattern of the households around the presumed poverty line and yielding the threshold caloric requirements. Typically, food is by far the most important commodity in this BN basket. If we denote the basket of BN by π com, then the poverty line is essentially Σ com π com . Pcom, where Pcom is the endogenously derived poverty line prices. The estimates of poverty incidence in each socioeconomic group can therefore be generated by using the respective poverty lines derived in equation 13. Having completed income and price specifications, one can capture the impact of macroeconomic financial shocks (e.g., an exchange rate shock) on income distribution and poverty. There are at least two transmission mechanisms. The most direct one is through a decline in nominal incomes or wages, related to collapsed domestic demand (increased numbers of laid-off workers). Another mechanism is through rising prices, especially those of basic commodities, leading to an increase in the monetary poverty line. Since prices are endogenously determined in the model, given a certain basket of Basic Needs made up of food and non-food commodities, a monetary poverty line is, in effect, also derived endogenously (see equation 13). Before arriving at a poverty measure, one has to determine first the intra-group distributions corresponding to the characteristics of each group. One example of such a distribution, e.g., used in Decaluwe et al. (1999), is the Beta distribution function.6 For a given household group, () 2 11 min)(max )(maxmin)( ),( 1 ,; −+ −− − −− ×= qp qp yy qpB qpYHHf (14) where , and max][min, ∈ YHH . Parameters min and max are the minimum and maximum incomes within a household group, respectively, and p and q are parameters that shape the distribution (when p and q are larger than unity, if p>q, p<q, and p=q, the distribution is skewed to the left, skewed to the right, and symmetric, respectively). Alternatively, one can also use the actual (parametric) distribution in each household category. Whichever distribution function is used, the resulting poverty measures such as headcount index and poverty severity can be determined through FGT specification (see below). For socioeconomic group ihh, the following applies: ∫     − = z ihhihhihhihh ihh ihh ihh dYHHqpYHHf PL YHHPL P 0 ),;( α α , (15) 6 The advantage of using such a function is the flexibility it provides in constructing a distribution that corresponds to the unique characteristics of each group. dy yy qpB qp qp ∫−+ −− − −− = max min 2 11 min)(max )(maxmin)( ),( 9 if Beta distribution is used. Alternatively, one can also use the actual distribution that gives: ∫     − = z ihhihh ihh ihh ihh dYHHYHHf PL YHHPL P 0 )( α α , (16) where PL is the poverty line, distinguished between rural and urban, and α is the poverty-aversion parameter. Based on the above formula, one can calculate the headcount index ( 0 P), poverty depth ( 1 P), and poverty severity ( 2 P).7 In this study, I use the actual (non-parametric) distributions. If one were to conduct counterfactual policy experiments, two alternative policies are worth exploring (Azis [2001] simulated precisely these two policy scenarios): preventing interest rates from being raised excessively, and this in combination with partial debt resolution. In the model, the policy-based interest rate is RSBI, which is the rate of the Central Bank’s certificate known as Sertifikat Bank Indonesia (SBI), and the debt service payments to be modified are labeled DEBSERV. The transmission mechanisms in such counterfactual scenarios are shown in Figure 3. Figure 3. Impacts of Higher Interest Rates and Debt Resolution (Counterfactual Policy Scenarios) 7 As is well known, the additively separable nature of the α P class of poverty measures permits one to measure poverty for each household group and then calculate national (social) poverty as the weighted sum of the group levels, ∑ = j j jPpopP αα , where j pop is the share of group j in the national population. DEBSERV RSBI FOREXDEB RISK PFCAPIN PFCAP EXPEXR EXR PINDEX SBI BANKF DOMPINV E D RGDP UNEMP YHH RT POVERTY POLRISK YF Transfers EQROW 10 Along with the (exogenous) rising political risk POLRISK and increased capital outflows EQROW, a surge in debt service DEBSERV would affect the expected and actual exchange rate (EXPEXR and EXR), and in turn raise the price index PINDEX. Following the aforementioned processes to arrive at the price of poverty line, changes in PINDEX will eventually affect poverty indicators (equation 13). On the income side, the household income YHH is affected by both the rising interest rate (through savers’ interest incomes) and the exchange rate depreciation (through surging local currency values of dollar savings). This is in addition to factor incomes and various types of transfers. In a crisis situation, the severity of poverty is usually far more important to observe than simply the headcount index. In this context, I will apply the FGT method for the poverty measures (explained below).8 But like in most SAM-based economywide models, the number of households in the SAM classification is usually limited, making the resulting income distribution less meaningful, since it only depicts the distribution between SAM-listed household categories. Therefore, one ought to measure the intra-category (intra-household) distribution of income to yield a poverty estimate. Once done, a comparison between the pre- and post-crisis intra-category distributions can be made. Such a comparison is subsequently confronted with the endogenously derived poverty line in order to generate the evolution of endogenous poverty measures. Next are the specifications of labor market. A sector’s demand for different labor categories (eight in the model) is derived from the first order condition for firms’ profit maximization. Thus, sectoral labor demand will depend on its product price, wages, and the prices of intermediate inputs. A composite labor demand function for each sector is postulated as a function of the various labor categories. This is the composite labor input, which appears as an argument in the sectoral domestic output functions. In