Challenges of Global Change for Agricultural Development and World Food Security

Transcription

1 Food Security Program in the Time of Economic Crisis: a lesson to learn from Indonesia Evita Hanie Pangaribowo evita.pangaribowo@uni-bonn.de Department of Economic and Technological Change Center for Development Research University of Bonn, Germany 2012 Paper on the occasion of the Abstract from the program. 52 nd GEWISOLA Annual Conference Keywords: Impact evaluation, food security, Indonesia Challenges of Global Change for Agricultural Development and World Food Security Copyright 2011 by Evita Hanie Pangaribowo. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. Universität Hohenheim, September26 28, 2012 Copyright 2012 by Evita Hanie Pangaribowo. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

2 Food Security Program in the Time of Economic Crisis: a lesson to learn from Indonesia Abstract This study evaluates the impact of food security program an almost universal program of Indonesian Social Safety Net Program in the time of economic crisis. Food security program aimed to protect poor households from the negative effects of economic crisis by means of highly subsidized rice. To assess the impact of the program, this study utilizes matching estimator approach combined with difference in difference method. The rich longitudinal dataset used in this study enables matching estimator and difference in difference approach to provide accurate estimate of the program s impact on its beneficiaries. Results indicate the positive impact of the food security program on the expenditures of richer nutrient food which include meat, fish and dairy products. The program has also substantial impact on health expenditure. In addition, this study finds fly-paper effects where food security program has increased rice and staple food expenditures in the short period. Aid-fungibility is also evident since the program has positive effect on adult goods expenditure. The implication is that improved targeting or applying conditionality might mitigate the unintended effects resulting from the program. Keywords: Impact evaluation, food security, Indonesia 1. Introduction Indonesia had experienced considerable variation in its economic performance and recorded as one of the new emerging economies in South-East Asia. Until mid 1990s, Indonesia recorded around seven per cent of annual income per capita growth (ADB, 2007). However, the economic crisis in the late of 1997 dampened the national economy resulting to high poverty rate and deteriorating living standard. In order to protect Indonesian households from the economic crisis, Indonesian government launched social safety net (SSN) programs. SSN consisted from five major programs: food security (Operasi Pasar Khusus OPK), employment creation (Padat Karya), education scholarship, health card, and community empowerment. Food security program became the main sector in the SSN package. The purpose of this program was to ensure that the poor households were able to access basic food at an affordable price (Sumarto, 2006). The program provided a highly subsidized rice price. In macro context, several studies mentioned that food security program contributed to poverty reduction through reducing poverty gap (Tabor and Sawit, 2005). This study aims to measure the impact of food security program. Food security program was chosen since it was an almost universal program and had largest coverage compared to other SSN programs. Food security program also absorbed sizeable share from government budget compared to the rest of SSN programs. The impact evaluation conducted on this study assesses the impact of rice for the poor program at the micro level. So far, evaluations are conducted at the aggregate level and limited to the program implementation (Hastuti, 2008; Tabor and Sawit, 2005). Evaluating the impact of food security program only at the program implementation might mask the real 1

