Introduction

The expansion of public health systems remains a priority for governments around the world in line with the recommendations of the Sustainable Development Goals (SDGs) initiative of 2015. Policy efforts have focused on contributing to goal number three, which aims to “ensure healthy lives and promote well-being for all, at all ages”. Target 3.8 of the SDGs states: “Achieve universal health coverage, including financial risk protection, access to high-quality essential health-care services and access to essential medicine that is safe, effective, high-quality, and affordable, and vaccines for all.” Through free or subsidized health insurance programs, governments attempt to ensure that vulnerable populations obtain access to health services.

Once individuals have gained access to publicly provided insurance, however, not only health, but other socioeconomic variables such as savings and consumption are likely to be affected. Lessons from the literature suggest that an insurance policy helps to reduce out-of-pocket and catastrophic expenditure (Barros, 2008[3] Barros, R. (2008). Wealthier but not much Healthier: Effects of a Health insurance Program for the poor in Mexico. Stanford Institute for Economic Policy Research.
; Doubova et al., 2015[24] Doubova, S., Perez-Cuevas, R., Canning, D., & Reich, M. (2015). Access to healthcare and financial risk protection for older adults in Mexico: secondary data analysis of a national survey. BMJ Open. doi: 10.1136/bmjopen-2015-007877
; Galarraga et al., 2010[32] Galarraga, O., Sosa-Rubí, S., Salinas-Rodriguez, A., & Sesma-Vazquez, S. (2010). Health insurance for the poor: impact on catastrophic and out-of-pocket health expenditures in Mexico. Springer. doi: 10.1007/s10198-009-0180-3
; Grogger et al., 2014[38] Grogger, J., Arnold, T., Sofia Leon, A., & Ome, A. (2014). Heterogeneity in the effect of public health insurance on catastrophic out-of-pocket health expenditures: the case of Mexico. Oxford University Press in association with The London School of Hygiene and Tropical Medicine. doi: 10.1093/heapol/czu037
; Knaul et al., 2006[52] Knaul, F. M., Arreola-Ornelas, H., Mendez-Carniado, O., Bryson-Cahn, C., Barofsky, J., Maguire, R., . . . Sesma, S. (2006). Evidence is good for your health system: policy reform to remedy catastrophic and impoverishing health spending in Mexico. Lancet. doi: 10.1016/S0140-6736(06)69565-2
; Knox, 2008[54] Knox, M. (2008). Health Insurance for All: An Evaluation of Mexico's Seguro Popular Program. Research gate.
; Leininger et al., 2010[56] Leininger, L., Levy, H., & Schanzenbach, D. (2010). Consequences of SCHIP Expansions for Household Well-Being. Forum for Health Economics & Policy, Article 3.
; Sommers et al., 2017[76] Sommers, B. D., Maylone, B., Blendon, R. J., Orav, J., & Epstein, A. M. (2017). Three-Year Impacts Of The Affordable Care Act: Improved Medical Care And Health Among Low-Income Adults. Health Affairs. doi: 10.1377/hlthaff.2017.0293
; Sosa-Rubi et al., 2011[78] Sosa-Rubi, S., Salinas-Rodriguez, A., & Galarraga, O. (2011). Impacto del Seguro Popular en el gasto catastrófico y de bolsillo en el México rural y urbano, 2005-2008. Salud Publica Mex. Retrieved from https://www.saludpublica.mx/index.php/spm/article/view/5066
), lessens the need to maintain precautionary savings either financial or asset-based (Chou et al., 2003[11] Chou, S.-Y., Liu, J.-T., & Hammitt, J. K. (2003). National Health Insurance and precautionary saving: evidence from Taiwan. Journal of Public Economics. doi: 10.1016/S0047-2727(01)00205-5
; Chou et al., 2004[12] Chou, S.-Y., Liu, J.-T., & Huang, C. J. (2004). Health insurance and savings over the life cycle—a semiparametric smooth coefficient estimation. Journal of Applied Econometrics. doi: 10.1002/jae.735
; Wagstaff & Pradhan, 2005[82] Wagstaff, A., & Pradhan, M. (2005). Health Insurance Impacts on Health and Nonmedical Consumption in a Developing Country. Washington: World Bank Policy Research Working Paper 3563, World Bank.
), and alters consumption levels (Cheung & Padieu, 2015[9] Cheung, D., & Padieu, Y. (2015). Heterogeneity of the Effects of Health Insurance on Household Savings: Evidence from Rural China. World Development. doi: 10.1016/j.worlddev.2014.08.004
; Gruber & Yelowitz, 1999[39] Gruber, J., & Yelowitz, A. (1999). Public Health Insurance and Private Savings. Journal of Political Economy. doi: 10.1086/250096
; Leininger et al., 2010[56] Leininger, L., Levy, H., & Schanzenbach, D. (2010). Consequences of SCHIP Expansions for Household Well-Being. Forum for Health Economics & Policy, Article 3.
). In line with this literature, we investigate possible changes in the consumption of food.

The final effect of the insurance programs on dietary choices is unknown a priori as an increased level of nutritional awareness or an expansion of disposable income derived from gaining access to medical assistance could lead to the transition to a more nutritionally enriched diet, but it could also incentivize individuals to reduce preventive care habits, including adding unhealthy products to their regular diet, creating a usual case of health hazard, as a result, the health status of the beneficiaries would not necessarily improve, but they would become more reliant on publicly provided medical care, which, under extreme circumstances, could trigger the collapse of public health systems.

Some high-income countries have created public insurance programs with relative success. For example, in Canada, Australia, Finland, Sweden, Norway, and Germany, at least 80% of the population is covered through social protection schemes ¹ According to the Universal Health Coverage (UHC) Service Coverage Index (SCI) of 2019, as reported in the SDGs monitor indicators by WHO (2021). . On the other hand, there are examples in low and middle-income countries such as Mexico, Guatemala, Nicaragua, Nigeria, Costa Rica, Ghana, and Colombia, where similar programs have produced moderate coverage results.