turn, it has been empirically determined over an extended period in the context of Indonesia that sectoral wage rates are strongly influenced by prices of valueadded (PV), labor productivity growth, and the inflation rate. Hence, sectoral wage rates are endogenously derived in the present model (see Thorbecke et al., 1992): p p pflpp vp p p vp pPDL FACDEMX PV PV PINDEXWAGES π           ×         ×= ∑ − 00 , )1( / (17) where FACDEM and PDL0 are, respectively, factor (labor) demand and labor productivity at the initial period. A key implication that underlies the form of the wage equation is the prevalence of labor market segmentation with wages being strongly sector-specific. The average wage rates for each labor category are arrived at on the basis of the sectoral wage rates, WAGESp, and the wage shares of each type of labor in each sector (wsharep,fl): 8 FGT stands for Foster-Greer-Thorbecke. It is a poverty measure that can be used to estimate not only the incidence of poverty but also its severity (see Foster, Greer, and Thorbecke, 1984). Incidentally, because of its advantageous features, FGT has been adopted as the standard poverty measure in developing countries such as Mexico, as stipulated in Chapter V Article 34 of its Constitution. 11 ∑××= pflppflfl wshareWAGESWFWF , 0 (18) In a standard model, the labor supply of each category is usually assumed to be fixed in the base year. In the current model, it is assumed that some labor slack prevails (in the form of unemployment or underemployment), and rural-urban migration factors play a role. In a crisis setting, it is expected that labor would migrate from urban to rural areas (a reverse migration), especially when the urban sector is hardest hit. This is particularly true in Indonesia as the labor market is flexible and most urban dwellers have close ties with their extended families in rural areas. As will be shown subsequently, there is indeed evidence of a major reverse migration. During the crisis, real wages in the rural non-farm sector declined less than in urban activities. This factor, combined with the reverse migration, mitigated partially the potential unemployment consequences of a 14 percent drop in real gross domestic product (GDP) in 1998. The decline in real wages in the farm sector was largely because of the excess supply induced by the urban-rural migration. It is revealing that, largely due to the agricultural sector’s role in absorbing these reverse migrants, even during the crisis the employment rate continued to increase, albeit at a slower pace. The massive urban-rural migration (Table 1) did change the rural-urban composition of the labor supply, causing the spatial unemployment as well as incomes to change. Table 1. Number of Households and Populations, 1995-1999 # Household # Pop # Household # Pop # Household # Pop 1. Agricultural Workers 5,064,667 20,794,316 5,893,304 24,196,504 7,099,082 30,608,337 2. Small Farmers 8,024,174 32,990,982 8,358,655 34,366,184 10,097,924 40,009,288 (land < 0.5 ha) 3. Medium Farmers 3,076,379 13,796,229 3,204,615 14,371,313 2,915,904 13,694,954 (land 0.501 - 1 ha) 4. Large Farmers 2,190,677 10,697,076 2,281,994 11,142,975 2,379,946 10,618,552 (land > 1 ha) 5. Rural Low (Non-Farm) 6,843,656 28,701,887 7,180,472 30,114,475 7,309,818 29,933,080 6. Non Labor Force (Rural) 2,795,633 9,097,513 2,933,223 9,545,255 3,051,457 9,877,266 7. Rural High (Non-Farm) 3,263,466 15,267,947 2,909,464 13,611,768 3,201,555 13,805,324 8. Low Urban 7,708,983 33,835,022 8,418,047 36,947,134 7,386,730 30,856,354 9. Non-Labor Force (Urban) 2,660,015 10,197,213 2,904,680 11,135,142 4,130,884 10,131,141 Household Category 1995 1998 1999 Source: CBS, based on SAM tables. 12 In most standard migration specifications, the Todaro model is normally used, in which labor movements are determined by the growths of earning differentials and employment opportunity. Despite its widespread use, however, such a specification does not necessarily fit well with the actual migration pattern in a country such as Indonesia. In particular, either due to imperfect information or other peculiarities, wage differentials do not always explain the observed labor movements. As shown in Azis (1997), this has indeed been the case in Indonesia. The fact that considerable numbers of people moved from urban to rural areas in 1999 does not seem to match with the trend in wage differentials, e.g., wages in the agricultural sector remained much lower than in urban-related activities, even after the crisis. It is likely that the bulk of the reverse migration consists of temporary migrants who decided to move for reasons other than wage differentials, e.g., loss of jobs, disappearance of income-generating opportunities, and in some cases the flight to safety due to increased crime rates and deteriorating security conditions in urban areas, especially after the riots of May 1998. The latter may have been the more compelling explanation. On the basis of this argument, I model the migration by making use of the changes in labor demand, DFL, to represent labor opportunity, as the explanatory variable: }1 0/ 0/ {0 −         ××= L xx yy DFLDFL DFLDFL LSMIG τ τ (19) where DFL0 is the labor demand at the initial period. As shown in the above equation, the labor demand probability is measured by the growth ratio of labor demand in category “y” to labor demand in category “x,” where “y” is the expected migrationdestination category and “x” is the expected migration-origin category. The model specified above is used to simulate the benchmark (actual) scenario and some counterfactual experiments. Before discussing the results of model simulations, let me first discuss some poverty trends and related policies in Indonesia. 