Trade barriers, openness, and economic growth.
Madsen, Jakob B.
1. Introduction
A recurring theme in international economics is the relationship
between openness and economic growth. Based on postwar data that
typically span the period from 1970 to 1990, there has, until recently,
been a consensus of a negative relationship between trade barriers and
growth and a positive relationship between growth and import
penetration. However, these findings have been challenged by Harrison
and Hanson (1999), Rodrik (1999), Yanikkaya (2003), and particularly,
Rodriguez and Rodrik (2000). (1) Rodriguez and Rodrik (2000) seriously
question the empirical method underlying the regression analysis in the
most important studies that find a positive relationship between
openness and growth. Rodriguez and Rodrik (2000) demonstrate that the
positive correlation between growth and openness found by Dollar (1992),
Ben-David (1993), Sachs and Warner (1995), and Edwards (1998) is not
robust to various measures of openness and important control variables.
Similarly, studies using pre World War II data consistently fail to
uncover a robust positive relationship between openness and growth (see
Bairoch 1972; Capie 1994; Foreman-Peck 1995; O'Rourke 2000;
Clements and Williamson 2001; Irwin 2002; Irwin and Tervio 2002;
Vamvakidis 2002). The empirical study of Vamvakidis (2002) is one of the
few studies that consider the relationship between openness and growth
over a long historical period. Using cross-section data over the periods
1870-1910, 1920-1940, 1950-1970, and 1970-1990, Vamvakidis (2002) finds
either a negative or no relationship between growth and openness before
1970 and a positive relationship thereafter.
A problem associated with most empirical studies is that
cross-sectional data, as opposed to panel data, are used. This prevents
them from controlling for fixed effects. More importantly, very little
attention has been given to growth versus level effects of openness and,
particularly, to the channel through which openness influences growth.
Endogenous growth theories have highlighted trade as the principal
channel through which knowledge is transmitted internationally (Grossman
and Helpman 1991). The early endogenous growth models have been
developed within the first generation endogenous growth framework, in
which the level of research and development (R&D) activity and
growth vary proportionally.
Since the seminal paper of Jones (1995), however, it has been
widely believed that first generation growth models are not consistent
with the empirical evidence. In response to Jones's critique,
endogenous growth theories have evolved into two distinct second
generation growth models, namely, semi-endogenous and Schumpeterian
growth models. Policies that seek to promote productivity have only
temporary growth effects in the semi-endogenous growth models of Jones
(1995) and Kortum (1997). In the Schumpeterian models of Aghion and
Howitt (1998) and Howitt (1999), R&D can have permanent growth
effects so long as R&D is increased along with income in the economy
to counteract the increasing product proliferation. To allow for this
possibility, knowledge spillovers have to be modeled following the
Schumpeterian framework.
The contribution of this article is twofold. First, an annual data
set for a panel of 16 relatively homogeneous industrialized countries,
which spans 137 years, is used to examine the productivity growth and
productivity level effects of trade barriers and import penetration. (2)
Because trade barriers and import penetration have fluctuated
substantially over the last 137 years, the data yield ample identifying
movements and, at the same time, enable one to control for country
characteristics. Furthermore, it is tested whether openness has
permanent or temporary output-growth effects.
Second, the article tests whether openness influences growth
because it enables countries to import knowledge that is produced in
other countries. Recent developments within endogenous growth theory
suggest that openness influences growth through the channel of imports
(Romer 1990, 1992; Grossman and Helpman 1991; Rivera-Batiz and Romer
1991; Aghion and Howitt 1998; Baldwin and Forslid 2000). Although some
studies have investigated the relationship between growth and knowledge
spillovers, very few, if any, have explicitly investigated the issue in
the context of openness. Domestic patent applications are used in this
article to construct domestic and foreign stocks of knowledge, and
bilateral trade shares are used to quantify trade-related spillover
effects. Spillover effects through the channel of imports follow the
theories described in Grossman and Helpman (1991), whereby productivity
is enhanced by imports of intermediate products that embody
technological knowledge. Similarly, in the model of Rivera-Batiz and
Romer (1991), countries can tap into world knowledge through the channel
of imports. It follows that imports that contain technological knowledge
may increase productivity in the importing country; whereas, imports of
products that do not embody technology might not influence growth at
all. A problem associated with the empirical estimates of these models,
following the seminal paper of Coe and Helpman (1995), is that knowledge
spillovers are assumed to have only level effects as opposed to
permanent growth effects along a balanced growth path. The possibility
that knowledge spillovers may have permanent growth effects following
the predictions of Schumpeterian growth theories is allowed for in the
empirical estimates in this article.
The literature is briefly surveyed in the next section, and the
empirical framework and the empirical estimates are presented in
sections 3 and 4. Sensitivity analysis is carried out in section 5, and
section 6 concludes the article.
2. Trade Barriers, Openness, and Growth
Why should openness impact positively on growth? The traditional
development literature considered exports as growth-enhancing because of
the positive productivity spillovers from the tradable to the
nontradable sector and because exports encourage more efficient
investment projects (Edwards 1993). The recent endogenous growth
literature has reoriented the argument as to how openness enhances
growth from focusing on exports to emphasizing imports of knowledge
(Romer 1990, 1992; Grossman and Helpman 1991; Rivera-Baltiz and Romer
1991; Baldwin and Forslid 2000). Barro and Sala-i-Martin (1995) argue
that imports give domestic producers access to a wider variety of
capital goods, thus effectively enlarging the efficiency of production.
The theories described in Grossman and Helpman (1991) suggest that
the quality of intermediate products positively influences the
efficiency of production. The new technology embodied in imported
intermediate products renders imported products more productive and,
therefore, increases labor productivity and total factor productivity
(TFP). As a consequence, trade will enhance growth only to the extent
that a country trades with research-intensive economies. The model of
Barro and Sala-i-Martin (1995, ch. 8) considers a two-country world,
where the technologically less advanced country taps into the knowledge
of the technologically more advanced country. Provided that the costs of
imitation are lower than the costs of innovation, the less advanced
country will catch up to the more advanced country.
Although most theories predict that growth is impeded by trade
barriers, some models predict that, under certain circumstances, trade
barriers may be good for growth (see, for instance, the discussion by
Rodriguez and Rodrik 2000). Grossman and Helpman (1991) and Matsuyama
(1992) show examples in which countries that are sufficiently far behind
the technological frontier may, through imports, be driven toward
production of traditional goods and, consequently, experience a lower
growth rate. A closely related argument is that the host country needs a
sufficiently high capacity to absorb the technology developed in the
technologically more advanced countries (see, for instance, Howitt
2000). These models underscore the importance of using a sample of
countries that are technologically not too far apart. The countries used
in this article are quite homogenous in terms of economic development,
length of schooling, and technological knowledge. We would, therefore,
expect the theoretical prior to go in the direction in which trade
barriers are bad for economic growth.
Empirically, some studies find a positive relationship between
growth and openness; whereas, others do not. The studies of Dollar
(1992), Ben-David (1993), Sachs and Warner (1995), Edwards (1998),
Vamvakidis (1998), and Frankel and Romer (1999) are well-known studies
that find a negative relationship between trade barriers and growth.
Studies that fail to find a negative relationship between trade barriers
and economic growth are the studies of Harrison and Hanson (1999),
Rodrik (1999), O'Rourke (2000), Rodriguez and Rodrik (2000), Irwin
(2002), Yanikkaya (2003), and, to some extent, Vamvakidis (2002).
Harrison (1996) and Rodriguez and Rodrik (2000) argue that the results
are sensitive to measurement of openness and inclusion of control
variables. Furthermore, Vamvakidis (2002) argues that most studies find
a positive relationship between growth and openness because the
estimates rely predominantly on post-1970 data. Vamvakidis (2002) shows
that the positive relationship between growth and openness is limited to
the post-1970 period, and that no such relationship can be found in
earlier data.
3. Empirical Framework
The empirical estimates in this article seek to be as inclusive as
possible by including important control variables and time dummies that
capture the effects of omitted variables that change by the same
magnitude across countries over time. Furthermore, in addition to
estimates covering the whole sample period, the estimation period is
decomposed into three subperiods to examine whether the nexus between
growth and openness is period specific. Instruments for openness are
used to overcome potential endogeneity problems and error-invariables
biases.
