Tuesday, November 17, 2009
As a general trend, there is no doubt that the new administration is seriously committed to evidence-based policy. Peter Orszag of the White House budget office has recently called for (and made the case for, in his blog) systematic evaluations of federal policies consistent with the president’s promise to “restore science to its rightful place.” But how does this play out with ED?
First, we are seeing a major shift from a static notion of “scientifically based research” (SBR) to a much more dynamic approach to continuous improvement. In NCLB there was constant reference to SBR as a necessary precondition for spending ESEA funds on products, programs, or services. In some cases, it meant that the product’s developers had to have consulted rigorous research. In other cases, it was interpreted as there having to be rigorous research showing the product itself was effective. But in either case, the SBR had to precede the purchase.
Evidence of a more dynamic approach is found in all of the competition-based stimulus programs. Take for example the discussion of “instructional improvement systems.” While this term usually refers to classroom-based systems for formative testing with feedback to the teacher allowing differentiation of instruction, it is used in a broader sense in the current RFAs and guidance documents. The definition provided in the R2T draft RFA reads as follows (bullets and highlights added for clarity):
“Instructional improvement systems means technology-based and other strategies that tools that provide
* and administrators
with meaningful support and actionable data to systemically manage continuous instructional improvement, including activities such as:
* instructional planning;
* gathering information (e.g., through formative assessments (as defined in this notice), interim assessments (as defined in this notice), summative assessments, and looking at student work and other student data);
* analyzing information with the support of rapid-time (as defined in this notice) reporting;
* using this information to inform decisions on appropriate next steps;
* and evaluating the effectiveness of the actions taken.”
It is important to notice, first of all, that tools are provided to administrators, not just to teachers. Moreover, the final activity in the cycle is to evaluate the effectiveness of the actions. (Joanne Weiss, who heads up the R2T program, uses the same language inclusive of effectiveness evaluation by district administrators in a recent speech).
We have pointed out in a previous entry that the same cycle of needs analysis, action, and evaluation that works for teachers in the classroom also works for district-level administrators. The same assessments that help teachers differentiate instruction can, in many cases, be aggregated up to the school and district level where broader actions, programs, and policies can be implemented and evaluated based on initial identification of the needs. An important difference exists between these parallel activities at the classroom and central office level. At the district level, where larger datasets extend over a longer period, evaluation design and statistical analysis are called for. In fact this level of information calls for scientifically based research.
Research is now viewed as integral to the cycle of continuous improvement. Research may be carried out by the district’s or state’s own research department or data may be made available to outside researchers as called for in the SLDS and other RFAs. The fundamental difference now is that the research conducted and published before federal funds are used is not the only relevant research. Of course, ED strongly prefers (and at the highest level of funding in i3 requires) that programs have prior evidence. But now the further gathering of evidence is required both in the sense of a separate evaluation and in the sense that funding is to be put toward continuous improvement systems that build research into the innovation itself.
Our recent news item about the i3 program takes note of other important ideas about the research agenda we can expect to influence the reauthorization of ESEA. It is worth noting that the methods called for in i3 are also those most appropriate and practical for local district evaluations of programs. We welcome this new perspective on research considered as a part of the cycle of continuous instructional improvement. — DN
Tuesday, September 22, 2009
With that amount at stake, it makes sense for government funders to look for some track record of scientifically documented success. The frequent references in ED documents to processes of “continuous improvement” as part of innovations suggest that proposers would do well to supplement the limited evidence for their innovation by showing how scientific evidence can be generated as an ongoing part of a funded project, that is, how in-course corrections and improvements can be made to the innovation as it is being put into place in a school system.
In his speech to the education industry, Shelton complained about the low quality of the evidence currently being put forward. Although some publishers have taken the initiative and done serious tests of their products, there has never been a strong push for them to produce evidence of effectiveness.
School systems usually haven’t demanded such evidence, partly because there are often more salient decision criteria and partly because little qualified evidence exists, even for programs that are effective. Moreover, district decision makers may find studies of a product conducted in schools that are different from their schools to have marginal relevance, regardless of how “rigorously” the studies were conducted.
The ED appears to recognize that it will be counter-productive for grant programs such as i3 to depend entirely on the pre-existing scientific evidence. An alternative research model based on continuous improvement may help states and districts to succeed with their i3 proposals—and with their projects, once funded.
Now that improved state and district data systems are increasing the ability of school systems to quickly reference several years of data on students and teachers, i3 can start looking at how rigorous research is built into the innovations they fund—not just the one-time evaluation typically built into federal grant proposals.
