Chapter 5 The Role of Trade in Technology Diffusion
- International Monetary Fund
- Published Date:
- August 2001
Trade is considered a major channel of international technology transfer. This chapter investigates its role in transferring technology from industrial to developing countries1. Defined broadly, technology covers production methods, product design and organisational methods. According to Grossman and Helpman (1991), trade can foster technology transfer through two main channels: production and information. Through trade with countries that are technological leaders, developing countries can gain access to intermediate products and capital equipment of higher quality (vertical differentiation) and broader variety (horizontal differentiation). They can also gain access to more open channels of communication about production methods, product design, organisational methods and market conditions. Finally, they can adapt to their use the foreign technologies used in their imported products, often at lower cost than innovation would require.
Recent research has tested empirically the role of trade in cross-country technology transfer. Coe, Helpman and Hoffmaister (1997) and Jaumotte (1998), for example, confirm its significance. Building on that evidence, this chapter investigates which type of trade—intra-industry or inter-industry—operates more effectively in the transfer process. Intra-industry trade refers to two-way trade in a given sector, while inter-industry trade refers to one-way trade in a sector. The chapter tests the hypothesis that intra-industry trade is more effective for technology transfer because countries are more likely to absorb foreign technologies when their imports are from the same sectors as the products they produce and export. Indeed, the possibility of using foreign technology in domestic production will likely be greater when the country already is a large producer of the same types of goods as it imports, particularly if it is to maintain its competitiveness in international markets.
The chapter extends the theoretical framework used in Jaumotte (1998), where growth of total factor productivity (TFP)2, a proxy for absorption of technology, is specified as a function of the technological gap of the country, weighted by the country’s degree of exposure to foreign technologies. That exposure is captured by the ratio of imports to GDP. The Grubel-Lloyd intra-industry trade index (IIT) is calculated to determine each sector’s involvement in intra-industry trade.
The chapter estimates both linear and non-linear regression specifications. In the linear regression, the ratios of imports to GDP are split into intra- and interindustry components on the basis of a specific cut-off for IIT. The import shares are then aggregated separately for the two components in order to estimate separately the effect of each sector’s openness on growth of TFP. The robustness of the results is tested by excluding sectors that are net exporters from those classified as having interindustry trade. Indeed, net exporters could bias the results to show that inter-industry trade is less efficient in transferring technologies, because they are presumably technologically advanced and thus less likely to learn from the technologies inherent in their imports. In the non-linear specification, each sector’s imports are weighted by some function of its IIT index.
The data sample covers intra- and inter-industry trade in 87 countries during 1970–93. The tests yield three findings. First, they confirm that trade with industrial countries enhances the technological development of developing countries. Second, in both the linear and non-linear regression specifications, intra-industry trade had a stronger effect on TFP growth than did inter-industry trade. Finally, certain country-specific factors could, if unchanged, keep developing countries from reaching the steady-state level of technology that OECD countries have reached. Evidence for sub-Saharan Africa confirms this conclusion.
Framework of Analysis
Technology is measured by TFP, defined as the residual part of output once the contributions of factor inputs have been accounted for. The relationship between the TFP growth of a country and its degree of openness to the technological leader is modelled as follows:
and where l denotes the technological leader, i the importing country, g the growth rate of TFP, m imports and y output. The first part of the model, derived from Barro and Sala-i-Martin (1995), relates the deviation of the importing country’s TFP growth from that of the leader to the technological gap between the two countries. The specification embodies two important assumptions. First, it assumes, all else equal, that technologically backward countries tend to have faster TFP growth than technological leaders. Indeed, gi > gl if and only if TFPi < TFPl, because the cost of imitation is less than the cost of innovation. Second, the specification assumes that the discrepancy between the TFP growth of the backward country and that of the leader is increasing the technological gap. This would be the case if, for example, the costs of imitation declined as the technological gap grew larger. Intuitively, it makes sense that, as the technological gap expands and the pool of innovations to imitate increases, the cost of imitation falls.
The parameter μ denotes the speed of convergence of country i towards the leader. In accordance with the theoretical literature that emphasises trade as a major channel of technology transfer across countries, Jaumotte (1998) specifies μ as a function of the country’s degree of openness to trading with the leader. She finds empirical evidence that trade plays a significant role in the technological catch-up of follower countries.
