77381 the world bank economic review, vol. 16, no. 1 23–48 Imported Machinery for Export Competitiveness Ashoka Mody and Kamil Yilmaz This article analyzes the relationship between export competitiveness and investment in machinery, allowing for imperfect substitution between domestically produced and imported machinery. A translog export price function is estimated for developed, export- oriented developing, and import-substituting developing economies in a panel data setting. Between 1967 and 1990 imported machinery helped lower export prices for export-oriented developing economies. Moreover, throughout the period imported machinery was not a substitute for domestic machinery. Import-substituting developing economies were unable to harness imported machinery to reduce costs early in the period, but from about the early 1980s, with the opening of their trade regimes, they were able to benefit from the cost-reducing effect. The results imply that innovative effort based on imported technologies can be a precursor to the development of domestic innovation capabilities. In this article we build on two recent lines of research: investment in equipment as a source of economic growth and imported goods as conduits for the interna- tional diffusion of technology. We combine these two themes to assess the effec- tiveness of imported machinery in increasing export competitiveness and thus in stimulating growth.1 Underlining the importance of machinery in the development process, De Long and Summers (1991, 1992a, 1992b, 1993) find strong empirical support for a causal relationship between equipment investment and economic growth in a cross-section of developing and developed countries. In particular, they find that a 1% increase in the share of equipment investment in gross domestic product Ashoka Mody was at the World Bank when this paper was written and is presently in the Research Department at the International Monetary Fund; his e-mail address is amody@imf.org. Kamil Yilmaz is with Koç University, Istanbul, Turkey; his e-mail address is kyilmaz@ku.edu.tr. The authors grate- fully acknowledge extensive comments from two referees and the editor. The views expressed here are those of the authors and should not be attributed either to the International Monetary Fund or the World Bank. 1. Several studies suggest that greater trade is associated with faster productivity growth (for ex- ample, Pack and Page 1994, Srinivasan 1995, 1999). However, Rodriguez and Rodrik (1999) and Rodrik (1999) are skeptical of such results when they are based on cross-sectional growth regressions. Srinivasan and Bhagwati (1999) express general concern about cross-sectional growth analyses and conclude that “nuanced and in-depth studies� of individual countries over a period of time provide the clearest evi- dence in favor of the beneficial effects of greater trade orientation. © 2002 The International Bank for Reconstruction and Development / THE WORLD BANK 23 24 the world bank economic review, vol. 16, no. 1 (gdp) raises the gdp growth rate by 0.34%.2 They infer that the domestic re- search and development (R&D) and learning activities associated with produc- ing and installing equipment create generalized benefits for the economy. The De Long–Summers results are also consistent with the possibility that the foreign knowledge embodied in imported equipment is of significant value to the economy buying the equipment. Studying the spillovers of knowledge across na- tional boundaries, Coe and Helpman (1995) and Engelbrecht (1997) find that such international spillovers are mediated through imported goods: the greater the imports, the greater the benefit from the stock of foreign knowledge. Engelbrecht notes that these two articles do not distinguish between different types of imports; such a distinction is likely to be important because consumer goods, intermediate inputs, and equipment are likely to convey spillovers to differing degrees. Extend- ing these earlier studies, Coe, Helpman, and Hoffmaister (1997) find that imported capital goods are critical conduits of international knowledge. In this article we examine empirically the different efficiency of domestically produced and imported machinery. With information freely accessible to all there would be no difference in efficiency. In practice, however, information is not freely available. Even in the absence of formal protection of intellectual prop- erty rights, domestic producers can be at a disadvantage relative to international producers because knowledge is tacit (for a review of the economic and man- agement literature on tacitness see, for example, Mody 1989). As a consequence, domestic and imported machinery trigger different forms of learning in the do- mestic economy. Domestic production and installation of machinery may in some instances be associated with considerable innovative activity in a developing economy. More often, however, the domestic production of machines is associ- ated with adaptive R&D—that is, tailoring foreign machinery to local require- ments and upgrading domestic equipment of earlier vintages. In contrast, imported machinery is bundled with “knowledge� in various forms: blueprints, installation support, quality control software, and services of trained engineers and supervisors. Absorbing knowledge through these means is less glamorous than developing or even adapting machines. However, because imported machinery forms a more comprehensive package, it can potentially lead to greater efficiency in the short run and stronger absorptive capacity in the long run. Imported machinery will also be more efficient because it is typically of newer vintage than domestically produced machinery.3 2. In the traditional neoclassical model an increase in the investment rate raises output but has no long-run effect on growth rates. The endogenous growth literature identifies conditions under which increased investment has external effects and thus raises growth rates. De Long and Summers go fur- ther and find evidence that the external effects are strongest when the investment is in machinery rather than in buildings and structures. 3. Other mechanisms for knowledge transmission can also be important, including the provision of technical and marketing support by foreign buyers of exported goods in the context of long-term rela- tionships (Westphal, Rhee, and Pursell 1981, Egan and Mody 1992, Mody and Yilmaz 1997). Mody and Yilmaz 25 We use the country’s trade regime to proxy the incentives to deploy knowl- edge (table 1). Not all economies (or firms) are able to take advantage of bundled software and training or the greater efficiency built into the new vintages of imported machinery. In countries with a strong export orientation, however, firms are likely to have strong incentives—driven by the need to stay competitive—to exploit the knowledge flows associated with imported machinery. Import sub- stitution was based on the premise that, with temporary protection, domestic producers would have the incentive to tap into and internalize internationally available knowledge. The extent to which this actually occurred, however, is an empirical issue. Incentives in more protected “import-substituting� economies were likely to be weaker because of the relatively small size of domestic markets and the less demanding domestic users of that machinery (see Srinivasan and Bhagwati 1999 for a review of how incentives are blunted in an import substi- tuting regime). Thus, although alternative explanations are possible, the results of our analysis are consistent with the proposition that import-substituting re- gimes create weaker incentives to invest in technological improvements that can help expand their export presence in international markets. Plotting the change in the volume of exports against the change in the capital stock in the previous year, we find that in export-oriented developing econo- mies an increase in exports is strongly associated with an increase in the stock of imported machinery (figure 1). A positive relation also exists between export growth and an increase in the stock of domestic machinery. A similar set of re- lationships is found for developed countries. In contrast, for import-substituting developing economies imported equipment and export growth are negatively related. Our empirical analysis focuses on the price of exports rather than on their volume, specifying a link from imported machinery to reduced costs and prices, which in turn leads to greater exports. We analyze the relationship between machinery investment and export competitiveness using a model of imperfect competition in international markets, allowing for imperfect substitutability between domestically produced and imported machinery. We specify an export Table 1. Trade Regimes and the Effects of Imported and Domestic Machinery Trade regime Export-oriented Import substituting Imported machinery Has access to the pool of inter- Can access the pool of inter- national knowledge and the national knowledge, but small incentives to exploit it. domestic markets may blunt incentives. Domestic machinery With strong incentives, can over- Both the knowledge pool and come the limits of the domestic incentives may be limited. knowledge pool as domestic capa- city improves. 