Rather than studying factors that determine the success or lack thereof of these programs, which are complex and multifaceted such as the quality, type, and number of health services provided (Buchmueller et al., 2005[5] Buchmueller, T., Grumbach, K., Kronick, R., & Kahn, J. (2005). The Effect of Health Insurance on Medical Care Utilization and Implications for Insurance Expansion: A Review of the Literature. Med Care Res Rev. 2005 Feb;62(1):3-30. doi: 10.1177/1077558704271718
; Chen et al., 2007[7] Chen, L., Yip, W., Chang, M., Lin, H., Lee, S., Chiu, Y., & Lin, Y. (2007). The effects of Taiwan's National Health Insurance on access and health status of the elderly. Health Econ. doi: 10.1002/hec.1160
; Currie & Gruber, 1996[20] Currie, J., & Gruber, J. (1996). Health Insurance Eligibility, Utilization of Medical Care, and Child Health. The Quarterly Journal of Economics. doi: 10.2307/2946684
; Escobar et al., 2010[27] Escobar, M.-L., Griffin, C., & Shaw, R. (2010). The Impact of Health Insurance in Low- and Middle-Income Countries. Washington, DC: Brookings Institution Press. Retrieved from http://www.jstor.org/stable/10.7864/j.ctt12638q
; Finkelstein et al., 2012[30] Finkelstein, A., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J., . . . Group, O. H. (2012). The Oregon Health Insurance Experiment: Evidence From The First Year. The Quarterly Journal, Q J Econ. doi: 10.1093/qje/qjs020
; Ghosh et al., 2017[34] Ghosh, A., Simon, K., & Sommers, B. D. (2017). The Effect of State Medicaid Expansions on Prescription Drug Use: Evidence from the Affordable Care Act. National Bureau of Economic Research. doi: 10.3386/w23044
; Guindon, 2014[40] Guindon, G. (2014). The impact of health insurance on health services utilization and health outcomes in Vietnam. Health Econ Policy Law. doi: 10.1017/S174413311400005X
; Hadley, 2003[42] Hadley, J. (2003). Sicker and poorer--the consequences of being uninsured: a review of the research on the relationship between health insurance, medical care use, health, work, and income. Med Care Res Rev. 2003 Jun;60(2 Suppl):3S-75S; discussion 76S-112S. doi: 10.1177/1077558703254101
; Knox, 2008[54] Knox, M. (2008). Health Insurance for All: An Evaluation of Mexico's Seguro Popular Program. Research gate.
; Parker et al., 2018[66] Parker, S. W., Saenz, J., & Wong, R. (2018). Health Insurance and the Aging: Evidence From the Seguro Popular Program in Mexico. Demography. doi: 10.1007/s13524-017-0645-4
; Sommers et al., 2017[76] Sommers, B. D., Maylone, B., Blendon, R. J., Orav, J., & Epstein, A. M. (2017). Three-Year Impacts Of The Affordable Care Act: Improved Medical Care And Health Among Low-Income Adults. Health Affairs. doi: 10.1377/hlthaff.2017.0293
; Sosa-Rubi et al., 2009[77] Sosa-Rubi, S. G., Galarraga, O., & Lopez-Ridaura, R. (2009). Diabetes treatment and control: the effect of public health insurance for the poor in Mexico. Bull World Health Organ. doi: 10.2471/BLT.08.053256
; Trujillo et al., 2005[80] Trujillo, A., Portillo, J., & Vernon, J. (2005). The Impact of Subsidized Health Insurance for the Poor: Evaluating the Colombian Experience Using Propensity Score Matching. Int J Health Care Finance Econ. doi: 10.1007/s10754-005-1792-5
; Wagstaff & Pradhan, 2005[82] Wagstaff, A., & Pradhan, M. (2005). Health Insurance Impacts on Health and Nonmedical Consumption in a Developing Country. Washington: World Bank Policy Research Working Paper 3563, World Bank.
; Wagstaff et al., 2009[83] Wagstaff, A., Lindelow, M., Jun, G., Ling, X., & Juncheng, Q. (2009). Extending health insurance to the rural population: an impact evaluation of China's new cooperative medical scheme. Health Econ. doi: 10.1016/j.jhealeco.2008.10.007
), the focus here is on examining the dietary choices of the beneficiaries before and after they received access to health services. An increase in unhealthy food consumption in a regular diet may indicate early signs of moral hazard.

We study the case of rural Mexico for three reasons. First, despite having a rich history of public health systems, its primary institutions were designed to serve the working population registered by a formal employer (mainly through IMSS and ISSSTE), the more vulnerable, those who were not eligible for any other social security program (excluding those able to afford private health insurance), were only offered access in 2004, when the Mexican Federal government launched Seguro Popular. The program registered 5.3 million beneficiaries at its onset, covering about 38% of Mexican municipalities. By the end of 2019, when Seguro Popular was reformed and replaced, the number of beneficiaries had grown to 51 million, covering 99.8% of municipalities. Taking advantage of the initial structure of the program we study the food choices of households in the municipalities where the program began.

Second, Mexico’s rates of overweight and obesity have remained the highest in the world for decades. In 2002, 57.1% of the adult population was overweight or obese. At the time of Seguro Popular, in 2004, it was at 58.3% (WHO, Global Health Observatory). Data from 2018, indicates that the prevalence of being overweight and being obese, for the population aged 20 years and more, has reached 39.1% and 36.1%, respectively, that is, 75.2% in total ² Estimated using the Mexican National Health and Nutrition Survey (ENSANUT) of 2018. . Results from this research contribute to understanding how publicly provided programs such as Seguro Popular may influence the population’s health status via food choices.

Third, when concentrating on vulnerable populations, the rural poor become a natural focus. Just prior to the implementation of Seguro Popular, 50% of the Mexican population lived in poverty ³ This is defined as insufficient income to acquire a basic basket of food and meet the necessary expenses on healthcare, clothing, housing, transport, and education despite the entire household income being used to acquire these goods and services. , and the income of 20% of the total population was insufficient to acquire a basic basket of food⁴ This is according to CONEVAL, the Mexican agency in charge of measuring poverty and evaluating it by different income dimensions. . Poverty remained rampant, and by 2004, 47.2% of the Mexican population lived in poverty, while 17.4% still could not afford a basic food basket. More recent data show that between 2018 and 2020, the population living in poverty conditions increased from 51.9 to 55.7 million, which is about 43.9% of the population. Of particular interest are the rural poor, of whom 56.8% lived in poverty in 2020, and about 75% had no access to health services or social security (CONEVAL, 2021[15] CONEVAL. (2021). Consejo Nacional de Evaluación de la Política de Desarrollo Social. Retrieved from Medición de la Pobreza, Pobreza en México: https://www.coneval.org.mx/Medicion/Paginas/PobrezaInicio.aspx
). Moreover, evidence suggests that these communities allocate a higher percentage of their resources toward unhealthy foods than those in urban regions (ENSANUT, 2020[49] Instituto Nacional de Salud Pública. (2020). Encuesta Nacional de Salud y Nutrición 2018-19, Resultados nacionales. Cuernavaca, México: Instituto Nacional de Salud Pública. Retrieved from https://ensanut.insp.mx/encuestas/ensanut2018/informes.php
). Given the issues of malnutrition outlined above, this exacerbates their vulnerability status and the urgency to better comprehend their dietary choices.

