Exchange Rate Pass-through in Canadian Manufacturing: Its Direct and Indirect Components

Transcription

1 Exchange Rate Pass-through in Canadian Manufacturing: Its Direct and Indirect Components S. W. Kardasz and K. Stollery University of Waterloo

2 Exchange Rate Pass-through in Canadian Manufacturing: Its Direct and Indirect Components 1. Introduction In recent years, the demise of the Bretton Woods agreement and the development of models of international trade with imperfect competition have led researchers to examine empirically how product prices respond to fluctuations in exchange rates. 1 In an important paper, Feinberg (1989) studied the industry determinants of exchange rate pass-through into domestic prices for the U.S. Subsequently, Yang (1997) and Lee (1997) examined the pass-through into import prices for the U.S. and Korea, respectively. More recently, we (Kardasz and Stollery 2001) dealt with passthrough into both prices for a broad sample of Canadian manufacturing industries using a twostage procedure. In the first stage we estimated pass-through elasticities by regressing domestic and import prices against the exchange rate and a number of control variables, the most important of which is the price in the matching U.S. industry or industries. Using annual data for the period from 1971 to 1989, satisfactory estimates of the elasticities were obtained for 31 industries at the L-level (essentially the 3- or 4-digit SIC level) of industry aggregation. In the second stage, the inter-industry variation in these elasticities was examined in a cross-section analysis. The present study extends our earlier work by first dividing the total price effect of exchange rate changes into two components, which we call the direct and indirect effects, and then by separately analyzing the industry-level determinants of elasticities measuring these two effects for both domestic and imported goods. In a model where domestic and import prices are simultaneously determined, the direct effects occur because the exchange rate is a determinant of the marginal cost of imports, expressed in domestic currency. The indirect effects, on the other hand, originate with the impact of the exchange rate on the prices of both domestically produced and imported materials used by domestic producers, and hence on their marginal costs. 1 See, for example, the survey article by Goldberg and Knetter (1997). -1-

3 This paper was motivated by a finding in our earlier study which proved to be robust, namely that the (total) domestic pass-through elasticity increases across industries with a variable measuring the responsiveness of domestic costs to the exchange rate. This result suggests that the exchange rate has an indirect effect on domestic prices that is important in at least some Canadian manufacturing industries. However, it does not provide any indication of the relative contribution of the direct and indirect effects to the total pass-through elasticities. The rest of this paper is organized as follows. In section 2 we present a simple conceptual framework. In section 3 we describe the estimation at the industry level of a threeequation model which yields separate estimates of the direct and indirect pass-through elasticities for both domestic and imported goods. In section 4 we attempt to explain the inter-industry variation in these elasticities. Our conclusions appear in section A Simple Conceptual framework We assume that international arbitrage is not possible and that unit costs do not vary with output. These assumptions allow the behaviour of firms in the domestic market to be examined in isolation. In addition, we imagine that the prices of domestic and foreign goods in this market (P d and P m ) are determined by representative domestic and foreign firms whose products are differentiated. The profit functions (in domestic currency) of the representative firms are i = (P i - C i ) Q i (P d, P m, P, Y) for i = d,m (1) where Q i ( ) represents the demand functions, P is the general price level (the price of all other goods), and Y is nominal income. C d, the unit cost of domestic producers, depends on the price of their material inputs (P mat ) and this, in turn, depends on E, the nominal exchange rate (the domestic currency price of foreign currency), i.e., C d = C d (P mat (E)). For simplicity, other determinants of P mat and C d are suppressed until the following section. In contrast, C m, the unit cost of imports, equals, where is the unit cost of foreign firms in foreign currency. -2-

4 Given the assumption that the demand functions are homogeneous of degree zero, we can perform the rest of the analysis in terms of real prices, costs and income. The best-response functions of the two firms can be written as p d = p d (p m, c d (p mat (e)), y) (2) and p m = p m (p d, e,, y) (3) where p d, p m, p mat and y represent the corresponding nominal variables deflated by P, e is the real exchange rate defined as EP* / P where P* is the foreign price level in foreign currency, and = / P*. A change in e has direct effects on both p d and p m which originate with the response of p m to e in equation (3). In contrast, the indirect effects arise from the impact of e on p mat and thereby p d in equation (2). Using Cramer s rule, the total effects of a change in e on both prices can be written in terms of elasticities as follows: (4) (5) where de = d ln p d / d ln e, me = d ln p m / d ln e, mate = ln p mat / ln e, and A = 1- ( ln p d / ln p m ) ( ln p m / ln p d ). Equations (4) and (5) divide the total pass-through elasticities, de and me, into their direct and indirect components. 3. Estimation of the pass-through elasticities -3-