26 Figure 1. Growth of Manufactured Exports on Growth of Total, Imported, and Domestic Machinery Stock D evelo ped C o untries Expo rt-o riented develo ping eco no mies Impo rt s ubs tituting develo ping 120 eco no mies 35 6 100 30 5 80 25 4 60 20 3 40 2 20 15 1 0 10 0 -20 5 -1 -40 0 -2 -60 -5 -3 -40 -20 0 20 40 60 80 0 10 20 30 40 50 0 5 10 15 20 25 Lagged to tal sto ck o f machinery (%) Lagged to tal sto ck o f machinery (%) Lagged to tal sto ck o f machinery (%) D evelo ped C o untries Expo rt-o riented develo ping eco no mies Impo rt s ubs tituting develo ping 120 eco no mies 35 6 100 5 30 80 4 25 60 3 20 40 2 26 15 20 1 0 10 0 -20 5 -1 -40 0 -2 -60 -5 -3 the world bank economic review, vol. 16, no. 1 -10 0 10 20 30 0 10 20 30 40 -5 0 5 10 15 Lagged impo rted sto ck o f machinery (%) Lagged impo rted sto ck o f machinery (%) Lagged impo rted sto ck o f machinery (%) D evelo ped C o untries Expo rt-o riented develo ping eco no mies Impo rt s ubs tituting develo ping 120 eco no mies 35 6 100 30 5 80 25 4 60 20 3 40 2 15 20 1 0 10 0 -20 5 -1 -40 0 -2 -60 -5 -3 -40 -20 0 20 40 60 80 0 5 10 15 20 0 5 10 15 20 Lagged do mestic sto ck o f machinery (%) Lagged do mestic sto ck o f machinery (%) Lagged do mestic sto ck o f machinery (%) Source: Authors’ calculations, based on data from U.N. Commodity Trade (Comtrade) database. Mody and Yilmaz 27 price function based on the demand for exports and the costs of producing the exported goods. We use a short-run cost function with variable labor and mate- rials costs and fixed stocks of imported and domestically produced machinery. Larger stocks of capital are expected to lower short-run production costs and thus export prices. Higher productivity of imported machinery would be reflected in the greater cost reduction than can be achieved using domestic machinery. We estimate the export price equation for developed countries, for develop- ing economies, and, within the second group, for export-oriented and import- substituting economies. For each group we use the fixed-effects procedure on panel data. But first we use the t-bar test recently developed by Im, Pesaran, and Shin (1996) to identify (in a panel data context) the presence of stochastic trends in export price and explanatory variables. We find that the variables do have stochastic trends (unit roots). Thus, because of the potential colinearity in the movement of variables of interest, we estimate the export price function in first differences. In the empirical application, to allow for lagged effects, we distinguish be- tween the effects due to the past year’s new investment and those due to the stock of capital at the start of the previous year. For the entire sample period, 1967– 90, the flow of new imported equipment in the previous year is seen to be asso- ciated with a decline in export prices in developed and export-oriented develop- ing economies, but not in import-substituting developing economies. The results also show that the relation between imported equipment and export prices has evolved over time. Throughout the period, we find that domestic machinery is not a substitute for imported machinery. I. A Model of Imperfect Competition in Export Markets In setting up the model we are guided by the following intuition: the significance that some developing countries, especially the newly industrializing countries of East Asia, attached to investment in machinery was not accidental. Instead, it was dictated by the adoption of an export-oriented strategy and the resulting discipline of international competition. To maintain market presence, exporters had to reduce production costs continually or enter into the production of higher- quality goods. Both strategies required substantial investment in new vintages of machinery and equipment. Initially, domestic machinery had lower produc- tivity, so the scope for substituting domestic for imported machinery was small. Over time the more advanced developing countries progressed to the point where their technological capability to produce machinery could compete with imports from developed countries. This intuition can be tested by estimating a cost function that includes machin- ery stock as an explanatory variable. However, data on production costs are dif- ficult to obtain. For this reason we estimate an export price function based on both demand and cost function parameters. In a model of imperfect competition, manu- facturers arrive at their export price given demand and cost conditions. Though 28 the world bank economic review, vol. 16, no. 1 demand depends on competitors’ prices and on incomes in target markets, pro- duction costs depend on input prices, output levels, and other variables that shift the cost function, such as the stocks of imported and domestic machinery. We assume that production for domestic and export markets are two inde- pendent decisions and focus on exports.4 Firms produce export goods through a homothetic production function using two variable inputs, labor and materi- als (including fuel, electricity, and raw materials) and a quasi-fixed input, the capital stock. Firms are assumed to be price takers in input markets. Conse- quently, the short-run cost function can be separated into variable input prices on the one hand and the quasi-fixed input and output on the other. We assume that each firm exports a differentiated product and chooses its export price to maximize its profit at a point in time, given the demand curve for its product and the cost of production.5 When the second-order condition for maximizing profit is satisfied, it is possible to solve for the profit maximizing price, by inverting the first-order condition. The profit maximizing price is a function of all variables that enter the cost function—the wage rate (w), price of materials (pm), and capital stock (K)—plus variables that shift the demand func- tion, the competitors’ average price (pc) and world income (Y). The export price may also be a function of the exchange rate (e), as will be discussed. (1) p = p(pc,Y,w,pm,e,K) The elasticities of the export price with respect to variable input prices, the prices of competing products, and the capital stock depend on the parameters of the cost and demand functions. When the second-order condition for maximiz- ing profit is satisfied, a positive elasticity of the marginal cost with respect to input prices is sufficient to generate a positive elasticity of the export price with respect to input prices. In other words, the exporter will increase its price fol- lowing an increase in input prices. With the second-order condition satisfied, a decreasing marginal cost with an increasing machinery stock is both a necessary and a sufficient condition for the 4. For a similar assumption and empirical implementation see Feenstra (1989). If the marginal costs of production for the domestic market and export markets are not flat, influences in one market will influence the other. Essentially, an omitted variable bias would arise where the omitted variables refer to demand influences in the domestic economy. If domestic demand were to shift exogenously, the marginal costs of production would change, leading to a change in prices charged in both the domestic and the international markets. We believe that these exogenous shifts would be reflected in the prices of domestic inputs (wages and materials costs). A bias may still remain, however, though the direction of it is unclear. If increased domestic activity leads to more investment but also higher marginal costs, a larger stock of private capital would be associated with higher export prices—the opposite of the relationship that we are hypothesizing. 