Assessing Socioeconomic Inequalities in Alcohol Consumption and Other Key Variables

The concentration index (CI) was used in this analysis to assess socioeconomic inequalities in alcohol consumption (current and binge drinkers) in SA. The CI is a well-known and widely used index to assess socioeconomic inequality in health outcomes and indicators (22). The CI is derived from the concentration curve. Its values can vary from −1.0 (where all current drinking or binge drinking is concentrated in the poorest households) to +1.0 (where all current drinking or binge drinking is concentrated in the richest households).

The concentration indexes for alcohol consumption (current drinkers and binge drinkers) that measure the extent to which alcohol consumption is concentrated among the wealthier (pro-rich) or poorer (pro-poor) group, were calculated using the “convenient regression” approach to control for other variables (for example gender, race, age, and urban) in addition to SES (25). The “convenient regression” was performed in Stata 12 (24) to compute the concentration index (CI) using the following equation:

where σ² is the variance of the fractional rank (r) of household per capita consumption expenditure (SES), z is the vector of control variables, and the Ordinary Least Squares estimate, β, is the CI. Analyses conducted with adjustment for a binary outcome variable did not yield results that were materially different.

In addition to the CI for alcohol consumption, the Distributive Analysis Stata Package (DASP) (26) was used to assess the concentration indexes of various equity stratifying variables (sex, age groups, rural, and urban) among current and binge drinkers. This was used to assess, for example, whether female binge drinkers are more prevalent among poorer or wealthier groups. The DASP was also run using Stata 12 to obtain the CI as follows:

where ξ^=∑i=1n[(Vi)2−(Vi+1)2(V1)2]hi;Vi=∑j=inwj (the summation of sampling weights, w_j) is such that the vector of total household consumption expenditure (SES), x, is arranged from the richest (x₁) to the poorest individual or household (x_n). μ^ represents the weighted average of the variable of interest such as the different population groups (sex, age, rural, and urban, etc.). h_i represents the value of the variable of interest for individual i.

The difference in the concentration indexes between two periods was computed using the DASP menu in Stata (26), accounting for the full sampling design. This difference can result in a pro-poor “shift” or a pro-rich “shift” (see Table 2 for details). Briefly, a pro-poor “shift” occurs when the change (that is the difference) in the CI, between two time periods, is negative; while a pro-rich “shift” occurs if this change is positive. Table 2 summarizes the broad scenarios that can cause pro-poor and pro-rich “shifts” (27).

Explaining a pro-poor and a pro-rich shift in the concentration index between two time periods.

Adapted from Ataguba (27).

(1) Pro-rich and pro-poor, as used here, do not relate to any specific income thresholds of income or consumption. They are used to describe the direction of shifts in distributions either toward wealthier households (pro-rich) or poorer households (pro-poor).

(2) For a previously pro-poor distribution, the original concentration index (CI_t−1) is negative.

(3) For a previously pro-rich distribution, the original concentration index (CI_t−1) is positive.

The prevalence of alcohol consumption and CI estimates for all the surveys years are also reported. Only the 2008 and 2014/15 data are used to analyze the changes in socioeconomic inequalities. In fact, the results for shorter time periods (for example between 2008 and 2010) were not different from those presented in this paper.