3 impact of the program. Moreover, with government s limited resource, a credible impact evaluation is needed to ensure that the resource is not wasteful. As what have been present in various social programs, identifying the impact of food security program might not be straightforward. In particular, food security program participation does not follow a random process 1. Using propensity score matching and difference in difference method, the impact of food security program will be evaluated in several outcomes: food and non food expenditure. From these specific expenditures, issues of aid-fungibility and fly-paper effects can be examined. 2. An Overview of Indonesian Food Security Program during Economic Crisis Indonesian government introduced food security program as part of SSN packages to prevent severe nutritional effect from the economic crisis. Food security intervention became the main component in the SSN program. Officially, the food security program named Operasi Pasar Khusus Beras (OPK) rice special market operation. The purpose of this program was to ensure that the poor households were able to access basic food at an affordable price (Sumarto 2006). The eligible households were selected based on Badan Koordinasi Keluarga Berencana Nasional (BKKBN) National Family Planning Agency in Indonesia. Tabor and Sawit (2001) mentioned that the program authorities were aware of BKKBN welfare criteria which were not designed to identify food insecure households. BKKBN categorizes households based on these following indicators: whether all household members have at least two meals a day, whether household members have different set of clothes for each type of activities (at home, working or going to school and going out), whether the house has a dirt floor, whether they bring the children to health center and receive medical treatment when they are sick, whether the households use family planning methods, and whether the household members are able to practice their religious duties 2. BKKBN involved volunteers and family planning cadres and located them in village health post. The cadres collected and updated those household data along with family planning monitoring. BKKBN welfare criteria gained many criticisms regarding its accuracy. Nevertheless, BKKBN welfare criteria were the only available list at that time meanwhile government needed to immediately mitigate the financial crisis. The implementation of food security program was under Badan Urusan Logistic (BULOG) the National Food Logistic Agency, a government agency responsible for food supply and food price stabilization. The program provided highly subsidized rice at a price of Rp 1,000 per kg compared to the average market rice price at Rp 3,000 per kg. The amount of which could be purchased by the target group was 20 kg per households per month but then reduced to between kg in 2000 (Tabor and Sawit, 2005; Hastuti, 2008). During the first twelve month of food security program, around Rp 3.3 trillion was transferred to more than 9 million 1 Food security program participation is based on certain eligibility criteria. Self-selection issues might rise in program assignment. See Abebaw et al (2010) in the case of Food Security program in Northwestern Ethiopia for comparison in the case of other developing countries setting. 2 BKKN classified households into four categories: keluarga pra-ks (pre prosperous households), KS 1 (prosperous 1), KS 2 (prosperous 2) and KS 3 (prosperous 3). Eligible household for food security program was pre prosperous and prosperous 1 families. 2

4 households as beneficiaries of the program. The program hence functioned as a sort of income transfer to the households. 3. Methodology 3.1 Data The data used in this study are from 1997 and 2000 waves of the Indonesia Family Life Survey (IFLS) which capture periods before and after economic crisis and the implementation of food security program. IFLS collects longitudinal data on individual respondents, their families, their households, the communities in which they live, and the health and education facilities they use. Furthermore, information on community characteristics and food security program are available which therefore enable this study to observe the program heterogeneity. The sample is restricted to panel households and excluded split off households since they might have different characteristics when they stayed in original households. The total sample used in this study is 7193 panel households. 3.2 Dependent and Explanatory Variables To assess the impact of food security program, certain outcomes are measured. The outcomes intended to examine are household food and non food expenditures. More specifically, outcomes of food expenditures are classified into rice, staple food, meat and fish, dairy products and adult goods expenditure. Non food expenditures will focus on two vital expenditures: medical and education. By analyzing the outcomes on specific expenditures, the impact of food security program on consumption smoothing can be assessed. Information on food and non food expenditures are derived from expenditure module and all expenditures adjusted to 2000 prices so that the real expenditure values between two waves are comparable 3. The explanatory variables are used to calculate the probability of receiving food security program in the matching estimator. Therefore, conditions that influence program eligibility will be used to calculate propensity score matching. As mentioned earlier, program eligibility criteria are based on BKKBN welfare criteria including household s welfare conditions as indicated by housing characteristics and income. Hence, covariates involved in propensity score estimator are household head and housing characteristics since these variables are observed and influence program eligibility. In more detail, housing characteristics are observed from the type of walls and floors in the house, whether the house has piped water or house owners. In addition, village characteristics and provincial dummy variables are also included to control for regional heterogeneity. To control for economic conditions in the community, major commodity prices such as rice and chicken price as well as average village per capita expenditure are entered in the matching estimator. 3 The detail calculation of deflators is available at Strauss et al. (2004). 3

