A number of information-theoretic alternatives to GMM have recently been proposed in the literature. For practical use and general interpretation, the main drawback of these alternatives, particularly in the case of conditional moment restrictions, is that they rely on high dimensional convex optimization programs. The main contribution of this paper is to analyze the informational content of estimating equations within the unified framework of least squares projections. Improved inference by control variables, shrinkage of implied probabilities and information-theoretic interpretations of continuously updated GMM are discussed in the two cases of unconditional and conditional moment restrictions.