The Medium-run Effects of a Cash Transfer on Beneficiary Health: Evidence from the Philippines

How to cite

Kandpal, Eeshani; Filmer, Deon; Friedman, Jed. 2023. "The Medium-run Effects of a Cash Transfer on Beneficiary Health: Evidence from the Philippines." AEA RCT Registry. AEARCTR-0011622. https://doi.org/10.1257/rct.11622-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Eeshani Kandpal* , Deon Filmer , Jed Friedman

* Primary investigator

Registry ID
AEARCTR-0011622
DOI
https://doi.org/10.1257/rct.11622-2.0
First registered
Last updated
Status
In development
Trial dates
2023-07-01 – 2025-12-31
Countries
Private
JEL codes
I18, I38, O12, O15, O20
Registry record
http://www.socialscienceregistry.org/trials/11622

Abstract

While conditional cash transfers (CCTs) have been well documented to improve school enrollment and attendance (Fiszbein and Schady, 2009; Hanlon, Barrientos, and Hulme, 2010; Saavedra and Garcia, 2012), the evidence of their impact on health outcomes is mixed. Virtually all CCTs have increased the use of preventive health and nutrition activities (Ranganathan and Lagarde, 2012; Gaarder, Glassman, and Todd, 2010; Lagarde, Haines, and Palmer, 2007). However, these programs have generally not led to improvements in anthropometry, although a few have reported meaningful impacts on selected nutritional outcomes (Ruel and Alderman, 2013). A meta-analysis of 17 programs that combined CCTs and unconditional cash transfers shows a mean impact of 0.025 on height-for-age z scores (HAZs), an effect size that is neither statistically significant nor biologically meaningful (Manley, Gitter, and Slavchevska, 2013). Perhaps unsurprisingly, the evidence on the long-term effects of CCTs largely focuses on the persistence of education impacts rather than cognitive, socioemotional or indeed health and nutrition outcomes (Molina Millán et al., 2019). In contrast, studying the conditional cash transfer in the Philippines, Pantawid, Kandpal et al. (2017) find substantial decreases in extreme stunting rates among beneficiary children in their first 1000 days of life.

Building on the large-scale randomized impact evaluation of Pantawid, this proposed study will contribute evidence on the persistent impacts of CCTs during the developmentally critical first 1000 day period on medium-run (approximately 10 years later) outcomes. Specifically, it will ask whether beneficiary children who received the program in utero or within the first 1000 days of life still differ from those who did not receive it, approximately ten years after the initial impact evaluation survey. We will seek to understand the impacts of the Pantawid program “shock”, operating chiefly through early life nutrition, on educational attainment, cognitive and socioemotional skills, and child work and disease histories. While control communities have entered the program since the survey for the first-round impact evaluation in 2011, exogenous variation in a cohort of children who received this shock at a critical stage in their growth has been maintained. We will exploit this variation in comparing cohorts that received the program in utero or the first 1000 days of life to the same cohorts in control communities which received the program after this malleable window had passed for these cohorts. This comparison will allow us to causally identify any medium run effects of Pantawid that operated through the discussed nutrition and health channels specific to early life. This component of the study would thus test the hypothesis that it is investments during a critical period in early child development that lead to sustained impacts. Alternatively, if sustained impacts aren’t identified, then benefits from catch-up investments made later in the child’s life are an attainable possibility.

Intervention

In this new research, we propose to return to the 130 treated and control communities that constituted the experimental sample of Pantawid (Chaudhury, Friedman, and Onishi 2013) to assess the height of children in both treated and control areas and from both beneficiary and non-beneficiary households, as well as cognitive, educational and health related outcomes. Doing so will allow us to carefully and comprehensively assess the medium run effects of CCTs on cognitive, socioemotional, and health outcomes among beneficiaries and the consequences of early life stunting using well-identified exogenous variation in stature among non-beneficiaries.