We use the Mexican National Survey of Income and Expenditure (ENIGH) of 2002 and 2004 to quantify and classify the total food expenditure of the rural household into nine categories: (1) animal protein, (2) cereals, (3) fruit and vegetables, (4) milk and derivatives, (5) processed sugars, (6) oils and fats, (7) alcoholic beverages and tobacco, (8) food consumed outside the household, and the remainder is grouped in (9) others. Through Difference-in-Difference estimations we compare the expenditure on these categories of rural households located in states where Seguro Popular started with that of rural households within the same states that were without the program, before and after the implementation. We find strong evidence that Seguro Popular increased the consumption expenditure on the categories of oils and fats, and processed sugars.

The rest of this document is organized as follows, the next section provides a brief background on Seguro Popular; section 3 presents a review of the literature that examines similar issues; section 4 characterizes the conceptual framework behind household decision-making and gathers the estimation strategy regarding the data and model; section 5 discusses the results; and section 6 concludes.

1. Background on Seguro Popular

Seguro Popular was launched in 2004 with the aim of providing financial protection to the population lacking social security and access to health care by incorporating them into a public and voluntary insurance program. In 2002 the uninsured segment accounted for approximately 57.8% of the country’s total population ⁵ Official Journal of the Federation (2002). . At its onset, the program was implemented only in selected regions of a few states, namely, Colima, Jalisco, Aguascalientes, Tabasco, and Campeche. These areas were chosen based on specific criteria related to their capacity to offer health services.

The only requirement to get enrolled was that one had not already signed up for another social security program. By joining, the beneficiary would commit to adhere to the operation rules of the program (2002), which primarily tried to encourage the insured to adopt health promotion and disease prevention behaviors. In practice, however, there were no enforceable mechanisms in place.

The program was largely financed by the federal government through annual contributions, which were determined by three parameters. The first parameter was a social fee or quota, calculated as a percentage of a daily general minimum wage in the labor market, based on the individual’s income level or decile. The second parameter was a Federal Solidarity Contribution, which represented at least one and a half times the amount of the social fee. The third parameter was a State Solidarity Contribution, which equaled at least half the social fee per affiliated person.

The interventions covered by the insurance were defined in the Universal Catalog of Health Services (CAUSES). The number of interventions was adjusted annually based on priority criteria and the structural capacity of the state public health network, health centers, and general hospitals. In 2019, the Seguro Popular program was replaced by the Instituto de Salud para el Bienestar (INSABI). That year, Seguro Popular guaranteed access to 294 interventions, including 1,807 medical diagnostics, 618 medical procedures, 633 medicines, and 37 medical supplies for general and specialized treatment, urgencies, general surgery, and obstetric consultations.

2. Literature Review

The effects of offering public health insurance have been explored in a variety of outcomes, here we describe first some studies that focus on utilization of health facilities as we have presumed the beneficiaries indeed take advantage of the access to a health service gained once they have been insured. Next we examined studies focused on the relationship with financial variables such as savings and consumption, exploring whether the literature supports the claim that financial decisions are indeed altered. Finally, we describe works that focus on the effects on food consumption and nutritional choices.

A review of the existing literature about the relationship between health insurance and households’ medical use for high-income countries can be found in Hadley (2003)[42] Hadley, J. (2003). Sicker and poorer--the consequences of being uninsured: a review of the research on the relationship between health insurance, medical care use, health, work, and income. Med Care Res Rev. 2003 Jun;60(2 Suppl):3S-75S; discussion 76S-112S. doi: 10.1177/1077558703254101
and Buchmueller et al. (2005)[5] Buchmueller, T., Grumbach, K., Kronick, R., & Kahn, J. (2005). The Effect of Health Insurance on Medical Care Utilization and Implications for Insurance Expansion: A Review of the Literature. Med Care Res Rev. 2005 Feb;62(1):3-30. doi: 10.1177/1077558704271718
, concluding that, overall, the studies consistently report positive and significant impacts of insurance on measures of utilization. Other studies have also shown that health insurance increases health care utilization in adults and children (Currie & Gruber, 1996[20] Currie, J., & Gruber, J. (1996). Health Insurance Eligibility, Utilization of Medical Care, and Child Health. The Quarterly Journal of Economics. doi: 10.2307/2946684
; Finkelstein et al., 2012[30] Finkelstein, A., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J., . . . Group, O. H. (2012). The Oregon Health Insurance Experiment: Evidence From The First Year. The Quarterly Journal, Q J Econ. doi: 10.1093/qje/qjs020
; Ghosh et al., 2017[34] Ghosh, A., Simon, K., & Sommers, B. D. (2017). The Effect of State Medicaid Expansions on Prescription Drug Use: Evidence from the Affordable Care Act. National Bureau of Economic Research. doi: 10.3386/w23044
; Sommers et al., 2017[76] Sommers, B. D., Maylone, B., Blendon, R. J., Orav, J., & Epstein, A. M. (2017). Three-Year Impacts Of The Affordable Care Act: Improved Medical Care And Health Among Low-Income Adults. Health Affairs. doi: 10.1377/hlthaff.2017.0293
). The evidence from low- and middle-income countries such as China, Colombia, Taiwan, and Vietnam, indicates that insurance programs have increased outpatient and inpatient utilization in rural and impoverished households (Chen et al., 2007[7] Chen, L., Yip, W., Chang, M., Lin, H., Lee, S., Chiu, Y., & Lin, Y. (2007). The effects of Taiwan's National Health Insurance on access and health status of the elderly. Health Econ. doi: 10.1002/hec.1160
; Guindon, 2014[40] Guindon, G. (2014). The impact of health insurance on health services utilization and health outcomes in Vietnam. Health Econ Policy Law. doi: 10.1017/S174413311400005X
; Trujillo et al., 2005[80] Trujillo, A., Portillo, J., & Vernon, J. (2005). The Impact of Subsidized Health Insurance for the Poor: Evaluating the Colombian Experience Using Propensity Score Matching. Int J Health Care Finance Econ. doi: 10.1007/s10754-005-1792-5
; Wagstaff & Pradhan, 2005[82] Wagstaff, A., & Pradhan, M. (2005). Health Insurance Impacts on Health and Nonmedical Consumption in a Developing Country. Washington: World Bank Policy Research Working Paper 3563, World Bank.
; Wagstaff et al., 2009[83] Wagstaff, A., Lindelow, M., Jun, G., Ling, X., & Juncheng, Q. (2009). Extending health insurance to the rural population: an impact evaluation of China's new cooperative medical scheme. Health Econ. doi: 10.1016/j.jhealeco.2008.10.007
).

