Hindawi Publishing Corporation

Economics Research International

Volume 2012, Article ID 584869, 19 pages
doi:10.1155/2012/584869

Research Article
Productivity of Nations: A Stochastic Frontier Approach to
TFP Decomposition

Jorge Oliveira Pires1 and Fernando Garcia2

1 Escola de Economia de São Paulo, Fundação Getulio Vargas, Rua Itapeva, 474-11 São Paulo, SP, Brazil
2 Confederação Nacional de Serviços, Rua Professor Tamandaré Toledo, 69-3 São Paulo, SP, Brazil

Correspondence should be addressed to Jorge Oliveira Pires, [email protected]

Received 31 August 2011; Accepted 26 October 2011

Academic Editor: Magda E. Kandil

Copyright © 2012 J. O. Pires and F. Garcia. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.

This paper tackles the problem of aggregate TFP measurement using stochastic frontier analysis. We estimate a world production
frontier for a sample of 75 countries over a long period. The “Bauer-Kumbhakar” decomposition of TFP is applied to a
smaller sample in order to evaluate the effects of changes in efficiency (technical and allocative), scale effects, and technical
change. Estimated technical efficiency scores are compared to productivity indexes offered by nonfrontier studies. We conclude
that differences in productivity are responsible for virtually all the differences of growth performance between developed and
developing nations and that a large part of this is due to allocative efficiency.

1. Introduction allocative efficiency change as determinants of long-term

growth. We are not aware at this time of any other SFA study
This paper uses an alternative way of measuring total factor that has produced quantitative results showing that allocative
productivity based on stochastic frontier analysis (SFA). The efficiency plays an important role in the economic growth of
great advantage of SFA is the possibility that it offers of de- nations.
composing productivity change into parts that have straight-
The stochastic frontier model used in this article assumes
forward economic interpretation.
the existence of technical inefficiency which evolves following
Previous studies have attempted to evaluate the efficiency
a particular behavior. This allows one to split productivity
and productivity growth of nations in the context of the so-
changes into (i) the change in technical efficiency, which
called technical inefficiency literature, most of which using
measures the movement of an economy towards (or away
data envelopment analysis (DEA) techniques for a sample of
from) the production frontier, and (ii) technical progress,
OECD countries (e.g., Färe et al. [1]). Deliktas and Balcilar
which measures shifts of the frontier over time.
[2, 3] have used SFA but in the context of a somewhat res-
tricted time period of analysis (1991–2000) which permits a When applied to a flexible technology (e.g., translog),
more comprehensive sample of countries (130-country fron- this technique further allows one to evaluate the presence
tier). However, they do not analyze or provide results about of scale efficiency. The Bauer-Kumbhakar decomposition
efficiency levels or rates of change of efficiency (or any other (Bauer [4], Kumbhakar [5]) may then be applied, allowing
component of TFP change) for the great majority of such the additional measurement of changes in allocative effi-
countries, focusing their analysis only on 25 transition coun- ciency. Our results show that this last component, together
tries. They do not study the role of allocative efficiency either. with technical change, explains a large portion of the differ-
The main contribution of our paper is to show that a ences in economic growth between developed and develop-
suitable decomposition of TFP can be applied to a fairly large ing countries.
sample of heterogeneous countries for an extensive period In the next section, we present the hypotheses behind
of time in order to evaluate not just the roles of technical the stochastic frontier estimation and TFP decomposition.
progress and technical efficiency change, but also scale and Section 3 presents the estimates of the world stochastic
2 Economics Research International

production frontier and discusses the technical efficiency Once it is assumed that v ∼ iid N(0, σ 2 ); u ∼
scores obtained comparing them to productivity indexes sug- NT(μ, σu2 ), that is, u has a normal-truncated distribution
gested by Islam [6] and Hall and Jones [7]. Technical progress (with a nonnull average μ) the two error components are
and returns to scale estimates are also discussed. In Section 4, independent of each other and x is supposed exogenous, the
we use the estimates of the previous section in order to model can be estimated by maximum-likelihood (ML) tech-
decompose TFP change from 1965 up to 2000. The role niques and the restriction of a half-normal distribution (μ =
of technical progress and allocative efficiency change in 0) can be tested. Given these conditions, the traditional asy-
economic growth of both developed and developing nations mptotic properties of the ML estimators hold. In addition,
is highlighted in Section 5. At last, in Section 6, we discuss we take the technical inefficiency component as time-variant,
the contribution of these results for the recent debate about as suggested by Battese and Coelli [12]:
the sources of economic growth and the nature and role of  
TFP components. uit = exp −η(t − T) · ui , uit ≥ 0, i = 1, . . . , N, t ∈ τ(i).
(3)
2. SFA and TFP Decomposition Other parameterizations of u are possible but we will
The approach adopted in this paper is that developed in not pursue them here (see, e.g., Kumbhakar [14], Corn-
the literature on technical efficiency and productivity, more well et al. [15], Lee and Schmidt [16]).
specifically in the “statistical” and “parametric” branches of In expression (3), τ(i) represents the Ti periods of
this literature, which is known as Stochastic Frontier Analysis time for which we have available observations for the i-nth
(SFA). The focus of SFA is to obtain an estimator for one country, among the available T periods in the panel (i.e.,
of the components of TFP, the degree of technical efficiency. τ(i) may contain all periods in the panel or only a subset
Technical efficiency is estimated in addition to technical of periods). The sign of η dictates the behavior of technical
change which in its turn is captured (as usual) by a time inefficiency over time. When η is not significantly differ-
trend and interactions of the regressors with time. The model ent from zero, technical inefficiency does not vary in time
used here is essentially that developed (independently) by (persistent inefficiency). This specification of the behavioral
Aigner et al. [8] and by Meeusen and van den Broeck [9]. pattern of inefficiency is somewhat inflexible, as the model’s
Their formulation was extended by Pitt and Lee [10] and architects themselves admit. According to the formulation,
Schmidt and Sickles [11] for the panel data case. Since these technical inefficiency must; grow at decreasing rates (η > 0)
two last mentioned studies, a number of enhancements have or decrease at increasing rates (η < 0). Moreover, the estimat-
been suggested, such as that of Battese and Coelli [12], in ed value for η is the same for all countries in the sample,
which the technical inefficiency is modeled so as to be time which means that the pattern of inefficiency rise or reduction
variant. A thorough compilation of this literature is found in is the same for all countries. Despite these limitations, we
Kumbhakar and Lovell [13]. believe the model can still bring very interesting insights into
The general stochastic production frontier model is de- the patterns of economic growth of nations.
scribed by the equations below, where y is the vector for the Assuming a translog technology with two production
quantities produced by the various countries, x is the vec- factors, namely, capital (K) and labor (L), the model can be
tor for production factors used, and β is the vector for the expressed in the following way:
parameters defining the production technology.
ln yit = β0 + βt · t + βK ln Kit + βL ln Lit + 0.5 · βtt · t 2
 
y = f t, x, β · exp(v) · exp(−u), u ≥ 0. (1)
+ 0.5 · βKK (lnKit )2 + 0.5 · βLL (ln Lit )2
The v and u terms (vectors) represent different error
components. The first one refers to the random part of the + 0.5 · βKL (ln Kit ) · (ln Lit ) + βKt [(ln Kit ) · t]
error, while the second is a downward deviation from the + βLt [(ln Lit ) · t] + vit − uit .
production frontier (which can be inferred by the nega- (4)
tive sign and the restriction u ≥ 0). Thus, f (t, x, β) ·
exp(v) represents the stochastic frontier of production and The output elasticities with respect to K and L can be
v has a symmetrical distribution to capture the random ef- obtained from (4), working out the derivatives. Due to the
fects of measuring errors and exogenous shocks that cause use of a translog technology, these elasticities are country and
the position of the deterministic nucleus of the frontier, time specific. The technical progress measure is also specific
f (t, x, β), to vary from country to country. The level of tech- for each country and period of time and can be obtained by
nical efficiency (TE), that is, the ratio of observed output partial differentiation of the deterministic part of (4) with
to potential output (given by the frontier) is captured by respect to time.
the component exp(−u) (note that TEit = yit / exp(xit β) = Bauer [4] and Kumbhakar [5] suggested a quite inge-
exp(xit β − uit )/ exp(xit β) = exp(−uit ) and, therefore, 0 < TE nious, yet simple, type of productivity decomposition which
< 1). For each country i and each time period t, we have goes beyond the division of productivity changes into a
  catchup effect and a technical innovation effect. Such frame-
yit = f t, xit , β · exp(vit ) · exp(−uit );
(2) work also accounts for scale effects and inefficient allocation
i = 1, . . . , N, t = 1, . . . , T. of productive factors. To perform this decomposition, we
Economics Research International 3

must first estimate the model depicted by (3) and (4). Then, reduces that of capital will necessarily bring a change in
it is possible to “compose” the rate of total factor productivity allocative efficiency. Only when there are no inefficiencies or
change from the results. In the expressions that follow, dots scale effects is the measure of productivity change identical
over variables indicate time derivatives, gTFP denotes the rate to technical progress.
of TFP growth, sK and sL are the shares of capital and labor in
aggregate income, and εK and εL are output elasticities with
respect to the factors of production. 3. Estimation of the World Stochastic
The components of productivity change can be identified Frontier (1950–2000)
from algebraic manipulations from the deterministic part of
the production frontier depicted in (2) combined with the The model estimation was conducted using statistics soft-
usual expression for the productivity change Divisia index: ware STATA, which includes among its preprogrammed
models that of Battese and Coelli [12]. The database for this
ẏ K̇ L̇ study consists of a nonbalanced panel for aggregated output
gTFP = − sK − sL . (5)
y K L and production factors (K and L) of a sample of countries
that includes both wealthy as well as poor nations. These
From the deterministic part of (2), we have data were basically obtained from Penn World Tables (PWT),
  version 6.1, for years 1950 to 2000. Data for factor shares
ẏ ∂lnf t, K, L, β K̇ L̇ ∂u
= + εK + εL − . (6) were obtained from the System of National Accounts 1968
y ∂t K L ∂t (SNA68, United Nations [17]) and from the Annual National
In the expressions that follow, RTS denotes returns to Accounts of OECD (OECD [18]). Since the construction
scale with RTS = εK + εL , gK is the growth rate of capital of a database that proved useful for conducting aggregate
(K̇/K) and gL is the growth rate of labor (L̇/L); λK = εK /RTS productivity analysis using SFA techniques seems to be an
and λL = εL /RTS are defined as normalized shares of capital important contribution of this study, we choose not to just
and labor in income. Combining (5) and (6), we have briefly describe the data and sample here, but to do it more
extensively in Appendix A.
 
