wordpress-themes.org wordpress themes wordpress themes

GLOBAL FACTOR TRADE: Implications for Net Factor Trade 5

Corrections on ROW Technology: T6

We have seen that production model P5 works quite well for most countries. There are a few countries for which the fit of the production model is less satisfying. There are relatively large prediction errors (ca. 10 percent) for both factors in Canada, for capital in Denmark, and for labor in Italy. Given the simplicity of the framework, the magnitude of these errors is not surprising. Since we would like to preserve this simplicity, neither do these errors immediately call for a revision of our framework. Here

There is one case, however, in which a closer review is appropriate. For the ten OECD countries, we have data on technology which enters into our broader estimation exercise. But this is not the case for ROW. The technology for ROW is projected from the OECD data based on the aggregate ROW endowments and the capital to labor ratio. Because the gap in capital to labor ratios between the ten and the ROW is large, there is a good measure of uncertainty about the adequacy of this projection. As it turns out, the prediction errors for ROW are large: the estimated technology matrix under-predicts labor usage by 9 percent, and over-predicts capital usage by 12 percent. Moreover, these errors may well matter because ROW is predicted to be the largest net trader in both factors and because its technology will matter for the implied factor content of absorption of all other countries.

Hence we will consider specification T6, which is the same as T5 except that we force the technology for ROW to match actual ROW aggregate endowments, i.e. BROW yrow = vrow. 22 A plot appears as Figure 13. This yields two improvements over specification T5. The slope coefficient rises by over one-third to 0.59 and the trade variance ratio doubles to 0.38. This suggests that a more realistic assessment of the labor intensity of ROW production materially improves the results.

Adding Gravity to the HOV Demand Model: T7

As we note in the theory section, one of the more incredible assumptions of the HOV model is costless trade. With perfect specialization and zero trade costs, one would expect most countries to be importing well over half of all goods they absorb. Simple inspection of the data reveals this to be a wild overestimate of actual import levels.

By estimating the log form of the gravity equation introduced earlier, we can obtain estimates of bilateral import flows in a world of perfect specialization with trade costs. We then use these estimates of import and own demand in order to generate the HOV factor service predictions. The results are presented in column T7 and illustrated in Figure 14. By almost every measure, this is our best model of net factor trade. The slope coefficient rises from 0.59 under T6 to 0.82 under T7.

That is, measured factor trade is over 80 percent of that predicted. The standard errors are small and the R2 is 0.98. Signs are correctly predicted over 90 percent of the time. The variance ratio rises to nearly 0.7. The results look excellent for each factor considered separately, and especially for capital, which has a slope coefficient of 0.87 and correctly predicts the direction of net factor trade in all cases. These results strongly endorse our use of the gravity equation to account for the role of distance or trade frictions in limiting trade volumes and net factor contents.