The B.E. Journal of Macroeconomics: Topics

The B.E.
Journal of Macroeconomics
Topics
Volume 9, Issue 1 2009 Article 16
Estimates of the Marginal Product of Capital,

1970-2000
Marcelo Mello∗
∗
Ibmec Rio de Janeiro, [email protected]
Recommended Citation
Marcelo Mello (2009) “Estimates of the Marginal Product of Capital, 1970-2000,” The B.E. Jour-
nal of Macroeconomics: Vol. 9: Iss. 1 (Topics), Article 16.
Brought to you by | Rice University

Authenticated
Download Date | 5/19/15 1:36 PM
Estimates of the Marginal Product of Capital,
1970-2000∗
Marcelo Mello
Abstract
We compute the marginal product of capital (MPK) taking into account cross-country dif-
ferences in the relative price of capital in a large panel including 95 countries over the period
1970-2000. Estimates of the price-uncorrected MPK suggest that the marginal product of cap-
ital is much larger in poor countries than in rich countries throughout the entire sample period.
However, estimates of the price-corrected MPK suggest that differences in the marginal product
of capital between rich and poor countries were significant in the 1970’s, decreased substantially
in the 1980’s, and were negligible in the 1990’s. In fact, the correlation between price-corrected
MPK and GDP per worker is close to zero for most of the 1990’s. Furthermore, we find that in
both rich and poor countries the capital-output ratio is increasing over time, and while the relative
price of capital is slightly decreasing in rich countries it is strongly increasing in poor countries.
Counterfactual reallocations of capital that equalize its rate of return around the world would have
yielded significant output gains in poor countries in the 1970’s, but gains in the 1990’s would have
been negligible. These findings are robust to alternative rich/poor cut-offs. Our results suggest
that international capital markets are largely integrated since the early 1990’s.
KEYWORDS: MPK, relative price of capital, capital-output ratio
∗
Department of Economics, Ibmec/RJ, Av. Presidente Wilson 118/1101, 20030-020, Rio
de Janeiro, RJ, Brazil; phone: (55-21)-4503-4161; fax: (55-21)-4503-4168; e-mail:
[email protected]. I thank seminar participants at the 76th meeting of the Southern Economic
Association in Charleston, South Carolina, and the 63rd, European meeting of the Econometric
Society in Milan, Italy. I would like to thank the editor, Peter Ireland, for encouragement, and two
anonymous referees who provided excellent comments, which led to substantial improvements in
the paper. I would also like to thank Miguel Lebre de Freitas, Marcelo Rezende, Louis Pavia,
and Fernando Veloso for carefully reading the paper and making many useful suggestions. All
remaining errors are mine.

Authenticated
Mello: Estimates of the MPK, 1970-2000
1. Introduction
In the neoclassical growth model with constant returns to scale and perfect
competition, the marginal product of capital (MPK) is computed as
/ , where α is the share of (reproducible) capital in total output, Y is output,
and K is physical capital. If the price of capital differs from the price of output,
then the expression for the MPK should be corrected for differences in the relative
price of capital. More precisely, the expression for the marginal product of capital
is given by / , where PMPK denotes the price-corrected
MPK, / is the inverse of the relative price of capital, and / is the inverse
of the capital-output ratio.
Caselli and Feyrer (2007), henceforth CF, compute the marginal product
of capital with corrections for cross-country differences in the relative price of
capital and the reproducible capital share. Their results suggest that the marginal
product of capital is largely equalized across countries. This finding is important
because it helps explain some fundamental issues. Notably, it addresses why
capital does not flow from rich to poor countries, a question posed by Lucas
(1990).
CF compute the MPK for a single cross-section of 53 countries from the
Penn World Tables (PWT) dataset. As mentioned above, their measure of the
MPK is corrected for cross-country differences in the relative price of capital and
the share of reproducible capital in total output. They observe that the concept of
capital in growth models refers to reproducible capital, not total capital. However,
available measures in the data refer to total capital, which includes reproducible
capital and natural capital. Natural capital includes inputs such as land and natural
resources, which are more abundant in poor countries. Since poor countries tend
have a larger share of wealth in the form of natural capital, the use of total capital
income as a measure of the capital share overestimates poor countries’ MPK.
CF face two constraints on their dataset. First, to generate estimates of the
reproducible capital share they use data from Bernanke and Gurkaynak (2001),
which is only available for 53 countries. Second, in order to construct a measure
for the reproducible capital share, they use a new dataset from the World Bank
which contains estimates of the components of wealth, including natural capital,
which is available for only one year.1
In this article, we construct a large panel including 95 countries with data
from the PWT 6.1 dataset over the period 1970-2000. We then compute the MPK
using its traditional one-sector formula, and its price-corrected version, i.e.,
corrected for differences in the relative price of capital. Instead of relying on a
1
The data on the components of wealth is taken from the World Bank publication “Where is The
Wealth of Nations?”, released in 2006.

Authenticated
The B.E. Journal of Macroeconomics, Vol. 9 [2009], Iss. 1 (Topics), Art. 16
single cross-section, the time series dimension and the broader set of countries
afforded by the panel allows us to examine the dynamics of the marginal product
of capital, the capital-output ratio, and the relative price of capital.
Our main findings are the following: First, over the period 1970-2000, we
find large differences in price-uncorrected MPK across countries. Second, we find
that these differences are substantially reduced once we take into account cross-
country changes in the relative price of capital. In particular, differences in price-
corrected MPK between rich and poor countries were relatively large in the 1970s
however they were negligible in the 1990s. Third, we find that the capital-output
ratio is increasing over time in both rich and poor countries. Fourth, over the
period 1970-2000, the relative price of capital decreases slightly in rich countries,
and it increases strongly in poor countries. Fifth, these findings are robust to
alternative rich/poor cut-offs.2
More specifically, our estimates suggest that the average MPK in poor
countries in the 1970s was 30.4%, while in rich countries it was 13.0%. Over the
1980s and the 1990s, MPK estimates were relatively stable across countries,
being, on average, 12% in rich countries and 24% in poor countries. Furthermore,
we find large differences between rich and poor countries’ price corrected MPK
during the initial years of the sample period. However, these differences decrease
over time, and disappear at the end of the sample period. The average price-
corrected MPK in poor countries in the 1970s was 18.9%, against 13.4% in rich
countries. In the 1980s and the 1990s, the average price-corrected MPK in poor
countries was 14.1% and 12.3%, respectively, against 12.3% and 12.2% in rich
countries, respectively.
The average capital-output ratio increases in both rich and poor countries
over the period 1970-2000. In 1970, the average capital-output ratio in poor
countries was 0.90. It grew at an average rate of approximately one percent per
year, reaching 1.23 in 2000. In rich countries, the initial average capital-output
ratio is 2.11, and it grew at an average rate of three-quarters of one percent per
year, reaching 2.66 in 2000. In addition, the relative price of capital is strongly
increasing in poor countries, with an average growth rate of 0.95% per year, while
it decreases slightly in rich countries, with an average negative growth rate of
0.15% per year.
In poor countries, the combination of an increasing capital-output ratio and
a strongly increasing relative price of capital implies that the ratio / is
decreasing over time. In rich countries / also decreases over time,
however at a much slower rate given the slow growth in the capital-output ratio,
and the smooth path of the relative price of capital. The dynamics of the capital-
2
In section 2, we discuss the criteria to determine the rich/poor cut-off.

