Section 1: Introduction
The impact of socio-economic circumstances on health status is one of the most important areas of discussion in public health and still a major subject of interest and extensive investigation to both economic researchers and policy makers. Some studies have strongly established that those with low incomes have lower health status than those with higher incomes. However, there is complexity in this relationship since the direction of causality between the two variables – income and health remains debatable. In addition, given the immediate indicators of or other factors that play into the link between income and health status, it is difficult to measure the impact of income on health status. This is because some factors might actually be contributing a larger proportion to health status than income. Possible factors include – environmental conditions, nutrition, genetics, level of education, and accessibility to health care services.
The economic questions of my papers as a group involve the casual effect of income (or country’s income) on health status of individuals. The policy relevance of these questions suggests that, if truly income affects health status, then the government should employ income and health policies that are designed to support persons with low income which in turn alleviates the consequences of low income; poor health status being a key consequence in this paper. The implementation of such policies will enable low income persons gain access to health care services. Possible policies include income policies such as employment insurance – that offers temporary financial assistance to those that are unemployed while they look for work, welfare assistance – that offers financial aid to those in need of basic necessities as well as health policy, precisely public health insurance such as Medicaid – for low income and needy people.
In this paper, I will be addressing two major questions. The initial question that I will address in this research is, “the effect of high income and low income on the health status of individuals respectively”. It will assist to verify whether persons with very low income are at higher risk of experiencing deteriorating health and vice versa. The next question to be addressed is the direction of causality between income and health – income to health status or health status to income level?
This paper is structured as follows: Section two reviews the article “The Casual Effect of Income on Health: Evidence from Germany Reunification” by Paul Frijters et al. The authors used a large scale natural experiment created by the “falling of the Berlin Wall” and the consequent reunification of Germany to examine the casual effect of income changes on the health satisfaction of East and West Germans. The third section looks at the findings from the article, “Estimating the Effect of Income on Health and Mortality Using Lottery Prizes as an Exogenous Source of Variation in Income” by Mikael Lindahl. This article builds on the vast literature that has established a strong positive relation between income and health status and a negative relation with mortality.
The fourth section reviews the article, “Does Money Protect Health Status? Evidence from South African Pensions” by Anne Case. The author uses pension income as an instrument to study the relationship between income and health. The fifth section reviews the article by Jerome Adda et al titled, “The Impact of Income Shocks on Health: Evidence from Cohort Data. The author studies the effect of income shocks on health for a prime-aged population. Section six examines the findings from the article, “Wealthier is Healthier” by Lant Prichett and Lawrence H. Summers. The authors estimated the effects of income using cross-country, time-series data on health and income per capital. The final section of this paper provides a summary of the conclusions from the articles mentioned above as well as unanswered questions and suggestions for future research. The major areas of discussion will come from the datum, natural experiments and econometric models that these articles use in order to answer questions about the impact of income on health.
Section 2:
This article investigates the causal effect of income changes on the health satisfaction of
East and West Germans in the years following reunification. The policy relevance in this article affirms that understanding the causal pathways between income and health is vital for policy design aimed at improving general health or narrowing health inequalities in society (Frijters, 2005). The key literature on which this article builds is the literature that has attempted to establish the casual effect of income on health using both cross-sectional and longitudinal survey data. The key findings in this literature are that there is a weak link between wealth, income and health (Frijters, 2005). The authors used a large scale natural experiment created by the “falling of the Berlin Wall” and the consequent reunification of Germany in 1990, in order to better establish the causal effect of income changed on health satisfaction (Frijters, 2005). It was widely acknowledged that the falling of the Berlin Wall was totally unexpected by the vast majority of East and West Germans, and resulted to a large income transfers to virtually all of the population of East Germany (Frijters, 2005).
The data source used by the authors is the German Socio-Economic Panel (GSOEP) between 1984 and 2002 which contains the GSOEP for East Germans and West Germans from 1990 to 2002 and 1984 to 2002 respectively. The GSOEP sampled individuals, aged 18 and over; this sample consisted of 46,953 persons per year observations (22,641 males; 24, 492 females) on 6198 East Germans and 176,770 persons per year observations (86,773 males; 24,492 females) on 20,617 West Germans (Frijters, 2005). The methods of analysis used are fixed-effects ordinal estimator – to control for unobserved individual heterogeneity that might determine both income and health satisfaction; and causal decomposition technique to account for panel attrition that allows for the identification of changes as respondents drop out of the sample and new respondents enter the sample (Frijters, 2005).
