Exploring the heterogeneous impacts of Indonesia's conditional cash transfer scheme (PKH) on maternal health care utilisation using instrumental causal forests (2501.12803v1)

Abstract: This paper uses instrumental causal forests, a novel machine learning method, to explore the treatment effect heterogeneity of Indonesia's conditional cash transfer scheme on maternal health care utilisation. Using randomised programme assignment as an instrument for enroLLMent in the scheme, we estimate conditional local average treatment effects for four key outcomes: good assisted delivery, delivery in a health care facility, pre-natal visits, and post-natal visits. We find significant treatment effect heterogeneity by supply-side characteristics, even though supply-side readiness was taken into account during programme development. Mothers in areas with more doctors, nurses, and delivery assistants were more likely to benefit from the programme, in terms of increased rates of good assisted delivery outcome. We also find large differences in benefits according to indicators of household poverty and survey wave, reflecting the possible impact of changes in programme design in its later years. The impact on post-natal visits in 2013 displayed the largest heterogeneity among all outcomes, with some women less likely to attend post-natal check ups after receiving the cash transfer in the long term.

• The paper employs instrumental causal forests on experimental data to explore the heterogeneous impacts of Indonesia's PKH on maternal healthcare.
• Positive effects were found for assisted deliveries and temporary health check-up gains, but no significant impact on facility deliveries, with outcomes varying by beneficiary.
• The results suggest PKH effectiveness could be improved through targeted policies considering local conditions and healthcare supply, highlighting the value of machine learning for evaluation.

Analyzing Heterogeneous Impacts of Indonesia's Conditional Cash Transfer Scheme on Maternal Health Care Utilization

The research paper explores the nuanced impacts of Indonesia's Program Keluarga Harapan (PKH), a conditional cash transfer (CCT) scheme, on maternal health care utilization using an advanced econometric methodology called instrumental causal forests. This paper provides insights into how enrolment in the PKH program influences health care behaviors among mothers and how these effects differ across various socio-economic and geographic dimensions.

Methodology and Data

This paper employs instrumental causal forests, an innovative machine learning technique designed to estimate heterogeneous treatment effects. The researchers leverage a unique dataset derived from a large-scale randomized experiment conducted during the PKH rollout. The experimental design included a baseline survey and follow-up surveys in 2009 and 2013, enabling a robust evaluation of short-term and long-term program impacts.

The methodology hinges on using instrumental variables (IV) to correct for selection bias, given the non-random enroLLMent in PKH. Specifically, the random assignment of the program serves as an instrumental variable to distinguish the causal effects of the program from other confounding factors.

Key Results

The findings from the analysis highlight substantial heterogeneity in program impacts:

1. Good Assisted Delivery: The paper finds significant positive effects on the probability of good assisted delivery both in the short term (2009) and long term (2013), with local average treatment effects (LATE) estimated at 0.15 and 0.16, respectively.
2. Health Check-ups: There are notable gains regarding prenatal visits in 2009, yet the effects diminish by 2013. Similarly, positive impacts are noted for postnatal visits in 2009 but not sustained in the longer-term analysis.
3. Facility Delivery: The results suggest that PKH did not have a significant effect on increasing deliveries in healthcare facilities at either time point.

The researchers stress that these effects are heterogeneous, with variability influenced by factors such as urban-rural residence, health care worker availability, and socio-economic status.

Implications and Future Directions

The findings imply that the efficacy of programs like PKH could be enhanced through more targeted policy design that accounts for local socio-economic and supply-side conditions. Specifically, aligning CCT initiatives with improvements in healthcare infrastructure and accessibility could bolster effectiveness.

The paper highlights the role of machine learning in unveiling heterogeneity in treatment effects, advocating its integration into policy evaluation frameworks. Future research could build upon these insights by investigating the application of AI in dynamically adjusting CCT conditions or targeting based on evolving socio-economic landscapes, thus optimizing policy impacts.

Conclusion

This paper contributes significantly to understanding CCT's nuanced impacts, challenging one-size-fits-all policy implementations by recognizing inherent heterogeneity among beneficiaries. The use of instrumental causal forests presents a methodological advancement in evaluating such heterogeneity, offering other researchers a blueprint for integrating machine learning in policy impact assessments.