5 3.3. Matching Method In the field of impact evaluation, an appropriate method is needed in assessing program s performance against a counterfactual, a situation when the program is absence (Ravallion, 2009). In the observational studies setting, program evaluation should consider the issue of non-random selection which might not be found in the experimental studies. Furthermore, in the situation where randomized experiment is absent, more complex and careful methods are needed. In the non experimental setting, the appropriate methodology of impact evaluation depends on three factors: the type of information available, the underlying model and the parameter of interest (Blundell and Dias, 2000). Richer datasets such as longitudinal or repeated cross-section allow less restrictive estimators. According to Blundell and Dias (2000), instrumental variables and Heckman selection estimators are relevant for cross sectional dataset while double difference can be applied in longitudinal or repeated cross section datasets. Another approach that can be implemented in non experimental setting is matching method. Method of matching is relevant both for cross sectional and longitudinal data. Based on the rich information of the program and the nature of panel data employed in this study, propensity score matching (PSM); a method which mimics experimental studies (Rubin and Thomas, 1996; Rosenbaum, 2002) is implemented. The underlying assumption of matching estimators is that the outcomes are independent of program participation conditional on a set of observed covariates. PSM is not a panacea for evaluation problem but with sound data and ample knowledge on the program, PSM may produce reasonable results 4. Results and Discussion 4.1 Food Security Program Coverage Food security program is the most critical program in SSN since it is dealt with basic food need that should be buffered in the time of crisis. The program was known as an almost universal program since it had the highest program coverage (Suryahadi and Sumarto, 2003). Table 1 presents the food security program distribution based on IFLS from 1998 to The program was launched in April 1998, but the timing of program disbursement varied across community due to different local capacity. It is shown that the program coverage was lower in 1998 compared to the next years. Based on IFLS sample, there were only per cent of communities received food security program and the program was distributed more in rural communities in The program coverage expanded in 1999 in which 95 per cent of IFLS communities were exposed to food security program. Interestingly, urban communities were exposed more to the program in the expansion period (1999 and 2000). Higher urban coverage might associate with some findings that urban households suffered more due to the crisis than the rural counterpart (Suryahadi and Sumarto, 2003). Local capacity which is suspected better in urban areas might also be the reason why the food security program placed more in urban areas. Even though food security program covered slightly more urban communities, recipient households were found to be more in rural areas (Table 1). The percentage of rural households received the program was almost double to the urban households. In addition, the program beneficiaries also increased in the second year of program implementation. This figure might associate to the expansion of food security program particularly when BKKBN 4

6 relaxed the eligibility criteria of food security program beneficiaries to pre-prosperous and prosperous I households. The number of beneficiaries then decreased in 2000 where most of households have been able to recover from the economic crisis. Table 1 Distribution of Food Security Program across Communities based on IFLS Sample Rural Urban Total Community with Food Security Program (%) Proportion of Recipient Households (%) Notwithstanding, the program was also subjected to some leakages (Olken et al., 2001; Olken, 2001). This study also found that a number of households in the highest income quintile also enjoyed the program though it was already known that the program was intended for the poor though it is shown that the percentage of households receiving the program falls as households per capita expenditure rises 4. This result confirms previous study by Hastuti (2008) using SUSENAS data which finds that more than ten per cent of highest income quintile households also became food security program beneficiaries. This finding is also in line with existing literatures on SSN programs where even rich households were benefiting from anti-poverty program the period of crisis (Sumarto et al., 2002; Cameron, 2010). 4.2 Matching Results Covariates involved in program participation are based on BKKBN eligibility criteria and geographical characteristics. Based on the sample, about 38 per cent of households are recipients of food security program. There are big difference between program recipient and non recipient. Based on the descriptive statistic of IFLS 2000 (post exposure) data, the share of food, education and health expenditure among the recipient households were 63 per cent, five per cent and 1.7 per cent, respectively. The gap of per capita expenditure between recipient and non recipient of food security program was almost ten folds. The years of schooling of the household head were higher in non recipient households (ten years) than recipient households (six years). The recipient households tend to have lower quality of housing with dirt-floor and non-brick wall and also limited access to piped water. The program selection is estimated using Probit model. Based on the sample, about 38 per cent of households are recipient of food security program. Beside propensity score, balancing 4 Detail information on food security program targeting performance is available from the author upon request. 5