Mixing results come from evaluations of the Mexican Seguro Popular. Rivera-Hernandez et al.’s (2019)[70] Rivera-Hernandez, M., Rahman, M., & Galarraga, O. (2019). Preventive healthcare-seeking behavior among poor older adults in Mexico: the impact of Seguro Popular, 2000-2012. Salud Publica Mex. doi: 10.21149/9185
reported that Seguro Popular had no significant effect on the use of preventive services (including screening for diabetes, hypertension, breast cancer, and cervical cancer) among adults aged 50 to 75 years, while Sosa-Rubi et al. (2009)[77] Sosa-Rubi, S. G., Galarraga, O., & Lopez-Ridaura, R. (2009). Diabetes treatment and control: the effect of public health insurance for the poor in Mexico. Bull World Health Organ. doi: 10.2471/BLT.08.053256
found that adults with diabetes who were enrolled had significantly more access to blood glucose control tests compared to uninsured adults. Knox (2008)[54] Knox, M. (2008). Health Insurance for All: An Evaluation of Mexico's Seguro Popular Program. Research gate.
also found increased health care utilization, especially in health center visits and hospitalization, and decreased usage of private care providers such as private doctors and pharmacies. Parker et al. (2018)[66] Parker, S. W., Saenz, J., & Wong, R. (2018). Health Insurance and the Aging: Evidence From the Seguro Popular Program in Mexico. Demography. doi: 10.1007/s13524-017-0645-4
investigated how the program affected the use of health services and diagnostic tests among population aged 50 and older, using the longitudinal Mexican Health and Aging Study from 2001 to 2012. They examined how the impact of the program varied depending on the availability of health services before the program started and the evidence indicates notable disparities in the effects of Seguro Popular, depending on how accessible health services were. Findings imply that the population with greater access to health services experiences more substantial and widespread benefits when there is the presence of an illness.

A large body of literature has found that reductions of catastrophic and out-of-pocket health expenditures result from the implementation of health insurance programs (Barros, 2008[3] Barros, R. (2008). Wealthier but not much Healthier: Effects of a Health insurance Program for the poor in Mexico. Stanford Institute for Economic Policy Research.
; Doubova et al., 2015[24] Doubova, S., Perez-Cuevas, R., Canning, D., & Reich, M. (2015). Access to healthcare and financial risk protection for older adults in Mexico: secondary data analysis of a national survey. BMJ Open. doi: 10.1136/bmjopen-2015-007877
; Galarraga et al., 2010[32] Galarraga, O., Sosa-Rubí, S., Salinas-Rodriguez, A., & Sesma-Vazquez, S. (2010). Health insurance for the poor: impact on catastrophic and out-of-pocket health expenditures in Mexico. Springer. doi: 10.1007/s10198-009-0180-3
; Grogger et al., 2014[38] Grogger, J., Arnold, T., Sofia Leon, A., & Ome, A. (2014). Heterogeneity in the effect of public health insurance on catastrophic out-of-pocket health expenditures: the case of Mexico. Oxford University Press in association with The London School of Hygiene and Tropical Medicine. doi: 10.1093/heapol/czu037
; Knaul et al., 2006[52] Knaul, F. M., Arreola-Ornelas, H., Mendez-Carniado, O., Bryson-Cahn, C., Barofsky, J., Maguire, R., . . . Sesma, S. (2006). Evidence is good for your health system: policy reform to remedy catastrophic and impoverishing health spending in Mexico. Lancet. doi: 10.1016/S0140-6736(06)69565-2
; Knox, 2008[54] Knox, M. (2008). Health Insurance for All: An Evaluation of Mexico's Seguro Popular Program. Research gate.
; Leininger et al., 2010[56] Leininger, L., Levy, H., & Schanzenbach, D. (2010). Consequences of SCHIP Expansions for Household Well-Being. Forum for Health Economics & Policy, Article 3.
; Sommers et al., 2017[76] Sommers, B. D., Maylone, B., Blendon, R. J., Orav, J., & Epstein, A. M. (2017). Three-Year Impacts Of The Affordable Care Act: Improved Medical Care And Health Among Low-Income Adults. Health Affairs. doi: 10.1377/hlthaff.2017.0293
; Sosa-Rubi et al., 2011[78] Sosa-Rubi, S., Salinas-Rodriguez, A., & Galarraga, O. (2011). Impacto del Seguro Popular en el gasto catastrófico y de bolsillo en el México rural y urbano, 2005-2008. Salud Publica Mex. Retrieved from https://www.saludpublica.mx/index.php/spm/article/view/5066
), the probability that households will incur impoverishing expenditures also lowers (Knaulet al., 2018[53] Knaul, F., Arreola-Ornelas, H., Wong, R., Lugo-Palacios, D., & Mendez-Carniado, O. (2018). Efecto del Seguro Popular de Salud sobre los gastos catastróficos y empobrecedores en México, 2004-2012. Salud Publica Mex. doi: 10.21149/9064
). Consequently, the disposable income of the newly insured might rise; the evidence points out that health insurance reduces uncertainty, enabling households to reduce precautionary savings (Chou, Liu, & Huang, 2004[12] Chou, S.-Y., Liu, J.-T., & Huang, C. J. (2004). Health insurance and savings over the life cycle—a semiparametric smooth coefficient estimation. Journal of Applied Econometrics. doi: 10.1002/jae.735
; Wagstaff & Pradhan, 2005[82] Wagstaff, A., & Pradhan, M. (2005). Health Insurance Impacts on Health and Nonmedical Consumption in a Developing Country. Washington: World Bank Policy Research Working Paper 3563, World Bank.
). For example, the Medicaid program in the United States led to a reduction in savings and an increase in consumption (Gruber & Yelowitz, 1999[39] Gruber, J., & Yelowitz, A. (1999). Public Health Insurance and Private Savings. Journal of Political Economy. doi: 10.1086/250096
; Leininger et al., 2010[56] Leininger, L., Levy, H., & Schanzenbach, D. (2010). Consequences of SCHIP Expansions for Household Well-Being. Forum for Health Economics & Policy, Article 3.
). Chou et al. (2003)[11] Chou, S.-Y., Liu, J.-T., & Hammitt, J. K. (2003). National Health Insurance and precautionary saving: evidence from Taiwan. Journal of Public Economics. doi: 10.1016/S0047-2727(01)00205-5
studied the effect of health insurance on households’ precautionary savings using Taiwan’s 1995 introduction of National Health Insurance and found a reduction in savings by an average of 8.6–13.7%.