3. Evolution of Poverty During the Crisis There have been several studies attempting to produce consistent estimates of poverty in Indonesia. A methodologically consistent measure implies that the poverty basket is calculated using the same procedure each time, whereas a welfare consistent approach means that an individual is at the same material standard of living in any two periods. By comparing poverty measures based on the two approaches, Suryahadi et al. (2000) claim that the welfare-consistent approach is preferable. Figure 4 shows the comparative trends of poverty in Indonesia using official numbers and welfare-consistent estimates.9 Although the size of poverty incidence at any time is different for the two estimates, the trend is similar, i.e., rising poverty from February 1997 (9.4 percent) to February 1998 (14.8 percent), peaking in December 1998 (17.9 percent), before declining in February 1999 (16.6 percent). 9 I do not include the methodologically-consistent estimates in Figure 4. It is important to note that, while the welfare-consistent estimates may be preferred because the price index share being used represents the actual consumption pattern of (some of) the poor, as argued by Suryahadi et al. (2000), the fact that it ignores the substitution effects still tends to result in an overestimation of poverty incidence. 13 0 2 4 6 8 10 12 14 16 18 Percent Figure 4. From Pre- to Post-Crisis Poverty: Indonesia Official (Actual) Consistent Est. Feb 96 Feb 97 Feb 98 Dec 98 Feb 99 Comparing data collected during different periods of the survey is not valid. Arguably, therefore, one should use a consistent time (month) of the year. This is the reason why February is consistently used in Figure 4. The number for December 1998 is presented in the figure only to indicate the peak poverty rate.10 A dramatic surge in inflation, especially if the food component has the largest weight in the bundle, can raise the poverty line significantly. This holds true even if there is no decline, or there is a nominal increase, in consumption expenditures. After enjoying a long period of single-digit inflation, Indonesia’s CPI jumped by 78 percent in 1998. More important, as shown in Figure 5, the rate fluctuated sharply. The highest monthly rate was recorded during June-August. Comparing the composition of the official poverty line and the components of inflation, food has indeed the largest weight, and its inflation was continuously highest among all components during August- September 1998. The tragedy of May 1998 that led to the downfall of Suharto caused prices of many basic goods to go up sharply. This raised the poverty line (in current prices) significantly, i.e., its annual growth during 1993-1996 and 1996-1998 jumped from 16 to 41 percent in urban areas, and from 13 to 39 percent in rural areas (see Figure 6). Between 1998 and 1999, the overall poverty line changed slightly (the increase was due to a small upward trend in rural areas). This is also confirmed by Figure 7, showing the evolution of the poverty line across subnational regions. 10 It is important to note, however, that the December 1998 data were obtained from the 100 villages survey (mini SUSENAS), suggesting that they are not exactly comparable to other poverty figures. Notes: 1996 & 1999: CBS, Susenas; 1997 & Feb 1998: Gardiner, Susenas Core; 1998: CBS, Mini Susenas. 14                      Figure 5. Fluctuating Monthly Inflation Rate in 1998 -5% 0% 5% 10% 15% 20% Dec 97-Mar 98 Apr May June Jul Aug Sept Oct Nov Dec Food PrepFood, B, T Housing Clothing Health Educ, Rec, Sp Trasp, Comm   General (monthly average) Food General 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 1993-96 1996-99 1993-96 1996-99 Urban Rural Source: LPEM-UI, "Menghitung Kembali Tingkat kemiskinan di Indonesia, 1990-1999 ," final report 2000 Figure 6. Annual Growth of Poverty Line: Urban and Rural 15 0 20000 40000 60000 80000 100000 120000 Java-Bali Sumatera Outer Islands Java-Bali Sumatera Outer Islands Figure 7. Poverty Line By Regions: 1996-1999 1996 1998 1999 Urban Rural The surge of inflation (78 percent) and poverty line (more than 40 percent) would have been enough to increase the poverty incidence in 1998, even with rising nominal income and consumption. In terms of wage income, nominal wages increased by 17 percent during 1997-1998, but real wages in both tradable and non-tradable sectors plummeted by 34 percent. The largest drop occurred in the manufacturing sector (more than 38 percent, see Figure 8). Combined with the fact that the change in employment remained positive even after the crisis (growing by 2.7 percent in 1997- 1998), and the unemployment rate increased by “only” less than 1 percentage point (around 0.8 percent according to the Labor Force Survey, Sakernas), this suggests that there has been a fairly high degree of flexibility in the labor markets, something that was not entirely expected by most observers, given the country’s stage of development and industrialization.11 In terms of consumption, the growth of nominal consumption of the lowest two quintiles was as high as 115-120 percent, but in real terms it dropped 6-9 percent. The increase in the nominal consumption of the middle and upper income groups (the remaining three quintiles) was lower, ranging from 102 to 110 percent, and their real consumption also declined more sharply, i.e., between 11 and 14 percent. In turn, real consumption of the top quintile fell by an impressive 24 percent. This has prompted the well-known conclusion that the hardest hit group during the crisis was the country’s urban middle class, most of which are on the main island of Java (Azis, 1998 and 2000b). 