Twenty-five year trends in body mass index by education and income in Finland

In the developed world those with lower socioeconomic position have a higher body mass index (BMI) and are more likely to be overweight or obese than those with higher socioeconomic position [1, 2]. The results obtained from Europe [3, 4] are in line with results from other continents.

Education and income are among the most common socio-economic indicators used in studies on obesity and overweight. European trend studies suggest that educational differences in BMI have persisted during recent decades [5‐8]. The European studies [5‐8] have analysed educational differences only, while both education and income have been included in some North American and Canadian studies. The trends of educational and income-related differences in BMI were divergent in the USA in 1960–1980. Among men the educational gradient turned from positive to negative, among women the negative association between education and BMI became stronger. Income was positively associated with BMI among men but negatively among women [9, 10]. Later North American studies observed parallel increases in BMI in all educational and income groups in 1986–2006 [11] and 1960–2008 [12] or did not find consistent changes in educational or income gradients (1970–2008) [13]. In Canada education and income related differently to BMI in 1978 and 2005. Education was inversely associated with BMI and there was no evidence of narrowing. The association was stronger for women than men. No clear association between income and BMI was observed for men whereas for women the association was inverse in 1978 but positive in 2005 [14]. The previous trend studies do not give a consistent picture of educational and income-related gradients of BMI. The associations of education and income with BMI seem to change over time and vary by populations and their subgroups.

European studies have analysed in cross-sectional settings the relative importance of several dimensions of socio-economic position as determinants of BMI. Educational level seems to have the most systematic association with BMI. Two studies compared variation in BMI or overweight in several European countries [3, 15]. They showed that educational level was a stronger determinant of overweight than occupation or income. In particular, the association with income and overweight diminished after adjusting for education.

Education and income are both independent and interrelated indicators of socio-economic position: education enhances knowledge and non-material resources while income provides material resources. Educational qualifications contribute to later income. Educational level is more stable than income over the life course. To understand the interrelations between the socio-economic determinants of health it is not sufficient to search for just one statistically strongest socio-economic determinant [16]. The same can be applied to studies on socio-economic determinants of BMI. The previous studies have not paid attention to interactions between the determinants. The analysis of interactions between education and income over time could illuminate the origin of socio-economic variation in BMI and contribute to identifying population subgroups that are most vulnerable to weight gain.

Educational and income differences in BMI may change according to the economic situation of the study population. The economic situation changed greatly in Finland during the 1990s. After a period of growth in the late 1980s, the Finnish economy experienced a recession. Income inequalities started to increase and this trend has continued since. The increasing income equalities may have contributed to the increase observed in health inequalities [17, 18]. According to a Finnish review, BMI has increased remarkably but the socioeconomic differences seem to have remained rather stable [19]. However, it could be assumed that during a period when the distributions of overweight, education and income have changed, the associations of education and income with BMI and the interactions between them might also have changed.

The overall aim of this study is to analyse the trends of socioeconomic differences in BMI among Finnish men and women from 1978 to 2002, during a period of increasing income and health inequalities. In order to analyse the importance of education and income, a specific emphasis will be put on a comparison of the two measures of socioeconomic position and on their possible overlap. We assume that changes in income-related differences in BMI are more obvious than the educational ones because the overall income-inequality became more prominent during the study period. Finally, we will cross classify education and income at different time periods in order to identify the population subgroups that have been most vulnerable to weight gain.

The specific research questions were to examine whether:

This paper is based on cross-sectional surveys repeated annually among 15–64-year-old Finns since 1978. The survey data was complemented with register data on household income. We will first present the associations of educational level and income with BMI from 1978 to 2002, and then analyse whether income differences in BMI are stable and similar in each educational group.

