Introduction

Top Abstract Introduction Methods Results Discussion Conclusion References

Does relative income become more important than absolute income in determining population health after a country has changed from a developing to a developed economy? By comparing countries that are members of the Organisation for Economic Cooperation and Development, Wilkinson showed that: (a) a strong relation exists between life expectancy and income distribution, whereas its relation with gross national product per capita is weak; and (b) a decrease in the prevalence of relative poverty is significantly related to a rapid improvement in life expectancy. ^{1 2} Kaplan et al³ and Kennedy et al⁴ reported independently that in the United States the relation between income distribution and mortality remained highly significant even after controlling for absolute income. Nevertheless, most of the studies on population health and relative income versus absolute income are cross sectional studies that were done in post-industrialised countries where the epidemiological transition is complete. I aimed to examine the changing relation between income inequality and mortality through different stages of economic development in Taiwan, using data at county and city level.

Taiwan, with a population of 21.3 million in 1995, is a newly industrialised country; its achievement in economic development has been frequently termed an "economic miracle" (table 1). In the early 1950s, Taiwan was a poor country with a gross national product per capita of no more than US$200 (£72 valued at 1950s prices).⁵ Since 1953 a series of economic development plans has been effectively implemented, and Taiwan's economy has shifted from agricultural to industrial and from import oriented to export oriented.⁶ As a result, gross national product per capita has increased rapidly to US$1132 (£627 valued at 1976 prices) in 1976 and to US$12 396 (£7853 valued at 1995 prices) in 1995.⁵

The economic development of Taiwan has been identified not only with rapid growth but also with improved income distribution.⁷ Like most developing countries Taiwan had a large income gap in the early years of economic transition. The ratio of income share of the richest 20% to that of the poorest 20% in Taiwan reached 20.5 in 1953, but it decreased substantially to 4.2 in 1976 and then slightly increased to 5.3 in 1995.⁵ Similarly, the Gini coefficienta commonly used measure of the degree of income inequality, ranging from zero to a maximum of onefor Taiwan has decreased from 0.56 in 1953 to 0.28 in 1976 and gradually increased to 0.32 in 1995.⁵

Besides being an "economic miracle" Taiwan has also achieved a "health miracle" (table 1). Since the 1950s, mortality has declined remarkably among all age groups in Taiwan. Improvement is especially significant in age specific mortality in children under age 5, which has decreased from 21.1 per 1000 population in 1953 to 1.9 per 1000 in 1995. Accordingly, life expectancy at birth in Taiwan has increased for males and females respectively from 58.2 years and 61.4 years in 1953 to 71.9 years and 77.8 years in 1995.⁸

With the decline in death rates Taiwan has experienced an epidemiological transition. In the early 1950s most of the leading causes of death in Taiwan were infectious diseases including gastroenteritis, pneumonia, tuberculosis, nephritis, bronchitis, and malaria. These began to give way to non-infectious diseases as the leading cause of death, and by the 1990s non-infectious diseases such as cancer, stroke, heart disease, hypertensive diseases, and diabetes mellitus have become dominant health problems in Taiwan.⁸

	Methods

Top Abstract Introduction Methods Results Discussion Conclusion References

Source of data
I collected data for three index years: 1976, 1985, and 1995. Data on income were obtained from the family income and expenditure survey, conducted by the directorate-general of budget, accounting, and statistics, Republic of China. The family income and expenditure survey, which aimed to address the general conditions of livelihood and to present the status of family income and expenditure in Taiwan, was conducted mainly through personal interview. In order to check the validity of results from the interview a small number of households were asked to keep accounts. The survey covered the civilian non-institutionalised population of Taiwan, with about 15 000 registered households selected through a two stage stratified random sampling in a calendar year. The information collected included family composition, housing conditions, family income and expenditure, fixed assets, mutual saving funds, and miscellaneous items. The earliest available electronic data file on the family income and expenditure survey was for 1976.