Authenticated
output ratio and the relative price of capital explain much of the convergence in
price-corrected MPK across countries.3
Following the methodology in CF, we compute the world price-corrected
MPK, that is, the marginal product of capital that would emerge in a world
characterized by frictionless international capital markets. We then calculate
output gains/losses associated with counterfactual reallocations of capital so as to
obtain MPK equalization across countries.
We find that the gain in world GDP from the equalization of price-
corrected MPK is small, about 0.5% in 2000. Furthermore, the equalization of the
PMPK would have yielded substantial GDP gains in poor countries in the 1970s,
on average, GDP would have increased by 6.1%. However, GDP gains in the
1980s and the 1990s would have been quite small; on average GDP would have
increased by 1.4% and 0.3%, respectively.
Our article is closely related to the literature that highlights the role of the
relative price of capital in the development process. It is particularly related to
Hsieh and Klenow (2007), who show that cross-country differences in the relative
price of capital are the key driving force behind a well-known stylized fact in
growth, namely, the positive correlation between real investment rates and real
income levels across countries. In fact, CF extends the work of Hsieh and Klenow
(2007) by looking at implications of their findings for rates of return on physical
capital across countries.
This article is structured as follows. In section 2, we describe the
construction of the panel. In section 3, we present individual estimates of the
PMPK and its components. In section 4, we present panel estimates of the MPK
and its price-corrected version in both rich and poor countries over the period
1970-2000. In section 5, we present estimates of the gains from MPK and PMPK
equalization. Section 6 presents estimates of the MPK and the PMPK from
alternative panels with different rich/poor cut-offs as a robustness check. Section
7 concludes.
2. The Data
Our benchmark panel contains data on 95 countries over the period 1970-2000
from the PWT dataset version 6.1. We call this panel PWT61_1970. Using their
PWT codes, the input variables are the following: (population),
(real GDP per worker), (real GDP per capita: chain series),
3
Since a time series for the reproducible capital share is not available we have to be silent with
respect to its role in the convergence in PMPKs across countries.

Authenticated
(real GDP per capita: Laspeyres), (investment share of the ), (price

4
level of GDP), and (price level of investment).
We follow the methodology in Caselli (2005) to construct the following
time series for each country: / , relative price of output, constructed as the
price level of GDP divided by the price level of investment, the number of
workers is constructed as · / , and the level of
investment constructed as · · /100.
We use the perpetual inventory method to construct the aggregate capital
stock. Again, following Caselli (2005), we set the initial value of capital at
/ , where is investment in the first year of the sample, is the average
growth rate in investment for the first year in which data is available until 1970,
and is the depreciation rate which is set at 6%. The capital-labor ratio is
obtained by dividing aggregate capital by the number of workers. Similarly, the
capital-output ratio is obtained by dividing the capital-labor ratio by GDP per
worker.
In order to ensure the quality of the capital stock series, for each country,
we initiate the aggregate capital series on the first year for which data is available.
For instance, the first year of the aggregate capital stock series for Argentina is
1950, however, the first observation we use is the 1970 observation. That is, we
discard the first 20 observations. In total, we initialize the aggregate capital stock
series in 1950 for 48 countries. For the remaining 47 countries, the break down of
the initial year is as follows: For 26 countries the initial year is 1960, and for the
remaining 21 countries the initial year is in the interval 1951-1959, with the
notable exception of Tunisia for which the initial year is 1961.5
For each year, countries are ranked according to their GDP per worker and
classified as either rich or poor. The top 25 countries in each year are considered
rich, and the remaining 70 countries are considered poor. Since the panel contains
95 countries this rich/poor cut-off roughly corresponds to the third quartile of the
GDP per worker distribution. Note that according to this definition, rich and poor
countries may switch states over time. CF classify rich countries as those with
GDP per worker greater than or equal to Portugal’s GDP per worker.6 As
mentioned above, there are 53 countries in CF’s sample and Portugal is the 24th
highest GDP per worker. Thus, roughly speaking, the dividing line between rich
4
The reader is referred to Heston et al. (2004) for details on the definition of these variables. All
our data is available upon request.
5
Ferreira et al. (2008) also construct estimates of the capital stock using the perpetual inventory
method over the period 1970-2000. They perform a number of robustness exercises to check
whether their capital stock estimates are contaminated by a bad initial guess. They find no
evidence that the initial guess compromises estimates of the capital stock series.
6
To be precise, instead of rich vs. poor CF classify countries as either developed (up from
Portugal) or developing (up to Malaysia).

Authenticated
and poor in CF’s sample is the median, which tends to over-represent rich
countries. In section 6, we discuss alternative rich/poor cut-offs.
We assume a constant capital share in order to calculate the MPK and its
price-corrected version throughout the entire panel. We assume α=0.30 in all
countries at all time periods. We justify the assumption of a constant capital share
as follows. First, the benchmark steady-state value of the capital share used in the
literature is α=0.30. Second, the evidence in Gollin (2002) suggests that total
labor share is constant across time and space. Therefore, by residual, we deduce
that the total capital share is constant across time and space. Third, what matters is
whether or not the reproducible capital share is systematically related to GDP per
worker. The correlation between GDP per worker and reproducible capital share
in CF’s sample is 0.35, which is quite low.7 Hence, any bias resulting from
assuming a constant alpha should not be large. Fourth, by assuming a constant
alpha we are able to isolate the effects of two key macroeconomic variables,
namely, the relative price of capital and the capital-output ratio, on the dynamics
of the marginal product of capital.
If indeed poor countries have a smaller share of reproducible capital, then,
as mentioned above, our calculations will overestimate poor countries’ MPK. In
fact, in CF’s sample, the average total capital share in rich countries is 0.305, and
in poor countries is 0.382, while the average reproducible capital share in rich
countries is 0.206, and in poor countries is 0.168. Thus, correcting for the
reproducible capital share reduces MPK estimates in both rich and poor countries.
However, on average, the reduction in poor countries is much larger than the
reduction in rich countries. That is, the MPK in poor countries is overestimated if
computed with the total capital share instead of the reproducible capital share.
3. Individual Estimates
We first present individual estimates of the price-corrected MPK and its
components for selected countries. Presenting individual estimates can serve as a
simple robustness check, because if we observe any unusual pattern, we have
reason to be suspicious. We select four countries classified as rich (U.S.,
Switzerland, Japan, and U.K.), and four countries classified as poor (Mexico,
Brazil, India, and Turkey).8 Figures 1 and 2 display, respectively, the PMPK in
the four rich countries, and in the four poor countries.
7
Table 1 in CF suggests that the correlation between reproducible capital and GDP per worker is
0.70. However, as pointed out by CF, this correlation was computed using data from the World
Bank (2006), and it differs from the correlation between the reproducible capital share and GDP
per worker computed from the data they present in the appendix, which equals 0.35.
8
More precisely, these countries are classified as either rich or poor for most of the 1980s and the
1990s, but not throughout the entire sample period.

