22 , 2018-01-18 , Institute of Comparative Economic Studies, Hosei University , 法政大学比較経済研究所
We estimate trade costs under large zero trade by using daily data on agricultural goods trade within a country. Because of the nature of daily data, there is a prominent zero daily trade between regions and daily delivery is subject to noisy demand and supply shocks, which tends to create heteroskedasticity of the data. Hence, we use Poisson Pseudo Maximum Likelihood (PPML) to estimate gravity model and investigate non-linear nature of trade costs. Empirical analysis shows a statistically significant, but economically subtle non-linearity in trade costs. We also aggregate daily data to monthly level to examine whether shocks are smoothed and thus those impacts are dampened. Our estimation shows that the difference is minor. Comparison of the results with other estimation methods such as the least squares of linear-in-log model and various Tobit procedures is also conducted. There is a large difference in the results between simple least squares and PPML, suggesting the significant heteroskedasticity. We also calculate outward and inward multilateral resistance terms to derive the incidence of trade costs and find that a large portion of trade costs is the buyers' burden.