Has inequality increased over the last several generations? The answer depends upon the “currency” for inequality assessment. An item has been distributed among the population of interest, and we are using a number to summarize that distribution. But which item is it?
Consider the United States. Income inequality among the US population has gone up since the 1970s. If we apply a standard inequality metric (such as the Gini coefficient or a variance-based metric) to the population-wide distribution of annual income, the time trend since the 1970s shows a clear increase in inequality. On the other hand, it appears that inequality of longevity in the US has decreased over the same period. This can be calculated by taking a mortality table for a given year, showing the number of deaths during that year of individuals in their first year of life, second year of life, etc.; applying an inequality metric to this distribution of age-at-death; and then looking at the time trend in the inequality scores.
Inequality of happiness in the US also appears to have decreased. A large body of work now seeks to quantify happiness by asking individuals survey questions such as: “How happy are you on a scale of 1 to 5?” Betsey Stevenson and Justin Wolfers looked at the variance in the distribution of these happiness scores in the US, for each year starting in the 1970s, and found that the variance of happiness has gone down.
The lesson of these examples is that the empirical project of measuring inequality depends upon a prior normative determination, regarding the appropriate “currency” for inequality assessment. The same is true for the project of assessing a society’s overall condition. GDP is the most widely used indicator of how countries are faring. The GDP calculation is based on the money value of marketed goods and services produced in the country during a given year. GDP per capita is thus a measure of the average flow of market value to the country’s population. But we could, instead, quantify overall social condition by calculating the average happiness, health, life expectancy, or educational level of individuals in the population, or the average level of environmental quality (pollutant load) to which they are subjected.
Analogous points hold true, once more, with respect to the measurement of poverty. Doing so means identifying some threshold (the poverty level); and determining the fraction of the population with holdings below this level and the below-threshold distribution. But first we must ask: Poverty of what? The traditional approach (as with inequality measurement and GDP) is to focus on material well-being —specifically, the percentage of the population whose incomes are below an income-poverty level, and the distribution of income among this income-deprived group. However, the burgeoning literature on so-called “multidimensional” poverty identifies a plurality of goods; the degree of poverty in a population is measured as a function of dimension-specific cutoffs, and the estimated distribution among poor individuals (those below the cutoff on at least one dimension) of multidimensional bundles of the referenced goods.
In short, the assessment of inequality, poverty, and overall social condition requires a prior determination regarding the “currency” for such assessment. But how exactly to move beyond the traditional approaches?
One, straightforward, possibility is the so-called “dashboard.” This means specifying a plurality of goods (income, health, longevity, education, environmental quality, etc.), and then applying an inequality, poverty, or overall-condition metric to the population distribution of each good, taken separately. Yet the dashboard ignores the correlation among goods. For example, a given distribution of income is less fair if those with high incomes are also highly educated, healthy, long-lived, etc.
A more attractive possibility is to measure each individual’s well-being, as a function of their multidimensional bundle (their holdings of each good); and then assess the population distribution of individuals’ well-being numbers. Various specific well-being measures are possible, corresponding to different normative positions regarding the nature of well-being: (1) the individual’s level of happiness, as produced by her holdings of the goods (the hedonic view of well-being); (2) a “utility” number, that takes account both of the individual’s holdings and of her preferences (the preference-based view of well-being); and (3) an “objective” number, determined just by the individual’s holdings, as opposed to her happiness or preference-satisfaction (the objective-good view of well-being).
Yet a third possibility is a correlation-sensitive multidimensional metric: one that takes account of the correlation among goods, but does not attempt to measure well-being at the individual level. This, indeed, is the structure of the main methodologies used in the literature on multidimensional poverty mentioned above.
Researchers Koen Decancq and Dirk Neumann examine a large data set, the German Socio-Economic Panel (SOEP), a large representative sample of German citizens that records information about the individuals’ income, health condition, employment status, stated “life satisfaction” (a marker of happiness), and various demographic attributes. For each individual, they calculate five different indicators: (a) income; (b) a composite objective measure of individual well-being, combining information about the three goods income, health, and employment status; (c) two preference-based measures of well-being, combining information about those goods and individual preferences for combinations thereof (these preferences themselves estimated in a subtle way from the data); and (d) life satisfaction. Decancq and Neumann then compare the worst off individuals (the lowest decile according to each of the five indicators).
They find large divergences. For example, the worst off individuals according to the income measure are, by definition, those in the lowest decile of the income distribution; by contrast, the average income of the worst off according to the life-satisfaction measure is barely lower than the average income of the entire population; and the worst-off according to the preference measures tend to be much less healthy than the income- or satisfaction-poor (reflecting a strong and unfulfilled preference for health among this group). Decancq and Neumann also quantify the overlap of the five measures—finding that very few individuals are worst off according to all five measures, and surprisingly few even according to two or three.
Their conclusion: “measurement matters.” We need not only to debate the causes and remedies for inequality and poverty but also (and indeed first) to ask: inequality and poverty of what?
Featured Image Credit: homelessness charity poor difference by ptrabattonui. Public domain via Pixabay.