ELSEVIER
Health Policy 40 (1997) 237-255
Household income and health care expenditures in Mexico Susan W. Parker ‘,*, Rebeca Wong b a El Colegio de Mexico, CEDDU, Camino Al Ajusco No. 20, 01000 Mexico D.F., Mexico b Department of Population Dynamics, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, USA
Received 27 January 1997; accepted 27 February 1997
Abstract The purposeof this paper is to examinethe determinantsof householdhealth expenditures in Mexico. Our analysisinvolves the estimationof householdmonetary health care expenditures, using the economicand demographiccharacteristicsof the householdas covariates. We pay particular attention to the impact of householdincome on health expenditures, estimating the elasticity of health care expenditureswith respectto income for different income groups and according to health insurancestatus. For the empirical analysis,we use the Mexican National Survey of Income and Expendituresof 1989.Our principal findings show that monetary health expendituresby Mexican householdsare sensitiveto changesin householdincomelevelsand that the group which is most responsiveto changesin income levels is the lower-incomeuninsuredgroup. This suggeststhat in times of economiccrisis, these householdsreduce cash expenditureson health care by proportionately more than higher-incomeand insured households.0 1997 Elsevier Science Ireland Ltd. Keywords:
Health expenditures;Health insurance;Mexico; Social security
* Corresponding author. Tel.:
[email protected]
+ 52 5 6455955, ext. 4065; fax:
+ 52 5 6450464; e-mail:
0168-8510/97/$17.00 0 1997 Elsevier Science Ireland Ltd. All rights reserved PIISO168-8510(97)00011-O
238
S. W. Parker,
R. Wong / Heulth
Policy
40 (1997)
237-2.55
1. Introduction Mexico experienced a severe economic crisis during the decade of the 1980s during which real wages fell by 50%. As of December, 1994, Mexico re-entered economic crisis with real wages falling by approximately 35% during 199.5. Economic shocks of this nature have consequences on all components of three overall welfare of the population-health being one of them. Further more, because of these economic shocks, various strata of the population can experience changes in income with varying consequences. In this context, it is particularly important to understand how changes in income levels affect health care expenditures by the population, and how the health care responses may vary across income groups. This is relevant because health care expenditures affect the level of health of the population in the long run, and in turn, the aggregate health status imposes a potential for economic development through the ability of the population to adapt new technologies and take advantage of other investments in human capital such as education and on-the-job training. The purpose of this paper is to examine the determinants of household expenditures on health care in Mexico. Our analysis involves the estimation of household monetary health care expenditures, using the economic and demographic characteristics of the household as covariates. We pay particular attention to the impact of household income on health care expenditures, and we estimate the income elasticity of health care expenditures separately for lower and upper income households. The paper is organized as follows: First, we describe the health status of the population in Mexico and we provide background on the Mexican health care system. Second, we use data from the National Income and Expenditures Survey (ENIGH)’ from 1989 to describe the patterns of expenditures on health care across households by socioeconomic characteristics of their members. Next we use multivariate regression techniques to examine the determinants of health expenditures among households to assess the effect of income; we also compare how these determinants vary across income groups. We end with a discussion on the results and considerations for further research in this area.
2. Background The disease profile of the Mexican population includes causes of death experienced in poor countries and those common in developed nations. At the end of the 1980s the main causes of death were cardiovascular diseases, accidents, malignant tumors, and infectious intestinal diseases. The prevalence of acute diseases is highest among the poorest sub-groups of the population [l], which tend to be those living in rural areas, those living in dwellings ’ Encuesta National de Ingresos y Gastos de 10s Hogares
S. W. Parker,
R. Wang/Health
Policy
40 (1997)
239
237-255
in poor conditions, those with large numbers living in the same dwelling, and those with heads of households with low educational status. The main causes of death among the rural poor are infections and malnutrition, while chronic and degenerative diseases and injuries are the most common causes of death in the urban population [2]. Thus Mexico has been experiencing an epidemiological transition of the ‘prolonged and polarized model.’ It is called prolonged because comunicable diseases are still the predominant cause of death. And it is called polarized because the health status differs greatly across population sub-groups due to existing social inequality. We are interested in one of the possible explanatory factors of such disparities in disease prevalence and causes of death-the use of health care services-and more specifically the resources allocated by households towards health care. We argue that the level of health care expenditures (in-kind and in-cash) represents an important policy variable that merits consideration in health research, particularly among those interested in assessing how vulnerable certain sub-groups of the population are when changes in income occur. The deterioration of macroeconomic conditions in Mexico during the 1980s was historically important. The Mexican economy had grown steadily between 1940 and 1980, with the gross national product more than tripling in these four decades. The oil boom of the five years prior to 1982 produced growth in the gross national product of an average 8.4% per year, contributing to a relative economic improvement for the population, particularly those who live in urban areas. During this period, health continued to improve; for example, Table 1 shows drastic increases in life expectancy and declines in infant mortality which have occurred since 1950. Table I Historical measures of health and well-being in Mexico Concept
1950
1960
1970
1980
1990
Infant mortality rate per 1000 registered births
96.2
74.2
68.5
38.9
23.9
49.57 48.09 51.04
58.91 57.61 60.32
62.00 60.05 63.95
66.22 63.20 69.40
69.69 66.43 73.05
56.8 60.4 53.4
66.5 70.5 62.7
76.3 79.5 73.1
83.0 86.2 79.9
87.6 90.3 85.0
33 172
61.0 41.5 58.9 45 744
70.7 51.0 74.8 64 173
79.4 63.6 87.5 64 880
Life expectancy Men Women Literacy rate” Men Women Percentage of Percentage of Percentage of Gross national
households with water households with sewage system households with electricity product per-capita in 1980 pesos
Sources: Compendio historic0 de estadisticas vitales, 1893-1993, secretaria de salud. La economia Mexicana en cifras 1990, national financiera estadisticas historicas 1993, I.N.E.G.I. a 1950-1970 figures, include individuals 10 years and older; 1980-1990 figures, include individuals 15 years and older.