gTFP = TP − u̇ + (RTS − 1) · λK · gK + λL · gL A number of alternative specifications were tested, im-
  (7) posing different restrictions on the parameters of the translog
+ (λK − sK ) · gK + (λL − sL ) · gL . technology. Likelihood ratios tests allow us to check if such
restrictions are valid or not. These statistics are presented in
That is, total factor productivity growth can be split into Appendix B. As a general result of these tests, we can say that
four elements: the statistics favor the (complete) translog functional form.
(i) technical progress, measured by TP = ∂ ln f (t, K, L, Results for the translog specification are presented in
B)/∂t; Table 1. All parameters are significant at 5%, except for the
capital elasticity of output, which is significant at 6.5%. The
(ii) change in technical efficiency, denoted by −u̇; mean inefficiency μ is significantly different from zero at
(iii) change in the scale of production, given by (RTS − 1) · 1%, showing that the normal truncated distribution is an
[λK · gK + λL · gL ]; appropriate assumption (if it were not significant, we would
(iv) change in allocative efficiency, measured by [(λK − fall back to the case of a half-normal distribution). The
sK ) · gK + (λL − sL ) · gL ]. estimated value of η is positive, which means that technical
efficiency grows at decreasing rates (catchup).
We can now study the impact of each of the components βKL is negative, revealing the possibility of substitution
of TFP. If the technology is immutable, it does not contribute between the production factors. The βt and βtt coefficients
to productivity gains. The same happens with technical indicate that the neutral part of technical progress has nega-
inefficiency. If it does not vary in time, it also does not have tive effects on production and in order to achieve (positive)
any impact on the rate of change of productivity. technical progress, it is necessary that the nonneutral part
The contribution of economies of scale depends both on of technical progress offsets these effects. The signs of βKt
technology as well as on factor accumulation. The presence and βLt indicate, respectively, that the nonneutral part of
of constant returns to scale (RTS = 1) cancels out the third technical progress goes hand in hand with capital accumu-
component on the right of (7). In the case of increasing lation (positive sign of βKt ), and inversely with labor supply
returns to scale (RTS > 1) and an increase in the amount (negative sign of βLt ), that is, technical progress is laborsaving
of productive factors, we have a higher rate of productivity and is more intense in countries where capital is abundant.
growth. If the amounts of production factors diminish, then Inspection of the results for returns to scale, techni-
we would have a reduction in the rate of productivity change. cal change, and technical efficiency reveals that these are
An inverse analogous reasoning can be made for decreasing economically meaningful. Table 2 shows country ranks for
returns and reduction (increase) in the amount of productive RTS, TE, and TP. The technical efficiency ranking must be
factors. viewed with caution. Although the presence of countries like
Since λK + λL = 1, the distances (λK − sK ) and (λL − sL ) Nicaragua, Venezuela, and El Salvador in the first positions
are symmetric and have opposite signs. Therefore, a factor does indeed seem odd, two aspects must be kept in mind:
reallocation that, say, increases the intensity of labor and (i) these results are “conditional” on the capital-labor ratio;
4 Economics Research International

Table 1: Time-variant inefficiency model (Battese and coelli [12]).

Number of observations = 746 Observations per country: minimum = 3

Number of countries = 75 Average = 9.9, maximum = 11
Wald χ 2 (8) = 14,540.41
Log likelihood = 272.07096
Prob > χ 2 = 0.0000
Confidence interval 95%
ln y Coefficients Standard error z P>z
Lower Upper
βt −0.1198 0.0455 −2.6300 0.0080 −0.2089 −0.0307
βK 0.2457 0.1330 1.8500 0.0650 −0.0149 0.5064
βL 0.3767 0.1883 2.0000 0.0450 0.0077 0.7458
βtt −0.0075 0.0015 −5.1300 0.0000 −0.0103 −0.0046
βKK 0.0275 0.0111 2.4900 0.0130 0.0058 0.0492
βLL 0.0572 0.0216 2.6500 0.0080 0.0150 0.0995
βKL −0.0605 0.0272 −2.2200 0.0260 −0.1138 −0.0072
βKt 0.0106 0.0023 4.6100 0.0000 0.0061 0.0152
βLt −0.0063 0.0030 −2.1100 0.0350 −0.0121 −0.0004
β0 8.8115 1.8366 4.8000 0.0000 5.2119 12.4111
μ 0.2074 0.0626 3.3100 0.0010 0.0846 0.3302
η 0.0652 0.0116 5.5900 0.0000 0.0423 0.0880
ln σ 2 −2.7946 0.2291 −12.2000 0.0000 −3.2437 −2.3456
ilgtγ 0.6735 0.3514 1.9200 0.0550 −0.0153 1.3623
σ2 0.0611 0.0140 0.0390 0.0958
γ 0.6623 0.0786 0.4962 0.7961
σu2 0.0405 0.0140 0.0131 0.0679
σv2 0.0206 0.0011 0.0184 0.0229

(ii) the estimations took place using PPP adjusted figures for 1, 000). Islam’s scores are averages for the period 1960–85.
GDP. In other words, the first aspect mentioned means that, For a detailed explanation of the indexes calculated by these
in a traditional Farrel diagram (Farrel [19]), a country such authors, we refer the reader to the original papers (Hall and
as Nicaragua is closer to the frontier, yet it is placed at the Jones [7] Islam [6], and also to the survey provided by Islam
“edge” of the unit isoquant closest to labor axis (lots of labor, [20], which compares the two of them.) Our main con-
scarce capital), at the same time that a country like Norway cern here, however, is not to analyze the absolute value of
would be further from the frontier, but on the opposite edge the indexes, but to compare how countries rank in each ap-
of the isoquant (abundant capital, scarce labor). The sec- proach.
ond aspect or caveat means that the productivity rank reflects As for the argumentation above, our adjusted index of
the efficiency in producing nontradables. When the value technical efficiency seems better suited for such comparisons,
of technical efficiency is converted by the PPP factor, we because it displays the efficiency scores in international US$.
have production efficiency evaluated at tradables’ prices. This This index is highly correlated with those suggested by the
adjusted TE ranking is displayed next to the first one in authors mentioned above, with the advantage that the rank-
Table 2 and shows a distinct ordering, in which developed ing seems to be more intuitive. The less productive nations
countries are positioned at the top, led by the United States. remain practically unchanged at the bottom of the ranking,
These adjusted scores would better translate international but the top of the productivity ranking no longer brings
competitiveness of countries. The numbers refer to the last less developed nations, as in Hall and Jones [7], for whom
subperiod in our sample (1995–2000). countries like Syria, Jordan, Mexico, and Brazil are listed
An interesting exercise is to compare our ranking to among the most productive economies, or as in Islam [6],
the productivity indexes suggested by Islam [6] and Hall in which Hong Kong is considered the most productive na-
and Jones [7], which are traditional studies that do not use tion: 53.7% more productive than the United States.
the frontier approach. Such indexes are also displayed in The results for RTS are also very intuitive. The countries
Table 2. The numbers of Hall and Jones [7] are obtained at the top of the ranking depict increasing returns to
using a variation of the Solow residual calculation. They do scale. These are large countries from the population and
a cross-sectional comparison with a base country, namely, territorial perspective. The bottom positions in the ranking
USA, for year 1988 (USA = 1,000). The numbers from Islam are occupied by basically very small (in size and population)
[6] are obtained using a nonfrontier panel data approach, countries. Another fact that comes to our attention is
which also stipulates USA as a base of comparison (USA = that Germany, GreatBritain, Italy, and France, all of them
Economics Research International 5

Table 2: Rankings: technical efficiency, returns to scale, and technical progress.

