The Free Exchange column in last week’s Economist (I’m a bit behind since I listen to the podcast version) cites a working paper by David Bloom and coauthors that shows that inequality increases as countries undergo the demographic transition. First, a quibble with Free Exchange’s summary. It states that “DHS [Demographic and Health Surveys] data from 60 developing countries enable them to divide households into five income groups and to show that in every continent.” The DHS allow them to do nothing of the kind. For reasons that are still confusing to me, it does not collect any information on income at all. If you’re curious about the reasons for this, allow me to direct you to a 77-page document describing and justifying their alternative, the DHS Wealth Index, which is constructed from observations about household assets. The dataset contains that wealth index plus the underlying assets used to create it.
What Bloom et al. actually do is to estimate each household’s permanent income by assuming that the probability of owning each asset the DHS measures is a linear function of local prices for all assets and also of permanent income. (Permanent income is roughly equivalent to an individual’s lifetime income divided by the number of years he or she will be alive. I guess you can do something similar for households.) The working paper says they use logistic regressions to back out the permanent income factor from observed asset holdings. I’m assuming that this is a multinomial logit, but this early version of the paper doesn’t go into much detail. It also doesn’t appear to say where they got the asset prices they used, which seem important and somewhat annoying to collect.*
Lack of detail (and oversimplification by journalists) aside, this seems like a decent approach given the available data, since censuses also often don’t ask about income. Which makes it a great example for why I don’t care about inequality per se. Focusing on countries with small fertility declines – those early in the demographic transition – the authors write
The point estimates reported in column (6) of Table 5 suggest that the number of dependent children increased on average by 0.04 children among households in the poorest quintiles, while the number of dependent children decreases by 0.19 and 0.14 children in households of the fourth and fifth wealth quintiles respectively.
The increase of 0.04 they mention is statistically insignificant, and is roughly the size of the estimated standard error. That’s about as precise an estimated zero as you can get. In other words, for countries just starting the demographic transition, the rich have benefited from the demographic transition and the poor have not.** Many times, when people talk about inequality, they really mean poverty – the rich getting richer and the poor getting poorer, something like that. Bloom et al. find a pure rise in inequality: the poor aren’t any worse off at all! But inequality went up, so shouldn’t we oppose the demographic transition on equity grounds? As it turns out, no: as the demographic transition proceeds, the benefits accrue to everyone, even the poor.
Now, if this process involved the poor becoming initially worse off, then I’d be a bit more concerned. “Suck it up, future generations will do well if you suffer now” doesn’t rub me the right way. It doesn’t, though, it just sees them falling further behind the rich. That’s a little unpleasant, and seems unfair, but strikes me as a third- or fourth-order concern relative to the main issue of absolute levels of poverty and well-being.
EDIT: Actually, since their error term follows the normal rather than the type II extreme value distribution, they are certainly not running multinomial logits, but they probably should be doing something in that vein since my purchase of e.g. a cow depends on the prices of other stuff I could buy with the same money.
*It’s a nifty approach, but I wonder if they would have gotten very different results by just doing something simple like counting the total number of assets. That depends on how good their price data is, probably.
**This assumes that having fewer kids makes you better off, which seems a bit dubious, but that’s a standard assumption in a lot of policy circles so let’s run with it.