We thank George A. Krause and Jeff Gill for helpful comments. All statistical analyses were conducted in STATA and replicated in limdep. All data and documentation to replicate this analysis including the complete STATA output can be found at http://web.polmeth.ufl.edu/ or www-bushschool.tamu.edu/pubman/ Zen and the Art of Policy Analysis:
A Response to Nielsen and Wolf
AbstractNeilsen and Wolf (N.d.) lodge several criticism of Meier, Wrinkle and Polinard (1999).Although most of the criticisms deal with tangential issues rather than our core argument, their criticisms are flawed by misguided estimation strategies, erroneous results, and an inattention to existing theory and scholarship. Our re-analysis of their work demonstrates these problems and presents even stronger evidence for our initial conclusion-both minority and Anglo students perform better in schools with more minority teachers.
1
Zen and the Art of Policy Analysis:
A Response to Nielsen and WolfPolicy analysis is an art not a rote exercise. It requires informed judgement at every step of the process and an understanding of the implications of methodological choices. This exchange illustrates the importance of that art. In "Representative Bureaucracy and Distributional Equity: Addressing the Hard Question" we presented two key findings-in school districts with more minority teachers both minority and Anglo students perform better. We also addressed several interesting but tangential issues including the anomaly of overall test scores, the linearity of the relationships, and how equity and performance were linked. Nielsen and Wolf (N.d.) lodge six specific criticisms of our work:1. The appropriate estimation technique is a fixed effects pooled regression.2. We omitted a key variable, percent Anglo students, in our model.
3.The relationship between minority teachers and student performance is not nonlinear.4. Minorities should be disaggregated into blacks and Latinos; when this is done, there is no relationship between minority teachers and minority student performance.5. The appropriate unit of analysis is the student not the school district.6. Substantively weighted least squares (SWLS) is used incorrectly.Even if Nielsen and Wolf were correct (and they are not), only one of these criticisms bear on our key findings; the others focus on peripheral issues. Our response demonstrates that Nielsen and Wolf are incorrect in assertions 1 through 5 and misinterpret assertion 6 because they failed to read the relevant work in this area. We deal with each criticism in order.
The Use of Fixed Effects Models
2Nielsen and Wolf contend that we have underspecified models, specifically that based on a Hausman specification test we should have estimated our equations as fixed effects models by including a series of dummy variables to represent the unique characteristics of each individual school district. With a panel data set of dimensions N x T (where N is the number of unique cases and T is the number of time periods), there are a myriad of possible models (Balt...