11 The positive growth of employment is almost entirely due to the increase of employment in the agricultural sector. For all other sectors, employment has actually declined. Meanwhile, the increase in unemployment rate (0.8 percent) is clearly lower than that in Thailand and the Republic of Korea, i.e., from 2.3 to 4.8 percent, and from 2.6 to 6.8 percent, respectively (World Bank, 2000). Figure 7. Poverty Line by Regions: 1996-1999 16 -40 -30 -20 -10 0 10 20 30 40 Nominal Real Nominal Real Nominal Real Nominal Real Figure 8. Annual Growth of Nominal and Real Wages 1990-1997 1997-1998 Agriculture Manufacturing Services Total This is also consistent with the finding that, although all FGT poverty indicators (particularly P2) were significantly higher in rural than in urban areas, these indicators increased significantly more in the latter during the crisis. The amount of resources needed to alleviate poverty, as estimated through the poverty gap measure P1, would also be larger. This is consistent with the greater downward trend of real wages in essentially urban activities (manufacturing and services) compared to agriculture, as observed in Figure 8, and the trend of reverse migration discussed earlier. At the same time, the facts that a large number of rice workers reside in Java and the decline of real wages in this region was sharper than in non-Java (Papanek and Handoko, 1999) suggest that poverty conditions in Java must have deteriorated relatively more.12 Unlike farmers in export-oriented agricultural products, the sharp depreciation of the rupiah created compounded difficulties for rice farmers who depend heavily on imported vital inputs such as quality seeds and fertilizer. This prompted a doubling of rice prices in 1998. Although the incidence of poverty might have gone up relatively less in many regions outside Java, especially in the eastern part of the country, e.g., East Timor, Irian Jaya, Maluku, and East Nusa Tenggara, the actual depth and severity of poverty in these regions have been much greater than in Java. Another important explanation for a sharp increase in poverty is the large concentration of population whose income is just marginally above the poverty line (the “near poor”). This is particularly true in Indonesia. At the onset of the crisis, the situation was such that with only a 20 percent increase in the poverty line, the number of poor would easily double (Azis, 1998). This re-emphasizes the critical role of poverty 12 Indeed, the FGT measure of poverty severity from 1996 to 1998 shows that P2 in Java’s rural areas increased considerably, i.e., from lower to above unity, except in West Java. But even in the latter, the increase was significant, i.e., from 0.26 to 0.66. Changes in poverty severity in Java’s urban areas were even more dramatic, e.g., in Central Java and Yogyakarta, P2 went up from between 0.4 and 0.5 to 2.4 (see Irawan and Romdiati, 2000). 23 this could potentially raise the per capita incomes of rural-based (e.g., “Agemp”) and reduce the incomes of urban-based households (e.g., “Urbanlow”). In other cases, the reverse may be true. As shown in Figure 13, the model simulation suggests that virtually all categories suffer from declining real incomes. Combined with the sharp rise in the poverty price index, this made a major contribution to raising the poverty incidence. Figure 12. Labor Real Income 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 Benchmark Jul-97 Aug-97 Sep-97 Nov-97 Jan-98 May-98 Dec-98 May-99 Index Ag-Paid Ag-Unpaid Man-Rur Man-urb Clerk-Rur Clerk-Urb Prof-Rur Prof-Urb clerical urban (“Clerk-Urb”) categories to agricultural workers (“Agemp”) will bring their labor incomes to their new (rural) destination by adding the household income of the “Agemp” category with per-labor income of “Man-Urb” and “Clerk-Urb” times the number of migrants from these two labor categories to the paid agriculture workers (“Ag-Paid”). To arrive at the per capita household income, the aforementioned income is divided by the number of population in “Agemp,” including the additional number due to natural growth and migration. Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Ag-Paid Ag-Unpaid Man-Urb Man-Urb Clerk-Rur Clerk-Urb Prof-Rur Prof-Urb 24 Figure 13. Household Real Income 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 Benchmark 35612 35643 35674 35735 35796 35916 36130 36220 Index Lfarm Mfarm Ruralhi Sfarm Agemp Urbanhi Urbanlow Rurallo Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 The dynamics of household real income are important to observe since not all incomes are derived from wage earnings. Various forms of transfers are received by low-income groups during the crisis, either through the Government’s social safety net and anti-poverty programs, or prompted by a mutual-help process (e.g., gotong royong, which is an important institution among rural communities). But from the perspectives of model specification, the most important additional source of earnings is the interest income received by savers, who expectedly belong to the “Urban High” group. Their incomes rise along with an increased interest rate and depreciated exchange rate. The relatively better position of this group at an early stage of the crisis worsens the overall inequality (Figure 14).23 Indeed, published data on income distribution also points in this direction. In the subsequent stages, inequality improves. Interestingly, increased (reduced) inequality occurs when the interest rates move upward (downward). The income distribution worsens in Stage 6 (May 1998), when the interest rates are sharply raised in response to massive pressures on the rupiah. At a later stage (Stage 8), the inequality index declines again as the interest rates begin to drop.24 Hence, there appears to be a fluctuation in inequality, consistent with the Gini coefficient calculated from the SAM 1999. 23 The index denotes the income ratio of high-income groups (“FarmLargeLand,” “Rural High,” and “Urban High”) and lower-income households (“FarmWorkers,” “FarmSmallLand,” and “Rural Low”). 24 The simulation sets the interest rates on SBI to increase dramatically from Stage 5 (January 1998) to Stage 6 (May 1998), and to decline from Stage 7 (December 1998) to Stage 8 (March 1999). This pattern follows the actual trend of SBI rates. 