5 There are four differences in specification between the simplified model presented in the previous section and the one we estimate empirically. First, in order to obtain industry estimates of mate, we add a third equation in which p mat is assumed to depend on e and two control variables, y and the price of materials in matching U.S. industries (p matu ). Second, c d is assumed to depend not only on p mat but also on the real wage rate (w) and an index of total factor productivity (tfp). Third, because of limitations in the available data, the price of domestically produced goods in the domestic market (p d ) is replaced by the price of all domestically produced goods ( ~ p d ) which includes exports. Finally, the impact of income on pricing is measured by y err, the trend-free residual obtained when the first difference in the logarithm of real GDP is regressed against time and time squared. This variable is intended to capture the effects of the business cycle on pricing. Our data set, which makes use of new series recently developed by the Input-Output Division of Statistics Canada, covers the period from 1961 to 1995 for 73 L-level industries in the Canadian manufacturing sector. 2 The variables used in the empirical analysis, along with their sources, are shown in Table 1. We express all price, cost and productivity variables (p d, p m, p mat, e, w, tfp, c US and p matu ) as natural logarithms. Prior to estimation, these variables were checked for unit roots using Phillips-Perron tests. For most of the variables, the null hypothesis of a unit root could not be rejected at the conventional 5% level of significance. However, finding cointegrating relationships in a system where p d and p m are simultaneously determined requires using the method developed in Johansen (1995) which, given the large number of industries under consideration, is clearly beyond the scope of the present study. Consequently, we differenced the data, and then checked for stationarity using the KPSS test with stationarity as the null. The one 2 The Input-Output Division distinguishes between 94 L-level manufacturing industries. Twenty of these industries were dropped from our study because of the difficulties associated with matching Canadian and U.S. data. Another industry (asphalt roofing) was dropped because the import price series appeared to be unreliable. -4-

6 industry (office, store and business machine industry) which failed this test was deleted from our data set. Since unit roots have low power and since differencing a stationary series creates an MA(1) component (Maddala and Kim, 1998, p.13), the price equations were estimated allowing for MA(1) errors. Assuming that the resulting price equations are linear, they can be written as follows: ln ~ p d = ln p m + 2 ln p mat + 3 ln w + 4 ln tfp + 5 y err (6) ln p m = ln ~ p d + 2 ln e + 3 ln c US + 4 y err (7) ln p mat = ln e + 2 ln p matu + 3 y err (8) Equations (6) and (7), which satisfy the rank and order conditions for identification, were estimated using 2SLS with the lagged exogenous variables as instruments. Equation (8), on the other hand, was estimated using OLS since all of its explanatory variables are exogenous. Satisfactory results for all three price equations were obtained for thirty-seven industries. These regressions are reported in Tables A1 to A3. turned out to be negative but insignificant in three of these industries (meat and meat products, veneer and plywood, and sash, door and other millwork), as did (fruit and vegetables, copper rolling, casting and extruding, and hardware, tools and cutlery). These coefficients were set equal to 0 for the purposes of calculating the pass-through elasticities presented in Table 2. Summary statistics relating to these elasticities appear in Table 3. The average values of the direct and indirect elasticities are and for domestic goods and and for imports. For domestic goods, the frequency distributions of the direct and indirect elasticities are quite similar, with over 80% of both elasticities having values less than 0.3. In contrast, for imports, about 73% of the direct elasticities exceed 0.4 while 84% of the indirect elasticities are less than 0.2. Taken together, these results indicate that, at least for our sample of thirty seven industries, the direct and indirect effects of exchange rate changes are about of equal importance for domestic goods, whereas the direct effect is clearly dominant for imports. -5-

7 When combined, the direct and indirect elasticities yield average total pass-through elasticities of for domestic goods and for imports. These values are more than double those obtained in our early study (0.125 and 0.255, respectively). There are at least three possible explanations for these differences. First, the estimates in the present study are based on a longer time period ( versus ). Second, less than half of the industries (specifically, sixteen out of a total of thirty-one in the previous paper and thirty-seven in the current one) are common to both studies. Third, the specifications and methods used to estimate the p d and p m equations in the two studies are quite different. In our earlier work, reduced form equations were estimated using e, p US (the price in matching U.S. industries), y err and time as explanatory variables. In this study, we use 2SLS with lagged values of p mat, w, tfp, c US and y err as the exogenous variables. We have not yet attempted to investigate the roles played by these three possible explanations although we suspect that the third one is the most important. While there may be no fully satisfactory way of determining which estimates of the total elasticities are more reliable, it should be pointed out that a value of for me is close to the value of 0.5 which, according to Goldberg and Knetter (1997, p.44), is the mean value of this elasticity found in studies for the U.S. We now turn to the second stage of our empirical analysis, an examination of the interindustry variation in the pass-through elasticities we have estimated. 4. Determinants of the pass-through elasticities For our sample of thirty-seven Canadian manufacturing industries, the simple correlations between the direct and indirect elasticities are for domestic goods and for imports. These low correlations support the view that (our estimates of) the direct and indirect elasticities are distinct variables that need to be examined separately. However, to facilitate comparison with earlier studies, we also analyze the determinants of the total domestic and import elasticities. The independent variables include three elements of market structure which are standard for an open economy like Canada s: concentration which we measure by the Herfindahl index -6-