Linear Regression Specification
The distinction between intra- and inter-industry trade is based on the Grubel-Lloyd intra-industry trade index, defined as
where s denotes sector s, X denotes exports and M denotes imports. The index measures the share of intra-industry trade in sector s. If there is no intra-industry trade—i.e. if the country imports or exports exclusively—the IIT index is zero. Conversely, if all trade is intra-industry—if Xs = Ms—the IIT index takes a value of one.
In the linear approach, the sectors of each country are classified as intra- or inter-industry trade sectors, depending on the value of their IIT indexes. Let b denote a cut-off, IR the set of inter-industry trade sectors and IA the set of intra-industry trade sectors.
The import shares are then aggregated separately for each sector, and a different coefficient is estimated for each aggregate. Thus, the following specification is estimated:
The chapter explores IIT cut-offs ranging from 0.1 to 0.9, in increments of 0.1. If both intra- and inter-industry trade have the same effects on technology transfers, their coefficients should not be significantly different, irrespective of the cut-off. If, instead, intra-industry trade has a significantly larger impact than inter-industry trade, two results can be expected. First, the coefficient on intra-industry trade should be larger than the coefficient on inter-industry trade, irrespective of the cut-off. Moreover, the difference between the two should become more significant as the chosen cut-off nears the “true” one. Second, as the cut-off is raised, the coefficients on both intra-and inter-industry trade should increase. Figures 5.1a-d illustrate four different ways in which the technological benefits from importing in a given sector can relate to the sector’s degree of intra-industry trade. In accordance with the main hypothesis, all four schemes show that the benefits from trade are increasing, although not necessarily strictly so, in line with the degree of intra-industry trade of the sector. In all four schemes, the coefficients on inter- and intra-industry trade are increasing, at least over a range, in the cut-off for the IIT index.
Figure 5.1a shows how the benefits from intra-industry trade can increase continuously. In this first scheme, both coefficients increase continuously as the cutoff is raised. In the second scheme (Figure 5.1b), the benefits can take only two values: a constant low value for sectors with low degrees of intra-industry trade, and a constant high value for sectors with high degrees of it. As the cut-off is raised, this scheme shows two phases. In the first, the coefficient on inter-industry trade holds constant while the one on intra-industry trade increases. In the second, the coefficient on inter-industry trade increases while the one on intra-industry trade stays constant. The “true” cut-off, Cm, occurs where the coefficient on inter-industry trade stops being constant and the one on intra-industry trade starts. In the third scheme (Figure 5.1c), the coefficient on intra-industry trade increases continuously, while the one on inter-industry trade is at first constant and then increases. The point at which the coefficient on inter-industry trade starts increasing identifies the true cut-off, Cm. Finally, in the fourth scheme (Figure 5. Id), the coefficient on inter-industry trade increases continuously while the one on intra-industry trade at first increases and then is constant. In this case, the true cut-off lies at the point where intra-industry trade starts being constant.
Figure 5.1.How Technological Benefits from Importing in a Sector Can Relate to the Degree of Intra-industry Trade
Testing for Robustness: Excluding Net Exporters from Inter-industry Traders
Inter-industry trade includes two types of sectors: net importers and net exporters. Net exporters are presumably technologically advanced and thus less likely to adopt the technologies used in their imports. Including them with net importers would bias the results to show that inter-industry trade is less efficient in transferring technology. To test the robustness of the results, therefore, the sectors are classified into three groups on the basis of the value of their export-to-import ratios: no base sector (NB), base sector (B) and good sector (G). Let b1 and b2 denote two cut-offs.
Note that there is a direct correspondence between the cut-off for the IIT index, b, and the two cut-offs for the ratio of exports to imports, b1 and b2, which can be expressed as:
With the corresponding cut-offs for the export-to-import ratio, the robustness of the results obtained on the basis of the distinction between inter- and intra-industry trade can be verified using the following specification:
Non-linear Regression Specification
The non-linear regression specification is the continuous version of the cut-off-based approach. Instead of splitting the sectors into two groups based on the value of their IIT index, the imports of each sector are weighted by a function of their IIT index.
The IIT index is entered in a flexible form, namely a quadratic, which will allow explicit testing of its role.