This research asks whether beneficiary children who received the program in utero or within the first 1000 days of life still differ from those who did not receive it, approximately ten years after the initial impact evaluation survey. We will seek to understand the impacts of the Pantawid program “shock”, operating chiefly through early life nutrition, on educational attainment, cognitive and socioemotional skills, and child work and disease histories. While control communities have entered the program since the survey for the first-round impact evaluation in 2011, exogenous variation in a cohort of children who received this shock at a critical stage in their growth has been maintained. We will exploit this variation in comparing cohorts that received the program in utero or the first 1000 days of life to the same cohorts in control communities which received the program after this malleable window had passed for these cohorts. This comparison will allow us to causally identify any medium run effects of Pantawid that operated through the discussed nutrition and health channels specific to early life. This component of the study would thus test the hypothesis that it is investments during a critical period in early child development that lead to sustained impacts. Alternatively, if sustained impacts aren’t identified, then benefits from catch-up investments made later in the child’s life are an attainable possibility.

Experimental design

The proposed study will survey approximately 650 children from treated and control villages. Of these 650 children, we have complete anthropometry data in 2011 for about half the sample. These children were between 6 and 36 months old at the time of the 2011 survey and will be between 12 and 15 years of age at the time of the proposed follow up. As described above, the assignment to treatment or control villages was randomized and community characteristics across treatment and control exhibited a high degree of balance. Household eligibility for the transfer was determined by a 2008 listing database on household wealth. Therefore, this sample remains experimental and will enable us to rigorously identify the medium-run effects of the CCT.

We estimate effects by comparing eligible households (poor with young children) located in treated villages to otherwise eligible households in control villages. In 2012, after the first-round impact evaluation, the control villages were included in the Pantawid program. Thus, the identification of impact between treatment and control groups concerns the three-year experimental period when the cohort of children born 2008 – 2011 were differentially exposed to the program.

The data we intend to use are from the 2011 survey, a 2023-24 follow up, administrative data on school enrollment and test scores, and the Philippine’s household expenditure survey, the FIES. The 2011 survey was conducted by the World Bank, including members of the team proposing the current study. All best practices in data collection were followed, and the survey yielded high quality data, including on anthropometry. Cost permitting, we will attempt to work with the same data collection firm for the proposed 2020 survey, and will switch to tablet-based data collection, further improving data quality. Philippine administrative data and FIES are known to be high-quality datasets. FIES is the official survey conducted by the Philippines Statistical Authority to measure poverty incidence and consumption aggregates. Conducted every three years, it has been used by a large number of studies and researchers to date (including this team in its analysis of changes in unit values of food over time).

Several categories of outcomes are of potential relevance to the proposed study. The first set, of course, are anthropometric outcomes, which will determine whether early life changes in stunting persist. In addition, evidence suggests that early life investments in, or indeed insults to, health and nutrition lead to sustained effects on cognitive ability. Therefore, we will measure cognitive outcomes as well as possibly relevant socioemotional and other non-cognitive outcomes. Cognitive and socioemotional tests appropriate for this age group include Raven’s Progressive Matrices to measure non-verbal cognitive ability, reading comprehension and vocabulary tests as well as specially designed games to assess behavioral dimensions such as persistence and curiosity. We will also seek to measure recent disease incidence to understand whether the CCT or early-life stunting affect the predisposition to chronic disease.

In addition, we will exploit all available administrative data on the size of transfers received, the age of school start, any interruptions or grade repetitions, scores on standardized school tests taken by these children. Using recall data, we will seek to construct detailed consumption and health histories, and any labor force participation.

Design details

Randomization method
14 year follow up of a randomization in office by a computer
Randomization unit
Village
Clusters
130 villages
Observations
650 children
Arms
65 treated villages; 65 control villages. Approximately 350 treated children and 300 children in control villages.
Primary outcomes
Height-for-age; school enrollment and attendance; other measures of health; cognitive outcomes and socioemotional skills

Cited by

No citation data available for this registration yet.

Keywords