14. Consumption Behavior of the Poorest and Policy Implications in Indonesia

Understanding the characteristics of the poorest is an appropriate starting point for designing policies and programs that improve their welfare. Following Ahmed et al. (2007), the poorest households were defined as the bottom 20 % of the household income range, whose members were living on less than US$0.50/day. I focused on the characteristics of the poorest households that highlight several aspects of the relationships between food policy and household food consumption behavior, including expenditure patterns of food consumption, geographical aspects, and social conditions. Understanding the diversity of food consumption is an important key for addressing the food and nutritional problems of the poor. The poorest households are often found in relatively remote locations that have the least access to public infrastructure and facilities such as roads and markets, which disadvantages them by limiting food variety and increasing food prices. This geographical association with poverty also hinders access to other basic needs, such as health care and education. With respect to human and social capital, members of the poorest households are less likely to be educated and more likely to be systematically excluded from certain opportunities.

As reported above, the poorest households consumed greater shares of good from the “staple foods” and “tobacco or alcohol products” categories, and lesser shares of nutrient-rich foods from the “meat and fish” or “dairy products” cateories than other households. An important aspect of understanding the behavior of the poorest households for policy making purposes is their responsiveness to changes in income and prices, which can be observed using empirical data from the surveys. How the poor respond to changes in prices and income should help direct food policy with respect to policy emphasis on either income or price, or on a combination of the two (Sadoulet and de Janvry 1995). Particularly when policies are needed to leverage nutrition and health, side interventions should be considered. Policy to improve household food consumption should therefore be grounded on sound knowledge of household responses to these key determinants of food consumption.

Changes in consumption among the poorest households in response to changes in income can be evaluated through expenditure elasticity. Elasticity estimates for expenditures of the poorest households on eight consumer goods categories are presented in Table 14.1. All of the expenditure elasticity estimates were calculated based on the IFLS data using the Quadratic Almost Ideal Demand System (QUAIDS), specific demand system model. For the “vegetables and fruit,” “oils,” and “snack and dried foods” categories, the expenditure elasticity estimates were close to unity and therefore these goods are deemed as necessities, whereas “staple foods,” “meat and fish,” “dairy products,” and “alcohol and tobacco products” categories were all classified as luxuries. Expenditure elasticity for the alcohol and tobacco products category was the highest. For this category the model indicated that a 10 % increase in total food expenditures would lead to a 13 % share increase in consumption. Comparing “staple foods” expenditures across annual surveys, these appeared to be elastic for the poorest households in 1997, but exhibited less elasticity in subsequent annual surveys. This finding was consistent with Bennet’s law, which states that households switch consumption from cheaper to more expensive calorie sources as income rises (Timmer et al. 1983; Fuglie 2004). For the poorest households, the “dairy products” category remained expenditure elastic for over a decade. As expenditures on tobacco and alcohol products increased over the last decade and considering the elastic nature of expenditures in this category, caution should be taken before drawing conclusions due to the possible crowding-out effect of alcohol and tobacco product expenditures on the consumption of more nutritious food items.

Price elasticity estimates reflect the responsiveness of consumption to changes in prices. Table 14.2 presents the own- (with respect to changes in the price of goods within each category) and cross-price (with respect to changes in the prices of goods from other categories) elasticity estimates for the poorest households based on a QUAIDS analysis of the IFLS data. Similar to expenditure elasticity, price elasticity for the “alcohol and tobacco products” category showed the most substantial response to price changes. The model estimated an elasticity value of 1.25 in absolute magnitude, therefore predicting that if the prices for alcohol and tobacco products increase by 10 %, then consumption would fall by 12.5 %. The poorest households also exhibited price elasticity for the “dairy products” and “meat and fish” categories. The own-price elasticity of the dairy products category was similar to that for “alcohol and tobacco products,” indicating that price changes for “dairy products” affected consumption less substantially than price changes for “alcohol and tobacco products.” The price elasticity value for the “staple foods” category was nearly close to unity among the poorest households, indicating that the demand for staple foods is relatively responsive to price changes.