8 (H), the (domestic) tariff (TF) which we supplement by a dummy variable (NTB) to take account of non-tariff barriers, and product differentiation. To these variables, we add the import share (MSR) and a regional dummy variable (REGD). As argued by Feinberg (1989, 507), Yang (1997, 96-97) and Kardasz and Stollery (2001, ), the import share is an observable variable which influences the demand elasticities of domestic and foreign firms and thereby the passthrough elasticities. We use REGD to proxy natural barriers to trade since markets become localized when transport costs and related factors such as product perishability are sufficiently important to limit the area served by a producing entity. Estimates of the elasticity of substitution between domestically produced and imported goods in the Canadian market have been calculated by Létourneau and Lester (1988). Conceptually, this variable appears to be a good measure of product differentiation for the purposes of this study. Unfortunately, it is available for only eighteen of the thirty-seven industries in our sample. In empirical studies of industrial organization, it is common to proxy product differentiation by the advertising-sales ratio. However, this practice appears to be questionable, given the expectation that advertising expenditures depend not only on the extent of product differentiation but also on other factors such as price-cost margins and, in the case of oligopolistic industries, strategic considerations. In the end, we decided to measure product differentiation by a dummy variable (PDIF) which is based on the industry taxonomy developed by the OECD and employed by Baldwin and Rafiquzzaman (1995) and by Martins et al (1996). All of the explanatory variable discussed so far apply to all three foreign elasticities and to the domestic elasticities provided they refer to the prices of domestically produced goods sold in the domestic market (p d ). However, as noted earlier, the available data, and hence the domestic pass-through elasticities we have estimated, refer to the prices of domestically produced goods whether sold at home or aborad (p ~ d). ~ p d can be viewed as being a weighted average of p d and the export price (p x ), both of which are unobserved, with the weights depending on the export share (x): -7-

9 ~ pd = (1-x) p d + x p x. Consequently, we add the export share in 1981 (XSR) as a determinant of the (observed) domestic elasticities since there is no reason to expect the exchange rate effects on p d and p x to be equal. From equations (4) and (5), it can be seen that the indirect elasticities ( di and mi ) depend on the responsiveness of the price of materials to the exchange rate ( mate ) and on the responsiveness of the domestic price to the price of materials ( ln p d / ln p mat ). To take account of these two factors, we include the estimated value of mate and the (domestic) ratio of materials to the value of output (MASR) as determinants of the indirect and total elasticities. Linear and, as described below, non-linear versions of the six pass-through elasticity equations were estimated using OLS with the HETCOV option in the SHAZAM econometrics software package. This option corrects the standard errors of coefficients for heteroskedasticity using White s heteroskedastic-consistent covariance matrix. As can be seen from equations (4.1) and (4.10) in Table 4, there is a significant inverse relationship between the two direct elasticities ( and ), on the one hand, and MRS, NTB and REGD, on the other. In addition, increases with PDIF. Some years ago, Harry Bloch (1974) found that the effects of concentration and the tariff on prices in Canadian import-competing industries are interdependent. For this reason, we added the interactive term H*TF to the list of explanatory variables. The coefficient on this variable proved to be positive and significant in equation (4.2), implying that increases with H and TF and, furthermore, that the marginal effects of each of these variables on increases with the level of the other. In other words, the same increase in concentration (the tariff) leads to an increase in which is larger in industries where the tariff (concentration) is high than where it is low. -8-

