Intended for healthcare professionals


Kaplan and colleagues' reply

BMJ 1996; 313 doi: (Published 09 November 1996) Cite this as: BMJ 1996;313:1207
  1. George A Kaplan,
  2. John W Lynch,
  3. Richard D Cohen,
  4. Jennifer L Balfour,
  5. Elsie R Pamuk
  1. Chief Associate research scientist Senior research scientist Graduate assistant Human Population Laboratory, California Department of Health Services, 2151 Berkeley Way, Annex 2, Berkeley, CA 94704, USA
  2. Health statistician National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD 20782, USA

    EDITOR,—Ken Judge claims that our study about the relations between income inequality and mortality in the United States is methodologically flawed and cannot contribute much to the debate on income inequality and health. We agree that an understanding of the impact of taxes, benefits, and household composition on this relation is important. However, Judge's criticisms are factually in error (our income measure did include government cash benefits) and illogical. He is incorrect in stating that an overestimate of the extent of income inequality necessarily calls into question the association between inequality and mortality. This would be true only if adjustment for taxes, benefits, and household composition reordered the states with respect to the extent of income inequality.

    We examined these issues in two ways, by accounting for taxes, benefits, and household size at both the aggregate and individual levels. In our paper we reported an association (adjusted for income) between the share held by the less well off 50% of the population and mortality of r = −0.57. This association was r = −0.49 after additional adjustment for average household size (a proxy for household composition), maximum welfare and food stamp payments, and the maximum difference between state income tax brackets (a proxy for the redistributive effect of state income tax). Thus, subject to the limitations of ecological analysis, these adjustment factors make little difference.

    To capture the effects of taxes, benefits, and household composition on the actual distribution of disposable income we used data from the Luxembourg income study, provided by Professor Tim Smeeding. The association between inequality and mortality in the United States was examined by use of disposable income distributions, adjusted at the household level for cash benefits and food stamp benefits and for income taxes and payroll taxes and adjusted with a household size equivalence scale. The data were based on the United States' current population survey and included the share of household income received by the most and least well off 20% of the population. We calculated directly comparable percentage shares from our data on gross income obtained from the 1990 United States census (table 1). The association between income inequality and mortality was virtually identical, although the absolute magnitude of income inequality in the states was reduced in the data from the Luxembourg study. The relations between income inequality and mortality were not affected by taking account of taxes, benefits, and household size.

    Table 1

    Association (adjusted for median income) between share of total household income received by least and most well off 20% of population and mortality adjusted for age, based on data on disposable income from Luxembourg income study and US census data, for 50 states in United States in 1990

    View this table:

    Judge refers to the “contentious” issue of the relation between income distribution and mortality, but we are aware of only one criticism of the data supporting this relation, and that paper was by Judge.1 Real progress on the relation between income distribution and health will be made by research at a variety of geopolitical levels and by thoughtful examination of the pathways by which income distribution can influence the health of individuals and societies.


    1. 1.
    View Abstract