Geographical inequalities of mortality by income in two developed island countries: a cross-national comparison of Britain and Japan

https://doi.org/10.1016/j.socscimed.2004.11.007Get rights and content

Abstract

In this paper we examine the ecological relations between household income distribution and age-grouped mortality in Britain and Japan. Comparable datasets were prepared in terms of age intervals of mortality, household income intervals and geographical units for years around 1990. Then we conducted a series of regression analyses to associate absolute and relative income indices with age and sex-specific standardized mortality ratios (SMRs). The results are as follows: (1) In Britain mortality is lower where inequalities in income are lower, while in Japan there is no obvious relationship. It is, however, apparent that—just as in the case of the USA and Canada—Britain and Japan appear to merge and appear part of a greater pattern when considered as a series of city regions. Thus an overall global relationship between income inequality and mortality may exist. To assess such global relationship, further studies using cross-national regional datasets covering a wide rage of rich nations are desirable. (2) Income–mortality relations are consistent among different age–sex groups in Britain, but there are substantial differences in the relationships as revealed between different demographic groups in Japan. In particular, while absolute income levels are correlated negatively with mortality of working-age men in both countries, mortality of elderly people in Japan is higher where absolute income is higher. This indicates the different historical contexts to the health divides these two different geographical contexts, but further consideration of a more historically nuanced understanding of income–mortality relations is required.

Introduction

There has been considerable debate concerning the relationship between income distribution and health outcomes for many years which escalated following Rodgers’ (1979) seminal paper. The essence of recent debates is summarised by Wilkinson's (1996) relative income hypothesis which suggests that a wider income distribution within a population group leads to a worse standard of overall population health. This hypothesis has been tested mainly in the US where a variety of aggregated and disaggregated datasets are available. In that country, even if controlling for individual income effects on individual health at the micro scale, income distribution variables, such as the Gini coefficient of inequality, are often significantly associated with mortality and self-rated health measures (Kennedy, Kawachi, & Prothrow-Stith, 1996; Kawachi, 2000; Wolfson, Kaplan, Lynch, Ross, & Backlund, 1999).

Such evidence is limited in the rest of the world. In the European context, while a social gradient to health is widely acknowledged (Machkenbach & Bakker, 2002), systematic analysis to assess regional population health outcomes in relation to variations in income distributions is rare, possibly due to data limitations. Moreover, other recent studies (Lynch et al., 2001; Osler et al., 2002; Shibuya, Hashimoto, & Yano, 2002; Pearce & Davey Smith, 2003) have presented counter arguments to the relative income hypothesis. Given this, Machkenbach (2002) thus suggested that “the idea that the evidence for a correlation between income inequality and the health of the population is slowly dissipating.” However, to date, we are still not sure of how well relative and absolute levels of income can partly explain regional population health collectively under different contexts. Blakely and Woodward (2000) insist that we need to conduct a variety of “natural experiments” by using datasets from different countries to test if consistent relationships between income and mortality are observed in different contexts.

In this paper we examine ecological relations between household income distributions and age standardised mortality by sex in Britain and Japan to add to this debate. For this study, we prepared comparable datasets in terms of age intervals of mortality, household income intervals and geographical units for years around 1990. Following previous studies (Kennedy et al., 1996; Kawachi & Kennedy, 1997), we calculate a series of absolute and relative indices of income inequality to compare with age-specific SMRs for the assessment of the income–mortality relationships.

Our research design is similar to that of Ross et al. (2000) who conducted cross-sectional comparative ecological studies of the importance of relative income between the US and Canada. Despite cultural similarity and close interactions between the two countries, the association of mortality with income inequality was only seen in the US where income inequality was much higher than Canada. Since the setting in terms of the difference in income inequality is well matched to our comparison between Britain and Japan, we can say that our design is another “natural experiment” to test the regional relative income hypothesis.

Britain and Japan share similar geographic settings among rich nations in their geographic extents and urban systems. Both are dominated by world cities, London and Tokyo. Both countries have provided health services to their population universally (National Health Insurance in Japan and National Health Service in Britain). However, standards of health and its historical–geographical structure differ between the two countries. In Britain, evidence showing strong relationships between material deprivation and mortality has been accumulated over many decades (Shaw, Dorling, Gordon, & Davey Smith, 1999). This evidence provided the basis for the social gradient and material wealth theory of health inequality. Deprived areas with worse health status, observed mainly in large cities, have persisted over many years, and became especially apparent from late 1970s to early 1980s when the nation experienced a rapid widening of income differentials under Thatcherism that has yet to be reversed under previous administrations (Shaw et al., 1999).