The equation used in the fixed-effects ordered logit model is as follows:
H*it = Xi,tβ + δt + fi + εit Hit = k ó H*it ϵ [γk, γk+1] where H*it is the latent health satisfaction; Hit is the observed health satisfaction; Xit is the observable time-varying characteristics; γk denotes the kth cut-off point for the categories; δt represents the unobserved time-varying general circumstances; fi is an individual fixed characteristics; and εit is the error term that is orthogonal to all characteristics. In the econometric framework, the endogenous variable, H ϵ {0,…10}, represents an ordinal indicator of health satisfaction as evaluated by the individual. This measure is available for a set of individuals indexed by i, where i is a number from 1 to 10; each number is observed over some contiguous subset of years indexed t, where t = 1,…..T (Frijters, 2005). The drawback of this model is that it uses a small amount of the total information available in the sample because the model reduces all the ordinal health satisfaction observations to (0, 1). The importance of this model is that it attempts to explain in which years an individual had a relatively high health satisfaction status. A positive effect of income would then mean that individuals’ relatively healthier periods occur when their incomes are relatively high (Frijters, 2005).
The causal decomposition model decomposes the changes in the expected latent health satisfaction for males and females separately in the post-reunification period and also in the pre-unification period for West Germans using the estimates from the fixed-effects models. The total changes in latent health satisfaction was decomposed into changes in: real household income, job-related variables such as maternity leave, unemployed and unemployed, family related variables such as marital status, household health related variables such as death of spouse, unobserved individual effects distribution and the unobserved average variables such as time parameters (Frijters, 2005).
The major result from this investigation is that increased income leads to improved health satisfaction, but the quantitative size of this effect is very small when changes in current income and a measure of permanent income are used (Frijters, 2005). The policy implication of these results is that increase in household income improved health satisfaction.
The strength in this article comes from the panel data that is used; the GSOEP covers a large sample of individuals and it is a longitudinal data since it compares individual differences over time. In addition, the sample is taken over a long time period – 19 year period. Also, the methods of analysis used are very comprehensive which makes it one of the strengths; the author makes use of the fixed-effects ordinal estimator to control for unobserved individual heterogeneity and causal decomposition technique to account for panel. Individuals in the study were also categorized based on socio-demographic characteristics. Another strength identified is within the external validity of the study since the author examines a unique period – Germany reunification. The Germany reunification was an ideal setting to observe the effect of income on health satisfaction since this action was completely unanticipated by the Germans. Strengths were also found within the internal validity of the study because this study entails an excellent natural experiment, therefore there were no biases in the way people behaved.
Weaknesses, on the other hand come from the self-reported health statuses which might create bias since it would be difficult to deny or confirm many claims. Lastly, the results of this study cannot be applied to other countries or cities since the study was done on a unique period in Germany.
Section 3:
The economic question that is studied by the author is the effect of income on health and mortality by using information on monetary lottery prizes to create exogenous variation in income. The policy relevance of this article involves distinguishing an association from a casual relation which suggests taking in to account the effects of income policies on the health of individuals. The key literature on which this article builds is the vast literature which has established a strong positive relation between income and health status and a negative relation with mortality.
This study comprises of an observational study in which the author uses data sets from the Swedish Level of Living Surveys (SLLS) for 1968, 1974, and 1981. The SLLS follows individuals across waves so that many individuals are included in all years and new individuals are often added in each wave to maintain a representative sample (Lindahl, 2005). An advantage of using this data set is that they contain extensive questions on health and the matched data on income and death dates from administrative registers; it also contains a question on the amount of money won on lotteries (Lindahl, 2005). The author uses OLS and IV models as the method of analysis to estimate the regressions of health on average lottery, average income and other covariates.