The following model is estimated for a panel of 16 Organisation for
Economic Cooperation and Development (OECD) countries:
[DELTA] 1n TF[P.sub.it]= [[alpha].sub.0] + [[alpha].sub.2]
[DELTA]O[p.sub.it] + [[alpha].sub.3] [DELTA] ln [S.sup.f.sub.it] +
[[alpha].sub.4] [DELTA] ln [S.sup.f.sub.it] + [[alpha].sub.5]
[(X/Q).sub.it] + [[alpha].sub.6] [(X/Q).sup.f.sub.it] (1)
+ [[alpha].sub.7] [m.sub.it] [(X/Q).sup.f.sub.it] + TD + CD +
[[epsilon].sub.1,it],
where TFP is total factor productivity, Op is trade openness,
[S.sup.d] is the domestic stock of knowledge, [S.sup.f] is the foreign
stock of knowledge spillovers through the channel of imports, X is the
productivity-adjusted innovative activity, Q is product variety,
[(X/Q).sup.d] is domestic research intensity, [(X/Q).sup.f] is foreign
research intensity spillovers through the channel of imports, m is
nominal imports of goods as the share of nominal GDP, CD is fixed effect
country dummies, TD is time dummies, [epsilon] is a disturbance term,
[DELTA] is a five-year difference operator, and the subscripts t and i
signify time and country, respectively. (3) Trade openness, Op, is
proxied by m or by the macro-tariff rate, Tr, which is measured as
nominal import duties divided by the nominal import values of goods.
Tariffs and the import penetration ratio are measured in decimal points.
See Madsen (2008c) for unit root tests of the variables involved in the
model, which reveal that all the variables in the model are stationary.
Also see Madsen (2008c) for methods used to estimate technology
spillovers through the channel of imports.
In addition to imports of research intensity, the interaction
between m and imports of research intensity has been added as a
regressor in Equation 1. Although [(X/Q).sup.f] consists of
import-weighted knowledge stocks, as shown below, these weights are
fractions that add up to one and, therefore, do not reflect openness.
[(X/Q).sup.f] is, therefore, multiplied by m to capture the role of
international trade. The more open an economy is, the easier it is to
tap into international knowledge. Note, however, that the stock of
foreign knowledge, [S.sup.f], is not multiplied by m because it is
already based on import weights that do not add up to one, as shown
below. A term that allows for the interaction between m and Sr is
included in the estimates in section 5.
The model is estimated in five-year differences to filter out
business cycle influences. Quah and Rauch (1990) find that short-run
cyclical fluctuations in import penetration and growth are highly
correlated without necessarily showing any structural relationship.
Inclusion of country dummies is not essential; the principal results are
insensitive to whether or not they are included (the results are
available from the author). Product variety is measured by employment
because Schumpeterian models assume that product varieties follow the
size of the population at the steady state. X is measured by the number
of domestic patent applications, following Madsen (2008a). Patents are
used because statistics on R&D expenditures or R&D workers are
available for only a couple of countries before World War II and because
the long historical R&D data are of bad quality.
Equation I is estimated using TFP, per capita output, Y/Pop, and
output per hour worked, Y/H, as dependent variables. The advantage of
using TFP as a dependent variable is that the growth effects of factor
accumulation have been taken into account in the TFP estimates. Thus,
TFP should, in principle, measure productivity. The advantages of using
per capita output and output per hour worked is that no theory-dictated
restrictions have been imposed on them and they are not influenced by
measurement errors of land and capital. Furthermore, output per hour
worked allows for TFP that has been derived under assumptions that may
not apply over the whole period, such as perfect competition and
constant returns to scale (see below for computation of TFP). Finally,
the growth in TFP and in labor productivity is the same along the
balanced growth path, provided that land is insignificant in production.
(4) Per capita output is also used as a dependent variable to compare
the results with those of most studies that measure productivity as per
capita output. Per capita output is the least useful productivity
measure of the three because it does not acknowledge the marked changes
in annual working hours and labor force participation rates that have
taken place over the past 137 years.
Equation 1 follows the predictions of second generation models of
economic growth extended to allow for international trade (see Madsen
2008a for derivation). In the semi-endogenous growth models developed by
Jones (1995), Kortum (1997), and Segerstrom (1998), a positive growth in
R&D inputs is required to maintain sustained growth in TFP because
of the assumption of diminishing returns to knowledge. This would
suggest that growth in innovative activity is the key variable in
explaining growth. Depreciation of the stock of knowledge is allowed for
in Equation 1, following Grossman and Helpman (1991) and Coe and Helpman
(1995), so that it is the change in net research activity that is
essential for growth. The Schumpeterian growth models of Aghion and
Howitt (1998), Young (1998), and Howitt (1999) assume that R&D
spreads more thinly across product varieties as the economy grows. To
ensure sustained TFP growth, R&D has to increase over time to
counteract the increasing range of products that lowers the productivity
effects of R&D activity. The normalization of R&D by product
varieties ensures that highly populated countries do not grow faster
than small countries. It also ensures that proportionally higher R&D
outlays, in terms of R&D expenditures, are required to maintain
growth as the economy becomes richer. The larger an economy is, in terms
of total output, the more likely it is that the R&D-induced product
lines for a representative firm are replications of R&D-induced
product lines of other firms. In other words, it is not the product
variety but product quality that generates growth in Schumpeterian
growth models.
Equation 1 is extended to allow knowledge spillovers through the
channel of imports following Coe and Helpman (1995) and Madsen (2008a).
Again the model distinguishes between the predictions of semi-endogenous
growth theories and Schumpeterian growth models. Following the early
endogenous growth literature, import penetration may influence growth
positively because it enhances the potential for an importing country to
tap into the world stock of knowledge (Romer 1990, 1992; Grossman and
Helpman 1991; Rivera-Batiz and Romer 1991). The theories described in
Grossman and Helpman (1991) suggest that the quality of intermediate
products positively influences the efficiency of production. The new
technology embodied in imported intermediate products increases
productivity. As a consequence trade will enhance growth only to the
extent that knowledge is embodied in the intermediate products that are
imported from elsewhere. Several papers have found that the foreign
stock of knowledge enhances TFP through the channel of imports (Coe and
Helpman 1995; Coe, Helpman, and Hoffmaister 1997; Engelbrecht 1997;
Lichtenberg and van Pottelsberghe de la Potterie 1998; Frantzen 2000;
Guellec and van Pottelsberghe de la Potterie 2001, 2004; del
Barrio-Castro, Lopez-Bazo, and Serrano-Domingo 2002; Lumenga-Neso,
Olarreaga, and Schiff 2005; Madsen 2007, 2008b). The method suggested by
Lichtenberg and van Pottelsberghe de la Potterie (1998) is used to
measure [S.sup.f.sub.it] (see Madsen 2008c for details).
The Coe-Helpman model has been extended to allow for research
intensity spillovers through the channel of imports by Madsen (2008a).
This allows for permanent growth effects of importing goods embodying
research intensity following the predictions of Schumpeterian growth
theory. This has important implications for the growth effects of
imports of knowledge. Although the importation of knowledge has only
level effects on productivity in the CoeHelpman model, imports of goods
embodying research intensity have permanent growth effects in the
Schumpeterain model. See Madsen (2008c) for the measurement of
[(X/Q).sup.f].
Tr (tariff rates) and m (import penetration) are used as proxies
for trade openness, noting that there is no universal measure of trade
openness. The theoretical literature gives more attention to the
relationship between trade policies and income growth rather than the
relationship between trade and growth (Yanikkaya 2003). Furthermore,
there is no clear consensus as to what represents openness or what is
meant by openness and trade liberalization (Yanikkaya 2003). Anderson
and Neary (1992) have constructed a trade restrictiveness index, which
incorporates tariffs and nontariff trade barriers. Unfortunately, the
data required to construct such a historical index are not available.
The literature has predominantly focused on m; however, Rodriguez and
Rodrik (2000) argue that m is not a good measure of trade barriers and
recommend the use of Tr instead. Although Tr specifically measures trade
barriers, m is influenced by several factors other than trade barriers
and, as such, may be a bad proxy for trade barriers. As discussed by
Edwards (1998), there are several ways in which m influences income,
and, in that sense, the coefficient of m may not be very informative
about causal factors.
Another issue is whether Tr and m are exogenous. The endogenous
tariff literature has long advocated that tariffs are an outcome of
endogenous forces (see, for example, Magee, Brock, and Young 1989).
Furthermore, the propensity to import is heavily influenced by economic
conditions. Madsen (2001) shows that the sharp reduction in world trade
during the Great Depression was a result of increasing tariffs and
nontariff trade barriers and decreasing income, which are all
endogenous. Furthermore, Frankel and Romer (1999) argue that countries
implementing free-market trade policies may also implement free-market
policies in addition to implementing stable monetary and fiscal
policies. The income effects of tariff policies in regression analysis
may, therefore, be disguised by the income effects of other policies
that are not controlled for in the regressions. Finally, Irwin (2002)
argues that the negative relationship between growth and tariff rates,
as often found in the growth literature, does not indicate anything
about causality. To address this issue Tr and m are instrumented based
on some of the instruments recommended by Trefler (1993) and Frankel and
Romer (1999). The choice of instruments is discussed in the following
subsection.
Equation 1 allows for the possibility that openness has permanent
direct growth effects through the variable Op (openness), following most
of the empirical literature on the nexus between growth and openness. In
any event, it cannot automatically be taken for granted that openness
has permanent growth effects. To satisfy the permanent growth-effects
criteria in endogenous growth models, foreign knowledge must
continuously produce a flow of ideas. In other words, there must be
constant returns to the imported foreign knowledge stock. However,
R&D knowledge spillovers through the channel of imports are already
accounted for in the estimates. Thus the residual knowledge spillovers
must stem from sources other than R&D or patents, such as human
capital, or other sources unaccounted for, such as increasing
competition and efficiency.