5. Because the analysis is restricted to the cost-reducing effect of the technology embodied in exist- ing machinery, the model is static and does not incorporate investment demand for domestic and im- ported machinery. Analytically it is not difficult to incorporate the demand for machinery through a dynamic model. However, because of a lack of data on production costs and the rental price of capital stock, it is not possible to estimate factor demand functions of the long-run model. Mody and Yilmaz 29 price to be a decreasing function of the machinery stock. Consequently, if the estimated price elasticity with respect to the machinery stock is negative, it fol- lows that the technology embodied in new machinery has a cost-reducing effect. For the purpose of empirical estimation and following Mann (1986, 1989), we simplify the demand function by substituting the world price (pw) for the competitors’ price (pc) and world income (Y). The world price variable reflects the influence of the pricing decisions of all competitors and of changes in world income. Thus, using the reduced-form price equation, we analyze the elasticity of the export price with respect to the world price, two input prices, and the two kinds of machinery stock.6 In the empirical analysis we assume that the export price decision is best sum- marized by the translog price function (2) log p = λ + Σ β i logXi + 0.5Σψi(logXi)2, + Σ Σ ψ i,j (logXi log Xj), i i i j >i where Xi = pw,w,pm,e,I–1 m ,I d ,Km ,K d . Id is the investment flow for domestic capi- –1 –2 –2 m tal goods, I is the investment flow for imported capital goods, and Kd and Km are the corresponding stocks. A variable with the subscript –1 is lagged one pe- riod; the subscript –2 implies a two-period lag. By considering the past year’s investment and the stock prior to that, we are able to obtain some sense of the lags with which the effects operate.7 Previous studies analyzing export price behavior under imperfect competition have noted that exchange rates often exercise an independent influence on the price of traded goods. In other words, even if all variables on both sides of the equation are measured in the same currency, exchange rate movements seem to have a significant effect on the price of exports (Feenstra 1989, Ohno 1989, Mann 1986). By representing the input prices in local currency terms and including the exchange rate as a separate variable, we allow for the possibility that changes in exchange rates are not perfectly passed through to export prices. Our primary results remain unchanged if we instead measure the input prices in dollars and drop the exchange rate variable. II. Empirical Specification and the Data We estimate the export price equation for a cross-section of 14 developed coun- tries and 25 developing economies. (For the definitions of variables and the data sources, see appendix table A-1; for the descriptive statistics, see appendix table A-2.) Because we derive the model based on profit maximizing assumptions for an individual firm, it would be best to use firm- or industry-level data to estimate 6. Local currency wages and the price of material inputs were obtained by dividing the correspond- ing variables denominated in U.S. dollars by the annual average exchange rate. 7. Of course, we are not decomposing the stock of capital in a strict sense, with this year’s stock equal to new investment plus the previous stock. That simple identity does not carry forward when we take logs. 30 the world bank economic review, vol. 16, no. 1 the price function in equation (2). However, data constraints preclude that route. Though data on export prices, input prices, and investment can be found for some manufacturing subsectors in some countries, it is not possible to obtain data on domestic and imported components of investment by each industry. We are there- fore forced to aggregate all manufactured exports from a country. Aggregation can be justified by assuming either a representative firm (as in Feenstra 1989, Ohno 1989) or a translog aggregate production technology for manufacturing exporters (Pindyck and Rotemberg 1983). Aggregation presents its own problems, however. The higher the level of aggregation, the more diffi- cult it becomes to obtain price indexes that reflect firm-level pricing decisions. An aggregate price measure incorporates changes in the composition of the com- modity basket as well as in the market price of each commodity in the basket. Is this a problem for our proposed empirical analysis? No, because our focus is on cost reduction. To the extent that changes in the composition of exports from one year to the next are important, the cost-reduction effect will be blurred. Indeed, if products were moving up the quality ladder, we would expect to find no cost-reduction effect. Thus a finding of cost reduction despite that possibility provides somewhat greater confidence in our results. Because our sample of developing economies represents substantially differ- ent development strategies, we divide the economies into two groups, export- oriented and import-substituting, based on the World Bank classification (World Bank 1986; see also Balasubramanyam, Salisu, and Sapsford 1996). Between 1967–73 and 1973–85 no major shift occurred in the outward orientation of the developing economies in our sample (appendix table A-3). However, although the economies remained differentiated in their broad policy stance, their trade policy regimes did not remain fixed. The reduction in trade barriers continued apace, with several of the import-substituting economies adopting more export- oriented policies in the 1980s. Thus the differences in policy regimes narrowed. Before estimating the export price function, we test for nonstationarity of the variables using the t-bar statistic proposed by Im, Pesaran, and Shin (1996) for heterogeneous panels. This is a well-known crucial first step in time-series mod- els. When a time-series equation contains a nonstationary variable, the results based on this estimation will be spurious. Im, Pesaran, and Shin (1996) recently extended the stationarity tests to cross-section, time-series models. The test procedure is simple. It extends the widely used augmented Dickey- Fuller (adf) test to a panel data framework and allows for heterogeneity across groups in the panel. First, the average adf unit root test statistic for the panel is obtained as the mean of individual adf unit root statistics. Next, the expected value and the standard error of the average adf test statistic under the null hypothesis of a unit root are obtained through Monte Carlo simulation. The t-bar statistic is calculated as the average adf test statistic minus its expected value divided by its standard error. Im, Pesaran, and Shin (1996) show that under the null hypothesis of a unit root, the t-bar statistic has a standard normal distribution for a sufficiently large number of countries, N, and number of periods, T, while Mody and Yilmaz 31 √N/T goes to zero. Using the Monte Carlo method, they show that the t-bar test has more power than adf tests applied separately to each individual in the panel. Based on the results of the Im-Pesaran-Shin test, we cannot reject the null hypothesis of a unit root for all variables of the price function for all country groups (appendix table A-4). Consequently, estimating the export price func- tion in levels (equation [2]) would generate spurious results. Next, we test for unit roots in the first-differenced variables and reject nonstationarity. This al- lows us to estimate the equation in first differences. III. Empirical Results We estimate the first-differenced export price equation using the fixed-effects procedure. This amounts to assuming that countries do differ in terms of the trend coefficient, which could be interpreted as disembodied technical change.8 For all country groups in our analysis we estimate the translog parameters