Welfare effects of technological convergence in processed food industries.

Given the estimates of equation (4), we now identify the contribution of technological convergence to aggregate follower relative TFP growth. On average, followers' relative TFP grew 4.03% per year during the 1993-2001 period on account of their technological catch-up with the leader. During the same period, factors other than catch-up reduced their relative TFP growth by an annual 5.81%. The net effect is that followers' relative TFP fell 1.78% between 1993 and 2001 (table 2; see also table 5). A follower's relative TFP changes on account of events either in the leader or follower nation indicated, respectively, in the denominator and numerator of equation (4)'s dependent variable. Research intensity and investment in technical development generally are higher in developed (leader) economies than in developing (follower) ones (Helpman 1997). Such factors shift the leader's technological frontier relative to the follower's, with effects that can linger for decades. Nevertheless, followers' relative TFP growth would have declined by 5.81% in the absence of technological convergence.

Equation (5) enables us to derive the mean rate of technological convergence in each of the seventeen industries even though coefficient [[delta].sub.i] in equation (4) is not significant in four industries. Convergence rates, given in table 3, vary from 2.5% (ISIC 1531) to 9.5% (ISIC 1543) per year, pointing to the public-good nature of technology (Grossman and Helpman 1990). Our convergence rates are higher than those reported by Bernard and Jones (1996a) for OECD countries. In the latter study, the annual speed of TFP convergence is 6.50% in agriculture and 1.68% in manufacturing.

What explains the differences between the Bernard-Jones results and our own? First, productivity convergence rates likely have risen in recent years, as information technology and economic integration have accelerated. Second, productivity convergence probably has been slower among OECD countries, which already are comparatively developed, than elsewhere. Our findings certainly are consistent with Bernard and Jones' (1996a) expectation that nations adopting existing technology likely catch up much more quickly than do those inventing their own. Third, observed productivity in food sub-industries tends to catch up more rapidly than in the industry as a whole, since global trade--and associated cross-border technological transfers--increasingly have occurred between firms within a given ISIC four-digit industry aggregate.

Empirical Test of Convergence Effects

We turn now to testing our hypotheses about TFP convergence effects (Results 1-4). Tables 4a-d provide estimates of the effects of followers' relative TFP growth on their shares of global value-added, imported shares of consumption, relative wage, and welfare. Estimates of the leader's welfare equation are given in table 4e. For each Result 1-4 above, we present three sets of estimates, one corresponding to OLS, the second to FGLS with groupwise heteroskedasticity in the industry dimension, and the third to FGLS with groupwise heteroskedasticity in the country dimension. Heteroskedasticity in the country dimension is more evident than that in the industry dimension, so the following discussion focuses on the FGLS results accounting for the former. Moreover, our estimation includes country-specific fixed effects. Replacing country-specific effects with industry fixed effects does not alter the results in tables 4a-e.

Beginning with table 4a, note that a follower's relative TFP growth, and to a lesser extent the growth rates in its shares of global capital and labor, significantly enhances its share of global value-added. Boosting the follower's relative TFP growth 1% raises the growth in its share of global value-added by 0.915%. Similarly, a 1% growth in the follower's capital-share and labor-share growth, respectively, lifts its growth in global value-added share by 0.239% and 0.771%. Relative TFP growth's comparatively large impact on value-added share growth suggests total factor productivity is especially important for the follower's share in global production. The sign and significance of the estimated parameters are robust across the three estimators. (14)

Using equation (4), we next identify productivity convergence's effect on the growth of the follower's share of global value-added. Table 5 shows convergence increased followers' share of value-added by an average 3.69% per year during the 1993-2001 period. In the absence of convergence, followers' shares would have fallen by as much as 5.32% per year. All else constant, the follower's gain in global value-added share implies a corresponding loss in the leader's share. However, factors other than convergence have increased the leader's share in global value-added.

Table 4b reports, on the basis of equation (7), productivity convergence's effects on followers' relative wages. Both right-hand-side coefficient estimates have the expected positive sign and, in all three specifications, are statistically significant at the 1% level. [R.sup.2] in the OLS fit is 53.9%. If a follower's relative TFP rises 1%, its relative wage goes up 0.224%. Elasticity of the follower's relative wage with respect to its capital-labor ratio is 0.112, underscoring capital's impact on the marginal product of labor. Table 5 gives productivity convergence's contribution to the growth in followers' relative wages (0.90% per year). The mean wage gap between follower and leader would have widened by -1.30% per year in the absence of productivity convergence.

Effects of productivity convergence on followers' imported share of consumption are presented in table 4c. Consistent with Result 3, a 1% increase in the follower's relative TFP growth leads to a 0.819% fall in the growth rate of its imported consumption share. The follower's relative capital has no significant effect, but growth in its relative labor significantly reduces growth in its imported consumption share, with an elasticity of -0.455. (15) The latter result is consistent with our earlier finding that processed food industries are labor-intensive. Table 5 shows a follower's productivity convergence would decrease its imported share of consumption by an annual 3.30%, although owing to factors other than convergence, the growth in its imported share of consumption would rise 1.46% per year.

Table 4d provides estimates of equation (10), the effect of productivity convergence on followers' welfare. A 1% rise in a follower's TFP growth improves its welfare by 0.652%, suggesting that technological convergence has a strong, positive, real-income effect on its welfare. Capital and labor growth make their own positive contributions to welfare, with elasticities of 0.193% and 0.645%, respectively. But as a proxy for terms-of trade, a follower's relative TFP does not reduce the follower's welfare significantly. Absolute TFP growth has in all three specifications a significant welfare-enhancing effect. Table 5 suggests technological convergence's positive real-income effect (2.63%) dominates the follower's welfare improvement, so followers realize net gains from convergence.

Table 4e shows, following equation (9), technological convergence's effects on the leader's welfare. The leader's absolute TFP growth significantly boosts its welfare growth rate (elasticity 0.342). Factor accumulation has similar effects: a 1% rise in capital and labor growth lifts welfare growth by a respective 0.330% and 0.524%. Unlike in the follower's welfare equation, we find statistical evidence of terms-of-trade effects, i.e., the follower's relative TFP growth significantly enhances the leader's welfare in the country-groupwise FGLS estimates. Expressed differently, a 1% increase in a follower's relative TFP growth improves the leader's welfare by 0.016%, a finding consistent with the terms-of-trade effect in our Result 4.

Summary and Conclusions

We have investigated the welfare effects of technological convergence in the processed food industries by extending Krugman's monopolistic competition model. Convergence is reflected in a narrowing inter-country gap in fixed or marginal costs. Comparative statics indicates convergence between technological leader and follower enhances the follower's competitiveness--as reflected in its share of global production--but weakens the leader's. By improving the leader's terms of trade, convergence also improves leader welfare. The follower's welfare change depends upon convergence's positive income effect relative to its negative terms-of-trade effect.

Data from seventeen processed food industries in thirty developed and developing nations were assembled to estimate, through a value-added equation, cross-country and cross-industry productivity level and growth. Estimates indicate significant cross-country variation in productivity level and growth rate. Technological convergence was then identified in each food industry through a regression of relative TFP growth rate on initial relative TFP level. Evidence of convergence is found in thirteen of the seventeen industries, and at rates generally higher than in earlier studies. Differences between our and earlier results likely can be attributed to aggregation and timing: our study focuses on the information-technology era in a setting with intra-industry trade.