Growing gaps: The importance of income and family for educational inequalities in mortality among Swedish men and women 1990–2009

Introduction

Widening social inequalities in mortality have been observed and analyzed in Denmark, Norway, and Finland [1–3]. During recent decades, educational inequalities in mortality grew in Sweden [4,5]. Mortality, expressed as life expectancy at age 30 years, has continuously improved for both men and women on all levels of educational attainment, with the exception of low-educated women, where life expectancy at age 30 has remained largely stagnant since the 1980s [4,5]. The reasons for the development in Sweden have remained largely unexplored, concerning both trends in educational inequalities in mortality and gender differences.

Education is an important socioeconomic factor for several reasons. Educational attainment is a determinant of other socioeconomic factors, influencing income and occupational class [6,7]. Education also provides individuals with resources needed to control their own life circumstances, in that it helps them find information, make decisions, and maintain relationships [6]. Because education helps individuals throughout the life course, it has been proposed that it has a cumulative effect on health, where well-educated individuals continuously make use of resources, thus resulting in a cumulative advantage, while low-educated individuals continuously fail to do so, thus resulting in a cumulative disadvantage [6,8]. Mirowsky and Ross [6] suggested that, in a society with efficient programs that protect its members from material risks and with accessible and high-quality healthcare, health will increasingly depend on individual resources [6], implying that the importance of education as a determinant of health increases as protection against occupational and environmental health risks are provided for all. This could explain why increasing health inequalities may develop alongside improving absolute levels of health.

Studies have shown that different dimensions of socioeconomic stratification, for example education, income and occupational class, have independent associations with mortality [3,7,9–11]. Because education determines and continually influences the impact of other dimensions of socioeconomic stratification, the contribution of other socio-demographic factors should be explored in an attempt to understand educational inequalities in mortality. The educational distribution is changing in Sweden, such that younger cohorts have higher educational attainment [4], which could influence the relationship between education and other dimensions of socio-economic stratification. Educational inequalities in mortality could then be partly explained by the development in inequalities in related dimensions of socioeconomic stratification as well as by the development in the relationship between education and other dimensions of socioeconomic stratification. It has been suggested that income and family type are part of pathways between education and health [6,8]. Income and family type reflect two social phenomena in which gender is constructed and practiced, the labor market and the family, which indicates that income and family type could be important factors in shaping not only educational inequalities in mortality, but also gender differences in educational inequalities in mortality.

Income has been found to be negatively associated with mortality in Sweden [11]. Education may be considered an important determinant of income, for example through the specific training that is required for various occupations. It is likely that part of the association between education and mortality goes through income.

In their systematic review, Mustard and Etches [12] found that a majority of studies have shown that income inequalities in mortality are larger among men than among women [12]. Furthermore, men’s and women’s conditions in the labor market differ [13,14], and Swedish women receive lower incomes than men do, even at high educational levels [15]. Income could then contribute to gender differences in educational inequalities in mortality.

The importance of own income may have increased for several reasons. Income inequality in Sweden has increased [16]. Furthermore, since the 1980s, the replacement rates for social insurance for sickness, work accidents, and unemployment have decreased in Sweden [17]. Real incomes in public sector occupations dominated by women decreased during the study period [18]. Kondo et al. [19] found that income inequalities in mortality widened among both men and women during the study period and that the inequalities between high- and low-income earners increased faster among women than among men [19]. Income may not only contribute to growing educational inequalities in mortality, but also to gender differences in educational inequalities in mortality.

Mortality has been found to be lower for married and cohabiting men and women. The association remained after controlling for socioeconomic factors [20], and it was stronger among men. It has been suggested that education provides individuals with resources that help them form and maintain positive social relationships [6], which indicates that some of the association between education and mortality may go through family type. In terms of gender, family type may have different implications for men and women. Men may benefit from women being taught to take responsibility for the health of others [21], for example by engaging in less risky lifestyles [22], while women rely on their own resources regardless of family type.

Family type is also connected to income, especially for single mothers, who have been found to have both an increased poverty risk compared with cohabiting mothers [23] and higher mortality than cohabiting mothers, an association that remains after controlling for social class [24]. Single mothers have been identified as one of the groups most inflicted by the economic crisis in the 1990s in terms of increased unemployment, decreasing income, and deteriorating psychosocial working conditions [18]. Family type may thus contribute to increasing educational inequalities in mortality, and the effect may be different for men and women.

The aim of the study is to disentangle the role of income and family type in the increase in educational inequalities in mortality in Sweden during the period 1990–2009, with a specific focus on gender differences.

