The measurement of PPPs and the measurement of poverty AN GU S D E - - PowerPoint PPT Presentation
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,
The new PPPs, why are they so different from what was
Work jointly with Bettina Aten, Bureau of Economic Analysis Paper is on my website
NBER Working Paper later this week
As well as PPPs and poverty more generally
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
We don’t need to deal with the trade balance, government
As well as estimates of the size of the problem And how to correct it
ICP2005 raised PPPs in poor countries relative to rich
ICP2011 results do the opposite, decreasing PPPs in poor
Discussion has mostly been about how implausible 2011
Little criticism of ICP 2005 because ICP 1993/ 96 had so many
No China, no India, piecemeal not integrated, and many new methods
So it was easy to dismiss the 2005 changes, which might have
Would account for jump up in 2005 Corresponding jump back down in 2011
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
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
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
Usual ICP issues of representativity versus comparability Not the same in selection of items for the CPI
Old weights, and relative price of food increased from 2005 to 2011 But this doesn’t do it, see paper
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
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
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
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
Extrapolation from 2005 would overestimate 2011 results
With a huge leap of faith at this stage, something like 2005
Correcting for ZAF, 22 percent
Confirms strong positive relationship between pc GDP and size
We can use the Balassa-Samuelson theorem, or its empirical
Price level index (the ratio of PPP to exchange rate) is lower in
Because non-traded goods are relatively cheap in poor countries
If the BS relationship is stable over time, particularly 2005
Then in 2005, countries in W. Asia, Asia, and Africa should be
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
Two corroborating sources of evidence
Also evidence for a ring-based explanation
Effectively revised only the ring linking factors
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
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
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
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
ICP 2011 lowered PPPs for poor countries relative to
If we hold global line at $1.25 (updated for US inflation since
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
The corresponding response would today be “the inebriating
Presumably not, though interesting to ask just why not
PPPs changed, but nothing actually happened to global poverty Perhaps we should recognize this, and adjust global line to keep
Around $2 in 2011 prices Still be major changes in regional and country structure of poverty
SSA much less sensitive to changes than is South Asia Large numbers of people in the latter near global lines These things matter: Africanization of poverty in 1993 revision of 1985
Major choice
Hold the line, and big reduction in count Hold the count, and increase the line
Statistical arguments are surely for the latter
But $1-a-day line has had considerable rhetorical value in development
discussions
Few people understand these matters, and even some who do, or should, will
accuse the Bank of self-interested dishonesty, whatever it chooses to do
Multiplies confusion
Guinea Bissau can determine the fate of India Weight all country lines by the number of poor people in the country
Better still weight by consumption patterns near the global line Makes less difference than one might think, see my paper with
Olivier Dupriez
In a world where many people live near the global line, headcount
PPPs have substantial standard errors: magnified by use in HCR Global poverty, as currently defined, perhaps cannot be measured with useful
precision
Not just ICP, but reporting periods in surveys, for example
Divorce global counts from the ICP
Might be possible to price a basic food bundle with local functional substitutions Bob Allen has argued for this, and has used it effectively in historical work Even costs $1.25 in the US
Don’t use PPPs at all
Remember global count does not depend on PPPs of poor countries relative to
the US
Only on relative PPPs among poor countries: not so different from exchange
rates: worth making calculations along these lines
Balassa Samuelson works mostly between rich and poor Would lose only $ value of global line, but perhaps time to lose that anyway