10 In our preferred regressions for the indirect elasticities ( and ), and MASR enter multiplicatively, a specification which seems to be consistent with the chain of causation associated with the indirect effect. The determinants of the indirect elasticities in equations (4.5) and (4.14) are quite similar. Specifically, the coefficients on H, TF, H*TF, NTB, PDIF and *MASR all have the same signs and are statistically significant in both equations. In fact, the only material differences between these equations occur with respect to XSR, which has a significant effect in (4.5) but does not appear in (4.14), and REGD, whose coefficient is significant only in (4.14). In these equations, the derivatives of and with respect to H increase with TF and they are negative when TF is less than and 9.672, respectively, and positive otherwise. These conditions are satisfied in 29 or 78% of the industries in our sample. Similarly, the derivatives with respect to TF increase with H and they are negative when H is less than and 0.102, respectively, and positive otherwise. The first condition holds for 16 (43% of the total) industries; the second, for 21 (57%). In sum, while increases in concentration lower the indirect elasticities in most (but not all) of our industries, the effects of tariff increases are more mixed. As might be expected, the regression results of the total elasticities ( and ) reflect those of the corresponding direct and indirect elasticities. The primary reason for reporting these results is to facilitate comparison with earlier studies, all of which deal with total elasticities. Not surprisingly, the present paper is most closely related to our 2001 study. However, it is also related to those by Feinberg (1989), Yang (1997) and Lee (1997) referred to in the introduction to this paper. Our earlier study leads to three important conclusions. Using our current notation, the first is that increases with * MASR and XSR. The present paper is consistent with that result and also with Feinberg s conclusion that the exchange rate has a larger impact on domestic -9-

11 prices in U.S. manufacturing industries with a heavy reliance on imported inputs. The second conclusion of our earlier study is that increases with the rate of price protection, which takes account of tariff and non-tariff barriers to trade, and with MSR. From equation (4.17), it follows that increases with TF when H exceeds 0.151, which is the case for only 12 of our industries. Furthermore, the coefficients on MSR and NTB are negative and significant, not positive. By and large, then, our present results do not support the second conclusion of our earlier paper. The third conclusion of that study is that the impacts on and of two variables which can be treated as measures of product differentiation, the elasticity of substitution between imported and domestic goods and the advertising-sales ratio, are not robust. The present study provides a plausible explanation for this finding: since PDIF affects the direct and indirect elasticities in opposite directions, a variety of results become possible when total effects are considered, depending on equation specification, time period of analysis and industry sample. Feinberg (1989) finds that the capital-sales ratio is a significant determinant of the total domestic pass-through elasticity in U.S. manufacturing. When we added the capital-output ratio (KQR) to our regressions, it turned out to be significant for both direct elasticities and for the total domestic elasticity. However, in contrast to Feinberg, we find that capital intensity has a positive effect on exchange rate pass-through. Yang (1997) and Lee (1997) analyze the determinants of the total import elasticity for U.S. and Korean manufacturing, respectively. Yang finds that this elasticity decreases with the capital-to-labour ratio. However, when we added this variable to our regressions in lieu of KQR, it proved to be insignificant. Lee s main result is that concentration reduces the import elasticity. In contrast, we find that decreases with H in only the six industries where the tariff is less than Thus, some of our current results, especially those relating to the role played by *MASR, appear to be consistent with those of earlier studies. At the same time, however, -10-

12 there are a number of inconsistencies. This study suggests two possible explanations for the disagreements. First, by focusing on total pass-through, earlier studies may have aggregated effects which work in opposite directions. Second, the relationship between pass-through elasticities, on the one hand, and concentration and the tariff, on the other, may be more complex than previously considered in earlier studies. TABLE 1 Variables used in the empirical analysis Price equations ~d P Implicit price index for gross output, 1981=1 a jt Implicit price index for imports, 1981=1 a -11-

13 Implicit price index for intermediate inputs, 1981=1 a W jt Compensation per person hour, 1981= 1 a (Payroll + cost of materials in current dollars)/real gross output in matching U.S. industries, 1981=1. Real gross output = (shipments + inventories in current dollars)/price of shipments b Cost of materials in current dollars/real gross output in matching U.S. industries, P t E t ~ P d jt 1981=1 b GDP deflator for Canada, 1981c GDP deflator for the U.S., 1981=1d Canadian dollars per U.S. dollar, average of noonday rates c Equals ~ P d / P t jt Equals Equals w jt Equals Equals Equals tfp jt e t Fisher Ideal Indexes of multifactor productivity based on gross output, 1981 = 1c Equals y t Index of real GDP, 1981= 1c Trend-free residual obtained by regressing the first difference of against time and time squared Cross-section equations H j TF j NTB j MSR j XSR j PDIF j Herfindahl index of concentration for 1980 e Tariff rate, circa 1987 f Equals 1 for the distilleries, 0 otherwise f Imports/Canadian market, 1981 a Exports / shipments, 1981 a Equals 1 for industries where product differentiation is high, 0 otherwise g REGD j Equals 1 for industries with regional markets, 0 otherwise h -12-

