The measurement of PPPs and the measurement of poverty AN GU S D E ATON W OR LD B AN K J U N E 16 , 2 0 14
Two topics ICP 2011 The new PPPs, why are they so different from what was anticipated? Work jointly with Bettina Aten, Bureau of Economic Analysis Paper is on my website http:/ / www.princeton.edu/ ~deaton/ papers.html NBER Working Paper later this week Implications for measuring poverty As well as PPPs and poverty more generally
Introduction ICP 2011 summary report was published on April 29 th , 2014 Large decrease in PPPs for non-OECD countries relative to the US Makes the world much less unequal China is not larger than the US, but close India is bigger than Japan These received lots of attention in the media Also, media and blogs noted that there were major implications for global poverty If $1.25 is held constant, about half of global poverty will vanish About 2/ 3 for India, from 300 million to about 100 million
Consumption PPPs PPPs for individual consumption by households are used for poverty calculations I focus on them here Somewhat simpler to ask what happened than for GDP We don’t need to deal with the trade balance, government services, education, and other comparison resistant items Document what happened And an explanation for how we got here As well as estimates of the size of the problem And how to correct it
Need to think about 2005 and 2011 together ICP2005 raised PPPs in poor countries relative to rich countries, compared with ex ante extrapolation from 1993/ 96 ICP2011 results do the opposite, decreasing PPPs in poor countries relative to rich countries, compared with ex ante extrapolation from 2005 Discussion has mostly been about how implausible 2011 results are conditional on 2005 being correct Little criticism of ICP 2005 because ICP 1993/ 96 had so many things wrong with it No China, no India, piecemeal not integrated, and many new methods So it was easy to dismiss the 2005 changes, which might have been a mistake
The hypothesis Perhaps part of what is going on is that 2005 numbers were too high? Would account for jump up in 2005 Corresponding jump back down in 2011 The 2005 regional linking could possibly have done this ICP 2005 priced a ring list of 1000 items in 18 countries Much less cross checking than for the regional lists All of the ring prices get aggregated into four “tectonic” indexes Africa, Asia, Western Asia, LAC versus OECD countries If the ring goods were hard to find (unrepresentative) and overpriced in Africa, say All African countries would have PPPs that were too high No ring in 2011, but common global list So the ring effect would be undone
Documenting what happened Some graphs comparing actual ICP 2005 and ICP 2011 with previous extrapolations Extrapolations come from previous rounds using CPI from the country and CPI from the US With country c and US country 0 Except for Eurostat, which updates continuously
Failures of extrapolation Many reasons why extrapolation can fail PPPs are multilateral indexes Each bilateral is superlative and so has weights of both countries CPIs have only weights of one country So you can’t get one from the other ICP and CPI collect different prices Usual ICP issues of representativity versus comparability Not the same in selection of items for the CPI Standard errors are likely to be large Poor quality of CPIs in many countries Old weights, and relative price of food increased from 2005 to 2011 But this doesn’t do it, see paper
Is it the ring? Certainly not all the ring The difference with extrapolation in 2011 is not uniform by region Changes of relative positions within regions This cannot be the ring, because it doesn’t work this way A test for the ring itself Simple case, the ring collected food prices in Cameroon in 2005 Global list collected food prices in Cameroon in 2011 We have an estimate of food price inflation from successive ICPs We have food price inflation from Cameroon’s own CPI If ring story is right, ICP prediction of food inflation should be substantially lower than CPI change itself Because of the overstatement in ICP ring in 2005 Ideally repeat for each basic head of each ring country
In reality. . . Insufficient overlap in ring list and new global list to allow this to be done at basic head level We do it instead using PPPs Our procedure Recalculate PPPs for each of 18 ring countries using only ring country data for consumption in both 2005 and 2011 We use GBR as numeraire, as in the original ring. We can check extrapolation, by comparing change in log PPPs relative to GBR from 2005 to 2011 with change in log CPI minus change in log CPI for GBR If ring overpriced in poor countries, ICP calculation of CPI change should be substantially lower than reality
What do we see? GBR goes through (0,0) by construction Extrapolation for other rich ring countries is OK Japan an exception, but not if the ring list were unrepresentative there South Africa in the other direction, but ring list likely representative there, for at least part of the population For poorer ring countries, actual CPI change is much larger than predicted from ICP Extrapolation from 2005 would overestimate 2011 results The gap between the lines is 25 percent With a huge leap of faith at this stage, something like 2005 overestimation for Africa, Asia, and W Asia Correcting for ZAF, 22 percent
Summary so far Last slide shows plot of underestimation against log per capita income Confirms strong positive relationship between pc GDP and size of discrepancy Hardly slam-dunk evidence But it is plausible, and together with standard errors and CPI deficiencies could take us quite a long way to understanding the difference between 2005 and 2011 Not clear what else could explain systematic reduction in PPPs in poorer countries
A cross check We can use the Balassa-Samuelson theorem, or its empirical counterpart, the Penn effect Price level index (the ratio of PPP to exchange rate) is lower in poor countries than in rich countries Because non-traded goods are relatively cheap in poor countries If the BS relationship is stable over time, particularly 2005 and 2011 Then in 2005, countries in W. Asia, Asia, and Africa should be above the line This turns out to be true Average gap between the two lines is 20 percent Consistent with direct evidence from the ring Gap between actual and extrapolations a little larger
End of Part One Africa, Asia, W. Asia 20-30 percent overstated in ICP 2005 Two corroborating sources of evidence Within region PPPs and not so different from extrapolations Also evidence for a ring-based explanation ICP 2011 looks OK (for now) Could revise 2005 by back-casting from 2011 using CPIs, and then adjust to preserve regional fixity Effectively revised only the ring linking factors
Poverty and the ICP H O W I T W O R K S ? A L T E R N A T I V E S ? W H A T T O D O ? L A T T E R I S M U C H M O R E A M A T T E R O F O P I N I O N
Starting points Bank’s poverty count uses consumption PPPs at several points Poorest countries’ poverty lines are averaged to get global line Having first been converted using PPPs Currently $1.25 per person per day The line is converted back to LCU at country PPPs to give LCU versions of global line Number of people below the line counted country by country NB If there were only one poor country (“India”) India’s national line would be the global line India’s PPP in USD is irrelevant for global count Changing India’s PPP affects the USD value of global line, but not the number of people in poverty More generally, with many poor countries, global count depends only on relative PPPs of poor countries PPPs of poor countries relative to the US matter for the global line, not count All assuming that the poorest countries are not changed at the same time
When PPPs change ICP 2011 lowered PPPs for poor countries relative to extrapolations If we hold global line at $1.25 (updated for US inflation since 2005) Local lines are now much lower, and many fewer poor Global poverty falls by around a half, Indian poverty by about two-thirds Post 2005, chief economist noted “the sobering news—that poverty is more pervasive than we thought—means that we should redouble our efforts, especially in sub-Saharan Africa” The corresponding response would today be “the inebriating news—that poverty is less pervasive than we thought—means that we should halve our efforts, especially in sub-Saharan Africa” Presumably not, though interesting to ask just why not
Recommend
More recommend