Examination of the relationships among these consumption categories through the cross-price elasticity estimates revealed a mixture of complementary and substitutive relationships. Cross-price elasticity values were generally much lower than own-price elasticity values, implying that consumption of each category among the poorest households was more responsive to changes in the price of goods within each category. Though all cross-price elasticity estimates were found to be inelastic, the notion of substitutes and complements plays a significant role in understanding consumption patterns. The poorest households somewhat freely substituted between the “meat and fish” and “vegetables and fruit” categories, suggesting that they seem to have similar roles in their diet. The comparison of the “vegetables and fruit” category prices to those of “meat and fish” yielded a demand elasticity value of 0.24, which predicts that a 10 % increase of vegetables and fruit prices would lead to a 2.4 % increase in the consumption of meat and fish.

A similar relationship was also found between the “meat and fish” and “staple foods” categories. The model predicted that the poorest would increase their staple foods consumption by 2.3 % if the price of meat and fish increased by 10 %. This finding suggests that poor households would cut expenditures on nutrient-rich foods such as green leafy vegetables or meat and fish, and shift to cheaper calorie goods such as staple foods. During the 1997 financial crisis there were dramatic price increases for most consumer goods, particularly for nutrient rich foods belonging to the “meats and fish,” “dairy products,” and “vegetables and fruits” categories, which had negative consequences on the nutritional welfare of Indonesian households.

Another striking substitutive effect, though inelastic, was found in the relationship between the demand for goods from the “alcohol and tobacco products” and “meat and fish” categories. The model predicted that if alcohol and tobacco product prices were to rise by 10 % then meat and fish demand would increase by 3.0 %. A substitutive relationship between “alcohol and tobacco products” with “meat and fish” is challenging for national food policy. What is important to poor households with respect to alcohol and tobacco products demand is the price. Only when the price of alcohol and tobacco products significantly increased, was a shift to greater meat and fish consumption among poor households predicted. Nevertheless given the large expenditure share of alcohol and tobacco consumption and its elasticity, it should be cautiously considered. The loss in terms of nutrition could be considerable due to the possible crowding-out effect of alcohol and tobacco products expenditures on those for nutrient rich foods such as eggs, meats, dairy products, and therefore household well-being would also likely be affected by decreases in nonfood expenditures such as education and medical expenses (BAPPENAS 2006; Mukherjee 2006).

The Indonesian government provides food subsidies to poor households for the purpose of maintaining access to basic food items. Indonesia operates a food security program called RASKIN, which is an abbreviation of beras miskin, literally meaning “rice for the poor.” Historically the program was part of the Social Safety Net introduced by the government in 1998 just after the financial crisis hit Indonesia in mid-1997. The purpose of this program was to ensure that poor households (both the chronically poor and those newly poor due to the crisis) were able to access basic food at affordable prices (Sumarto 2006). Eligible households were selected by the National Family Planning Agency (Badan Koordinasi Keluarga Berencana Nasional [BKKBN]).

Tabor and Sawit (2001) pointed out that the program authorities were aware that the BKKBN welfare criteria were not designed to identify food insecure households. The BKKBN evaluated households based on the following indicators: whether all household members regularly consumed at least two meals a day, whether household members had different sets of clothes for different activities (home, work, school, and public), whether houses had dirt-floors, whether households were able to bring their children to health centers to receive medical treatment when they were sick, whether the households used family planning methods, and whether household members were able to practice their religious observances. The program provided highly subsidized rice at a price of Rp 1,000/kg, compared to average market prices of Rp 3,000/kg. The amount of rice that could be purchased by beneficiary households was originally 20 kg per month, but this was reduced to between 10 and 20 kg in 2000 (Tabor and Sawit 2005; SMERU 2008).

Since the program implementation was rushed and the eligibility criteria were not designed properly (Tabor and Sawit 2001), leakage has been a significant problem with the Rice for the Poor program. Table 14.3 presents indicators of the effectiveness of the food security program targeting. The percentage of households that received benefits from the program fell as household per capita expenditures rose in both Java and the rest of the country. Only about 30 % of the Rice for the Poor program beneficiary households were among the poorest 20 % of society. The targeting performance was slightly better outside of Java where almost 60 % of the program beneficiaries were among the poorest 40 % of households.