Cheung and Padieu (2015)[9] Cheung, D., & Padieu, Y. (2015). Heterogeneity of the Effects of Health Insurance on Household Savings: Evidence from Rural China. World Development. doi: 10.1016/j.worlddev.2014.08.004
pointed out that the New Cooperative Medical Scheme’s (NCMS) allowed households to lower savings and boost consumption in rural China. Kirdruanga and Glewwe (2018)[51] Kirdruang, P., & Glewwe, P. (2018). The Impact of Universal Health Coverage on Households’ Consumption and Savings in Thailand. J Asia Pac Econ. doi: 10.1080/13547860.2017.1359893
studied the impact of Thailand’s Universal Health Coverage Scheme (UCS) on households’ savings, and they found that, in the short run, the UCS had little or no impact on either households’ savings or households’ consumption expenditures. No effect on savings was found in the long run (unless savings is defined to include consumption of durable goods). The increased disposable income can also be associated with changes in labor supply. The literature has produced mixed results, depending on gender, age, and other specific socioeconomic characteristics. Contractions in labor supply can be found in Knox (2008)[54] Knox, M. (2008). Health Insurance for All: An Evaluation of Mexico's Seguro Popular Program. Research gate.
and Chou and Staiger (2001)[13] Chou, Y., & Staiger, D. (2001). Health insurance and female labor supply in Taiwan. J Health Econ. doi: 10.1016/s0167-6296(00)00075-8
, while evidence of increases can be found in Garthwaite et al. (2014)[33] Garthwaite, C., Gross, T., & Notowidigdo, M. J. (2014). Public Health Insurance, Labor Supply, and Employment Lock. The Quarterly Journal of Economics. doi: 10.1093/qje/qju005
and Valle (2014)[23] del Valle, A. (2014). From caring to work: The labor market effects of noncontributory health insurance. Job Market Paper.
.

The evidence described so far generally supports the notion that providing health insurance fosters the use of health facilities, reduces catastrophic and out-of-pocket health expenditures, and decreases precautionary savings. We now explore the literature that offers insights into how consumption decisions are altered. Gruber and Yelowitz (1999)[39] Gruber, J., & Yelowitz, A. (1999). Public Health Insurance and Private Savings. Journal of Political Economy. doi: 10.1086/250096
documented that eligibility to the program Medicaid in the USA, was strongly associated with consumption expenditures. Leininger et al. (2010)[56] Leininger, L., Levy, H., & Schanzenbach, D. (2010). Consequences of SCHIP Expansions for Household Well-Being. Forum for Health Economics & Policy, Article 3.
focused on studying the Children’s Health Insurance Program (CHIP), which provides health coverage to eligible children through Medicaid, using the Consumer Expenditure Survey (from 1996 to 2002) they found that eligibility for CHIP is associated with an increase in overall expenditure, most of which is allocated to consumption of basic needs (housing, food, and transportation).

Evidence from low- and middle-income countries shows comparable results. Wagstaff and Pradhan (2005)[82] Wagstaff, A., & Pradhan, M. (2005). Health Insurance Impacts on Health and Nonmedical Consumption in a Developing Country. Washington: World Bank Policy Research Working Paper 3563, World Bank.
studied the effects of the introduction of Vietnam’s Health Insurance (VHI) program on health outcomes and nonmedical household consumption. Using propensity score matching with a double-difference estimator ( representing households with partial or full family coverage), they found that the program increased nonmedical household consumption, including food consumption. The program also impacted favorably on the height-for-age and weight-for-age of young school children and the body mass index among adults.

Kirdruanga and Glewwe (2018)[51] Kirdruang, P., & Glewwe, P. (2018). The Impact of Universal Health Coverage on Households’ Consumption and Savings in Thailand. J Asia Pac Econ. doi: 10.1080/13547860.2017.1359893
studied Thailand’s Universal Health Coverage Scheme (UCS) on households’ consumption using data from the Socio-Economic Survey (SES) and the Health and Welfare Survey (HWS). They found evidence of increased consumption, especially of durable goods, over time (from 2001 to 2007). The UCS’s increased consumption was identified as both an income effect (by reducing out-of-pocket medical costs) and a risk reduction effect.

Analysis of the New Cooperative Medical Scheme (NCMS) in rural China has also shown that consumption increases among insured individuals (Bai & Wu, 2014[1] Bai, C.-E., & Wu, B. (2014). Health insurance and consumption: Evidence from China’s New Cooperative Medical Scheme. Journal of Comparative Economics. doi: 10.1016/j.jce.2013.07.005
; Cheung & Padieu, 2015[9] Cheung, D., & Padieu, Y. (2015). Heterogeneity of the Effects of Health Insurance on Household Savings: Evidence from Rural China. World Development. doi: 10.1016/j.worlddev.2014.08.004
; Zhao, 2018[91] Zhao, W. (2018). Does health insurance promote people's consumption? New evidence from China. China Economic Review. doi: 10.1016/j.chieco.2018.08.007.
). Using data from the China Health and Nutrition Survey (CHNS), Cheung and Padieu (2015)[9] Cheung, D., & Padieu, Y. (2015). Heterogeneity of the Effects of Health Insurance on Household Savings: Evidence from Rural China. World Development. doi: 10.1016/j.worlddev.2014.08.004
showed that higher middle-income participants tended to reduce their savings and increase their consumption. For the poorest households, however, they found no effects, likely due to their considerable dissaving and borrowing constraints, as their consumption expenditures were higher than their average income. The share of the food consumption budget was estimated at around 145%.

Zhao (2018)[91] Zhao, W. (2018). Does health insurance promote people's consumption? New evidence from China. China Economic Review. doi: 10.1016/j.chieco.2018.08.007.
studied the specific impact of the critical illness insurance (CII), an expansion of the NCMS program, on the consumption of rural households and found that the CII increased per capita daily household consumption by 15%. The study also identified heterogeneity in the consumption smoothing effects of CII across households of different income levels as the policy exacerbated consumption inequality among rural households.

Panchalingam (2020)[65] Panchalingam, T. (2020). Effects of Public Health Insurance Expansions on the Non-Healthcare Consumption Expenditures of Low-Income Households. The Ohio State University, 44. doi: 10.2139/ssrn.3740775
examined the Medicaid expansion program, focusing on the patterns of non-healthcare consumption of insured households. The author found that eligible families spent less on fresh food per adult and more on health and beauty products. He et al. (2020)[43] He, X., Lopez, R., & Boehm, R. (2020). Medicaid expansion and non-alcoholic beverage choices by low-income households. Health Economics. doi: 10.1002/hec.4133
investigated the impact of the 2010 Patient Protection and the Affordable Care Act (ACA) on non-alcoholic beverage choices in low-income households. Their results indicate that diet-carbonated soft drinks and bottled water purchases increased, while carbonated soft drinks, fruit juice, fruit drinks, milk, and tea remained constant. They also found that the policy decreased sugar purchases and increased purchases of non-alcoholic beverage products with lower sugar content.