The estimates are corrected for serial correlation and
heteroscedasticity using feasible least squares. Furthermore, to gain
efficiency, the contemporary correlation between error terms across
countries is allowed for in the estimates that cover the entire period
from 1875 to 2006. Finally, the economy-wide TFP data are based on the
three-factor homogenous Cobb-Douglas production technology (capital,
land, and labor). See Madsen (2008c) for details on data computations.
Instruments for Tariffs and the Propensity to Import
As mentioned above, Op is proxied by Tr or m. Tariffs and the
propensity to import are instrumented because they need not be exogenous
and because they are proxies for trade openness and, as such, are
measured by an error. Some of the instruments used here follow the
suggestion of Trefler (1993) and Frankel and Romer (1999). Not all of
their instruments are used here either because they are fixed over time
for each individual country or because they are not available over the
last 137 years. The following instruments are used for Op: population
density (ratio of population to land area), time dummies, population
size, rate of unemployment, change in the rate of unemployment, per
capita mining gross domestic product (GDP), per capita agricultural
production, per capita arable land, and rate of inflation. The change in
population density and the population size are much influenced by
fertility and death rates that are not strictly determined by growth.
Sachs and Warner (1995) argue that countries with higher population
densities are more likely to be open and have more international
contacts. Frankel and Romer (1999) find that once population is
controlled for, variables representing geography account for only a
small proportion of the variations in trade.
The literature on endogenous tariffs suggests that unemployment is
an important determinant of tariffs (see Magee, Brock, and Young 1989).
Trefler (1993) also notes that politicians insist that trade protection
safeguards the livelihood of the potential unemployed in industries that
are particularly threatened by international competition. Per capita
agricultural GDP is included as an instrument because agricultural
products have traditionally been subject to much higher tariff rates
than other tradables, presumably because the agricultural lobby has been
particularly strong (Madsen 2001). Per capita GDP in mining is included
in the estimates because it is likely to be exogenous and a commodity
that is often traded internationally. Finally, the rate of inflation
serves as a potential important instrument for Tr because tariffs are
often in fixed nominal values, which implies that tariff rates are
reduced in periods of inflation and vice versa (see discussion below).
Furthermore, a country that experiences inflation in excess of inflation
among its trade partners will experience lobbying among firms for tariff
escalations to be able to compete with the outside world (Magee, Brock,
and Young 1989).
Data
The data cover 16 countries (G16) that have consistent data on
macro-tariff rates, import penetration ratios, and variables used to
compute TFP over the period from 1870 to 2006. These countries are
Canada, the United States, Japan, Australia, Belgium, Denmark, Finland,
France. Germany, Italy, the Netherlands. Norway, Spain, Sweden,
Switzerland, and the United Kingdom.
Macro-tariff rates are measured as revenues from import duties
divided by nominal imports of goods. Although macro-tariff rates suffer
from the index number problem, where substitution away from imports of
high- to low-tariff items mutes the mean and the variance of the average
tariff rate, it is widely agreed that macro-tariff rates are good
measures of tariff rates (Irwin 1998; O'Rourke 2000: Madsen 2001).
Furthermore, Rodriguez and Rodrik (2000), who are among the strongest
critics of the empirical literature on the nexus between growth and
trade barriers, recommend tariff rates as measures of trade barriers.
Figure 1 displays the weighted average of the macro-tariff rate for
G16 countries. The figure shows that tariff rates have changed
substantially over the course of history, particularly when the index
number problem is taken into account. Increasing tariffs can be
identified over the following four periods: 1875-1890, 1918-1923,
1930-1935, and 1951-1960. The increasing tariffs during some of these
periods were partly deflation induced. Madsen (2001) finds that almost
half of the variations in the macro-tariff rates were price induced in
the interwar period. This finding suggests that the tariff escalations
in the 1880s, the beginning of the 1920s, and the 1930s were in part or
entirely deflation induced, assuming that price changes were equally
influential for tariff rates before World War l and during World War II.
The finding also suggests that the tariff reductions during the world
wars were inflation induced.
[FIGURE 1 OMITTED]
The import penetration ratio is displayed in Figure 2. The ratio
grew slowly during the globalization period from 1870 to 1913.
Thereafter it declined markedly as a result of increasing nationalism,
increasing trade barriers, and reduced or negative income growth (Madsen
2001). The 1913 level was not reestablished until the end of the 1970s.
The post-1914 decline occurred in three abrupt stages: World War I, the
Great Depression, and World War II. Measured in terms of import
penetration, it is remarkable that the level of globalization, which
resumed after World War II, has only recently reached the level that
prevailed in 1913. Although the growth in services partly explains why
the G16 countries are not more open today than they were a century ago,
the figure nevertheless indicates that today's globalization is not
historically unique. The decline in import penetration over the period
from 1913 to 1945 and its slow recovery is more unusual than the level
of import penetration that prevails today.
The average TFP for the G16 countries is displayed in Figure 3
(construction of TFP is detailed in Appendix 2). TFP has been increasing
at a relatively constant rate over the whole period, except during the
period 1948-1973, where the growth rate was exceptionally high. The
growth in TFP coincides only partially with increasing import
penetration (see Figure 2). The increasing openness over the period from
1870 to 1913 is associated with a steady increase in TFP. However, TFP
continues its steady increase from 1913 to 1930 despite a significant
reduction in openness during the same period. Conversely during the
period from 1948 to 1973 the strong growth in TFP is associated with a
significant increase in openness. The periods of rapid TFP growth are
also only weakly associated with tariff reductions (Figure 1). During
the period from 1890 to 1930 the macro-tariff rate followed a U-shaped
path despite a relatively stable upward trend in TFP: however, the
average tariff rate was reduced in the high growth period from World War
II to 1973. Overall, the figures indicate a blurred relationship between
TFP, tariffs, and import penetration, which suggests that factors other
than openness have been influential for TFP growth.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
The stock of knowledge and research activity are estimated using
patent application count data because R&D data, which are used by
Coe and Helpman (1995) and most of the subsequent literature, have
become consistently available for the OECD countries only over the last
couple of decades. Patent data, by contrast, have been consistently
collected on an annual basis for almost all OECD countries since 1870
and are considered useful indicators of new knowledge (see, for
instance, Griliches 1990). Because the time lag between the time at
which a patent application is filed and eventually granted and the
chance of success varies substantially over time and across countries,
the estimates of the stock of knowledge and research activity are based
on patents applied for. The stock of knowledge is computed using the
inventory perpetual method on the domestic patent applications with a
20% geometric depreciation rate following the estimates by Pakes and
Schankerman (1984). Patent data for 21 OECD member countries are used to
construct [S.sup.f] (data sources are available from the author upon
request). These countries have, over the entire period considered, filed
more than 90% of the patent applications in the world (WIPO 2002).
4. Estimation Results
Simple Regression Results
First, consider the estimates in Table 1, where the coefficients of
all the research-related variables are restricted to zero. Tr and m are
instrumented in all the estimates in this article except the estimates
in the last table (Table 8). The tests for overidentifying restrictions
do not reject the null hypothesis that the instruments are exogenous at
the 7% level. (5) Consider the level effects of openness. Openness is
measured by Tr in the first three columns in the table. The estimated
coefficients of the level of Tr are negative: however, they are
statistically significant only in the estimates in which productivity is
measured in per capita terms. As discussed above, per capita is the
least reliable measure o1" productivity advances of the three
productivity measures because it fails to acknowledge changes in per
capita hours worked and labor force participation. These results suggest
that per capita output measures of productivity can give misleading
results about the nexus between openness and growth. In the estimates in
columns 4-6, where openness is measured by import penetration, the
estimated coefficients of openness are statistically insignificant,
except for the estimates where pet capita output is a regressor. Overall
these results suggest that openness does not have permanent productivity
growth effects in this simple framework.
Turning to the estimated coefficients of the change m openness, the
results contradict each other. The estimated coefficients of the change
in Tr are significantly positive, suggesting that increasing tariffs
bring productivity up to a higher level. Conversely, the estimated
coefficients of the change in m are significantly positive in two of the
three regressions, which is more in line with the predictions of
standard models of growth and openness. These contradictory results
indicate that Tr or m may be bad proxies for trade openness, that
important control variables are omitted from the estimates, or that
specification problems are present in the estimates. Furthermore, the
coefficient estimates of Tr and m do not give any indication as to how
openness influences growth. It is now investigated whether these results
hold against the inclusion of control variables and changes in
estimation periods.
Unrestricted Estimates of Equation 1
Unrestricted estimates of Equation 1 over the periods 1875-2006,
1956 2006, t915 1951, and 1870 1910 are presented in Tables 2 5,
respectively. The tests for overidentifying restrictions do not reject
the null hypothesis that the instruments are exogenous at the 10% level.