The data were derived from a series of repeated cross-sectional health behaviour surveys. A nationwide random sample from the Finnish population aged 15 to 64 years was drawn annually during the period 1978–2002, with some 5000 Finns receiving a mailed questionnaire each year. The response rate has varied between 86 and 75% among women and 83 and 62% among men. The rate has declined over time [20]. The data have been collected according to the general ethical rules applied in the National Institute for Health and Welfare (and previously by the precursor National Public Health Institute). Individual participants cannot be identified from the data and only authorised persons have access to the data. Register linkages were not possible after the 2002 questionnaire because of changes in the ethical rules.

Employed respondents aged over 25 years were included in this study (N=57 351). All cases of missing data on income (4731), education (852) and BMI (666) were excluded. The proportion of missing income data was high during the first study period (9.4% for women, 16% for men) but decreased systematically over time. During the last study period 4.4% of income data was missing among women, 5.0 % among men. The study comprises data for a 25-year period (1978–2002) relating to 25 339 men and 25 390 women.

Educational level refers to years spent in full-time education as reported in the questionnaire. To take into account the increase in the general level of education in Finland since 1978, we divided the educational years into gender-specific tertiles per study year. Consequently education was classified into three groups: low, intermediate and high education (Table 1).

The information on respondents’ income was linked to the data from the national tax register of Statistics Finland. The linkage to data from the year 1985 was not possible due to missing ID-numbers in the survey data, meaning that the survey year 1985 could not be included in the study. Income data for the period 1978–1983 were based on participants’ income in the year 1980 and for the period 1984–1986 from the year 1985. Thereafter, data were available for each study year. The average household income per consumption unit was calculated from the total household income (gross income without income transfers). Consumption units were calculated according to OECD guidelines (=ECD 1982), with the household’s first adult person receiving a weight of 1.0, the other adults receiving a weight of 0.7, and children a weight of 0.5. Income was divided into tertiles for the analyses, representing low, intermediate and high income groups, which were calculated separately per study year and gender to represent the relative change in income (Table 1).

The tertiles of education and income were formed along the same principles in order to improve the comparability of the two indicators of socioeconomic position.

Body mass index (BMI) was calculated from self-reported height and weight (kg/m2). Persons whose BMI was 25 or over but below 30 were considered overweight, those with a BMI of 30 or more were classified as obese. The preliminary analyses were carried out by using both the prevalence of overweight and obesity as dichotomous variables and the mean BMI as a continuous variable. However, the trends and socioeconomic variations did not differ according to the outcome variable. As the main focus of this study was on the changes of BMI in socioeconomic groups and not on the proportions of obese or overweight persons, mean BMI was chosen for the final analyses.

The respondents were divided into 10-year- age groups 25–34, 35–44, 45–54 and 55–64. The research years were divided into five periods: 1978–1982, 1983–1987, 1988–1992, 1993–1997 and 1998–2002.

Our study showed a remarkable increase in BMI among 25–64-year-old Finns between 1978 and 2002. In women both education and income showed strong inverse associations with BMI. The magnitudes of the associations fluctuated but stayed statistically significant over time. Among the Finnish men, socioeconomic differences in BMI were more complicated. Educational differences were weaker than among the women and BMI did not seem to vary by income. However, income differences varied according to educational level. At the turn of the century, the high income men in the lowest educational group had the highest BMI whereas the income pattern in the highest educational group was the opposite. Our study did not show a systematic overall increase or decrease in the socio-economic differences between 1978 and 2002 but demonstrated diverging trends in the sub-groups of the studied population.

When interpreting the results, some methodological considerations need to be taken into account. Weight and height were self-assessed by the respondents. Self-assessed BMI is an underestimate of the measured BMI [21]. A recent Finnish study [22] that was based on measured BMIs from 1997 to 2002 demonstrated about 7% higher BMI in women and 4% higher BMI in men compared to our data. However, the increase was similar in both studies. Educational differences in measured BMI’s were also in line with our observations, especially among men. Among women the educational differences were smaller than in our study. The comparison of our results with the previous Finnish study suggests that women with higher education had a tendency to underestimate their body weight. Similar results have been reported from Australia [23]. As the focus of our study was not in absolute figures but on trends and variations between population groups, the bias caused by self-assessment barely disturbs our main conclusions.