240
S. W. Parker,
R. Worzg ; Health
Policy
40 (1997)
237- 255
In 1982, marking the beginning of severe economic crisis, the gross national product fell for the first time in four decades, and unemployment doubled with respect to the previous year [3]. The most severe years of the crisis were between 1982 and 1986 with real wage income falling by about 50% [3]. Sluggish growth characterized the period between 1986 and 1992, although real wages remained constant [4]. Mexico once again entered a period of economic crisis in December 1994. The inflation rate in 1995 reached about 52%, while real Gross Domestic Product fell by 6.9% (GEA, 1996). The minimum wage fell further, losing 13.3% of value in real terms during 1995, while the real manufacturing wage fell by 27% [5]. One of the major concerns for policy makers is that the fall in purchasing power of the population will hinder their utilization of health care to the extent of producing a significant loss in their health status. There is little evidence on how economic crises affect health2. Our research question deals with how households allocate resources towards health across income groups, in an effort to understand the likely effect of macro-economic shocks on the overall household allocation of limited resources. These relationships inform the likely impact of economic recessions on the welfare of the population. In particular, and given the inequalities in disease patterns mentioned above, this paper focuses on the sensitivity of households health care expenditures to income levels and how this sensitivity varies across sub-groups of the population.
3. The health care system in Mexico The health care system in Mexico has a public-sector orientation, with the underlying philosophy that individuals and households should be protected by the public sector. The health system does however, have both a public and a private component. The public sector includes institutions that provide health care for the population working outside of the formal sectors of employment or uninsured. These are the Ministry of Health (SSA), the National Institutes of Health, the Social Security System (IMSS) Solidarity Program, the National System for Integral Family Development (DIF) and the Health Services of the Federal District Department (DDF). The public sector also covers population with health insurance: those belonging to the Mexican Institute of Social Security (IMSS), the Institute of Social Security and Services for State Workers (ISSSTE), the Armed Forces Social Security, the Mexican Oil Workers social security (PEMEX), and ’ The reaction of various groups of the population to large changes in income due to economic crises has received some research attention. A longitudinal study of families in Guadalajara, Mexico (one of the three major Metropolitan areas in Mexico) between 1982 and 1987, [6] reports that the proportion of household expenditures allocated to food and housing increased during the years of crisis (1982 and 1985) whereas the share of expenditures allocated towards health and education declined. There is also evidence that families coped with income shifts by increasing the labor force participation in part-time and non-paid activities, by increasing the participation of women and children and by increasing participation of the unskilled labor force during the 1980s [7,8]
S. W. Parker,
R. Wang/Health
Policy
40 (1997)
237-255
241
other health services for state and federal government employees. On the other hand, the private sector includes a variety of individuals and institutions working in a range of traditional and alternative medicine, mobile units, hospitals and clinics, private practices and private medical insurance. In 1995, almost half of the Mexican population was covered by a social security institution, 40% was covered by institutions for the non-insured, 5% used private services, and 11% had no access to the health system’s facilities [9]. Private pharmacies comprise a major source of informal care; pharmacists generally prescribe medicines and treatments in Mexico. In summary, the provision of health services is largely the responsibility of the public sector. Using the World Bank’s classification of countries [lo], Mexico can be regarded as a middle-income country, with the government using general revenues to pay for health care for middle- and low-income groups of the population, while upper-income households (a minority) use private insurance or resources to pay for private health care. Coverage by the social security system is determined largely by participation in the formal labor market; hence this system tends to discriminate against agricultural workers (and hence those living in rural areas), domestic workers and small-commerce industry workers. These are also activities in which the poorest sector of the population tends to be over-represented. It is unclear, however, whether changes in family income due to macroeconomic adjustments have a real impact on health care utilization in countries such as Mexico, where the provision of health care is largely the responsibility of the public sector. The brief summary provided above suggests that in Mexico, the population has access to health services based on two main factors: the level of urbanization of the community where they live and whether there are services available in it, and whether or not they form part of the social security system. Given this system of provision of services, we would expect that health care utilization would exhibit a large variation across the population, and this is supported by the literature [I], which indicates that the utilization of health services differs vastly by income group.
4. Data and descriptive statistics
We use the 1989 Mexican National Household Survey of Income and Expenditures, a nationally representative data set including detailed information on all types of income and expenditures for approximately 11 500 households, as well as socioeconomic and demographic characteristics of the individual members. Each household reports expenditures, auto-consumption3 and in-kind medical services received for the previous 3-month period. The survey includes 38 types of health expenditures, and these can be grouped into seven different categories which include primary care, hospital care, pregnancy, child birth, non-prescription medi‘The category of auto-consumption refers to self-treatment, for instance, herbal medicinal plants grown and used as treatment. Given the low incidence of this type of health treatment, we exclude it from discussion in the paper, but is included in the tables for completeness.