Data and methods

Data on Swedish-born individuals between the ages of 30 and 74 years were collected from total population registries. The use of data was approved by the KI regional ethics committee (Ref. 02-481). The data were divided into four periods, 1990–1994, 1995–1999, 2000–2004, and 2005–2009, covering a total of 529,275 deaths and 728.5 million person-months. The mortality follow-up ended in April 2009, making the 2005–2009 period somewhat shorter. The study applied an open cohort approach, allowing individuals who turned 30 during the follow-up to be included in the analysis, which means that the exact study populations varied during the different periods. Person-months at risk were calculated and divided into 5-year age categories. Data on education, income, and family type were collected for the first year (1990, 1995, 2000, and 2005) within each time period and treated as fixed covariates. Mortality was then assessed during the following 4 years in each respective period, with age treated as a time-varying covariate. Education was divided into three categories: lower secondary education or lower (ISCED 0–2), upper secondary education (ISCED 3–4), and tertiary education (ISCED 5–6). In order to capture two aspects of family formation, couple formation and parenthood, family type was divided into four categories: living alone without children, living alone with one or more children, cohabiting without children, and cohabiting with children. In the registries, cohabiting couples who were unmarried and did not have any children were classified as living alone. Readers are advised to interpret the point estimates based on family type with caution, although the general pattern and time trends are likely reflections of the true associations. Disposable household income was defined as the total income available to the household after transfers and adjusted for household composition using the Oxford method [25]; it was then divided into quintiles.

Temporary life expectancy, indicating the expected number of years lived between 30 and 74, was calculated using life tables [26]. Rate ratios were estimated using Poisson regression with robust standard errors in Stata 12. The outcome was number of deaths, and the offset was time under risk measured in person-months. All regressions were controlled for age. Analyses were carried out for men and women separately.

Results

The aim of the present study was to disentangle the role of income and family status in the increase in educational inequalities in mortality among Swedish men and women during the period 1990–2009.

The educational distribution changed during the follow-up; there was an increase in educational level among both men and women. During the 1990–1994 period, men and women had a similar educational distribution and both men and women became increasingly well-educated during the follow-up; this development was more rapid among women. During the 2005–2009 period, 18% of women and 23% of men had a low education, while 35% of women and 30% of men had a high education (Table I). Temporary life expectancy between 30–74 years, indicating the absolute level of mortality, increased in all educational groups and for both genders, with the possible exception of low-educated women, where life expectancy only improved slightly (+0.17 years) (Table II). These results indicate that the stagnation in life expectancy at age 30, previously presented for life expectancy at age 30 by Statistics Sweden [4], is also observed when enforcing an upper age limit of 74. Absolute inequalities increased among women, while the trend among men was somewhat unclear. The absolute difference between high- and low-educated men in temporary life expectancy decreased slightly from 2.99 years in 1990–1994 to 2.78 years in 2005–2009 (peaking in 2000–2004 at 3.08 years). For women, absolute inequalities increased every period from 1.82 years in 1990–1994 to 2.42 years in 2005–2009.

OPEN IN VIEWER

Table I.

Description of the data.

Year	Men			Women
	n	Deaths	Person-months	n	Deaths	Person-months
1990–1994	2,230,256	102,696	92,376,953	22,31,315	63,175	92,863,537
1995–1999	2,270,641	89,641	94,531,605	2,252,433	55,878	94,030,728
2000–2004	2,303,683	77,250	96,222,193	2264649	50,612	94,825,644
2005–2009	2,300,955	54,151	82,750,517	2246144	35,872	80,939,796
	Education (%)
	Men			Women
	ISCED 0–2	ISCED 3–4	ISCED 5–6	ISCED 0–2	ISCED 3–4	ISCED 5–6
1990–1994	39%	40%	20%	40%	40%	21%
1995–1999	33%	44%	23%	31%	44%	25%
2000–2004	27%	47%	27%	24%	47%	30%
2005–2009	23%	48%	30%	18%	46%	35%
	Family type (%)
	Men				Women
	Cohabiting	Cohabiting with children	Living alone	Living alone with children	Cohabiting	Cohabiting with children	Living alone	Living alone with children
1990–1994	25%	42%	29%	4%	27%	40%	25%	9%
1995–1999	24%	40%	31%	4%	26%	38%	26%	10%
2000–2004	24%	37%	35%	4%	26%	36%	28%	10%
2005–2009	24%	37%	35%	5%	26%	36%	28%	11%

OPEN IN VIEWER

Table II.

Temporary life expectancy.