Consider first the estimates over the period 1875 2006 in Table 2. The
estimated coefficients of Tr and [DELTA]Tr (columns 1-3) are generally
of low statistical significance and with conflicting signs, which
suggests that there is no clear direct relationship between growth and
tariffs when knowledge and research intensity are allowed for in the
estimates. The estimated coefficients of m in the estimates in columns 4
6 are statistically insignificant: whereas, the estimated coefficients
of [DELTA]m are positive and mostly statistically significant. Overall,
these results suggest that openness does not have any direct permanent
growth effects but may have direct temporary positive growth effects.
The estimated coefficients of [DELTA]ln [S.sup.t] are consistently
positive and highly significant. Because [DELTA]ln [S.sup.f] is based on
the interaction between import penetrations and foreign stock of
knowledge, this result suggests that openness has temporary growth
effects, provided that the country trades with countries that have
positive knowledge stocks. The estimated coefficients of [(X/Q).sup.f]
are in most cases insignificant: however, the estimated coefficients of
m[(X/Q).sup.f] are positive and often highly significant. This result
suggests that foreign research intensity, conditional on import
penetration, is influential for growth. The estimated coefficients of
[(X/Q).sup.d] are positive and highly significant, pointing toward
strong growth effects of domestic research intensity. Finally, the
estimated coefficients of [DELTA]ln [S.sup.d] are positive and, in most
cases, statistically significant.
In the estimates over the period from 1956 to 2006 in Table 3, the
levels of tariffs have either positive or negative growth effects:
whereas, the estimated coefficients of [DELTA]Tr are statistically
insignificant. The estimates in columns 4 6 in Table 3 indicate that the
level and change in import penetration have not been influential for
growth over the last 50 years. The estimates show that research is
essential for growth and that knowledge and research intensity
spillovers through the channel of imports are important for growth. In
summary, the estimates in Table 3 suggest that openness has been
important for growth in the post-World War II period but that the
beneficial effects of openness stem from knowledge spillovers.
The estimates in Table 4 over the period from 1915 to 1951 are
interesting because they cover a period during which the world was
exposed to two depressions and two world wars.
The regressions suggest that tile tariff escalations during the
period 1929 1932 had strong negative growth effects. The estimated
coefficients of Tr are consistently highly significant and negative:
whereas, the estimated coefficients of [DELTA]Tr are insignificant, thus
pointing toward permanent negative growth effects of tariffs.
Conversely, the estimated coefficients of m are insignificant: whereas,
the estimated coefficients of [DELTA]m are positive and mostly
significant. The estimated coefficients of [DELTA]ln [S.sup.d] and
[DELTA]ln [S.sup.f] are positive and highly significant: whereas, the
estimated coefficients of research intensity are insignificant.
Overall, these estimates give insight into an unusual growth
process during the world wars and the interwar period. First, growth was
significantly adversely affected by the escalating trade barriers
shortly before the Great Depression and during the first half of the
Great Depression. Second, the negative growth effects of trade barriers
were reinforced by the reduced knowledge spillovers through the channel
of imports. Third, research activity did not have permanent, but
temporary, growth effects during that period, which may be a result of
the large gyrations in output and productivity during that period. This
may have blurred the genuine relationship between productivity growth
and foreign and domestic research intensity.
Finally, consider the estimates over the period from 1875 to 1913
in Table 5. Only one of the estimated coefficients of [DELTA]Tr, Tr,
[DELTA]m, and m is significant at conventional significance levels. This
result is consistent with the results from pre World War I estimates in
the literature (Bairoch 1972; Capie 1994; Foreman-Peck 1995;
O'Rourke 2000, Clements and Williamson 2001; Irwin 2002; Irwin and
Tervi6 2002; Valnvakidis 2002). Furthermore, the estimates indicate that
domestic and foreign research activity was influential for growth during
that period. The estimated coefficients of the domestic and the foreign
stock of knowledge and imports of research intensity are almost all
highly significant.
Overall the estimates suggest the absence of direct productivity
effects of openness over the last 137 years except during the interwar
period and the world wars, during which tariff had permanent adverse
productivity growth effects. The estimates, however, show that knowledge
spillovers through the channel of imports were influential for growth
for all periods considered. Furthermore, the estimates indicate that
R&D intensity has permanent growth effects, which are consistent
with the predictions of Schumpeterian growth theories and the findings
of Madsen (2008a).
5. Sensitivity Analyses
To test for the sensitivity of the results to model specification,
Equation 1 is extended to allow for the investment ratio and the
interaction between import penetration and international knowledge
spillovers. Furthermore, ordinary least squares (OLS) estimates of
Equation 1 are undertaken in this section. First, the following two
equations are estimated:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
and
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
where I is real nonresidential investment and Y is real GDP.
Equation 2 is Equation 1 augmented with the investment ratio to allow
for transitional dynamics, noting that the growth effects of knowledge
and knowledge spillovers are derived from growth models along their
balanced growth paths. Equation 3 is Equation 1 augmented to allow for
the interaction between import penetration and foreign knowledge.
Following the lead of Coe and Helpman (1995), most of the subsequent
empirical literature has included this interaction term in their
regressions. Coe and Helpman (1995) multiplied [S.sup.f] and In Ss to
capture the role of international trade. They argued that although
[S.sup.f] consists of import-weighted knowledge stocks, these weights
are fractions that add up to one and, therefore, do not reflect
openness. The weights used in this article to construct [S.sup.f], which
are based on the weighting scheme of Lichtenberg and van Pottelsberghe
de la Potterie (1998), do not add up to one; however, they are
influenced by import penetration, as argued above. Thus, the variable
[m.sub.it] In [S.sup.f.sub.it] counts import penetration twice. Here m
and In [S.sup.f] are multiplied to investigate whether knowledge
spillovers through the channel of imports influence growth in a
nonlinear fashion.
The results of estimating Equation 2 are displayed in Table 6. The
tests for overidentifying restrictions do not reject the null hypothesis
that the instruments are exogenous at the 5% level. The estimated
coefficient of the investment ratio is significant only in the estimates
where the dependent variable is per capita output, which suggests that
I/Y is not capturing transitional dynamics but is correlated with the
labor force participation rate and hours worked. The parameter estimates
of the other regressors are almost unaffected by the inclusion of the
investment ratio.
The results of estimating Equation 3 are presented in Table 7. The
tests for overidentifying restrictions do not reject the null hypothesis
that the instruments are exogenous at the 5% level. The estimated
coefficient of m ln [S.sup.f] is statistically significant in most
cases. Coupled with the fact that the estimated coefficient of In St
remains statistically highly significant, this result suggests that an
independent effect on growth comes from import penetration. In other
words, the more open is the economy, the stronger is its capacity to
absorb knowledge that is produced elsewhere. Furthermore, the estimates
suggest that the spillover effect is increasing more than
proportionately with import penetration, since import penetration
influences knowledge spillovers in a nonlinear fashion.
The estimated coefficients of m[(X/Q).sup.f] are highly significant
in the estimates where productivity is measured as TFP or output per
hour worked. This suggests that the capacity of a country to take
advantage of the research intensity among its trade partners is an
increasing function of import penetration. Note that the estimated
coefficients of [(X/Q).sup.f] have gone from being statistically
significant in Table 1 to being insignificant in Table 7. This
reinforces the result that the opportunity to import research intensity
from trade partners depends positively on imports. Finally, the
estimated coefficients of openness have hardly been affected by the
inclusion of the interaction terms, which shows that the results
obtained in the previous section are robust to inclusion of other
control variables.
Finally, Equation 1 is estimated using OLS over the period from
1875 to 2006. The tests for overidentifying restrictions do not reject
the null hypothesis that the instruments are exogenous at the 5% level.
The results, which are reported in Table 8, are very close to the
instrumental variables (IV) estimates in Table 1. The estimated
coefficients of Tr and [DELTA]Tr (columns 1-3) are insignificant except
for the estimated coefficient of [DELTA]Tr in the third column. The
estimated coefficients of m in the estimates in columns 4-6 are
statistically insignificant, while the estimated coefficients of
[DELTA]m are positive and mostly statistically significant. The
knowledge variables remain significant, as in the other regressions.
Overall the estimation results are quite similar to the estimates in
Table 1, and this suggests that the results are fairly insensitive to
whether or not instruments are used in the regressions.
6. Conclusion and Implications of the Findings
This article argues that simple models relating per capita GDP
growth to the level of tariffs and openness may not uncover a genuine
relationship between growth and trade barriers. This is because
important conditional variables may be omitted from the regressions and
because the interaction between imports and the foreign stock of
technology is not allowed for in such models. Estimates in this article
based on a simple model relating productivity and openness confirm the
findings in the literature that productivity growth tends to be
unrelated to tariffs and the propensity to import.
