Abstract
Chapter 6 revealed not only that implementation varied among schools in the 10 ES districts, but also that three variable groups (derived from the general systems framework which had guided our research) predict with considerable accuracy the degree to which implementation occurred. In this chapter, our intention is to explore in greater detail some empirical and theoretical issues surrounding implementation of change in schools. Specifically, the following questions are addressed:
-
1.
How do the different groups of independent variables—structure, culture, and input—relate to each other and interact in ways that predict implementation?
-
2.
How does a particular type of system linkage, that is, linkage through the bureaucratic authority structure, affect program implementation?
-
3.
What can we say, more generally, about the impact of system linkages on the implementation of change in schools?
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Notes
We analyzed the commonality of the structure, culture, and input variable groups as they relate to the three dependent variables as follows:\( X_u = T - Y Y_u = T - X XY = T - \left( {X_u + Y_u } \right) \) where Xu and Yu are unique contributions of variable groups in their respective regressions, XY is the common contribution, and T is the total variance explained by both sets. For these regressions, we used only structure, culture, and input variables that met the criterion of contributing at least 2% to the multiple R2 of the regressions presented in Chapter 6.
Hage and Aiken (1967) were able to explain 55% of the variance in “number of program changes.” However, their results should be interpreted with some caution since they had only 16 cases, and entered seven variables into the regression analysis. This points up the problem we faced of “shrinkage” in the R2 when degrees of freedom are used up by entering a large number of variables. We have discussed the unadjusted R2, following existing conventions in the sociological literature. However, the adjusted R2s in Table 17 indicate that “shrinkage” due to diminishing degrees of freedom is not a major problem, as the adjusted R2s are still higher than those generally found in the literature.
For this reason, we do not include a description of the cluster analysis and its results in this chapter. However, the interested reader can refer to Jastrzab, Louis, and Rosenblum (1977) for a description of cluster analysis, the results of the cluster procedure, and the regression analyses performed.
The selection was made on the following basis: using the three dependent variables, nine regressions were computed. Three involved entering all structure and culture variables stepwise, three only structure variables, and three only culture variables. Structure and culture variables that contributed at least 2% to the explained variance in at least three of the six regressions where they were included were classified as having consistently strong predictive power. Several colleagues have noted that they were concerned about the fact that educational level was not included among our group of powerful variables, and wondered whether further exploratory analysis might have revealed it as a significant factor in explaining change patterns. We responded to this concern with additional attempts to look for impacts of whether a school was an elementary or secondary school (Jastrzab et al., 1977), but were unable to locate any results which systematically illuminated our more powerful findings.
In order to avoid a problem of multicollinearity between main and interaction terms, the mean was subtracted from each score prior to its multiplication.
It is important to emphasize that the adjusted multiple R2 continued to increase through each step of the regression rather than declining, as it would if the “shrinkage” due to lost degrees of freedom outweighted the increased fit obtained through the addition of new variables. While the number of variables entered is large compared to the degrees of freedom, the increase allows us to be confident that we are not, in fact, “over-predicting” the equation.
While four interaction terms which included individualization as a factor emerged in the regression analysis, two of these did not produce a clear interaction pattern when they were examined in the dichotomized cross-tabular presentations. These two, the interaction of individualization and precentage male (in an equation with quality) and of individualization and tension (in an equation with quality), are therefore not discussed.
For the purposes of this analysis we introduce three new variables that have not been included up until this point. These are superintendent influence over planning, principal influence over planning, and teacher influence over planning. These variables are derived from a survey question (fall, 1973) which asked respondents to indicate on a five-point scale the level of influence that these and other groups had over the planning of the ES project in their district (see Appendix B).
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1981 Plenum Press, New York
About this chapter
Cite this chapter
Rosenblum, S., Louis, K.S. (1981). Further Exploration of Implementation in Schools. In: Stability and Change. Environment, Development, and Public Policy. Springer, Boston, MA. https://doi.org/10.1007/978-1-4613-3234-3_7
Download citation
DOI: https://doi.org/10.1007/978-1-4613-3234-3_7
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4613-3236-7
Online ISBN: 978-1-4613-3234-3
eBook Packages: Springer Book Archive