This is a preview of subscription content, log in via an institution to check access.
References
William A. Niskanen, Jr.,Bureaucracy and Representative Government (Chicago & New York: Aldine-Atherton, 1971), pp. 3–9.
In addition to Niskanen's book, see also Gordon Tullock,The Politics of Bureaucracy (Washington, D. C.: Public Affairs Press, 1965) and Anthony Downs,Inside Bureaucracy (Boston: Little, Brown & Co., 1967).
Niskanen,Bureaucracy, p. 23.
Richard S. Sterne, Helen A. Scott, and Albert Rosen,Evaluation of Low Cost Meals Program (Miami: 1968; mimeograph).
Preparation of variables for analysis involved various steps. Data were first coded, if necessary. Frequency distributions were prepared and examined. Cutting points and groupings were chosen to normalize, to the degree possible, the data. For dichotomous variables, it is desirable to have no imbalance greater than 3 to 7 between the two sides. Some items weighted 1 to 4 were admitted, since they were logically important variables. Multivariate analyses used, were based on Dean J. Clyde, Elliot M. Cramer, and Richard J. Sherin,Multivariate Statistical Analyses (University of Miami: 1966), Ch. 5. It is the nature of the Manova program used to subtract first the entire relationship of all other variables than those in the hypothesis. The covariates, therefore, include all of the variables other than the lunch score and hypothesis variables for a particular analysis as shown in Table 1, 2. In each case, the covariates were highly correlated to the lunch score and with a high degree of significance. It is also a characteristic of the Manova program used to test in sequential order each of the specific variates in the analysis being tested against the lunch score. This is done after subtracting the effects of the covariates (given at the bottom of each table). Inasmuch as placing any one first might obscure the effect of one which came later, where there were six variables, six different tests were made, placing each one first in turn, this being shown by the column labeled “order” in each of the tables. In Table 1, where there were four variates, only four different orders had to be tested. In Tables 1 & 2 above the material is presented in this manner: the variables tested in each hypothesis are listed on the left as A, B, C, D, E, and F in the case of a six factor rotation, or A, B, C, and D in a four factor rotation. The “F” score for each variable is given, and the probability that such an “F” score might be obtained by chance is also shown. Those probability scores for significant variates are marked with an asterisk, (i.e.,*). It is only in Table 2 that such scores were found. In certain cases, probabilities of .07 or .09 occurred with the variables which reached .05, and these higher figures approach significance. Factor analysis was based upon Clyde et al., (Chapter 4). Factor analysis expanded the scope of information provided by the Manova program. The factor analysis used was in the form of Varimax rotation. The output provided fourteen factors among the entire forty variables, including the Lunch Score. Careful examination of the figures was made. All factor loadings below .400 were eliminated in the interpretation of data. In the analyses, cases from the six centers were pooled, though sampling ratios were uneven. All data, including the attributes of centers to which individuals belonged, were reduced to individual case form. From data analyses of the relation of meals, program participation to individual situations could be made; e. g. relationships of morale scores belonging to organizations other than the centers, but no extrapolation from obtained the sample to the overall propulation (e. g., percentage married, etc.).
Additional information
The authors are, respectively, Director of Social Research, Citizens Planning Council, Rochester, N. Y.; Department of Political Science, University of Rochester; and Administrative Assistant to the County Manager, Dade County, Florida.
Rights and permissions
About this article
Cite this article
Sterne, R.S., Rabushka, A. & Scott, H.A. Serving the elderly?—an illustration of the niskanen effect. Public Choice 13, 81–90 (1972). https://doi.org/10.1007/BF01718853
Issue Date:
DOI: https://doi.org/10.1007/BF01718853