Authenticated
20.0%
18.0%
16.0%
14.0%
12.0%
10.0%
8.0%
6.0%
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
US Switzerland Japan UK
Figure 1: PMPK in Japan, Switzerland, U.K., and U.S.
35.0%
32.0%
29.0%
26.0%
23.0%
20.0%
17.0%
14.0%
11.0%
8.0%
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
Mexico India Brazil Turkey
Figure 2: PMPK in Brazil, India, Mexico, and Turkey
Over the period 1970-2000, the PMPK decreases in both set of countries.
The decrease in rich countries is smooth, while in poor countries it exhibits

Authenticated
marked oscillations. Overall, the pattern shown in figures 1 and 2 is consistent

with the estimates we present in next section.9
Table 1 displays estimates of the capital-output ratio for the above selected
countries. The capital-output ratio increases in all eight countries. The annual
average growth rate of the capital-output ratio in Japan is 2.3%, which is by far
the fastest growth rate among the four rich countries. The U.S. exhibits the second
fastest growth rate of the capital-output ratio, with annual average growth of
0.94%. The fastest growth rate in the capital-output ratio among the four poor
countries is Turkey with average annual growth of 2.3%. Alternatively, the
capital-output ratio in Brazil and India exhibits the slowest growth in the group
with average annual growth of 0.59% and 0.58%, respectively.
Table 2 displays estimates of the relative price of capital for the above
selected countries. The relative price of capital is decreasing in all rich countries,
except in the U.K. where it is stable over time. The relative price of capital falls at
an average annual rate of 0.72% in the U.S., 0.53% in Japan, and 0.36% in
Switzerland. On the other hand, the relative price of capital is strongly increasing
in Brazil and India, with average annual growth rate of 0.74% and 1.54%,
respectively. The relative price of capital in Mexico and Turkey exhibits marked
oscillations. However, in Mexico it ends up near stability with close to zero
average growth, and in Turkey it decreases continually starting in the mid 1980s.
Table 3 compares our estimates of the MPK and the capital-output ratio
for China with those from Hsieh et al. (2006). These authors compute the capital-
output ratio and the return to capital in China using disaggregated data from
national accounts. As shown in table 3, our estimates of the MPK and the capital-
output ratio and those from Hsieh et al. (2006) are close. The mean absolute error
in the MPK estimate is 2.7%, and in the case of the capital-output ratio it is 0.07.
In any case, it is reassuring that our estimates are so close to estimates from
disaggregated data.
As shown in the next section, the dynamics of the capital-output ratio and
the relative price of capital in this set of countries are consistent with the
dynamics of the aggregate series in the panel.
9
Although we do not attempt to do this here, it would be interesting to relate the marginal product
of capital with measures of the financial return to capital. Furthermore, it would also be interesting
to relate PMPK differentials with measures of bilateral financial integration. Perhaps, a starting
point in this direction could be the literature on empirical international finance such as Lane and
Milesi-Ferretti (2004), and Portes and Hey (2005).

Authenticated
Table 1: Estimates of the Capital-Output Ratio,

selected countries 1970-2000
Year US CHE JPN UK MEX IND BRA TUR
1970 1.70 2.70 1.84 1.59 1.51 0.92 1.70 0.94
1971 1.70 2.77 2.00 1.66 1.52 0.96 1.67 0.94
1972 1.68 2.84 2.08 1.77 1.49 1.00 1.65 0.94
1973 1.67 2.90 2.17 1.78 1.47 1.04 1.63 0.98
1974 1.75 2.99 2.42 1.93 1.51 1.10 1.70 1.10
1975 1.83 3.36 2.54 1.94 1.54 1.05 1.83 1.18
1976 1.78 3.47 2.61 1.92 1.57 1.08 1.83 1.23
1977 1.77 3.41 2.66 1.89 1.60 1.07 1.90 1.26
1978 1.77 3.42 2.69 1.93 1.58 1.07 2.00 1.26
1979 1.82 3.35 2.72 1.99 1.57 1.17 2.00 1.29
1980 1.90 3.24 2.80 2.14 1.60 1.14 1.98 1.35
1981 1.95 3.27 2.87 2.24 1.64 1.14 2.17 1.33
1982 2.07 3.36 2.92 2.07 1.71 1.14 2.22 1.33
1983 2.05 3.39 2.98 2.03 1.79 1.10 2.31 1.29
1984 1.99 3.33 2.99 2.07 1.76 1.09 2.23 1.24
1985 2.02 3.28 2.98 2.03 1.76 1.09 2.12 1.22
1986 2.03 3.29 3.02 2.03 1.83 1.08 2.06 1.19
1987 2.04 3.34 3.03 2.01 1.83 1.07 2.05 1.20
1988 2.04 3.31 3.00 1.97 1.91 1.05 2.11 1.27
1989 2.05 3.26 3.01 2.06 1.88 1.04 2.10 1.37
1990 2.09 3.24 3.03 2.14 1.86 1.06 2.23 1.39
1991 2.16 3.36 3.11 2.09 1.87 1.09 2.22 1.46
1992 2.15 3.44 3.24 2.04 1.91 1.10 2.27 1.49
1993 2.16 3.49 3.36 1.98 1.96 1.08 2.19 1.54
1994 2.15 3.51 3.45 1.94 1.98 1.07 2.10 1.69
1995 2.17 3.54 3.51 2.00 2.10 1.07 2.07 1.70
1996 2.19 3.59 3.51 2.00 2.06 1.04 2.05 1.71
1997 2.20 3.58 3.56 1.97 2.03 1.07 2.04 1.72
1998 2.22 3.55 3.70 1.98 2.04 1.06 2.14 1.77
1999 2.25 3.56 3.76 2.03 2.07 1.07 2.11 1.93
2000 2.27 3.52 3.76 2.00 2.00 1.10 2.04 1.92
Note: Author’s estimates using data from PWT 6.1, and following the methodology described in
section 2.