Extending the Framework to Several Technological Leaders
The model is specified with a unique technological leader. In practice, however, the technological leader is the group of OECD countries, and the TFP growth of the importing country is assumed to depend on the sum of the technology transfers from each technological leader. Thus, for example, equation 1 becomes
This aggregation procedure excludes the possibility of duplication or synergy amongst the technological transfers from different leaders. This assumption usually is made in the literature. Jaumotte (1998) tested and could not reject it.
The sample contains 87 countries, of which 63 are developing countries and 24 are OECD countries. The developing countries are grouped into five regions: East Asia (8 countries), Latin America (22), Middle East and North Africa (8), South Asia (5) and sub-Saharan Africa (20). See Appendix 1 for a complete list of countries. The data cover 1970–93.
To measure TFP, the chapter uses the growth-accounting approach, which imposes conventional values for factor shares. It then uses three alternative measures of TFP to test the robustness of the results to a particular specification of the aggregate production function. These are given by
where Y denotes GDP, K denotes the total stock of physical capital, L denotes the labour force and H denotes the stock of human capital. Note that the last specification exhibits increasing returns to scale, while the other two feature constant returns to scale. The data needed to measure TFP are from a revised version of the data set compiled by Bosworth et al. (1995). The definition and the original source of the data for each variable are given in Appendix 2. To make the TFP levels comparable across countries, the data on output and physical capital were converted into 1987 international prices, using 1987 purchasing-power parities for GDP and investment3.
The trade data for measuring the import-to-GDP ratios, IIT indexes and export-to-import ratios are from Feenstra et al. (1997), who report manufacturing trade flows disaggregated by trade partners and sectors in 34 industries classified according to the Bureau of Economic Analysis Manufacturing Industry Classification. The trade data are aggregated into 10 sectoral categories matching the International Standard Industrial Classification system. The data for nominal GDP are from the World Economic Outlook (IMF, 1997). The ratios of imports to GDP are calculated using imports from OECD countries only, whereas the IIT indexes and ratios of exports to imports are based on trade with the world.
Tables 5.1 and 5.2 summarise the TFP data for the sample of countries examined in the chapter. Table 5.1 reports the average annual growth rate of TFP during the period 1970-93 by region. Table 5.2 reports the average TFP gap of each region with respect to OECD countries in 1970 and 1993 and the TFP growth rate during 1970-93. An increase in the gap indicates that the region has been diverging from the OECD countries, while a decrease reflects catch-up.
|Middle East and North Africa||0.01||0.01||0.001|
Table 5.1 shows the TFP growth rates of OECD countries as significantly positive over the entire period although, not surprisingly, East Asia had higher ones. TFP growth was also positive for Middle East/North Africa and South Asia but less significantly so. Strikingly, sub-Saharan Africa and Latin America had significantly negative TFP growth rates. As Table 5.1 suggests, Table 5.2 shows East Asia catching up with the OECD countries, while sub-Saharan Africa and Latin America diverged significantly from them.
Tables 5.3-5.5 summarise the trade data for the sample. Table 5.3 reports the share of imports from the OECD countries in GDP, averaged for 1970-90 by region. Apart from South Asia, the data are similar across regions, ranging from 14 per cent to 21 per cent. Tables 5.4 and 5.5 report the percentages of countries that had intra-industry trade indexes greater than 0.7, by region and sector in 1970 (Table 5.4) and 1990 (Table 5.5). Two main facts emerge from these tables. First, as the sector totals indicate, no sector is an inter-industry or intra-industry trader by nature. The proportion of countries in which intra-industry trade is considered to predominate in a given sector is similar across all sectors. Second, the regional totals show great variation across regions. South Asia and Latin America started in 1970 with more intra-industry trade sectors than the Middle East and North Africa, East Asia and sub-Saharan Africa. By 1990, however, East Asia had more intra-industry trade sectors than the Middle East and North Africa, South Asia and sub-Saharan Africa.
