Summary of findings

The findings of the present study suggest that the economic trends during the past 35 years have been accompanied by a widening of health inequalities between occupational classes as well as geographic areas of the country. The post-bubble economy has been characterised by lacklustre growth combined with a dramatic shift in the workforce away from life-long employment towards more precarious employment.4 This economic restructuring has increased pressure on workers in managerial and professional workers (primarily men) who are being squeezed to raise their productivity. The changing pattern of health inequalities across occupational groups is consistent with this interpretation, that is, the stalled decline in premature mortality among white-collar workers relative to other occupational classes.

Comparison with other studies

The present findings suggest that the health effects of the changing economic conditions depend on individual's socioeconomic circumstances. A previous study in Japan demonstrated that, although self-rated health improved for both sexes throughout the economic crisis of the 1990s, health disparities in relation to occupations widened, especially among men.25 They also reported that middle-class male workers and female home makers seemed to be particularly adversely affected by the crisis.25 The present study, however, provides a different pattern of widening health disparities in both sexes. For men, absolute health status improvement was observed only among some lower occupational groups (eg, production process and related workers, sales workers and clerical workers), whereas higher occupational classes (eg, specialist and technical workers and administrative and managerial workers) apparently obtained no benefit throughout the period. Indeed, although they were advantaged with regard to mortality risk in 1970s and 1980s, they were overtaken in the 1990s by those in lower occupational classes who benefited more during the same period. Of note, this ‘cross-over’ almost coincided with the collapse of the economic bubble in the early 1990s. We note at the same time that neither male service workers nor agricultural, forestry and fishery workers experienced improvements in premature mortality throughout the period.

By contrast, for women, we observe that absolute health status improved roughly to the same extent across occupational groups and that changes in ranking were less pronounced in women compared with men. We should note that relatively few women were represented in the three occupational groups with higher risk of mortality (ie, administrative and managerial workers, security workers and transport and communication workers). Even excluding these occupational groups, however, health inequalities appeared to have increased in women. These findings may be explained by differences between men and women according to the type of work and industrial sector of employment. Men are more likely to be engaged in work in the private sector as well as in parts of the economy that are more vulnerable to economic downturns (such as finance and business services, manufacturing, construction).26

Potential mechanisms of social inequalities in mortality

The present findings provide a marked contrast to the evolution of health inequalities described in other industrialised countries. In industrialised western European and North American countries, health status typically follows a hierarchical pattern: that is, the lower the socioeconomic position, the worse the health status.5 8 10 11 We show that this ‘typical’ pattern of health inequalities does not necessarily apply to Japan. In contrast to Western countries, previous studies in Japan have yielded inconsistent results with regard to the relationship between socioeconomic status and health outcomes, and lower non-manual or manual workers do not necessarily exhibit less healthy behaviours compared with those in higher occupational classes.27–32 Nevertheless, a recent study of a nationally representative sample in 2001 showed that men in lower occupational classes, such as service work, transportation and labour work, were significantly more likely to engage in health risk behaviours compared with professional workers.33 They also showed that there is a cumulation of risky behaviours in lower female occupational classes.33 Further, another cross-sectional study in Japan demonstrated that occupation was not significantly associated with psychological distress among men or women by using a nationally representative sample in 2007.34 Thus, the pattern of health inequalities in the present analysis is not consistent with occupational class differences in health behaviours or psychosocial stress.

As a possible explanation for the present findings, we note that occupation-based socioeconomic position may reflect social networks,14 which enables its members to access a wide variety of resources. In this respect, recent research from Japan has emphasised the evaluation of social capital as well as social networks in the workplace to explain variations in workers' health.35–37 We thus hypothesised a posteriori that, following the collapse of the economic bubble, workers of higher occupational classes were more likely to experience a breakdown of social cohesion within companies, which could cancel out the potential positive benefits among them. We also note that there is a possibility that the ‘compositions’ of each occupational group went through a (substantial) change throughout the study period, which might have led to different patterns of occupational hazards, especially among lower occupational groups. In other words, there is a possibility that work environment have improved markedly among them throughout the study period, which now requires less labour load. Finally, a possibility of healthy worker effect cannot be ruled out among some lower occupational groups. This could be induced by the following two processes: (1) healthy people might have selectively entered these occupations and (2) unhealthy workers might have selectively exited these occupations. Further studies are warranted to examine these possible explanations of the present findings.38

It is worth mentioning that typical occupational hierarchy does not necessarily apply to the occupation (major group) of the Japan Standard Occupational Classification. Indeed, there is inherently more ambiguity in the ranking of occupations compared with education and income.39 In addition, as noted by Galobardes et al,14 the decrease in manual occupations with concomitant increase in low-level service occupations has altered the stratification that occupation generates in terms of socioeconomic position and so classification such as manual and non-manual workers may lose some of their meaning in economies which include a large number of low-paid non-manual service jobs. Importantly, the occupational classification in the present study yields reasonably consistent occupational grouping throughout the study period, and each group has a reasonably large data. We therefore examined the time trend of social inequalities by using the finest occupational classification available in the census. By using a fairly detailed occupational classification, it is likely that we could adjust for other omitted compositional variables (eg, education), to the extent that the cross-tabulation of age and occupation correlate with them.