Technical efficiency Technical efficiency

Rank Hall and Jones [7] Islam [6] Returns to scale∗ Technical progress∗
(US$, PPP)∗ (US$)∗
1 NIC 0,975 USA 0,955 SYR 1,256 HKG 1,537 IND 1,148 JPN 0,79%
2 VEN 0,974 JPN 0,899 JOR 1,181 CAN 1,041 IDN 1,108 USA 0,44%
3 CAN 0,971 CHE 0,872 MEX 1,143 USA 1,000 USA 1,107 GER 0,27%
4 SLV 0,970 GBR 0,820 ITA 1,093 NOR 0,861 PAK 1,101 FRA −0,11%
5 MEX 0,968 ISR 0,811 HKG 1,090 BEL 0,787 BRA 1,098 CHE −0,31%
6 TUR 0,958 SWE 0,778 FRA 1,029 ESP 0,787 JPN 1,087 ITA −0,34%
7 USA 0,955 CAN 0,770 BRA 1,002 FRA 0,787 MEX 1,085 GBR −0,55%
8 ZAF 0,939 HKG 0,763 USA 1,000 JPN 0,787 PHL 1,081 NLD −0,66%
9 CHL 0,932 DNK 0,761 CAN 0,987 DNK 0,748 EGY 1,078 AUS −0,72%
10 IRN 0,929 NOR 0,740 ESP 0,983 GBR 0,712 TUR 1,077 AUT −0,76%
11 GTM 0,928 ISL 0,740 PRT 0,980 NLD 0,712 IRN 1,076 CAN −0,76%
12 TTO 0,924 SYR 0,723 GBR 0,962 SWE 0,712 THA 1,075 ESP −0,79%
13 HKG 0,890 FIN 0,722 AUT 0,958 AUT 0,677 GER 1,075 NOR −0,79%
14 TUN 0,883 IRL 0,717 BEL 0,948 GER 0,677 GBR 1,071 SWE −0,82%
15 CRI 0,882 FRA 0,685 NLD 0,926 CHE 0,619 ITA 1,070 DNK −0,83%
16 ISR 0,877 BEL 0,682 SWE 0,911 ISR 0,619 FRA 1,070 BEL −0,89%
17 ZWE 0,867 VEN 0,672 GER 0,900 TTO 0,619 KOR 1,066 KOR −1,13%
18 ECU 0,865 NLD 0,665 AUS 0,898 AUS 0,589 ZAF 1,066 FIN −1,14%
19 LKA 0,858 AUT 0,656 CHE 0,873 ITA 0,589 COL 1,064 HKG −1,28%
20 MYS 0,851 GER 0,643 VEN 0,873 NZL 0,589 ESP 1,062 NZL −1,55%
21 COL 0,849 ITA 0,631 ISR 0,840 VEN 0,533 KEN 1,061 GRC −1,61%
22 PRY 0,848 CRI 0,625 TTO 0,834 FIN 0,507 ARG 1,060 BRA −1,62%
23 GBR 0,833 IRN 0,615 GTM 0,825 MEX 0,487 MAR 1,059 ARG −1,68%
24 EGY 0,832 MEX 0,588 COL 0,800 SYR 0,463 NPL 1,056 PRT −1,69%
25 URY 0,831 AUS 0,587 FIN 0,800 BRA 0,419 UGA 1,056 ISR −1,83%
26 SEN 0,829 ESP 0,583 NOR 0,780 CRI 0,383 CAN 1,055 IRL −1,89%
27 JOR 0,816 GRC 0,558 DNK 0,778 GRC 0,383 PER 1,053 MEX −2,00%
28 PHL 0,816 JAM 0,558 IRL 0,770 IRL 0,383 VEN 1,051 MYS −2,38%
29 GRC 0,809 ARG 0,511 TUN 0,762 KOR 0,383 GHA 1,051 ISL −2,39%
30 NPL 0,800 URY 0,486 NZL 0,754 MYS 0,383 MYS 1,050 THA −2,42%
31 PAN 0,796 PRT 0,480 TUR 0,751 URY 0,383 LKA 1,049 ZAF −2,62%
32 ESP 0,790 NZL 0,466 JPN 0,744 ZAF 0,383 AUS 1,042 TUR −2,64%
33 AUS 0,788 KOR 0,448 GRC 0,742 PRT 0,347 SYR 1,041 VEN −2,69%
34 ARG 0,778 SLV 0,443 CRI 0,736 PER 0,330 CHL 1,041 CHL −2,80%
35 PER 0,776 CHL 0,416 ARG 0,730 PRY 0,330 ZWE 1,039 IRN −2,80%
36 ITA 0,776 PAN 0,415 URY 0,696 GTM 0,314 ECU 1,038 PER −2,92%
37 IRL 0,774 TTO 0,404 KOR 0,664 MAR 0,314 NLD 1,038 COL −2,95%
38 PRT 0,773 TUR 0,386 DOM 0,651 NIC 0,301 GTM 1,037 SYR −3,01%
39 MAR 0,772 GAB 0,371 ZAF 0,645 COL 0,287 MWI 1,036 IDN −3,16%
40 DOM 0,768 GTM 0,358 EGY 0,595 PAN 0,287 RWA 1,032 GAB −3,21%
41 BEL 0,764 MYS 0,348 MAR 0,576 TUN 0,273 SEN 1,032 URY −3,29%
42 FRA 0,755 ZAF 0,345 PER 0,565 TUR 0,273 TUN 1,030 PHL −3,36%
43 FIN 0,752 BRA 0,339 MYS 0,560 ARG 0,259 TCD 1,030 IND −3,51%
44 IDN 0,752 PER 0,337 SLV 0,557 JOR 0,259 GRC 1,029 TUN −3,52%
45 BRA 0,750 JOR 0,325 PRY 0,541 SLV 0,247 PRT 1,029 EGY −3,58%
46 HND 0,745 DOM 0,319 PAK 0,527 THA 0,247 BOL 1,028 TTO −3,62%
47 BOL 0,744 HND 0,318 CHL 0,522 ECU 0,237 DOM 1,028 PAN −3,71%
6 Economics Research International

Table 2: Continued.
Technical efficiency Technical efficiency
Rank Hall and Jones [7] Islam [6] Returns to scale∗ Technical progress∗
(US$, PPP)∗ (US$)∗
48 SWE 0,742 EGY 0,286 THA 0,513 CHL 0,225 BEL 1,028 MAR −3,71%
49 NLD 0,738 COL 0,281 ECU 0,504 DOM 0,214 SWE 1,023 CRI −3,73%
50 CHE 0,738 BOL 0,252 LKA 0,481 PAK 0,194 HND 1,022 JAM −3,80%
51 RWA 0,737 TUN 0,252 BOL 0,469 PHL 0,186 AUT 1,021 ECU −3,95%
52 PAK 0,734 ECU 0,250 PAN 0,463 BOL 0,169 SLV 1,020 DOM −4,02%
53 TCD 0,722 PRY 0,242 HND 0,449 JAM 0,169 HKG 1,018 PRY −4,10%
54 GAB 0,720 NIC 0,237 NIC 0,443 EGY 0,153 CHE 1,017 JOR −4,25%
55 DNK 0,713 SEN 0,226 JAM 0,410 LKA 0,153 NIC 1,017 SLV −4,29%
56 LSO 0,708 LSO 0,210 PHL 0,389 HND 0,126 PRY 1,016 GTM −4,39%
57 NZL 0,703 MAR 0,209 IND 0,344 NPL 0,120 ISR 1,015 LKA −4,48%
58 AUT 0,700 PHL 0,203 SEN 0,316 SEN 0,110 JOR 1,014 PAK −4,53%
60 KOR 0,692 LKA 0,188 ZWE 0,275 UGA 0,104 DNK 1,010 BOL −4,58%
61 JAM 0,678 ZWE 0,185 NPL 0,244 ZWE 0,104 FIN 1,010 HND −4,90%
62 GER 0,677 THA 0,176 RWA 0,242 IND 0,071 CRI 1,006 ZWE −4,90%
63 NOR 0,659 KEN 0,161 KEN 0,237 KEN 0,071 NOR 1,006 LSO −5,19%
64 ISL 0,654 RWA 0,159 GHA 0,215 RWA 0,065 IRL 1,004 NIC −5,24%
65 IND 0,640 UGA 0,158 UGA 0,162 MWI 0,058 URY 1,004 KEN −5,32%
66 GHA 0,637 PAK 0,146 TCD 0,151 GHA 0,053 NZL 1,003 GHA −5,49%
67 UGA 0,635 TCD 0,138 MWI 0,130 TCD 0,042 PAN 1,000 SEN −5,50%
68 SYR 0,632 IDN 0,136 JAM 0,998 NPL −5,94%
69 JPN 0,621 GHA 0,119 LSO 0,994 MWI −6,10%
70 KEN 0,611 NPL 0,118 TTO 0,981 UGA −6,19%
71 THA 0,595 IND 0,109 GAB 0,975 RWA −6,43%
72 MWI 0,502 MWI 0,102 ISL 0,939 TCD −6,56%
Source: Islam [20], Table 11.3 page 485–488], and own estimations. Hall and Jones indexes depicted here are equal to exp (log A), in Table 9 of Hall and Jones
(1996) [7].
(∗ ) Annual average rates 1996–2000. Three countries were left out due to lack of information for this subperiod.

European nations of very homogeneous characteristics, are sight, a positive relation between these variables. Figure 1(b)
placed next each other in the ranking. plots the measure of technical progress against the average
The results for technical progress seem at first sight rather education level of the population revealing another intuitive
odd, with almost all of them being negative. (Although most relation: countries with better educated population are also
of the numbers for Technical progress (TP) in Table 2 are the ones with the highest levels of technical progress.
negative, they refer only to the last subperiod of our sample, In addition to the analysis of production frontiers using
that is, 1996–2000. TP presented a pattern in which it starts data at five-year intervals, two other experiments were car-
positive and then shows a decaying trend along the sub- ried out to evaluate the relative performance of the model: (i)
periods in the sample.) Nonetheless, the ordering seems to the estimation of the stochastic frontier model using annual
match our intuition regarding the technological performance data, and (ii) the estimation of traditional panel data models
of nations. At the top positions are Japan, United States, Ger- (fixed effects and random effects). The results are presented
many, and France. Among the countries at the bottom are in Appendix C and suggest that frontier models may be better
the African nations, wellknown for their lack of technological suited for the analysis of productivity in comparison with
knowledge. traditional econometric methods.
A simple exercise of casual empiricism provides an inter-
esting “test” of the existence of economic intuition behind
the estimations of technical progress performed by the 4. The Behavior of TFP and Its Components
model. The idea is to evaluate if the measure of technical
progress produced is related in any way to the effort to inno- With the results obtained in Section 3 for the 75-country
vate carried out by countries in recent years. Figure 1(a) sample, and the data on functional distribution of income
brings a scattered diagram for the technical progress measure (sK and sL ), it is possible to decompose productivity change
and the natural logarithm of R&D expenses (average for in the manner shown in Section 2. However, data for factors
1990–2000) and shows us what seems to be, at least at first shares in income are not available for all these economies.
Economics Research International 7

0.01 JPN them, Japan’s performance stands out, with an averageannual

USA productivity growth rate of 2.42%. The countries that follow
GER
are Austria (1.77%), France (1.75%), Norway (1.53%), Swit-
0 zerland (1.51%), and USA (1.49%). In the middle block, we
CHE ITA

AUS find some Latin American countries, such as Jamaica, Brazil,

DNK SWE
Peru, Venezuela, and Bolivia, all of them with relatively low
Technical progress (p.a. % )

−0.01 KOR
TFP growth rates. Brazil showed an average rise in producti-
PRT ARG
BRA vity of 0.39% per year. Among the other Latin American
MEX countries in the sample, we see that Mexico, Costa Rica, and,
−0.02
ISL THA MYS
surprisingly, Chile had reductions in productivity. Greece
VEN and Turkey are the only OECD members with negative
CHL
productivity growth during this time.
−0.03
PHL Countries for which technical progress (TP) showed
IND
the greatest impact on productivity growth were Japan, the
−0.04
United States, France, Switzerland, Italy, United Kingdom,
The Netherlands, and Australia, in this order. Contributions
of TP for this group of countries ranged from 0.56 to 0.30
−0.05 percent per year, on average. As we can see, they are all de-
veloped nationsthat invest substantial amounts in R&D.
10
20
30
40
50
100
200
300
400
500
1.000
2.000
3.000
4.000
5.000
10.000
20.000
30.000
40.000
50.000
100.000
200.000
300.000
Among the 19 countries that presented positive contri-
butions of technical progress, 18 are OECD members (see
R&D expenditures (Millions of US$)—average 1990–2000 Table 3). As mentioned before, for many countries, the last
(a) subperiod (1996–2000) reveals negative technical progress
0
(as shown in Table 2). Notwithstanding that, in Table 3, we
depict positive average annual contributions of TP to growth
JPN for 19 countries (of the total 36). These numbers are averages
USA
GER calculated for the longer period 1970–2000.
FRA
0 ITA CHE Brazil is the only member of OECD that managed to
ESP DNK AUS SWE have technical progress contributing for higher productivity,
KOR
Technical progress (p.a. % )