using the data for 1967–90 (table 2). The specification test for functional form indicates that the translog function provides a better approximation of the ex- port price decision than does the Cobb-Douglas function. However, the parameters of the translog function cannot be interpreted di- rectly. Instead, one needs to derive the elasticity estimates of the export price function with respect to input prices, the exchange rate, and imported and do- mestic machinery using the underlying parameters of the translog function. These elasticities take the following form: (3) Ei = βi + ψi logXi + jΣ ψ logXj, ≠i i,j where Xi = pw,w,pm,e,I–1 m ,Id ,Km ,Kd and a bar over a variable denotes its average –1 –2 –2 value for the country group throughout the sample period. The standard error of each elasticity is estimated using the δ method (for a more detailed treatment see Rao 1973, p. 388–90). One can write the elasticities in the following matrix notation: E = Z Ψ , where Ψ is the 44 × 1 vector of translog function parameters and Z is an 8 × 44 matrix of zeros, ones, and the means of log variables, as given in equation (3). Using this matrix notation, we obtain the variance-covariance matrix of the elasticity matrix E, ΣE = ZΣΨ Z', where Σ Ψ is the variance-covariance matrix of the parameter estimates, excluding the intercept. In the rest of the article the results focus on the elasticities. We present the results in two parts. First, we discuss the results for the full sample period, 1967– 90, the period for which we have complete data for the variables of interest.9 8. Alternatively, one could assume that individual effects occur randomly rather than being fixed. This implies that the individual effect is part of the random disturbance rather than the constant term specific to a country. However, this assumption is not justified here because we did not sample countries randomly. 9. The binding data constraint is imposed by the use of machinery investment data from the Penn World Tables, which end in 1990. (See Heston and Summers, 1991.) Table 2. Export Price Equation: Translog Estimates, 1967–90 Export-oriented Import-substituting Parameter Developed Countries Developing economies developing economies developing economies βPw 2.418 (1.62) –2.595 (1.39)* –1.951 (2.31) –4.187 (2.02)** βw –0.517 (0.89) –0.003 (0.52) –1.536 (1.23) 0.286 (0.78) βPm 1.608 (1.15) 1.878 (0.35)*** 2.451 (0.59)*** 1.798 (0.56)*** βe –0.340 (1.71) –2.140 (0.64)*** –1.221 (1.52) –2.170 (0.97)** βIm –1.699 (0.73)** 1.156 (0.35)*** 0.990 (0.78) 1.936 (0.49)*** βId –0.124 (0.46) 0.474 (0.32) 0.676 (0.45) 0.414 (0.48) βKm 2.782 (1.33)** –1.446 (0.79)* –2.389 (1.47) –3.146 (1.43)** βKd 0.428 (0.68) –0.652 (0.98) 0.696 (1.53) 0.554 (1.57) ΨPw –0.217 (0.27) 0.319 (0.31) 0.352 (0.46) 0.589 (0.47) Ψw –0.138 (0.14) –0.116 (0.05)** –0.255 (0.13)** –0.098 (0.08) ΨPm 0.226 (0.18) 0.175 (0.04) 0.259 (0.05)*** 0.161 (0.07)** Ψe 0.392 (0.31) 0.135 (0.07)* –0.039 (0.20) 0.203 (0.11)** ΨIm 0.072 (0.08) 0.026 (0.05) –0.059 (0.06) 0.201 (0.09)** 32 ΨId –0.010 (0.02) 0.011 (0.01) 0.004 (0.02) 0.010 (0.02) ΨKm –0.176 (0.18) 0.099 (0.12) 0.037 (0.20) 0.593 (0.25)** ΨKd –0.060 (0.07) 0.076 (0.08) –0.060 (0.12) 0.234 (0.13)* ΨPw,w –0.027 (0.17) 0.037 (0.10) 0.161 (0.17) 0.032 (0.16) ΨðPw,Pm –0.222 (0.18) –0.393 (0.08)*** –0.502 (0.12)*** –0.405 (0.14)*** ΨPw,e 0.094 (0.26) 0.423 (0.14)*** 0.418 (0.24)* 0.424 (0.21)** ΨPw,Im 0.163 (0.10)* –0.172 (0.07)** –0.167 (0.11) –0.306 (0.10)*** ΨPw,Id 0.118 (0.07)* –0.044 (0.06) 0.018 (0.07) –0.011 (0.10) ΨPw,Km –0.012 (0.16) 0.295 (0.11)*** 0.170 (0.19) 0.515 (0.16)*** ΨPw,Kd –0.259 (0.08)*** 0.077 (0.08) 0.061 (0.17) –0.009 (0.12) Ψw,Pm 0.199 (0.09)** 0.033 (0.03) –0.011 (0.06) 0.071 (0.04)* Ψw,e –0.138 (0.13) 0.084 (0.05)* 0.284 (0.15)** 0.024 (0.08) Ψw,Im –0.066 (0.06) 0.085 (0.03)*** 0.118 (0.06)** 0.049 (0.06) Ψw,Id –0.058 (0.05) –0.028 (0.03) –0.031 (0.05) 0.000 (0.05) Ψw,Km 0.095 (0.11) –0.088 (0.04)** –0.097 (0.10) –0.037 (0.10) Ψw,Kd 0.117 (0.08) 0.026 (0.04) 0.113 (0.13) –0.052 (0.09) ΨPm,e –0.302 (0.23) –0.216 (0.05)*** –0.255 (0.08)*** –0.231 (0.08)*** ΨPm,Im –0.053 (0.09) 0.009 (0.03) 0.036 (0.03) 0.018 (0.04) ΨPm,Id –0.062 (0.07) 0.049 (0.02)** 0.052 (0.03)* 0.032 (0.03) ΨPm,Km –0.109 (0.11) 0.005 (0.03) –0.064 (0.05) 0.018 (0.05) ΨPm,Kd 0.111 (0.08) –0.056 (0.02)** –0.026 (0.04) –0.050 (0.03) Ψe,Im 0.249 (0.11)** –0.100 (0.03)*** –0.164 (0.06)*** –0.059 (0.06) Ψe,Id 0.083 (0.07) –0.030 (0.03) –0.033 (0.05) –0.043 (0.06) Ψe,Km –0.209 (0.15) 0.069 (0.05) 0.172 (0.11) 0.010 (0.09) Ψe,Kd –0.122 (0.09) 0.055 (0.04) –0.085 (0.13) 0.098 (0.08) ΨIm,Id 0.013 (0.04) 0.012 (0.02) –0.012 (0.03) 0.003 (0.04) ΨIm,Km –0.009 (0.09) –0.049 (0.06) 0.109 (0.09) –0.184 (0.09) ΨIm,Kd –0.027 (0.05) –0.027 (0.03) –0.081 (0.06) –0.041 (0.06) ΨId,Km –0.056 (0.05) –0.013 (0.02) –0.011 (0.04) –0.032 (0.04) ΨId,Kd 0.029 (0.03) –0.037 (0.02)* –0.065 (0.04) –0.021 (0.03) ΨKm,Kd 0.074 (0.09) –0.021 (0.08) 0.058 (0.12) –0.287 (0.17)* Adjusted R2 0.87 0.35 0.49 0.22 Durbin-Watson statistic 1.95 2.21 2.22 2.21 Sum of squared residuals 0.332 5.044 1.365 3.397 33 Degrees of freedom 268 502 218 240 H1 87.9 [<0.001] 2.13 [0.15] 0.65 [0.42] 0.29 [0.59] H2 2.81 [0.20] 0.03 [0.87] 1.30 [0.26] –0.15 [0.70] H3 34.9 [<0.001] 18.4 [<0.001] 15.5 [<0.001] 12.9 [<0.001] H4 43.1 [<0.001] 11.9 [0.98] 6.4 [0.85] 8.4 [0.76] H5 154.7 [<0.001] 96.2 [<0.001] 101.7 [<0.001] 69.2 [0.001] *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level. Note: Figures in parentheses are heteroskedasticity-consistent standard errors (White 1980). Figures in square brackets are the marginal significance levels for the corresponding hypothesis. H1: World price elasticity is equal to one (joint test for market power and economies of scale). H2: Davidson and MacKinnon J-test: imported and domestic machinery are imperfect substitutes. H3: Davidson and MacKinnon J-test: imported and domestic machinery are perfect substitutes. H4: Country fixed effects do not differ from each other. H5: Cobb-Douglas and translog functional forms do not differ. Source: Authors’ calculations (see appendix table A-1 for data sources). 34 the world bank economic review, vol. 16, no. 1 Next, to undertake a more detailed analysis of the data, we repeat the estima- tions of the translog function for subsample windows, dropping one observa- tion from the beginning of the sample period each time. Full Sample Period: 1967–90 World price elasticity is high when own-price elasticity of demand is high, when there are significant diseconomies of scale in production, or when both condi- tions exist. Indeed, world price elasticity approaches one as own-price elasticity approaches infinity, that is, when the demand curve for the country’s products is infinitely elastic. As expected, the world price elasticity estimate is lowest for developed countries (0.36), which face the least elastic demand curve and where diseconomies of scale are likely to be weakest (table 3). The test result supports the hypothesis that world price elasticity differs significantly from that for de- veloped countries. World price elasticity for developing economies is 0.94, quite close to 1. World price elasticity for export-oriented economies is about the same as that for import-substituting economies and in both cases does not differ sta- tistically from one. The lower wage and material price elasticities for developing economies are consistent with their price-taking role in the world market. A price-taking firm cannot increase its prices to fully reflect increases in unit costs. In contrast, for a firm with market power, which can influence the export price of its prod- ucts, wage and material price elasticity would differ significantly from zero. Wage elasticity is highest for developed countries, at 0.24. Wage elasticity for developing economies is 0.02 and does not differ significantly from zero. This result is driven mainly by import-substituting developing economies. Though their wage elasticity is –0.07 and not significantly different from 0, wage elas- ticity for export-oriented economies is 0.11 and statistically significant. Mate- Table 3. Export Price Equation: Elasticity Estimates, 1967–90 Export-oriented Import-substituting Developed Developing developing developing Elasticity countries economies economies economies EPw 