25 Figure 14. Income Distribution 1 1.05 1.1 1.15 1.2 1.25 Benchmark Jul-97 Aug-97 Sep-97 Nov-97 Jan-98 May-98 Dec-98 Mar-99 Index Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Results from the simulation also indicate that the unemployment rate increases considerably. Yet, the allowance of wage decline and labor mobility (from urban to rural areas, and from the formal to informal sector), which is a prominent sign of a flexible labor market, prevented an even bigger catastrophe from occurring. While during the crisis’ peak the unemployment rate may have increased significantly, towards the end of the simulation the recorded unemployment rate shows only a slight increase from the pre-crisis level (1.79%, see Table 3 shown later). Indeed, from the recorded data, the increase in unemployment is surprisingly small, i.e., less than 1 percentage point, according to Sakernas data.25 Nonetheless, the combined forces of unemployment, declining real wages (incomes), and a surging poverty price could potentially raise the poverty incidence. There is, however, some evidence of consumption smoothing that could lead to a poverty incidence lower than originally predicted.26 Also, as the poverty price and the CPI dropped in the later stage (Stage 8), the poverty line should have declined as well. The detailed results of poverty measures are discussed below. As stated earlier, information about intra-group distribution is needed before arriving at poverty measures. In this particular instance, I use a parametric measure of distribution to estimate the poverty incidence, in which the intra-group distribution is directly generated from the SUSENAS-core data with a 206,597 sample size.27 25 Note that the quoted unemployment figures are not exactly comparable, since the figures under the “End of Simulation” column in Table 3 refer only to Stage 8 (roughly March 1999) of the simulation. 26 Many households have either changed their food menu (e.g., eating rice once a day, using other less desirable foods the rest of the time), switched to lower priced food (e.g., from imported to domestic produce), or used their accumulated savings to purchase food (dis-saving). There is widespread evidence showing that a smoothing process also takes place in non-food consumption. But the impact on poverty, more particularly on diets, is less serious compared to the case when the smoothing is in food consumption (especially among the poor). It is also important to note that the economic crisis was not the only culprit. During 1997/98, Indonesia also suffered from crop failures due to the fickle global weather (El-Niño) and a massive haze problem. Subsistence farming areas were the worst affected. 27 Note that since the SUSENAS-core does not distinguish between farmers according to different land sizes (small, medium, and large land owners are lumped together), we have to use only six, instead of eight, household categories in the analysis: four in the rural areas, i.e., agricultural employee (“Agemp”), (cont.) Benchmark Jul-97 Aug-97 Sep-97 Nov-97 Jan-98 May-98 Dec-98 Mar-99 26 Based on the limited number of economic sectors in SAM, in the model simulation the BN basket is limited to only four commodities, i.e., food (rice), other food, textiles, and social services. After approximating the consumption share of each of these goods in the base year’s BN consumption for both urban and rural areas (the share of food in the rural BN basket is higher than in the urban BN basket; the share parameters are denoted by α pr,u in equation 13), I apply equation 13 to generate the price of poverty line PL. The Indonesian statistical office (BPS) provides urban and rural poverty lines based on two extensive baskets of commodities and services (comprising more than 50 items) consumed by the poor. I adopted these BPS urban and rural poverty lines for the base year 1996 (rather than the more limited SAM-based BN basket) and multiplied it by the above derived price indices to obtain the 1999 (post-crisis) poverty lines that are used in the simulation experiments. Recall from earlier discussions that the poverty line is essentially Σ com π com . Pcom, where π com is a basket of quantities of commodities reflecting basic needs. In the above calculation, the πcom used to obtain the price deflator consisted of only four commodities, whereas the πcom used to derive the actual poverty line in 1996 is based on more than 50 items. With the above specifications, the fluctuations of monetary poverty lines in urban and rural areas are commensurate with the movements in the poverty line prices endogenously determined within the model. Subsequently, the changing poverty incidence in each of the socioeconomic household groups can be estimated by applying the respective poverty lines. As indicated in Section 3, based on SUSENAS core data in 1996 and 1999, the intra-group distributions in both years for each group appear quite similar, suggesting that the distribution in each of the household categories can be assumed to be unchanged. Each group distribution is then adjusted by the change in the mean income of that household group generated in the simulation. Hence, if the mean income of a specific group were to rise by 100 percent in nominal terms, each individual’s income within the same group is assumed to increase, likewise, by 100 percent. Therefore, both the poverty line and the household income are derived endogenously in the model. Table 3 shows the resulting poverty trend (FGT measures) along with the income distribution index, unemployment rates, and selected macroeconomic indicators. Note that the “End of Simulation” column refers to Stage 8 of the simulation, representing roughly March 1999 (see Azis, 2001). It is clear that as the crisis evolves, the extent of the deterioration of poverty conditions is larger in urban than in rural areas.28 While the head-count poverty among farmers owning land increases from 21 to 26 percent, and among rural non-farm farmers with land (“Farm”), non-farm rural low income household (“Rurallo”), non-farm rural high income household (“Ruralhi”), and two in the urban areas, i.e., low income urban household (“Urbanlo”), and high income urban household (“Urbanhi”). Also, it is important to note that the SUSENAS survey reflects the conditions at around March of the respective year. 28 Note that the headcount ratio (α=0) increases from 14.85 to 19.65 percent, while the actual data based on SUSENAS point to 9.4 and 14.8 percent in February 1997 and February 1998, respectively, and 16.6 percent in February 1999 (see again Figure 4). While the trend of the simulation results is fairly comparable with these data, the discrepancy is due to the fact that I use the core SUSENAS with 200,000- plus sample data, while the official data shown in Figure 4 are based on the SUSENAS