Abstract
Since aggregation's first in-depth treatment by Theil (1954), it has been widely recognized that testing and evaluating micro-theory with aggregated data had a direct influence on parameter estimates, attendant conclusions, and policy implications of such theories. Yet, because the magnitude of its effects have not been sufficiently demonstrated, aggregation remains a common problem. Aggregate data are often substituted when microdata is not available or when micro-theory is used in the context of a macro problem even though the estimated coefficients are biased estimated of the specific micro-coefficients. As an example, banking reserve theories are generally developed on the basis of an individual bank's behavior. However, the theories are tested with aggregate banking data. In most cases a micro-evaluation, which would be proper, is never performed. In this dissertation several free and excess reserve theories are re-evaluated. Particular emphasis is placed upon the proper micro-evaluation and upon the effects of aggregation on the estimation process and attendant conclusions. Initially this work develops several frameworks relating to aggregation over individuals, aggregation over time, and aggregation introduced specification errors. These frameworks are presented to demonstrate some probable pitfalls encountered when aggregated data is used to test micro-theory. After a review of excess and free reserve literature and theory, several prominent hypotheses are tested on individual bank data. These results, which for the most part are unsupportive of the theories, differ substantially with the originally reported results which used aggregate data..
Rowe, Robert David (1975). A micro-empirical analysis of aggregation using commercial bank reserve functions. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -184316.