Elsevier

World Development

Volume 113, January 2019, Pages 246-258
World Development

Universal health coverage: A (social insurance) job half done?

https://doi.org/10.1016/j.worlddev.2018.09.004Get rights and content

Highlights

  • We assess the economic impacts of illness with health but no earnings insurance.

  • Substantial economic illness risks remain under the Thai Universal Coverage Scheme.

  • Severe illnesses increase out-of-pocket medical spending and cause earnings loss.

  • Spending on non-medical goods is protected by informal transfers, credit and savings.

Abstract

Evidence on households’ ability to smooth consumption over health shocks is mostly obtained from environments where there is little or no formal insurance of either medical expenses or sickness-related earnings losses. To establish whether households remain economically vulnerable to illness after the introduction of universal health coverage (UHC), we examine the impact of health shocks of different severity on informal workers in Thailand who are entitled to comprehensive public medical care but lack social protection of earnings. Using three years of panel data, we find that the most severe illness that strikes an initially healthy worker reduces household earnings by almost one third and, despite UHC, raises out-of-pocket spending on medical care by around two thirds. However, households are able to protect spending on goods other than medical care by drawing on informal insurance, credit and savings. These coping strategies substitute for the lack of formal earnings insurance and fill gaps in the effective health care coverage. On average, the combination of UHC and informal insurance of the residual risks does a reasonably good job of protecting living standards from the economic impact of illness, at least in the short term.

Introduction

A global campaign to extend health care coverage in the developing world was launched in the last decade (World Health Organization, 2010). This push for universal health coverage can potentially improve protection against the risk of unforeseen medical expenses, but it will do little or nothing to reduce exposure to the other major economic risk associated with illness – lost earnings. Nevertheless, if households that acquire health care coverage can cope with the earnings risk by drawing on support networks, credit and savings, then formal health insurance combined with this informal insurance of earnings may be an appropriate policy response to health risks in economies characterized by a large informal sector.

Whether the effective health care coverage in such economies provides sufficient protection against the economic risks associated with illness is currently difficult to establish because most of the existing evidence on households’ ability to smooth consumption over health shocks is obtained from environments in which there is little or no formal insurance of either medical expenses or sickness-related earnings losses (Townsend, 1994, Gertler and Gruber, 2002, Asfaw and Von Braun, 2004, Wagstaff, 2007, Genoni, 2012, Islam and Maitra, 2012, Mohanan, 2013, Khan et al., 2015). Inability to smooth consumption over both risks does not necessarily imply incapacity to cope with one of them. This paper contributes to knowledge of the economic impact of health shocks by examining the extent to which illness continues to threaten living standards in a population that has acquired cover for health care but still completely lacks formal earnings insurance. We establish whether households with health insurance can cope with remaining risks related to sickness. If they cannot, then the social insurance job done by universal health coverage (UHC) is only half complete. If they can, then the value of additional formal insurance is low because it would crowd out informal mechanisms that are sufficient to cope with the residual risks.

We address this question by studying the economic impact of health shocks in Thailand, a country that took a major stride towards UHC in 2001 by extending access to reasonably comprehensive medical care benefits to the four fifths of the population not insured through formal sector employment.1 This landmark reform has been shown to have increased access to medical care sufficiently to produce marked gains in population health (Gruber et al., 2014, Limwattananon et al., 2015) and to have substantially reduced out-of-pocket payments for health care (Limwattananon et al., 2015). But earnings insurance in Thailand, as elsewhere, remains restricted to formal sector employees who form less than a third of the workforce. We estimate the impact of illness on earnings in the other two-thirds of the Thai workforce that is on the margins of poverty and assess the extent to which households can deploy informal insurance mechanisms to maintain expenditure on goods other than medical care despite experiencing a drop in income.

Legislating UHC does not necessarily eliminate the threat to living standards generated by medical expenditure risk. The health system may be overstretched by the attempt to cover a large fraction of the population for a wide array of health services from a very limited budget. Those entitled to health care in the public sector may choose to pay out-of-pocket (OOP) in order to avoid queues, obtain off-list medicines and access care considered to be of higher quality. In Thailand, universal coverage is estimated to have reduced OOP health payments by 28% – well short of 100% (Limwattananon et al., 2015). Our second contribution is to determine the medical expenditure risk that can remain after the introduction of UHC by estimating the extent to which OOP payments increase when illness strikes a Thai informal sector worker who, at least in principle, has cover for public health care.

We use panel data to identify informal sector workers who are struck by the onset of a new health condition and compare their earnings, OOP medical expenses and non-medical expenditures with those of other informal workers who do not experience any deterioration in health. To increase the plausibility of our main identification strategy, which relies on the assumption that the health change is exogenous conditional on the lagged value of the outcome, we select only workers who are initially healthy. We demonstrate robustness to an alternative identification strategy that assumes illness is exogenous conditional on an individual fixed effect.