Measure of absolute and relative income
In the analysis absolute income was defined as median household disposable income after payment of taxes and receipt of benefits. Relative income was defined as the proportion of household disposable income received by households whose disposable income was below a specified centile on the distribution of household disposable income. For example, the relative income for the 50th centile in Taiwan in 1995 was 28.2% because the less well off 50% households received 28.2% of disposable income. Using data from the family income and expenditure survey I calculated relative income for the 20th, 50th, 70th, and 90th centiles for all 21 counties and cities and for Taiwan as a whole.

Statistical analysis
To determine the association between absolute and relative income and measures of mortality I calculated Pearson correlation coefficients. Ordinary least squares multiple regression was further used to discover whether median household disposable income or the share of household disposable income received by the less well off 50% had more influence on mortality. For the regression analysis I presented squared multiple correlation coefficients as well as partial regression coefficients. The squared multiple correlation coefficient or R² was interpreted as the proportion of the variance of the measure of mortality which was "explained" by the model comprising two predictors: median household disposable income and the share of household disposable income received by the less well off 50%.¹¹ Because R² must increase as further variables are introduced into a regression, I further presented adjusted R² to take into account the chance contribution of each variable included.¹⁰ The partial regression coefficient or  was the amount by which the measure of mortality changed on the average when median household disposable income or the share of household disposable income received by the less well off 50% changed by one unit and the other remained constant.

	Results

Top Abstract Introduction Methods Results Discussion Conclusion References

Association between mortality and income measures
Table 2 shows Pearson correlation coefficients for the association between mortality and absolute income as well as income distribution. Although the association between all age mortality and median household disposable income remained highly significant (P<0.001) across the three index years (Pearson correlation coefficient 0.66 to 0.71), the association between mortality in children under 5 and median household disposable income tended to weaken from 1976 to 1995 (r=0.79; P<0.001 v r=0.50; P<0.05). The relation, however, between mortality and income distribution tended to strengthen over the study period, and the relation patterns were broadly consistent across the four measures of income distribution. In 1995 the share of household disposable income received by the less well off 20%, 50%, 70%, and 90% were all strongly correlated with all age mortality (Pearson correlation coefficient 0.57 to 0.64); the association with mortality in children under 5 was even stronger (0.64 to 0.75). To the contrary, the association between mortality and income distribution was very weak in 1976, especially for all age mortality (0.08 to 0.22). Only the association between mortality in children under 5 and the share of household disposable income received by the less well off 20% and 50% was statistically significant in 1976 (P<0.05).

Relative importance of level of income v income distribution
Results from the ordinary multiple regression analysis of mortality on median household disposable income and the share of household income received by the less well off 50% show the relative importance of absolute income versus relative income in determining mortality (table 3). Overall, the proportion of the variance of mortality accounted for by the two absolute and relative income measures in any of the six multiple regression models was significantly high (at least P<0.01). The value of R² ranged from 46% to 58% for all age mortality and from 38% to 63% for mortality in children under 5. The adjusted R² tended to have a value slightly lower than that of R² across different measures of mortality and different study years.

	Discussion

Top Abstract Introduction Methods Results Discussion Conclusion References

During the past two decades Taiwan has undergone critical stages of economic transition and become a newly industrialised country. This study shows that, contrary to the weak relation of income inequality to mortality in 1976, counties and cities with more equal income distribution in 1995 were more likely, with the adjustment of median household disposable income, to have a lower mortality (table 3). Especially, in 1995 the effect of income distribution on mortality in children aged under 5 became highly significant and the effect of median household disposable income became non-significant. Therefore, the Taiwan case examined here supports Wilkinson's proposition that relative income is more important than absolute income in determining population health in developed countries.¹²

In attempting, however, to generalise from this study it should be noted that Taiwan's success in economic development is unique. Firstly, Taiwan has achieved economic transition in a very short period. Secondly, it has succeeded in combining a rapid growth of national economy with improved income distribution, which is different from the experience of Western industrialised countries.¹³ Thirdly, county and city income inequality has been significantly associated with the level of economic development as measured by median household disposable income (Pearson correlation coefficient 0.51 to 0.66). These and other factors such as drastic political reform in Taiwan since the 1980s might have conditioned the changing relation of income inequality to mortality.