The author estimated the regressions of health in 1981 on the average lottery prize from 1969 to 1981 using the following equation: Hi81 = α + βLi81,13 + ϴ’Xit + ŋi81, where Hi81 represents the various measures of poor health in 1981 for individual i; Li81,13 is the average lottery prize in 1969 to 1981; Xit is a vector of demographic and family background variables, as well as socioeconomic variables measured in 1968 and lastly ŋi81 denotes a random error term. The author controlled for socio-economic variables measured as early as 1968 and not later because variables measured later are potentially endogenous with respect to lottery prizes before 1969 to 1981; the dependent variable is in bad health (Lindahl, 2005). Subsequently, the author estimated the OLS and IV regressions of health in 1981 on the logarithm of average income in 1967 to 1981 using the equations below: (1) Hi81 = α + β log (Ii81,15) + ϴ’Xit + εi81
(2) log (Ii81,15) = ∏O + ∏1 Li81,13 + τ’Xit + vi81, where Ii81,15 is the average income in 1967 to 1981; εi81 and vi81 are the random error terms. According to Lindahl, the reasons for indicating Hi81 as a function of log income are that health variables and log income often are approximately linearly related and that the usage of log income facilitates interpretation. The magnitude of the estimated income effect is β. Therefore if β = – 1, then a 10 percent increase in income yields roughly 10 percent of a standard deviation increase in good health, on average (Lindahl, 2005).
The main results of this study are that higher income causally generates good health; and income is not protective against bad health for older people. The author also found out that income causally produces fewer symptoms of poor mental health and decreases the chance of a person being overweight (Lindahl, 2005). The policy implications of these results is that income redistribution had a positive effect on health status.
After a clear review of this article, it is apparent that it exhibits both strengths and weakness. The strengths of this article lie in the source of data utilized (SLLS) since it follows individuals across waves so that many individuals are included in all years and new individuals are often added in each wave to maintain a representative sample. Strengths are found in the data source used since it contained extensive questions on health and matched data on income and death status from tax registers. Another strength identified is the method of analysis used – the way in which average disposable family income is calculated, the estimation of health regressions on average lottery prizes as well as the OLS and IV estimation of health regressions on the logarithm of average income. The author also controlled for several confounding variables such as the socio-economic variables and the standardized index of bad health in order to prevent bias in estimation.
One major weakness of this article is found in the internal validity of the study in that the data source covers a small time period – only three periods. Another weakness is that the number of individual in the study is unknown; therefore it is difficult to make generalizations. Similar to other articles, weaknesses also come from the self-reported health statuses.
Section 4:
The author quantifies the impact of a large, exogenous increase in income on health status that is associated with the South African state old age pension. This study represents a good natural experiment; the state old age pension constitutes an excellent natural experiment in South Africa because those who received this income never expected it when they were younger and when the apartheid was still strong, therefore it represents an exogenous increase in income similar to lottery winnings (Dr. Dooley). The policy relevance of this article involves pensions that are designed by the government for persons that have retired or are of old age when they are no longer earning a regular income from employment. The key findings in the literature on which this paper builds is that socioeconomic status has a large impact on health outcomes.
The data source used is the Langeberg survey which asks information on individuals’ health, mental health, social connectedness and economic status. This survey was run in 1999 on racially stratified random sample – blacks, coloureds and whites of 300 households (1300 individuals) in the Langeberg health district (Case, 2001). The data used is cross-sectional since it interviews individuals at one point in time (1999) to compare differences amongst them. The survey was developed over a four year period and was the joint product of various researchers at the University of Capetown, South Africa that included economists, gerontologist, physicians and public health experts (Case, 2001). The survey consisted of four modules; the first module was a household module which collected information from the person in the household identified as “most knowledgeable about how income is spent by the household”. The next module was for younger adults, aged 18 to 54, which collected information on work histories, earnings, health status and social connectedness. The third module was for older adult, aged 55 or greater which asked additional questions on the activities of daily living and about South Africa’s unique old age pension. The fourth module collected information on vaccines from children’s health cards and information on breastfeeding practices as well as the weights and heights of the children (Case, 2001).