Authenticated
Table 2: Estimates of the Relative Price of Capital,

selected countries 1970-2000
Year US CHE JPN UK MEX IND BRA TUR
1970 1.07 0.88 1.03 0.90 1.15 1.45 1.70 1.35
1971 1.05 0.90 0.99 0.91 1.14 1.49 1.67 1.55
1972 1.04 0.94 0.98 0.96 1.14 1.49 1.65 1.41
1973 1.01 0.92 1.00 0.98 1.13 1.43 1.63 1.30
1974 0.96 0.88 1.00 1.01 1.09 1.48 1.70 0.82
1975 1.03 0.83 0.94 0.98 1.12 1.60 1.83 0.87
1976 1.00 0.85 0.92 0.97 1.14 1.57 1.83 0.94
1977 0.98 0.87 0.91 0.95 1.24 1.58 1.90 1.46
1978 0.96 0.92 0.90 0.97 1.22 1.60 2.00 1.68
1979 0.95 0.90 0.93 0.99 1.24 1.63 2.00 1.60
1980 0.91 0.91 0.94 1.02 1.22 1.68 1.98 1.73
1981 0.92 0.92 0.92 1.05 1.17 1.69 2.17 1.38
1982 0.94 0.87 0.90 0.97 1.30 1.71 2.22 1.56
1983 0.94 0.85 0.90 0.98 1.58 1.79 2.31 1.69
1984 0.92 0.89 0.91 0.99 1.48 1.87 2.23 1.89
1985 0.92 0.88 0.90 0.99 1.46 1.91 2.12 1.85
1986 0.93 0.85 0.89 1.00 1.52 1.96 2.06 2.06
1987 0.91 0.87 0.89 1.00 1.50 1.92 2.05 1.35
1988 0.90 0.90 0.89 0.99 1.33 1.90 2.11 1.62
1989 0.89 0.92 0.89 1.01 1.35 1.96 2.10 1.38
1990 0.88 0.89 0.91 1.02 1.30 1.95 2.23 1.20
1991 0.86 0.86 0.90 0.96 1.21 1.95 2.22 1.11
1992 0.87 0.86 0.91 0.92 1.13 1.98 2.27 1.06
1993 0.86 0.82 0.90 0.90 1.05 1.94 2.19 1.01
1994 0.87 0.81 0.89 0.90 1.03 1.92 2.10 1.08
1995 0.86 0.79 0.89 0.93 1.37 1.93 2.07 1.08
1996 0.86 0.78 0.89 0.93 1.37 1.94 2.05 1.31
1997 0.86 0.78 0.89 0.92 1.31 1.83 2.04 1.19
1998 0.89 0.77 0.89 0.92 1.17 1.76 2.14 1.08
1999 0.89 0.77 0.89 0.93 1.15 1.79 2.11 0.90
2000 0.85 0.79 0.87 0.91 1.11 1.82 2.04 1.05
Note: Author’s estimates using data from PWT 6.1, and following the methodology described in
section 2.

Authenticated
Table 3: Comparing Estimates of the Marginal Product of Capital in China

MPK K/Y MPK K/Y
Year Mello (2009) Mello (2009) Hsieh, Bai, Hsieh, Bai,
and Qian and Qian
(2008) (2008)
1978 22.4% 1.34 23.2% 1.39
1979 23.3% 1.29 22.0% 1.37
1980 22.9% 1.31 25.4% 1.36
1981 23.0% 1.30 20.9% 1.44
1982 23.6% 1.27 23.7% 1.45
1983 23.8% 1.26 24.5% 1.43
1984 25.0% 1.20 23.9% 1.34
1985 23.2% 1.29 25.7% 1.24
1986 23.2% 1.29 27.8% 1.32
1987 23.0% 1.31 26.6% 1.34
1988 21.9% 1.37 27.5% 1.28
1989 19.7% 1.52 24.6% 1.41
1990 20.0% 1.50 21.9% 1.49
1991 20.9% 1.44 26.1% 1.44
1992 21.6% 1.39 26.4% 1.36
1993 21.7% 1.38 27.9% 1.31
1994 21.2% 1.42 24.9% 1.39
1995 20.6% 1.46 24.1% 1.37
1996 19.9% 1.51 21.4% 1.39
1997 19.2% 1.56 22.0% 1.48
1998 18.8% 1.60 20.2% 1.57
1999 18.3% 1.64 19.6% 1.64
2000 18.7% 1.61 18.7% 1.63
Note: Columns 1-2 display author’s estimates using data from PWT 6.1 following the
methodology described in section 2. Columns 3-4 are taken from Hsieh, Bai, and Qian (2008),
table 1, pp. 71-72.
4. Estimates of the MPK, 1970-2000

Table 4 displays the average MPK and PMPK in rich and poor countries over the
period 1970-2000.10 First, we observe large differences between rich and poor
countries’ MPK. Second, the decrease in MPK in poor countries is quite rapid,
while in rich countries it is relatively slow. The MPK in rich countries falls from
10
We compute geometric averages.
10

Authenticated
14.2% in 1970 to approximately 11% in 2000, while in poor countries it falls from
33.5% in 1970 to approximately 24% in 2000.11
Table 4: The MPK and the price-corrected MPK, 1970-2000

Rich Poor Rich Poor
Year MPK MPK PMPK PMPK
1970 14.2% 33.5% 14.6% 22.2%
(6.5%) (42.9%) (3.3%) (13.3%)
1975 12.6% 29.6% 13.1% 18.8%

(3.2%) (38.5%) (2.1%) (9.4%)
1980 12.1% 26.1% 12.2% 15.7%

(3.9%) (27.3%) (1.8%) (7.1%)
1985 11.8% 24.0% 12.4% 13.9%

(3.1%) (23.4%) (2.1%) (6.7%)
1990 11.8% 24.3% 11.9% 12.5%

(2.2%) (19.3%) (2.3%) (4.7%)
1995 11.3% 24.1% 12.4% 12.1%

(1.9%) (17.4%) (1.9%) (4.2%)
1996 11.2% 24.3% 12.5% 12.1%

(1.9%) (17.6%) (1.9%) (4.4%)
1997 11.6% 24.2% 12.0% 12.6%

(3.3%) (18.1%) (2.4%) (3.6%)
1998 11.2% 24.2% 12.2% 12.3%

(1.9%) (17.9%) (1.8%) (3.5%)
1999 11.2% 24.2% 12.0% 12.0%

(2.0%) (17.8%) (1.7%) (3.6%)
2000 11.3% 24.3% 12.1% 12.0%

(2.0%) (17.5%) (1.8%) (3.6%)
Note: Standard deviation in parentheses.
11
The literature on dynamic Hecksher-Olin models suggests that large cross-country income
differences can be consistent with small interest rate differentials. In this sense, our MPK
differentials might be seen as upper bounds. See Cuñat and Maffezzoli (2004) for details.
11

Authenticated
Moreover, we observe that differences in marginal product of capital

between rich and poor countries are substantially reduced once we take into
account cross-country differences in the relative price of capital. In particular,
PMPK differences between rich and poor countries were significant in the 1970s,
considerably reduced in the 1980s, and negligible in the 1990s. In particular, the
average difference in PMPK between rich and poor countries in the 1970s is
5.52%, in the 1980s is 1.85%, and in the 1990s is 0.37%.
It is informative to examine the dynamic behavior of the components of
the PMPK. Figure 3 displays the average capital-output ratio in both rich and poor
countries. This ratio is increasing in both rich and poor countries. Over the period
1970-2000, the capital-output ratio increases by 38% in poor countries, against
26% in rich countries.12 Despite faster growth, at the end of the sample period the
capital-output ratio in poor countries still remains quite low when compared to
rich countries; it equals 1.23 against 2.66 in rich countries.
3.0
2.5
2.0
1.5
1.0
0.5
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
K/Y - RICH K/Y - POOR
Figure 3: The capital-output ratio in Rich and Poor countries
12
One may ask whether the increase in the capital-output ratio comes from an increase in the
saving rate. However, data from the PWT 6.1 suggest that the saving rate in both rich and poor
countries is stable over the period 1970-2000.
12

