Little is known about the small sample distributions of t-ratios of estimates of coefficients of dynamic simultaneous equation models. This paper investigates the Edgeworth and Nagar approximate distributions of the t-ratios of 2SLS estimates of an artificial two-equation dynamic model. The approximations are compared with probabilities estimated by a Monte Carlo experiment. The results show that the Edgeworth and Nagar approximations are better than the asymptotic normal.