Given the changes caused in consumption, health has also been associated with changes in obesity and overweight rates. Studies based on the Affordable Care Act (ACA) show mixed results, while some studies find that overweight and obesity rates decrease (Barbaresco et al., 2015[2] Barbaresco, S., Courtemanche, C. J., & Qi, Y. (2015). Impacts of the Affordable Care Act dependent coverage provision on health-related outcomes of young adults. Journal of Health Economics. doi: 10.1016/j.jhealeco.2014.12.004
; Courtemanche & Zapata, 2013[19] Courtemanche, C., & Zapata, D. (2013). Does Universal Coverage Improve Health? The Massachusetts Experience. Journal of Policy Analysis and Management. doi: 10.1002/pam.21737
; Rhubart, 2018[69] Rhubart, D. C. (2018). Disparities in individual health behaviors between medicaid expanding and non-expanding states in the U.S. SSM - Population Health, 36-43. doi: 10.1016/j.ssmph.2018.08.005
). There are also findings that body mass index and obesity tend to increase (Bhattacharya et al., 2009[4] Bhattacharya, J., Bundorf, K., Pace, N., & Sood, N. (2009). Does Health Insurance Make You Fat? National Bureau of Economic Research. doi:10.3386/w15163
). Bhattacharya et al. (2009)[4] Bhattacharya, J., Bundorf, K., Pace, N., & Sood, N. (2009). Does Health Insurance Make You Fat? National Bureau of Economic Research. doi:10.3386/w15163
argued that health insurance induces a moral hazard effect by weakening incentives to lose weight. The moral hazard effects on the behavior of insured households have also been examined by Rashad and Markowitz (2009, 2010), who found that having insurance is associated with a higher body mass index but not with a higher probability of being obese.

Evidence from less developed countries is more specific on consumption across food groups. Fan et al. (2021)[28] Fan, H., Yan, Q., Liu, S., Cai, J., & Coyte, P. C. (2021). Childhood Nutrition in Rural China: What Impact Does Public Health Insurance Have? The Professional Society for Health Economics and Outcomes Research, Value in Health. doi: 10.1016/j.jval.2020.06.017
studied the impact of the public health insurance New Cooperative Medical Scheme (NCMS) on childhood nutrition in poor rural households in China (2004, 2006, 2009, and 2011), aiming to identify the mechanisms through which health insurance coverage affects nutritional intake. The study showed that NCMS was associated with a decline in calories, fat, and protein intake and an increase in carbohydrates. Increments in out-of-pocket medical expenses were identified as the primary channel through which the NCMS affected children’s nutritional intake, as NCMS coverage tended to encourage the use of higher-level medical providers.

Chen et al. (2022)[8] Chen, Q., Pei, C., Huang, J., & Tian, G. (2022). Public health insurance and enrollees’ diet structure in rural China. Heliyon. doi: 10.1016/j.heliyon.2022.e09382
studied the impact of enrollment in the NCMS program on the insured’s diet diversity and balance. Their results revealed benefits in diet diversity, overall diet balance, and nutritional intake. For those enrolled, they found evidence of under-consumption of animal products and fruits, and of over-consumption of grains, pointing out what they refer to as a potential health risk on the insured.

The work of Costa-Font et al. (2020)[18] Costa-Font, J., Gyori, M., & Saenz de Miera, B. (2020). Do health insurance extensions affect nutritional choices and outcomes? – A case study of Mexico’s Seguro Popular. Department of Health Policy of the London School of Economics and Political Science. PhD thesis, London School of Economics and Political Science.
is, to our knowledge, the only study that investigates the effects of Seguro Popular on health and nutritional choices. They analyzed the effect of the program on individuals who are overweight and obese, and food consumption using three waves of the Mexican Family Life Survey (MxFLS): one pre-treatment (2002) and two covering the expansion of the program (2005 and 2009). The study primarily focuses on the nutritional choices and outcomes of households benefiting from the program. Their findings indicate that Seguro Popular had no discernible impact, as their coefficients on all outcomes are remarkably close to zero and not statistically significant. Their choice of methods, surveys and geographic focus differ from ours, which may explain the different results obtained.

3. Estimation Strategy

The econometric analysis begins with the estimation of systems of Seemingly Unrelated Equations (SUR) introduced by Zellner (1962)[90] Zellner, A. (1962). An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias. Taylor & Francis, Ltd. doi: 10.2307/2281644
. The explained variables here are the expenditures in each of the food groups. In the SUR models, the equations are linked, as their disturbances are allowed to be correlated, feeding the system with additional information that would be missed if the expenditure equations were considered separately. The correlation in disturbances among the equations that explain household expenditure could come from the same sources, such as income, price levels, or household characteristics, gaining efficiency in the estimation by combining the information on different equations.

There are nine regression equations each for the nine discrete categories of food. Although the demand for each category is represented in individual equations, any income shock will likely affect the demand for all categories. A SUR system is then appropriate to capture this relationship among the equations through the error term. Consumption of the food f of household h is expressed in equation (1) as follows:

for $f=1,\dots ,F$ and $h=1,\dots ,H$ . Where $Y_{fh}$ is the real per capita expenditure of household h on the food category f, X’ represents a set of explanatory variables including income, demographic structure of the household (total male and female, minors, and senior adults) and characteristics of the head of the household (age, sex, educational levels, and work formality, for example), SP is a dummy variable taking a value of one if the household was insured, so ${\beta }^{+}$ will capture the short-run effects (in 2004) of Seguro Popular.

The matrix form of the regression model is:

where $X_f$ is the set of regressors for the equation of the f category of food, including $SP$ .