14 MASR j KQR j Ratio of materials to the value of output, 1981a Ratio of net capital stock to gross output, 1981 i NOTES All variables refer to Canada unless otherwise specified. The subscripts t and j index time and industries, respectively. Canadian and U.S. industries were matched using Statistics Canada and the U.S. Bureau of the Census, Concordance Between the Standard Industrial Classifications of Canada and the United States, 1980 Canadian SIC United States SIC. SOURCES a Input-Output Division, Statistics Canada. b Bureau of the Census, dbase III+ data files; Bureau of Labour Statistics, Public Database. c CANSIM Database. d Council of Economic Advisors, Economic Report of the President, e Statistics Canada (1983), Industrial Organization and Concentration in the Manufacturing, Mining and Logging Industries, f Department of Finance (1988), The Canada-U.S. Free Trade Agreement: An Economic Assessment, Table A2.1. g Baldwin and Rafiquzzaman (1995), Appendix B, and Martins et.al. (1996), Table 6. h Department of Consumer and Corporate Affairs (1971), Concentration in the Manufacturing Industries of Canada, Tables IV-1 and A-5. i Microeconomics Analysis Division, Statistics Canada TABLE 2 Pass-through elasticities Industry Domestic prices Import prices Direct Indirect Total a Direct Indirect Total a -13-

15 Meat and meat products Fish products Fruit and vegetables Vegetable oils (except corn oil) Bread and other bakery products Soft drinks Distilleries Plastic products Footwear Miscellaneous leather products Man-made yarns and cloth Wool yarns and cloth Hosiery Veneer and plywood Sash, door and other millwork Other wood products Paper boxes and bags Primary steel Steel pipes and tubes Non-ferrous metals smelting and refining Copper rolling, casting and extruding Power boilers and structural metals Wire and wire products Hardware, tools and cutlery TABLE 2 (Continued) Pass-through elasticities -14-

16 Industry Domestic prices Import prices Direct Indirect Total a Direct Indirect Total a Heating equipment Commercial refrigeration and air conditioning equipment Motor vehicles Truck and bus bodies and trailers Record players, radio and TV receivers Communication and other electronic equipment Communication and energy wires and cables Batteries Hydraulic cement Concrete products Refined petroleum and coal products Plastics and synthetic resins Toilet preparations a Totals need not equal the sum of components due to rounding. TABLE 3 Summary statistics for pass-through elasticities -15-

17 Domestic elasticities Import elasticities Direct Indirect Total Direct Indirect Total Mean Standard Deviation Minimum Maximum

18 TABLE 4 Determinants of the exchange-rate pass-through elasticities Equation No. Domestic elasticities Direct Constant H TF H*TF NTB MSR XSR PDIF REGD MASR MASR KQR * (1.88) (1.33) (0.73) * (2.67) * (1.89) (1.00) (1.10) * (2.34) * (2.44) (0.32) (0.62) 0.139* (2.05) * (3.17) * (2.18) (1.30) (1.06) * (2.60) * (2.09) (0.99) (0.46) 0.157* (2.73) * (3.01) * (2.27) (1.51) (1.54) * (2.85) 0.376* (2.05) 0.25 Indirect (0.92) * (1.92) (0.27) * (4.52) (1.26) (1.49) * (2.70) (0.91) 0.281** (1.54) 0.521** (5.19) * (2.05) * (5.48) * (3.27) 0.144* (5.90) * (7.05) (1.67) 0.265* (2.36) * (4.60) (0.23) 0.822** (10.66) (1.59) * (4.97) * (3.37) 0.162* (5.67) * (6.30) (1.69) 0.306* (2.45) * (4.09) (0.01) (9.84) (1.68)

19 Total (0.04) (0.05) (0.41) * (3.22) * (2.17) (1.44) (0.13) (1.68) 0.579** (1.68) 0.289** (1.73) * (3.57) * (2.06) (1.65) 0.269* (3.64) * (7.71) * (2.70) 0.638* (2.53) (0.85) * (2.49) * (2.98) * (4.18) (1.34) 0.327* (5.11) * (7.99) * (2.92) 0.774* (3.60) (0.24) * (3.11) TABLE 4 Continued Determinants of the exchange-rate pass-through elasticities Equation No. Constant H TF H*TF NTB MSR XSR PDIF REGD MASR Import elasticities Direct * (4.12) (1.27) (0.66) * (3.94) * (2.92) 0.260* (2.74) * (1.85) * (4.37) (0.07) (1.09) (1.24) * (4.48) * (2.86) 0.242* (2.70) * (1.97) * (3.76) (0.51) (1.01) 0.156* (1.75) * (3.13) * (2.69) 0.295* (3.34) * (2.19) Indirect (0.07) (1.65) (0.76) * (1.88) (0.84) * (2.56) 0.088* (2.03) (0.11) 0.306** (2.63) * (2.18) * (4.57) * (3.52) 0.128* (5.44) * (2.71) (1.11) * (4.05) 0.073* (1.78) (1.50) * (3.97) * (3.27) 0.138* (4.76) * (2.10) (1.18) * (3.73) (1.60) Total * (2.94) (1.00) (1.14) * (2.19) * (2.42) 0.209* (2.19) (0.69) (0.13) (0.98) * (7.12) (0.47) * (1.82) 0.212* (1.85) * (5.08) * (2.35) 0.168* (2.08) (0.96) * (5.42) (0.93) * (1.73) 0.242* (2.09) * (4.36) * (2.21) *(**) Indicates statistically significant at 10 percent, two-tailed (one-tailed) test * (2.33) (1.22) TABLE A1 Domestic Price Equations -18-