The author uses ordered probits of self-reported health status as a method of estimation to examine the effect of pension income on health status. Ordered probits basically assumes that rankings of health statuses are meaningful but cardinal differences are not meaningful (Dr. Dooley). For example, if an individual rates his health on a scale of 5 and another rates his health on a scale of 1; 5 being excellent health and 1 being very poor health; it does not imply that the individual is 5 times as healthier than the one who ranks his health as very poor.
The main finding from this study is that income, in the form of an old age pension, improves the health status of all household members, in households that pool income (Case, 2001). The policy implication of this result is that there is a true effect of pension income on child health. Hence, governments should consider cash transfers as one means of improving child health.
The strength of this paper is apparent in the type of survey it uses because it interviewed individuals separately in order pull out private information to which other household members do not have access. Another strength is that the survey is designed by reliable authorities which make the survey authentic. Also, the survey took into account various races such as blacks, whites and coloured and controlled for various confounding variables such as sex, race, age and number of pensioners that could create bias in the results. Furthermore, for the purpose of comparison, the author presented ordered orbits for blacks, whites and coloured in the U.S, using data from the National Health Interview Survey (NHIS) from 1986-1995. Hence, the findings from this study can be applicable to U.S.A. In addition, strengths were found in the time period for the Langeburg survey because it was run 9 years from the time the apartheid ended in 1990. This implies that most of the pensioners never expected such a good pension. Thus, this represents a truly "exogenous" increase in old age income like winning a lottery and shows a causal effect of an increase in income. Similar to the article above, strengths were also found within the internal validity of the study because the study entails an excellent natural experiment; hence there were no biases in the way people behaved.
However, weaknesses are seen in the survey used because it samples a small number of individuals (1300 individuals). Secondly, the data was self-reported, therefore they might have been misreport which makes it difficult to confirm or deny many claims. Weaknesses are also found in the external validity of this study since the results from this study may not be representative of other areas of South Africa and other countries in Africa or the wider developing world because South Africa is an exceptional society. Another major weakness of this article is the problem of estimating the impact of income on child health due to omitted variables such as parenting skills.
Section 5:
This article studies the effect of permanent income innovations (shocks) on health for a prime-aged population with particular focus on the effect of income shocks on health over the life-cycle. Income shocks signify changes in the income of cohorts to uncover causal effects of income shocks on health. The policy relevance of this article is the extent to which income policies actually lead to improvements in health status. The key findings in the literature on which this article builds is that those with greater levels of economic resources have better health.
The data sources used are three different cross-sectional surveys that sampled more than half a million individuals over a twenty-five year period (1978 to 2003) and reported detailed information on individual’s health (both subjective and objective measures), health behaviours, income, expenditure and socio-economic factors (Jerome et al, 2005). The first survey is The Family Expenditure Survey (FES) which contains detailed information on household and consumption. This data covers the period from 1978 to 2003 and the sample size consists of 148,517 individuals. The second survey is The General Household Survey (GHS) which contain questions on health measures and risk behaviours; it covers the period from 1971 to 2003. The third survey, The Health Survey for England (HSE) unlike the previous two, sampled a small number of people and covers a small time period from 1991 to 2003 (Jerome et al, 2005). The methods of analysis used by the authors consisted of three steps; the first step was to regress the health and income variables on a suitable set of regressors capturing cohort and age effects, the second step was to use the first-differences residuals and the last was to use the GMM technique (Jerome et al, 2005). Initially, the author models income and health as stochastic processes that evolve over the lifecycle.
The main result from this study is that, income shocks has little effects on health status, but do affect health behaviours (such as eating habits) and mortality (Jerome et al, 2005). The policy implication of these results provides evidence that permanent income shocks lead to poorer health behaviour and no evidence that it directly affects health measures (such as blood pressure) .
Clearly, the strength of this article is found in the data source used since it covers the life-cycle – a twenty-five year period from 1978 to 2003 and reports comprehensive information about the individuals in study. Secondly, the survey samples a large number of individuals – more than half a million individuals. Thirdly, the surveys used are gotten from two countries U.S and England which helps for comparison, applicability and generalization.
Conversely, weakness of this article comes from the methods of estimation because it is very unclear and difficult to understand. For example, the author does not provide the meaning for the GMM technique. Weaknesses are also seen from the self-reported health status which might create bias in the results because individuals might provide wrong information. Also, this study is an observational study which involves a purely descriptive data; therefore it does not make any predictions regarding causality.