Another methodological problem typical for surveys was also observed in our study. The response rate has decreased systematically over the twenty-year period, and more in low educated groups [20]. Thus, the socioeconomic differences are probably greater in the sample than among the respondents especially during the later study periods.

The use of household income data obtained from the national tax register was one of the strengths of our study. Unfortunately, the proportion of missing income data was high during the first study period. Subjects with missing income information had higher mean BMI (p<0.001) than the others during 1978–1982; later the difference was no more statistically significant. This suggests that our results overestimate the increase of BMI over time. As a consistent increase of BMI has been observed in other Finnish studies the possible overestimation may not be significant. Anyway, the results concerning the period 1978–1982 need to be interpreted with caution.

An advantage of a mailed questionnaire is its broad coverage and representativeness of the whole adult population in Finland. Another methodological strength of our study was the construction of comparable classifications of education and income for the analyses. The statistical power of the large number of respondents allowed stratified analyses of educational and income groups over a long time period.

Educational differences in BMI, overweight and obesity have been shown to exist in practically all Western countries during the last two decades [6, 15, 24‐27]. Another common finding confirmed by our study is that educational differences are more systematic than those between income groups [3, 9, 10, 15, 28, 29].

In line with previous studies, our study showed that socioeconomic patterns of BMI are more clear-cut among women than men and educational differences can be observed among both genders [30‐32]. Roskam and Kunst [15] compared socioeconomic variation in European countries and analysed differences in the prevalence of overweight by educational attainment, occupational class and household income. Among men no occupational variations in overweight were observed after adjusting for the other socioeconomic variables. After controlling for education and occupation, household income was negatively associated with overweight in women but positively, although weakly, in men. Researchers have given several explanations for the larger socioeconomic variation among women. Women in higher socioeconomic position might be more concerned about their body shape and make more efforts to lose weight [30]; they also tend to have better employment status and less family demands than women in low socioeconomic position [4]. Among men, weight concerns and family demands might play a smaller role in all socioeconomic groups.

Our study did not show a consistent decrease or increase in educational differences of BMI. This finding is in line with previous Finnish (17, 20) and European (5–8) studies. We assumed that changes in income-related differences in BMI would be more obvious than educational ones because the overall income-inequality became more prominent during the study period. We did not observe an increasing income gradient across the studied subgroups. However, we identified diverging income trends among men in the highest and lowest educational group. In the 1990s, the upward-trend in BMI was the clearest among high-income men with low education and among low-income men with high education. By the end of the century the income gradient was negative in the lowest educational group but positive in the highest. These groups of men are in a contradictory situation, they belong to a low and high socioeconomic group at the same time. Maintenance of normal body weight seems to involve both the cultural resources associated with high education and the material resources associated with high income.

The low-educated high-income Finnish men have probably started with physically demanding jobs, without the motivation and skills to maintain an ideal weight as do the higher educational groups. High income has probably led to a decrease in physical activity and higher consumption of alcohol. A similar explanation does not apply to high-educated but low-income men. Their situation might reflect the consequences of economic downturn and the fact that they did not reach the income level of their high-educated peers. Picket et al. [33] and Wilkinson et al. [34] refer to psychosocial consequences of low socioeconomic position. The low socioeconomic position reduces people’s control over their life and work and has effects on health behaviours and body weight.

During a period of economic downturn and an overall increase in income inequalities the mean BMI increased significantly but no systematic changes in educational or income-related differences were observed. Among Finnish men the association of income with BMI depended on educational level. At the turn of the century the income gradient was negative in the highest educational group but positive in the lowest. Our results encourage further research on the pathways between income, education, living conditions and the increasing BMI.