Authenticated
Figure 4 displays the average relative price of capital in both rich and poor
countries over the period 1970-2000. It increases by 34% over the period 1970-
2000 in poor countries, while it decreases by 4.5% in rich countries.13
The evidence presented above allows us to examine the role of the relative
price of capital and the capital-output ratio in promoting convergence in PMPKs.
In both rich and poor countries the ratio / decreases over time, however it
decreases much faster in poor countries. In rich countries, the capital-output ratio
increases and the relative price of capital decreases slightly. The increase in the
capital-output ratio more than offsets the decrease in the relative price of capital,
so that the ratio / decreases over time. In poor countries, the capital-
output ratio increases by 38% over the period 1970-2000, and the relative price of
capital increases by 34%, which causes a huge decrease in the / ratio.
2.1
1.9
1.7
1.5
1.3
1.1
0.9
0.7
0.5
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Pk/Py - RICH Pk/Py - POOR
Figure 4: The relative price of capital in Rich and Poor countries
Another way of looking at the dynamics of the components of the

marginal product of capital is to look the temporal evolution of the correlation
between GDP per worker and relative price of capital, and GDP per worker and
capital-output ratio. In the 1970s, the correlation between GDP per worker and the
relative price of capital averaged -0.49, and in the 1980s and the 1990s it averaged
-0.60 and -0.58, respectively. That is, over time, the correlation between GDP per
worker and relative price of capital increased in absolute value. Similarly, the
13
This by itself is an interesting fact which warrants further investigation. This is, in fact, a
research project companion to the present article.
13

Authenticated
correlation between GDP per worker and capital-output ratio became stronger
over time. It increased from an average of 0.64 in the 1970s, to 0.70 in the 1980s
and the 1990s.
This finding, namely, the convergence in PMPKs across countries,
suggests that capital transfers across countries could have had a strong impact on
poor countries’ GDP in the 1970s and maybe in the 1980s, but were doomed to
have little economic impact in the 1990s. In the next section, we provide estimates
of the potential gains from counterfactual reallocations of capital, which confirms
this intuition.
5. Estimates of the Welfare Gains from MPK Equalization

Under perfect international capital mobility the marginal product of capital should
be equalized across countries. In this case, there would be no gains in reallocating
capital across countries. We follow the methodology in CF in order to compute
the welfare gains from the equalization of the marginal product of capital. First,
we compute the world marginal product of capital, that is, the MPK that would
emerge in a world in the absence of capital markets frictions. We denote the
equalized world price-uncorrected MPK by MPK*, and the equalized world price-
corrected MPK by PMPK*. Second, given estimates of the MPK* and the
PMPK*, we estimate the gains in terms of world GDP from counterfactual
reallocations of capital, as well as GDP gains in rich and poor countries. Below,
we reiterate the methodology in CF, adapting it to the assumption of a constant
alpha.
Assume that the production function in sector j in country i is given by a
Cobb-Douglas technology, , and let denote the price of
output in country , and denote the price of capital. The rates of return within
country are equalized if for j=1, 2, …, J, that is, we must have:
j=1, 2, …, J (1)
Consider a situation in which capital was reallocated across countries such that
the PMPK is constant and equal to PMPK*.14 Therefore, the new value of capital
must satisfy
(2)
14
We follow CF and assume that labor is unchanged in the reallocation process.
14

Authenticated
Combining equations (1) and (2), we obtain an expression for the counterfactual
stock of capital in sector in country .
(3)
Aggregating across sectors to obtain the amount of capital in country , we have
∑ (4)
Where ∑ is the actual total capital in country . The resource constraint

requires the world capital stock to be fixed, that is, ∑ ∑ . Given the
15
resource constraint, we obtain an expression for the PMPK*:
∑ /
∑
(5)
Given the PMPK*, we obtain the counterfactual capital stock in country i from
equation (4). With the counterfactual capital stock in hands, we obtain the
counterfactual level of output in country . Further manipulation shows that the
counterfactual level of output in sector in country is given by
(6)
Summing over , we obtain the counterfactual level of GDP in country .
(7)
Gains from counterfactual reallocations of capital can be measured by the

percentage difference between the world counterfactual GDP and the world actual
GDP, that is
15
Because CF have country-specific alphas, they obtain a slightly more complicated expression.
They have to recur to a numerical procedure to obtain the PMPK*. In our case, since alpha is the
same for all countries, we can solve directly for the PMPK*.
15

Authenticated
∑
∑ 1 (8)
Figure 5 displays the counter-factual equalized MPK and PMPK over the
period 1970-2000. The MPK* and the PMPK* decline smoothly throughout the
entire sample period, reflecting the strong capital deepening process documented
in section 4. The MPK* and the PMPK* reach, respectively, 13.9% and 12.8% in
2000.
Figure 6 displays the gain in world GDP from the equalization of MPK
and PMPK. The gain in world GDP from the equalization of the MPK is between
2.5% and 3% per year. Gains in world GDP from the equalization of the PMPK
are smaller. On average, they would have been 1.38% in the 1970s, 0.72% in the
1980s, and 0.61% in the 1990s.
19%
18%
17%
16%
15%
14%
13%
12%
11%
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
MPK* PMPK*
Figure 5: Equalized MPK and PMPK
16

Authenticated
4%
3%
3%
2%
2%
1%
1%
0%
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
% Gain - MPK % Gain - PMPK
Figure 6: % Gain in World GDP
Figure 7 displays GDP gains/losses from the equalization of the PMPK in

rich and poor countries. The GDP gain in poor countries would have been
substantial in the 1970s, with an average annual gain of 6.1%. However, the
average annual GDP gain in the 1980s and the 1990s would have been
substantially smaller, respectively, equal to 1.4% and 0.3%.
To sum up, our estimates suggest that poverty alleviation schemes based
on capital transfers could have been effective in the 1970s and to some extent in
the 1980s. However, they were bound to have had little economic impact in the
1990s.
17

Authenticated
12%
10%
8%
6%
4%
2%
0%
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
-2%
-4%
Rich Poor
Figure 7: % GDP gain in Rich and Poor countries from PMPK equalization
6. Robustness Checks
In this section, we perform a series of robustness checks. First, we look at MPK
and PMPK estimates for the year 1996 and compare them with those in CF.
Second, we construct a panel with 49 countries from CF’s sample, and compute
the MPK and the PMPK over the period 1970-2000 using country-specific but
time-fixed estimates of the capital share from Bernanke and Gurkaynak (2001),
and CF’s estimates of the reproducible capital share. Third, we check whether
alternative rich/poor cut-offs alter our findings. For instance, we look at panels in
which countries can be classified as rich, middle-income, or poor, and panels in
which countries are divided into OECD and non-OECD, as alternatives to the
rich/poor classification of section 4. As it is shown below, none of our
conclusions are altered with these alternative rich/poor cut-offs.
Table 5 compares our estimates from section 4 for 1996 with those from
CF. Column (1) reproduces estimates of the PMPK and the PMPK-k from CF,
where PMPK-k denotes the PMPK corrected for the reproducible capital share.
Columns (2) and (3) display PMPK estimates imposing α=0.30 and α=0.185 to
the 53 countries in CF’s sample, respectively.16 Column (4) displays PMPK
estimates from our benchmark panel in section 4. Column (5) displays estimates
from our base panel imposing α=0.185.
16
Recall that 0.185 is the average value of the reproducible capital share in CF’ sample.
18