This kind of research for continuous improvement is an innovation in itself—an innovation that may start with the “data-driven decision making” mode in which data are explored to identify an area of weakness or a worrisome trend. But the real innovation in research will consist of states and districts building their own capacity to evaluate whether the intervention they decided to implement actually strengthened the area of weakness or arrested the worrisome trend they identified and chose to address. Perhaps it did so for some schools but not others, or maybe it caught on with some teachers but not with all. The ability of educators to look at this progress in relation to the initial goals completes the cycle of continuous improvement and sets the stage for refocusing, tweaking, or fully redesigning the intervention under study.
We predict that i3 reviewers, rather than depending solely on strong existing evidence, will look for proposals that also include a plan for continuous improvement that can be part of how the innovation assures its success. In this model, research need not be limited to the activity of an “external evaluator” that absorbs 10% of the grant. Instead, routine use of research processes can be an innovation that builds the internal capacity of states and districts for continuous improvement.
Friday, September 11, 2009
Easton made five points.
First, he is not retreating from the methodological rigor, which was the hallmark of his predecessor, Russ Whitehurst. This simply means that IES will not be funding poorly designed research that does not have the proper controls to support conclusions the researcher wants to assert. Randomized control is still the strongest design for effectiveness studies, although weaker designs are recognized as having value.
Second, there has to be more emphasis on relevance and usability for practitioners. IES can’t ignore how decisions are made and what kind of evidence can usefully inform them. He sees this as requiring a new focus on school systems as learning organizations. This becomes a topic for research and development.
Third, although randomized experiments will still be conducted, there needs to be a stronger tie to what is then done with the findings. In a research and development process, rigorous evaluation should be built in from the start and should relate more specifically to the needs of the practitioners who are part of the R&D process. In this sense, the R&D process should be linked more directly to the needs of the practitioners.
Fourth, IES will move away from the top-down dissemination model in which researchers seem to complete a study and then throw the findings over the wall to practitioners. Instead, researchers should engage practitioners in the use of evidence, understanding that the value of research findings comes in its application, not simply in being released or published. IES will take on the role of facilitating the use of evidence.
Fifth, IES will take on a stronger role in building capacity to conduct research at the local level and within state education agencies. There’s a huge opportunity presented by the investment (also through IES) in state longitudinal data systems. The combination of state systems and the local district systems makes gathering the data to answer policy questions and questions about program effectiveness much easier. The education agencies, however, often need help in framing their questions, applying an appropriate design, and deploying the necessary and appropriate statistics to turn the data into evidence.
These five points form a coherent picture of a research agency that will work more closely through all phases of the research process with practitioners, who will be engaged in developing the findings and putting them into practice. This suggests new roles for the Regional Education Labs in helping their constituencies to answer questions pertinent to their local needs, in engaging them more deeply in using the evidence found, and in building their local capacity to answer their own questions. The quality of work will be maintained, and the usability and local relevance will be greatly increased.
Thursday, July 9, 2009
“I am trying to put much more emphasis on evidence-based policy decisions here at OMB. Wherever possible, we should design new initiatives to build rigorous data about what works and then act on evidence that emerges — expanding the approaches that work best, fine-tuning the ones that get mixed results, and shutting down those that are failing.”
This suggests a continuous process of improving programs based on evaluations built into the fabric of program implementations, which sounds very valuable. Our concern, however, at least in the domain of education, is that Congress or the Department of Education will contract for a national experiment to prove a program or policy effective. In contrast, we advocate a more localized and distributed approach based on the argument Donald Campbell made in the early 70s in his classic paper “The Experimenting Society” (updated in 1988). He observes that “the U.S. Congress is apt to mandate an immediate, nationwide evaluation of a new program to be done by a single evaluator, once and for all, subsequent implementations to go without evaluation.” Instead, he describes a “contagious cross-validation model for local programs” and recommends a much more distributed approach that would “support adoptions that included locally designed cross-validating evaluations, including funds for appropriate comparison groups not receiving the treatment.” Using such a model, he predicts that “After five years we might have 100 locally interpretable experiments.” (p.303)
Dr. Orszag’s adoption of the “top tier” language from the Coalition for Evidence Based Policy is buying into the idea that an educational program can be proven effective in a single large scale randomized experiment. There are several weaknesses in this approach.
First, the education domain is extremely diverse and, without the “100 locally interpretable experiments,” it is unlikely that educators would have an opportunity to see a program at work in a sufficient number of contexts to begin to build up generalizations. Moreover, as local educators and program developers improve their programs, additional rounds of testing are called for (and even the “top tier” programs should engage in continuous improvement).