BRA PRT ARG

mainly in the 1965 to 1985 period. This trend matches that
MEX of three other Latin American countries that underwent a
−0.0 THA MYS ISL
VEN CHL
marked import substitution process, namely, Mexico, Peru,
COL
PHL
and Venezuela. The fall in the pace of technical progress of
IND
these countries coincides with the debt crisis and economic
−0.0
PAK
liberalization, periods during which the industrialization
ZWE process slows down its pace.
SEN
LSO An important aspect pertains to the interpretation of
technical regress (negative technical progress) that appears in
−0.1 UGA
the results of this study (other authors also report this kind
of result using frontier techniques—Rao and Coelli [21] is
an example). First, it should be pointed out that a frontier
−0.1 was not estimated for each country and, therefore, it is not
0 2 4 6 8 10 12 14 a matter of saying that this or that country had “inward”
Average years of schooling of population over 25 years shifts to their frontiers. The interpretation is quite difficult
(b) in light of the way that technical progress was achieved, by
including a time trend in the model (and interactions of
Figure 1: R&D Expenditures, human capital and technical time with capital and labor). According to Arrow [22], this
progress. Sources: own estimations and World Development Indica- procedure, which is rather common in the literature, is most
tors (World Bank) for R&D and Schooling. Technical progress rates of all a confession of ignorance. As discussed in Section 3, the
are averages for 1996–2000. underlying idea here is that countries closer to the frontier
(and on the forefront of technical progress) are responsible
for the actual shift in the world production frontier.
We managed to collect data for only 36 of the 75 countries in One way of interpreting technical regress in less devel-
the sample, and just from 1970 up to 2000. The full decom- oped nations is that it may be the result of circumstances or
position of TFP is then restricted to this group and period. decisions that end up halting the production of some high-
Table 3 brings the results. technology products and encouraging the manufacturing of
Examination of average productivity change numbers low-technology products. Since GDP is the aggregation of
along the 30 years in question shows some interesting results. value added in a number of industries, this sliding perfor-
All top positions are occupied by OECD countries. Among mance could be the result of production shifting from some
8 Economics Research International

Table 3: Sources of economic growth 1970–2000: average annual % change.

1 2 2
Economic Capital Labor Productivity change Random3
Country
growth accumulation expansion Technical Technical Scale Allocative shocks
TFP
progress efficiency effects efficiency
AUS 4.16 1.17 1.06 0.93 0.30 0.46 0.06 0.12 1.00
AUT 3.74 1.70 0.30 1.77 0.25 0.69 0.01 0.81 −0.03
BEL 3.29 1.60 0.31 1.40 0.23 0.52 0.03 0.61 −0.02
BOL 2.73 1.68 1.00 0.05 −0.55 0.57 0.00 0.02 0.00
BRA 5.59 4.51 1.13 0.39 0.01 0.55 0.41 −0.58 −0.44
CAN 4.14 1.87 1.10 0.98 0.26 0.06 0.12 0.54 0.19
CHE 2.03 1.31 0.52 1.51 0.40 0.59 −0.01 0.53 −1.30
CHL 4.98 3.88 0.87 −0.41 −0.30 0.13 0.11 −0.35 0.64
COL 4.93 3.73 1.05 −0.20 −0.30 0.31 0.20 −0.41 0.34
CRI 4.89 4.01 1.63 −0.32 −0.47 0.24 −0.12 0.03 −0.43
DNK 2.61 0.89 0.35 1.39 0.27 0.65 −0.01 0.47 −0.02
ESP 3.97 2.79 0.54 1.30 0.23 0.45 0.17 0.44 −0.66
FIN 3.65 1.76 0.37 1.45 0.18 0.55 −0.02 0.74 0.07
FRA 3.43 1.79 0.47 1.75 0.41 0.54 0.15 0.64 −0.57
GBR 2.73 0.99 0.27 1.33 0.32 0.35 0.10 0.55 0.13
GRC 3.67 4.37 0.30 −0.46 0.05 0.41 0.04 −0.95 −0.55
IRL 6.03 2.18 0.61 0.97 −0.05 0.49 −0.04 0.56 2.27
ISL 4.13 1.22 0.93 0.97 −0.09 0.82 −0.21 0.46 1.01
ITA 3.53 2.14 0.29 1.24 0.35 0.49 0.15 0.25 −0.14
JAM 1.80 2.06 0.86 0.41 −0.43 0.75 −0.08 0.18 −1.54
JOR 6.23 5.95 2.17 −0.83 −0.63 0.39 −0.14 −0.45 −1.06
JPN 5.26 3.54 0.58 2.42 0.56 0.92 0.34 0.57 −1.28
KEN 5.17 3.22 1.48 0.07 −0.79 0.95 0.16 −0.24 0.39
KOR 9.31 6.93 0.88 0.62 −0.03 0.71 0.40 −0.46 0.87
MEX 5.00 4.57 1.27 −0.18 −0.07 0.06 0.34 −0.51 −0.66
NLD 3.59 1.20 0.65 1.27 0.30 0.58 0.05 0.34 0.46
NOR 4.09 1.75 0.40 1.53 0.25 0.81 −0.03 0.50 0.41
NZL 2.39 0.47 0.77 0.76 0.15 0.68 −0.02 −0.05 0.39
PER 3.15 2.75 1.00 0.21 −0.20 0.49 0.10 −0.17 −0.81
PRT 4.71 3.09 0.39 1.20 −0.02 0.50 0.05 0.67 0.03
SWE 2.57 0.96 0.36 1.44 0.28 0.57 0.01 0.57 −0.20
THA 8.01 6.51 0.62 −0.73 −0.29 1.00 0.37 −1.79 1.60
TTO 3.62 2.65 0.83 −0.23 −0.41 0.15 −0.16 0.18 0.36
TUR 5.38 5.93 0.79 −1.33 −0.26 0.08 0.33 −1.48 −0.01
USA 3.97 1.70 0.84 1.49 0.52 0.09 0.28 0.59 −0.07
VEN 1.82 2.24 1.41 0.06 −0.10 0.05 0.08 0.04 −1.90
(1) Growth of GDP; (2) Growth rates adjusted by income shares; (3) Obtained as a residual.

highly productive sectors to others, where productivity is In the sample including all 75 countries, the estimated
lower. To test this hypothesis is beyond the scope of this value for this parameter was positive, which resulted in a
paper and would require an analysis disaggregated by sectors catchup pattern for all countries: technical efficiency grows at
of activity (an interesting example of a disaggregate study decreasing rates. The countries that had the largest catchups
is Kim and Han [23], which analyses several South Korean were Thailand, Kenya, Japan, Iceland, Norway, Jamaica, and
industries using the Bauer-Kumbhakar decomposition). South Korea.
All countries enjoyed rising technical efficiency as shown It is quite intuitive that Thailand, Japan, and South Korea
by the positive numbers in Table 3. That is a characteristic should appear at the top here, since they have made great
of the estimated model. The Battese and Coelli [12] model effort to absorb technology and educate their people. For
imposes the restriction of a common η to all countries. the other countries, however, this conclusion does not seem
Economics Research International 9

to be so obvious. Nonetheless, Kenya, Iceland, and Jamaica efficiency remains above the measure that excludes this
enjoyed very high rates of growth during some periods in component. The opposite happens in Brazil for most of the
the sample, revealing a movement towards the frontier whose sample period and in all of it for Mexico. South Korea, on the
cause could only be understood following a deeper inves- other hand, has a distinct pattern, in which the curves cross
tigation of the history of these economies (something beyond each other, that is, allocative efficiency inverts its impact,
the scope of this study). Among countries with low technical becoming a driver for productivity gains in that country.
efficiency improvements are the United States and Canada For Brazil, TFP computed without allocative efficiency
which is reasonable, since both these nations are already close is usually superior showing the effects of “ill allocation” of
to the frontier. They are in fact pushing the frontier further. production factors. From the mid 1980s to the mid 1990s,
It is plausible to suppose that countries with vast masses this tendency reverses and begins to contribute to producti-
of population are those set to gain the most from scale effects. vity growth, even if very little. After that period, the contribu-
That is exactly the result obtained, with Brazil, South Korea, tion turns negative again, but not as large as in the first five-
Thailand, Mexico, Japan, Turkey, and the USA showing large year periods under study. Mexico also reduces the negative
numbers for this component. All of them but Japan and the allocative effects as of the mid 1980s, but never enough to
USA are usually referred to as “developing” nations and have contribute to a rise in productivity. For those two countries,
surely experienced leaping growth during some subperiods the improvement in allocative efficiency roughly coincides
of the sample, most of which based on factor accumulation. with market-oriented reforms. A significant relationship be-
This accumulation also paved the way to productivity growth tween such phenomena remains to be tested though. On the
based on scale. It is also rather intuitive that countries with other hand, Figure 2 also shows that rich nations such as Fra-
small population have gained less, or even lost productivity, nce, the United States, and Japan have persistent gains with
as witnessed by the results of Ireland, Jamaica, Costa Rica, allocative efficiency.
Jordan, Trinidad and Tobago, and Iceland.
The estimated model produces scores that reflect the
levels of technical efficiency of these nations, but not levels 5. The Role of Technical Progress and
of allocative efficiency. The effects of allocative efficiency are Allocative Efficiency in Economic Growth
only evaluated in dynamic terms and reflect either an approx-
imation or a departure of the value of the estimated shares We will now take a closer look at the differences in eco-
of income factors (λK and λL ) from their competitive values nomic growth patterns of developed and developing nations.
(i.e., factor remuneration from its marginal products). As Table 4 and Figure 3 bring results on the sources of growth
shown in Table 3, countries that had the largest allocative for those two groups of countries. The group of developed
efficiency gains were Austria, Finland, Portugal, France, nations consists of OECD member countries except Mexico,
Belgium, the United States, and Japan. At the other end are Greece, and Turkey, which are in the developing nations
countries that lost out with the dynamics of factor allocation. group. This last group includes, in addition to the above
Most of the Latin American countries fall within this group, three, all other countries in the sample (total of 36 countries).
as well as South Korea (until 1985) and Thailand. Some of Table 4 displays annual averages for each five-year period and
OECD’s poorest members are also among those countries also for the whole thirty-year period.
that had poor performance in allocative efficiency terms, We see that developing nations grew more than devel-
such as Greece and Turkey. oped ones (18.2%). This happened because both capital
We see systematic gains with factor allocation in richer accumulation and labor expansion were larger in developing
economies and losses (or very modest gains) in poorer ones. countries. However, the growth of GDP per worker was
It is interesting to point out that the differences between greater in developed countries, which can be attributed basi-
these two groups of countries regarding changes in allocative cally to two factors: (i) the difference between the growth
efficiency are even more marked in the first three five-year rates of capital and labor was greater in developed nations,
periods of the sample. It is well known that both Brazil and thus providing higher growth of capital per worker; (ii) the
Thailand decided on a strategy of “growth without adjust- change in TFP in developed nations was considerably higher
ment” in response to the oil shock of 1973, with increasing than in developing ones (yet it should be said that in the sec-
debt during this time. In Brazil the II National Development ond group this change pushed down GDP’s growth). The dif-
Plan was being implemented and government played an ferences between the two groups in regard to growth of cap-
important role in resource allocation in the economy and was ital per worker are well below the differences in TFP growth.
responsible for large infrastructure investments. The impor- This suggests that productivity plays a role of great impor-
tance of government in resource allocation is also a charac- tance in the development of nations, better yet, that it might
teristic of South Korea during the early years in the sample. explain a significant part of the differences in GDP per cap-
Figure 2 shows the evolution of total factor productivity ita growth between rich and poor countries.
change in six economies, calculated in two ways: (i) with If we take a look at the relative importance of the com-
allocative efficiency and (ii) without this component. The ponents of productivity, we see that developed nations have
first aspect to be highlighted is the distinct patterns of behav- some advantages, even if minor, in regard to technical effici-
ior displayed by developed and developing nations. France, ency. On the other hand, we also see that this difference is
the United States, and Japan present dynamic gains with in part offset by positive scale effects enjoyed by developing
resources allocation, and for them TFP computing allocative countries. Judging by the magnitude of the differences
10 Economics Research International

