Pooled Mean Group Estimation of the Bilateral Trade Balance Equation

What is it about?

The autoregressive distributed lag model (ARDL), even though it distinguishes between the short run and the long run effect, allows both the intercepts and slopes to vary across countries. Static panel estimations, such as fixed-effects estimation (FE), cannot distinguish between the short run and the long run behavior. To address the issue of short run heterogeneity as well as long run homogeneity of the estimated coefficients in a panel framework, the pooled mean group (PMG) estimator has gained popularity since 1999. In this paper, we estimate the bilateral trade balance model for the USA vis-à-vis her 19 OECD trading partners for the period 1973q1–2004q4 using the PMG estimator and find that PMG performs better than ARDL, FE, and MG estimators and provides significant and theoretically consistent results.

Featured Image

Why is it important?

The bilateral approach in estimating the trade balance equation has gained momentum in recent years to address the problem of aggregation bias in earlier studies. In the bilateral approach, researchers mostly use the ARDL model. One potential problem with the ARDL approach is that the sign and the significance of the parameters are much too sensitive to the selection of partners and time periods. Because of insufficient information about each country, increasing the number of Estimation of the Bilateral Trade Balance Equation 525 partner countries does not solve the problem of omitted variables, misspecification and multicollinearity among the RHS variables. As a result, estimation of individual country equation by the ARDL approach may not provide us with precise estimates of all the short run and the long run parameters including the speed of adjustment coefficient. The estimated coefficients in many cases produce counterintuitive sign and insignificant coefficients for many countries. In addition to this, in ARDL approach, all the short run and long run coefficients are assumed to be different across countries and thus may not reflect the situation of recent globalization and increasing cooperation among trading countries. To address this issue, we use quarterly data for 19 OECD trading partners of the USA over the period 1973q1–2004q4 and use PMG estimator developed by Pesaran et al. (1999) that assumes long run homogeneity and allowing for short run heterogeneity. We check the sensitivity of our result by allowing maximum three lags or four lags and use the SBC and find that there is no significant change in the result and the PMG estimation performs the best. Using different algorithms in ML procedure and the introduction of trend term also do not cause noteworthy change in result. This is the first panel study in estimating the bilateral trade balance equation using the PMG approach to resolve many of the potential problems of the group-specific ARDL estimation as well as the traditional MG or FE estimation. We find that the results from the PMG estimator are more theoretically consistent and free from many limitations that are contained in the ARDL, FE or MG estimators.

Perspectives