The ADB Says Poverty Is Rising in Asia: I Have My Doubts

Would you believe that half of Asia lives in poverty and the absolute number is rising? That’s what the new Key Indicators report by the Asian Development Bank (ADB) would have you believe—49.5 percent in 2010, to be precise, with a poverty count of 1.8 billion, up from 1.6 billion in 2005.

This is surprising. Indeed, it runs counter to every other assessment I have seen.  Absolute approaches to measuring poverty over a very wide range of poverty lines (including $1.25 a day) have shown a decline in Asia’s poverty measures. And (weakly) relative poverty measures also show declining poverty rates, as is shown here. It takes a marked upward revision in the real value of the poverty line over time to deliver rising poverty in Asia. This calls for a closer inspection.

At the heart of the ADB’s new poverty counts is a set of three “adjustments” they make to the international line of $1.25 a day in 2005 prices. Taken separately, none of the three is enough to give rising poverty in Asia, but together they manage it. However, when I look more closely at the three adjustments, I have concerns about two of them.

The first adjustment is to switch to a regional line of $1.51 a day. To get this, the ADB has applied essentially the same method used to set the global $1.25 a day line, but solely on data for Asia. (The $1.25 line is the mean of the national lines found in the poorest 20 or so countries in the world; the ADB did roughly the same for the poorest countries in Asia.)  The $1.25 line was never meant to be the only line, and higher lines are defensible. Switching to $1.51 adds 10 percentage points to the poverty rate for Asia in 2010.

The second adjustment is to use the rate of food price inflation for adjusting over time whenever this is greater than the overall rate of inflation. This is odd. The authors acknowledge that people living around the median do not just consume food. In Asia, close to half of their expenditures go to nonfood goods. It can be granted that the official Consumer Price Indices for some countries put too low a weight on food for the purposes of poverty measurement. But the standard corrective is to reweight the index appropriately, not to simply switch to the food CPI when it rises more than the ordinary CPI.  This second adjustment adds 4 percentage points to the poverty rate for Asia.   

The third adjustment is for “vulnerability.” The aim here is to find the increment to the poverty line in a risky world that gives the same level of welfare as a risk-free line. This third adjustment adds 12 percentage points to the 2010 poverty rate.

There are a number of issues here. While this type of “welfare-consistency” is conceptually appealing for setting poverty lines, the report is not internally consistent. It argues for welfare consistency in dealing with risk, but implicitly rejects that principle for relative poverty lines, at different levels of development. Welfare consistency almost certainly requires setting a higher line in richer countries to compensate for the welfare loss from relative deprivation and the higher costs of social inclusion (as argued here).

It’s also not clear why the base poverty line ($1.25 or $1.51) should be treated as risk free; the underlying national lines were clearly not developed in a riskless world. As the report acknowledges, national poverty lines reflect the prevailing view of what it means to be “poor” in a given country. Such assessments are unlikely to ignore country-specific risk.  

I have other concerns about the report’s assumptions, and the problems will not be evident to many readers. Using some math, and numerous assumptions buried in quite technical boxes, the report derives a formula for the risk-adjusted poverty line corresponding to any given risk-free absolute line. Let me try to explain what all those equations in the report’s boxes mean.

“Risk” is assumed to take a rather special form, namely that it rescales personal income levels (it is “multiplicative”) and in a way that it is uncorrelated with the long-run income levels. The latter appear to be measured by overall regional averages. A single parameter reflecting relative risk aversion is assumed.

Are the assumptions plausible? Some observers will think that the ADB’s assumed parameter for the extent of risk aversion is on the high side. But for me the more worrying feature is that any rise over time in the overall interpersonal variance of incomes is taken to reflect an increase in their riskiness. Rising variance is passed onto their risk-adjusted lines. 

That is hard to accept. Plainly the rising income dispersion that we see in Asia and elsewhere stems in no small measure from rising inequality of wealth. Rising risk is no doubt a contributing factor, but it is not the whole story. To put it another way, the risk-free long-run interpersonal variance of incomes in Asia is almost certainly rising. Thus the ADB’s upward revisions to poverty lines overcompensate for rising vulnerability.

The ADB’s researchers have made a valiant, even heroic, effort to deal with what they see as the shortcomings of standard absolute poverty lines. In the end, however, I am not convinced by their calculations, or their implication that poverty is rising in Asia.