Extending the model to allow for the influence on productivity
growth of the growth in domestic and foreign knowledge stock and
research intensity through the channel of imports changes the results in
important Ways. The estimated coefficients of openness, measured by
tariff rates and import penetration, remained mostly insignificant,
except for the period 19151951, during which the level of tariff rates
had significant negative effects on productivity growth. However,
openness is influential for productivity growth once the interaction
between the propensity to import and foreign knowledge stock or research
intensity is allowed for in the estimates. First, the estimated
coefficients of the interaction between the propensity to import and
trade-weighted research intensity were mostly highly significant.
Second, the growth in foreign knowledge through the channel of imports
was highly influential for productivity growth. Furthermore, the
interaction between the growth in the import of knowledge through the
channel of imports and the propensity to import was influential for
growth. This reinforces the findings that openness is important for
growth when conditioned on knowledge spillovers. These results are
powerful because they show (i) that the positive relationship between
imports of technology and growth is not spuriously driven by openness
and (ii) that openness is not a virtue in its own right but that trade
needs to be targeted at products that embody technological knowledge
before a country can benefit from trade.
Another noteworthy result is that tariff rates had markedly
negative effects on growth during the period 1915-1951, a period that
covers two depressions and two world wars. The tariff escalations over
the period from 1916 to 1932 contributed significantly to a reduction in
the growth rates during that period. Simulations of the models suggest
that tariffs reduced productivity growth by 0.7% on an annual basis.
This figure suggests that, although the tariff escalations at the onset
of the Great Depression contributed to the Depression, they cannot stand
alone as factors that were responsible for the Great Depression, as
claimed by Meltzer (1976).
The findings that the interaction between the propensity to import
and research intensity were influential for productivity growth suggest
that openness can have permanent growth effects as long as a
country's trading partners undertake at least some research. The
measure used for research intensity in this article, namely, patents
applied for per worker, has been relatively constant over the past
century. This implies that the increasing propensity to import since the
end of World War II has increased the contribution of foreign research
intensity to productivity growth. This may partly explain the
productivity convergence among the OECD countries that has intensified
since the end of World War II.
The question is whether the results are limited to the OECD
countries, or whether they can be generalized to developing countries,
particularly the poorest developing countries. Coe, Helpman, and
Hoffmaister (1997) find significant knowledge spillovers through the
channel of imports from North to South, which suggests that the findings
in this article, to some extent, operate effectively for developing
countries. However, whether the interaction between the propensity to
import and research intensity influences growth in developing countries
has thus far not been tested. A problem facing the low-income developing
countries is that they do not as yet have the educational- and
research-related capacity to exploit effectively the technology that has
been developed elsewhere. Future research in this area should shed some
light on this issue.
Appendix 1: Instrument Variable Regression
Table A1 shows the results from the first-stage IV regressions. The
estimated coefficients are generally significant in the regressions in
which Tr and m are measured in levels and the [R.sup.2] are of
acceptable levels. This suggests that the instruments for Tr and m are
potentially useful. The [R.sup.2] are on the lower side in the
regressions in which [DELTA]Tr and [DELTA]m are dependent variables, and
most of the estimated coefficients of the nondeterministic variables are
insignificant. These results suggest that the instruments used for
[DELTA]Tr and Am are not of high quality and that they may potentially
give misleading coefficient estimates of [DELTA]Tr and Am. However,
because the OLS estimates in Tables 1-8 are almost identical to the IV
estimates, the potential bias introduced by the use of bad instruments
is unlikely to be significant.
Appendix 2: TFP Data Source
The economy-wide TFP data are based on the three-factor homogenous
Cobb-Douglas production technology. Following the Divisa-Tornqvist
method, the land shares are allowed to vary over time and across
countries:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (A1)
where [Y.sup.r] is real GDP, L is labor inputs (annual hours worked
times economy-wide employment), K is capital stock, T is land area under
cultivation, (1 - [alpha]) is labor's income share for country i,
and s is the agricultural sector's share of the economy-wide GDP,
which is allowed to vary across countries and over time. Labor's
income share is calculated as the economy-wide compensation to employees
divided by nominal GDP, where labor's compensation is corrected for
imputed payments to the self-employed and the data are calculated as far
back in history as income share data are available (data are available
from the author upon request). This correction is essential because
earnings from self-employment in national accounts are counted as
profits and, consequently, do not count as labor income earned by the
employment of self-employed. To correct for this bias the average
earning per employee, multiplied by the number of self-employed, is
added to the compensation of employees. Labor inputs are measured as
annual hours worked per worker multiplied by economy-wide employment, as
opposed to population, to cater for the fact that the labor force
participation rate and annual hours worked have changed substantially
over the past 137 years.
The division of output elasticities between land and capital
follows the method suggested by Denison (1967, p. 41), in which the
output elasticity of land is measured as the share of agriculture in
total GDP. While land is not an important factor of production for the
industrial countries today, it was an essential production factor before
the mid-20th century. The unweighted average of the share of agriculture
in total GDP has declined from 37% in 1870 to 2% in 2002 for G16
countries. This underscores the importance of including land as a factor
of production in the TFP estimates that go far back in history.
For other data sources see Madsen (2008c).
Table A1. Absolute t-Ratios Associated with the Instruments in the
First Round Regression
Tr [DELTA]Tr M [DELTA]m
U 1.84 0.65 1.97 0.31
[DELTA]U 1.54 0.20 0.22 0.20
[DELTA]Min 2.31 0.61 1.29 0.92
[DELTA]Landp 2.84 1.87 2.99 0.65
[DELTA]Agr 0.79 0.10 1.87 1.14
[pi] 0.36 6.19 1.42 0.74
[DELTA]Pden 2.06 0.89 3.61 0.04
Pop 1.03 0.50 9.22 0.15
[R.sup.2] 0.32 0.21 0.47 0.29
Agr = per capita agricultural production, Landp = per capita arable
land, Min = per capita mining GDP, Pden = population density,
Pop = population size, U =the rate of unemployment, n = the five-year
inflation rate, and 4 = five-year difference estimator. Time dummies
and constant terms are included in the regressions but not shown.
Received November 2007; accepted December 2008.
References
Aghion, Philippe, and Peter Howitt. 1998. Endogenous growth.
Cambridge, MA: MIT Press.
Anderson, James E., and J. Peter Neary. 1992. Trade reform with
quotas, partial rent retention, and tariffs. Econometrica 60:57-76.
Bairoch, Paul. 1972. Free trade and European economic development
in the 19th century. European Economic Review 3:211-45.
Baldwin, Richard E., and Rikard Forslid. 2000. Trade liberalization
and endogenous growth: A q-theory approach. Journal of International
Economics 50:497-517.
Burro, Robert J., and Xavier Sala-i-Martin. 1995. Economic growth.
New York: McGraw-Hill.
Ben-David, Dan. 1993. Equalizing exchange: Trade liberalization and
income convergence. Quarterly Journal of Economics 108:653-79.
Capie, Forest. 1994. Tariffs and growth: Some insights from the
worM economy, 1850-1940. Manchester, UK: Manchester University Press.
Clements, Michael A., and Jeffrey G. Williamson. 2001. A
tariff-growth paradox? Protection's impact the world around
1875-1997. NBER Working Paper No. 8459.
Coe, David T., and Elhanan Helpman. 1995. International R&D
spillovers. European Economic Review 39:859-87.
Coe, David T., Elhanan Helpman, and Alexander W. Hoffmaister. 1997.
North-south R&D spillovers. Economic Journal 107:134-49.
Del Barrio-Castro, Thomhs, Enrique Lopez-Bazo, and Guadalupe
Serrano-Domingo. 2002. New evidence on international R&D spillovers,
human capital and productivity in the OECD. Economics Letters 77:41-45.
Denison, Edward F. 1967. Why growth rates differ. Washington, DC:
Brookings Institution.
Dollar, David. 1992. Outward-oriented developing economies really
do grow more rapidly: Evidence from 95 LDCs, 1976-85. Economic
Development and Cultural Change 40:523-44.
Edwards, Sebastian. 1993. Openness, trade liberalization, and
growth in developing countries. Journal of Economic Literature
111:135843.
Edwards, Sebastian. 1998. Openness, productivity and growth: What
do we really know? Economic Journal 108:383-98.
Engelbrecht, Hans-Jurgen. 1997. International R&D spillovers,
human capital and productivity in the OECD economies: An empirical
investigation. European Economic Review 41:1479-88.
Foreman-Peck, James. 1995. A model of later nineteenth century
European economic development. Revista de Historia Economia 13:441-71.
Frankel, Jeffrey A., and David Romer. 1999. Does trade cause
growth? American Economic Review 89:379-99.
Frantzen, Dirk. 2000. R&D, human capital and international
technology spillovers: A cross-country analysis. Scandinavian Journal of
Economics 102:57-75.
Griliches, Zvi. 1990. Patent statistics as economic indicators: A
survey. Journal of Economic Literature 107:1661-1707.
Grossman, Gene, and Elhanan Helpman. 1991. Innovations and growth
in the global economy. Cambridge, MA: MIT Press.
Guellec, Dominique, and Bruno van Pottelsberghe de la Potterie.
2001. The internationalisation of technology analysis with patent data.
Research Policy 30:1253-66.