	Temporary life expectancy 30–74 years
	Men				Women
	All	Low	Mid	High	All	Low	Mid	High
1990–1994	39.4	38.3	39.8	41.3	41.5	40.8	41.9	42.6
1995–1999	40.1	38.8	40.3	41.7	41.9	40.8	42.1	42.9
2000–2004	40.7	39.1	40.8	42.2	42.1	40.7	42.3	43.0
2005–2009	41.5	39.9	41.5	42.7	42.6	41.0	42.7	43.4

Tables III and IV show age-adjusted rate ratios with robust standard errors for education, income, and family type across the four time periods. Relative educational inequalities in mortality increased for both men and women. The increase was greater among women than among men. During the 1990–1994 period, the level of relative inequality, measured as rate ratios between the high and low educated, was similar among men and women, 1.79 among men and 1.78 among women. During the 2005–2009 period, the rate ratio was 1.98 for men and 2.10 for women. There was also an increase in inequalities among the mid- and high-educated, among both men and women (Tables III and IV).

OPEN IN VIEWER

Table III.

Age adjusted RRs for all-cause mortality, men, 30–74 years.

	Men
	1990–1994		1995–1999		2000–2004		2005–2009
	RR	CI	RR	CI	RR	CI	RR	CI
Education
Low	1.79	1.75–1.83	1.84	1.80–1.88	1.92	1.88–1.96	1.98	1.93–2.03
Mid	1.39	1.36–1.42	1.42	1.39–1.45	1.46	1.43–1.49	1.51	1.47–1.54
High	1		1		1		1
Income
1	2.26	2.21–2.31	2.28	2.23–2.33	2.84	2.78–2.91	3.19	3.10–3.28
2	1.91	1.87–1.95	1.99	1.95–2.04	2.23	2.18–2.29	2.28	2.22–2.35
3	1.55	1.52–1.59	1.59	1.55–1.62	1.62	1.58–1.66	1.61	1.56–1.66
4	1.27	1.25–1.30	1.26	1.24–1.29	1.28	1.25–1.31	1.16	1.13–1.20
5	1		1		1		1
Family type
Living alone no children	2.44	2.38–2.49	2.56	2.50–2.62	2.84	2.77–2.91	3.03	2.94–3.13
Living alone children	1.86	1.78–1.94	1.98	1.90–2.07	1.92	1.83–2.01	1.94	1.83–2.05
Cohabiting no children	1.43	1.40–1.46	1.44	1.40–1.47	1.48	1.44–1.52	1.50	1.45–1.55
Cohabiting children	1		1		1		1

OPEN IN VIEWER

Table IV.

Age adjusted RRs for all-cause mortality, women, 30–74 years.

	Women
	1990–1994		1995–1999		2000–2004		2005–2009
	RR	CI	RR	CI	RR	CI	RR	CI
Education
Low	1.78	1.73–1.84	1.95	1.90–2.01	2.04	1.98–2.09	2.10	2.04–2.17
Mid	1.32	1.28–1.37	1.41	1.37–1.45	1.44	1.40–1.48	1.47	1.43–1.52
High	1		1		1		1
Income
1	2.28	2.21–2.35	2.14	2.08–2.21	2.64	2.56–2.72	2.88	2.78–2.99
2	1.83	1.77–1.89	1.90	1.84–1.96	2.09	2.02–2.15	2.15	2.07–2.23
3	1.43	1.39–1.48	1.52	1.47–1.57	1.46	1.41–1.51	1.47	1.41–1.53
4	1.25	1.21–1.29	1.22	1.18–1.26	1.20	1.16–1.24	1.17	1.12–1.22
5	1		1		1		1
Family type
Living alone no children	2.09	2.02–2.15	2.21	2.14–2.29	2.36	2.27–2.44	2.56	2.45–2.68
Living alone children	1.66	1.59–1.73	1.66	1.59–1.74	1.59	1.51–1.66	1.71	1.61–1.81
Cohabiting no children	1.40	1.35–1.44	1.41	1.36–1.46	1.44	1.39–1.49	1.51	1.44–1.58
Cohabiting children	1		1		1		1

Income inequalities in mortality tended to be larger among men than among women. While income inequalities in mortality decreased between the two top income quintiles, rate ratios between the top quintile and the three bottom quintiles increased for both men and women. The largest increases were between the richest and the poorest quintiles, increasing from 2.26 to 3.19 among men and 2.28 to 2.88 among women (Tables III and IV).

The pattern of mortality differences by family type was similar among men and women, although the point estimates should be interpreted with caution due to imprecise classifications in the registries. Both among men and women, living alone without children was associated with the highest risk of mortality compared to cohabiting with at least one child. The rate ratio increased during the follow-up from 2.44 to 3.03 among men and from 2.09 to 2.56 among women. Living alone with a child was associated with the second highest risk of mortality. The rate ratio was higher among men than among women, while the trend was inconclusive for both genders. Cohabiting without children was associated with the smallest rate ratio compared with cohabiting with children, increasing from 1.43 to 1.50 among men and from 1.40 to 1.51 among women (Tables III and IV).

