The world of development economics (and its constituency of development economists) may seem strange to outsiders. Stranger still is the place our work, as economists living and working in the developing world, meets the institutions, policies, and products that may someday trace their intellectual ancestry to our research.
In part because the economics of poverty are hard to understand and in part because people are lazy in communicating their ideas, per diem per capita income statistics are thrown around as a way to describe quality of life in the developing world. Not only are these "dollars per day" statistics dubious as a metric, they are misleading as an average.
Most of the farmers I know in northern Uganda (which is, incidentally, most of the farmers I know) experience three problematic phenomena that are difficult to mitigate, but easy to understand. All have to do with uncertainty.
The first problem is the simplest: A farmer finds it hard at the time of planting to estimate his income at the time of harvest. Indeed, many farmers' estimates will not be in the right order of magnitude, due to crop failures, disease, inflation, or changes in the prices offered for the crops in question. To give a sense of the variance here, the price of beans in north-central Uganda has varied in the past three years from about five-hundred shillings per kilo up to slightly over thirteen-hundred shillings per kilo (this variance is, in percentage terms, seven times the wholesale price variance a farmer of beans in France experienced during the same period).
The second problem is one of probability: It is difficult for humans to estimate the probability that a high-magnitude, low-frequency event will occur. People vastly overestimate or underestimate the likelihood that unlikely, disruptive events will occur. People made very poor estimates post-9/11 of the likelihood of additional terrorist attacks and made even poorer estimates post-Columbine of the likelihood of additional deadly school shootings. The current fashion is to disproportionately worry about the likelihood of additional nuclear meltdowns in 2011. Because something like an equatorial hailstorm is rare but disastrous, planning for it in the absence of savings or insurance is challenging.
The third problem is one of smoothness: A farmer's income is notoriously sporadic. In the absence of credit, it is incredibly difficult to smooth the income profile of the farmer, meaning life quickly becomes a series of income spikes punctuated by costly disasters. I've met with African farmers living on very small incomes, many of whom I consider friends, and their reply when asked what happens when you have a bad harvest is very simple: You don't eat. It is difficult for WorldOne westerners to understand what DevWorld people mean when they say "You don't eat," but it means just that.
Describing a farmer's life by squishing his estimated income from two harvests (in much of equatorial Africa, the majority of crops experience biannual harvests) into an aggregate number and then stretching the result back out over a denominator of 365 is like trying to describe the geography of London in relation to the bowling crease at Marylebone Cricket Club. It might be logically defensible as the centre of something, but also the next best thing to meaningless.
At this point, many will raise the concern of international donors. These are people the DevWorld-focused NGO community often presumes unable to visualise the cost of anything except in relation to the price of their per capita per diem Starbucks intake. Surely, average income per day must be the best way to describe the complex concept of poverty to these people, no? No. This is partly because the income profiles of those in poverty are very complex, as illustrated most effectively and recently by Collins, but as early as Popkin. It is also partly because it is a too-easily manipulated number. Should foreign aid be included, divvied-up, and averaged in? In compiling these statistics, should formal employment be treated differently (as noted by Wade and others, these formal concepts such as "non-farm payrolls" have little meaning in other societies)?
If people are offered the mean incomes of the poor, people should at least - as a matter of statistical honesty - also insist upon seeing the variance and getting a sense of what one standard deviation above (or below) the mean income really looks like in a place like Malawi, Rwanda, or Uganda. The location of the curve on the x-axis matters just as much as (not more than) the curve's fatness and skew. Reporting a more holistic set of income data including variance would do two things. It would give a more legitimate sense of the income profile of the poor by helping to isolate dependable income from sources like foreign aid initiatives (which are income streams that suffer from the three issues, supra, to a far lesser extent) from locally-derived, often-less-dependable income from sources like agriculture. Second, and more importantly, it would force better year-to-year consistency (honesty) from NGOs and IGOs as to how income statistics for the poor are created.
We're a long way from being able to describe to the average person near Geneva what it's like to live near Gwanda (Zimbabwe). Providing average income per day with no other point of reference may provoke pity, but does little to promote understanding.