242
S. W. Parker,
R. Wmg/Health
Policy
40 (1997)
237-255
Table 2 Average level of household health expenditures by type of medical care” Concept
Monetary expenditure
Autoconsumption
Primary Care % With expenditure greater than zero
115.40 (260.90) 40.31
37.90 (68.50) 0.49
136.60 (302.00) 16.26
Hospital Care Excluding Pregnancy % With expenditure greater than zero
801.60 (1846.00)
0.00 (0.00)
1747.30 (4628.00)
Medical Care During Pregnancy % With expenditure greater than zero Medical Care During Child birth % With expenditure greater than zero Medicine Without Prescription % With expenditure greater than zero
2.05
0.00
In-kind medical services received
1.52
83.40 (120.00) 2.65
0.90 (0.70) 0.03
139.00 (214.00) 1.57
402.60 (690.10)
0.00(0.00)
693.00 (751.00)
1.36
0.00
0.75
9.50 (19.20) 41.45
8.05 (12.00) 0.22
19.00 (38.70) 1.08
Orthopedic Devices % With expenditure greater than zero
197.70 (261.60) 3.14
0.00 (0.00) 0.00
229.70 (441.90) 0.39
Health Insurance % With expenditure greater than zero
115.00 (229.80) 1.87
0.00 (0.00) 0.00
86.00 (114.00) 0.05
Total expenditures % With expenditure greater than zero
130.60 (461 .lO) 63.4
27.40 (57.70) 0.07
300.80 (146.30) 19.0
Standard deviations in parentheses. Total sample size: 11 531. Source: Mexican National Survey of Income and Expenditures, 1989, authors’ calculations. a Includes only households reporting a positive expenditure or service received.
cation, orthopedic devices and health insurance. For the households reporting that medical services were received free of charge or in-kind, the survey includes an estimate of the value of the services received, based on regional and institutional costs4. Households that report receiving in-kind services include those visiting a health center in which they can receive services as part of health benefits through formal employment, as well as households without benefits visiting a health clinic at which free services are provided. 4 The individual responding to the questionnaire is asked to provide an estimate of the value of the service, medicine, and supplies that were received free of charge. If the individual is unable to provide this information, an estimate is obtained from medical units in the area of residence of the household.
S. W. Parker,
R. Wong / Health
Policy
40 (1997)
237-255
243
Table 2 shows the average level of household monetary health care expenditures and the value of in-kind health services received by category of expenditures during the 3-month reference period. About two-thirds of households reported making a monetary expenditure on health care during the reference period, while about one-fifth of all households reported receiving in-kind medical services. For those reporting an expenditure greater than zero, the average amount of in-kind services received, however, was much greater than the average value of monetary expenditures, 300 new pesos compared to 130 new pesos’ Turning to the different categories of health expenditures, the most common expenditures were for primary care and medicines, while in-kind services were most likely for primary care. The amount of expenditures and in-kind services, however, was much larger for hospital care and child birth than for primary care and medicines. Less than 2% of the sample reported making a payment for health insurance. Table 3 shows the distribution of household health expenditures and in-kind services for those reporting an expense or service received during the reference period. Regarding monetary expenditures, 54% of the total were on primary care and 34% on medicines. For in-kind health care services, on average, 80% of services were for primary care. Table 4 provides characteristics of households reporting cash medical expenditures versus those receiving in-kind services. Rural households are less likely to Table 3 Percentage distribution Concept Primary Care
of household health expenditure by type of medical care” Monetary expenditure
Autoconsumption
In-kind medical services received
54.27% (45.44)
65.85% (47.71)
79.63% (38.31)
Hospital Care Excluding Pregnancy
2.26% (13.62)
0.00% (0.00)
6.37% (22.80)
Medical Care During Pregnancy
2.51% (14.17)
3.66% (18.89)
5.64% (21.74)
Medical Care During Childbirth
1.61% (11.60)
0.00% (0.00)
3.13% (16.23)
34.31% (43.83)
30.49% (46.32)
3.54% (17.73)
Medicine Without Prescription Orthopedic Devices
3.17% (15.42)
0.00% (0.00)
1.48% (11.31)
Health Insurance
1.87% (12.40)
0.00% (0.00)
0.21% (4.35)
Total
100%
100%
100%
Source: Mexican National Survey of Income and Expenditures, 1989, authors’ calculations. a Includes only households reporting a positive expenditure or service received. Standard deviations in parentheses. 5 In 1989, one US dollar was approximately equivalent to three Mexican pesos.
244
S. W. Parker,
R. Wong /Health
Policy
40 (1997)
237.-255
Table 4 Percentage of households with monetary expenditures on health care and percentage of households receiving in-kind medical care by socio-economic characteristics Characteristic
Monetary expenditures (“/;I)
In-kind care received (%)
Total
64.2
19.6
Residence Area Urban Rural Head with health benefits Head with no health benefits
64.4 63.9 62.4 66.1
21.2 16.7 30.0 9.2
Household size l--2 members 3-- 5 members 5 or more
56.1 64.7 66.9
18.9 21.6 16.9
Education of head One Some primary Completed primary Primary and more
60.9 63.9 62.0 68.0
13.6 18.0 20.0 24.6
Income Bottom 20% Middle 60% Upper 20%
52.0 65.6 72.2
14.2 19.8 24.5
Household has water electricity and sewage Yes 65.4 No 63.0
23.3 15.9
Total sample Size: 11 53 1. Sample size with monetary expenditures: 7319 Sample size with in-kind services received: 2194 Source: Mexican National Survey of Income and Expenditures, 1989. authors’ calculations,
receive in-kind services, but they are about equally as likely as urban households to make a cash expenditure on health care. Those in which the household head has health benefits are less likely to make a health expenditure and, not surprisingly, households in which the household head has health benefits are much more likely to receive in-kind care than those in which the household head lacks benefit@. Turning to income levels, lower-income households are less likely to make a monetary expenditure on health care and less likely to receive in-kind services than upper-income households. Similarly, higher education levels are positively related to making a health care expenditure and to receiving in-kind/free services. 6 We also constructed variables measuring the percentage of adults in the household reporting health benefits. However, health benefit status of the household head is highly correlated with the health benefit status of other members of the household. Thus, for simplicity, we use the health benefit statusof the household head to proxy coverage of benefits for the household.