Guellec, Dominique, and Bruno van Pottelsberghe de la Potterie.
2004. From R&D to productivity growth: Do the institutional setting
and the source of funds of R&D matter? Oxford Bulletin of Economics
and Statistics 66:353-78.
Harrison, Ann. 1996. Openness and growth: A time-series,
cross-country analysis for developing countries. Journal of Development
Economics 48:419-47.
Harrison, Ann, and Gordon Hanson. 1999. Who gains from trade
reform? Some remaining puzzles. Journal of Development Economics
50:125-54.
Howitt, Peter. 1999. Steady endogenous growth with population and
R&D growing. Journal of Political Economy 107:715-30.
Howitt, Peter. 2000. Endogenous growth and cross-country income
differences. American Economic Review 90:829-46.
Irwin, Douglas A. 1998. Changes in U.S. tariffs: The role of import
prices and commercial policies. American Economic Review 88:1015-26.
Irwin, Douglas A. 2002. Interpreting the tariff-growth correlation
of the late 19th century. American Economic Review, Papers and
Proceedings 92:165-9.
Irwin, Douglas A., and Marko Tervio. 2002. Does trade raise income?
Evidence from the twentieth century. Journal of International Economics
58:1-18.
Jones, Charles I. 1995. R&D based models of economic growth.
Journal of Political Economy 103:759-84.
Kortum, Samuel. 1997. Research, patenting, and technological
change. Econometrica 65:1389-1419.
Lichtenberg, Frank R., and Bruno van Pottelsberghe de la Potterie.
1998. International R&D spillovers: A comment. European Economic
Review 42:1483-91.
Lumenga-Neso, Olivier, Marcelo Olarreaga, and Maurice Schiff. 2005.
On 'indirect' trade-related R&D spillovers. European
Economic Review 49:1785-98.
Madsen, Jakob B. 2001. Trade barriers and the collapse of world
trade during the great depression. Southern Economic Journal 67:848-68.
Madsen, Jakob B. 2007. Technology spillover through trade and TFP
convergence: 135 years of evidence for the OECD countries. Journal of
International Economics 72:464-80.
Madsen, Jakob B. 2008a. Semi-endogenous models versus Schumpeterian
theory: International evidence over a century. Journal of Economic
Growth 13:1-26.
Madsen, Jakob B. 2008b. Economic growth and world exports of ideas:
A century of evidence. Scandinavian Journal of Economics 110:145-67.
Madsen, Jakob B. 2008c. Trade barriers, openness, and economic
growth. Working Paper No. 27-08, Department of Economics, Monash
University.
Magee, Stephen P., William A. Brock, and Leslie Young. 1989. Black
hole tariffs and endogenous policy theory: Political economy in general
equilibrium. Cambridge, UK: Cambridge University Press.
Matsuyama, Kiminori. 1992. Agricultural productivity, comparative
advantage, and economic growth. Journal of Economic Theory 58:317-34.
Meltzer, Allan H. 1976. Monetary and other explanations of the
start of the great depression. Journal of Monetary Economics 2:455-71.
O'Rourke, Kevin. 2000. Tariffs and growth in the late 19th
century. Economic Journal 110:456-83.
Pakes, Ariel, and Mark Schankerman. 1984. The rate of obsolescence
of patents, research gestation lags, and the private rate of return to
research resources. In R&D, patents, and productivity, edited by Zvi
Griliches. Chicago: University of Chicago Press, pp. 73-88.
Quah, Danny, and James E. Rauch. 1990. Openness and the rate of
economic growth. Unpublished paper, University of California, San Diego.
Rivera-Batiz, Luis, and Paul M. Romer. 1991. Economic integration
and endogenous growth. Quarterly Journal of Economics 106:531-55.
Rodriguez, Francisco, and Dani Rodrik. 2000. Trade policy and
economic growth: A skeptic's guide to cross-national evidence. In
NBER Macroeconomics Annual, edited by Ben S. Bernanke and Kenneth S.
Rogoff. Cambridge, MA: MIT Press, pp. 261-325.
Rodrik, Dani. 1999. The new global economy and developing
countries: Making openness work. Washington, DC: Overseas Development
Council.
Romer, Paul. 1990. Endogenous technological change. Journal of
Political Economy 98:S71-S102.
Romer, Patti M. 1992. Two strategies for economic development:
Using ideas and producing ideas. World Bank Annual Conference on
Economic Development. Washington, DC: World Bank.
Sachs, Jeffrey, and Andrew Warner. 1995. Economic reform and the
process of global integration. Brookings Papers on Economic Activity
1:1-118.
Segerstrom, Paul S. 1998. Endogenous growth without scale effects.
American Economic Review 88:1290-1310.
Trefler, Daniel. 1993. Trade liberalization and the theory of
endogenous protection: An econometric study of U.S. import policy.
Journal of Political Economy 101:138--60.
Vamvakidis, Athanasios. 1998. Regional integration and economic
growth. Worm Bank Economic Review 12:251-70.
Vamvakidis, Athanasios. 2002. How robust is the growth-openness
connection? Historical evidence. Journal of Economic Growth 7:57-80.
World International Property Organization (WIPO). 2002. 100 years
of industrial property statistics. Geneva: WlPO. Yanikkaya, Halit. 2003.
Trade openness and economic growth: A cross-country empirical
investigation. Journal of Development Economics 72:57-89.
Young, Alwyn. 1998. Growth without scale effects. Journal of
Political Economy 106:41-63.
Jakob B. Madsen *
* Department of Economics, Monash University, 900 Dandenong Road,
Caufield East, Melbourne, Victoria 3145, Australia; E-mail
Jakob.Madsen@buseco.monash.edu.au.
Support from an Economic Policy Research Unit grant from the Danish
government is gratefully acknowledged. Keeli Hennessey and Phillip
Oksnes provided excellent research assistance. Helpful comments and
suggestions from seminar participants at the University of Copenhagen
and Brunel University and particularly three referees and John Pepper,
the coeditor of the Southern Economic Journal, are gratefully
acknowledged.
(1) Often-cited studies finding a negative relationship between
trade barriers and economic growth include Dollar (1992), Sachs and
Warner (1995), Harrison (1996), Edwards (1998), and Frankel and Romer
(1999).
(2) Edwards (1998) and O'Rourke (2000) are among the few
studies that condition their output growth regressions on factors of
production. Edwards (1998) uses TFP as the dependent variable for a
cross-section sample of 93 countries over the period from 1950 to 1990.
O'Rourke (2000) allows for land under cultivation and capital stock
in his regressions; however, capital stock is proxied by coal
consumption.
(3) Only direct knowledge transfers are allowed for in the
estimates. Lumenga-Neso, Olarreaga, and Schiff (2005) advocate the
importance of allowing for indirect knowledge transfer through the
channel of imports. The indirect effect arises from the fact that
imports of knowledge from country X to country Y are transmitted to
country Z when country Z imports from country Y. Although indirect
spillover effects are potentially important, it is beyond the scope of
this article to consider these effects. More importantly, following
endogenous growth theory, it is only the intermediate products from
country X that are passed on to country Z, through country Y, that are
relevant for knowledge spillovers. These product types cannot easily be
identified by the historical data that are available.
(4) Consider the constant returns to scale production function used
in this article to estimate TFP: g =
B[K.sup.[alpha]][T.sup.[beta]][L.sup.I-[alpha]-[beta], where B is
productivity, K is capital, Tis land area, and L is labor. Since Y/K is
constant along a balanced growth path, the growth in output per worker,
[g.sub.y], follows the following growth path: [g.sub.y = 1/1-[alpha] g-
[beta]/1-[alpha] x n, where n is the population growth rate and g is the
rate of technological progress. Thus the growth in output per capita or
output per hour worked is lower than TFP growth because of diminishing
returns introduced by land as a fixed factor of production.
(5) The tests of overidentifying restriction regress the residuals
from the IV regressions on the exogenous variables. The [R.sup.2] from
these regressions is used to estimate n[R.sup.2], where n is the number
of observations, Here, n[R.sup.2] is distributed as [[chi].sup.2] under
the null hypothesis that the instruments are exogenous.
