This study investigated the hypothesis that education might protect against the obesogenic effects of having an occupation in the agricultural sector compared with having an occupation in the industrial and service sectors in a country undergoing rapid economic transition, by examining whether education modified the association between occupation and central obesity in a population of Chinese women. The results showed that having no education more than doubled the odds of central obesity associated with having a non-agricultural occupation compared with an agricultural occupation in this population (OR; 95%CI: 2.25; 1.55, 3.25). Differences in health behaviours including meat, fruit and vegetable consumption may have a role in explaining this. In women with any level of education no evidence of an association between occupation and obesity was found.
While the data are limited by their cross-sectional nature this study contributes new understanding to the SES-obesity association, demonstrating different associations with central obesity for two key SES indicators.
The need to examine the inter-relationships between SES more closely in transition settings. In high income countries, education, occupation and income levels tend to be collinear but this may not be the case in lower income countries. In particular, this study contributes to the growing evidence that education may have different properties as an SES indicator in relation to obesity in low- and middle-income countries compared with indicators that are more closely linked to material circumstances such as occupation and wealth
Comparison with prior studies
Many of the key SES-adiposity studies in lower income countries use multi-country data. Few have incorporated occupation as an SES indicator, favouring education and wealth instead as they are considered more reliable measures of SES in lower income contexts and, therefore, more comparable across time and space
. There are currently no studies in the epidemiological literature investigating the inter-relationship between education and occupation in relation to female obesity in lower income settings to our knowledge, although independent effects of education and wealth on obesity have been reported in single country studies. In Peru, the Philippines, China and Brazil
[21, 35–38] a positive association has been found between wealth and obesity together with an inverse or protective association between education and obesity (usually among women but not men) when both are taken into account in the analysis. Recent data from China corroborate the emergence of a protective association between education and obesity at least among urban residents
 and women
[22, 23]. These patterns are comparable to findings in the early 1990s in Eastern Europe during its economic transition, when education and material circumstances acted differently as SES indicators of health outcomes
Occupation has been considered to be a good measure of material circumstances
 and found to be negatively associated with obesity in women (higher occupational status-lower obesity)
 in high income countries. However, the nature of the association in low- and middle-income countries is likely to vary according to the specification of the occupational variable and the level of development of the country. The specification in this study was intended to capture the variation resulting from the differences between agricultural-based jobs and jobs in industry and services, in terms of the daily physical and food environment that affect health behaviours.
The results suggested that dietary behaviours including food and alcohol consumption accounted for part of the association between occupation and central obesity in the group with no education but it was not possible to fully assess the independent role of physical activity. Other studies from China suggest that diet may be more important than physical activity in explaining the association between occupation and markers of excess adiposity. In a study of 7,011 Chinese women 50 years and older examining the association between education and occupation (specified as manual vs. non-manual) and a composite index of the metabolic syndrome (including WC) adjustment for physical activity had little impact
. Another study
 examining urban/rural differences in central obesity in 8,014 women attributed a higher proportion (43.8%) of the excess risk of central obesity in urban areas to diet and a lower proportion to physical activity.
The findings from the present study corroborate others in documenting the high levels of central obesity in China. Estimates from the China Health and Nutrition Survey from 1993 to 2009, representing a total of 52,621 participants, showed that there was a significant increase in central obesity in Chinese women 60 years and over – the age group examined in this study - from 47.4% (SE: 2.4) in 1993 to 66.5% (SE:1.3) in 2009
Plausible and competing explanations
Excess adiposity is increasingly viewed as a mismatch between biology and the environment
. Economic transition to a higher income economy is usually associated with a move from a predominantly agrarian and/or subsistence economy to a predominantly industrial and/or service-based economy resulting in changes in dietary composition, occupational patterns and leisure time activities conducive to excess body fat storage
[45–47]. But the mechanisms that explain the SES-adiposity association are complex and not fully understood
: the association may be bidirectional and confounded by other factors such as heredity, health behaviours and general socio-cultural norms
, as well as show period variation
At its most basic, higher status occupations might influence obesity risk through levels of physical activity
, however they are also likely to be associated with living in an urban environment and, therefore, the consumption of higher levels of foods rich in fat and sugar and possibly lower leisure time physical activity. Furthermore, women’s entry into the labour force can lead to an increased reliance on processed or ready-made foods as well as a greater number of visits to restaurants and other prepared food outlets
On the other hand, women’s education is known to be protective for a variety of health outcomes
[15, 16, 53–55]. In the case of obesity, it may allow women to make better dietary and exercise choices through the cognitive advantages that can operate in a number of ways including improved access to and understanding of health related information, clearer risk perception related to lifestyle choices, altered time preferences and better self-control
. However, there is evidence that these cognitive advantages are unrelated to time-preferences and personality
. Alternative explanations include that more educated women may conform to different cultural norms of physical beauty that favour slimness
[48, 56] and that better education may operate through psychosocial pathways by affording better job-control and therefore lower stress levels which modulate inflammatory responses linked to obesity
It is important to remember that countries undergoing rapid transition may experience an influx of new food products including high-calorie and nutrient-poor processed foods alongside other changes in lifestyle. In other words, the nature of disease risk changes. Yet public health infrastructure may not be equipped to deal with these nor engage the public in managing these risks. The combination of longstanding food insecurity, aggressive commercial marketing and inadequate public health systems may result in a large asymmetry of information between consumers and sellers when assessing consumption-related health risks. This could give those with higher education levels an advantage because they may be able to correct cognitive biases within this imbalanced information environment more easily.
