Transcript Divergence, Big Time

Divergence, Big Time
Lant Pritchett
Journal of Economic Perspectives
Summer 1997
Main Idea
• The predictions of the Solow model are not
supported by data.
• Countries have experienced vastly different
rates of economic progress since the 1870s.
• Convergence does not seem to be
happening.
The Facts
• Lant starts with 17 “advanced capitalist” countries
as defined by Maddison.
• In Table 1, he documents their levels of income in
1870 and shows their average rates of growth in
three subsequent periods: 1870-1960, 1960-80 and
1980-94.
• Three facts jump out: (a) convergence, (b) but
rates of growth around a narrow band; (c) growth
rates pretty stable.
The Facts
• Making an inference about the evolution of the
WID on the basis of these data would be wrong:
de Long (AER, 1988).
• There are two kinds of bias, both of which aids the
“convergence hypothesis.”
• One, sample selection (only rich and poor but fastgrowth countries in the dataset).
• Two, measurement error (if either your initial
GDP data is “too high” or “too low” you help the
hypothesis).
How Can We Expand the Data?
• How about we estimate an initial GDP per
capital level for “poorer” countries outside
the 17 that we just examined?
• But how do we do that?
• Lant comes up with some clever
alternatives:
– Estimate a reasonable subsistence level of GDP
per capita ($250 in 1985 dollars).
How Can We Expand the Data?
– You can come up with such subsistence levels
of income per capita by running simple OLS
regressions with caloric intake data from the
FOA and per-capita income data from the Penn
World Tables (footnote 8). Then back-out the
income per capita needed for subsistence
caloric intake (under 2000/day). Turns out to be
P$250.
– How about the lowest five-year average GDP
per capita observed in the Penn World Tables
data (P$275 Ethiopia and P$278 Uganda).
Putting things together
• You need to buy these first: (a) the percapita income data for the set of all countries
in Penn World Tables (b) the estimates of
growth for the set of now-rich countries and
(c) all countries had to have at least P$250
back in 1870.
• Then, there had to have been a lot of
divergence in the WID between 1870 and
1960!
How So?
• Look at Figure 1.
• If there had been no divergence, then the
poorest countries in the world today ought
to have grown at rates at least as high as the
U. S. since 1870. Imputing backwards in
time, gives us incomes well below
subsistence for many countries (around
P$100 or less).
How So? (continued)
• The U.S. per capita income grew fourfold
between 1870 and 1960. In 1960, there were
42 countries (out of 125) whose per-capita
incomes were $1000 or less. All of these
countries must have grown at slower rates
than the U.S. if they had started out at $250
in 1870 (as we have assumed).
How So? (continued)
• In Table 2, Lant shows us some estimates
based on the $250 assumption and actual
data.
• The main point remains: developed
countries have stable, predictable patterns of
sustained growth and some convergence
among them; less-developed countries are
all over the map (poverty traps, takeoffs and
convergence, and meltdowns.