20 Dependent variable: ln p d _ 2 Industry constant ln p m ln p mat ln w ln tfp y err R D.W. Meat and meat products (0.21) (0.39) (26.4) (2.11) (0.57) (0.59) (0.83) 1.85 Fish products (0.12) (1.17) (8.26) (1.89) (2.43) (0.10) (0.54) 1.94 Fruit and vegetables (0.85) (2.51) (1.61) (1.03) (4.13) (1.21) (3.00) 1.96 Vegetable oils (except corn oil) (0.28) (0.93) (0.47) (1.16) (0.93) (1.53) (0.13) 2.00 Bread and other bakery products (0.06) (2.53) (5.90) (3.21) (3.81) (0.17) (1.15) 1.84 Soft drinks (1.20) (0.53) (5.16) (1.34) (1.20) (0.82) (0.13) 1.93 Distilleries (0.66) (1.61) (0.31) (0.51) (3.52) (0.30) (0.82) 1.99 Plastic products (0.26) (0.97) (7.50) (0.56) (3.87) (1.36) (2.22) 1.86 Footwear (2.51) (3.79) (3.95) (4.10) (0.22) (0.23) (3.41) 2.10 Miscellaneous leather products (6.62) (5.60) (0.83) (2.59) (0.45) (1.64) (3.40) 1.85 Man-made yarns and cloth (1.08) (3.14) (3.49) (0.17) (0.35) (0.28) (1.90) 1.89 Wool yarns and cloth (0.64) (3.53) (2.48) (0.57) (0.07) (1.97) (8.18) 1.98 Hosiery (2.41) (3.36) (0.79) (2.63) (0.21) (1.77) (3.70) 1.74 Veneer and plywood (1.30) (0.48) (5.66) (0.80) (0.68) (1.44) (3.80) 1.74 Sash, door and other millwork (1.34) (0.26) (4.12) (0.36) (1.80) (4.59) (0.06) 1.95 Other wood products (1.35) (4.55) (1.07) (2.69) (2.89) (2.89) (0.27) 1.91 TABLE A1 (Continued) _ 2 Industry constant ln p m ln p mat ln w ln tfp y err R D.W. -19-

21 Paper boxes and bags (0.21) (0.60) (10.4) (0.63) (1.11) (0.40) (1.14) 2.12 Primary steel (0.35) (1.91) (2.38) (1.15) (0.09) (0.77) (2.07) 1.52 Steel pipes and tubes (0.91) (3.17) (1.66) (0.55) (2.30) (0.29) (0.22) 1.90 Non-ferrous metals smelting and refining (0.22) (0.22) (8.33) (0.08) (0.47) (2.57) (1.66) 1.84 Copper rolling, casting and extruding (0.58) (2.00) (4.82) (0.11) (1.53) (0.65) (0.47) 2.02 Power boilers and structural metals (0.02) (1.28) (3.98) (0.26) (0.77) (0.38) (0.69) 1.94 Wire and wire products (0.88) (1.54) (4.65) (2.18) (0.65) (1.69) (1.06) 2.03 Hardware, tools and cutlery (0.16) (2.13) (0.24) (5.88) (0.25) (0.47) (9.72) 2.28 Heating equipment (1.21) (0.87) (2.47) (1.34) (0.84) (1.72) (3.27) 1.98 Commercial refrigeration and air conditioning equipment (2.09) (2.02) (0.85) (1.36) (1.09) (2.06) (2.21) 1.80 Motor vehicles (0.87) (2.32) (2.64) (1.57) (0.56) (0.58) (7.24) 2.04 Truck and bus bodies and trailers (3.76) (4.24) (1.32) (2.68) (1.29) (1.42) (0.02) 1.99 Record players, radio and TV receivers (1.84) (5.23) (0.00) (1.82) (7.45) (1.11) (6.63) 1.96 Communication & other electronic equipment (1.36) (1.10) (2.14) (1.54) (3.51) (2.79) (0.92) 2.04 TABLE A1 (Continued) -20-