Strikingly almost 20 % of the program beneficiaries were from the highest income quintile households, even though the program was intended to help the poor. The misadministration of the program was also evident in Java, where almost 10 % of the wealthiest households benefitted from the program. This indicated a failure of the food security program operators to provide oversight of the selection of beneficiaries. Improved targeting would be an effective means for the Indonesian government to improve program benefits for the poor, but the costs and feasibility of more narrow targeting need to be assessed in order to identify economically optimal targeting methods.

The results presented in Table 14.3 show that even a targeted social protection effort for the poor can be subject to errors of inclusion. The errors of inclusion in the implementation of the Rice for the Poor program have several implications. Based on the latest IFLS community data, the program was the most well-known among the poorest and was considered the most beneficial social protection program. However, it was also the second most criticized program due to the lack of program socialization and a suitable mechanism for filing complaints.³ There were other consequences related to the limited program impact on the poorest households. In a similar case of a conditional transfer program in Honduras, there was no positive impact on household nutritional status due to implementation problems (Adato and Hoddinott 2009). In the case of the Rice for the Poor program in Indonesia, targeting seems to be a bottleneck that limited the program’s impacts.

The evidence that the Rice for the Poor program had limited impacts on the poorest is presented in Table 14.4. The program’s impacts on food consumption were evaluated at the household level after program implementation and the resulting changes in consumption are presented. The program benefitted the poorest households by increasing meat consumption by 50 % at the time of post-exposure program disbursement. “Dairy products” category expenditures increased by about 90 % among the poorest households. Though animal source foods such as meat and dairy products contain more nutrients, the increased consumption of these categories by the poorest households was attributed to a shift to “better-tasting” food (Jensen and Miller 2008; Banerjee and Duflo 2011).

As previously mentioned, the food security program seems to have had many loopholes. By design the Rice for the Poor program provided generous support to the households that were suffering most from the crisis. Although it was found that the program reached the intended households, the mistargeting of households was also clear and lead to unintended program effects. Since the poorest households have limited understanding of dietary nutrition, accompaniment of the food subsidy intervention seems to be a prudent option. Changing food consumption behavior of the poor might require time and education rather than be an immediate response to improved food availability. In addition, the fact that extra resources from cash transfers are often used to purchase unhealthy goods that are consumed by adult males such as cigarettes is another indication that government intervention to improve household food consumption should consider an accompanying educational component.

Although the Rice for the Poor program beneficiaries were mostly determined by BKKBN welfare criteria, local village authorities were responsible for the distribution of benefits at the village level. Addressing the political and bureaucratic causes of exclusion might be an important pathway towards more effective targeting (von Braun et al. 2009). The Rice for the Poor program had little bottom-up process in the implementation, particularly the involvement of households as the ultimate point of distribution. Analysis of the IFLS community data revealed that the village head and associated staff, along with the BKKBN cadre and the head of neighborhood associations determined which households were eligible for the Social Security Net programs, including Rice for the Poor. The data analysis results further revealed that only 50 % of the villages had mechanisms in which a household could have applied to be a program recipient if they felt that they qualified for receiving subsidized rice but were not selected. The long bureaucratic distribution process combined with local and cultural contexts might have restricted the eligibility of some community members, particularly the ones with the lowest local socioeconomic status, from being involved in determining the program’s application.

Many studies have highlighted the occurrence of elite capture in development programs and have recommended community-based development program approaches to reduce this problem (Mansuri and Rao 2004; Beard and Dasgupta 2006; Fritzen 2007). This problem appears to have affected the Rice for the Poor program. Making the program more reflective of people’s needs and involving bottom-up participation are potential methods of improving the design of food security programs, just as they have benefitted other community-driven development programs (Rao and Ibanez 2005). An inclusive and transparent process of recipient eligibility identification is one of the operational aspects that would help mitigate this problem.