S. W. Parker.
R. Wang/
Health
Policy
40 (1997)
237-255
245
In summary, the results of Table 4 suggest a close relationship between having formal sector jobs, high income levels and the use of health care services. Individuals who live in urban areas are more likely to have formal sector jobs than those in rural areas and thus are more likely to have health benefits, which provide in-kind services. Rural households, the majority of which are poor, are more likely to be employed in the agricultural sector or in informal sector jobs which lack health benefits. Urban-rural residence also measures availability of health care, as medical services such as health clinics and pharmacies are less available in rural areas. Finally, we turn to the percentage of household expenditures which are devoted to health. Table 5 provides three different ratios measuring the extent to which households allocate expenditures to health care. The first ratio shows total health expenditures (including the value of in-kind services received) out of total expenditures, while the second includes only monetary health expenditures. The third ratio measures health care expenditures as a percentage of total expenditures minus expenditures on food. All the ratios included in the table refer only to households having positive expenditures on health care. Overall, about 5% of total expenditures during the reference period are on health care. As a share of non-food expenditures, households spend about 10% on health care. Rural households spend a higher percentage on health care than urban households while lower income households spend proportionally more on health care than upper income households. For instance, households in the lowest 20-percentile income group spend over 13% of non-food expenditures on health care, compared with 7% for the highest 20-percentile income group. Finally, the results also indicate that smaller-sized households allocate a higher share of expenditures to health care than larger households.
5. Household
expenditures
on health: theory and empirical
specification
The new household economics assumes that households derive utility or satisfaction from consuming commodities and services, and household members must produce many of the commodities that are desired for consumption. Households face constraints of monetary and time resources as well as psychic costs. For our case, we assume that households derive utility from their household members’ health and other consumption. In order to consume health, households produce it by combining inputs of production such as health care services and the household members’ time. In this sense, the utilization of health services is assumed to be a demand derived from the demand for health. We formulate a simplified utility function that represents all members in the unit, and assume that households derive utility from overall consumption and health, conditional on the composition of the household (demographic factors) and constrained by the social and economic context (socioeconomic factors). The derived demand for health care is a function of the level of income of the household, prices, and preferences for health relative to other consumption. Similar to previous research, we incorporate the household demographic variables in our model [l 1,121.
246
S. W. Parker.
R. Wong / Health
Policy
40 (1997)
237-255
We specify an empirical model in which market prices are not-observable. Several authors have dealt with the issue of estimating demand functions or expenditure functions when prices are unobserved [l 1,131. One convention is to use budget share equations as an alternative approach instead of estimating demand equations for health care [I 1,141. Another approach uses expenditures and quantities to derive estimates of prices. In our case, information on quantities of health care is not reported, and the prices of health care services are mostly unobserved because most Table 5 Percent share of health expenditures in total expenditures:various characteristicsa
share measures by household
Characteristics
Total health/ (Total expend)b
Total mon. health/ (Total mon. expend)
Total health/ (Total food)d
Total
5.32%
(8.9)
4.27%
9.71%
Residence Urban Rural
4.98% 5.95%
(8.4)
(10.0)
3.72% (10.0) 5.7% (9.1)
11.26% (1.53)
Head with health benefits Head with no health benefits
5.87% 4.73%
(9.7) (8.1)
3.68% 4.83%
(6.8) (8.3)
10.16% 9.23%
Household size l-2 members 3 - 5 members 5 or more
5.33% 4.63%
(8.9) (7.9)
5.61% 4.09% 4.05%
(9.6) (7.1) (7.5)
13.86% (1.84) 9.31% (1.35) 8.74%
(1.23)
Education of Head None Some primary Completed primary Primary and more
5.31% 5.91% 5.11% 4.95%
(8.9) (9.6)
5.03% 4.82% 4.00% 3.52%
(9.2) (7.9) (7.6) (6.0)
10.74%
(1.48)
7.95%
(1.18)
Income Bottom 20% Meddle 60% Upper 20%
6.04% 5.59% 4.85%
(1.3) (8.9) (8.4)
5.54% 4.13% 3.75%
(9.9) (7.0) (7.2)
13.48% 9.68% 6.96%
(1.73) (1.13) (1.13)
Water, sewage and electricity in household Yes No
5.33% 5.30%
(9.0) (8.9)
3.86% (7.0) 4.669% (8.2)
9.04% 10.42%
(1.44)
7.10% (1.14)
(8.9) (8.3)
(7.6)
8.89%
(14.2) (6.6) (1.45)
(1.33)
11.54% (1.57) 9.11% (1.34)
(1.34)
Total Sample size: 11 531. Source: Mexican National Survey of Income and Expenditures, 1989, authors’ calculations. a Includes only households reporting a positive expenditure or service received. Standard deviations in parentheses. b Total health/Total Expend is defined as total health expenditures (including non-monetary expenditures) divided by total expenditures (including non-monetary). ’ Total Mon. Health/Total Mon. Expend. is defined as total monetary (cash) health expenditures divided by total monetary expenditures. d Total Health/Total-food is defined as total health expenditures (including non-monetary) divided by total expenditures (including non-monetary) minus expenditures on food.