Second, the information value of local experiments is much higher for the decision-maker who will always be concerned with performance in his or her school or district. National experiments generate average impact estimates, while giving little information about any particular locale. Because concern with achievement gaps between specific populations differs across communities, it follows that, in a local experiment, reducing a specific gap—not the overall average effect—may well be the effect of primary interest.
Third, local experiments are vastly less expensive than nationally contracted experiments, even while obtaining comparable statistical power. Local experiments can easily be one-tenth the cost of national experiments, thus conducting 100 of them is quite feasible. (We say more about the reasons for the cost differential in a separate policy brief). Better yet, local experiments can be completed in a more timely manner—it need not take five years to accumulate a wealth of evidence. Ironically, one factor making national experiments expensive, as well as slow, is the review process required by OMB!
So while we applaud Dr. Orszag’s leadership in promoting evidence-based policy decisions, we will continue to be interested in how this impacts state and local agencies. We hope that, instead of contracting for national experiments, the OMB and other federal agencies can help state and local agencies to build evaluation for continuous improvement into the implementation of federally funded programs. If nothing else, it helps to have OMB publicly making evidence-based decisions. —DN
Campbell, D. T. (1988). The Experimenting Society. In E. S. Overman (Ed.), Methodology and epistemology for social science: Selected Papers. (pp. 303). Chicago: University of Chicago Press.
Tuesday, June 9, 2009
Hanushek and Lindseth propose that, where administrators and teachers are accountable and rewarded for results, they will consider the trade-offs in efficiency of spending money one way or another. For example, smaller class size may lead to better results but, if the same money were spent to increase teacher quality, the results may be much more substantial. This proposal, of course, depends on there being sufficient evidence that various programs, policies, or approach have a measurable impact. And they further acknowledge that getting this information is not a matter of running one-time experimental evaluations. The wide variation of populations, resources, and standards in US school systems means that a large number of smaller scale evaluations are called for. If states and school districts were to get into the habit of routinely pilot testing programs locally (and collecting and analyzing the data systematically) before scaling up within the district or state, the gains in efficiency could be substantial.
Hanushek and Lindseth do not address the question of how local evaluations of sufficient quality and quantity can be paid for. If one depends solely on the Institute of Education Sciences for grants and contracts, the process will be slow and the resources inadequate. Setting aside a certain percentage of federal grants to states and districts for evaluations is often unproductive because the evaluations are not designed or timed to provide feedback for continuous improvement. Too often, educators and administrators treat the evaluation as a requirement that takes money from the program. We have argued elsewhere that integrating research into program implementation at the local level calls for building local school district capacity for rigorous evaluations. It also calls for a reform agenda that changes how decisions are connected both to explorations of district data and to locally generated evidence as to whether programs and policies are having the desired impact. This is different from contracting with the evaluator once program is under way because the plan for the evaluation is part of the plan for implementation. Directing a good portion of the program funds to a process of continuous improvement will make the program more efficient and provide educators with the hundreds of studies that will begin to accumulate the kind of evidence that they need to make a rational choice about what programs are worth trying out in their own locale.
Educators, especially those who spend their days engaged with children in a classroom, may find the rational economic model on which the authors’ proposals are based unsatisfying and perhaps simplistic. Most people don’t go into education because they are maximizing their economic return. Nonetheless, it is hard to find a rationale for retaining teachers who are demonstrably ineffective beyond the traditional practice of union solidarity that militates against differentiation of skills among its members. The authors' arguments are thought provoking in that they demonstrate in rich narrative detail the obvious irrationality of considering only the amount of money put into schools and not considering the effectiveness of the programs, policies, and approaches that the money is spent on. —DN
1 Hanushek, E.A. & Lindseth, A.A. (2009). Schoolhouses, courthouses and statehouses: Solving the funding-achievement puzzle in America’s schools. Princeton NJ: Princeton University Press.
Wednesday, May 13, 2009
Over the past few weeks, in webinars and similar venues, educators have been asking what this means. Many are hesitant to commit funds without knowing what evidence of compliance will be called for. The following quotes were compiled by Jennifer House, Ph.D., Founder of Redrock Reports:
Superintendent of a large suburban district: “We just need to know what kind of data needs to be collected for the accountability portion of ARRA—especially funds in the State Fiscal Stabilization Fund.”
Superintendent of an urban district: “When is the Department of Education going to tell us what data they need for the accountability and reporting requirements of ARRA?”