Our third contribution is to estimate a dose-response relationship by discriminating between the effects of health changes of varying degrees of severity that are distinguished by the extent of any drop in reported health that occurs simultaneous to the onset of the health condition. We demonstrate that greater drops in reported health are observed for more serious conditions. We are aware of only one other study that has examined heterogeneity in the economic impact of illness by severity (Gertler & Gruber, 2002).

Evidence from low- and middle-income countries (LMICs) on the ability of households to smooth consumption in the short term over the economic impact of illness is mixed.2 Townsend (1994) finds that households in three southern Indian villages can smooth consumption. But this may be due to use of a health measure (sickness days) that does not discriminate major from minor illness. Some later studies that examine more serious health events reject complete smoothing of consumption (Gertler and Gruber, 2002, Wagstaff, 2007, Mete et al., 2008, Gertler et al., 2009). The validity of this finding rests on the assumed exogeneity of the health measures used. Two studies with designs intended to deal with endogeneity do not reject full consumption smoothing (Genoni, 2012, Mohanan, 2013).3 Recently, attention has shifted to how households smooth consumption over illness. If short-term shocks can be absorbed only by depleting long-term productive capacity, for example, by reducing investment in education, or if risk exposure and aversion causes households to divert valuable resources to self-insurance, then ability to keep consumption constant while health fluctuates does not imply that there are no gains to be had from social insurance (Chetty & Looney, 2006). Mohanan (2013) finds that in Karnataka (India) the accumulation of debt is the primary means of maintaining expenditures on food and housing while incurring large medical expenses resulting from health shocks. Mitra, Palmer, Mont, and Groce (2016) find that Vietnamese households experiencing ill-health and hospitalization are able to hold their non-medical expenditure constant, but they spend less on education. Liu (2016) finds a similar result for Chinese households that do not have health insurance. Those that do are able to smooth their consumption over health shocks without sacrificing their children’s education.

Most of this literature examines policy environments in which there is little or no formal insurance of health risks. We contribute evidence on the extent to which poor and near-poor Thai households continue to be economically vulnerable to illness after health care coverage is universally provided. The studies of Vietnam (Mitra et al., 2016) and China (Liu, 2016) do examine households’ ability to cope with the economic impact of health shocks in contexts where there is some formal health insurance. But compared with Thailand, coverage in these countries during the periods studied was either less universal across the population (Vietnam) or less comprehensive with respect to health services included in the benefit package (China). Informal sector workers in Vietnam who were insufficiently poor to qualify for a tax-financed scheme were covered only if they enrolled voluntarily. And enrolment was very low.4 The Chinese insurance scheme studied by Liu was restricted to covering mainly inpatient treatments during the period examined.

We find that the onset of a new health condition reduces employment and hours of work. Effects rise substantially with greater drops in reported health. Consistent with previous studies using data from Indonesia (Genoni, 2012) and China (Liu, 2016), we find that reduced labor supply of the person struck by illness is partially compensated by increased work effort of other adults in the household. Fadlon and Nielsen (2018) argue that such intra-household substitution of labor indicates gains that could be obtained from (more) social insurance. Partly as a result of the reallocation of work effort within the household, on average, there is no significant impact on household earnings. However, the most severe health conditions result in a significant 27 percent drop in earnings. Illness raises household OOP medical spending by an estimated 42 percent, although this is from a low absolute level due to universal coverage. Nevertheless, illness burdens some households with substantial medical expenses. It increases the probability that OOP spending absorbs at least a tenth of the household budget by two percentage points, and this effect rises to nine points for the most severe conditions.

Despite the substantial rise in spending on medical care and the fall in earnings resulting from the most severe health conditions, there is no significant impact on non-medical expenditure. Thai households engaged in the informal sector of the economy appear to be able to smooth their consumption over illness. This finding cannot be attributed to a failure to capture the impact of major health problems. The estimated labor supply effect of the most severe health conditions is substantially larger than the effect of a limitation on an Activity of Daily Living (ADL) estimated by Gertler and Gruber (2002) using data from Indonesia. As with other evidence that households are able to smooth consumption over health shocks (Genoni, 2012, Islam and Maitra, 2012, Mohanan, 2013, Liu, 2016, Mitra et al., 2016), interpretation depends on how households are achieving this. In addition to labor substitution within the household, we find a substantial positive impact on non-labor income, which is mainly from remittances and informal transfers from other households. As a result of the intra-household transfer of work effort and the inter-household informal transfers of money, both of which have been found to play important roles in providing protection from the economic impact of health shocks also in Indonesia (Genoni, 2012), there is no significant impact of even the most severe health conditions on total household income. This is evidence of informal insurance through mechanisms that, other than intra-household labor substitution, are considered efficient. We provide indirect evidence that suggests informal transfers are crowded out by formal government transfers. Besides informal transfers, households cope with illness-induced earnings losses and medical expenses by cutting back on saving and by borrowing. On average, the effect on borrowing is similar in magnitude to that on non-labor income. But households experiencing the most severe health conditions, as those in Karnataka studied by Mohanan (2013), rely much more heavily on borrowing. Whether or not this is efficient depends on the terms on which households are borrowing. Access to credit through microfinance programs has been shown to increase the extent to which Indonesian households can smooth consumption over health shocks (Gertler et al., 2009) and to reduce the need to sell livestock to protect consumption when hit by illness in rural Bangladesh (Islam & Maitra, 2012). But borrowing on unfavourable terms could lead to accumulating debt and eventually the sale of productive assets. Overall, our analysis suggests that in Thailand, the combination of formal health care coverage and informal insurance of the residuals risks does a reasonably good job of protecting living standards from the economic impact of illness, at least in the short term.