Why is greater income equality associated with better health? Psychosocial stress from relative deprivation, disrupted social cohesion, disinvestment in social capital, and under investment in human resources all have been suggested as pathways through which income inequality affects population health, ^{2 14 15} but only a small number of empirical studies to date have attempted to clarify the relation. ^{3 16-19} Taiwan, as well as other newly industrialised countries blessed with a rapid economic transition, should provide a good opportunity for further work to understand mechanisms linking income inequality to health.

	Conclusion

Top Abstract Introduction Methods Results Discussion Conclusion References

The health of the population is affected more by income distribution than by the level of income after a country has changed from a developing to a developed economy. Thus, a newly developed country such as Taiwan should pay more attention to the consequences on population health of the gap between the rich and the poor and not merely healthcare reform.²⁰

	Acknowledgments

I thank Shu-Chen Liu, Shao-I Lin, and Tsung-Hsueh Lu for their help in data processing.

	Footnotes

Funding: None.

Competing interests: None declared.

	References

Top Abstract Introduction Methods Results Discussion Conclusion References

1.	Wilkinson RG. Income distribution and life expectancy. BMJ 1992; 304: 165-168.
2.	Wilkinson RG. Unhealthy societies: the afflictions of inequality. London: Routledge, 1996.
3.	Kaplan GA, Pamuk E, Lynch JW, Cohen RD, Balfour JL. Inequality in income and mortality in the United States: analysis of mortality and potential pathways. BMJ 1996; 312: 999-1003[Abstract/Free Full Text].
4.	Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and mortality: cross sectional ecological study of the Robin Hood index in the United States. BMJ 1996; 312: 1004-1007[Abstract/Free Full Text]. [Important correction BMJ 1996;312:1194.]
5.	Republic of China Council for Economic Planning and Development. Taiwan statistical data book, 1998. Taipei: Council for Economic Planning and Development, 1998.
6.	Li KT. The evolution of policy behind Taiwan's development success. New Haven: Yale University Press, 1988.
7.	Kuo SWY, Ranis G, Fei JCH. The Taiwan success story: rapid growth with improved distribution in the Republic of China, 1952-1979. Boulder, CO: Westerview, 1981.
8.	Republic of China Department of Health. Health and vital statistics, vol I. General health statistics, 1997. Taipei: Department of Health, 1998.
9.	Republic of China, Ministry of the Interior. Taiwan-Fukien demographic fact book (published annually). Taipei: Ministry of the Interior. (Sources: 1973, 1976, 1985, 1995.)
10.	Armitage P, Berry G. Statistical methods in medical research, 3rd ed. Oxford: Blackwell Scientific, 1994.
11.	Goldstone LA. Understanding medical statistics. London: William Heinemann Medical, 1983.
12.	Wilkinson RG. Health inequalities: relative or absolute material standards? BMJ 1997; 314: 591-595[Free Full Text].
13.	Kuznets S. Economic growth and income inequality. Am Econ Rev 1955; 45: 1-28.
14.	Davey Smith G, Egger M. Commentary: understanding it allhealth, meta-theories, and mortality trends. BMJ 1996; 313: 1584-1585[Free Full Text].
15.	Kaplan GA, Lynch JW. Whither studies on socioeconomic foundations of population health. Am J Public Health 1997; 87: 1409-1411[Free Full Text].
16.	Kawachi I, Kennedy BP, Lochner K, Prothrow-Stith D. Social capital, income inequality, and mortality. Am J Public Health 1997; 87: 1491-1498[Abstract/Free Full Text].
17.	Wilkinson RG, Kawachi I, Kennedy BP. Mortality, the social environment, crime and violence. Sociol Health Ill 1998; 20: 578-597.
18.	Daly MC, Duncan GJ, Kaplan, Lynch JW. Macro-to-micro links in the relation between income inequality and mortality. Milbank Q 1998; 76: 339.
19.	Wallberg P, McKee M, Shkolnikov V, Chenet L, Leon DA. Economic change, crime, and mortality crisis in Russia: regional analysis. BMJ 1998; 317: 312-318[Abstract/Free Full Text].
20.	Chiang TL. Taiwan's 1995 health care reform. Health Policy 1997; 39: 225-239[Medline].

(Accepted 29 July 1999)