Table 1. Parameter Estimates of the Restricted Version of Equation 1
in the Period 1875-2006
1 2 3
Y/Pop TFP YlH
Tr -1.763 (10.8) -0.177 (1.18) -0.212 (1.16)
[DELTA]Tr 0.870 (7.40) 0.234 (1.93) 0.543 (4.46)
m
[DELTA]m
[chi square] 0.02 0.01 0.06
DW 2.05 1.95 1.98
[R.sup.2](B) 0.90 0.77 0.89
4 5 6
Y/Pop TFP YlH
Tr -0.245 (2.85) -0.022 (0.27) -0.124 (1.38)
[DELTA]Tr 2.000 (3.46) 2.380 (4.50) 0.055 (0.10)
m 0.02 0.03 0.05
[DELTA]m 2.00 2.00 1.98
[chi square] 0.69 0.76 0.85
DW
[R.sup.2](B)
The numbers in parentheses are absolute t-statistics. [R.sup.2](B)
is Buse's raw-moment [R.sup.2]. [X.sup.2] is the p-values of tests for
overidentifying restrictions. The following instruments are used for
tariffs and openness: The rate of unemployment, the change in the rate
of unemployment, the growth in per capita agricultural GDP, the growth
in per capita mining GDP, growth in consumer prices, the growth in per
capita arable land, population growth, the change in population
density, and time dummies. Time dummies and country dummies are
included in the estimates but are not shown. The following data points
are used: 1875, 1880, 1885, 1890, 1895, 1900, 1905, 1910, 1915, 1920,
1925, 1930, 1935, 1940, 1951, 1956, 1961, 1966, 1971, 1976, 1981, 1986,
1991, 1996, 2001, and 2006. m = propensity to import, TFP = total
factor productivity, Tr = macro-tariff rate, Y/H =output per hour
worked, and Y/Pop = per capita income.
Table 2. Parameter Estimates of Equation 1 in the Period 1875-2006
1 2
Y/Pop TFP
Tr -1.387 (7.31) -0.159 (0.79)
[DELTA]Tr 0.324 (2.25) 0.073 (0.50)
m
[DELTA]m
[DELTA]1n [S.sup.d] 0.023 (2.41) 0.029 (3.13)
[DELTA]1n [S.sup.f] 0.044 (9.94) 0.029 (6.860
(X/Q) (d) 0.007 (2.18) 0.009 (3.41)
(X/Q) (f) 0.006 (2.41) -0.004 (1.59)
m(X/Q) (f) 0.01 (1.39) 0.022 (3.08)
[chi square] 0.01 0.01
DW 2.07 2.01
[R.sup.2](B) 0.90 0.79
3 4
Y/H Y/Pop
Tr 0.614 (2.57)
[DELTA]Tr 0.167 (1.11)
m -0.185 (2.04)
[DELTA]m 1.857 (3.21)
[DELTA]1n [S.sup.d] 0.018 (1.67) 0.002 (0.15)
[DELTA]1n [S.sup.f] 0.027 (5.83) 0.044 (10.83)
(X/Q) (d) 0.016 (4.42) 0.015 (5.05)
(X/Q) (f) 0.003 (1.20) 0.005 (1.66)
m(X/Q) (f) 0.028 (3.84) 0.011 (1.57)
[chi square] 0.04 0.02
DW 1.98 2.08
[R.sup.2](B) 0.88 0.89
5 6
TFP Y/H
Tr 0.008 (0.10) -0.081 (0.91)
[DELTA]Tr 2.577 (4.71) 0.355 (0.61)
m 0.019 (2.16) 0.02 (1.93)
[DELTA]m 0.03 (7.58) 0.029 (6.16)
[DELTA]1n [S.sup.d] 0.010 (3.99) 0.013 (3.88)
[DELTA]1n [S.sup.f] -0.003 (1.46) 0.002 (0.90)
(X/Q) (d) 0.023 (3.53) 0.031 (4.09)
(X/Q) (f) 0.02 0.05
m(X/Q) (f) 2.05 1.99
[chi square] 0.78 0.88
DW
[R.sup.2](B)
See notes to Table 1. m = propensity to import, [S.sup.d] = domestic
knowledge stock, [S.sup.f] = foreign knowledge spillovers through the
channel of imports, TFP = total factor productivity, Tr = macro-tariff
rate, (X/Q) (d) = domestic research intensity, (X/Q) (d) = foreign
spillovers of research intensity through the channel of imports,
Y/H =output per hour worked, and Y/Pop =per capita income.
Table 3. Parameter Estimates of Equation 1 in the Period 1956-2006
1 2 3
Y/Pop TFP Y/H
Tr -0.978 (2.45) 0.682 (1.74) 1.244 (2.68)
[DELTA]Tr 0.270 (0.72) 0.342 (0.92) 1.514 (0.84)
m
[DELTA]m
[DELTA]ln [S.sup.d] 0.103 (3.01) 0.038 (1.14) 0.023 (0.59)
[DELTA]1n [S.sup.f] 0.097 (7.08) 0.051 (3.82) 0.055 (3.55)
(X/Q) (d) 0.005 (0.40) 0.032 (2.62) 0.001 (0.10)
(X/Q) (f) 0.019 (0.64) 0.007 (0.77) 0.019 (1.65)
m(X/Q) (f) 0.009 -0.002 (0.18) 0.016 (1.24)
[chi square] 0.03 0.03 0.07
DW 1.85 1.78 1.81
[R.sup.2] (B) 0.89 0.77 0.88
4 5 6
Y/Pop TFP Y/H
Tr 0.136 (0.42) 0.066 (0.420 0.049 (0.26)
[DELTA]Tr 1.392 (1.99) 1.450 (1.99) -0.124 (0.13)
m 0.095 (1.23) 0.040 (1.23) 0.048 (1.27)
[DELTA]m 0.095 (6.30) 0.061 (6.300 0.104 (10.1)
[DELTA]ln [S.sup.d] 0.000 (0.02) 0.027 (2.21) -0.005 (0.39)
[DELTA]1n [S.sup.f] 0.023 (2.33) 0.012 (1.24) 0.032 (2.68)
(X/Q) (d) 0.009 (0.66) 0.012 (1.25) 0.020 (1.21)
(X/Q) (f) 0.01 0.02 0.07
m(X/Q) (f) 1.92 1.80 1.84
[chi square] 0.88 0.74 0.90
DW
[R.sup.2] (B)
See notes to Table 1. m = propensity to import, [S.sup.d] = domestic
knowledge stock, [S.sup.f] = foreign knowledge spillovers through the
channel of imports, TFP =total factor productivity, Tr = macro-tariff
rate, (X/Q) (d) = domestic research intensity, (X/Q) (f) = foreign
spillovers of research intensity through the channel of imports,
Y/H = output per hour worked, and Y/Pop = per capita income.
Table 4. Parameter Estimates of Equation 1 in the Period 1915-1951
1 2 3
Y/Pop TFP Y/H
Tr -5.169 (3.52) -5.774 (3.64) -6.452 (3.86)
[DELTA]Tr -0.030 (0.07) -0.414 (0.94) -0.304 (0.61)
M
[DELTA]m
[DELTA]ln [S.sup.d] 0.264 (4.390 0.187 (2.83) 0.132 (1.80)
[DELTA]1n [S.sup.f] 0.027 (3.13) 0.019 (2.06) 0.012 (1.01)
(X/Q) (d) 0.008 (0.25) 0.024 (0.65) 0.044 (1.06)
(X/Q) (f) 0.006 (0.41) 0.000 (0.07) 0.005 (0.28)
m(X/Q) (f) 0.026 (0.71) 0.004 (0.10) 0.049 (1.05)
[chi square] 0.05 0.02 0.07
DW 1.90 1.85 1.82
[R.sup.2] (B) 0.77 0.60 0.73
4 5 6
Y/Pop TFP Y/H
Tr
[DELTA]Tr
M -1.222 (2.68) -0.072 (0.16) 0.204 (0.39)
[DELTA]m 7.654 (3.24) 7.786 (3.37) 2.019 (0.73)
[DELTA]ln [S.sup.d] 0.246 (3.75) 0.281 (3.92) 0.178 (2.13)
[DELTA]1n [S.sup.f] 0.036 (5.17) 0.028 (3.43) 0.018 (1.65)
(X/Q) (d) 0.014 (0.41) 0.010 (0.30) 0.038 (0.95)
(X/Q) (f) -0.008 (0.63) 0.002 (0.11) 0.011 (0.60)
m(X/Q) (f) 0.037 (1.15) -0.010 (0.26) 0.039 (0.89)
[chi square] 0.04 0.04 0.02
DW 1.95 1.95 1.85
[R.sup.2] (B) 0.83 0.65 0.75
See notes to Table 1. m = propensity to import, [S.sup.d] = domestic
knowledge stock, [S.sup.f] = foreign knowledge spillovers through the
channel of imports. TFP =total factor productivity, Tr = macro-tariff
rate, (X/Q) (d) = domestic research intensity, (X/Q) (f) = foreign
spillovers of research intensity through the channel of imports,
Y/H =output per hour worked, and Y/Pop = per capita income.