Title I director of a major urban district: “I know what I need to do for Title I reporting. Is there any other data I need to collect to report on the use and impact of the ARRA funds?”
IDEA director of a large suburban district: “What other data is needed about ARRA funds”
Paraphrase of seven questions from a single MDR webinar: “When will we hear what the accountability requirements are for ARRA?”
CIO of a large suburban district: “We need to accommodate the data that needs to be collected in our system for ARRA. When do we get the word?”
These educators need to know what is meant by “accurately measure and track results.” Will this information just be used to audit who was paid for what? Or will ED be calling for a measure of results in terms of impact on schools, teachers, and student achievement?
State Education Agencies are asked for “baseline data that demonstrates the State’s current status in each of the four education reform areas.” Will the states and districts be asked for subsequent data showing an improvement over baseline?
Educators have heard that, in the near future, ED will describe specific data metrics that states will use to make transparent their status in the four education reform areas for the purpose of “showing how schools are performing and helping schools improve.” They expect that this will not be a one-time data collection; instead, they expect an element of tracking to help them with continuous improvement.
ED has a one-time opportunity to move education toward an evidence-based enterprise on a massive scale by calling for evidence of outcomes—not just the starting baseline. Conditions are ripe for quickly and easily promoting a major reform in how districts measure their own results. Educators already expect this. A simple time series design is all that is needed. Training and support for this can be readily supplied through existing IES funding mechanisms.
Monday, April 6, 2009
Campbell, D. T. (1988). The Experimenting Society. In E. S. Overman (Ed.), Methodology and epistemology for social science: Selected Papers. (pp. 303). Chicago: University of Chicago Press.
Tuesday, March 10, 2009
Mike Smith, a senior advisor to Secretary Duncan, recently outlined the goals of the stimulus plan this way: (1) get the money out fast; (2) create jobs; and (3) stimulate reform. While some hope that this funding amount will constitute a new default level, we are being warned to expect a “cliff” when the one-time funding is exhausted. Thus a wise use of this money would be for new jobs that have the effect of creating something that won’t have to be paid for at the same level on an ongoing basis. Repairing a school building illustrates the idea. The work can start quickly and generate short-term jobs and, once completed, the repaired building will be around for a long time before another infusion of repair money is needed.
What is the analogy in the domain of education reform? If a school system uses stimulus money to purchase services with a fixed annual cost, the service may have to be discontinued when the money is spent. On the other hand, if the school system purchases two years of professional development, then teachers may be able to carry their new practices forward without continuing operational costs. Similarly, investing in data systems and building school district capacity for analyzing local data and for evaluating programs could pay off in sustained systemic improvements in district decision processes.
The stimulus package does contain funds for data systems-an investment that will pay off beyond the initial implementation. By themselves, data systems are technical capacities, not reforms; alone they do not change how educators can generate useful evidence that can improve instruction. With the stimulus funding, we now have an opportunity to put in place local district capacities for data use-not just at the classroom level but also at the level of district and state policies and strategies.
To harness the stimulus funds for reforms involving data systems, we need to look at the instances in the stimulus package where data would actually be used. This leads us to the places calling for evaluations. For example, the $650 million in technology grants will flow through the mechanism of Title II D, which includes the suggestion to use funds to “support the rigorous evaluation of programs funded under this part, particularly regarding the impact of such programs on student academic achievement…” Thus these funds may be used to enhance the use of data systems to conduct local evaluations of the technologies, thereby building the capacity of districts to generate useful evidence.
In a section on teacher incentive programs, the stimulus bill specifically calls for “a rigorous national evaluation.” We don’t think this should be interpreted as a single large national evaluation. In our view, a large number of local evaluations would be a more productive use of the funds, as long as each is rigorous. Doing so—that is, distributing the available funds to the local districts implementing the incentive programs—stimulates district hiring (or retaining district staff who specialize in evaluations).
We should be viewing data use and evidence generation in districts as an important part of the reform agenda and not just as an isolated technical problem of data warehousing. By distributing program evaluation to districts, we can fulfill the goals of creating (or retaining) jobs while stimulating reform that will endure beyond the two years when the extraordinary funds are available. —DN
Friday, February 13, 2009
The headline in the January 28 issue of Education Week suggests that the pendulum is swinging from obtaining rigorous scientific evidence to providing greater freedom for development and innovation.
Is there reason to believe this is more than a war of catch phrases? Does supporting innovative approaches take resources away from “scientifically based research”? A January 30 interview with Arne Duncan, our new Secretary of Education, by CNN’s Campbell Brown is revealing. She asked him about the innovative program in Chicago that pays students for better grades. Here is how the conversation went:
Duncan: ...in every other profession we recognize, reward and incent excellence. I think we need to do more of that in education.