S. W. Parker, R. Wong/HealtA
Policy 40 (1997) 237-255
241
of the services are public and hence offered free-of-charge, or are inaccurately measured. When the data available presents severe limitations because of the lack of price data and quantities purchased, it is possible to approximate the variation in prices due to regional location in the country and due to health health insurance coverage of household members. For example, in Mexico, it is reasonable to assume that households that lack coverage by the social security system will face higher prices for health care than those that have coverage. Households will also face different prices if they are covered by private insurance or have access to a publicly subsidized service than if they do not. In our empirical model, we include coverage of health insurance by household members, and we use region of residence7. As an indicator of regional variation in prices, and although more-accurate estimates can be obtained with full price data, this procedure affords us estimates of socioeconomic and demographic determinants of health care expenditures that can provide useful estimates for our purposes. We formulate an equation for household expenditures as follows: Mh-exPh*=~+P,ln
[‘:I [:I, .:, T
+p2 In n-
+ 1 pknik+ 6Z,+e,
where h-exp? are the expenditures on health care by household-i, Yi is the household income, n, is the number of household members, nik is the number of household members in age-group k, zi represents a vector of other household socioeconomic and demographic characteristics, and e, is a random error term. We adopt the convention of using logarithmic transformations of expenditures and income in order to capture possible non-linearity in the relations. We assume that h-exp* is the unobserved desired level of expenditures by the household. We only observe h-exp when it is positive and beyond a certain threshold; otherwise we observe zero. As is common in these types of specifications, we estimate the expenditure equation including only those households with positive expenditures, using a Heckman correction for the selectivity bias introduced by this convention [15]. We use the state-level expenditures on health and education as our identifying variable. This variable is assumed to affect the likelihood that a household reports positive expenditures on health care, but not the amount of expenditures*. We estimate the regression equations for four different groups. First, we divide our sample into households where the household head has health benefits (insured) and households where the household head does not (uninsured). Second, for each of these two groups, we construct two income groups (upper 50th percentile and lower 50th percentile), as other authors have found that the demand for health care is less elastic among higher-income households [16].
’ In Mexico, the minimum wage varies according to region, such that areas with higher costs of living have higher minimum wage levels. We use these government definitions of region to control for differences in prices of health care. s We assume that the level of development of the community affects the type and density of health services available, but not the relative price of health care.
248
S. W. Parker.
R. Wang , Health
Policy
40 (1997)
237 -25.5
5.1. Dependent variable Our dependent variable represents the total monetary household expenditures devoted to health during the previous 3 months, excluding health insurance’. 5.2. Explanatory
variables
We include three groups of explanatory variables: household income; household composition; and socioeconomic variables. We adopt the relatively-common convention of using total household expenditures as a proxy for total household income. This convention is justified because total expenditures give a more-accurate idea of permanent income, and because in household Surveys from developing countries, expenditures are reported more-accurately than income [17]. We use this variable as an explanatory variable, and also to form the low and high ‘income’ groups of the population according to percentiles. We include total monetary and in-kind expenditures separately in our regressions, as we assume that the impacts of monetary versus in-kind income on a household’s health expenditures may be different. Among our household composition variables, we include total household size, a squared-term to capture possible non-linear effects of household size, the percentage of women of fertile ages (15-49 years), and the percentage of children with respect to the total number of household members. As other socioeconomic explanatory variables, we use whether the household resides in an urban or rural areato, whether the household head has health insurance, whether the dwelling has electricity, water, and sewage services or not, and education of the household head”. In addition, all regressions include region dummies, to proxy for the difference in cost of living across regions. Table 6 shows the means and standard deviations of the variables for the four groups of analysis. Both in the lower- and upper-income groups, households in which the head lacks health benefits are: (1) less likely to have electricity, water and sewage services in their place of residence; (2) more likely to live in rural areas; and (3) more likely to have lower levels of education, as compared with the groups in which the household head has health benefits. These findings confirm that households without health benefits have worse socioeconomic conditions than households with health benefits.
y We exclude health insurance from the dependent variable measure because the wording of the expenditures question makes it difficult to distinguish between cash payments towards private insurance plans and mandatory payments for social security. In addition, the proportion of houshelds reporting expenses for health insurance is negligible (less than 2%). “’ An urban area is defined a a community with 100 000 people or more ” Because education is reported in discrete categories in the survey, we measure the education of the household head as a series of dummy variables: no formal education, some primary education completed, primary education completed, and education above the primary level.