Brazil Japan
0.08 0.18

0.06 0.16

0.14
0.04
0.12
0.02
0.1
0
0.08
−0.02 0.06

−0.04 0.04
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

Mexico France
0.04 0.14

0.03 0.12

0.02 0.1

0.01 0.08

0 0.06

−0.01 0.04

−0.02 0.02

−0.03 0
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

Korea USA
0.15 0.12

0.1
0.1

0.08
0.5
0.06
0
0.04

−0.05
0.02

−0.1 0
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

With allocative efficiency With allocative efficiency

Without allocative efficiency Without allocative efficiency

Figure 2: Total factor productivity change, with and without allocative efficiency.

between the groups of countries regarding the pace of sizable 12.5% in allocative efficiency improvement, at the
technical progress and the evolution of allocative efficiency, same time that in poor countries this variable fell 11%. Here,
we are able to conclude that these two components explain we have an accumulated difference of 26.4% in this com-
most of the differences in productivity existing between the ponent, which places this figure at the forefront in explaining
two groups. the differences in productivity among the two groups of
While developed nations enjoyed technical progress of countries, and consequently the differences in the rates of
7.2% in the 30 years analyzed here, developing countries in output growth.
fact suffered a 9.8% drop in that component, a gap that adds Lower rates of growth of output per worker in develop-
up to 18.8%. We also notice that rich countries accumulated ing nations in comparison with developed ones lead to
Economics Research International 11

Table 4: Sources of economic growth per group of countries and subperiods: % change.

Group of Annual averages in the sub-periods Annual

Variable Accumulated
countries∗ 1966–1970 1971–1975 1976–1980 1981–1985 1986–1990 1991–1995 1996–2000 average
Developed 5.33 3.92 3.44 2.30 3.62 2.04 3.57 4.04 228.22
GDP growth Developing 5.29 5.14 5.64 1.93 3.38 4.12 2.90 4.74 301.10
Difference 0.03 −1.16 −2.08 0.37 0.23 −2.00 0.65 −0.67 −18.17
Developed 5.58 5.84 4.34 3.14 2.99 2.44 2.53 4.48 272.60
Capital
accumulation Developing 6.54 6.72 6.59 5.04 3.85 4.13 3.59 6.09 489.67
Difference −0.90 −0.83 −2.11 −1.81 −0.82 −1.63 −1.02 −1.52 −36.81
Developed 0.97 1.19 0.89 0.77 0.59 0.85 0.65 0.98 34.14
Labor
expansion Developing 2.95 2.82 2.81 2.68 2.41 2.15 1.91 2.96 139.92
Difference −1.93 −1.59 −1.87 −1.86 −1.78 −1.27 −1.24 −1.92 −44.09
Change in Developed 4.32 2.70 2.53 1.52 3.01 1.17 2.91 3.03 144.68
GDP per Developing 2.27 2.25 2.76 −0.73 0.95 1.92 0.98 1.73 67.18
worker Difference 2.00 0.43 −0.22 2.27 2.04 −0.74 1.91 1.28 46.36
Change in Developed 4.57 4.59 3.42 2.35 2.39 1.57 1.87 3.46 177.76
capital per Developing 3.49 3.79 3.68 2.30 1.40 1.94 1.65 3.04 145.78
worker Difference 1.05 0.77 −0.25 0.05 0.98 −0.36 0.22 0.41 13.02
Developed 1.32 1.56 1.34 1.04 0.97 0.68 0.59 1.25 45.14
Change in
Developing 0.07 − 0.10 − 0.23 − 0.11 − 0.16 −0.35 − 0.37 −0.21 − 6.11
TFP
Difference 1.25 1.66 1.58 1.15 1.14 1.03 0.97 1.46 54.58
Developed 0.54 0.44 0.33 0.21 0.09 −0.04 −0.17 0.23 7.22
Technical
Developing 0.04 −0.06 −0.16 −0.28 −0.41 −0.53 −0.66 − 0.34 −9.76
progress
Difference 0.50 0.50 0.49 0.49 0.50 0.49 0.49 0.58 18.82
Change in Developed 0.56 0.53 0.49 0.46 0.43 0.40 0.38 0.54 17.63
technical Developing 0.43 0.40 0.37 0.35 0.33 0.31 0.29 0.41 13.13
efficiency Difference 0.13 0.13 0.12 0.11 0.10 0.10 0.09 0.13 3.98
Change in Developed 0.05 0.06 0.06 0.06 0.07 0.07 0.07 0.07 2.25
scale Developing 0.01 0.07 0.10 0.10 0.13 0.14 0.14 0.12 3.53
efficiency Difference 0.04 0.00 −0.04 −0.04 −0.07 −0.07 −0.07 −0.04 −1.24
Change in Developed 0.17 0.52 0.45 0.30 0.38 0.24 0.31 0.39 12.50
allocative Developing −0.40 −0.50 −0.54 −0.28 −0.21 −0.26 −0.14 −0.39 −10.99
efficiency Difference 0.57 1.02 1.00 0.59 0.59 0.49 0.45 0.78 26.40
∗
The values in this table were calculated by taking simple arithmetic averages over the countries comprising each group of the rates of change for each
subperiod. The accumulated effects as well as the annual average numbers were computed by compounding (or discounting) rates. The relative factor prices
and relative marginal productivity of each period used in our calculations refers to the last year of the 5-year interval; for example, the values of sK , sL , λK ,
and λL used to estimate the allocative efficiency changes between 1966 and 1970 refer to 1970.

divergence between the standards of living of the two the economies with better governance in 1998 enjoyed less
groups. In light of this, common aspects among countries distortions and consequently greater allocative gains between
having similar growth patterns (and similar behavior for 1995 and 2000 as depicted in Figure 4(a). Figure 4(b) plots
the difference between the two measures of TFP mentioned the allocative efficiency measure against the openness index
before—with and without allocative efficiency) should be mentioned (for 33 of the 36 economies) from 1970 to 2000.
sought and their motivation explored. The remarks made Here, we have five-year changes in these two measures, and
at end of Section 4 are speculations on this topic and relate we can see that reduction in barriers to capital flow seems to
these common aspects to policy. be associated with improvements in allocative efficiency.
The allocative nature of policy effects is captured in a
rather informal way in Figure 4, where the measure of alloca-
tive efficiency estimated here is plotted against two indi- 6. SFA and Recent Issues in the
cators (that, hopefully, do reflect policy). One is the Gover- Economic Growth Literature
nance Index developed by Kaufmann et al. [24] and the other
is the degree of capital account openness, according to the Two excerpts from contemporary remarks made by Robert
measure suggested by Santana and Garcia [25]. As expected, Solow reveal that much remains to be clarified regarding
12 Economics Research International

0.25 0.25

Contribution of factor accumulation

0.2 0.2

Contribution of TFP
0.15 0.15

0.1 0.1

0.05 0.05

0 0

−0.05 −0.05

−0.1 −0.1
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

0.08 0.08

0.06 0.06

Technical efficiency
Technical progress

0.04 0.04

0.02 0.02

0 0

−0.02 −0.02

−0.04 −0.04
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

0.08 0.08

0.06 0.06
Allocative efficiency

0.04 0.04
Scale efficiency

0.02 0.02

0 0

−0.02 −0.02

−0.04 −0.04
1970 1975 1980 1985 1990 1995 2000 1970 1975 1980 1985 1990 1995 2000

Developing countries Developing countries

Developed countries Developed countries

Figure 3: Sources of growth by group of countries (% changes).

the determinants of economic growth and their relative originates in people and processes that are not
importance: usually connected with R&D,”
Solow [26],
“. . .Bits of experience and conversation have sug- “. . .the nontechnological sources of differences in
gested to me that it may be a mistake to think TFP may be more important than the technologi-
of R&D as the only ultimate source of growth cal ones. Indeed they may control the technological
in total factor productivity. I do not doubt that ones, especially in developing countries,”
it is the largest ultimate source. But there seems
to be a lot of productivity improvement that Solow [27],
Economics Research International 13

0.04 The evidence presented here also seems to corroborate

SWE
AUTGBRFIN
the second statement made by Solow. As argued in the previ-
0.02
BELFRA USA IRL DNKCHE
PRT ISL NOR ous section, allocative efficiency is, among the components
CAN NLD
JPN ESP
JAM TTO ITA
AUS
of TFP, the one that most contributes to the gap between
VEN BOL CRI the two groups of nations (with respect to income per
Allocative efficiency