0.355*** (0.069) 0.942*** (0.079) 0.923*** (0.095) 0.925*** (0.139) Ew 0.240*** (0.054) 0.020 (0.030) 0.112*** (0.041) –0.070 (0.066) EPm 0.187*** (0.032) 0.062*** (0.013) 0.062*** (0.016) 0.056** (0.022) Ee –0.712*** (0.050) –0.092*** (0.032) –0.213*** (0.052) –0.022 (0.056) EI m –0.052*** (0.019) –0.024 (0.022) –0.072** (0.033) 0.001 (0.010) EI d –0.016 (0.998) –0.041 (0.480) –0.042 (0.860) –0.058 (0.760) E Km 0.097 (0.115) –0.017 (0.084) –0.029 (0.130) 0.068 (0.185) EKd –0.078 (0.089) –0.170 (0.120) –0.171 (0.150) –0.188 (0.220) **Significant at the 5 percent level. ***Significant at the 1 percent level. Note: Figures in parentheses are heteroskedasticity-consistent standard errors (White 1980). Source: Authors’ calculations (see appendix table A-1 for data sources). Mody and Yilmaz 35 rials price elasticity differs significantly from zero for all groups. It is highest for developed countries, at 0.19, and about 0.06 for all three groups of devel- oping economies. What is the evidence for a cost-reducing role for the stock of machinery? Elas- ticity estimates for the entire period show that imported machinery has a cost- reducing effect for developed countries and export-oriented developing economies. But this effect is significant for the imports of equipment in the past year, not for the stock of imported equipment at the start of the previous year. Thus the evi- dence suggests that the technology embodied in new imported equipment helps competitiveness and, moreover, acts relatively quickly. For export-oriented developing economies the coefficient on the lagged capital stock term is negative but statistically insignificant. This implies that the gains from new investment in imported capital goods are not reversed. For developed countries the coefficient on the lagged capital stock term is positive but never statistically different from zero, implying some persistence in the cost-reducing effects. Do domestic and imported machinery substitute for each other? We use the nonnested Davidson and MacKinnon (1981) J-test to determine whether im- ported and domestically produced machinery are perfect or imperfect substitutes in terms of their cost-reducing effect (for a description of the test see also Greene 1997). If they are imperfect substitutes, we need to consider their cost-reducing effects separately, and the price equation with imported and domestic machin- ery as separate right-hand-side variables (equation [2]) is appropriate. However, if they are perfect substitutes, we need to include their sum, the total stock of machinery, as a right-hand-side variable. The usual nested test does not apply here because an alternative to the null hypotheses cannot be constructed by re- stricting the parameters implied by the null. Because of this property of the model, imperfect and perfect substitution are nonnested hypotheses. The J-test is used in such situations, but because it is a two-way test its use may lead to inconclusive results. In the first stage (hypothesis test H2) imperfect substitution is the null hypothesis and perfect substitution is the alternative hy- pothesis.10 The procedure works as follows. First we obtain the predicted ex- port price under the assumption of perfect substitution (combining domestic and imported machinery to form one capital stock variable). Then we include this predicted export price as an additional variable in the export price estimation under the assumption of imperfect substitution. If the coefficient on the predicted export price variable differs significantly from zero, we can reject the hypothesis of imperfect substitution. The J-test amounts to testing whether the estimate of the dependent variable obtained under the alternative specification of perfect substitution has any explanatory power in the null specification of imperfect substitution for the export price function. If it does, we can reject the hypothesis of imperfect substitution. The p-values in tables 2 and 4–6 refer to the statistical 10. We thank an anonymous referee for suggesting the use of nonnested hypothesis tests. 36 the world bank economic review, vol. 16, no. 1 significance of the coefficient on the predicted price estimated from the alterna- tive hypothesis. Next we take perfect substitution as the null hypothesis and imperfect substi- tution as the alternative and again conduct the J-test (H3). If the test fails to reject the null hypothesis of perfect substitution, we can conclude that the two types of machinery are perfect substitutes. If instead the test rejects the null hypoth- esis of perfect substitution, we need to look at the result of the test in which imperfect substitution is the null hypothesis (H2). If the null hypothesis of im- perfect substitution cannot be rejected, we can conclude that the two types of machinery are imperfect substitutes. In this instance the J-test results are quite clear. The null hypothesis of perfect substitution is rejected, but the null hypothesis of imperfect substitution cannot be rejected even at very high levels of significance. Subsample Windows Considerable changes occurred in the market power and the technology absorp- tion capacity of different countries in 1967–90. To help us study the evolution of elasticity estimates over this period, we use subsample windows regressions. We start with the full sample, 1967–90. Then we drop the observation for 1967 and estimate the model for the subsample 1968–90. Next we drop the observa- tion for 1968 and estimate the model for 1969–90, and so on up to the subsample window 1979–90. In this fashion we obtain 13 different estimates of elasticity. As we move from the first window (1967–90) to the last (1979–90), we obtain a better fit for the regressions (the adjusted R2 increases) for developed coun- tries and for import-substituting developing economies, and the quality of the fit remains relatively unchanged for export-oriented developing economies. The J-test continues to strongly reject the null hypothesis of perfect substitution be- tween imported and domestic machinery but not the null hypothesis of imper- fect substitution. For developed countries the cost-reducing effect of new investment in imported machinery declined quite rapidly, and although the sign continued to be nega- tive in all but one period, by the early 1970s the effect had become statistically insignificant (table 4). Soon thereafter, by the mid-1970s, the stock of domestic machinery had a cost-reducing effect. One could interpret this shift as implying that domestic capabilities matured by the early 1970s in developed countries and thus that the leading edge of the innovation process shifted from a reliance on external sources to a locus in domestic research and adaptation. This does not necessarily mean that domestic machinery embodied more sophisticated tech- nologies than imported machinery. Instead, it suggests that domestic machinery came to play a more central role in a broader process of technological innova- tion, one that had persistent effects. For export-oriented developing economies there was a similar pattern of evolu- tion (table 5). The cost reducing effect of new investment in imported machinery remained statistically significant throughout the period, though there is some sug- Table 4. Elasticity Estimates for Developed Countries, Subsample Windows Hypothesis tests Elasticity estimates (marginal significance levels) Subsample Degrees window pw w pm e m I–1 d I–1 m K–2 d K–2 Adjusted R2 of freedom H1 H2 H3 1967–90 0.355 *** 0.240*** 0.187*** –0.712*** –0.052*** –0.016 0.097 –0.078 0.87 268 [<0.001] 0.09 [<0.001] 1968–90 0.333 *** 0.258*** 0.200*** –0.724*** –0.051*** –0.017 0.061 –0.067 0.87 255 [<0.001] 0.10 [<0.001] 1969–90 0.280 *** 0.269*** 0.223*** –0.766*** –0.053*** –0.019 –0.015 –0.108 0.88 242 [<0.001] 0.97 [<0.001] 1970–90 0.277 *** 0.254*** 0.218*** –0.761*** –0.039** –0.015 0.054 –0.115 0.88 229 [<0.001] 0.93 [<0.001] 1971–90 0.254 *** 0.263*** 0.222*** –0.774*** –0.035* –0.017 0.128 –0.113 0.89 216 [<0.001] 0.19 [<0.001] 1972–90 0.214 *** 0.288*** 0.238*** –0.795*** –0.019 –0.023 0.142 –0.124 0.89 202 [<0.001] 0.04 [<0.001] 1973–90 0.180 ** 0.244*** 0.239*** –0.812*** –0.017 –0.017 0.129 –0.136 0.90 188 [<0.001] 0.10 [<0.001] 1974–90 0.147** 0.187*** 0.250*** –0.831*** –0.009 –0.018 0.216 –0.112 