Richard Rothstein

Economic Policy Institute 

In response to Michael McPherson's inquiry, permit me to make reference to five research reports or projects which I believe have had an important influence. These are all well-known, so I am presuming that I do not need to provide precise bibliographic references. 

  1. The report by Schweinhart et. al. on the 27 year follow-up of the Perry Pre-School experiment, along with the recently released report by a similar group of authors on the 40 year follow-up. This report has had extraordinary influence on research, policy, and practice. It has made it almost conventional wisdom that investments in very early childhood have the most important impact. It has also illustrated that the "gold standard" for educational research has two rarely (almost never) fulfilled requirements: first, that it is based on randomized experimentation; second (and perhaps even more important), that it follows the treatment and control groups for very long periods of time.
  2. The STAR ( Tennessee class size) research, and in particular the recent paper(s) by Alan Krueger and Diane Whitmore. This experiment has had great influence on class size policy and, like the Perry research, has also legitimized the importance of randomized experimentation. However, much of the policy that was influenced by this research was misguided, because the application of lessons learned from the experiment were in many cases over-simplified. There are many lessons to be learned from studying these misguided applications, regarding the challenges of scaling up, distinguishing aggregated from sub-population results, distinguishing policy-significant from statistically significant results, etc. I mention the positive importance of the Krueger and Whitmore papers because, along with a few other papers that came out of the project itself, these emphasized long-term outcomes of the early-grade treatments, and it is these long-term outcomes with which we should be interested.
  3. David Grissmer's 1994 report on "Student Achievement and the Changing American Family." Grissmer used an unusually creative combination of NELS, NLSY, and NAEP data to disaggregate the influence of demographic from other factors in changes in student achievement. It is remarkable that there has been little interest in following-up on this research, to attempt further to isolate the demographic factors and to specifically identify the residuals which might have included school as well as social policy changes. One suspects that the reason there has been so little interest in follow-up is that researchers are mostly, unlike Grissmer, unwilling to focus on "big ideas" and instead are focused on localizable and short-term policy interventions.
  4. Daniel Koretz's CBO reports from the mid-1980s that explored, more deeply than had previously been done, conventional views about the SAT test score "decline," utilizing a variety of other test databases (including NAEP and Project Talent) to estimate the extent to which this decline was attributable to school or demographic factors. These reports had great influence, although they did not completely offset the conventional misinterpretations of the "decline." The Koretz reports followed, and built upon, the report of the Wirtz Commission established by (I believe it was) the College Board in about 1972.
  5. Paul Peterson's school voucher experiments of the last decade. These experimental findings had important policy implications because they showed that, even an experiment designed and controlled by partisan voucher advocates could not detect significant effects of the voucher policy. The main lesson to be learned from this, however, is the danger of permitting partisans, without an independent oversight board, and without proper peer review, to fully control the reporting of research results. Although Peterson's findings were paradigm-shattering, he and his co-authors have consistently misrepresented their own findings, and only the most determined outside statisticians (again, most significantly, Alan Krueger) have been able to penetrate the data to describe their true meaning. Had one of Peterson's own researchers (David Myers from Mathematica) not protested Peterson's misrepresentation of the findings, few policymakers would have become aware of the significance of this research.

These are the five that immediately come to mind. Perhaps if I considered this problem further, my list would change.