0 NZL
KEN PER
JORKOR worker and TFP)—roughly about 60% of these differences.
COL MEXBRA CHL
The pattern of technical progress is behind the other 40%
−0.02
of the gap. These facts suggest that economic policies that
directly affect factor allocation are extremely relevant in ex-
−0.04 GRC plaining the differences in the growth performance between
developed and developing countries.
−0.06
TUR THA Regarding the changes in productivity associated with
technological diffusion and inefficiency reduction, the results
of this article allow us to identify the importance of the
−0.08
0.2 0.4 0.6 0.8 1 1.2 increase in technical efficiency estimated by the stochastic
Governance index frontier model, which contributes both to the growth
of developed nations as well as developing ones. This
(a)
perception seems to be in part shared by Robert Solow.
0.1 The first statement quoted above goes on to identify and
briefly describe the nontechnological sources of productivity
mentioned (see Solow [26], page. 8). The examples given by
0.05 him are typical of what the production frontier literature calls
efficiency improvement, both technical and allocative. Since
Allocative efficiency

he does not consider the explicit possibility of inefficiency

in his thinking, Solow seems to consider these phenomena
0 some sort of innovation, yet not related to R&D expenses
(design, marketing, etc.). However, we clearly see that he feels
that technical progress leveraged by R&D spending is not the
−0.05 only driver of productivity.
Recently, Easterly and Levine [28] added fuel to the exist-
ing controversy among the scholars currently discussing the
−0.1
“sources of economic growth,” on the relative importance of
0 0.2 0.4 0.6 0.8 1 factor accumulation and productivity. The underlying ob-
Capital account openness jective of their work is to demonstrate that, unlike what is
preached by the “neoclassical revival,” the focus of investi-
(b)
gation of economic growth should be productivity and its
Figure 4: Governance, financial liberalization and allocative effi- determinants. The terminology neoclassical revival is due to
ciency. Sources: Kaufmann et al. [24] for the governance index; Alwyn Young. It is used by Klenow and Rodrı́guez-Clare [29]
Santana and Garcia [25] for openness index; and own estimations to qualify a body of studies that tries to counter the new
of allocative efficiency. growth theory and is associated to the hypothesis that the
differences among nations in levels and in per-capita income
The results presented in the previous sections readily change are caused by differences in physical and human
provide information to clarify the apparent contradiction capital accumulation. Some examples of this line of work
between these two statements. Even if restricted to a relatively are Mankiw et al. [30], Young [31, 32] and Barro and Sala-
small sample of countries, the results presented in Table 4 i-Martin [33]. Easterly and Levine [28] list five stylized facts
reveal that, in fact, for developed nations, technical progress regarding economic growth to underpin their idea. Some of
and technical efficiency changes are responsible for the larger their findings are corroborated by the results of this study,
part of TFP change accumulated in the last 30 years: of the but others are not.
45.1 percentage points increase in TFP, 26.1 can be attributed The first stylized fact presented by these authors states
to the joint effect of these two components (i.e., around 58% that differences in TFP growth explain the differences among
of all change). Yet for developing nations, for which TFP the various countries in per-capita income levels and per-
had a 6.1 percentage points decrease during the same period, capita income growth rates. Although factor accumulation
the component that contributed the most to this result is may be important to trigger growth and be responsible for a
allocative efficiency change, which reduced productivity in sizable share of this growth in a number of countries, it is not
nearly 11 percentage points. Together, technical progress and able to explain the differences in level of income or in rates
changes in technical efficiency contributed with a small accu- of income change among nations. In relation to this fact, it
mulated growth of 2.1 percentage points, an unsatisfactory should be pointed out, first of all, that the great importance
performance which is at least in part due to relatively small of capital accumulation in the countries’ growth rate also ap-
investments in R&D made by poor nations. pears in the results for the reduced 36-country sample. In
14 Economics Research International

fact, we find that 80% of the growth would be attributable 0.02

to the accumulation of capital and labor, and only the
0.015

Average of random error component

20% remaining would come from productivity gains. The
results vary when we calculate separately the average for 0.01
the group of 24 developed countries (factor accumulation
is lower, close to 63% of the growth) and for the group 0.005
of 12 developing nations (contribution to productivity is
0
negative and, therefore, the factor accumulation is behind all
the economic growth). −0.005
If the reference is output per worker, the importance of
capital accumulation remains high. With some additional −0.01
calculations based on the numbers listed in Table 4, we
−0.015
conclude that on average 62.1% of the GDP growth is due
to capital accumulation. Klenow and Rodrı́guez-Clare [29] −0.02
reach a similar result for a sample of 98 countries: on average 1970 1975 1980 1985 1990 1995 2000
70% of economic growth is the result of physical and human
Developed
capital accumulation. The comparison is obviously limit-
Developing
ed, because human capital is not considered in this study and
because the samples are different. Nonetheless, the reduced Figure 5: Average random error component by group of countries,
sample used in this article contains 24 of the 30 OECD 1970 to 2000.
members, while the sample used by Klenow and Rodrı́-
guez-Clare [29] contains all of them. Consequently, we can tent. A consequence of this is that productivity measures
conclude that most of the nations that differentiate the two would necessarily have a volatile behavior. This could be
samples are developing economies, which generally have a avoided if production data to be explained reflected the pot-
higher share of capital in income. Thus, the inclusion of these ential, rather than the actual output.
countries would tend to raise the participation of factors in Robert Solow (Solow [26]) argues that growth theory is
the growth of GDP per worker (above 62.1%), bringing the a theory of the evolution of potential product. This is justi-
results of the two studies closer to each other. fied by the fact that the countries’ growth paths do not
Regarding income per capita, the difference between rich resemble at all the concept of steady state. In economies
and poor nations is the second stylized fact pointed out by where agriculture has a considerable weight, sudden weather
Easterly and Levine [28]. According to them, this phenome- changes or pests can bias the traditional TFP measure. Con-
non is not very consistent with the analytical apparatus that sequently, either we work with potential output as a depen-
emphasizes factor accumulation with diminishing returns dent variable or we add explanatory variables that control for
and lack of economies of scale. It would be more appropriate weather changes or pests. Demand fluctuations are another
to emphasize productivity growth based on technology and source of deviation of output from its balanced growth
increasing returns. Klenow [34] argues, however, that insti- path. If we return to Figure 3 and examine the evolution
tutional frameworks (such as tax structure, protectionism, of productivity change, we see that it has an absolutely ser-
lack of property rights,) may reduce the accumulation of ene behavior. Here probably lies the greatest contribution of
physical and human capital. In line with the ideas suggested the approach combining stochastic production frontier esti-
by Easterly and Levine [28], the results of this article reject mation and the Bauer-Kumbhakar decomposition: it allows
an interpretation based on factor accumulation for this dis- us to separate the effects of random shocks from the other
crepancy. TFP components. (In fact, it is even possible to evaluate if the
The divergence between developed and developing na- assumptions of a normal truncated distribution for the tech-
tions was one of the results found using the empirical model nical efficiency component and of normal distribution with
applied in this study. Moreover, it is clear that the differ- zero mean for the random component describe well the be-
ences in the rates of productivity change are behind all the havior of the observed data. The analysis of the residuals
differences in the rates of growth of GDP per worker (the obtained in our estimations reveals that the presumption
accumulation of factors contributed towards reducing such of normal distribution with zero mean seems to suit the
differences). Note that this result was obtained within the data well.) All the other TFP components have clear trends,
traditional framework of an aggregated production function with little fluctuation, except perhaps for allocative efficiency,
with diminishing returns. It was not necessary to incorporate which responds to policies.
in the analysis a new sector (a knowledge production sector In evaluating the random shocks obtained as a residual
presenting increasing returns). and depicted in Figure 5, one can easily provide an interpre-
The third stylized factor in the list of Easterly and Levine tation for their behavior in terms of demand fluctuations
[28] suggests that the accumulation of factors is persistent, following unexpected events. We can associate, for instance,
at the same time that economic growth is not. Considering the downturn in the second half of the seventies to the oil
that changes in the rate of growth depend both on changes in crises of 1974 and 1979 and the long recession period of the
factor accumulation as well as on changes in productivity, the eighties in developing countries to the debt crisis following
validity of this stylized fact implies that TFP cannot be persis- the Mexican default in 1982.
Economics Research International 15