Section 6:
The authors of this article examine the effect of income on health using cross-country, time series data on health (infant and child mortality and life expectancy) and per capita income. The policy relevance of this article involves increasing a country’s per capital income to see its effect on child health. The key literature on which this article builds is the one that has estimated an income-health relationship using cross-national data. The key findings in the literature on which this article builds is similar to the results gotten in this study but the researchers who conducted these studies were unable to address issues of causality (Pritchett et al, 1996).
The data source used is the one at five-year intervals over the period from 1960 up to 1985, for a maximum of five observations per country (Pritchett et al, 1996). The author uses the OLS and IV models as the method of estimation; the OLS results on infant mortality was first reported and then the robustness of the OLS estimates with respect to variations of timing of observation, data quality and income definition were verified. The IV estimates for infant mortality for a single specification and sample was also reported. Similarly, OLS and IV estimates for total child (under 5) mortality and life expectancy were accounted for (Prichett et al, 1996). The author estimated the five year log differences for countries with GDP per capita below $6000 using observations for the years 1960 to 1985. The author utilized instrumental variables as an estimation strategy to identify the causal effect of income on health. Instrumental variables in this context are variables that are not influenced by an unobserved variable suspected to be causing both income growth and health improvement (Prichett et al, 1996).
The main conclusion of this study is that increases in country’s income raises health status (Prichett et al, 1996). The policy implication of this result is that much of the improvement in child health is due to the adoption of low-cost interventions that exists for reducing infant mortality not attributable to income changes. Hence, these low-cost interventions should be implemented along with income policies for overall improvement in child health.
The strengths of this article lie in the data source since it covers a long time period from 1960 to 1985 using a five-year interval. Strength comes from the author’s use of instrumental variables and other health-status indicators such as mortality and life expectancy. Instrumental variables are determinants of income growth but exogenous with respect to health. The importance of using mortality as an indicator of health status is that, it is available for a large number of years and countries (Pritchett et al, 1996). Furthermore, it avoids the potentially more severe reverse causation problems associated with the relationship between adult health and income growth (Pritchett et al, 1996). Another strength found was that confounding variables such as education and income were controlled for so as to prevent bias in estimation. The weakness of this article is that it does not provide information about socio-demographic characteristics of individuals and does not specify the number of individuals in the study. In addition, some of the data sources discussed in this article did not come from reliable sources.
Section 7: Summary, Unanswered Questions, and Suggestions for Future Research
Having evaluated the findings and conclusion in these articles, we see that the overall conclusions in these articles are that income has a significant impact on health status. Given that the five articles utilized different estimation methods as well as dissimilar explanatory variables, it is difficult to comment on the overall effectiveness of the policies involved in these articles. Evaluating the articles individually, Case Anne used pension income as an instrument to examine the relationship between income and health status. He found out that pension income improved the health status of all individuals in the household that pool income. This provided the evidence that there is a true effect of pension income on child health. Therefore, we can come into a conclusion that pension policy was effective in this case.
In resolving the issue with regards to the unanswered questions, it is essential to assess the strengths and weaknesses of the articles. For example, the use of instrumental variables and the control of confounding variables strengthened the internal validity of the studies. The omission of important variables weakened the internal validity of the study. For instance, in the article, “Does Money Protect against Health Status” Evidence from South African Pensions, there were omitted variables such as parenting skills which might have been correlated with health and income. Thus, an unanswered question will be, “what are other possible omitted variables?” A next question will be in regards to external validity, for example can the results in these studies be applicable to Canada as well as other countries?
To conclude, since it is clear from the analysis in each article that the existence of a causal link between income and health is still uncertain, an agenda for future research will be to utilize random variations in income and make use of panel data models. Another area of future research will be to include omitted variables in future studies so as to strengthen the internal validity of the study. A subsequent area of research will be to avoid self-reported datum in the study so as to avoid misreport from respondents. Studies should also involve more of natural experiments since people in these types of studies are not aware that they are being studied. This would strengthen the internal validity of the study as well as prevent bias in results.
...(download the rest of the essay above)