S. W. Parker, R. Wang / Health Policy 40 (1997) 237-255
249
Table 6 Means and standard deviations of variables used in the analysis Insured Lower 50% 12.4 (1.02) 3.05 (9.91) 0.353 (0.478) 0.238 (0.426) 0.222 (0.415) 5.36 (2.52) Electricity, water. drainage in 0.415 (0.493) household (0,l) 0.243 Proportion of fertile women in household (0.160) 0.112 Proportion of children in household (0.147) Urban residence (0,l) 0.589 (0.492) N 2887 Logarithm of per-capita monet. HH expenditures Logarithm of per-capita inkind HH expenditures HH head educ. level Some primary (0,l) HH head educ. level Completed primary (0. I) HH head educ. level More than primary (0.1) Total members in HH
Uninsured Upper 50%
Lower 50%
Upper 50%
13.8 (0.836) 2.10 (10.5) 0.157 (0.402) 0.203 (0.492) 0.589 (0.492) 4.33 (1.96) 0.798 (0.401) 0.294 (0.194) 0.100 (0.148) 0.878 (0.326) 2886
11.9 (1.67) 3.34 (9.83) 0.399 (0.478) 0.164 (0.370) 0.082 (0.274) 5.49 (2.69) 0.200 (0.400) 0.215 (0. I 54) 0.128 (0.157) 0.348 (0.476) 2861
13.4 (1.05) 0.944 (10.5) 0.257 (0.437) 0.221 (0.415) 0.327 (0.469) 4.50 (2.22) 0.592 (0.492) 0.270 (0.199) 0.105 (0.155) 0.748 (0.434) 2890
Source: Mexican National Survey of Income and Expenditures, 1989, author’s calculations. Standard deviations in parentheses.
6. Results Table 7 Table 8 and Table 9 show the multivariate results. We first discuss the general regression results contained in Table 7 and Table 8 and then discuss more specifically the income elasticity estimates, which are shown in Table 9. Table 7 presents the binary probit estimates of the probability of making a monetary health expenditure. We find that total monetary income is positively and significantly related to the probability of making a health expenditure for all groups except the lower income group of the insured population. Total in-kind income is positively and significantly related to the probability of making a cash health expenditure for all groups. In general, urban residence is negatively related to the probability of making a health care expenditure. For all groups, the presence of children is positively related to the probability of making a health expenditure. Table 8 presents the two stage least squares (2SLS) selection-bias corrected estimates of the log of household health expenditures12. Total monetary income has r2 In several of the regression models, the coefficient on lambda, the selection correction term, is
250 Table 7 Binomial probit populationa
S. W. Purker.
of the probability
R. Wong / Health
of making
Policy
a cash health
40 (1997)
expenditure
Insured Lower Constant Logarithm of percapita monet. HH penditures Logarithm of percapita in-kind HH penditures Square of logarithm per-capita monet. expenditures
237-255
of the insured
and uninsured
Uninsured 50%
Upper
50%
Lower
- 1.56 ( - 3.80) 0.024 (0.373)
- 8.29 (- 5.50) 1.05 (4.95)
0.0029 (2.83)
0.0026 (2.59)
0.0069 (2.52)
- 0.033 ( - 4.40)
0.00041 (0.890)
-0.00016
0.0061 (0.222) 0.024 (-2.70)
0.067 (1.45) -0.084 (.529)
50%
Upper
-6.71 (-4.86) 0.874 (3.78)
50%
-6.30 (-5.65) 0.809 (5.06)
ex0.0044 (3.86)
0.0024 (2.45)
exof HH
Square of logarithm of per-capita in-kind HH expenditures HH head educ level Some primary (0, l)b HH head educ level Completed primary KUY HH head educ level (0.1) More than prima$ Square of total members in household Total members in HH Electricity, water drainage in HH (0,l) Percentage of fertile women in household Percentage of children in household Health insurance (0,l) Urban residence (0,l) State-level social spending per eapita N , Log - likelihood
- 0.084 (-2.51) -0.0024
(-0.424)
0.014
(-2.65)
(.315) -0.0036
-0.027
(- 2.89)
-0.026
0.0009 (1.95)
-0.014 (-0.607) -0.078
-0.00003
-0.0046 (0.177) -0.0129
(-0.0257)
- 0.026
( - 2.58)
(-0.637) -0.0015
(-4.44)
(-0.074)
(-0.475)
-0.0129
(- 1.37)
(-4.33)
0.058 (4.47) -0.075 (-3.36)
0.064 (4.07) -0.045 (-1.82)
0.045 (0.739)
0.018 (0.393)
0.125 (1.90)
0.045 (1.03)
0.305 (4.43)
0.209 (3.20)
0.283 (4.23)
0.324 (5.35)
-0.147 (-0.644) -0.062 (- 2.45) - 0.00023 (-0.296)
0.014 (0.157) -0.055 (-2.34) 0.0006 (1.24)
2861 - 1756.1
2890 - 1632
0.108 (1.78) -0.052 (-2.16) 0.0009 (1.56) 2887 - 1874
0.052 (1.02) 0.00017 (0.006) 0.0007 (1.60) 2886 - 1787
Source: Mexican National Survey of Income and Expenditures, Regional controls were included in all regressions. T-statistics in parentheses, *Marginal effects evaluated at the mean are reported instead
0.057 (4.03) - 0.026 ( - 0.925)
(-0.475) -0.0049
1989, authors’
of coefficients b Omitted education category is household heads with no formal education.