Figure 1 displays the trend in four estimates for men and women; the rate ratio for all-cause mortality between the high and low educated controlling for age, age and family type, age and income, and age, family status, and income. There were clear gender differences in how family type was related to educational inequalities in mortality. Variation in family type explained part of the level of inequalities among men, but it did not explain the increasing trend. For women, controlling for family type did not attenuate the educational inequalities in mortality at any point in the study period. The results indicate that family type was on the pathway between education and mortality among men, but not among women.

OPEN IN VIEWER

Figure 1.

Rate ratios for all-cause mortality between high and low educated and 95% confidence intervals among Swedish men and women, 30–74 years.

Controlling for income substantially attenuated the educational inequalities in mortality, although the remaining educational inequalities were larger among women. This indicates that income is a more important pathway connecting education and mortality among men than among women. Variation in income also explained the positive trend in educational inequalities among men during the entire study period. The increase in educational inequalities between 1990–1994 and 1995–1999 among women could not be explained by controlling for income, but between 1995–1999 and 2005–2009, the trend among women was attenuated. This suggests that the pathway connecting education and health through income can account for the increasing educational inequalities in mortality among both men and women.

Controlling for both family type and income further attenuated the level of educational inequalities among men, but did not influence the trend, compared with controlling only for income. This result indicates that income and family type represented different pathways connecting education and mortality. Among women, controlling for family type and income simultaneously resulted in more or less similar results as controlling for income did, both in terms of the magnitude and trends in educational inequalities.

Discussion

The results confirmed previous observations of decreasing overall mortality and an increase in educational inequalities in mortality. The study also supported the observation that Swedish low-educated women seem to have experienced little or no improvement in mortality during recent decades.

Mackenbach [27] and Bambra [28] have identified the comparatively large social inequalities in mortality in the Nordic countries as a paradox and a puzzle, respectively [27,28]. Although this study does not contrast the development in Sweden to that in other countries, it contributes empirical evidence on what mechanisms shape trends in educational inequalities in mortality in Sweden.

Although the estimates should be interpreted with caution, the results indicated that the mortality of both men and women was associated with family type, but while family type was part of a pathway connecting education and mortality among men, this was not the case among women. In a study carried out on Swedish total population data, Torssander and Eriksson [29] found that women’s education had an impact on both own and their spouse’s mortality in married couples, while no association was observed for men’s education their spouse’s mortality. Umberson [21] argued that women were more likely than men were to control the health of the immediate social environment [21]. Part of the educational gradient in mortality among men could then be the result of higher-educated men having partners with a high education and benefiting from their resources, while the educational gradient among women was not influenced by their partner’s education.

Income explained a larger part of the association between education and mortality among men than it did among women. It is possible that household income is a more direct assessment of the socioeconomic position of men than that of women, due to gender differences in labor market participation and wages [14]. Employment status was not possible to include in the empirical analysis due to data limitations; the registers are collected for administrative purposes and the available indicators of employment status are not consistent over the study period.

Because men generally contribute a larger share of the income, household income is a closer estimate of men’s than of women’s income. On the other hand, if more women work part-time, it is possible that individual income level may underestimate their positions in the occupational structure. Ross and Mirowsky [30] proposed that education is a more important determinant of health for women, as women generally have fewer resources than men do [30].

The increasing trends in educational inequalities in mortality were largely attenuated when controlling for disposable household income, both among men and women (Tables III and IV). Perhaps surprisingly, the increase in educational inequalities in mortality was not due to a higher concentration of individuals with a low disposable income among the low educated. Rather, the proportion of individuals with a low disposable income among the high educated seem to have increased somewhat during the study period (see web supplement). Instead, it is the relative death risks in the bottom income quintiles that have increased during the study period [19]. Although the concentration of individuals in the bottom income quintile did not increase among the low educated during the study period they are consistently more likely to be in the bottom income quintile than the high educated. This, in combination with the increase in relative death risk, resulted in wider inequalities in mortality by educational attainment. The growing income inequalities in mortality may also account for the faster increase in educational inequalities among women compared to men. Kondo et al. [19] found that income inequalities in mortality increased faster among women than among men [19], and low educated women are more likely to be in the bottom income quintile compared to low educated men (see web supplement). There are several possible explanations for the increasing income differences in mortality. Income is potentially connected to mortality through material, psychosocial, and behavioral pathways [19] and further research is needed to identify the specific mechanisms behind the increase in differences in mortality between income groups.

Understanding the specific social processes that generate health inequalities is important when formulating policy aimed at reducing health inequalities. The present study adds to our understanding of increasing educational inequalities in mortality in Sweden by examining the role of income and family type. Family type explained parts of the association between education and mortality for men only, but could not account for the observed trends. Income explained part of the association between education and income among both men and women, as well as the trend of increasing educational inequalities in mortality.