Transcript Document

Comments on “Human capital and the wealth of nations”
by R. Manuelli and A. Seshadri
 What the paper is about:
• Accounting of output per worker
• Novelty: quality of education is taken into account, at least
conceptually.
• Main finding: most of cross-country income differences due to factor
accumulation, not TFP.
 Details:
• y = z kq h1-q
• y = output,
h = human capital,
k = physical capital,
z = TFP
• In the paper:
In Hall and Jones :
h = h(s, investment)
h = ers
• Lowest quintile has TFP equal to 73% of the US level
Hall and Jones estimate it at 20% (?)
• Interpretation: ignoring differences in quality of education amplifies
differences in TFP.

Alternative interpretations:
1.
Differences in human capital across countries are exaggerated:
–
–
–
2.
Quality is not properly measured:
–
–
–
3.
Schooling quantity and quality (and thus h) are estimated from
calibration.
Top/bottom quintile quantity : 20% higher in calibration than data
Top/bottom quintile quality: almost 40% higher in calibration
Proxy is public spending in schooling per pupil/GDP per worker
It ignores private spending
Does a higher ratio really mean better quality?
Different PWT databases. Does it matter?

Two more caveats:
•
Calibration for the US around 2000 from steady state implications
of the model.
–
–
–
•
Worst year to assume steady state in the US (period of abnormally high
growth rates – “new economy”)
Estimates of human capital in the rest of the world also based on steady
state assumption.
In general, is a country ever in steady state?
Role of h is inflated because of its endogenous response to TFP
changes.
–
In equilibrium h is ultimately determined by TFP (through wages) and
life expectancy

Further research I: the “Becker Paradox”
Life
∆s
Years of
expectancy
/∆L5
schooling (s)
at age 5 (L5)
•
1960
1990
1960
1990
Rich countries
72.4
76.8
7.8
10.6
.65
Middle & lowincome countries (ex
SSA)
62.8
70.3
3.2
6.1
.39
Sub-Saharan Africa
(SSA)
54.4
59.8
1.3
3.1
.33
Convergence in life expectancy but not in years of schooling. What
can the model say on this?

Further research II: Macro-Mincer return
•
Better education quality implies a higher return on schooling, ceteris
paribus.
•
For each country r = ra + r
where ra = average world return; r = deviation from average
•
Standard growth regression:

•
Dy = c + aDk + rDs + e
Dy = c + aDk + raDs + e , where e = rDs +e
So omitting schooling quality from growth regressions would bias
the estimated ra up. In practice it is 0. Can the model explain this?