22 _ Industry constant ln p m ln p mat ln w ln tfp y err 2 R D.W. Communication and energy wires and cables (1.44) (5.54) (3.29) (0.96) (1.24) (0.56) (2.81) 2.01 Batteries (0.94) (0.93) (1.90) (0.04) (0.90) (0.04) (2.08) 1.85 Hydraulic cement (0.16) (2.54) (0.07) (0.83) (0.19) (1.49) (1.68) 2.04 Concrete products (0.14) (0.52) (3.18) (0.66) (0.76) (0.27) (0.18) 1.93 Refined petroleum and coal products (1.63) (0.87) (9.76) (1.71) (0.26) (1.32) (1.98) 1.98 Plastics and synthetic resins (2.02) (2.57) (5.15) (1.75) (1.25) (2.05) (0.95) 1.89 Toilet preparations (0.88) (2.05) (0.92) (0.52) (2.13) (0.46) (1.28) 2.06 NOTES: Absolute t-values in brackets. p m is the predicted value of the import price. is the estimated value of the MA(1) error parameter. D.W. is the Durbin-Watson d statistic. -21-

23 TABLE A2 Import Price Equations Dependent variable: ln p m _ Industry constant ln p d ln e ln c US y err 2 R D.W. Meat and meat products (0.04) (2.95) (0.18) (2.93) (2.34) (1.86) 1.90 Fish products (0.03) (2.99) (1.30) (1.64) (1.55) (0.89) 2.06 Fruit and vegetables (0.63) (0.52) (3.46) (7.14) (2.45) (2.12) 1.68 Vegetable oils (except corn oil) (0.02) (1.48) (1.66) (5.26) (1.27) (3.03) 1.82 Bread and other bakery products (0.08) (4.18) (0.41) (0.17) (0.37) (0.77) 2.00 Soft drinks (0.30) (2.50) (0.42) (2.77) (0.42) (2.00) 1.93 Distilleries (0.35) (1.43) (0.59) (0.14) (3.07) (2.89) 1.82 Plastic products (1.35) (1.84) (2.79) (1.49) (1.78) (3.76) 1.99 Footwear (0.87) (0.37) (0.25) (0.00) (2.28) (1.74) 1.90 Miscellaneous leather products (0.00) (3.16) (0.00) (0.18) (0.83) (5.65) 1.87 Man-made yarns and cloth (0.55) (2.43) (3.38) (2.19) (0.10) (0.67) 1.90 Wool yarns and cloth (0.95) (3.60) (0.38) (0.62) (2.51) (4.72) 1.60 Hosiery (0.33) (3.21) (4.80) (2.76) (0.01) (0.21) 1.98 Veneer and plywood (4.10) (2.22) (0.57) (7.23) (1.92) (5.72) 1.71 Sash, door and other millwork (1.22) (0.91) (3.39) (0.38) (1.01) (1.55) 1.82 Other wood products (0.91) (1.90) (9.36) (2.31) (2.49) (6.88)

24 TABLE A2 (Continued) _ Industry constant ln p d ln e ln c US y err 2 R D.W. Paper boxes and bags (0.78) (2.89) (1.51) (0.73) (1.31) (0.65) 1.83 Primary steel (0.62) (3.83) (2.55) (1.70) (0.69) (0.66) 1.92 Steel pipes and tubes (0.79) (3.98) (3.16) (1.78) (0.45) (2.07) 1.93 Non-ferrous metals smelting and refining (0.02) (0.03) (0.48) (3.31) (0.17) (6.03) 2.25 Copper rolling, casting and extruding (2.41) (0.04) (4.27) (7.36) (2.07) (7.39) 1.51 Power boilers and structural metals (0.07) (1.73) (2.68) (1.35) (1.14) (0.45) 1.94 Wire and wire products (0.52) (1.71) (5.81) (2.85) (0.47) (1.52) 1.99 Hardware, tools and cutlery (0.08) (0.53) (3.32) (1.45) (0.64) (0.61) 2.00 Heating equipment (3.16) (2.39) (6.40) (1.01) (1.44) (6.07) 1.37 Commercial refrigeration and air conditioning equipment (0.24) (0.71) (4.21) (0.61) (1.36) (1.60) 1.90 Motor vehicles (0.65) (1.83) (2.83) (1.24) (1.50) (2.25) 1.83 Truck and bus bodies and trailers (1.16) (1.43) (3.72) (1.40) (1.15) (0.80) 1.93 Record players, radio and TV receivers (0.17) (3.05) (3.01) (1.32) (0.68) (8.31) 1.92 Communication and other electronic equipment (1.08) (0.39) (3.38) (0.87) (1.53) (0.09) 2.00 Communication and energy wires and cables (0.63) (0.89) (2.48) (3.01) (1.57) (0.32) 1.99 Batteries (0.46) (1.04) (3.42) (0.24) (0.40) (0.46) 1.98 Hydraulic cement (0.57) (2.36) (6.02) (2.11) (5.02) (5.83)