The fourth stylized fact points out that production fac- The output variable used is GDP measured at constant
tors tend to flow towards the same direction and as a conse- prices (1996 US$), with purchasing power parity (PPP)
quence economic activity is quite concentrated. This is valid adjustment. It is obtained by taking the real GDP per capita
not only among countries but also within them (regions, chain series (RGDPPCH) from PWT 6.1 and multiplying it
states, and cities). If there were no productivity differences, by total population for each country.
the trend would be exactly the opposite, that is, that of an With respect to labor (L), we use a proxy, the population
even distribution of factors among the various countries, of equivalent adults (peqa), obtained from PWT. The concept
because of the presence of decreasing returns. Differences derives from population data: based on data for the total
in policies could explain factor accumulation (regulation, population (pop), an average is computed that attributes a
tax structure, legal systems, public education, etc.). However, weight of 1 to people older than 15 (pop15+ ) and 0.5 to
usually these policies have a nationwide scope and would not people aged up to 15 (pop15− ). That is, pequa = (pop15+ ) ×
be helpful in explaining concentration within the nations. 1 + (pop15− ) × 0.5. These data are obtained indirectly
Easterly and Levine [28] do not provide a single explanation from the PWT 6.1, by performing calculations using three
for this phenomenon and argue that such stylized fact is variables: real GDP per capita chain series (rgdpch) was
consistent with existing explanations in terms of poverty divided by real GDP per equivalent adult (rgdpeqa) and
traps, intragroup factors, or geographical externalities and then multiplied by the population (pop), that is, L =
is also consistent with explanations based on differences of (rgd pch/rgd peqa) · pop = (GDP/ pop) · (peqa/GDP) · pop.
productivity caused by technological differences. Another possibility would be to use data pertaining to the
The results of this article have shown that developing labor force. These can be obtained through a transformation
nations accumulate production factors at a much faster pace similar to the one described above, using the variable real
than that of developed nations and for this reason also grow GDP per worker (rgdpwok). A country-by-country detailed
faster. The model used presents a measure of scale effects for analysis of the two series suggests that peqa is more reliable,
the sample countries that is intuitive but not fully consistent which was the motivation of our choice.
with the notion of concentration of economic activity. The perpetual inventory method was used to compute a
Although the estimated measure of scale effects for develop- series for the stock of capital of each nation in the sample.
ing economies came up suggesting increasing returns to scale This method uses an initial capital stock estimate (com-
(for India, Indonesia, Brazil, and Mexico, to name a few), the puted from investment data), the supposition of a stable rate
magnitude of these effects is not up to the task of explaining of growth for a given period, and additional suppositions
the fourth stylized fact identified by Easterly and Levine [28]. regarding the depreciation rate. The measure of the initial
The fifth and last stylized fact states that policies capital stock is quite sensitive to the problems of measure-
implemented by nations have a relevant impact on long- ment error regarding the flow of investment (and also the
term growth rates of these nations. The authors try to show growth of GDP).
that variables related to policy decisions of nationwide The investment series used in computing the capital stock
scope, such as education, degree of trade and financial was obtained from multiplying the GDP, in constant 1996
liberalization, and size of the government, among other local currency, by the “current” investment rate, and then
factors, are related to countries’ growth rates and to TFP. This converting this result to US$ using the 1996 exchange rate.
is consistent with our results since changes in government GDP in 1996 local currency units was obtained by simply
policy have fundamental impacts on allocative efficiency. adding up all its components, which are available in the
We believe to have shown the great importance of allocative nafinalpwt spreadsheet of the PWT. The current investment
efficiency change in productivity change, and consequently rate was obtained dividing the value of investment in current
in growth rate differences. local currency by the current GDP. The exchange rate used
With a clear economic interpretation and the advantage is obtained from the series XRAT, found in the nafinalpwt
of separating random shocks from the regular behavior of the spreadsheet of the PWT 6.1.
economies, the stochastic frontier approach combined with The initial capital stock is computed using the investment
flexible functional forms (for production frontiers) and the series. To do so, we took as the reference year, the year
TFP decomposition described here seems to be a promising following that of the start of the investment series. We then
way of looking at aggregate productivity, one up to the task used the perpetual inventory method to build up the remain-
of providing a broad range of explanations in the field of der of the series. This procedure allowed each country to
economic growth. have its own capital stock series beginning in the first year
for which we have available data for aggregated investment.
The capital stock series used in this study was not
Appendices adjusted for purchasing power parity disparities. More speci-
fically, it is taken in constant 1996 US$. This reflects the per-
A. Data and Sample ception that investment decisions are taken considering rela-
tive domestic prices. Cohen and Soto [35] also notice this and
Below, we detail the definitions of each series used in the argue that PPP adjustment imposes on poorer countries rela-
econometric estimations. We also describe the procedures tive prices that are different from those of the market, and an
used in selecting the countries and the time periods that apparently high marginal productivity of capital. The price of
actually comprise the econometric estimations. investment goods has been decreasing over time in relation
16 Economics Research International

to the price of other products, a trend that has become more Of the remaining 112 economies, other 13 were excluded
evident with the growing production of the information because of lapses in the historical series caused by wars,
technology and communications industries. The quality of civil wars, or splitups. In these cases, the estimation of capi-
the products in these two industries has undoubtedly been tal stock using the perpetual inventory method can clearly
improving, with prices continually dropping and capital use not be applied. The countries rejected due to this criterion
continually increasing. The consequence of this is that the were the following: Angola, Ethiopia, Bangladesh, Gui-
importance of factor accumulation in the explanation of nea, Comoros, Haiti, Burundi, Central African Republic,
economic growth is increasing, making the part relative to Madagascar, Mozambique, Sierra Leone, Papua New Guinea,
productivity smaller. Once capital stock values undergo PPP and Congo (formerly Zaire). Eighteen other nations were
adjustment, these effects are exacerbated. excluded because of having highly volatile GDP per capita
Factor shares sK and sL were basically obtained from two and investment rate figures, which causes excessively high
databases: (i) Annual National Accounts from OECD, which deviations in the capital stock estimations (namely, Algeria,
brings information from 1970 to 2000 for 30 members of Benin, Botswana, Burkina Faso, Cameroon, Congo, Cote
that organization; and (ii) the System of National Accounts D’Ivoire, Fiji, Mauritius, Gambia, Guinea-Bissau, Guyana,
1968 (SNA68) from the United Nations. For OECD nations Mali, Mauritania, Namibia, Niger, Tanzania, and Togo).
belonging to the sample in this study, we have used only Note that all countries included in this last group are
this organization’s database (it is homogeneous and contains poor, most of them from Africa. A question could be raised
more information than the SNA, some of them estimates, here, arguing that this decision would create a biased analysis
though). Information pertaining to non-OECD countries through selection. We argue that this is not a problem, be-
were obtained mostly from SNA68. cause the purpose here is to describe a quite flexible produc-
Data for some countries were not available in SNA68 tion frontier (translog): in this case, output elasticities with
(usually those relative to the first and the last years of the respect to the productive factors can vary among countries
sample). For these countries, we tried other sources. Among and in time, which renders flexibility to the adjustments. In
them, we can name the Economic Commission for Latin the event we undertook an analysis using the Cobb-Douglas
America and the Caribbean (ECLAC) for data pertaining technology, elasticities would be constant and would express
to Bolivia (2000), Costa Rica (2000), Trinidad and Tobago sample averages subject to selection bias. In this analysis, the
(2000), Jamaica and Peru (1995 and 2000), and MIDEPLAN selection should actually favor more precise estimations, as
(Ministerio de Planificación y Cooperación) for data of Chile the excluded economies generally have a low “grade” in the
(1975 to 1985 and 2000). For Bolivia and Costa Rica the ranking provided by the PWT in regard to data quality (see
numbers for 2000 are actually those of 1999 (the closest Heston et al. [36], Table A, p. 13).
available). For Chile, there was no available information This leaves us with 75 countries with data spanning from
for 1970 in any sources used. We used then the numbers 1950 to 2000. The observations were taken for 11 different
for 1973, first year for which the national accounts of this time periods, every 5 years, starting in 1950 and finishing in
country display that information. For Brazil, data used are 2000. This type of procedure is rather common in the econo-
from the local official statistical bureau, Instituto Brasileiro mic growth literature and is justified by the interest of study-
de Geografia e Estatı́stica (IBGE). ing long-term effects, which can be better addressed by more
The selection of countries included in the sample spaced time observations. Forbes [37], to give one example,
followed some criteria. The first and obvious criterion was makes estimations with data gathered every five years and
availability of homogeneous data for the period in question. justifies this saying that yearly data contain short-term dis-
Nations that had a reduced number of observations were ex- turbances.
cluded. A minimum of 30 continuous observations per Before proceeding to the estimations, the data were care-
country was set. Therefore, of the 203 economies listed in fully reviewed, on a country per country basis. Special care
the PWT 6.1, 86 countries that did not have information on was taken with the series for capital stock. It is known that
either the labor force, GDP, investment, or exchange rate for estimations for initial capital stock can present problems
the last 30 years were excluded. This criterion essentially re- that render less reliable capital stock numbers for the first
moved from the sample a number of countries created or years of the series. We must remember that the initial capi-
split in the last 20 to 30 years. tal stock calculations presume a stable behavior for the eco-
Previously socialist economies, such as People’s Republic nomic growth rate (steady state), an assumption not very
of China, Hungary, Romania and Poland, or those nations realistic. In the event the growth rate for the initial period
that are protectorates of others, such as Puerto Rico and is too low (much lower than that of steady state), the initial
Taiwan (Hong Kong was kept in the sample, though), were capital stock tends to be overestimated, and consequently the
also excluded. The group of 86 excluded nations also com- numbers appear to be too small in the initial periods. The
prises those with a very small population—less than 500 opposite can occur when the rate is high. Based on scatter
thousand inhabitants in 2000. For this reason, countries like plots (capital x GDP), we noticed the presence of observa-
Barbados, Cape Verde, Equatorial Guinea, Luxembourg, and tions for some countries that could suggest inadequate esti-
Seychelles Islands were also left out. The only exception to mations of initial capital stock. This was the case of the fol-
this last rule was Iceland, a country with “good quality” in- lowing countries: Argentina, Australia, Denmark, Iceland,
formation dating back to 1950. The Netherlands, New Zealand, and Syria. Therefore, the first
Economics Research International 17

Table 5: Likelihood ratio tests.

Full Cobb-
Harrod Solow Hicks Translog
Model Translog Douglas
Neutral Neutral Neutral without TP
All βs =/0 with TP
Harrod Neutral 19.69 — — — — —
βKt = 0 χ 2 (1)
Solow Neutral 4.55 NC — — — —
βLt = 0 χ 2 (1)
Hicks Neutral 19.71 0.01 15.15 — — —
βKt = βLt = 0 χ 2 (2) χ 2 (1) χ 2 (1)
Translog without TP 33.43 13.74 28.88 13.72 — —
βt = βtt = βKt = βLt = 0 χ 2 (4) χ 2 (3) χ 2 (3) χ 2 (2)
Cobb-Douglas with TP 110.01 90.32 105.46 90.31 NC —
βtt = βKt = βLt = βKL = βKK = βLL = 0 χ 2 (6) χ 2 (5) χ 2 (5) χ 2 (4)
Cobb-Douglas without TP 152.21 132.51 147.65 132.50 118.78 42.20
βt = βtt = βKt = βLt = βKL = βKK = βLL = 0 χ 2 (7) χ 2 (6) χ 2 (6) χ 2 (5) χ 2 (3) χ 2 (1)
NC = not comparable; TP = technical progress.

