Inadequate provision of health services can form a major barrier to care (Filmer, Hammer and Pritchett, 2000;World Health Organization, 2005;Worid Health Organization, 2006b;Leshabari, et al., 2008). For example, Chaudhury, et al. (2006) found that Indonesia had one of the highest primary health provider absentee rates of countries sampled, with 40 percent of its local health workforce absent at unannounced visits. Other studies similariy find high rates of absenteeism in other developing countries (Banerjee, Deaton and Duflo, 2004). [A number of studies have shown that providers are responsive to incentives but that incentive systems can be undermined by those higher up in the bureaucracy who may be more interested in the welfare of providers than of patients (Banerjee, Duflo, and Glennerster, 2008). As a result there Is substantial policy Interest in involving communities more directly in delivery oversight and in providing incentives (WDR, 2004).] The Generasi program is designed to address these problems in several ways. Each year all participating villages In Generasi receive a block grant to improve maternal health, child health, and education in their villages. To give villages incentives to use the block grant efficiently, the size of each village's block grant in the subsequent year depends in part on its performance on 12 targeted indicators, such as, the number of children who receive immunizations, prenatal and postnatal care, child nutrition, and the number of children enrolled and attending school. The Generasi program may improve the provision of health services in several ways. First, communities may choose to use the block grant to improve the financial resources available for service provision [and to reduce the distance patients have to travel to get to them,] by subsidizing the hiring of additional service providers or providing transportation subsidies to providers so that they can increase outreach services to outlying villages. By allowing communities to decide how best to allocate the funds, block grants allow communities to take advantage of their local information as to where needs are greatest (Alderman, 2002;Bardhan, 2002;Faguet, 2004;Galasso and Ravallion, 2005). By providing explicit financial performance bonuses, the incentivized version of Generasi provides an incentive for communities to focus resources where they will most improve health indicators, rather than where they generate the most rents for elites (Chattopadhyay and Duflo 2004;Platteau 2004;Bardhan and Mookherjee 2006). Second, the Generasi program may improve community oversight of health providers. Several recent studies [{including studies by other members of the POI project team)] have suggested an important role for community oversight (Bjorkman and Svensson, 2009) and incentives (Banerjee, Duflo and Glennerster, 2008) in improving health services. By allowing communities, instead of service providers, control of funds, block grants give communities leverage over service providers (World Bank, 2004), which could potentially give communities the leverage to improve oversight. Moreover, in the incentivized Generasi program, communities have an incentive to closely monitor service providers to ensure that services are actually being implemented in their communities, which helps overcome the free rider problem in monitoring (OIken, 2007). B2. Estimating the impact of community block grants on maternal and child health-seeking behavior Existing evidence suggests that health-seeking behavior?the decision to seek care?is affected by both ease of access to health services and the quality of those services (e.g., Samrasinghe and Akin, 1994;Akin and Hutchinson, 1999;Lewis, Eskeland and Traa-Valerezo, 2004). In particular, several studies have shown that women are more likely to obtain reproductive health services if a provider is located close to their place of residence (e.g., Panis and Lillard, 1994;Frankenberg, 2005;Achadi, et al., 2007), and children are more than twice as likely to be fully immunized when regular monthly immunizations are available in their village (Banerjee, Duflo, Glennerster and Kothari, 2008). With respect to quality of care, Bloom, et al. (2006), show that contracting out the operation of Cambodian health clinics to nongovernmental organizations (NGOs) improved the quality of care and led to dramatic increases in antenatal care, childbirth in a health care facility, and vitamin A receipt?many of the same types of health-seeking behaviors that are the focus of this study. Since the community block grants and the increased community oversight that goes with them may improve health care quality and accessibility, the existing evidence suggests that these improvements may lead to an increase in health-seeking behavior. Existing research, [including studies by members of this POI research team, also suggests that healthseeking behavior is price sensitive (e.g., Cohen and Dupas, 2007;Kremer and Miguel, 2007).] In Indonesia, the IRMS experiment in the eariy 1990s, which increased prices at health clinics in a randomly chosen set of six districts, found price elasticities of demand between -0.4 and -1.0 (Gertler and Molyneaux, 1996;Dow, et al., 1997). To the extent the Generasi block grants are used to subsidize user fees, this could also increase service provision. CCTs, such as Mexico's Progresa program, examine similar health-seeking behaviors. Similar to Generasi, CCTs provide grants, and condition the size of the grants on health-seeking behavior (as well as educational enrollment and attendance);the key difference is that Generasi grants and incentives are at the individual level (and hence affect only demand). Rigorous evaluations of CCTs in several countries have shown dramatic impacts on health-seeking behavior (Rawlings and Rubio, 2005). For example, CCTs increased participation in growth monitoring by 18 percent in Mexico (Gertler, 2000) and by 22 percent in Colombia (Attanaslo, et al., 2005b). CCTs also increased immunization: measles coverage for children between 12 and 23 months old increased by 3 percentage points in Mexico (Barham, 2005), and DPT coverage for children under two increased by 9 percentage points in both Colombia (Attanaslo, et al., 2005b) and Honduras (Morris, et al., 2004).^ Existing randomized evaluations of CCTs do not separate out income effects (the cash transfer) from price effects (conditioning the transfer on health-seeking behavior).^ Although the intervention studied here is at the community level, not the individual level, the fact that Generasi subdistricts are randomly allocated to two versions of the program?incentivized and non-incentivized?means that we will be able to separately identify the program's income effects and incentive effects. [Another close parallel to the Generasi program are pay-for-pertormance contracting schemes, where nonstate actors (usually NGOs) are contracted by the government to Improve health delivery systems for a defined population. Loevinsohn and Harding (2005) review 10 cases of large-scale contracting for health outcomes in the developing world, and conclude that contracting out these services was associated with substantial improvements in targeted indicators, particularly on factors such as immunization, vitamin A, and antenatal care, all of which are included in the Generasi target indicators.]