module with only 60,000-plus sample data. 27 households the increase is from 11 to 20 percent (for low income rural) and from 9 to 12 percent (for high income rural), the poverty incidence within the category of agricultural workers drops from 18 to 14 percent. The latter can be explained by the fact that most agricultural workers are employed in the plantation export sector, which benefited from the currency depreciation. In contrast, in the two categories of urban households, the head-count poverty incidence increases, most dramatically among the urban low income, i.e., from 16 to 25 percent. Similar trends are observed for the poverty gap and the severity of poverty indicators. Overall, urban poverty appeared to rise from 12 percent in 1996 to 19 percent in 1999, and rural poverty from 16 to 20 percent. All these trends occur as the macroeconomic conditions deteriorate along with the imposition of a high interest rate policy, i.e., real GDP drops, the exchange rate collapses (by as much as 240 percent) and the country suffers from hyperinflation. Compared to other Asian crisis countries, the political factor played a much more compelling role in Indonesia. The political uncertainty and deteriorating economic conditions sparked riots, causing serious disruptions in the distribution of major food commodities. This, along with a tremendous amount of funds injected by the monetary authority to prevent the banking sector from collapsing (causing the base money to surge, despite high interest rates), resulted in hyperinflation in 1998. The CPI increased by more than 70 percent, close to what the model estimates, i.e., 74.96 percent (see Table 3). The rising poverty line price index (more-than 70 percent) eventually caused the poverty incidence to increase. When available resources needed to help alleviate poverty are limited, as is usually the case following a major financial crisis, policymakers should adopt a new priority. In this context, the FGT measures could provide a useful policy guide. The direct policy implications of the P2 ( α = 2) is that it is a distributionally sensitive measure. P0 (head-count measure) is totally insensitive to distribution, although this is easier to calculate (and hence more widely used). P1, known as poverty gap, is a measure of shortfall from the poverty line. In P0, a person who is Rp1 below the poverty line is counted exactly the same way as someone who is Rp1,000 below the poverty line. In contrast, under P1, each poor person contributes to overall poverty according to their distance (shortfall) from the poverty line. 28 Table 3. Impacts of High Interest Rate on GDP, Prices, Employment, Income Distribution, and Poverty Baseline End of Simulation Change Interest Rate (RSBI) 14.50% 25.00% 10.50% Real GDP 520615 491341 -5.62% Exhange Rate 2249 5523 245.58% Price Index 1 1.7496 74.96% Poverty Line Price 1 1.7065 70.65% Income Distribution 1.2081 1.4525 20.23% Unemployment Rate 7.20% 8.99% 1.79% Employment 72486 72215 -0.37% Poverty Measures 1. Head-count ( α =0) Agemp 0.1783 0.1417 -3.66% Farm 0.2096 0.2569 4.73% Rurallo 0.1136 0.1981 8.45% Ruralhi 0.0885 0.1212 3.27% RURAL 0.1612 0.2006 3.94% Urbanlo 0.1588 0.2520 9.32% Urbanhi 0.0803 0.1156 3.53% URBAN 0.1222 0.1863 6.41% TOTAL 0.1485 0.1965 4.80% 2. Poverty Gap ( α =1) Agemp 0.0299 0.0227 -0.72% Farm 0.0421 0.0541 1.20% Rurallo 0.0183 0.0362 1.79% Ruralhi 0.0151 0.0219 0.68% RURAL 0.0300 0.0392 0.93% Urbanlo 0.0308 0.0561 2.53% Urbanhi 0.0149 0.0232 0.83% URBAN 0.0234 0.0403 1.69% TOTAL 0.0278 0.0395 1.17% 3. Poverty Severity ( α =2) Agemp 0.0078 0.0057 -0.21% Farm 0.0132 0.0174 0.42% Rurallo 0.0048 0.0102 0.54% Ruralhi 0.0042 0.0063 0.21% RURAL 0.0088 0.0119 0.31% Urbanlo 0.0091 0.0183 0.92% Urbanhi 0.0044 0.0071 0.27% URBAN 0.0069 0.0129 0.60% TOTAL 0.0082 0.0122 0.40% The policy implication is that the poverty gap is the total amount of funds needed to eliminate poverty if perfect targeting were possible (each poor person would receive subsidies equal to their shortfalls). But a case can be made that the poor should be given a weight greater than simply their shortfalls. P2 gives them a weight equal to the square of these shortfalls. Hence, a person that can afford to consume only food that is 1,000 calories short of daily requirements might be four times more vulnerable to diseases than a person with a 500 calorie shortfall (not twice as would be the case if P1 had been adopted). 29 A policymaker who is concerned about achieving a more equal income distribution should adopt P2 as his/her poverty measure to be minimized.29 At the other extreme, a policymaker who is totally indifferent to distributional objectives would use P0. The poverty conditions worsen more in urban than in rural areas, i.e., 6.41 versus 3.94 percentage point, 1.69 versus 0.93, and .60 versus 0.31 according to, respectively, head-count ratio, poverty gap, and poverty severity. Given the limited resources available for poverty alleviation following the crisis, and the urgency to mitigate the impact upon the severity of poverty, the results shown in Table 3 should be of value for, say, targeting purposes. From the analysis of the poverty trend, therefore, one can surmise that the crisis has hit urban households more negatively than rural households. This result is the combined outcome of two forces, a lower increase in households’ nominal income and a faster increase in poverty line prices in urban areas than in rural areas. Figure 15 (summarizing Table 3) clearly exemplifies that the macroeconomic policy response during the crisis could not prevent macroeconomic conditions from deteriorating and produced worsening social conditions. Despite the tightening policy (high interest rates), the exchange rate collapsed. The economy fell into recession (falling GDP), although the resulting increase in unemployment was relatively small due to the flexibility of the country’s labor market. While the relative income distribution particularly between urban and rural areas could potentially improve due to more severe impacts on urban areas, the high interest rate policy and continued 29 For example, the Mexican Constitution