Unintended Spillovers of Targeted Health Insurance on Intra-household Resource Allocations

Abstract

The last three decades have seen an increase in the number of targeted health insurance programmes being implemented in developing countries. However, little is known about their intra-household impacts on household members who are not the intended beneficiaries. Using variation introduced by a universal health insurance programme targeted to children below the age of six in Vietnam, I assess the programme’s impacts on expenditures and labour supply of ineligible children. I find that beneficiary households decrease spending on health and education and decrease leisure time for ineligible children relative to non-beneficiary households after the programme is introduced. These results provide new evidence to the intra-household literature on unintended impacts of targeted programmes on resource allocations for children who are not eligible. They also call to attention, the need for concurrent interventions for older children when implementing early life health interventions in developing countries.

Acknowledgements

I wish to thank the Richard-Palmer Jones and the two anonymous referees who have provided constructive feedback in the development of the paper during the review process. I specially want to thank John Hoddinott, Joseph R. Cummins, Mindy Marks and Srinivas Venugopal for their inputs on developing the paper. I would also like to thank participants at research seminars from the Tata-Cornell Institute of Agriculture and Nutrition (Cornell University), the University of Vermont, Burlington and the Indian Institute of Management, Udaipur for their inputs into earlier versions of the paper. I would like to thank Hai Nguyen Thi Hong from the Statistical Documentation and Service Center in Vietnam for helping me purchase the data. Any errors here are mine alone.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary materials

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2020.1806243

Notes

1. In the Supplementary Materials section S1, I build on the household model proposed by Yi et al. (2015), to provide some intuition on how these unintended effects may occur.

2. More details on these programmes can be found in Supplementary Materials (section S2).

3. In 2008, the private health services sector was heavily regulated. Most private providers sold drugs and medicines and there were very few private practices.

4. Data from the Global Health Expenditures Database of the World Health Organisation. Accessed July 2019 – http://apps.who.int/nha/database/ViewData/Indicators/en.

5. Hansen and Le (2013) through a careful exposition show that, when compared to the census data, these surveys may represent only the better off communes. This has created some concern that the sample may not be nationally representative. Despite this limitation, the VHLSS is considered to be a high-quality data source for researchers interested in studying the country (Liu, Violette, & Barrett, 2016).

6. Sick days for individuals are collected only after the 2002 VHLSS.

7. Details on how these variables were constructed can be found in the Supplementary Materials section S3 of the paper.

8. This is because they may have had different labour market opportunities, health needs and education needs compared to households with children who are all older than 17 (Deaton, 1997).

9. The regressions are robust to adding commune-level fixed effects and clustering at the level of the commune as well. See table S8 in the Supplementary Materials (Section S6).

10. The CHI did not provide children under six in HCFP households with additional or more insurance. They were allowed to access health services through the CHI or the HCFP.

11. Details on the methodology I use to construct this measure and the sensitivity to the 50 per cent cut off on main outcomes is explored in the Supplementary Materials (section S4). Overall, I find no cause for concern that the current methodology is subject to measurement error could bias the main results.

12. The coefficient of 1.64 USD is calculated by converting coefficient into 1000 Vietnamese Dong (DNG) and then dividing it by 15,500. This was the Vietnamese DNG to USD exchange rate in 2008.

13. The coefficient of 1.12 USD is calculated by converting coefficient into 1000 Vietnamese Dong (DNG) and then dividing it by 15,500. This was the Vietnamese DNG to USD exchange rate in 2008.

14. I also run quantile regressions at different percentiles of the distribution. In line with intuition, I find that the results on education and health are driven changes in spending conditional on higher percentiles in the distribution (>75%). This would imply that these allocation effects are stronger in households that were spending more on their older children to start with compared to similar households in the spending distribution. Results can be presented on request.

15. It is not uncommon in these contexts that households and parents determine what types of household activities their older children should be involved in. Older girls often help with child care and older boys help as unpaid labour.