Geographic and temporal variation in mortality

By applying the novel multilevel methods, the present study shows that geographic inequalities in premature mortality have also widened since 1995. In an ecological study, Fukuda et al40 assessed the time trend of geographic health inequality in Japan, by examining the association of life expectancy and age-adjusted mortality with per capita income of prefectures and municipalities. While excluding Okinawa prefecture from the analyses, they found a possible increase in geographic health inequalities from 1995 to 2000, following a decrease from 1955 to 1995.40 Note that the present study examined geographic inequalities, conditional on individual age and occupation. The present findings thus provide suggestive evidence of ‘common ecologic effects’ of place where people live,41 although we should note that the seemingly ecologic effects might be due to an omitted compositional effect (eg, income). Broadly speaking, since 1995, higher mortality risk has been consistently observed in the northeastern region in the main island (Tohoku region) for both sexes. Overall, the economic conditions of the predominantly rural areas in the region may be characterised by population decline, population ageing and lower per capita income.19 42 Notably, however, not all rural prefectures have undergone the same transition; indeed some rural prefectures (such as Nara and Okayama) had moved up through the ranks as having significantly lower mortality for both sexes in 2005. In the supplementary analysis, no clear associations were found with prefecture-level socioeconomic variables, and it remains unknown what contributed to these distinct patterns. These patterns deserve further attention in future studies.

Limitations of the study

There are some limitations of our analysis. First, although we were able to conduct a fairly detailed analysis of trends by using occupations to measure certain aspects of socioeconomic position, neither the status in employment nor the predominant type of employment contract was available, and in particular, we lacked information on whether the individuals were in standard jobs or precarious jobs. Given the conspicuous increase in the proportion of the labour force engaged in non-standard work,4 as well as mounting evidence that precarious work is associated with worse health,43 future work needs to examine whether the changing character of the workforce in Japan is contributing to widening health inequalities. The use of more detailed indicators of socioeconomic position would provide further insight into the social inequalities of health. Indeed, greater attention to the theoretical as well as empirical aspects of measurement of socioeconomic position will likely enhance the rigour of research on occupational health inequalities, which would increase the possibility for meaningfully comparing results across studies.44

Second, occupation at the time of death was used in our numerator data, which may not necessarily reflect the individual's life-course socioeconomic position.44 45 If unhealthy workers selectively exited some occupations, this would have led to an underestimation of mortality in those sectors. The proportion of agricultural workers significantly decreased during the study period for both sexes, as well as that of administrative and managerial workers (for men). However, this may reflect real trends in the workforce.

Third, considering the possible discrepancies of the respondents on the two occasions (ie, the notification of deaths and the census), we should note the potential for numerator denominator bias between the two sources of information. In particular, the possibility of measurement error in occupation at the time of death cannot be ruled out—the person recording the notification of deaths may either promote the deceased to a higher status job or demote them because the respondents did not know the details of the deceased's job. Indeed, rapid changes in the occupational structure of Japan could give plausibility to the extremely large ORs resulting from the potential for numerator denominator bias.

Fourth, the smallest geographic unit available was the prefecture (of which there are 47), and we could not explore geographic inequalities in finer detail. However, the prefecture may be a useful and valid unit of analysis since it is the unit that has direct administrative authority in the economic, education and health sectors.1 Furthermore, the prefecture has specific jurisdiction over health centres, which is the locus of preventive healthcare activity in Japan.1 Note also that the boundaries between prefectures have not changed since the Meiji Restoration (1867), enabling long-term analysis.1 Since previous studies demonstrated that the choice of geographic units as well as area-based measures is critical in the investigation of geographic inequalities,46 47 these issues warrant further examination.

Conclusions

Despite several limitations associated with the use of secondary data, the present findings indicate that both social and geographic inequalities in premature adult mortality have increased during Japan's ‘Lost Two Decades’ following the collapse of the asset bubble. As a nation, Japan must grapple with the triple demographic trends of declining fertility, population ageing and overall population decline. These trends threaten to further dampen economic activity, escalating the load on the social security system. In addition, Japan now faces multiple challenges in the wake of the earthquake and tsunami on 11 March 2011, and this may further place downward momentum on the nation's struggling economy. Given these momentous challenges, it is imperative to continue to monitor future trends in health inequalities in order to avert the potential impacts on Japan's health security.