Authenticated
First, estimates in columns (2) and (4) in which we impose α=0.30 to both
samples, and in columns (3) and (5) in which we impose α=0.185 to both samples
are close, which suggests that correcting for cross-country differences in the
relative price of capital goes much of the way towards PMPK equalization
between rich and poor countries.17 Comparing PMPK estimates in columns (3)
and (5), we observe that our estimate of the PMPK in rich countries is 7.8%,
against 7.7% in CF’s sample, and that our estimate of the PMPK in poor countries
is 7.9% against 7.6% in CF’s sample.
Second, in CF’s sample, correcting the marginal product of capital for
cross-country differences in relative price of capital is not enough to generate
equalization in PMPKs, as can be seen in column (1). Their estimate of the PMPK
in rich countries is 12.6%, against 15.7% in poor countries. Similarly, when CF
corrects for the reproducible capital share only (not shown on table 5), their
estimate of the PMPK in rich countries is 7.5%, against 11.9% in poor countries.
Ultimately, in the case of CF, the correction for both the reproducible capital
share and the relative price of capital is necessary to obtain equalization in MPKs.
Table 5: Comparative Estimates of the MPK and PMPK for the year 1996
CF’s CF’s CF’s PWT_61 PWT_61
estimates sample sample α=0.30 α=0.185
(n=53) α=0.30 α=0.185 (n=95) (n=95)
(n=53) (n=53)
Col. (1) Col. (2) Col. (3) Col. (4) Col. (5)
PMPK
Rich 12.6% 12.4% -- 12.6% --
Poor 15.7% 12.4% -- 12.8% --
PMPK-k
Rich 8.4% -- 7.7% -- 7.8%
Poor 6.9% -- 7.6% -- 7.9%
Note: Estimates in column 1 are taken from CF. Estimates in column 2 are generated using CF’s
sample, and imposing α=0.30 to all countries. Estimates in column 3 are generated using CF’s
sample, and using the average reproducible capital share, α=0.185, from CF. Estimates in column
4 are taken from Panel PWT_61_1970, constructed by the author. Estimates in column (5) are
generated from our panel PWT61_1970, imposing α=0.185 as the capital share.
17
As mentioned above, in CF’s sample the correlation between reproducible capital share and
GDP per worker is 0.35. On the other hand, the correlation between the relative price of capital
and GDP per worker in their sample is -0.70, and the correlation between GDP per worker and the
capital-output ratio is 0.74. This suggests that the relative price of capital and the capital-output
ratio play a much bigger role in equalizing the marginal product of capital than the reproducible
capital share.
19

Authenticated
To further explore the role of the capital share in generating PMPK

estimates, table 6 displays estimates of the PMPK computed using data on capital
share from Bernanke and Gurkaynak (2001), which we denote by PMPK-BG, and
estimates of the PMPK using data on the reproducible capital share from CF,
which we denote by PMPK-CF, and compare them with our estimates of the
PMPK using α=0.30. As mentioned above, in both cases we assume that alpha is
country-specific but fixed over time. That is, we repeat the 1996 value of alpha
over time. Because we assume that alpha is country-specific, we have to limit the
number of countries in the panel to exactly match the data on capital
share/reproducible capital share. Therefore, we end up with a panel including 49
countries.18
In order to determine the rich/poor cut-off we follow CF, and assume that
countries with GDP per capita greater than or equal to Portugal’s GDP per worker
are considered rich. Since Portugal’s GDP per worker ranks 22nd in 1996, we
consider as rich any country among the top 22 of GDP per worker distribution.
Columns 1 and 2 display estimates of the PMPK-BG in rich and poor
countries. Estimates of the PMPK-BG suggest that there were large differences in
the marginal product of capital between rich and poor countries in the 1970s up to
the mid 1980s. For instance, in 1970 the PMPK-BG in poor countries was 22.7%,
and in rich countries was 14.9%. Around the mid 1980s and towards the end of
the sample period, PMPK-BG estimates in both rich and poor countries stabilize
around 12% and 15%, respectively.
Columns 3 and 4 display the PMPK computed assuming α=0.3. In this
case, we observe that equalization of the marginal product of capital is obtained
already around the mid 1980s. It is informative to reconcile these estimates with
the PMPK-BG. In Bernanke and Guykarnak’s (2001) data, the average capital
share in rich countries is 0.305, and in poor countries it is 0.382. That is,
according to their estimates the capital share in poor countries is greater than in
rich countries. Therefore, it is not surprising that PMPK and PMPK-BG estimates
for rich countries are so close, and that the PMPK-BG is higher than the PMPK in
poor countries.
Estimates of the PMPK-CF in rich and poor countries, respectively, are
shown in columns 5 and 6. Estimates suggest that the marginal product of capital
in rich and poor countries was already equalized in the 1970s. And, it remained so
until the mid 1980s. In the second half of the 1980s, the marginal product of
capital becomes higher in rich countries than in poor countries, and it remains so
towards the 1990s. Again, it is informative to reconcile these estimates with those
18
We initially started with the 53 countries from CF’s sample. However, CF has no data on
reproducible capital share for Hong Kong, and we could not obtain a complete set of data for
Algeria, Mauritius, and Singapore. Therefore, we also excluded these countries and end up with 49
countries.
20

Authenticated
generated assuming α=0.30. The average reproducible capital share in rich

countries is 0.206 while in poor countries it is 0.169. All else equal, this explains
why estimates of the PMPK-CF generate a higher marginal product of capital in
rich countries.
Table 6: The PMPK-BG and the PMPK-CF, selected years1970-2000

PMPK- PMPK- PMPK PMPK PMPK- PMPK-
Year BG BG Rich Poor CF CF
Rich Poor Rich Poor
Col. (1) Col. (2) Col. (3) Col. (4) Col. (5) Col. (6)
1970 14.9% 22.7% 14.9% 19.0% 9.0% 9.7%
(6.3%) (16.5%) (3.3%) (11.9%) (2.5%) (8.9%)
1975 13.4% 18.6% 13.3% 15.7% 8.4% 7.7%

(4.3%) (7.6%) (15.7%) (5.8%) (2.4%) (5.1%)
1980 12.0% 17.6% 12.2% 14.5% 7.4% 7.4%

