For the regression equation C = A,B + e where A is a p x q stochastic matrix whose elements are independently distributed and contemporaneously correlated with the elements of 8, the lth moment of the least squares estimator of (3 exists if and only if l < p - q + 1. In particular, this implies that the lth moment of the k-class estimator of the coefficients of the G1 -1 non-normalizing endogenous variables of an equation with K1 included and K2 excluded exogenous variables in a simultaneous system with N observations exists if and only if 1 < M where

This paper considers a problem faced by players who are involved in a sequence of games: not necessarily the same games, not necessarily with the same opponents, and not necessarily under conditions of complete information. The players are assumed to act in response to stationary Markovian hypotheses which they form about the actions of their opponents. Conditions are explored which require that these hypotheses be correct on average and that the players actions be optimal in response to their hypotheses.

This paper is concerned with testing for causation, using the Granger definition, in a bivariate time-series context. It is argued that a sound and natural approach to such tests must rely primarily on the out-of-sample forecasting performance of models relating the original (non-prewhitened) series of interest. A specific technique of this sort is presented and employed to investigate the relation between aggregate advertising and aggregate consumption spending. The null hypothesis that advertising does not cause consumption cannot be rejected, but some evidence suggesting that consumption may cause advertising is presented.