|Gap 1||Gap 2||Gap 3|
|Middle East and North Africa||1.41||1.45||0.14||3.23||2.42||-0.03||1.19||1.23||0.19|
|Regional Averages||Mean||Std. Dev.||Mean||Std. Dev.||Mean||Std. Dev.|
|Middle East and North Africa||0.21||0.09||0.14||0.04||0.24||0.11|
|Sector||East Asia||South Asia||Sub-Saharan Africa||Middle East/|
|Food, beverages, and tobacco||37.5||40.0||42.9||37.5||36.4||37.5||38.6|
|Textiles, wearing apparel and leather||0.0||0.0||9.5||37.5||18.2||54.2||25.0|
|Wood and wood products||25.0||20.0||9.5||0.0||18.2||25.0||17.1|
|Paper, printing, and publishing||25.0||0.0||0.0||0.0||4.6||37.5||13.6|
|Non-metallic mineral products, except||0.0||20.0||14.3||0.0||18.2||29.2||17.1|
|Basic metal industries||0.0||20.0||14.3||12.5||13.6||37.5||19.3|
|Fabricated metal products||12.5||0.0||4.8||0.0||4.6||41.7||14.8|
Includes the 24 OECD countries in the sample, plus Israel.
Includes the 24 OECD countries in the sample, plus Israel.
|Sector||East Asia||South Asia||Sub-Saharan Africa||Middle East/|
|Food, beverages, and tobacco||50.0||60.0||38.1||25.0||45.5||54.2||45.5|
|Textiles, wearing apparel and leather||37.5||20.0||33.3||62.5||27.3||50.0||38.6|
|Wood and wood products||25.0||0.0||14.3||0.0||22.7||37.5||21.6|
|Paper, printing, and publishing||25.0||0.0||9.5||12.5||9.1||54.2||22.7|
|Non-metallic mineral products, except||62.5||20.0||9.5||25.0||22.7||58.3||33.0|
|Basic metal industries||37.5||0.0||4.8||0.0||18.2||62.5||26.1|
|Fabricated metal products||62.5||0.0||0.0||12.5||4.6||66.7||26.1|
Includes the 24 OECD countries in the sample, plus Israel.
Includes the 24 OECD countries in the sample, plus Israel.
The structure of the data is as follows. The data for 1970-93 were split into five sub-periods: 1970-74, 1975-79, 1980-84, 1985-89 and 1990-93. The use of five-year intervals helps to smooth business-cycle effects and to isolate long-run evolutions. The dependent variable in the regressions is measured as the average annual TFP growth for each sub-period. The explanatory variables, however—the technological gap and the ratio of imports to GDP—are measured as the beginning-of-period values instead of the five-year averages. This helps minimise the risk of endogeneity. Because the time dimension of the panel is relatively small compared with the number of countries, one can ignore time-series issues, for which the techniques have not yet been fully developed for panel data.
To test the robustness of the results across regions, each equation was estimated first for the total sample and then by region. The two main regions considered were the OECD countries and the developing countries. The developing countries were further disaggregated into East Asia, Latin America, the Middle East and North Africa, South Asia and sub-Saharan Africa. The estimates for the full sample are reported both with and without country-specific fixed effects. For the regional estimates, fixed effects are included only when an F test indicated they were necessary. The F-test statistics are also reported in the tables. All estimates have heteroskedastic-consistent standard errors noted in parentheses.
The equations were estimated for each of the three TFP measures in the data section. Only the results for TFP1 are reported, however, because the results for the alternative measures of TFP were similar. Table 5.6 reports the estimation results of equation 1. The TFP gap, weighted by the share of imports from OECD countries in GDP, enters significantly in most regressions, confirming the finding by previous studies that trade with OECD countries plays an important role in the transfer of technologies. The model holds not only for the full sample but also for most of the regions4. The results suggest it is important to control for initial conditions that might affect the TFP growth potential of countries. Indeed, the results are stronger when country-specific fixed effects are introduced or when the regressions are estimated by region. For instance, in the regression for the total sample, the adjusted R2 increases from 0.006 without fixed effects to 0.18 when fixed effects are included. The size of the coefficient on the import-weighted gap also increases considerably, from 0.01 to 0.10. Similarly, the adjusted R2 and the size of the coefficient on the import-weighted gap are much larger for the regional regressions than for the full-sample regression without fixed effects.
|R2||Adjusted R2||F test, no|
|Middle East and North Africa||-0.021||0.751||0.180||No||0.355||0.320||0.398|
The difference between the two sets of results can be interpreted in terms of unconditional versus conditional convergence. The regression in the full sample without fixed effects assumes that all countries are converging towards the same steady-state level of technological development and measures the speed of convergence towards this unconditional steady state. In controlling for fixed effects or estimating the regression by region, however, countries are allowed to have their own, different steady states and the regression measures the speed of convergence towards them—hence the term “conditional convergence”. As the results show, conditional convergence is much faster than unconditional convergence.