Implications of the findings
The main implication of the findings in this study is that obesity risk in low- and middle-income countries may not solely be determined by changing material circumstances associated with working and living in a different economic environment but that having a better level of education may protect against the detrimental effects of these significant changes in living conditions. This may occur through cognitive mechanisms that promote better dietary and leisure-time physical activity choices, and empower individuals in navigating new disease risks resulting from economic transition. Data from China show that the prevalence of obesity has increased at a faster rate in poorer rural areas than in richer urban areas
 and that lower income groups have disproportionately increased their consumption of animal fat and edible oil and reduced their consumption of healthier traditional foods which may be a result of the penetration of global food corporations
. Improving education levels among these groups may contribute to improving health behaviours within the changing food environment experienced in rapidly changing economies like China. Evidence from Europe based on macro-level data show that national expenditure on education is inversely correlated with population levels of obesity
. Investments in education may be useful where legislation on commercial activity may be politically unfeasible, however, this should not be a substitute for the strengthening of public health systems and economic governance.
Strengths and limitations
The Four Provinces data consisted of nationally representative data. It randomly recruited older women from four provinces in China, included anthropometric and health behaviour data, and had a high response rate. However, the cross-sectional nature of the data limits the interpretation of the findings in terms of temporal and causal inferences. The four provinces were broadly comparable and the overall sample had levels of central obesity comparable to national levels reported elsewhere. The prevalence of obesity was 66.0% (see Table
2) which was almost exactly the same as the prevalence reported for the age, sex and year equivalent group in the Chinese Health and Nutrition Survey which was 66.5%
. However, although the provinces had comparable levels of economic development and modernisation to other provinces, caution should be exercised in generalising our findings to all of China’s 169 million older inhabitants. The age range was confined to women over 65 years and limits the generalisation of the findings to the rest of the population.
There are many mechanisms that could explain the role of education which could not be explored due to the data limitations. For instance, the effect of body shape preference and early life deprivation could not be assessed, and inferences regarding the impact of health behaviours were limited due to the dichotomous specification of the variables. In terms of occupation, the non-agricultural category was heterogeneous and included both manual and office-based workers and could, therefore, be further segmented to examine dose–response or gradient effects as well as non-linear associations. However, this definition was informed by a substantial body of theory and empirical work documenting the link between shifts away from agriculture (towards an industrial economy) with a rise in obesity and attribute the rise to changes in diet and physical activity levels
[7, 60, 61].
We excluded housewives from the sample as there was no theoretical basis underpinning the relationship between housewife status and obesity levels. This may have introduced bias in the analysis, however, the study did not aim to examine the association between occupational status and obesity but of the association between two specific occupational categories (agricultural and non-agricultural) in a context where there have been major shifts away from agricultural work and a parallel rise in obesity”.
Finally, the occupation variable may have been subject to reporting bias as those who classified themselves as having an agricultural occupation may have, in fact, been employed in other sectors as seasonal migrant workers. These issues require further exploration in the epidemiological literature examining socioeconomic status and non-communicable disease outcomes in lower income settings through improved data collection and measurement accuracy. Further investigation of the hypothesis and the mechanisms behind the observed associations would benefit from the use of longitudinal data.