Several important studies have shown an apparent paradox: even as income inequality has taken off over the last thirty years, differences in consumption between the rich and poor have changed slightly or not at all. Focusing on the bottom of the distribution, Bruce Meyer and James Sullivan (2009) actually conclude that consumption poverty has declined during time periods when income poverty was rising. Analysts know that both income and consumption data are measured with lots of error. Unless this error is completely random across groups, we should be concerned that either the trend in income, or consumption, is biased over time.
An ambitious new paper by Orazio Attanasio, Erik Hurst, and Luigi Pistaferri attempts to address measurement error in the Consumer Expenditure Survey (CEX), and comes to a different conclusion. The CEX is the most comprehensive source of information on spending and saving in the United States, but it has become increasingly worse at measuring certain types of non-durable purchases. If this bias is worse at the top of the spending distribution – a real possibility since high spenders may not track certain purchases as well as those living on tighter budgets – then the unadjusted trend in the CEX may bias the inequality trend downward.
The authors deal with this in a few innovative ways (consult the paper for exact details). One is that they focus on categories of consumption that are known to be measured with less error, such as money spent on food at home versus money spent on entertainment. The second is that they examine the value of durable purchases such as car ownership (which is very well measured in the CEX). The third is that they bring in other data sources, the Panel Study of Income Dynamics, which is thought to measure certain purchases with more accuracy (although the PSID covers fewer items in the family budget). They also consult another CEX data source that is independent of the quarterly survey – the daily diary (respondents keep a diary over two weeks and enter in their purchases every day) – because it covers a smaller time-slice there is more variance in the daily diary, but it is a good supplement for examining trends.
All of these data sources lead to the same conclusion: the trend in consumption inequality mirrors the trend in income inequality much more closely than previously thought.
Here are a few figures that emphasize the trend:
This figure just shows how much more variability there is over time in people’s consumption (although it doesn’t tell us where the variance is growing in the income distribution).
Similarly, we can see that the ratio of spending on food (which is income inelastic) to entertainment (which has an elasticity greater than 1.5) has been increasing over time. This is interesting circumstantial evidence, but not conclusive. The figure below I think makes the best case.
Here we see that there is a divergence in consumption measured in the diary component at all points in the income distribution.
This paper makes an important contribution from a policy standpoint, because it mutes the argument that the “real” effect of inequality on people’s standards of living is much lower than its impact on their incomes. Changes in income are transitory, but what people actually purchase reflects something about their permanent income (since people can smooth their consumption across time by borrowing and saving).