BUREAUCRATIC BEHAVIOR MODELED BY REDUCED-RANK REGRESSION - THE CASE OF EXPENDITURES FROM THE SOVIET STATE BUDGET

A bureaucratic hierarchy compresses information flowing upward and dis aggregates directives passing downward, thus making itself amenable to modeling by reduced-rank regression. Applying this technique to the S oviet government, we find that about eight latent variables are requir ed to characterize the mapping of predetermined variables into a vecto r of budgetary expenditures. The latent variables can be interpreted a s targets for expenditures in high-priority sectors. (C) 1998 Elsevier Science B.V.