7 property test is also conducted. In this case, balancing property test end up with 8 blocks of propensity score. The figure shows overlap area of common support. The overlap region of common support is encouraging for propensity score matching. Considering food security program is a targeted program, less significant area of overlap is common since the program selection is based on certain criteria. The test of equality of mean value of recipient and control group could not reject the equality of all covariates. This result is not surprising since food security program is a targeted program. The results from program participation are presented in Table 2. It is revealed that urban and household farm dummy has negative and significant impact on program participation. This finding confirms the descriptive statistics where the higher proportions of beneficiaries are rural households. Household income category also has a significant effect on program participation. The higher the income the less likely the household participates in the program. Low housing quality as indicated by dirt-flooring type has a positive and significant effect on food security program participation. This result shows that on average, the program has reached the targeted group. The regional factors as shown by provincial dummy has significant effect on program participation and provinces in Java are more likely to receive the program. Households in Lampung and West Nusa Tenggara were more likely to become program recipients. Community characteristics as measured by remoteness from vehicle stop station and district capital, average per capita expenditure and rice price indicate a significant impact. Program tends to be placed in relatively remote area and poorer community. The community with higher rice price also tends to receive food security program. 6

8 Table 2 Results for the Matching Estimator Coefficient Std. Error Household Head's Characteristics Age ** Education (years of schooling) ** Work (Work=1) Gender (Male=1) Household characteristics Under 6 years years years (male) *** years (female) years and over (male) years and over (female) * Income category *** Health Card *** Urban * Java *** Housing characteristics Ceramic floor *** Dirt Floor *** Bamboo wall Community Remoteness and Village Economy Nearest bus stop in the village ** District capital in the village Rice price (per kg) *** Chicken price (per kg) Community Average Per Capita *** Provincial Dummy North Sumatra *** West Sumatra *** Lampung *** West Java *** Central Java *** Yogyakarta *** East Java *** Bali *** West Nusa Tenggara *** N 7178 Pseudo R-squared Note: Standard errors are in parentheses. * Denotes statistically significance at 10% level. ** Denotes statistically significance at 5% level. *** Denotes statistically significance at 1% level. 7

9 4.2 Impact on Food and Non Food Expenditures In general, the evaluation conducted in this study investigates the impacts of food security program. The outcomes of food security program are food and non food expenditure. The food expenditure is broken down into rice, staple, dairy products, meat and fish and adult goods expenditure. Non food expenditure is focused on education and health expenditure. In fact, food security program enables household to have extra resource which allows household to allocate this extra resource into human capital investment such as better nutrient food, education and health expenditure. Matching estimator and difference in difference method are applied to examine the program effects. Table 3 reports food and non food expenditures of program households and presents the differences between program recipients and control group. From Table 3, it is shown that the difference of food and non food expenditure is significant, except medical expenditure in preexposure data. It can be seen that food security program recipients spend more on rice, staple food, meat and adult goods expenditures compared to their counterparts. It is also evident that food expenditures of recipients and control both slightly increased between 1997 and 2000, while non food expenditures shows the reverse figure during the period. Within food expenditures, it is found that dairy products, meat and fish and adult goods expenditure increased during the period. Table 3 Food and Non Food Expenditures (in log term) Based on Matched Sample Pre-exposure Post-exposure Recipient Control Diff Recipient Control Diff Food Expenditure *** *** Rice *** *** Staple *** *** Dairy Product *** *** Meat *** *** Fish *** *** Adult Goods *** *** Non Food Expenditure *** *** Medical *** Education *** *** N Note: *** Denotes statistically significance at 1% level. Using matching estimator, the average treatment effect are evaluated based on cross sectional data (IFLS 2000) which captures post-exposure period only as well as panel data which measure the change of consumption before and after the program. In conducting the average treatment effect, this study employs Kernel method since bootstrapping of standard errors procedures may not be appropriate for other matching method such as nearest neighbor matching due to non-smoothness of the method (Abadie and Imbens, 2006). The standard errors of the average treatment effects are given by bootstrapping with 150 replications. The results are presented in Table 3. The second column presents the average treatment effect based on post exposure data only (IFLS 2000) while the third column is derived from panel 8