The disturbance vectors ${\epsilon }_1$ to ${\epsilon }_F$ are assumed to have the following variance-covariance matrix:

for

where $\sum$ is the matrix variances and covariances for the F=9 individual equations. According to Moon and Perron (2006)[60] Moon, H. R., & Perron, B. (2006). Seemingly Unrelated Regressions. University of Southern California and Université de Montréal.
, in the classical linear SUR model, there is the assumption that for each $f=1,\dots ,F$ conditional on all the regressors $X$ , the errors ${\epsilon }_t$ are i.i.d with mean zero and homoscedastic variance. Furthermore, by applying least squares or generalized least squared methods (Srivastava & Dwivedi, 1979[79] Srivastava, V., & Dwivedi, T. (1979). Estimation of seemingly unrelated regression equations: A brief survey. Journal of Econometrics. doi: 10.1016/0304-4076(79)90061-7
), the $\beta$ estimators can be obtained as:

While the SUR model will help us capture the effect of Seguro Popular once it was implemented, we are aware of possible self-selection issues. To isolate the causal effect of the program considering a temporal dimension, before and after the intervention, we implement a quasi-experimental design and estimate the effect through a difference in differences (DiD) approach. The DiD technique compares the changes in food expenditure over time between two groups, treatment (population that received the insurance) vs control (the group that did not), while controlling for other socioeconomic characteristics. This estimation method is useful when the data stem from a natural experiment (or quasi-experiment) (Wooldridge, 2013[87] Wooldridge, J. M. (2013). Introductory Econometrics: A Modern Approach (Fifth Edition). United States: South-Western, Cengage Learning.
), like when an exogenous event, such as Seguro Popular, occurs. The control and treatment groups emerge naturally due to the policy change.

Simply measuring the impact of the program as the difference in the output before and after the intervention would not be an accurate estimation either since other individual and household factors might have also changed and influenced the magnitude of the effect. Changes in the expenditures would be incorrectly attributed only to the public intervention under study. The DiD approach helps to isolate the impact of the policy but requires a reliable approach to consider the possible selection bias. To illustrate the procedure, we follow Duflo et al., (2008)[26] Duflo, E., Glennerster, R., & Kremer, M. (2006). Using Randomization in Development Economics Research: A Toolkit. National Bureau of Economic Research. doi: 10.3386/t0333
, define:

Since a household either participates or not in the program, the estimate of interest is rather the average effect in the population, $E[Foodexp$ _hT - $Foodexp$ _hC]. With access to data on both groups, the effect can be obtained by taking the difference in expected consumption between the group of households with Seguro Popular, $E[Foodexp$ _hT|T], and the group without the health insurance, $E[Foodexp$ _hC|C ], that is:

The selection bias can be theoretically illustrated by subtracting and adding $E[Foodexp$ _hC|T] to equation (5), this is the expected consumption expenditure on the food category of interest for a household in the treatment group had it not been treated, thus:

where:

With a difference-in-difference approach we use data on consumption expenditures before (period 0, year 2002) and after (period 1, year 2004) the implementation of Seguro Popular to control for pre-existing differences between the two groups, and under the assumption that differences between the groups remained constant over time (followed parallel trends), the difference-in-difference estimator is:

If the parallel trends assumption holds, equation (7) provides an unbiased estimate of the effect of Seguro Popular on the consumption expenditure of the types of food of interest. It can be written as:

which indicates that the consumption expenditure in the treatment group, without access to public health insurance, would have followed the same time trend as the control group. The DiD estimator is then obtained by estimating the following linear regression model, for each food category f:

where $Period$ is a dummy variable taking values of one for the post-implementation period, 2004, and $SP$ is a dummy for the treatment group. The difference-in-difference estimate ${\widehat{\beta }}_3$ measures the effect of Seguro Popular (different changes over time), the difference between the calculated trends for the treatments and control group.

In the estimations, the treatment and control group were created based on the percentage of coverage of Seguro Popular within the municipality where the households resided. We classified the sample in four groups, starting with municipalities where the program was not offered (0% coverage), followed by a group with municipalities with low coverage (25 - 50%), the third group represents medium coverage (50 – 75%), and the fourth group contains those municipalities with high coverage (>75%). Households located in municipalities where the program's coverage was higher than 50% constitute the treatment group, and those with no coverage form the control group. Ideally, only municipalities with high or full coverage would form the treatment group. Unfortunately, the number of observations here is extremely low (see Table 2), which motivated us to add all municipalities with medium coverage. Results then would be seen as a lower bound approximation to the true effects.

4. Data Description

Data for the empirical analysis come from The Mexican National Household Income and Expenditure Survey (ENIGH) of 2002 and 2004. It distinguishes urban from rural communities (< 2,500 inhabitants), allowing us to focus only on the latter. The sample of 2002 represents 3,305,493 rural households and 3,339,657 in the sample of 2004. The survey is rich in information; it provides detailed data on consumption, including expenses and the amounts of food consumed, income, as well as demographic and other socio-economic characteristics of both the household and each household member.

The survey labels the different expenditures by group codes. The purposes of this research require the information labeled with code “A”, which identifies expenses on “food and drinks”. This group represents more than 80% of the total spending on household intake. All products included in food and drinks are further classified into the nine different categories shown in Table 1. Namely, (1) animal protein, (2) cereals, (3) fruit and vegetables, (4) milk and derivatives, (5) processed sugars, (6) oils and fats, (7) alcoholic beverages and tobacco, (8) food consumed outside the household, and the remainder is grouped in (9) others. Consumption expenditures in each one of these nine categories are the explanatory variables that form the system represented in equation (2) for the SUR model, and that will be individually regressed to obtain the DiD estimator shown in equation (8).

Table 2 shows how rural households distributed their food expenses in 2002 (left panel) and 2004 (right panel). The columns separate the municipalities according to the proportion of households that Seguro Popular insured. As in 2002 Seguro Popular had not been implemented, this comparison helps us identify changes in expenditure behavior before and after the policy at different levels of coverage. For example, column [1] in the left panel indicates that in 2002, in households where the policy would remain absent, 27.7% of the total expense was allocated to consumption of food and vegetables, 23% was spent on cereals, followed by animal protein with 16.8%, these three categories then accounted for nearly 70% of the total. In column [4], which shows the expenditure distribution of households located in municipalities where the coverage would be high (over 75%), a similar pattern of expenditure emerges, with food and vegetables accounting for 23.7%, followed by cereals 23.3%, and animal protein with 15.5%. The right panel shows the expenditure shares once the policy was introduced. Column [4] indicates that in households that were granted access to Seguro Popular, there was a decrease in the participation of fruit and vegetables of about 8 percentage points, to 15.8%, this change appears meaningful as in households that remained excluded, column [1], the proportion only reduced by 5.7 pp, to 22%. The intake of processed sugars appears to have increased among those covered by the policy since the share of expenditure in this category more than doubled (from 4.9% to 10.5%). Cereals do not show notable changes, while there was a small increase in the share of expenditure on Animal Protein (about 3pp).

Table 3 compares the average expenditure in food for the treated, before and after the policy. The t-test results support a statistically significant reduction in Animal Protein and Cereals, of about $400 in both cases. We cannot know from the survey what specific types of food households consume away from home, but the expenditure in this category, labeled Outside, increased and the change is statistically significant. The category of Others also shows a statistically significant increment.