0.90 174 [<0.001] 0.48 [<0.001] 1975–90 0.125* 0.134*** 0.212*** –0.830*** –0.020 –0.017 0.100 –0.231** 0.90 160 [<0.001] 0.48 [<0.001] 1976–90 0.194** 0.079 0.185*** –0.770*** 0.004 –0.027 0.283 –0.201** 0.92 146 [<0.001] 0.23 [<0.001] 1977–90 0.162* 0.084 0.184*** –0.786*** –0.012 –0.033 0.171 –0.347*** 0.92 132 [<0.001] 0.65 [<0.001] 1978–90 0.154* 0.100 0.204*** –0.800*** –0.018 –0.035 0.193 –0.346*** 0.93 118 [<0.001] 0.12 [<0.001] 1979–90 0.147 0.079 0.157*** –0.806*** –0.020 –0.023 0.147 –0.233** 0.93 104 [<0.001] 0.03 [<0.001] *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level. m is investment in imported machinery in t – 1, Id is investment in Note: pw is the world price, w is the wage rate, pm is the price of raw materials, e is the exchange rate, I–1 –1 m and Kd are the stocks of imported and domestic machinery at the end of t – 2. domestic machinery in t – 1, and K–2 –2 H1: World price elasticity is equal to one (joint test for market power and economies of scale). H2: Davidson and MacKinnon J-test: imported and domestic machinery are imperfect substitutes. H3: Davidson and MacKinnon J-test: imported and domestic machinery are perfect substitutes. Source: Authors’ calculations (see appendix table A-1 for data sources). Table 5. Elasticity Estimates for Export-Oriented Developing Economies, Subsample Windows Hypothesis tests (marginal significance Elasticity estimates levels) Subsample Degrees window pw w pm e m I–1 d I–1 Km d Adjusted R2 of freedom H1 H2 H3 –2 K–2 1967–90 0.923*** 0.112*** 0.062*** –0.213*** –0.072** –0.042 –0.029 –0.171 0.49 218 0.42 0.26 [<0.001] 1968–90 0.882*** 0.108*** 0.068*** –0.220*** –0.082** –0.044 –0.093 –0.177 0.50 209 0.20 0.60 [<0.001] 1969–90 0.925*** 0.102*** 0.066*** –0.201*** –0.071** –0.038 –0.006 –0.154 0.52 199 0.42 0.68 [<0.001] 1970–90 0.854*** 0.114*** 0.070*** –0.221*** –0.074** –0.037 –0.045 –0.209 0.52 188 0.11 0.71 [<0.001] 1971–90 0.844*** 0.109*** 0.072*** –0.207*** –0.079** –0.042 –0.072 –0.188 0.52 177 0.09 0.59 [<0.001] 1972–90 0.857*** 0.108*** 0.068*** –0.191*** –0.094*** –0.025 –0.081 –0.159 0.55 165 0.10 0.43 [<0.001] 1973–90 0.829*** 0.099*** 0.065*** –0.192*** –0.11*** –0.003 –0.243* –0.149 0.58 153 0.05 0.91 [<0.001] 1974–90 0.800*** 0.085*** 0.081*** –0.163*** –0.102*** –0.002 –0.116 –0.189 0.56 141 0.02 0.92 [<0.001] 1975–90 0.811*** 0.087*** 0.058*** –0.138*** –0.099*** 0.001 –0.138 –0.160 0.39 129 0.03 0.73 [<0.001] 1976–90 0.862*** 0.064* 0.088*** –0.134*** –0.055** –0.023 –0.095 –0.376*** 0.51 117 0.10 0.74 [<0.001] 1977–90 0.875*** 0.088*** 0.083*** –0.093** –0.067** –0.018 –0.189 –0.44*** 0.52 105 0.21 0.87 [<0.001] 1978–90 0.915*** 0.067* 0.061*** –0.050 –0.063** –0.006 –0.267 –0.261** 0.52 93 0.43 0.86 [<0.001] 1979–90 0.909*** 0.072* 0.056*** –0.072 –0.033* –0.002 –0.179 –0.207* 0.55 81 0.46 0.19 [<0.001] *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level. m is investment in imported machinery in t – 1, Id is investment Note: pw is the world price, w is the wage rate, pm is the price of raw materials, e is the exchange rate, I–1 –1 in domestic machinery in t – 1, and K– m and Kd are the stocks of imported and domestic machinery at the end of t – 2. 2 –2 H1: World price elasticity is equal to one (joint test for market power and economies of scale). H2: Davidson and MacKinnon J-test: imported and domestic machinery are imperfect substitutes. H3: Davidson and MacKinnon J-test: imported and domestic machinery are perfect substitutes. Source: Authors’ calculations (see appendix table A-1 for data sources). Mody and Yilmaz 39 gestion that the size of the effect declined in the 1980s. Also in the 1980s, with the development of domestic capabilities (not always in advanced research but typi- cally in rapid reverse engineering and adaptation of technologies), investment in domestic capital began to play a greater part in technological advance. Finally, for import substituting economies the effect of imported equipment was negligible until the late 1970s (table 6). The inevitable opening of markets began to occur in these economies in the early 1980s, accompanied by domestic deregulation and thus greater competition from both domestic and foreign sources. During this period investment in imported goods began to have a greater cost-reducing effect. However, the results also suggest that the period of techno- logical advance based on imported goods has not yet been followed by a shift to domestic sources of innovation. IV. Conclusions We have provided empirical evidence on the relationship between export com- petitiveness and the flows and stock of machinery, allowing for the possibility of imperfect substitution between domestically produced and imported machin- ery. Our results show that imported machinery has had an important cost- reducing effect in developed and export-oriented developing economies. This effect acted quickly and typically was not reversed. For developed countries the cost-reducing effect of imported capital goods faded by the early 1970s, presum- ably because the locus of innovation shifted increasingly to domestic sources. For export-oriented developing economies imports of capital goods continued to have an effect throughout the period, though the benefits from domestic in- novation also became tangible in the early 1980s. In contrast, in developing economies where import substitution had been the dominant trade strategy, exporters were unable or lacked the incentive to use imported machinery to improve their competitiveness until the late 1970s. There- after, as some of these economies became more open to international trade and less constrained by domestic regulation, imported capital goods began to play a greater role in innovation and thus to have a cost-reducing effect. Domestically produced machinery does not appear to have provided sustained aid to interna- tional competitiveness in such economies. One interpretation of De Long and Summers (1991, 1992a, 1992b, 1993) is that because additions to the stock of machinery spur growth, government policies should support rapid increases in this stock. The authors themselves were cautious about drawing such a conclusion, however, and were more inclined to favor a lib- eral import regime. While rewarding entrepreneurial behavior, a liberal regime would also facilitate the inflow of imported equipment and thus foster growth. Our results certainly support that view. But they also suggest a possibility for sequencing in innovative activities. Early innovation is most quickly achieved by importing technology. Domestic innovation capability can be built in paral- lel, however, ultimately becoming the principal locus of investment in innova- Table 6. Elasticity Estimates for Import Substituting Developing Economies, Subsample Windows Hypothesis tests (marginal significance Elasticity estimates levels) Subsample Degrees window pw w pm e m I–1 d I–1 Km d Adjusted R2 of freedom H1 H2 H3 –2 K–2 1967–90 0.925*** –0.07 0.056*** –0.022 0.0 –0.058 0.068 –0.188 0.22 240 0.59 0.70 [<0.001] 1968–90 0.853*** –0.058 0.067*** –0.030 0.013 –0.073 –0.028 –0.246 0.26 228 0.26 0.54 [<0.001] 1969–90 0.952*** –0.012 0.061*** –0.044 –0.016 –0.052 0.118 0.018 0.38 216 0.68 0.38 [<0.001] 1970–90 0.946*** –0.006 0.056*** –0.046 –0.023 –0.044 0.207 0.016 0.40 203 0.62 0.64 [<0.001] 1971–90 0.956*** 0.002 0.055*** –0.029 –0.018 –0.038 0.371* –0.076 0.44 191 0.68 0.43 [<0.001] 1972–90 0.867*** –0.02 0.052*** –0.045 –0.028 –0.029 0.276 –0.033 0.44 179 0.17 0.59 [<0.001] 1973–90 0.850*** –0.041 0.049*** –0.014 –0.020 –0.039 0.276 –0.125 0.45 166 0.11 0.65 [<0.001] 1974–90 0.818*** 0.080 0.049*** –0.063* –0.020 –0.007 0.367* 0.003 0.46 153 0.03 0.83 [<0.001] 1975–90 0.781*** 0.079 0.034*** –0.063* –0.026 –0.015 0.308 0.012 0.28 140 0.01 0.90 [<0.001] 1976–90 0.885*** 0.031 0.047*** –0.070** –0.018 –0.035 0.276 –0.086 0.37 127 0.17 0.82 [<0.001] 1977–90 0.832*** 0.052 0.047*** –0.084** –0.055** –0.013 0.105 0.150 0.45 114 0.05 0.35 [<0.001] 1978–90 0.904*** 0.042 0.029*** –0.059* –0.056** –0.044 0.139 0.196 0.53 101 0.25 0.46 [<0.001] 1979–90 0.918*** 0.001 0.036*** –0.028 –0.078** –0.016 –0.138 0.330 0.58 88 0.39 0.60 [<0.001] *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level. m is investment in imported machinery in t – 1, Id is investment Note: pw is the world price, w is the wage rate, pm is the price of raw materials, e is the exchange rate, I–1 –1 in domestic machinery in t – 1, and K– m and Kd are the stocks of imported and domestic machinery at the end of t – 2. 