During a long period before the mid 1990s life expectancy did not change much in Denmark . The stagnation in life expectancy among persons with a low educational level or income appears clearly from the figures. At the end of the century life expectancy began to increase for all groups although with a slower rate among the socially disadvantaged groups.
The overall conclusions were remarkably consistent transversely to the different methods. We expected that the considerably changes in the size of the educational subpopulations during a quarter of a century would result in different trends in social inequality depending on whether the four educational levels were defined by fixed categories or by “flowing” quartiles. One result leaps to the eye when trends are estimated on the basis of equivalent disposable income. Life expectancy among women in the lowest income quartile improved more than it did for women who belong to the two middle quartiles. It appears from Figure 4 that the middle-income half of the Danish female population lag behind, but a similar trend was not seen for educational differentials (Figure 2). This paradox might be related to changes in the labour market structure, the flexicurity model, and wage rate developments, family constellations and how these factors interact with the construct of equivalent disposal income.
No simple theory exists that explain the different developments, but to some extent mortality trends are related to health-related behaviour and the so-called “intervention generated inequalities” that characterizes the Scandinavian welfare societies might contribute to increasing social inequality in health [12, 13].
Health-related behaviours vary between social groups. For instance smoking is more prevalent among social disadvantaged people. But the social gradient in life expectancy is seen irrespectively of smoking category . The difference between 30-year-old male never smokers with a high and a low educational level was 2.3 years, whereas the similar difference among heavy smokers was 3.7 years. The educational disparities for women were 1.3 years for never smokers and 1.9 years for heavy smokers . Thus, the impact of smoking seems to be greater among people with a low educational level in spite of a shorter lifetime. This might in part be explained by higher tobacco consumption among heavy smokers or more clustering of risk factor exposures among lower educated smokers.
The decline in smoking prevalence is faster among socially advantaged than socially disadvantaged people . Thus, heavy smoking prevalence among men with less than 10 years of education increased from 26.0% in 1987 to 31.4% in 2005 whereas it decreased from 12.2% to 8.8% among men with more than 14 years of education. For women with less than 10 years of education the prevalence increased from 18.5% to 29.1% and for high educated women it decreased from 17.6% to 7.6%. The same direction of trends were seen in all age groups with the only exception that prevalence of heavy smoking among high educated men aged 65 and over increased from 5.2% in 1987 to 7.7% in 2005 . Prevalence of moderate smoking declined in all groups.
High alcohol consumption is characterized by a general upward trend, although stagnation among the highest educated persons results in almost equality between educational groups . Overweight and obesity increased in all educational groups, but the strong social gradient is maintained .
A result based on data from the British Whitehall II study suggests that health-related behaviours (smoking, alcohol consumption, unhealthy diet and physical inactivity) assessed longitudinally could explain 72% of social inequalities in mortality .
A recent Finnish study investigating trends in income differentials in life expectancy for 35 year-olds found that the difference in life expectancy between the lowest and the highest income quintile widened by 5.1 years for men and 2.9 years for women from 1988 to 2007, the difference being 12.5 years for men and 6.8 years for women in 2007 . For both sexes the main reason for the increased disparity was stagnation among those with the lowest income. For men this is in line with our results, whereas the change was more favourable among Danish women in the lowest income quartile. However, social classification based on educational level in our study also indicated poor improvement in survival among women with the shortest education.
The difference in life expectancy between higher non-manual and unskilled Swedish 20-year-old men increased from 2.1 years in 1980 to 3.8 years in 1997. For Swedish women the increase was from 1.6 years to 2.2 years .
Also in Belgium life expectancy increases more among people with a high educational level compared with people with a low educational level and for women without formal education life expectancy even seems to decrease .
Studies in the USA similarly demonstrate increasing social inequality in life expectancy [20–22]. Thus, only modest improvement in life expectancy during 1981–2000 was seen among less educated black women and white non-Hispanics men and women, while the more educated groups experienced substantial gains . Another US study estimated trends in life expectancy differentials by lifetime earnings and found that between 1990 and 2000 differentials in life expectancy between ages 35 and 76 has increased from 2.7 years to 3.6 years between men in the lowest and the highest lifetime earning quintile. For women the difference increased from 0.7 years to 1.5 years .
An important strength of our study was that death rates were based on data for all Danish citizens although information on educational level was only available for ages under 65 for the entire period. As sex- and age-specific death rates were calculated exactly within each educational and income level, life tables could be constructed, apart from educational-specific figures for the elderly. However, the lack of educational information on the elderly is due to the systematic data collection procedure at Statistics Denmark and not related to social characteristics. We assumed that the relative disparities in death rates in 2010 between educational groups were valid for ages above 65 for the entire period. If we made the conservative assumption that the mortality rates after age 65 were equal for all educational groups the difference between the lowest and the highest educational quartile in life expectancy at age 30 grew from 2.0 years in 1987 to 4.1 years in 2011 for men and from 1.2 years to 2.6 years for women.
The division into educational and income quartiles defines synthetic life tables because educational or income membership of quartiles changes during life, which is not reflected in the period life table birth cohorts. But this construct makes comparisons between social categories possible and is similar to the classical life table assumption of constant age specific death rates for all birth cohorts.