Our main estimates are obtained by comparing values of an outcome, such as earnings, of individuals who are struck by illness (treatment group) with the outcome values of those who remain healthy (comparison group), after conditioning on the lagged value of the outcome. They are robust to making the treatment and comparison groups observationally equivalent in terms of the pre-illness values of the outcome and covariates by applying inverse probability weights (IPW) based on the estimated propensity score of illness, and also to combining IPW with regression adjustment. In addition, the estimates are generally robust to adoption of a difference-in-differences identification strategy that rests on the counterfactual outcome being independent of illness conditional on time invariant unobservables, rather than the lagged value of the outcome.

The next section outlines the system of social protection of health risks in Thailand and describes the data. The third section explains the identification strategies and estimators employed. Results are presented in the fourth section and the final section concludes.

Section snippets

Social protection in Thailand

Social protection varies by employment sector. Separate schemes for public and private sector salaried employees each give entitlement to comprehensive health care benefits, as well as paid sick leave and medical retirement benefits (International Labour Organization, 2013, Social Security Administration, 2013). The two thirds of the workforce that does not get social protection through formal sector employment – mainly low-income households engaged in agriculture or small (family) businesses –

Identification strategy

We assume that conditional on the initial values of the outcome Yit-1 and the covariates Xit-1 (plus region-specific period effects, δrt), the onset of illness is not correlated with the counterfactual outcome that would have materialized in the absence of illness Y0it. Specifically, we identify the average effect of illness on those who succumb to illness under the following conditional independence assumption (Angrist and Pischke, 2008, Imbens and Wooldridge, 2009),EY0itYit-1,Xit-1,δrt,Illness

Results

Estimates from the LDV approach implemented by the least squares regression (2) are given in the Tables presented in this section. Estimates obtained from the alternative estimators and identification strategy are given in the online Appendix (Tables A7, A9–A11). Estimates obtained using both identification strategies without conditioning on covariates are presented in online Appendix Table A16.

Conclusion

Our analysis reveals that substantial economic risks associated with illness remain even after universal health coverage is put in place in a middle-income country. These risks do not only take the form of earnings losses that remain (formally) uninsured. There is also residual medical expenditure risk. This does not amount to failure of universal coverage in Thailand. Previous research has demonstrated that the extension of coverage to the informal sector greatly reduced out-of-pocket

Conflict of interest statement

None of the study authors have any conflicts of interest regarding the submitted study.

Acknowledgements

We thank two referees for valuable comments. The research reported in this paper was partly financed by the EU FP7 project entitled Health Equity and Financial Protection in Asia (HEALTH-F2-2009-223166). Sven Neelsen received funding from the EU FP7 Marie Curie Intra European Fellowship HEALTHEVENT (PIEF-GA-2012-331815). Owen O’Donnell was supported by the Greek General Secretariat for Research and Technology (Ministry of Education & Religion) and the Swiss Agency for Development and Cooperation

References (42)

  • A. Asfaw et al.

    Is consumption insured against illness? Evidence on vulnerability of households to health shocks in rural Ethiopia

    Economic Development and Cultural Change

    (2004)
  • I. Bonfrer et al.

    Health shocks, coping strategies and foregone healthcare among agricultural households in Kenya

    Global Public Health

    (2016)
  • J. Bound

    Self-reported versus objective measures of health in retirement models

    Journal of Human Resources

    (1991)
  • S. Dercon et al.

    In sickness and in health: Risk sharing within households in rural Ethiopia

    Journal of Political Economy

    (2000)
  • N. Duan

    Smearing estimate: A nonparametric retransformation method

    Journal of the American Statistical Association

    (1983)
  • I. Fadlon et al.

    Household labor supply and the gains from social insurance

    Journal of Public Economics

    (2018)
  • M.E. Genoni

    Health shocks and consumption smoothing: Evidence from Indonesia

    Economic Development and Cultural Change

    (2012)
  • P. Gertler et al.

    Insuring consumption against illness

    American Economic Review

    (2002)
  • P. Gertler et al.

    Do microfinance programs help families insure consumption against illness?

    Health Economics

    (2009)
  • U. Giedion et al.

    The impact of universal coverage schemes in the developing World: A review of the existing evidence

    (2013)
  • J. Gruber et al.

    The great equalizer: Health care access and infant mortality in Thailand

    American Economic Journal: Applied Economics

    (2014)
  • Cited by (0)

    View full text