2 –2 H1: World price elasticity is equal to one (joint test for market power and economies of scale). H2: Davidson and MacKinnon J-test: imported and domestic machinery are imperfect substitutes. H3: Davidson and MacKinnon J-test: imported and domestic machinery are perfect substitutes. Source: Authors’ calculations (see appendix table A-1 for data sources). Mody and Yilmaz 41 tion. In the wake of increasing labor costs, countries adopting an export-oriented strategy, especially the East Asian newly industrializing countries, relied heavily on machinery imports to acquire modern technology. Governments and private businesses supported the absorption and adaptation of imported technology through local R&D and engineering efforts. Over time these domestic efforts have become increasingly important. Further analysis along these lines would benefit from disaggregated time-series data on manufacturing subsectors. In particular, sectoral data on machinery investment and machinery prices would make it possible to endogenize the use of machinery. Our results also point to the importance of trade as a vehicle for the transfer of knowledge, identifying capital goods as the conduit. Recently, however, Keller (2000) and Branstetter (2001) have argued that knowledge spillovers within a country are quantitatively more important than international knowledge trans- fer. Our results suggest that the relative importance of internal and external knowledge spillovers may change as the international environment changes and as domestic incentives and absorptive capacity evolve. Thus further exploration of the determinants of internal and external knowledge spillovers is also likely to be a fruitful avenue of research. 42 the world bank economic review, vol. 16, no. 1 Appendix: Data Sources, Descriptive Statistics, and Unit Root Tests Table A-1. Definition of Variables and Data Sources Variable Meaning and data source Developed and developing economies pw Unit value index for manufactured exports from the rest of the world, 1987 = 100. Calculated from the export prices of all other countries weighted by their world market shares. (Source: World Bank, iectrade database) w Annual wage per employee in manufacturing, in thousands of U.S. dollars (total wage bill/number of employees). (Source: U.N. Industrial Development Organization (unido) sectoral database.) The dollar value was converted to local currency using the exchange rate e. e Exchange rate, units of local currency per U.S. dollar, annual average. (Source: Inter- national Monetary Fund, International Financial Statistics database) KT Stock of total machinery, in constant 1985 U.S. dollars. Calculated from machinery investment data using the perpetual inventory method and assuming a depreciation rate of 12 percent. (Source: Penn World Tables, version 5.6) Km Stock of imported machinery, in constant 1985 U.S. dollars. Obtained from imports of nonelectrical machinery (711, 712, 714, 715, 717, 718, 719) and electrical machinery (722, 723, 72491, 726, 7295, 7296, 7297, 7299), Standard Industrial Trade Classi- fication (sitc), rev. 1, using the perpetual inventory method with a 12 percent depre- ciation rate. (Source: United Nations, Commodity Trade (Comtrade) Database.) Note: To obtain imports in constant prices, import data in current U.S. dollars were deflated by the dollar price of investment goods from the Penn World Tables, version 5.6. Kd Stock of domestically produced machinery (KT – Km). Developed countries only p Price index for manufactured exports of country j, U.S. dollars, 1987 = 100. (Source: oecd) pm Price index for imported raw materials, local currency, 1987 = 100. (Source: oecd) Developing economies only p Price index for manufactured exports, U.S. dollars, 1987 = 100. (Source: World Bank, iectrade database) pm Price index for crude petroleum imports, U.S. dollars, 1987 = 100. (Source: World Bank, iectrade database.) The U.S. dollar value was converted to local currency using the exchange rate e. Mody and Yilmaz 43 Table A-2. Descriptive Statistics Capital stock (1990) Period average (standard deviation) (billions of 1985 U.S. dollars) p w pm e (1987 = 100) ($1000) (1987 = 100) (LC/US$) KT Kd Km Developed countries Australia 82.6 10.9 78.5 1.0 95.8 56.3 39.5 (23.9) (5.3) (27.8) (0.2) Austria 63.4 9.8 53.9 18.3 42.2 14.3 27.9 (26.5) (6.5) (32.2) (4.9) Belgium- 65.8 10.6 67.0 42.3 52.4 6.5 45.9 Luxembourg (29.0) (6.0) (34.8) (8.5) Denmark 62.7 14.1 51.8 7.1 31.3 11.6 19.7 (27.1) (7.9) (34.2) (1.4) Finland 61.6 10.3 65.2 4.3 36.2 17.2 19.0 (31.9) (7.2) (31.4) (0.7) France 63.4 9.9 60.9 5.7 339.2 213.2 125.9 (27.3) (5.8) (32.5) (1.3) Germany 61.9 13.5 61.9 2.6 390.7 248.5 142.2 (27.4) (8.1) (32.7) (0.8) Italy 62.4 8.5 67.0 1,007.2 307.0 230.8 76.2 (29.6) (5.5) (26.0) (407.5) Japan 65.4 11.1 51.0 253.5 860.7 805.3 55.4 (27.7) (8.4) (33.5) (74.8) Netherlands 67.0 12.7 57.1 2.7 76.2 12.6 63.7 (26.7) (7.1) (36.9) (0.6) New Zealand 70.1 8.8 55.0 1.2 17.1 8.9 8.2 (32.0) (4.5) (29.5) (0.4) Sweden 66.3 11.9 79.2 5.6 54.69 13.84 40.84 (29.1) (5.5) (26.1) (1.3) United Kingdom 66.4 9.0 65.6 0.5 335.9 196.2 139.7 (30.2) (5.8) (32.2) (0.1) United States 69.8 15.6 66.6 1.0 1603.1 1203.8 399.3 (28.9) (6.9) (30.9) (0.0) Export-oriented developing economies Brazil 88.0 3.4 151.5 0.0 101.2 80.3 20.9 (18.5) (1.0) (58.9) (0.0) Greece 93.5 5.9 146.7 85.4 20.1 10.2 9.9 (18.3) (2.1) (68.0) (51.2) Hong Kong 83.1 1.6 151.5 6.3 23.5 6.7 16.9 (China) (18.2) (0.7) (58.9) (1.4) Indonesia 91.0 0.8 151.5 936.2 75.7 53.7 22.0 (20.9) (0.2) (58.9) (526.3) Israel 93.2 11.6 150.3 0.6 23.1 12.8 10.3 (16.0) (5.5) (65.0) (0.8) Korea, Rep. of 91.9 3.6 149.8 653.6 84.8 38.4 46.5 (17.9) (2.4) (66.5) (156.0) Malaysia 83.7 2.3 151.5 2.4 27.6 8.6 19.0 (22.7) (0.7) (58.9) (0.2) (continued) 44 the world bank economic review, vol. 16, no. 1 Table A-2. (continued) Capital stock (1990) Period average (standard deviation) (billions of 1985 U.S. dollars) p w pm e (1987 = 100) ($1000) (1987 = 100) (LC/US$) KT Kd Km Portugal 81.0 3.5 161.0 92.1 16.9 3.4 13.4 (18.4) (1.1) (66.8) (54.3) Singapore 80.0 5.5 151.5 2.2 37.6 8.5 29.1 (21.0) (2.6) (58.9) (0.2) Thailand 86.6 1.4 152.1 22.9 39.1 19.0 20.1 (19.3) (0.6) (62.5) (2.5) Turkey 91.1 3.7 153.3 545.1 64.0 48.1 15.9 (28.3) (1.2) (66.1) (789.5) Uruguay 77.7 3.3 151.5 0.2 5.6 4.5 1.1 (23.6) (1.2) (58.9) (0.3) Import-substituting developing economies Argentina 76.2 5.5 160.5 0.03 12.0 3.7 8.3 (25.4) (2.3) (56.8) (0.1) Colombia 76.1 2.3 156.8 136.4 12.5 5.6 7.0 (23.3) (0.6) (75.4) (141.8) Guatemala 82.2 2.0 155.4 1.6 3.4 2.1 1.3 (16.9) (0.6) (54.8) (1.0) Honduras 83.1 2.7 155.4 2.1 2.4 1.6 0.8 (18.3) (0.9) (54.8) (0.5) India 91.0 1.0 162.9 10.8 165.9 150.5 15.4 (20.9) (0.3) (53.3) (3.0) Kenya 96.2 0.1 151.5 12.5 3.2 1.0 2.1 (14.6) (0.0) (58.9) (5.2) Mexico 81.0 4.3 130.7 0.6 98.6 68.6 30.1 (22.1) (1.1) (50.8) (1.0) Pakistan 89.7 1.1 151.5 13.4 68.6 62.2 6.5 (17.1) (0.4) (58.9) (4.1) Panama 89.7 4.5 173.4 1.0 1.4 0.5 0.9 (17.1) (1.1) (62.9) (0.0) Peru 91.0 2.9 155.8 0.01 10.1 6.3 3.9 (15.3) (1.0) (72.7) (0.0) Philippines 81.9 1.3 148.5 13.1 21.5 14.2 7.3 (15.4) (0.4) (63.4) (6.6) Sri Lanka 88.3 1.8 153.4 21.2 2.3 0.9 1.4 (16.6) (0.3) (56.5) (10.2) Venezuela, 93.8 7.6 151.5 10.4 28.9 11.5 17.3 R.B. de (26.0) (2.9) (58.9) (12.1) Source: Authors’ calculations (see appendix table A-1 for data sources). Table A-3. Trade Regimes of the Sample Developing Economies 1967–73 1973–85 Outward oriented Inward oriented Outward oriented Inward oriented Strongly Moderately Strongly Moderately Strongly Moderately Strongly Moderately Hong Kong (China) Brazil Honduras Argentina Hong Kong (China) Brazil Colombia Argentina Korea, Rep. of Colombia Kenya India Korea, Rep. of Israel Guatemala India Singapore Guatemala Mexico Pakistan Singapore Malaysia Honduras Peru Indonesia Philippines Peru Thailand Indonesia Israel Sri Lanka Turkey Kenya Malaysia Turkey Uruguay Mexico Thailand Uruguay Pakistan Philippines Sri Lanka 45 Note: Based on World Bank classification. Not included in the classification are Greece, Panama, Portugal, and R.B. de Venezuela. Countries in bold face are treated