25 TABLE A2 (Continued) _ Industry constant ln p d ln e ln c US y err 2 R D.W. Concrete products (1.29) (1.71) (3.84) (0.07) (1.07) (1.68) 2.02 Refined petroleum and coal products (1.74) (0.58) (1.35) (7.38) (1.38) (2.10) 1.95 Plastics and synthetic resins (2.29) (1.52) (4.28) (6.07) (0.63) (4.91) 1.84 Toilet preparations (0.41) (1.03) (6.83) (0.79) (0.93) (1.17) 1.93 NOTES: Absolute t - values in brackets. p d is the predicted value of the domestic price. is the estimated value of the MA(1) error parameter. D.W. is the Durbin-Watson d statistic. -24-

26 TABLE A3 Materials Price Equations Dependent variable: ln p mat _ Industry constant ln e ln p matu y err 2 R D.W. Meat and meat products (2.46) (8.72) (12.7) (0.44) (6.90) 1.57 Fish products (5.26) (4.83) (5.40) (4.34) (9.17) 1.26 Fruit and vegetables (0.29) (3.60) (4.38) (1.39) (0.43) 1.94 Vegetable oils (except corn oil) (0.79) (0.26) (2.14) (0.62) (6.39) 2.12 Bread and other bakery products (0.44) (2.86) (5.27) (0.99) (0.07) 1.95 Soft drinks (0.78) (1.41) (6.93) (0.43) (0.29) 1.97 Distilleries (0.09) (2.75) (1.59) (0.67) (0.25) 1.99 Plastic products (0.32) (1.52) (6.10) (0.56) (0.53) 1.96 Footwear (0.39) (4.66) (5.79) (1.86) (6.05) 1.87 Miscellaneous leather products (0.53) (3.42) (0.43) (1.22) (2.63) 1.85 Man-made yarns and cloth (2.06) (3.25) (2.17) (0.63) (3.05) 2.10 Wool yarns and cloth (1.02) (1.27) (1.49) (0.97) (2.65) 1.90 Hosiery (1.30) (0.17) (1.22) (1.65) (0.94) 1.91 Veneer and plywood (0.95) (4.24) (7.02) (1.84) (3.42) 1.78 Sash, door and other millwork (0.54) (1.92) (6.06) (1.97) (8.11) 1.50 Other wood products (0.50) (0.69) (2.02) (0.45) (0.07)

27 TABLE A3 (Continued) _ Industry constant ln e ln p matu y err 2 R D.W. Paper boxes and bags (0.51) (3.11) (9.05) (0.96) (1.30) 1.87 Primary steel (1.43) (2.25) (5.75) (0.96) (2.38) 2.01 Steel pipes and tubes (2.10) (1.71) (3.58) (4.67) (4.17) 1.66 Non-ferrous metals smelting and refining (0.11) (0.88) (6.46) (1.52) (1.34) 1.86 Copper rolling, casting and extruding (1.17) (3.72) (6.31) (2.59) (5.60) 1.64 Power boilers and structural metals (0.70) (2.46) (4.42) (1.37) (1.75) 1.82 Wire and wire products (13.2) (2.72) (13.0) (0.15) (7.99) 1.83 Hardware, tools and cutlery (1.12) (3.18) (2.73) (1.72) (5.80) 1.51 Heating equipment (1.86) (1.85) (0.54) (0.01) (8.40) 1.17 Commercial refrigeration and air conditioning equipment (0.21) (2.12) (0.74) (0.90) (0.90) 1.97 Motor vehicles (3.09) (5.13) (2.51) (0.49) (7.72) 2.21 Truck and bus bodies and trailers (0.33) (3.37) (0.66) (0.66) (1.68) 1.74 Record players, radio and TV receivers (1.90) (2.40) (1.52) (0.76) (9.12) 1.85 Communication and other electronic equipment (5.06) (0.26) (1.17) (2.35) (8.29) 2.15 Communication and energy wires and cables (1.82) (2.53) (4.41) (2.65) (4.64) 1.96 Batteries (4.18) (3.71) (5.95) (1.64) (6.13)

28 TABLE A3 (Continued) _ Industry constant ln e ln p matu y err 2 R D.W. Hydraulic cement (0.06) (2.30) (2.06) (0.60) (0.30) 1.93 Concrete products (0.05) (0.52) (1.81) (0.93) (0.13) 2.01 Refined petroleum and coal products (0.41) (1.60) (9.08) (0.48) (0.25) 1.99 Plastics and synthetic resins (0.48) (1.07) (4.25) (0.50) (0.11) 1.97 Toilet preparations (0.19) (5.34) (2.49) (1.62) (7.29) 1.92 NOTES: Absolute t-values in brackets. is the estimated value of the MA(1) error parameter. D.W. is the Durbin-Watson d statistic. -27-