In sub-Saharan Africa, the fixed effects are negative, suggesting that the region is characterised by conditions which, if unchanged, will prevent it in the long run from attaining the level of technological development that the OECD countries have achieved. Its steady-state level of technology, conditional on these factors, is lower.
Next, the ratio of imports from OECD countries to GDP is divided into two sub-aggregates. One groups imports in sectors classified as intra-industry traders and the other groups imports in sectors classified as inter-industry traders. Table 5.7 reports the estimation results of equation 2, for a range of cut-offs for the IIT index. First, the coefficient on IA (the term that interacts the import shares of inter-industry sectors with TFP gaps) is consistently larger than the coefficient on IR (the term that interacts the import shares of inter-industry sectors with TFP gaps). The difference between the two coefficients becomes more significant as the cut-off for the IIT index is raised. Table 5.7 also shows from the F tests of the null hypothesis that the coefficients are not significantly different.
|F test, β = γ||11.123||6.368||7.417||4.790||1.045||1.709||2.139||0.602||0.787|
Second, as the cut-off is raised, the coefficients on both IA and IR increase. The coefficient on IR at first holds stable at about 0.077, until the cut-off for the IIT index is raised above 0.7, when it starts increasing. The coefficient on IA, however, increases continuously. This pattern corresponds to the one described in Figure 5.1c. Both results indicate that intra-industry trade is a more efficient channel of technology transfer than inter-industry trade, and transfers through trade start increasing dramatically when the sector’s IIT index rises above 0.7, which appears to be the appropriate cutoff separating intra- and inter-industry trade sectors.
Table 5.8 reports the entire estimation results of equation 2 for a cut-off of 0.7 for the IIT index. Note that the coefficient on TFP growth in OECD countries has a point estimate close to one, as the theoretical model predicts. The null hypothesis that the coefficient is one cannot be rejected and the coefficient is generally significantly different from zero. Regarding the respective roles of intra- and inter-industry trade, the coefficient on intra-industry trade is three to four times larger than the coefficient on inter-industry trade, and significantly so. The results for the total sample are confirmed both for developing and OECD countries, but more strongly for developing countries. Among the latter, the results are particularly strong for sub-Saharan Africa. The difference between intra- and inter-industry trade takes a different form in East Asia, with a non-significant effect of IA but a significantly negative effect of IR. Thus, the null hypothesis that the two coefficients are the same can also be rejected with confidence.
|C||α||β||γ||Fixed Effects||R2||Adjusted R2||F test, no fixed effects||F test β = γ|
|Middle East and North Africa||-0.021||0.753||0.180||0.181||No||0.355||0.301||0.386||0.000|
Table 5.9 tests the robustness of these results by excluding net exporters from the inter-industry trade category. The classification of sectors as net importers or net exporters is based on their export-to-import ratios, with cut-offs at 0.5 and 1.9, corresponding to the cut-off of 0.7 for the IIT index. A sector is classed as a net importer if its ratio is below 0.5, indicating that it has no production base (NB). It is classed as a sector with intra-industry trade if its ratio falls between 0.5 and 1.9, indicating the existence of a production base (B). It is called a net exporter if its ratio is greater than 1.9, indicating a strong production base (G). Confirming a priori expectations, the coefficient on G is negative or not significant. The results for NB and B are similar to those obtained previously for inter- and intra-industry trade, confirming the greater importance of intra-industry trade.
|c||α||β||γ||δ||Fixed Effects||R2||Adjusted R2||No Fixed Effects||β = γ|
Table 5.10 reports the estimation results for the non-linear specification, the continuous equivalent of the cut-off-based approach. Instead of dividing the sectors into two subgroups based on their IIT index values, the imports of each sector are weighted by some—possibly non-linear—function of the IIT index. The index is entered in the form of a second-order polynomial, whose coefficients are estimated freely. The regression for the total sample when fixed effects are included clearly indicates a positive and increasing influence of the IIT index on TFP growth. The coefficient on the linear term y is negative but not significant, while that on the squared IIT index 8 is positive and strongly significant. Restricting the sample to developing countries or to OECD countries yields the same pattern of results, although less strongly for the OECD countries.
