The conditional effect of income distribution on mortality risk of men in Tunisia: Poverty effect or wealth effect?

ABSTRACT

The paper examines the conditional effect of income on mortality risk of men in Tunisia in a population based case-control survey study combining individual data from two databases on 11,231 survivors and 1,498 deceased. The main question addressed in the study is whether inequalities in health, as measured by mortality, in Tunisia mainly reflect the effects of poverty or are they caused by a socioeconomic gradient? Using logistic regression models with various income specification, our empirical findings provide strong evidence of an excess mortality associated with poverty, but also, unexpectedly, an under mortality associated with highest incomes. Beyond an absolute poverty effect, we fail to find support for the hypothesis of a concave relationship between income and mortality since we highlight a protective effect of the highest incomes. Besides, under the different income specifications estimation results confirmed the significant effect of income on mortality risk, regardless of the socio-professional categories effects. This effect is not limited to poverty, since the mortality risk decreases continuously throughout the income distribution calling into question the effects of socioeconomic gradient on mortality risk.

JEL CLASSIFICATIONS:

• I12
• C12
• D31

KEYWORDS:

• Income distribution
• Mortality
• Socioeconomic gradient
• Health capital model
• Tunisian population

Notes

¹ In the literature, the shape of the relationship between mortality and income is generally examined through three main hypotheses: the absolute income hypothesis, the relative income hypothesis and the income-inequality hypothesis.

² Both databases are not publicly available, but can be accessed by special request to the Tunisian National Institute of Statistics, for DHS, and to General Directorate of Taxes, for HITS. The use of two databases is due to the lack of income data from the DHS.

³ The equivalence scale considered here is independent of the age of the taxable household members. We were constrained in our choice by the availability of data provided by the DHS. We used, instead, the equivalence scale approximate formula estimated by the INS, in which the number of consumption units (CU) is calculated as follows: CU = 1 + 0.3(n − 1).

⁴ The “executive” social category can be considered as a good instrument of the level of education, since this category is distinguished mainly by higher level of training.

⁵ “Workers” occupation can be considered as a good proxy for working conditions, since it is recognized that manual workers experience more difficult working conditions than executives or intermediate employees.

⁶ The variables were introduced in continuous form, it is normal that the coefficients of the continuous variables are close to zero and that their odds ratios are very close to 1. The interpretation is therefore based on the sign of the coefficients, as well as on their statistical significance levels. The estimated coefficients on the linear and cubic terms are negative while the squared terms are positive, indicating a logistic relationship.

⁷ To conserve space, the distribution by income deciles has been omitted.