(2.1%) (7.6%) (1.9%) (5.0%) (1.9%) (4.6%)
1985 12.0% 14.9% 12.4% 12.3% 7.6% 6.1%

(2.4%) (6.2%) (2.0%) (4.9%) (1.8%) (4.1%)
1990 11.7% 14.4% 12.2% 11.7% 7.9% 5.6%

(2.3%) (5.7%) (2.1%) (3.9%) (1.8%) (3.9%)
1995 11.9% 15.1% 12.3% 12.3% 8.0% 5.9%

(2.1%) (5.7%) (2.0%) (3.1%) (1.6%) (3.6%)
1996 11.9% 15.0% 12.4% 12.2% 8.1% 5.9%

(2.0%) (5.6%) (2.0%) (2.9%) (1.5%) (3.6%)
1997 11.9% 15.0% 12.4% 12.2% 8.0% 5.8%

(2.0%) (5.5%) (2.0%) (2.9%) (1.5%) (3.8%)
1998 11.7% 15.0% 12.1% 12.2% 7.9% 5.9%

(1.8%) (5.1%) (1.8%) (2.8%) (1.3%) (3.7%)
1999 11.5% 14.6% 12.0% 11.9% 7.8% 5.7%

(1.8%) (4.6%) (1.7%) (2.7%) (1.3%) (3.6%)
2000 11.6% 14.6% 12.1% 12.1% 7.9% 5.6%

(1.9%) (4.4%) (1.8%) (3.1%) (1.3%) (3.2%)
Note: Standard deviation in parentheses.
21

Authenticated
In conclusion, our finding that the PMPK converges across countries

around the mid 1980s shows up in the data under different assumptions about the
reproducible capital share, and in different sets of countries.
Alternative Rich/Poor Cut-offs
The rich/poor cut-off initially considered in section 2, though useful, can be

considered arbitrary. In this sense, we generate PMPK estimates for alternative
rich/poor cut-offs to check whether our results are affected. As shown below,
none of our initial results are altered under plausible alternative rich/poor cut-offs.
First we divide the countries in our benchmark panel, PWT61_1970, into
three income categories: rich, middle-income, and poor. The classification is as
follows: In each year, we rank the countries by their GDP per worker, and
consider the top 25 as rich, the next 25 as middle-income, and the remaining 45 as
poor. Figure 8 displays the PMPK in rich, middle-income, and poor countries
from panel PWT61_1970.
25%
23%
21%
19%
17%
15%
13%
11%
9%
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1995
1995
1996
1997
1998
1999
2000
Rich Middle-Income Poor
Figure 8: Price-corrected MPK in Rich, Middle-Income, and Poor countries
According to figure 8, the convergence pattern in PMPKs observed

previously is also present in the three income groups. PMPK differences across
countries were large in the 1970s. However, these differences were substantially
reduced in the 1980s, and became insignificant in the 1990s. Furthermore, the
PMPK in rich and middle-income countries was already equalized in the early
1980s. Overall, the convergence pattern exhibited in figure 8 is consistent with
22

Authenticated
our findings in section 4. Again, it is useful to examine the dynamics in the

components of the PMPK. Figure 9 displays the capital-output ratio in rich,
middle income, and poor countries.
2.5
1.5
0.5
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1995
1995
1996
1997
1998
1999
2000
K/Y - Rich K/Y - Middle-Income K/Y - Poor
Figure 9: The capital-output ratio in Rich, Middle-Income, and Poor countries
The capital-output ratio exhibits steady growth in all income groups. In

poor countries, the capital-output ratio grows by 48% over the period 1970-2000.
Despite its fast growth, its average value in 2000 is quite low, standing at 1.1. In
middle-income countries, the capital-output ratio grows by 21%, and in rich
countries it grows by 26% over the period 1970-2000. The average value of the
capital-output ratio in middle-income countries in 2000 is 1.47. This value is quite
low if compared to the average value of the capital-output ratio in rich countries,
which equals 2.66. Figure 10 exhibits the evolution of the relative price of capital
in these three income groups.
An interesting aspect shown in Figure 10 is that the relative price of
capital increases more in middle-income countries than in poor countries. This
pattern is consistent with the low growth in the capital-output ratio exhibited by
middle-income countries. Over the period 1970-2000, the relative price of capital
in middle-income countries increased by 42%, while in poor countries it increased
by 29%. In rich countries, as stated in section 4, the relative price of capital
decreased by 4.5%.
23

Authenticated
2.5
2.3
2.1
1.9
1.7
1.5
1.3
1.1
0.9
0.7
0.5
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1995
1995
1996
1997
1998
1999
2000
Pk/Py - Rich Pk/Py - Middle-Income Pk/Py - Poor
Figure 10: The relative price of capital in Rich, Middle-Income, and Poor
countries
In the second panel we consider, we divide the panel into OECD and Non-
OECD countries. We use the same set of OECD countries as did Mankiw et al.
(1992). Therefore, our OECD sample consists of the following 21 countries: U.S.,
Canada, U.K., Ireland, Portugal, Spain, France, Belgium, Netherlands, Italy,
Austria, Switzerland, Denmark, Sweden, Norway, Finland, Greece, Turkey,
Japan, Australia, and New Zealand.19
Figure 11 displays estimates of the PMPK in OECD and non-OECD
panel. According to Figure 11, the PMPK in OECD and non-OECD countries
exhibit the same convergence pattern observed previously. In particular, there
were large differences in PMPK in the 1970s until the mid 1980s. However,
already in the second half of the 1980s, we observe converge in the PMPKs across
OECD and non-OECD countries.
Figures 12 and 13 display the evolution of the capital-output ratio, and the
relative price of capital, respectively, over the period 1970-2000. The capital-
output ratio increases by 26% in OECD countries and by 39% in non-OECD
19
Mankiw et al. (1992) also include Germany in their OECD sample. However, we do not have
data on GDP per worker for Germany before 1990, and therefore had to exclude Germany from
the OECD list. Since we also exclude Luxemburg (because Mankiw et al. , 1992, also did), we end
up with 93 countries in this panel.
24

Authenticated
countries. The relative price of capital decreases by 3% in OECD countries, and it

increases by 32% in non-OECD members. These findings are in line with our
previous results.
24%
22%
20%
18%
16%
14%
12%
10%
PMPK - OECD PMPK - Non-OECD
Figure 11: The PMPK in OECD and non-OECD countries
2.1
1.9
1.7
1.5
1.3
1.1
0.9
0.7
0.5
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Pk/Py - OECD Pk/Py - Non-OECD
Figure 12: The relative price of capital in OECD and non-OECD countries
25

