GLOBAL FACTOR TRADE: Implications for Net Factor Trade
The reason is that the trade literature is replete with proposed amendments to the HOV model that in the end do not help us to understand actual factor service flows. We have already evaluated the hypotheses statistically, so in this section we examine the extent to which these hypotheses help us to understand real world factor trade flows. In order to understand the economic significance of our models, we conduct tests of the HOV model of production and trade under a variety of specifications, as developed in Section II and summarized in Table 2. Here our tests are designed not for model selection, but rather to help us see the economic implications for the HOV model of each of the hypotheses that we have considered. We will begin by working primarily on the production side. Once we have made the major improvements we anticipate in that area, we move on to consider an amendment to the absorption model.
If you are at a point in your life when you would like to borrow some money, we have just what you need: speedy payday loans available to you all day long and in the amount you want. Come and get your one at Source without thinking twice about, it, the rates we offer as low and transparent as that!
Production and Trade Tests For each specification, we provide two tests of the production model. In all cases the technology matrices that we use are based on the fitted values obtained in the previous section. Furthermore, we express both measured and predicted factor content numbers as a share of world endowments in efficiency units. This adjustment eliminates the units problem and enables us to plot both factors in the same graph. The production Slope Test examines specifications P1 to P5 by regressing the measured factor content of production (MFCP) on the predicted factor content of production (PFCP). For example, in specification P1 this involves a regression of BYc on Vc. The hypothesized slope is unity, which we would like to see measured precisely and with good fit. The Median Error Test examines the absolute prediction error as a proportion of the predicted factor content of production. For example, for P1 this is \BUS Yc – Vc\ / Vc.
We provide three tests of the trade model. The first is the Sign Test. It asks simply if countries are measured to be exporting services of the factors that we predict they are exporting, i.e. is sign (MFCT = sign (PFCT) ? For example, in trade specification T1, it asks if sign(BfTc) = sign Vfc – sc V^. The statistic reported is the proportion of sign matches. The trade Slope Test examines specifications T1 to T5 by regressing the MFCT on the PFCT. For example, in specification T1 this involves a regression of BfTc on (V* – sc VfW). The hypothesized slope is again unity, which we would like to see measured precisely and with good fit. The Variance Ratio Test examines the ratio Var(MFCT)/Var(PFCT). One indicator of “missing trade” is when this ratio is close to zero, whereas if the model fit perfectly the variance ratio would be unity. We also consider several robustness checks.
Before turning to our own results, it is well to have in mind how the HOV model has fared under these tests in prior work. Results from the most relevant studies are summarized in Table 3. The results lend themselves to a simple bottom line: All prior studies on international data have failed disastrously by at least one of these measures.
We now turn to tests of our various production and trade specifications.