0.083 (6.32) -0.011 (-0.547)
calculations
S. W. Parker,
R. Wong / Health
Table 8 Two stage least squares coefficients uninsured by income group
Policy
of log of cash
health
Insured Lower
40 (1997)
237-255
expenditures:
251
separating
insured
and
Uninsured 50%
Upper
50%
Lower
50%
Upper
50%
Constant
3.85 (0.374)
Logarithm monet. tures
of per-capita HH expendi-
Logarithm in-kind tures
of per-capita HH expendi-
0.076 (0.059)
-0.019
Square of logarithm per-capita monet. expenditures
of HH
Square of logarithm per-capita in-kind expenditures
of HH
HH head educ. level Some primary (0,l)
(-253)
0.029 (0.642)
- 0.00047 ( - 0.022)
0.085 (0.710)
-40.9
(- 1.72)
6.01 (2.06)
-0.012
(- 1.54)
-0.174
(- 1.86)
0.0022 (- 1.21)
- 26.8 ( - 1.42) 3.87 (1.62)
(-
(-1.95)
4.55 (2.61)
0.010 (0.978)
-0.095
-27.8
1.09)
0.0039 (1.32)
-0.0067
(-1.14)
-0.134
(-2.37)
0.0036 (2.12)
-0.556
(-1.91)
-0.018
(-0.161)
0.145 (1.31)
-0.765
(-3.10)
-0.396
(-1.78)
0.175 (1.48)
(-0.881)
-0.715
(-3.09)
0.037 (0.185)
0.170 (1.46)
Square of total members households
0.0056 (0.875)
-0.013
(-1.17)
Total
0.0051 (0.394)
HH head educ. level Completed primary al)
-0.0024
HH head educ. level More than primary (O,l)
-0.183
members
Electricity, drainage @,I)
in HH
water, in household
Percentage of fertile women in household Percentage of children household
in
(-0.011.)
(-0.117)
0.081 (0.336)
0.067 (0.157)
0.306 (1.41)
- 0.038 ( - 0.058)
0.821 (1.45)
1.45 (2.17)
0.565 (0.923)
0.385 (1.43)
(-1.10)
Urban
residence (0,l)
-0.523
(-3.93)
- likelihood
0.245 (0.182) - 3056.3 1595
-0.11
(-,719)
1.76 (1.10) -3613.7 1878
(-1.30)
0.266 (1.54)
-0.034
-0.033
-0.197
(-0.751)
0.114
(0,l)
Log N
0.447 (3.00)
-0.0081
(-0.011)
insurance
(0.676)
(-2.27)
-0.0021
Health
Lambda
0.349 (2.11)
-0.014
-0.159
(- 1.33)
-0.316
(- 1.74)
2.20 (1.45) -3243.3 1727
Source: Mexican National Survey of Income and Expenditures, 1989, authors’ a Omitted education category is household heads with no formal education. b Regional controls were included in all regressions. c T-statistics in parentheses.
0.068 (0.764)
0.238 (0.715) -0.347
(-2.65)
0.556 (0.419) 3889 2058 calculations.
252
S. W. Parker,
R. Wang/
Health
Policy
40 (1997)
237..255
a positive and significant effect on health expenditures for all groups except for the lower income uninsured group. Total in-kind income only has a significant effect for the insured low-income group and the impact is, unexpectedly, negative. This may be because a large portion of the in-kind income of the population with health benefits corresponds to health care, and in-kind health care is likely a substitute for monetary health expenditures. Table 9 shows estimates of the elasticities of monetary health expenditures with respect to total household income, derived from the regression results” The Table also includes estimates of the change in the probability of making a health care expenditure with respect to total household income. Overall, the elasticity estimates are plausible, ranging (for those groups where the estimates are significant) between 0.96 and 1.6. The change in the probability of making a health expenditure associated with a 1% change in household income ranges between 0.09 and 0.23. The low-income group exhibits the largest estimated impacts of income changes on both the probability of making a health care expenditure and on the amount of health expenditures. This suggests that households in the low-income group are the most vulnerable to changes in income levels, and in times of economic crisis, tend to cut back on health care expenditures to a greater degree than households in the other groups. The elasticities of health expenditures with respect to total in-kind income are much smaller than the monetary income elasticities, illustrating that the marginal impact of in-kind income is quite different than the impact of monetary income. In particular, households spend on health a smaller share of an increase in in-kind Table 9 Impacts of household monetary and in-kind income on cash health expenditures Insured
-
Uninsured
Lower 50%
Upper 50%
Lower 50%
Upper 50%
0.795 -0.021
1.20” - 0.002”
1.60” 0.036
0.959” 0.0001”
Change in probability of monetary health expenditure with respect to: Total monetary income 0.093” 0.139” 0.231” Total in-kind income 0.005” 0.0019” 0.026”
0.112” 0.0015”
Elasticity with repect to: Total monetary income Total in-kind income
Source: Tables 6, 7, and 8 and authors’ calculations a significant at the lOa% level or lower. insignificantly different than zero, indicating that the OLS estimates are not biased due to selection of the sample. Nevertheless, we prefer to report the selection-corrected estimates. I3 The expression for the elasticities is$, + 2p2 (In x)where the regression coefficients correspond to those of the log-income and the squared log-income terms in the equation, respectively. The elasticity is obtained by evaluating the expression at the mean value of In x.
S. W. Parker,
income than in monetary low-income group.