mandates that health, education, and other welfare public expenditures among regions be allocated so as to minimize P2. Figure 15. Macroeconomic and Social Indicators: Simulation Results 0 1 2 3 4 5 6 Baseline End of Simulation Index RSBI*10 Real GDP/100.000 Exhange Rate/1000 Poverty Line Price Income Dist Index Unemployment Rate*10 Headcount Poverty/10 Poverty Gap Poverty Severity Baseline End of Simulation 30 depreciation of the rupiah disrupted the process by creating a windfall for large savers (mostly urban), causing the income distribution to worsen. In Azis (2001), I have shown the macroeconomic results of counterfactual tests by experimenting with two sets of policy response to the crisis, i.e., less-tight monetary policy, labeled “Less tight” and a combination of such a policy with partial debt resolution (labeled “Less tight & debt”). All these alternative policies are imposed at the time of the IMF arrival in October/November 1997 (Stage 4). The results of these alternative policies are subsequently contrasted with the results discussed above, which are based on the estimated actual trend, hereafter “Benchmark (IMF).” The generated income distribution from the exercise is rather interesting. Up to Stage 5, all scenarios produce an improvement, but in the subsequent stages, there is clearly a worsening trend of inequality (Figure 16). The resulting difference between “Less tight” and “Less tight & debt” is small and almost overlapping on the figure, but still favoring the latter. It is revealing that there is a close correlation between the worsening (improving) trend of income distribution and the increasing (lowering) interest rate. Surely, the interest incomes and the windfall from foreign asset holdings (e.g., time deposits in foreign currency) in an environment of super-high interest rates and a severe collapse of the exchange rate have contributed to such a relation. ∗ N.B. Lines almost overlapping Nevertheless, the relative income distribution under the “Benchmark (IMF)” is clearly the least preferred one; the outcome is persistently worst among the three scenarios.30 30 Note that the inequality measure used in this study is derived from the mean incomes of the SAM-based eight household groups. Each intra-group distribution is assumed to remain the same as in the base (1995/6) period. Yet, in the present context, this is a defensible assumption, as a comparison of the actual pre- and post-crisis group distributions in Appendix Figs.2a-2i suggest. Figure 16. Income Distribution: Benchmark & Counterfactuals 0.98 1.03 1.08 1.13 1.18 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Index Benchmark (IMF) Less tight Less tight & debt Nov 97 Jan 98 May 98 Dec 98 Mar 99 B ∗ ∗ ∗ 31 The numerical effects of the scenarios on prices show that up to Stage 7 the “Benchmark (IMF)” produces the highest price index. The gap is largest in Stage 6, though prices are converging in the remaining stages. In fact, in Stage 8 both the price and the poverty line price levels under the “Benchmark (IMF)” are slightly lower than under the “Less tight” scenario (Figures 17 and 18). Nonetheless, for the entire period, the poverty line price level is still highest under the “Benchmark (IMF)” experiment. Figure 17. Prices: Benchmark & Counterfactuals 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Index Benchmark (IMF) Less tight Less tight & debt Nov 97 Jan 98 May 98 Dec 98 Mar 99 Figure 18. Prices for Poverty Line: Benchmark & Counterfactuals 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Index Benchmark (IMF) Less tight Less tight & debt Nov 97 May 98Jan 98 Dec 98 Mar 99 32 With such a trend of poverty line prices, the FGT poverty measures can be derived endogenously. Table 4 shows the results. The headcount poverty measure clearly indicates that the “Less tight” scenario produces lower poverty incidence, 15.4 compared to 19.6 percent. In all household categories, the poverty incidence is lower. Comparing the “Less tight & debt” with “Less tight,” however, yields an interesting outcome. While a joint-policy of avoiding high interest rates and resolving some of the debts produces a slightly higher poverty incidence for “Farm” and “Ruralhi,” causing the increase of rural poverty to be higher, for all other categories the policy produces a lower headcount ratio. In urban areas, for example, the ratios for both types of household, i.e., “Urbanlo” and “Urbanhi” are smaller, i.e., 20.17 versus 20.30 percent, and 9.33 versus 9.40 percent, respectively. Overall, the headcount ratios under the two counterfactual scenarios are almost the same; under “Less tight & debt,” the ratio would have been only slightly lower. A different pattern is observed in terms of poverty gap (α=1) and poverty severity (α=2). Due to the greater weight of rural poverty for these two measures, the “Less tight & debt” scenario produces greater increases in the overall (total) poverty gap and severity compared to what the “Less tight” experiment yields. Although the difference in the ratios does not seem to be large, and they remain higher for urban compared to rural areas, the resource allocation for poverty alleviation could be slightly different between the two. Relatively speaking, had policymakers followed the FGT formula, the amount of resources allocated to rural areas should have been larger in “Less tight & debt” than in “Less tight,” although the amount would have still been smaller compared to those allocated to urban areas. If policymakers are also concerned with the issues of distribution among the poor, they should be aware of the fact that the increase in the rural poverty severity in “Less tight” is less than in the “Less tight & debt” scenario. On the other hand, the increase of poverty severity in urban areas under the former is larger than that under the latter. It is therefore clear that not only the macroeconomic indicators but also the poverty conditions would have been better if Indonesia had not stuck too rigidly with the IMF-style policy (for more detailed macroeconomic impacts, see Azis [2001 and 2002]; for broad discussions on the policy response to the crisis, see Hill [1999]). These counterfactual experiments demonstrate that the policy scenario of not raising interest rates would have created potential benefits in terms of poverty and income distribution. Further, the combination of non-high interest rate and partial debt resolution at the early stages of the crisis appear to be the most preferred scenario from the perspective of these social indicators. 