which the relationships are not exact, so that a set of ideal economic variables is assumed to be generated by a set of dynamic stochastic relationships, as in Koopmans [12], and the actual economic time series are assumed to differ from the ideal economic variables because of random disturbances or measurement errors. The asymptotic error variance matrix for the coefficients of one of the relationships is obtained in the case in which these relationships are estimated using instrumental variables. With this variance matrix we are able to discuss the problem of choice that arises when there are more instrumental variables available than the minimum number required to enable the method to be used. A method of estimation is derived which involves a characteristic equation already considered by Hotelling in defining the canonical correlation [10]. This method was previously suggested by Durbin [7]. The same estimates would be obtained by the maximum-likelihood limited

When estimating a single equation with an error generated by an autoregressive process of higher order than one using a sequence of likelihood ratio tests to determine the correct order, the asymptotic size of the tests will be biased because of multiple optima of the likelihood function. A new type is suggested similar to the Durbin test [2] which is not biased in this way. IN HIS ARTICLE on testing for serial correlation in the presence of lagged endogenous variables [2] Durbin proved a general theorem which gives a significance test shown to be generally asymptotically equivalent to a likelihood ratio test. This paper proposes a generalization which gives a test criterion that may be preferred to the existing test criteria insofar as it can be set up using a less arbitrary choice of the parameters to be re-estimated, and also has the advantage of being relatively simple to compute. It seems more appropriate than the general Durbin form of test for application to the dynamic specification problem discussed in the third section of this article.

This paper develops approximations of the Gram-Charlier type to the cumulative distribution function of the instrumental variables estimator on classical assumptions. In the special case where there are only two endogenous variables in the estimated equation, exact values of the cumulative distribution function are computed by numerical integration and compared with the approximations. Although the error in the approximation depends critically on the parameters of the stochastic model, the approximation is good for the special case even for small sample size over a wide range of values of the parameters. THIS PAPER was originally conceived as a study of the finite sample distribution of two stage least squares estimates. Since it was found that the distribution of a more general class of instrumental variables estimates can be discussed in the same way with a trifling complication of the algebra, the paper was modified to cover these estimates. The basic approach is somewhat similar to that of Nagar [15], since it involves expanding the formulae for the estimator as a series of terms of 0(1), O(T-+), O(T- 1), O(T- 1+), etc., and from this a similar expansion is found for the cumulative probability of the form

Article plaidant en faveur de l'utilisation d'estimateurs d'equations simultanees pour estimer les modeles d'effets aleatoires dynamiques d'un panel de donnees

This paper considers the null hypothesis that the errors on a regression equation form a random walk. By using the standard Durbin-Watson assumptions, we derive three test statistics that are uniformly most powerful against the alternative hypothesis that the errors are being generated by the stationary first order Markoff process. Unfortunately, the tabulated lower and upper bounds are too wide apart and so we compare the powers of the three tests using simulated as well as economic data. It is then recommended that the Imhof routine should be attached to standard regression programs to calculate the exact limit of the Berenblut-Webb statistic.