10 data. The results show that the food security program has no impact on total food and non food expenditure, but it helps the poor in smoothing within food consumption, particularly for poor household to afford meat, fish and dairy products. Table 4 Average Treatment Effect on Food and Non Food Consumption ATT (post exposure) Food Expenditure (0.027) Rice 0.367** (0.141) Staple 0.138* (0.083) Dairy Product (0.167) Meat 0.308** (0.120) Fish 0.454** (0.159) Adult Goods 0.466** (0.178) Non Food Expenditure ** (0.029) Medical 0.274** (0.142) Education (0.195) ATT (DID and matching) (0.026) (0.168) (0.112) 0.383** (0.195) 0.244* (0.142) 0.344* (0.188) (0.193) (0.037) 0.387** (0.187) (0.188) Note: Standard errors are in parentheses. * Denotes statistically significance at 10% level. ** Denotes statistically significance at 5% level. The results show (Table 4) that food security program has a positive and large effect on rice and staple food expenditure based on post exposure data. This means that in the short run, food security program does help the recipients to cover their basic food need though this evidence signs fly-paper effect of the program. Borrowing from Arthur Okun s term, it is found that the government program sticks where it hits. Extra resource from food security program is not a perfect substitute for rice and staple expenditures; rather it acts as a complement. Food security program indeed helps the program recipients since rice price from the program is much lower than the market price. Nevertheless, the increase of rice expenditure as a result from food security program should be explored more whether it is a shift of quality or quantity. This fact is supported by the situation where households tend to smooth their consumption and prioritize their primary consumption in the period of economic crisis. Based on post exposure data, it is also revealed that food security program has positive and substantial effect on meat and fish expenditure. Furthermore, the program also has a positive influence on medical expenditure though overall impact on non food expenditure is negative. The impact of food security program is even larger than the health card impact. Johar (2009) revealed that health card program had very limited impact on the consumption of primary health care. There is a possible explanation for the negative impact of food security on non food expenditure. Recipient of food security program had to increase their food expenditure to meet their basic need and put aside non food expenditure. 9

11 However, the post exposure data shows that there is also unintended effect from food security program. The extra resource as a result of the food security program is also transferred into adult goods consumption and the effect is substantial and even larger than meat and fish expenditure. It has been a long debate that the government support has led to aid-fungibililty in Indonesia. The Indonesian Consumer Foundation reported that there was a misused of direct cash aid from government. Instead of transferring into more human capital related expenditure, more than 50 per cent of the direct cash aid was spent for smoking (Kompas, 2009). In this case, the aid fungibility is evident. From panel data, DID and PSM estimation shows that food security has limited impact on food and non food expenditure. In particular, food security program also has no impact on education expenditure. Notwithstanding, it is consistently revealed that food security program has a positive and substantial effect on meat and fish and medical expenditure. In addition, based on panel data, there is a significant effect of food security program on dairy product expenditure and the impact is slightly larger than the impact on meat and fish expenditure. This means that food security program does support the recipients and contribute to the income of the recipients which enables the households to shift their consumption to more expensive nutrient sources. Accordingly, the program has enabled the program beneficiaries to invest in better in improving human capital. 5. Conclusion The Indonesian economic crisis has hit the poor households and forced them to smooth the consumption. Food security program has provided access for poor household to purchase rice with highly subsidized price. The matching estimator shows that the program had reached its target. Households characterized by low quality housing, who are less educated, residing in rural and Java areas, are most likely to participate in food security program. Notwithstanding, the program was still subjected to many loopholes, particularly related with targeting since geographical bias is also evident in the program implementation. Using propensity score matching, this study reveals that food security program has positive impact on selected food and non food expenditures. In particular, it is found that food security program has enabled the program beneficiaries to increase expenditures on better nutrient food such as meat-fish and dairy products. The program also has positive impact on health expenditures. However, aid-fungibililty is also evident since the extra resource resulted from the program also transferred to adult goods expenditure. Besides, this study also finds fly-paper effects where food security program has increased rice and staple food expenditures in the short period. To sum up, food security does support the program participants to smooth their consumption in period of economic crisis. Food security program indeed help the program recipient to invest more on rich nutrient food though the program is also misused. The findings suggest that improving targeting or applying conditionality might mitigate the unintended effects resulting from the program. 10

