McIsaac SJ, Wilkinson RG. Income distribution and cause-specific mortality. European Journal of Public Health 1997; 7: 45-53

This important paper considers the comparability of international data on income distribution, makes adjustments for differences in response rates, and tries to look at cause of death specific associations with income distribution. The discussion includes interesting speculations such as that road accident rates may reflect underlying social stress in society or that different predominating causes of death in countries reflect varying cultural patterns. Understanding the problems of making comparisons among countries make this important reading for those wishing to circumvent these difficulties. The paper is outlined in considerable detail below.

Abstract

The aim was to identify the age-, sex and cause-specific premature mortality rates contributing to the association between life expectancy and income distribution in developed countries. Income distribution was calculated for the 13 OECD countries and years for which the Luxembourg Income Study held data. The potential years of life lost (1-65 years) by sex and cause, as well as the age- and sex-specific all-cause mortality rates and standardized mortality ratios for children 1-19 years were calculated from data supplied by the WHO. On finding evidence suggesting that reported income distribution is strongly affected by low response rates in some income surveys, we used 2 measures of income distribution: that among households where the 'head of household' was aged less than 65 years (weighted by response rates) and that among households with children (among whom response rates were thought to be higher). Partial correlations and regressions controlling for the year were used to analyze the relationship between mortality and income distribution. Both measures of income distribution showed broadly similar results. A more egalitarian distribution of income was related to lower all-cause mortality rates in both sexes in most age groups. All 6 major categories of cause of death contributed to this relationship. The causes of premature mortality contributing most were road accidents, chronic liver disease and cirrhosis, infections, ischaemic heart disease among women and other injuries among men. Income distribution was associated with not only larger absolute changes in mortality from these causes, but also with larger proprotionate changes. Suicides and stomach cancer tended to be more common in egalitarian countries.

Introduction

Summary of studies to date finds:

• at least 8 different research workers or groups have reported statistically significant relationships between income distribution and measures of mortality
• a ninth has reported a relationship with height

Data sources for these:

• 1 study with data on developing countries
• 3 studies on a mixture of developed & developing countries
• 5 studies exclusively on developed countries

• fertility
• maternal literacy & education in developing countries
• average incomes
• absolute poverty
• racial differences
• smoking
• various measures of provision of medical services in developed countries
• no attempt yet to break down association between income distribution & life expectancy, or all-causes mortality into age-, sex- and cause-specific mortality rates

Data and Methods

• Income Distribution Data from LIS
• looked at shares of personal disposable income (after taxes & benefits)
• adjusted for household size
• 2 sets of income distribution data
• that among households where the 'head of household' was aged less than 5 years (weighted by response rates)
• that among households with children aged 18 and under(among whom response rates were thought to be higher) (CHILDHSE)

Death rates

• WHO data, 
• 14 age-specific all causes rates (male & female)
• infants 1-4 years
• 5 year intervals to age 64
• standardized mortality ratios for children 1-19 years
• potential years of life lost (deaths between 1-65 years) 
• cause specific
• infectious & parasitic diseases
• malignant neoplasms
• malignant neoplasms of stomach
• malignant neoplasms of trachea, bronchus and lung
• malignant neoplasms of female breast
• diseases of circulatory system
• ischaemic heart disease
• cerebrovascular disease
• diseases of respiratory system
• bronchitis, emphysema, asthma
• diseases of digestive system
• chronic liver disease & cirrhosis
• injury and poisoning
• MVA's
• suicide & self-inflicted injury
• controlled for year of data

Response rates

• expected previously found strong relation between small number of countries & life expectancy was found to be weaker, and when broken down into age-, sex- and cause-specific components, largely disappeared
• household income response rates sometimes fell as low as 50%, & disproportionately low among poor, & to lessor extent among rich, suggested by 4 pieces of evidence 
• 1. lower response rates associated with higher reported shares of income received by lower deciles & lower shares by upper deciles (would expect societies with genuinely narrow income distributions to be more cohesive and have higher response rates, but found inverse) suggests low response rates may lose poor +/- rich, leading to underestimation of income inequality
• 2. weighing countries by square of proportion responding (to discount countries with low response rates), got expected distribution between mortality & income distribution
• 3. Family Expenditure Survey (source of British data held by LIS) has been found to underrepresent poor & to lesser extent, the rich. low among people from Northern Ireland, 'new commonwealth' and Pakistan, & households with _2 cars, as well as self-employed
• 4. response rates of households with children are high in the Family Expenditure Survey, and correlations among countries are high between mortality rates & income distribution whether data weighted or not. 
• hence can deal with low response rates by:
• weighing data
• looking at CHILDHSE, but this will cover a smaller proportion of single men than women 

Results

Decile Shares

deciles of income share are correlated with mortality, negatively in the poorer deciles (mortality rates are lower where less well-off people receive a larger share of income), and positively in richer deciles.

the single measure of income share that has the strongest correlation with all-cause mortality is when looking at the proportion of income received by the poorest 30% of the population

when looking at cause specific mortality, the income proportion with the strongest correlation varies with the cause, with heart disease deaths being more correlated with the proportion of income received by the poorest 20 or 30%, while for cancers, it is the proportion received by the least well-off 90%. 

Age and sex

correlations of deaths with income received by the poorest 30% is strongest among children and younger adults, though present in all age groups

Causes of death

top 4 contributory causes for relationship with income distribution

• chronic liver diseases and cirrhosis
• traffic accidents
• infections
• ischaemic heart disease

Discussion

• internationally comparable data is limited
• it may not be possible to take this kind of analysis further on basis of data currently available
• association between income distribution and mortality is spread over most of life
• while within a population, male mortality is sometimes more correlated with socioeconomic status, this doesn't appear to be true with income distribution, it affects both sexes
• suicides and stomach cancers tend to be more common in egalitarian countries, and is consistent with the data from Japan, that is not included in this study using LIS data
• this has ramifications with Durkheim's different kinds of suicide related to the degree of social integration
• socioeconomic differences in mortality appear in different causes of death in different countries, perhaps finding expression through a variety of national cultural patterns
• France: alcohol & cirrhosis contribute to excess mortality, while heart disease does not
• Britain: respiratory & cardiovascular disease do
• Finland: accidents & cardiovascular disease do
• road accidents may reflect socioeconomic stress throughout society
• could amount of courtesy and cooperative driving behavior be a sensitive Indicator of people's attitudes to other members of society as unknown fellow citizens?

Keywords

• cause specific mortality
• economic growth
• gender differences
• Gini coefficient
• gross national product
• health inequalities
• heart disease
• homicide
• income
• income distribution
• income inequality
• inequality
• infant mortality
• international
• life expectancy
• mortality
• population health
• poverty
• psychosocial factors
• relative deprivation
• social class
• social stratification
• socioeconomic status
• suicide
• traffic accidents
• unemployment