two (or three) observations of the series pertaining to these cannot reject the hypothesis that the Harrod Neutral model
countries were eliminated. is nested in the Full Translog model).
A similar problem occurred for some countries when
scatter plots of production and population of equivalent
adults were analyzed. Ireland, Greece, and Cyprus experi- C. Alternative Estimations
enced, at different moments, considerable reductions in their
population of equivalent adults, presenting a behavior not In addition to the analysis of production frontiers using data
compatible with the premise of factors diminishing returns. at five-year intervals, two other experiments were carried out
For the first two nations, this happened at the beginning to evaluate the relative performance of the model: (i) the
of the series, a fact that could indicate problems with estimation of the stochastic frontier model using annual data,
different sources for population data (up to 1960 the and (ii) the estimation of traditional panel data models (fixed
PWT’s population data come from the United Nations Deve- effects and random effects). The results are presented below.
lopment Centre and after this year they come from the World Regarding the first experiment, it can be said that five-
Bank). Consequently, we decided to exclude the first two year interval data yield better results than annual ones, as
observations of the series for Ireland (1950 and 1955) and expected. In spite of being valid on the whole, the annual
the first (1955) for Greece. model generates nonsignificant coefficients associated with
time, capital stock, and the labor force: p(z) = 16.7%, 19.0%,
and 24.5%, respectively. Moreover, the model’s total variance
B. Likelihood Ratio (LR) Tests is larger, given the existence of short-term variability in
output (0.077 as opposed to 0.061 in the model with five-
The likelihood ratio statistic is given by λ = −2[L R − year data). The average technical inefficiency, given by μ,
L  R and L
 NR ], where L  NR are, respectively, the estimated log- is relatively high (0.249 compared to 0.207 in the five-year
likelihood of the described model and of the nonrestrict model), which explains the higher variance of technical
model. Table 5 summarizes the tests performed. The null efficiency and lower of technical progress (that turns out
hypothesis under question is always that the model identified negative for all countries after 1997). For this reason, the
in the matrix line is nested in the model of the matrix influence of μ in the total variance also rises, from 66.3% to
column. The λ statistic has a χ 2 (DF) distribution, where 77%. If the estimation of technical inefficiency was based on
DF is the difference in the degrees of freedom between the the Battese and Coelli [38] model, it might have been possible
models. If the value expressed in the cell of the matrix to control the effect of these short-term variations. There
below is greater than the critical value, then the null is also the possibility to follow the approach suggested by
hypothesis cannot be rejected, otherwise, it can be rejected. Cuesta [39], a task we intend to pursue in another enterprise.
For example, testing if the Harrod Neutral model is nested in The estimates produced by the traditional panel data
the Full Translog Model is the same as testing the restriction fixedeffects and randomeffects models turned up quite infe-
H0 : βKt = 0 in this last specification. Since the test statistic rior to those of the stochastic frontier models. The Hausman
for the comparison between the restricted and unrestricted test (χ 2 = 49, 03) favors the fixed effects model, although
model is λ = −2[262,224−272,071] = 19,69 and the critical 4 out the 9 coefficients of the translog specification ended
value at 1% is 6,64, we reject H0 (which in turn means that we up being not significant at 10%. Furthermore, the results
18 Economics Research International

are not intuitive at all. Some countries have zero or negative [15] C. Cornwell, P. Schmidt, and R. C. Sickles, “Production fron-
labor elasticities, such as Iceland and South Korea, and re- tiers with cross-sectional and time-series variation in effi-
turns to scale vary a lot: the United States, to give an exa- ciency levels,” Journal of Econometrics, vol. 46, no. 1-2, pp.
mple, would have an estimated RTS of 1.26, whereas Iceland 185–200, 1990.
would be a mere 0.49. The estimations of technical progress [16] Y. H. Lee and P. Schmidt, “A production frontier model with
likewise do not seem very reasonable: United States, Japan, flexible temporal variation in technical efficiency,” in The
and Germany show large technical regress at the same time Measurement of Productive Efficiency, H. Fried, C. A. K. Lovell,
and S. Schmidt, Eds., Oxford University Press, New York, NY,
that Trinidad and Tobago, Lesotho, and Jamaica have extra-
USA, 1993.
ordinary technical progress. These resultssuggest that fron-
[17] United Nations, “System of national accounts 1968,” CD-Rom,
tier models may be better suited for the analysis of produc-
2003.
tivity in comparison with traditional econometric methods.
[18] OECD, Annual National Accounts for OECD Member
Countries—Data from 1970 onwards. Table 3: GDP by Income,
2003.
References [19] M. J. Farrel, “The measurement of productive efficiency,”
Journal of the Royal Statistical Society A, vol. 120, pp. 253–290,
[1] R. Färe, S. Grosskopf, M. Norris, and Z. Zhang, “Productivity 1957.
growth, technical progress, and efficiency change in industrial- [20] N. Islam, “Different approaches to international comparison
ized countries,” American Economic Review, vol. 84, no. 1, pp. of total factor productivity,” in New Developments in Produc-
66–83, 1994. tivity Analysis, C. R. Hulten, E. R. Dean, and M. J. Harper, Eds.,
[2] E. Deliktas and M. Balcilar, “A comparative analysis of University of Chicago Press, Chicago, Ill, USA, 2001.
productivity growth, catch-up and convergence in transition [21] D. S. P. Rao and T. J. Coelli, “A cross-country analysis of GDP
economies,” in Proceedings of the 6th METU Conference on growth catch-up and convergence in productivity and ine-
Economics, Ankara, Turkey, September 2002. quality,” Working Paper 5/98, Centre for Efficiency and Pro-
[3] E. Deliktaş and M. Balcilar, “A comparative analysis of ductivity Analysis (CEPA), 1998.
productivity growth, catch-up, and convergence in transition [22] K. J. Arrow, “The economic implications of learning by doing,”
economies,” Emerging Markets Finance and Trade, vol. 41, no. Review of Economic Studies, vol. 29, pp. 155–173, 1962.
1, pp. 6–28, 2005. [23] S. Kim and G. Han, “A decomposition of total factor produc-
[4] P. W. Bauer, “Decomposing TFP growth in the presence of cost tivity growth in korean manufacturing industries: a stocha-
inefficiency, nonconstant returns to scale, and technological stic frontier approach,” Journal of Productivity Analysis, vol. 16,
progress,” Journal of Productivity Analysis, vol. 1, no. 4, pp. no. 3, pp. 269–281, 2001.
287–299, 1990. [24] D. Kaufmann, A. Kraay, and P. Zoido-Lobaton, “Governance
[5] S. C. Kumbhakar, “Estimation and decomposition of produc- matters,” World Bank Policy Research Working Paper 2196,
tivity change when production is not efficient,” Econometric 1999.
Reviews, vol. 19, pp. 425–460, 2000. [25] J. R. Santana and F. Garcia, “New evidence of the impact of
[6] N. Islam, “Growth empirics: a panel data approach,” Quarterly capital account liberalization on economic growth,” in Pro-
Journal of Economics, vol. 110, no. 4, pp. 1127–1170, 1995. ceedings of the Latin American Meeting of the Econometric Soci-
[7] R. Hall and C. Jones, “The productivity of nations,” NBER ety, Santiago, Chile, 2004.
Working Paper 5812, 1996. [26] R. M. Solow, “After technical progress and the aggregate pro-
[8] D. Aigner, C. A. K. Lovell, and P. Schmidt, “Formulation and duction function,” in New Developments in Productivity Analy-
estimation of stochastic frontier production function models,” sis, C. R. Hulten, E. R. Dean, and M. J. Harper, Eds., University
Journal of Econometrics, vol. 6, no. 1, pp. 21–37, 1977. of Chicago Press, Chicago, Ill, USA, 2001.
[9] W. Meeusen and J. van den Broeck, “Efficiency estimation [27] R. M. Solow, “Applying growth theory across countries,” World
from Cobb-Douglas production functions with composed Bank Economic Review, vol. 15, no. 2, pp. 283–288, 2001.
error,” International Economic Review, vol. 18, pp. 435–444, [28] W. Easterly and R. Levine, “It’s not factor accumulation:
1977. stylized facts and growth models,” World Bank Economic
Review, vol. 15, no. 2, pp. 177–219, 2001.
[10] M. M. Pitt and L. F. Lee, “The measurement and sources
[29] P. Klenow and A. Rodrı́guez-Clare, “The neoclassical revival
of technical inefficiency in the Indonesian weaving industry,”
in growth economics: has it gone too far?” in NBER Macroeco-
Journal of Development Economics, vol. 9, no. 1, pp. 43–64,
nomics Annual 1997, B. Bernanke and J. Rotemberg, Eds., pp.
1981.
73–114, MIT Press, 1997.
[11] P. Schmidt and R. Sickles, “Production frontiers and panel [30] N. G. Mankiw, D. Romer, and D. N. Weil, “A contribution to
data,” Journal of Business and Economic Statistics, vol. 2, no. the empirics of economic growth,” Quarterly Journal of Econo-
4, pp. 367–374, 1984. mics, vol. 107, pp. 407–437, 1992.
[12] G. E. Battese and T. J. Coelli, “Frontier production functions, [31] A. Young, “A tale of two cities: factor accumulation and tech-
technical efficiency and panel data: with application to paddy nical change in Hong Kong and Singapore,” in NBER Macro-
farmers in India,” Journal of Productivity Analysis, vol. 3, no. economics An-Nual 1992, O. J. Blanchard and S. Fischer, Eds.,
1-2, pp. 153–169, 1992. MIT Press, Cambridge, Mass, USA, 1992.
[13] S. C. Kumbhakar and C. A. K. Lovell, Stochastic Frontier [32] A. Young, “The tyranny of numbers: confronting the statistical
Analysis, Cambridge University Press, Cambridge, UK, 2000. realities of the East Asian growth experience,” Quarterly Jour-
[14] S. C. Kumbhakar, “Production frontiers, panel data, and time- nal of Economics, vol. 110, no. 3, pp. 641–680, 1995.
varying technical inefficiency,” Journal of Econometrics, vol. 46, [33] R. J. Barro and X. Sala-i-Martin, Economic Growth, McGraw-
no. 1-2, pp. 201–211, 1990. Hill, New York, NY, USA, 1995.
Economics Research International 19

[34] P. Klenow, “Comment on ’Its not factor accumulation: stylized

facts and growth models’,” World Bank Economic Review, vol.
15, no. 2, pp. 221–224, 2001.
[35] D. Cohen and M. Soto, “Why are poor countries poor? A
message of hope which in-volves the resolution of the Bec-
ker/Lucas paradox,” Manuscript (revised version of Dis-cus-
sion Paper 3528 of the Centre for Economic Policy Research,
2002), 2003.
[36] A. Heston, R. Summers, and B. Aten, Data Appendix for a
Space-Time System of National Accounts: Penn World Table Ver-
sion 6.1, Center for International Comparisons of Production,
Income and Prices at the University of Pennsylvania, 2002.
[37] K. J. Forbes, “A reassessment of the relationship between
inequality and growth,” American Economic Review, vol. 90,
no. 4, pp. 869–887, 2000.
[38] G. E. Battese and T. J. Coelli, “A model for technical inef-
ficiency effects in a stochastic frontier production function for
panel data,” Empirical Economics, vol. 20, no. 2, pp. 325–332,
1995.
[39] R. A. Cuesta, “A production model with firm-specific temporal
variation in technical inefficiency: with application to spanish
dairy farms,” Journal of Productivity Analysis, vol. 13, no. 2, pp.
139–158, 2001.