In this article we discuss the implementation of a portfolio choice framework and the inclusion of credit rationing by banks in several large-scale macroeconometric models built during the 1960s. We argue that the Fed-MIT-Penn model has a more transparent structure: the structure of the money market is clearer, as is the relationship of its equations with the microeconomic choices of banks. Regarding credit rationing, we found that modelers made important efforts to include it despite its nonobservable nature and to develop a measure of it. A succession of proxy variables was used and despite constant negative results modelers kept trying to find a place for credit rationing in their model. These results invite a deeper reflection on the idea of microfoundations in large-scale macroeconometric models and on the role of beliefs in macroeconometric modeling.