Authenticated
2.5
1.5
0.5
K/Y - OECD K/Y - Non-OECD
Figure 13: The capital-output ratio in OECD and non-OECD countries
In the third alternative cut-off we consider, we exclude all countries from

our benchmark panel with a population less than 1 million in 1985. This filter
eliminates 9 countries from the original 95, so we end up with 86 countries. Then,
in each year we classify as rich the 23 countries with the highest GDP per worker
in the distribution (top 27% of the distribution). In this case, as in the original
panel, in 1996, the country with the 23rd highest GDP per worker is Portugal,
which corresponds to the same cut-off point adopted by CF. The dynamics of
PMPK, capital-output ratio, and relative price of capital, are consistent with the
results obtained above. Therefore, to economize on space we do not report these
results here. They are, however, available upon request.
In conclusion, the stylized facts we document in section 4, namely,
convergence in PMPK between rich and poor countries starting in the mid 1980s,
growth in the capital-output ratio in both rich and poor countries, and high growth
in the relative price of capital in poor countries and stability in rich countries,
appear to be robust to plausible alternative rich/poor cut-offs.
7. Conclusion
We compute the MPK and the price-corrected MPK in a panel of 95 countries
over the period 1970-2000. We also calculate the MPK and its price-corrected
version under frictionless international capital markets. We then compute the
welfare gains associated with failure to attain MPK equalization.
26

Authenticated
We find evidence of convergence in price-corrected MPK across

countries. In particular, we find large differences in PMPKs across countries in
the 1970s until the mid 1980s. However, in the 1990s, PMPK differences between
rich and poor countries were negligible. In fact, the correlation between price-
corrected MPK and GDP per worker is close to zero for most of the 1990s. The
convergence in price-corrected MPK across countries implies that welfare gains
associated with capital transfers across countries are small. These results are in
line with previous findings in the literature that give support to the hypothesis that
international capital markets are integrated.
Furthermore, we find that the capital-output ratio is increasing in both rich
and poor countries, and the relative price of capital is increasing strongly in poor
countries and is slightly decreasing in rich countries.
Finally, because we have a time series dimension and a broader set of
countries, afforded by the panel, we are able to examine the dynamics of the
marginal product of capital and its components, the relative price of capital and
the capital-output ratio, instead of relying on a single cross-section as in the
previous literature.
References
Bernanke, B., and R. S. Gurkaynak, (2001) “Is Growth Exogenous? Taking

Mankiw, Romer, and Weil Seriously”, in B. Bernanke and K. Rogoff, eds.,
NBER Macroeconomics Annual, Cambridge, MA, MIT Press.
Caselli, F., (2005) “Accounting for Cross-Country Income Differences”, in

Handbook of Economic Growth, ed. P. Aghion, and S. Durlauf, Elsevier
North-Holland, vol. 1A: 679-741.
Caselli, F., and J. Feyrer, (2007) “The Marginal Product of Capital”, Quarterly
Journal of Economics, 122(2): 535-568.
Cuñat, A., and M. Maffezzoli, (2004) “Neoclassical Growth and Commodity

Trade”, Review of Economic Dynamics, 7(3): 707-736.
Ferreira, P., S. Pessoa, and F. Veloso, (2008) "The Evolution of International

Output Differences (1970-2000): From Factors to Productivity," The B.E.
Journal of Macroeconomics: Vol. 8: Issue 1 (Topics), Article 3.
Gollin, D., (2002) “Getting Income Shares Right”, Journal of Political Economy,
110 (2): 458-474.
27

Authenticated
Heston, A., R. Summers, and B. Aten, (2004) Penn World Table Version 6.1,
Center for International Comparisons at the University of Pennsylvania
(CICUP).
Hsieh, C., and P. Klenow, (2007) “Relative Prices and Relative Prosperity”,
American Economic Review, 97(2): 562-585.
Hsieh, C., C. Bai, and Y. Qian, (2006) “The Return to Capital in China”,
Brooking Papers on Economic Activity, 2: 61-101.
Lane, P., and G. M. Milesi-Ferretti, (2004) “International Investment Patterns”,

IMF Working Paper, 04/134.
Lucas, R. E., (1990) “Why doesn’t Capital Flow from Rich to Poor Countries?”,
American Economic Review, 80(2): 92-96.
Mankiw, N. G., D. Romer, and D. Weil, (1992) “A Contribution to the Empirics

of Economic Growth”, Quarterly Journal of Economics, 107(2): 407-437
Portes, R., and H. Rey, (2005) “The Determinants of Cross Border Equity Flows”,
Journal of International Economics, 65(2): 269-296.
World Bank, (2006), “Where is The Wealth of Nations?”, The World Bank.
28

Authenticated

The B.E. Journal of Macroeconomics: Topics

Uploaded by

Copyright:

Available Formats

The B.E. Journal of Macroeconomics: Topics

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The B.E. Journal of Macroeconomics: Topics

Uploaded by

Copyright:

Available Formats

The B.E.

Estimates of the Marginal Product of Capital,

Brought to you by | Rice University

KEYWORDS: MPK, relative price of capital, capital-output ratio

Brought to you by | Rice University

Brought to you by | Rice University

Brought to you by | Rice University

Brought to you by | Rice University

(real GDP per capita: Laspeyres), (investment share of the ), (price

Brought to you by | Rice University

Brought to you by | Rice University

Figure 1: PMPK in Japan, Switzerland, U.K., and U.S.

Mexico India Brazil Turkey

Figure 2: PMPK in Brazil, India, Mexico, and Turkey

Brought to you by | Rice University

marked oscillations. Overall, the pattern shown in figures 1 and 2 is consistent

Brought to you by | Rice University

Table 1: Estimates of the Capital-Output Ratio,

Brought to you by | Rice University

Table 2: Estimates of the Relative Price of Capital,

Brought to you by | Rice University

Table 3: Comparing Estimates of the Marginal Product of Capital in China

4. Estimates of the MPK, 1970-2000

Brought to you by | Rice University

Table 4: The MPK and the price-corrected MPK, 1970-2000

1975 12.6% 29.6% 13.1% 18.8%

1980 12.1% 26.1% 12.2% 15.7%

1985 11.8% 24.0% 12.4% 13.9%

1990 11.8% 24.3% 11.9% 12.5%

1995 11.3% 24.1% 12.4% 12.1%

1996 11.2% 24.3% 12.5% 12.1%

1997 11.6% 24.2% 12.0% 12.6%

1998 11.2% 24.2% 12.2% 12.3%

1999 11.2% 24.2% 12.0% 12.0%

2000 11.3% 24.3% 12.1% 12.0%

Brought to you by | Rice University

Moreover, we observe that differences in marginal product of capital

K/Y - RICH K/Y - POOR

Figure 3: The capital-output ratio in Rich and Poor countries

Brought to you by | Rice University

Pk/Py - RICH Pk/Py - POOR

Figure 4: The relative price of capital in Rich and Poor countries

Another way of looking at the dynamics of the components of the

Brought to you by | Rice University

5. Estimates of the Welfare Gains from MPK Equalization

Brought to you by | Rice University

Aggregating across sectors to obtain the amount of capital in country , we have

Where ∑ is the actual total capital in country . The resource constraint

Summing over , we obtain the counterfactual level of GDP in country .

Gains from counterfactual reallocations of capital can be measured by the

Brought to you by | Rice University

Figure 5: Equalized MPK and PMPK

Brought to you by | Rice University

Figure 6: % Gain in World GDP

Figure 7 displays GDP gains/losses from the equalization of the PMPK in

Brought to you by | Rice University

Brought to you by | Rice University

Brought to you by | Rice University