The struggle against poverty is a worthy social goal, and the absence of poverty is a human right. But poverty is also an obstacle to other social goals, particularly the full realization of other human rights. A necessary instrument in poverty reduction is data: how many people suffer from poverty? Without an answer to that question it’s very difficult to assess the success of poverty reduction policies (such as development aid).
And that’s were the problems start. We regularly post poverty data on this blog, but we’re also aware that this isn’t an exact science and that there’s some uncertainty in the data. The data may not reflect accurately the real number of people living in poverty. There are definition issues – what is poverty? – that may reduce the accuracy of the data or the comparability between different measurements of poverty (or between different measurements over time), and there are issues related to the measurements themselves. I’ll focus on the latter for the moment.
Poverty is often measured by way of surveys. These surveys, however, can be biased because of
- sample errors: underreporting of the very rich and the very poor (more on sample errors here), and
- reporting errors: failure of the very rich and the very poor to report accurately.
The rich are less likely than middle-income people to respond to surveys because they are less accessible (their houses for instance are less accessible). In addition, when they respond, they may tend to underreport a larger fraction of their wealth as they have more incentives to hide (for tax reasons for example).
The very poor may also be inaccessible, but for other reasons. They may be hard to interview when they don’t have a fixed address or an official identification. In poor countries, they may be hard to find because they live in remote areas with inadequate transportation access. And again, when they report, it may be difficult to estimate their “wealth” because their assets are often in kind rather than in currency.
Because we can have underreporting of the two extremes on the wealth distribution, we believe that income distribution is more egalitarian than it really is. Hence we underestimate income inequality and relative poverty.
We get this:
instead of this:
But apart from relative poverty we also underestimate absolute poverty since we’re often unable to include the very poor in the reporting for the reasons given above. Again we get something like this:
By “cutting off” the people at the poor end of the distribution, it seems like most people are middle class and society largely egalitarian.
However, absolute poverty can also be overestimated: if the poor respond, we may fail to accurately assess their “wealth” given that much of it is in kind. And it’s unlikely that these two errors – underestimation and overestimation – cancel each other out.
These and other problems of poverty measurement make it difficult to claim that we “know” more or less precisely how many poor people there are, but if we make the same errors consistently we may be able to guess, not the levels of poverty, but at least the trends: is poverty going up or down?