Conclusions and Policy Implications
This chapter has investigated the role of international trade in transferring technology from industrial to developing countries. It tested the hypothesis that intra-industry trade is more effective in transferring technology than is inter-industry trade. The rationale for this hypothesis is that a country is more likely to absorb the innovations embodied in foreign technology when it already produces and exports goods in the same product category as those it imports.
The chapter takes a general framework already developed by researchers and modifies it to test for the effects of inter-industry trade versus those of intra-industry trade. The tests used data for the absorption of technology (measured by growth of TFP) and trade of 87 countries during 1970-93. Of the countries in the sample, 20 were in sub-Saharan Africa.
|Coefficients||Fixed Effects||R2||Adjusted R2||F test,|
|Middle East and North Africa||-0.028||1.118||0.278||-0.979||1.253||No||0.372||0.300||0.900|
The tests, for both the full sample and the subgroup of 20 African countries, confirmed earlier research, which showed that developing countries acquire technology by trading with industrial countries. The findings indicate that, other factors being constant, developing countries that imported more from OECD countries (as measured by their import-to-GDP ratios) experienced faster TFP growth. The wider the initial technology gap, the larger the gain. Thus, countries technologically farther behind in 1970 gained more than those technologically more advanced.
Intra-industry trade played a larger and more significant role in transferring technology than did inter-industry trade. TFP growth was much more pronounced when a sector’s IIT index exceeded 0.7, a finding even more strongly evident in the subgroup of 20 African countries. The 0.7 cut-off for the IIT index was used to differentiate sectors according to their export/import intensity (XIM). Both the import-intensive sectors (with XIM< 0.5) and the export-intensive sectors (with XIM> 1.9) had an IIT below 0.7, while sectors with more significant two-way trade (0.5 < XI M< 1.9) had an IIT above 0.7. Tests repeated without data from export-intensive sectors reconfirmed these findings. This data exclusion was justified because export-oriented industries presumably are more advanced technologically and thus have less need to adopt the technologies of their import sectors.
Test results also showed the existence of country-specific factors that could prevent sub-Saharan Africa from attaining the same steady-state level of technological development as have OECD countries, but the coefficients calculated from the tests could not identify the precise factors. Nonetheless, the general economic literature suggests several factors that might affect a country’s long-run equilibrium level of technology. These factors may be grouped under “general productivity parameters”; they include political stability, institutional environments and human capital.
One can draw several policy implications from these results, including confirmation of the case for accelerating trade liberalisation to encourage technology transfers. The following recommendations emerge:
— Developing countries, when negotiating trade agreements with industrial countries, should seek to reduce trade barriers in sectors with high IIT at the outset of liberalisation. This is contrary to current developing-country practices, which usually seek to retain trade protection for goods these countries produce. The findings here suggest, however, that rapid liberalisation of such sectors offers greater benefit to developing countries.
— Developing countries should adopt domestic policies that actively promote intra-industry trade. This may include both devising policies to provide key infrastructure or vocational training that will enhance production and exports in new sectors, and adopting measures to encourage foreign direct investment (FDI). As other researchers have argued, FDI may lower the cost of adopting and producing new technologies, because foreign investors likely will already be familiar with them. Thus, FDI may lower the cost of producing and exporting new goods.
— Finally, developing countries should focus on identifying the specific factors that can prevent them from reaching their technological potential and adopting remedial actions. Policy reform itself would need to take a co-ordinated approach to address the entire mix of policies rather than focus on sequential change.
East Asia: China, Chinese Taipei, Indonesia, Malaysia, Philippines, Republic of Korea, Singapore and Thailand.
South Asia: Bangladesh, India, Myanmar, Pakistan and Sri Lanka.
Industrial countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom and United States.
Middle East and North Africa: Algeria, Cyprus, Egypt, Iran, Jordan, Malta, Morocco and Tunisia.
Latin America: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Trinidad and Tobago, Uruguay and Venezuela.