39 Appendix Cumulative Density Functions for Different Household Groups, 1996 and 1999 at Constant 1996 Prices Figure 1a: Agricultural Workers Figure 1b: Farmers with Land 00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8 2 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Farmers w/Land 96 99 Figure 1c: Rural Low Figure 1d: Rural Non-Labor Force 00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8 2 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Rural Non-LbF 96 99 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 x 105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Agricultural Workers 96 99 0246810 12 14 16 18 x 105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Rural Low 96 99 40 Figure 1e: Rural High Figure 1f: Urban Low Figure 1g: Urban Non-Labor Force Figure 1h: Urban High 00.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 33.3 3.6 3.9 4.2 4.5 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Urban High 96 99 00.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Rural High 96 99 00.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Urban Non-LbF 96 99 00.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 33.3 3.6 3.9 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Urban Low 96 99 41 Figure 1i: Rural Groups Figure 1j: Urban Groups 00.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 33.3 3.6 3.9 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Urban Groups 96 99 00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8 2 x 106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Income (Rupiah) Population Index Rural Groups 96 99 42 Intra-Group Distributions, 1996 and 1999 at Constant 1996 Prices Note that the inequality measure used in this study is derived from the mean incomes of the SAM-based eight household groups. Each intra-group distribution is assumed to remain the same as in the base (1995/96) period. Yet, in the present context, this is a defensible assumption, as a comparison of the actual pre- and post-crisis group distributions in Figs.2a-2i suggest. For each pair, the top figure is for 1996 and the bottom figure is for 1999. Figure 2a: Agricultural Workers Figure 2b: Farmers with Land Figure 2c: Rural Low Figure 2d: Rural Non-Labor Force 2 3 4 5 6 7 8 9 10 11 x 10 5 0 50 100 Po p ulation ( Person ) Agricultural Workers 2 3 4 5 6 7 8 9 10 11 x 10 5 0 50 100 Po p ulation ( Person ) Expenditure per Person (Rp) 2 4 6 8 10 12 14 16 18 x 10 5 0 200 400 600 800 Population (Person) Farmers w/Land 2 4 6 8 10 12 14 16 18 x 10 5 0 200 400 600 800 Population (Person) Expenditure per Person (Rp) 2 4 6 8 10 12 14 16 x 10 5 0 50 100 150 200 250 Population (Person) Rural Low 2 4 6 8 10 12 14 16 x 10 5 0 50 100 150 200 250 Population (Person) Expenditure per Person (Rp) 2 4 6 8 10 12 14 16 x 10 5 0 50 100 Population (Person) Rural Non-LbF 2 4 6 8 10 12 14 16 x 10 5 0 50 100 Po pulation (Person) Expenditure per Person (Rp) 43 Figure 2e: Rural High Figure 2f: Urban Low Figure 2g: Urban Non-Labor Force Figure 2h: Urban High 2 4 6 8 10 12 14 16 x 10 5 0 50 100 Po p ul ati on ( Person ) Rural High 2 4 6 8 10 12 14 16 x 10 5 0 50 100 Po p ul ati on ( Person ) Expenditure per Person (Rp) 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 6 0 50 100 Po p ulation ( Person ) Urban Non-LbF 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 6 0 50 100 Po p ulation ( Person ) Expenditure per Person (Rp) 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 6 0 50 100 150 200 250 Population (Pe rson) Urban High 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 x 10 6 0 50 100 150 200 250 Population (Person) Expenditure per Person (Rp) 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 x 10 6 0 50 100 150 200 Population (Pe rson) Urban Low 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 x 10 6 0 50 100 150 200 Population (Person) Expenditure per Person (Rp) 44 Figure 2i: Rural Groups Figure 2j: Urban Groups 2 4 6 8 10 12 14 16 x 105 0 500 1000 1500 Population (Person) Rural Groups 2 4 6 8 10 12 14 16 x 105 0 500 1000 1500 Population (Person) Expenditure per Person (Rp) 2 4 6 8 10 12 14 16 18 x 105 0 200 400 600 Population (Person) Urban Groups 2 4 6 8 10 12 14 16 18 x 105 0 200 400 600 Population (Person) Expenditure per Person (Rp) HOW TO CONTACT US? Asian Development Bank Institute Kasumigaseki Building 8F 3-2-5 Kasumigaseki, Chiyoda-ku, Tokyo 100-6008 Japan Tel: +81 (03) 3593-5500 Fax: +81 (03) 3593-5571 E-mail: [email protected]g www.adbi.org Papers are also available online at the ADBI Internet site: http://www.adbi.org/publications/ RESEARCH PAPER SERIES Family-Based Business Groups: Degeneration of Quasi-Internal Organizations  and Internal Markets in Korea December 2001 Code: 28-2001 by Sang-Woo Nam Can “Moral Hazard” Explain the Asian Crises? December 2001 Code: 29-2001 by Luiz A. Pereira da Silva and Masaru Yoshitomi Avoiding Double Mismatches and Withstanding Regional Financial Crises: The Singapore Experience December 2001 Code: 30-2001 by Khee-Giap Tan, T. Karigane, and M. Yoshitomi The Political Ecology of Famine: The North Korean Catastrophe and Its Lessons January 2002 Code: 31-2002 by Meredith Woo-Cumings Road from State to Market–Assessing the Gradual Approach to Banking Sector Reforms in India– February 2002 Code: 32-2002 by Sayuri Shirai Growth and Poverty: Lessons from the East Asian Miracle Revisited February 2002 Code: 33-2002 by M. G. Quibria Did East-Asian Developing Economies Lose Export Competitiveness in the Pre-Crisis 1990s?  Assessing East-Asian Export Performance from 1980 to 1996 March 2002 Code: 34-2002 A New Approach to Modeling the Impacts of Financial Crises on Income Distribution and Poverty March 2002 Code: 35-2002 by Iwan J. Azis ADB INSTITUTE RESEARCH PAPER 35