GLOBAL FACTOR TRADE: Implications for Net Factor Trade 3
If we exclude the US as well, the slope falls to about 0.90. The R2 in each case is respectably above 0.9. Also, in both cases, the median production errors are approximately 20 percent. The ROW continues to be a huge outlier, given its significantly lower productivity. These results suggest that use of an average technology matrix is a substantial improvement over using that of the US, since median production errors fall by one-third to one-half. Nonetheless, the fact that prediction errors are still on the order of 20 percent for the OECD group, and much larger for the ROW, suggests that there remains a lot of room for improvement.
The biggest concern people have when applying for payday loans with speedy cash is how they are going to repay the debt. Getting a loan at Reading here
, you do not need to worry about this aspect, as we take our job very seriously and provide you with sufficient assistance to make sure you have a clear repayment plan.
Examination of T2 can be brief. The sign test correctly predicts the direction of net factor trade only 45 percent of the time. The variance ratio continues to be essentially zero, again indicating strong missing trade. The Slope Test coefficient is -0.006. In short, factor abundance continues to provide essentially no information about which factors a country will be measured to export. These statistics are reinforced by the pictures in Figure 5 and Figure 6. Overall, this model is a complete empirical failure.
Hicks-Neutral Technical Differences: P3 and T3 Specifications P3 and T3 are predicated on the existence of Hicks-neutral differences in efficiency across countries.20 The estimation of these efficiency differences is discussed above in Section IV and here we view the implementation. A plot of P3 appears as Figure 7. There continue to be substantial prediction errors, the largest by far being for the ROW, but also sizable ones for the UK and Canada. Nonetheless, the median prediction error falls to about one-third of its previous level, now around 7 percent. The slope coefficient varies somewhat according to the inclusion or exclusion of the ROW, although typically it is around 0.9. When all data points are included, the R2 is about 0.9. When we exclude ROW, the R2 rises to 0.999.
There is an additional pattern in the production errors. If we define capital abundance as capital per worker, then for the four most capital abundant countries, we underestimate the capital content of production and overestimate the labor content. The reverse is true for the two most labor abundant countries. These systematic biases are exactly what one would expect to find when using a common or neutrally-adjusted technology matrix in the presence of a continuum of goods. Moreover these biases are not small. Quite often these biases in over- or under-predicting the factor content of production were equal to 20 percent of a country’s endowment. Thus, while allowance for Hicks-neutral efficiency differences substantially improves the working of the production model, prediction errors remain both sizable and systematic.
We have seen that the Hicks-neutral efficiency shift did give rise to substantial improvements for the production model. Will it substantially affect our trade results? The answer is definitely not. A plot of T3 appears as Figure 8. The sign test shows that factor abundance correctly predicts measured net factor trade exactly 50 percent of the time. The trade variance ratio is 0.008, indicating that the variance of predicted factor trade still exceeds that of measured factor trade by a factor of over 100. The slope coefficient is essentially zero. In sum, while the adjustment for efficiency differences is useful in improving the fit of the production model, it has done next to nothing to resolve the failures in the trade model.