Table 5. Parameter Estimates of Equation 1 in the Period 1875-1910
1 2 3
Y/Pop TFP Y/H
Tr 0.955 (0.95) 0.902 (1.08) 2.114 (2.43)
[DELTA]Tr -0.391 (0.65) -0.328 (0.69) -0.598 (1.09)
M
[DELTA]m
[DELTA]ln [S.sup.d] 0.021 (1.26) 0.051 (3.64) 0.040 (2.55)
[DELTA]1n [S.sup.f] 0.077 (3.94) 0.052 (2.74) 0.052 (2.41)
(X/Q) (d) 0.007 (1.02) 0.001 (0.27) 0.008 (1.38)
(X/Q) (f) -0.024 (1.88) -0.017 (1.42) -0.013 (0.88)
m(X/Q) (f) 0.029 (2.26) 0.052 (4.12) 0.032 (2.36)
[chi square] 0.02 0.03 0.03
DW 1.85 1.81 1.78
[R.sup.2] (B) 0.82 0.77 0.89
4 5 6
Y/Pop TFP YlH
Tr
[DELTA]Tr
M -0.347 (1.20) -0.23 (1.03) -0.494 (1.96)
[DELTA]m 2.972 (1.16) 2.936 (1.24) 3.165 (1.40)
[DELTA]ln [S.sup.d] 0.018 (1.10) 0.045 (3.03) 0.043 (2.65)
[DELTA]1n [S.sup.f] 0.067 (3.38) 0.043 (2.32) 0.040 (1.93)
(X/Q) (d) 0.009 (1.21) 0.002 (0.42) 0.012 (1.62)
(X/Q) (f) -0.026 (1.93) -0.018 (1.46) -0.014 (0.87)
m(X/Q) (f) 0.031 0.056 0.035 (2.54)
[chi square] 0.01 0.02 0.05
DW 1.74 1.74 1.76
[R.sup.2] (B) 0.82 0.75 0.88
See notes to Table 1. m = propensity to import, [S.sup.d] =
domestic knowledge stock, [S.sup.f] = foreign knowledge
spillovers through the channel of imports, TFP =total factor
productivity, Tr = macro-tariff rate, (X/Q) (d) = domestic
research intensity, (X/Q) (f) =foreign spillovers of research
intensity through the channel of imports, Y/H =output per
hour worked, and Y/Pop =per capita income.
Table 6. Parameter Estimates of Equation 2 in the Period 1875-2006
l 2 3
Y/Pop TFP Y/H
Tr 0.167 (3.38) -0.024 (0.54) 0.089 (1.88)
[DELTA]Tr -1.276 (6.59) -0.174 (0.85) 0.658 (2.70)
m 0.334 (2.35) 0.068 (0.48) 0.186 (1.24)
[DELTA]m
[DELTA]ln [S.sup.d]
[DELTA]1n [S.sup.f] 0.030 (3.14) 0.028 (2.99) 0.022 (2.03)
(X/Q) (d) 0.043 (9.79) 0.029 (6.84) 0.027 (5.84)
(X/Q) (f) 0.007 (2.30) 0.010 (3.39) 0.015 (4.09)
m(X/Q) (f) 0.005 (1.90) -0.003 (1.49) 0.003 (1.12)
[chi square] 0.016 (2.07) 0.021 (2.81) 0.03 (4.01)
DW 0.01 0.02 0.03
[R.sup.2] (B) 2.07 2.00 1.99
0.91 0.79 0.88
4 5 6
Y/Pop TFP Y/H
Tr 0.212 (4.38) -0.023 (0.47) 0.067 (1.01)
[DELTA]Tr
m
[DELTA]m -0.195 (2.18) 0.013 (0.15) -0.076 (0.84)
[DELTA]ln [S.sup.d] 1.811 (3.22) 2.550 (4.61) 0.245 (0.58)
[DELTA]1n [S.sup.f] 0.011 (1.10) 0.018 (1.99) 0.023 (2.24)
(X/Q) (d) 0.044 (10.9) 0.031 (7.55) 0.029 (6.16)
(X/Q) (f) 0.014 (4.81) 0.010 (3.87) 0.012 (3.42)
m(X/Q) (f) 0.003 (1.21) -0.002 (0.78) 0.019 (0.85)
[chi square] 0.018 (2.51) 0.024 (3.24) 0.032 (4.21)
DW 0.02 0.03 0.02
[R.sup.2] (B) 2.08 2.05 2.00
0.89 0.79 0.88
See notes to Table 1. m = propensity to import, [S.sup.d] =
domestic knowledge stock, [S.sup.f] = foreign knowledge
spillovers through the channel of imports, TFP =total factor
productivity, Tr = macro-tariff rate, (X/Q) (d) = domestic
research intensity, (X/Q) (f) = foreign spillovers of research
intensity through the channel of imports, Y/H =output per
hour worked, and Y/Pop =per capita income.
Table 7. Parameter Estimates of Equation 3 in the Period 1875-2006
1 2 3
Y/Pop TFP Y/H
[DELTA] (m ln[S.sup.f) 0.010 (1.79) 0.026 (4.42) 0.012 (1.82)
Tr -1.379 (7.30) -0.169 (0.84) 0.593 (2.48)
[DELTA]Tr 0.379 (2.55) 0.181 (1.27) 0.220 (1.44)
M
[DELTA]m
[DELTA]ln [S.sup.d] 0.025 (2.57) 0.032 (3.51) 0.019 (1.82)
[DELTA]1n [S.sup.f] 0.041 (7.94) 0.021 (4.44) 0.022 (4.18)
(X/Q) (d) 0.007 (2.05) 0.009 (3.35) 0.015 (4.09)
(X/Q) (f) 0.006 (2.34) -0.005 (2.07) 0.003 (1.14)
m(X/Q) (f) 0.011 (1.56) 0.023 (3.36) 0.029 (3.98)
[chi square] 0.03 0.02 0.05
DW 2.06 1.99 1.99
[R.sup.2] (B) 0.92 0.81 0.88
4 5 6
Y/Pop TFP Y/H
[DELTA] (m ln[S.sup.f) 0.006 (1.21) 0.020 (3.53) 0.011 (1.81)
Tr
[DELTA]Tr
M -0.184 (2.03) 0.001 (0.02) -0.089 (1.01)
[DELTA]m 1.836 (3.17) 2.268 (4.18) 0.231 (0.69)
[DELTA]ln [S.sup.d] 0.003 (0.26) 0.021 (2.48) 0.021 (2.06)
[DELTA]1n [S.sup.f] 0.041 (9.00) 0.024 (5.33) 0.025 (4.62)
(X/Q) (d) 0.015 (5.01) 0.010 (3.87) 0.012 (3.62)
(X/Q) (f) 0.005 (1.59) -0.004 (1.86) 0.002 (0.85)
m(X/Q) (f) 0.011 (1.65) 0.026 (3.75) 0.031 (4.18)
[chi square] 0.05 0.01 0.03
DW 2.08 2.04 2.00
[R.sup.2] (B) 0.89 0.79 0.88
See notes to Table 1. m = propensity to import; [S.sup.d] = domestic
knowledge stock, [S.sup.f] = foreign knowledge spillovers through the
channel of imports, TFP = total factor productivity, Tr = macro-tariff
rate, (X/Q) (d) = domestic research intensity, and (X/Q) (f) = foreign
spillovers of research intensity through the channel of imports,
Y/H = output per hour worked, and Y/Pop = per capita income.
Table 8. OLS Estimates of Equation 1 in the Period 1875-2006
1 2 3
Y/Pop TFP Y/H
Tr 0.010 (0.30) 0.032 (1.03) 0.053 (1.42)
[DELTA]Tr 0.073 (1.65) -0.053 (1.29) -0.184 (4.14)
m
[DELTA]m
[DELTA]ln [S.sup.d] 0.010 (0.88) 0.022 (2.47) 0.019 (1.77)
[DELTA]1n [S.sup.f] 0.046 (11.1) 0.029 (6.93) 0.025 (5.58)
(X/Q) (d) 0.016 (5.06) 0.009 (3.72) 0.012 (3.71)
(X/Q) (f) 0.006 (2.10) -0.003 (1.29) 0.003 (1.09)
m(X/Q) (f) 0.069 (0.97) 0.022 (3.00) 0.003 (4.40)
[chi square] 2.07 2.00 2.02
DW 0.88 0.81 0.89
[R.sup.2] (B)
4 5 6
Y/Pop TFP Y/H
Tr
[DELTA]Tr
m
[DELTA]m -0.033 (0.59) 0.009 (0.16) -0.116 (1.82)
[DELTA]ln [S.sup.d] 0.174 (3.37) 0.206 (3.66) 0.163 (2.50)
[DELTA]1n [S.sup.f] 0.012 (1.11 0.027 (3.19) 0.018 (1.68)
(X/Q) (d) 0.042 (10.6) 0.027 (6.78) 0.027 (5.52)
(X/Q) (f) 0.013 (4.21) 0.01 (3.78) 0.011 (3.29)
m(X/Q) (f) 0.008 (2.49) -0.004 (1.20) 0.006 (2.05)
[chi square] 0.005 (0.44) 0.023 (2.20) 0.014 (1.26)
DW 2.07 1.99 1.99
[R.sup.2] (B) 0.89 0.82 0.88
See notes to Table I. m = propensity to import, [S.sup.d] = domestic
knowledge stock, [S.sup.f] = foreign knowledge spillovers through the
channel of imports, TFP =total factor productivity, Tr = macro-tariff
rate, (X/Q) (d) = domestic research intensity, and (X/Q) (f) = foreign
spillovers of research intensity through the channel of imports,
Y/H = output per hour worked, and Y/Pop = per capita income.
