Income Inequality and Empirical Economics
I found this podcast less interesting for the findings presented (recent increases in income inequality are overstated) than for the issues addressed.

In particular, attempting to measure purchasing power by looking at before-tax reported income and dividing by the CPI is far from perfect. They also discuss the fact that looking at all demographic subsamples of households could have rising incomes while overall household incomes go down if the poorer subsamples grow in size. Of course, the average incomes of the subsamples aren't necessarily more important numbers than the average income of the total population, depending on precisely what question you're getting at, but it's tempting and wrong to infer from the aggregate number that some of the subsamples must have the same sign. An issue that wasn't addressed in this podcast that I've seen raised elsewhere is that rich people and poor people buy different things; measuring real income for a subset of the population by using a basket of goods for the whole population (or a different population) has problems beyond the ones raised in the podcast.

More generally, empirical proxies usually correspond somewhat with the theoretical concepts they are meant to represent, but they almost never do so perfectly. In this case CPI is specifically designed to measure inflation, and it does, but imperfectly. It irks me when a paper says, "we're going to use firm size as a proxy for liquidity" - they aren't usually quite that stupid, but I'm making up an example - and then refers to "liquidity" through the experimental portion of the paper when the author means "market cap".

If you want an introduction to the sort of thinking you have to do to do empirical social science research, that podcast would be a good place to start.
- posted by Steven @ 3:04 PM