Time-series estimation of the U.S. import demand for light vehicles assembled in Mexico
This paper estimates the U.S. import demand for automobiles assembled in Mexico—SITC 781: Motor cars and other motor vehicles principally designed for the transport of persons (not public transport), including station wagons and racing cars.
Relevance
In 2012, 2.9 million light vehicles were assembled in Mexico—up from 1.5 million in 2004. Over the period 2004-2012, on average Mexico exported 80 percent of its total production of light vehicles. In particular, roughly 65 percent of such exports go to the U.S. Given the importance of the car industry in Mexico and its dependence on the U.S. demand, this paper seeks to assess the determinants of U.S. import demand for automobile assembled in Mexico. No previous assessment was found in the literature.
Methodology
The econometric model follows the mainstream literature and has its theoretical grounds in Senhadji (1998); however, this paper includes relevant regressors, in addition to the income and relative price variables, to avoid biased elasticity estimates. Also, following the recent literature, this model determines the time-series properties of the variables and determines the estimation method accordingly. Time series from January 1996 to June 2012 are used in this study. The properties of the processes that generate the series are analyzed, in order to specify their order of integration. The series that are included in the cointegrating regression model are I(1), according to results from unit root tests—ADF, PP, and ERS Point-Optimal. The paper then estimates the cointegrating import demand regression using Dynamic OLS procedure—which addresses issues such as possible endogeneity among regressors and the short span of the time-series employed.
Results
According to—preliminary—estimation results, the U.S. import demand for automobiles assembled in Mexico is income inelastic, perfect price inelastic, and responsive to the included competition variables. Elasticity estimates are robust to alternative specifications of the cointegrating regression equation. (Residuals from the DOLS cointegrating regression are tested and found to be stationary, which confirms that estimates are not spurious.) Also, this paper points to the estimate bias from omitting relevant variables, as it is common in the mainstream literature on estimating import demand equations.