as export oriented; the others are treated as import substituting. Source: World Bank 1986. DEFINITIONS Strongly Outward Oriented. Trade controls are either nonexistent or very low in the sense that any disincentives to export resulting from import barriers are more or less counterbalanced by export incentives. There is little or no use of direct controls and licensing arrange- ments, and the effective exchange rates for imports and exports are roughly equal. Moderately Outward Oriented. Incentives favor production for domestic rather than export markets. But the average rate of effec- tive protection for the home market is relatively low, and the range of effective protection rates relatively narrow. The use of direct controls and licensing arrangements is limited. The effective exchange rate is higher for imports, but only slightly. Moderately Inward Oriented. Incentives clearly favor production for the domestic market. The average rate of effective protection for the home market is fairly high, and the range of effective protection rates relatively wide. Direct import controls are extensive. The exchange rate is somewhat overvalued. Strongly Inward Oriented. Incentives strongly favor production for the domestic market. The average rate of effective protection for the home market is high and the range of effective protection rates wide. Direct controls and licensing disincentives for the traditional export sector are pervasive, positive incentives for nontraditional exports are few or nonexistent, and the exchange rate is substantially overvalued. 46 the world bank economic review, vol. 16, no. 1 Table A-4. Im-Pesaran-Shin Unit Root Test Results Average Standardized average augmented augmented Dickey-Fuller test Expected Standard Dickey-Fuller statistic value a errora test statistic Developed countries Variables in first differences ∆p –2.5276 –1.4290 0.1688 –6.5060** ∆w –2.5814 –1.4492 0.1663 –6.8068** ∆pm –2.5718 –1.4810 0.1694 –6.4389** ∆e –2.3588 –1.4401 0.1675 –5.4833** ∆K T –2.2518 –1.6831 0.1724 –3.2990* ∆K m –2.3732 –1.5619 0.1683 –4.8201** ∆K d –2.6671 –1.6402 0.1680 –6.1125** Variables in levels p –1.8951 –1.4752 0.1645 –2.5530 w –1.6833 –1.5182 0.1660 –0.9947 pm –1.4419 –1.6494 0.1711 1.2127 e –1.2044 –1.5305 0.1658 1.9671 KT –1.9335 –1.5804 0.1671 –2.1130 Km –1.4902 –1.7306 0.1730 1.3891 Kd –1.4609 –1.7087 0.1697 1.4599 Developing economies Variables in first differences ∆p –2.5336 –1.5538 0.1860 –5.2669** ∆w –2.1085 –1.4809 0.1805 –3.4769* ∆pm –2.9571 –1.5463 0.1811 –7.7875** ∆e –2.1858 –1.5420 0.1893 –3.3992* ∆K T –2.3027 –1.5194 0.1749 –4.4792* ∆K m –2.1896 –1.5907 0.1783 –3.3589* ∆K d –2.2340 –1.4719 0.1707 –4.4646** Variables in levels P –1.1557 –1.6773 0.1928 2.7043 w –1.1079 –1.5930 0.1836 2.6425 pm –1.5940 –1.5661 0.1773 –0.1572 e –0.9703 –1.8957 0.1844 5.0184 KT –1.7476 –1.3673 0.1850 2.0562 Km –1.0986 –1.6987 0.1844 3.2539 Kd –1.2850 –1.6709 0.1795 2.1480 *The null hypothesis of a unit root in each country’s variable is rejected at the 5 percent level of significance. **The null hypothesis of a unit root in each country’s variable is rejected at the 1 percent level of significance. aThe expected value and standard error of the average augmented Dickey-Fuller test statistic are computed through stochastic simulations with 10,000 replications. Source: Authors’ calculations (see appendix table A-1 for data sources). Mody and Yilmaz 47 References Balasubramanyam, V. N., M. Salisu, and David Sapsford. 1996. “Foreign Direct Invest- ment in Export-Promoting and Import-Substituting Countries.� Economic Journal 106(434):92–105. Branstetter, Lee. 2001. “Are Knowledge Spillovers International or Intranational in Scope? Microeconomic Evidence from the US and Japan.� Journal of International Econom- ics 53(1):53–79. Coe, David T., and Elhanan Helpman. 1995. “International R&D Spillovers.� Euro- pean Economic Review 39(5):859–87. Coe, David T., Elhanan Helpman, and Alexander Hoffmaister. 1997. “North-South R&D Spillovers.� Economic Journal 107(440):134–49. Davidson, Russell, and James G. MacKinnon. 1981. “Several Tests for Model Specifica- tion in the Presence of Alternative Hypotheses.� Econometrica 49(3):781–93. De Long, J. Bradford, and Lawrence H. Summers. 1991. “Equipment Investment and Economic Growth.� Quarterly Journal of Economics 106(2):445–502. ———. 1992a. “Equipment Spending and Economic Growth: How Strong Is the Nexus?� Brookings Papers on Economic Activity 2:157–99. ———. 1992b. “How Robust Is the Growth-Machinery Nexus?� Rivista di Politica Economica 92(11):5–54. ———. 1993. “How Strongly Do Developing Economies Benefit from Equipment Investment?� Journal of Monetary Economics 32(3):395–415. Egan, M. L., and Ashoka Mody. 1992. “Buyer-Supplier Links in Export Development.� World Development 20(3):321–34. Engelbrecht, H. J. 1997. “International R&D Spillovers, Human Capital and Produc- tivity in oecd Economies: An Empirical Investigation.� European Economic Review 41(8):1479–88. Feenstra, Robert C. 1989. “Symmetric Pass-Through of Tariffs and Exchange Rates under Imperfect Competition: An Empirical Test.� Journal of International Economics 27(1– 2):25–45. Greene, William. 1997. Econometric Analysis, 3d ed. Upper Saddle River, N.J.: Prentice Hall. Heston, Alan, and Robert Summers. 1991. “The Penn World Table (Mark 5): An Ex- panded Set of International Comparisons, 1950–1988.� Quarterly Journal of Eco- nomics 106(2):327–68. Im, Kyung S., H. Hashem Pesaran, and Yongcheol Shin. 1996. “Testing for Unit Roots in Heterogenous Panels.� University of Cambridge, Department of Economics. Keller, Wolfgang. 2000. “Geographic Localization of International Technology Diffu- sion.� nber Working Paper 7509, National Bureau of Economic Research, Cambridge, Mass. Mann, Catherine L. 1986. “Prices, Profit Margins, and Exchange Rates.� Federal Re- serve Bulletin 72(6):366–79. ———. 1989. “Effects of Exchange Rate Trends and Volatility on Export Prices: Indus- try Examples from Japan, Germany, and the United States.� Weltwirtschaftliches Archiv 129(3):588–618. Mody, Ashoka. 1989. “Firm Strategies for Costly Learning.� Management Science 35(4):496–512. 48 the world bank economic review, vol. 16, no. 1 Mody, Ashoka, and Kamil Yilmaz. 1997. “Is There Persistence in the Growth of Manu- factured Exports? Evidence from Newly Industrializing Countries.� Journal of Development Economics 53(2):447–70. Ohno, Kenichi. 1989. “Export Pricing Behavior of Manufacturing: A U.S.–Japan Com- parison.� IMF Staff Papers 36(3):550–79. Pack, Howard, and John Page. 1994. “Accumulation, Exports, and Growth in the High- Performing Asian Countries.� Carnegie-Rochester Conference Series on Public Policy 40:199–236. Pindyck, Robert S., and Julio J. Rotemberg. 1983. “Dynamic Factor Demands and the Effects of Energy Price Shocks.� American Economic Review 73(5):1066–79. Rao, C. Radhakrishna. 1973. Linear Statistical Inference and Its Applications. 2d ed. London: Wiley. Rodriguez, Francisco, and Dani Rodrik. 1999. “Trade Policy and Economic Growth: A Skeptic’s Guide to Cross-National Evidence.� nber Working Paper 7081, National Bureau of Economic Research, Cambridge, Mass. Rodrik, Dani. 1999. Making Openness Work: The New Global Economy and the De- veloping Countries. Washington, D.C.: Overseas Development Council. Srinivasan, T. N. 1995. “Long-Run Growth Theories and Empirics: Anything New?� In Takatoshi Ito and Anne Krueger, eds., Growth Theories in Light of the East Asian Experience. Chicago: University of Chicago Press. ———. 1999. “Trade Orientation, Trade Liberalization, and Economic Growth.� In Gary Saxonhouse and T. N. Srinivasan, eds., Development, Duality, and the International Economic Regime: Essays in Honor of Gustav Ranis. Ann Arbor: University of Michi- gan Press. Srinivasan, T. N., and Jagdish Bhagwati. 1999. “Outward-Orientation and Development: Are Revisionists Right?� Yale University, Department of Economics, New Haven, Conn. Westphal, Larry E., Yung W. Rhee, and Garry Pursell. 1981. Korean Industrial Compe- tence: Where It Came From. World Bank Staff Working Paper 469, Washington, D.C. White, Halbert. 1980. “A Heteroskedasticity-Consistent Covariance Matrix Estimator and Direct Test of Heteroskedasticity.� Econometrica 48(4):817–38. World Bank. 1986. World Development Report. New York: Oxford University Press.