R. Wong /Health
Policy
40 (1997)
237-2.55
income. The largest elasticities
253
are for the uninsured
7. Conclusions
Our principle findings suggest that monetary health expenditures by Mexican households are sensitive to changes in household income levels, and the elasticity of monetary health expenditures with respect to income is largest for low-income uninsured groups. This suggests that in times of economic crisis and recession, these households reduce cash expenditures on health care by proportionately more than higher-income and insured households. The findings also indicate the importance of the health benefit status of the household in determining cash expenditures on health care, which is in turn determined by labor market participation in the formal or informal sector of employment. We find that monetary health expenditures also are sensitive to changes in in-kind income. Nevertheless, the size of the effects of these changes is much smaller than the estimated impacts of monetary household income. Rural households are more likely to make cash health expenditures than urban households, which may reflect the lower availability in rural areas of health clinics at which free services may be obtained, compared to urban areas. The results presented in this paper, based on a cross-section of Mexican households, indicate that with macro-economic shocks, households are likely to adjust their health-care expenditures, and the pattern of adjustment will differ by income group and by health-benefit coverage of the population. A major conclusion that we draw from these results is that with economic crisis, and despite the efforts of the public sector to make health care available to large sectors of the population, the health care expenditures of the low-income uninsured strata will adjust (decrease) by more than those in the upper-income and insured groups. One important limitation of the study is that we are unable to observe the health care needs of the household members. To the extent that the socioeconomic characteristics and dwelling conditions are highly correlated to the health care needs of household members, our estimated effects of income on health care expenditures may be partially capturing the effects of health care needs in the household. Our future research will involve the analysis of the 1984 and 1992 Mexican National Surveys of Income and Expenditures, to examine how patterns of adjustment may change over the period of economic crisis (through 1989) and (sluggish) recovery (after 1989). Additionally, given our findings on the importance of health benefits in the determination of household monetary expenditures, future research shall further examine the relationship between having health benefits and receiving in-kind services. Our results help explain the polarization that characterizes the health outcomes of the population. As decision-makers in Mexico prepare for the population to move through the epidemiologic and demographic transitions, it will be important
254
S. W. Parker,
R. Wong /Health
Policy
40 (1997)
237-255
to consider the inequalities that seem evident in access to health care, in particular during times of severe economic adjustments, and the consequences that these inequalities have on the human capital of the Mexican population.
Acknowledgements
The authors acknowledge the assistance provided by Maria Cristina Gutierrez Delgado. The research reported in this paper was partly supported by the University Development Linkages Project (USAID) and the Rockefeller Foundation, and benefitted from using the facilities of El Colegio de Mexico and the Johns Hopkins Population Center (NICHD grant 5P30HDO6268-21).
[I] Lozano, R.C. Infante, L. Schlaepfer and Frenk .I., Desigualdad, pobreza y salud en Mexico (Inequality, poverty and health in Mexico), Consejo Consult&o de1 Programa National de Solidaridad, Mexico City, 1993. [2] National Academy of Medicine, The North American Free Trade Agreement and Medical Services: Elements for Analysis and Negotiation, Mexico City, 1993. [3] Lustig, N., Mexico, the Remaking of an Economy, The Brookings Institution, Washington D.C.. 1992. [4] Parker, S.W., Male and Female Wages in Mexico: 1986-1992, Mimeo, El Colegio de Mexico, 1995. [5] Grupo de Economistas y Asociados, GEA Information: Carpeta de Information Economica Oportuna, Mexico City, 1996. [6] Gonzalez de la Rocha. M., Bienestar familiar, consumo alimentario y acceso a 10s servicios durante la crisis (Family welfare, food consumption and access to services during the crisis), In: CA. Denman, A. Escobar, C. Infante, F.J. Mercado and L. Robles (eds.), Familia, salud y sociedad (Familia, health and society), Universidad de Guadalajara: Guadalajara, 1993. [7] Wong, R., Jamil K. and Levine R., Determinants of mothers’ employment under conditions of economic deterioration: urban Mexico in the 198Os, Johns Hopkins Population Center Working Paper Series, No. 39, Baltimore, Maryland, 1993. [8] Rendon, T. and Salas C., El empleo en Mexico en 10s ochenta: tendencias y cambios, (Employment in Mexico in the eighties: trends and changes), Comercio Exterior, Mexico, August, 1993. [9] Secretaria de Salud, Vision de conjunto de la Refonna de1 Sistema National de Salud. Subsecretaria de Planeacion, Mexico City, 1995. [lo] The World Bank, Investing in Health, World Development Report, Oxford University Press, Washington D.C., 1993. [l 1] Deaton, A., Ruiz-Castillo J. and Thomas D., The influence of household composition on household expenditure patterns: theory and Spanish evidence, Journal of Political Economy, 97 (1989) 179-200. [12] Pollack, R.A. and Wales T.J., Demographic variables in demand analysis, Econometrica, 49 (1981) 1533-59. [13] Deaton, A. Estimation of own- and cross-price elasticities from household survey data, Journal of Econometrics, 36 (1987) 7-30. [14] Russo, G., Ogawa N., Wandaningsih D. and Sardjunani N., Household expenditures for health care in Indonesia: estimates from the 1987 socioeconomic survey, HOMES Indonesia Project. East-West Center, Honolulu, Hawaii, 1993
S. W. Parker,
R. Wong / Health
Policy
40 (1997)
237-255
255
[15] Maddala, G.S., Limited Dependent and Qualitative Variables in Econometrics, Cambridge: Cambridge University Press, Cambridge, 1983. [16] Gertler P., Locay L. and Sanderson W., Are user fees regressive? The welfare implications of health care financing proposals in Peru, Journal of Econometrics, 36 (1987) 67-88. [17] Russo, G. and Herrin A.N., The determinants of household health care expenditures in the Philippines, East-West Center Working Papers Population Series No. 68, East-West Center, Honolulu, 1993.