Growth incidence analysis involves calculating the amount of growth in some welfare-indicator (e.g. income, life expectancy) at each quantile (e.g. quartile, decile, or percentile) of the distribution of that indicator. This information, and its graphical representation in the form of a growth incidence curve (GIC), gives a much fuller picture of changes in social welfare and poverty over any given period than can be provided by a single figure, such as the rate of per capita income growth, or the reduction in the poverty headcount. However, although it is now widely accepted that welfare and poverty are multi-dimensional, and cannot be reflected accurately by any one single indicator, the vast majority of GICs calculated to date have been based on growth rates in income (or expenditure).
This paper begins filling this gap by providing estimates of GICs for non-income welfare indicators in Ghana and Uganda. The main practical difficulty is that most of such indicators found in standard household surveys are discrete, as opposed to continuous, variables – i.e. their values are limited to a fixed number of categories (e.g. good, poor), rather than a very large number (e.g. units of local currency). This limits the range of values the GIC can take; in the extreme, it limits it to one of only two values: growth or no growth. The best way to address this problem is by expanding and improving the coverage of non-income welfare indicators in standard household surveys. This will require an initial ‘bench-marking’ exercise to establish best-practice and standardise as much as possible across surveys in different countries. In the meantime, an alternative approach is to calculate so-called ‘conditional’ GICs, which involves calculating the growth in non-income indicators among different groups of the population ranked by income (or expenditure).
From the household surveys for Ghana and Uganda in 1992 and 1999 it was possible to construct nine non-income welfare indicators, in the dimensions of education, health, household amenities (e.g. use of drinking water, sanitation and electricity), and household assets. The results show some significant and important differences across income and non-income welfare indicators, both in terms of aggregate trends and distributional patterns. In Ghana, despite relatively low rates of income growth at the lower end of the income distribution, there have been significant improvements and catching up in other non-income indicators, such as primary school enrolment, use of good drinking water, and the value of assets owned (in proportional if not always in absolute terms). In Uganda, despite reasonably high rates of income growth, particularly at the lower end of the income distribution, there has been a deterioration of other non-income indicators. Overall, the analysis cautions against over-reliance on purely income-based measures of poverty and welfare. It also suggests the need to consider ways in which potential trade-offs between income and non-income welfare indicators can be incorporated into policy analysis.