12 Reference ABADIE, A., and G. IMBENS (2006): On the Failure of the Bootstrap for Matching Estimators. Mimeo, University of California, Berkeley. ABEBAW, D., FENTIE, Y., and B. KASSA (2010): The Impact of a Food Security Program on Household Food Consumption in Northwestern Ethiopia: a matching estimator approach. In: Food Policy 35: ASIAN DEVELOPMENT BANK, (2007): Key Indicators 2007: Indonesia. Available at: [Accessed March 12, 2008]. BLUNDELL, R., and M. C. DIAS (2000): Evaluation Methods for Non Experimental Data. In: Fiscal Studies 21(4): CAMERON, L. (2009): Can a Public Scholarship Program Successfully Reduce School Drop-Outs in a Time of Economic Crisis? Evidence from Indonesia. In: Economics of Education Review 28: HASTUTI, (2008): The Effectiveness of the Raskin Program, Social Monitoring and Early Response Unit (SMERU), Research Report, Available at [Accessed March 23, 2008] JOHAR, M., (2009): The Impact of the Indonesian Health Card program: a matching estimator approach. In: Journal of Health Economics 28: KOMPAS, (2009): Uang BLT Habis buat Merokok (Cash Transfer is Consumed Up for Smoking), Available at Merokok [Accessed April 25, 2010] OLKEN, B. A., NABIU, M., TOYAMAH, N., and D. PERWIRA, D. (2001): Sharing Wealth: how villages decide to distribute OPK Rice, SMERU Working Paper. SMERU Research Institute, Jakarta. OLKEN, B. A. (2001): Corruption and the Costs of Redistribution: micro evidence from Indonesia. In: Journal of Public Economics 90: RAVALLION, M (2009): Evaluating Anti-Poverty Programs. In: Schultz, P. T. and J. Strauss (Eds.): Handbook of Development Economics Volume 4. North-Holland, Amsterdam, RUBIN, D.B., and N. THOMAS (1996): Matching using estimated propensity scores: relating theory to practice. In: Biometrics 52: STRAUSS, J., BEEGLE, K., DWIYANTO, A., HERAWATI, Y., PATTINASARANY, D., SATRIAWAN, E., SIKOKI, E., SUKAMDI, and F. WITOELAR (2004): Indonesian 11

13 Living Standard: before and after the financial crisis, RAND Corporation, Santa Monica. SUMARTO, S., (2006): Social Safety Net: Indonesia, Policy Brief 6, Overseas Development Institute London. Available at: [Accessed March 12, 2008] SUMARTO, S., SURYAHADI, A., and W. WIDYANTI (2002): Designs and Implementation of Indonesian Social Safety Net Programs. In: The Developing Economies XL(1): SURYAHADI, A., and S. SUMARTO (2003): Evolution of Poverty during the Crisis in Indonesia. In: Asian Economic Journal 17(3): TABOR, S.R., and M. H. SAWIT (2005): RASKIN: A Macro-Program Assessment, Available at y_for_distribution.pdf. [Accessed February 27, 2009] TABOR, S. R., and M. H. SAWIT (2001): Social Protection via Rice: the OPK Rice subsidy program in Indonesia. In: The Developing Economies XXXIX(3):