Hysteresis in Japanese Export Market: A Dynamic Random-Effect Probit Approach to Panel Data of Japanese Machinery-manufacturing Firms

We estimate the hysteresis effect in export participation decision based on rich panel data of Japanese machinery-manufacturing firms from 1994 to 2015. We find that the "observed" hysteresis effect, difference between the probability of export for the firms that exported in the previous year and the probability of export for the firms that did not export in the previous year, is quite large (0.904 for machinery-manufacturing sector as a whole) Using the estimates of dynamic random-effects probit model, we calculate the degree of "genuine" hysteresis Then we find that the "observed" hysteresis in part reflects spurious effects that are caused by both observable (firm size) and unobservable firm characteristics that tend to persist over time. The genuine hysteresis effect, free from spurious effects, is much smaller than the observed one, but the magnitude of genuine hysteresis still exceeds 0.5.