Do financial institutions exhibit similar currency flow patterns within the Federal reserve system. Certain models indicate that the answer is yes. However, the answer not only depends upon model formation, but also upon the manner in which the raw data is structured. This structuring refers to how the raw data is collapsed, or what level of detail is used in the model. For example, should cash flows be averaged on a yearly basis. The choice of granularity becomes crucial in obtaining an affirmative answer to the question. This paper presents the results of various model formations using linear models and principal components, and using various granularities in the data. Other factors involved in the modeling effect include temporal effects, geographic effects, and characteristics of the financial institutions such as institution type (commercial bank, S&L, Credit Union, etc.) and ''size'' (deposits, loans, volume). The results of this study indicate similar currency flow behaviors at more general levels of detail, indicate how varied and different these behaviors can be at finer levels of detail, and indicate which data transformations and covariates are best for model construction.
展开▼