Sub-Saharan Africa: Cameroon, Côte d’lvoire, Ghana, Kenya, Madagascar, Malawi, Mali, Mauritius, Mozambique, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, Sudan, Tanzania, Uganda, Zaire, Zambia and Zimbabwe.
The definitions and the original sources of the data for each variable needed to measure total factor productivity (TFP), as described in the paper by Bosworth, Collins and Chen (1995), are listed below.
GDP: Definition: local currency, 1987 constant prices. Primary sources: OECD for the industrial countries, World Bank and IMF for the developing countries.
Stock of physical capital: Definition: local currency, 1987 constant prices. The measure of the capital stock is based on a perpetual inventory estimation with a common fixed annual geometric depreciation rate of 0.04. Primary source: Nehru and Dhareshwar (1993).
Labour force: Definition and source: actual employment for the industrial countries and estimates from the International Labour Organisation of the economically active population for developing countries.
where H denotes the stock of human capital, w denotes the wage weight of people at the jth education level, and Pj denotes the fraction of the population in theyth education level. The wage weights are standardised at 1.0 for those who have completed the primary level of education. The relevant wage weights are 0.7 for no schooling, 1.4 for completion of the secondary level, and 2.0 for completion of the third level. Note that the few studies that have examined the structure of relative wage rates by education find surprisingly little variation across countries. Source: Barro and Lee (1993) for the fractions of the population at the different education levels.
The authors thank Geert Almekinders, Robert Barro, Ehsan Choudhri, David Coe, Roland Daumont, Samir El-Khouri, Dominique Gross, Elhanan Helpman, Suheil Kawar, Mohsin Khan, Saleh Nsouli, Jean-François Ruhashyankiko and Abdelhak Senhadji for helpful comments. The authors are also grateful to Barry Bosworth for providing the data on output, physical capital, labour and education.
TFP is defined as the log of output minus the weighted logs of factor inputs, where the weights equal factor shares.
These data are provided in the Penn World Tables.
The absence of significant results for East Asia and South Asia might be due to the small sample size for these regions.
AgmonT. (1979) “Specialization in the European Coal and Steel Community” Journal of Common Market Studies Vol. 8.
BarroR.J. and J.-W.Lee (1993) “International Comparisons of Educational Attainment” Journal of Monetary Economics Vol. 32December.
BarroR.J. and X.Sala-i-Martin (1995) Economic GrowthMcGraw-HillNew York.
BayoumiT.D.T.Coe and E.Helpman (1996) “R&D Spillovers and Global Growth” NBER Working Paper No. 5628National Bureau of Economic ResearchCambridge, Massachusetts.
BosworthB.S.Collins and Y.-C.Chen (1995) “Accounting for Differences in Economic Growth” Brookings Discussion Papers in International Economics Vol. 115Brookings InstitutionWashington, D.C.
CoeD.T. and E.Helpman (1995) “International R&D Spillovers”. European Economic Review Vol. 39February.
CoeD.TE.Helpman and A.W.Hoffmaister (1997) “North-South R&D Spillovers” The Economic JournalJanuary.
CoeD.T. and A.W.Hoffmaister (1998) “North-South Trade: Is Africa Unusual?” IMF Working Paper 98/94International Monetary FundWashington, D.C.
FeenstraR.C.R.E.Lipsey and H.P.Bowen (1997) “World Trade Flows, 1970–1992, with Production and Tariff Data” NBER Working Paper No. 5910National Bureau of Economic ResearchCambridge, Massachusetts.
GreenawayD. and C.Milner (1986) The Economics of Intraindustry TradeBasil BlackwellNew York, N.Y.
GrossmanG.M. and E.Helpman (1991) Innovation and Growth in the Global EconomyMIT PressCambridge, Massachusetts.
International Monetary Fund (1997) World Economic OutlookWorld Economic and Financial SurveysWashington, D.C.
JaumotteF. (1998) “Technology Diffusion and Trade: An Empirical Investigation”unpublished, Harvard UniversityCambridge, Massachusetts.
KellerW. (1997) “Trade and the Transmission of Technology” NBER Working Paper No. 6113National Bureau of Economic ResearchCambridge, Massachusetts.
NehruV. and A.Dhareshwar (1993) “A New Database on Physical Capital Stock: Sources, Methodology and Results” Revista de Anàlisis Economico Vol. 8June.