On the other hand, until the early 1980s, Japan had experienced drastic economic developments with dramatic socio-economic changes and has simultaneously attained one of the lowest levels of income inequalities and the longest life expectancy seen worldwide. This Japanese experience was taken up as a demonstration of the relative income hypothesis (Wilkinson, 1996) and social gradient thesis (Evans & Stoddart, 1994). Cockerham et al. (2000), however, insisted that the social gradient explanation of health inequality was not adequate to explain the situation in Japan. They showed that Okinawa, a southern peripheral island part of Japan had experienced the worst income levels but had the longest life expectancy in Japan. During the rapid economic growth in 1960s and early 1970s in Japan, people living in the major metropolitan areas generally experienced better health conditions than those living in rural areas. After that period, the Japanese health gap has reduced and the major metropolitan areas no longer enjoy the best health standards (Hoshi, 2000). Considering several more facts defying the social gradient thesis, Cockerham et al. (2000) anticipated no regional correlation between income and mortality in Japan. Shibuya et al. (2002) associated absolute/relative income indices of 47 prefectures with self-rated health by using an individual dataset. They certainly found no effect of a regional relative income measure on measures of self-rated health when individual attributes including age, sex, marital status and income were controlled for. The geographical consistency between self-rated health and mortality at the prefecture level is however, only weakly associated to life expectancy in Japan at present (Kanda, Ojima, & Yanagawa, 2000).

Section snippets

Regional settings

Associations between absolute/relative income and mortality around 1990 are examined in 47 Japanese prefectures and 30 British pseudo-NUTS2 regions which are slightly different with formal NUTS2 regionalisation: the two London regions are integrated (to be comparable to Tokyo) and the NUTS2 regions are approximated by aggregations of function cities (which only pose slight problems on the borders of Cheshire). Northern Ireland is excluded because of the lack of consistent statistics. Mean

Results

Multiple regression analysis with mean income and decile ratios as explanatory variables was conducted to assess the relative importance of absolute and relative income for mortality in Britain and Japan. Fig. 3 shows standardised partial regression coefficients of mean income and decile ratio for each age–sex group. Except for Japanese females, income indices are significantly correlated with mortality, especially for working age groups between ages 20 and 64. The trends for both absolute and

Sensitivities of areal income to mortality

Despite the different spatial structures and historical backgrounds to the health gap in both countries, our analysis has shown that the regional gaps in absolute household income correspond well to the health gaps of working-age folk in both countries. Counter to the anticipation of Cockerham et al. (2000), the geographical distribution of mortality is significantly associated with regional absolute mean income levels in Japan as well as in Britain. Conversely, when the two countries are

Acknowledgements

An earlier draft paper was presented at the 10th International Symposium in Medical Geography, University of Manchester, Manchester, UK, 17 July 2003. Helpful comments from two anonymous referees are gratefully acknowledged.

References (29)

  • Migration rates by age group and migration patternsapplication of Rogers’ Migration schedule model to Japan, the Republic of Korea and Thailand

    (1990)
  • A. Kanda et al.

    Jikakuteki-kenkokan no kenko-sihyo toshiteno yukosei (Effectiveness of self rated health as a health indicator)

    Journal of Health and Welfare Statistics

    (2000)
  • I. Kawachi

    Income inequality and health

  • Cited by (43)

    • Social capital and use of assisted reproductive technology in young couples: Ecological study using application information for government subsidies in Japan

      2021, SSM - Population Health
      Citation Excerpt :

      One possible reason for the non-significant association between Gini coefficient and ART use is the threshold effect of income inequality (Kondo, 2012). Studies have reported a notable effect of income inequality in countries where such inequality is high (e.g., the United States and Britain); however, no such associations have been observed in countries with relatively low inequality (Nakaya & Dorling, 2005; Ross et al., 2000). Thus, the association between income inequality and ART use in different countries requires further investigation.

    • Associations of neighborhood socioeconomic conditions with self-rated health, mental distress, and health behaviors: A nationwide cross-sectional study in Japan

      2020, Preventive Medicine Reports
      Citation Excerpt :

      As previous neighborhood studies have limited international geographical scope (i.e., concentration in Western countries), this study may contribute enhancing the generalizability of the findings in neighborhood and health studies for it provides evidence from Japan. In addition, as mentioned in the introductory section, previous Japanese studies have often employed ecological designs, investigated limited cities/regions, or used larger spatial units for analysis (Fukuda et al., 2005, 2007; Ito et al., 2014; Nakaya and Dorling, 2005; Nakaya and Ito, 2019; Nakaya et al., 2014). Therefore, this nationwide study provides basic evidence on the independent associations between neighborhood socioeconomic conditions and health status and behaviors in Japan.

    • Social capital, income inequality and self-rated health in Chita peninsula, Japan: a multilevel analysis of older people in 25 communities

      2009, Social Science and Medicine
      Citation Excerpt :

      For example, Eckersley (2006) pointed out that the association of inequality and social trust with crime was conditional on the prevalent values of society, and that its effects became more pronounced in societies that were characterized by a high level of self-interested values. A second potential question about our results is why the relative income hypothesis was supported in this current study despite two unsupported results of prior studies conducted in Japan (Nakaya & Dorling, 2005; Shibuya et al., 2002). Because of the difference in the methods adopted in each study and the small number of studies that have examined the relative income hypothesis in Japan, we cannot say for certain but three possible reasons are presumable.

    View all citing articles on Scopus
    View full text