Comparing the expenditure distribution in food before Seguro Popular between the treatment group and the control group (as shown in Table 4), allows us to observe a notably different spending behavior between the groups. On average, the treated group had significantly higher expenditures in Animal Protein, Cereals, and Processed sugars, but lower in Fruit and vegetables and Milk and its derivatives.

Various explanatory variables ⁶ Table 13 in the appendix shows the list of all variables. will be employed in the estimations; these are used to control for socio-economic characteristics at the head of the household, household, municipality, and state levels that could have influenced the spending choices. Mean values for the two groups, before and after the intervention, and for all the variables considered are presented in Table 5. The values portray relatively similar groups, in both, the average number of male household members is around 2, equal to the average of 2 female members. The average number of older adults (> 65 years) is less than the unity, and of minors (<18 years) is 1. In 2002, the average monetary income for beneficiaries’ households was $13,154.65, and for the non-beneficiaries was $12,957.37 (a $197.29 difference), this gap widened in 2004, as beneficiaries’ households had a quarterly income increment of $56.40, while for those in the control group it grew by $2,671.89. The number of employed members per household remained between 1 and 2, and around 70% of households received a social transfer. The household head level of education with the highest proportion is basic education, with about 60% of households having an average head age of 48.

Municipal variables are chosen to reflect households' socio-economic and infrastructural aspects that may influence household food consumption, shaping dietary consumption patterns. The average of accredited years of schooling is 5.80. The percentage of the population aged 15 years and over without any school year completed is 16%, 80% of households are male-headed, the percentage of households without piped water, drainage, and electricity is around 4%, 70% of private dwellings households inhabited a floor made of a material other than dirt, and 10% of the population aged 5 and over speaks an indigenous language.

The parallel trends assumption in the DiD procedure means that with the absence of Seguro Popular, the food spending behavior of the two groups would have followed the same trend over time. In satisfying the assumption it is useful to examine how similar the groups were before the program. We resort to weighted t-tests for means to this end, the p-values (Table 11) indicate that the groups were statistically different in a handful of features; in particular, the proportion of older adults (>65) and male children (7-15) are larger in the control group. The variables for education suggest individuals in the treatment group completed more schooling years. The percentage of male-headed households and the proportion of households where the floor is not made of dirt are also higher for the treated.

Thus, despite the groups being similar in several features including total income, we cannot presume the two groups to be perfectly identical before the implementation of Seguro Popular. Recognizing these pre-existing disparities is essential, as they could introduce bias and confound the estimated treatment effect. Pre-existing differences between the groups may have potential implications for subsequent expenditure outcomes and decision-making processes. By identifying and accounting for these disparities, we can better comprehend the potential effects of these differences on the outcomes of interest, which allows us to mitigate the risk of drawing erroneous conclusions in our estimation of the Seguro Popular program’s impact.

In this analytical context, we further examine changes in the observed characteristics of the treated group from 2002 to 2004. The outcomes of the weighted t-tests, as presented in Table 7, reveal significant differences across demographic, educational, and municipal characteristics, observed in 16 out of the 23 variables. It is important to note that income levels remained statistically unchanged during this period, and therefore, if any changes in food consumption are found, they should not be attributed to an increase in income.

Considering the results of descriptive statistics, it is imperative to control the behavior of the treatment group over time in the estimations. Tracking the same treatment group over time controls individual disparities and heterogeneities within our data. Characteristics and circumstances of individuals in the treatment group may undergo temporal variations, which can bias the estimates of the impact of Seguro Popular. These individual disparities will be accounted for through the selected estimation methods, allowing the focus to remain on net changes resulting from the treatment.

In addition to controlling for individual-level variation, we also assessed broader contextual factors that could influence consumption. To explore whether changes in consumption patterns could be driven by fluctuations in food prices, we conducted t-tests comparing average food prices by food groups reported for 2002 and 2004 ⁷ Table 14 in the appendix shows detailed results. (for treatment and control). Significant increases were found in several categories, including animal protein, cereals, fruits and vegetables, and food consumed outside the household. Although price data are not directly included in the estimation models, these tests suggest that any observed substitution toward less healthy food cannot be fully attributed to price inflation. Additionally, as mentioned before, our models include municipal-level fixed characteristics that proxy for local economic and infrastructure conditions, which may partially absorb the effects of regional price dynamics. It is also important to note that food price data in the available sources contain important gaps and inconsistencies, which limit their inclusion as reliable covariates in the main models.

5. Results

Following the outline of the two methods presented in the estimation strategy, here we show the estimates of health insurance effect on food consumption, first with SUR models and then with the DiD approach.

Conclusions

Free or subsidized insurance programs aimed at promoting access to medical care for the vulnerable poor are ubiquitous around the world. The interconnectedness of financial insurance-savings-consumption decisions imply that these programs may also impact the choices of food. We investigated the final effect of access to medical care on dietary choices taking the Mexican program Seguro popular in rural regions as a case of study. A priori the effects of these programs were unknown as promoting a healthier diet or encouraging unhealthy habits are both possible.

The findings from our analysis, utilizing both Seemingly Unrelated Regressions (SUR) and Difference-in-Differences (DiD) models, highlighted significant shifts in food consumption patterns following the implementation of Seguro Popular. Households in municipalities with high coverage of the program exhibited a significant increase in spending on foods categorized as processed sugars, and oils and fats. This indicates that the provision of health insurance appears to inadvertently encourage poorer dietary choices. The robustness of these findings across different functional forms and income levels underscores the need for policymakers to consider the broader implications of health insurance programs on dietary habits.

Like many other similar programs across the globe, Seguro Popular was established with a clear and honorable objective. However, given the shifts in consumption patterns and nutritional preferences it causes among the recipients in the rural regions of Mexico, the risks of nutritional deterioration are tangible. These could fundamentally undermine the core rationale behind its creation. The results help to add valuable information on public health insurance programs about Mexican rural households’ consumption and spending structures. More generally, these findings are helpful in enriching the political debates on the possible unintended consequences of insurance programs in vulnerable communities.

Despite the strengths of our study, it is crucial to acknowledge its limitations. The classification of households based on municipal-level coverage percentages may have introduced some misclassification in the treatment group, potentially biasing our estimates. Additionally, the short-term nature of our post-implementation data limits the generalizability of our findings to longer-term outcomes. Future research should aim to incorporate longer follow-up periods and more precise measures of program coverage to discern whether these changes are transitory or indicative of enduring transformations. Nonetheless, our study provides valuable insights into the unintended dietary consequences of health insurance programs.

References