CNN: For the students specifically, you think money is the way to do that? It’s the best incentive?
Duncan: I don’t think it is the best incentive; I think it's one incentive. This is a pilot program we started this fall so it’s very early on. But so far the data is very encouraging—so far the students’ attendance rates have gone up, students’ grades have gone up, and these are communities where the drop out rate has been unacceptably high and whatever we can do to challenge that status quo. When children drop out today, Campbell as you know, they are basically condemned to social failure. There are no good jobs out there so we need to be creative; we need to push the envelope. I don’t know if this is the right answer. We’ve got a control group.
CNN: But is it something that you would like to try across the country, to have other schools systems adopt?
Duncan: Again, Campbell, this is...we are about four months into it in Chicago. We have a control group where this is not going on, so we’re going to follow what the data tells us. And if it’s successful, we’ll look to expand it. If it’s not successful, we’ll stop doing it. We want to be thoughtful but I think philosophically I am pro pushing the envelope, challenging the status quo, and thinking outside the box...
Read more from the interview here.
Notice that he is calling for innovation: “pushing the envelope challenging the status quo, thinking outside the box.” But he is not divorcing innovation from rigorously controlled effectiveness research. He is also looking at preliminary findings of changes such as attendance rates that can be detected early. The “scientifically based” research is built into the innovation’s implementation at the earliest stage. And it is used as a basis for decisions about expansion.
While there may be reason to increase funding for development, we can’t divorce rigorous research from development; nor should we consider experimental evaluations as activities that kick in after an innovation has been fielded. We are skeptical that there is a real conflict between scientific research and innovation. The basic problem for research and development in education is not too much attention to rigorous research, but too little overall resources going into education. The graphic included in the Ed Week story makes clear that the R&D investment in education is miniscule compared to R&D for the military, energy, and heath sectors (health gets 100 times as much as education, whereas the category “other” gets 16 times as much).
The Department of Defense, of course, gets a large piece of the pie, and we often see the Defense Advanced Research Projects Agency (DARPA) held up as an example of a federal agency devoted to addressing innovative and often futuristic, requirements. The Internet started as one such engineering project, the ARPANET. In this case, researchers needed to access information flexibly from computers situated around the country and the innovative distributed approach turned out to be massively scalable and robust. Although we don’t always see that research is built in as a continuous part of engineering development, every step was in fact an experiment. While testing an engineering concept may take a few minutes of data collection, in education, the testing can be more cumbersome. Cognitive development and learning take months or years to generate measurable gains and experiments need careful ways to eliminate confounders that often don’t trouble engineering projects. Education studies are also not as amenable to clever technical solutions (although the vision of something as big the Internet coming in to disrupt and reconfigure education is tantalizing).
It is always appealing to see the pendulum swing with a changing of the guard. In the current transition, what is coming about looks more like a synthesis in which research and development are no longer treated as separate— and certainly not seen as competitors. —DN
Thursday, January 1, 2009
The Consortium for School Networking (CoSN), of which Empirical Education is a member and long time supporter, advocates for technology for schools. In an article entitled “Why Obama Can’t Ignore Ed Tech”, Jim Goodnight, founder and CEO of SAS, and Keith Krueger, CEO of CoSN, argue for investing in education technology as a way to support “21st century learning” while creating jobs in the technology and telecommunications sectors. They also suggest that the investment will lead school districts to hire staff members specializing in technical and technology curricula, a function they note as currently being “vastly understaffed.”
As a research organization, we have to maintain a cautious attitude about claims, such as those in the CoSN article, that technology products will reduce discipline problems and dropout rates generally. We do agree that an investment in school technology will call for increased staffing—that is, creating jobs—which is the primary goal of the stimulus package.
But we believe there is a better argument for an investment in technical infrastructure. Network and data warehouse technologies inherently provide the mechanisms for measuring whether the investments are making a difference. Combined with online formative testing, automatic generation of usage data, and analytic tools, these technologies will put schools in a position to keep technology accountable for promised results. Using technology as a tool for tracking results of the stimulus package will, of course, create jobs. It will call for the creation of additional positions for data coaches, data analysts, trainers, and staff to handle the test administration, data cleaning, and communication functions.
The fear that a stimulus package will just throw money at the problem is justified. Yes, it will provide jobs and benefits to certain industries in the short term, whereas any lasting improvement may be elusive. While building a new bridge employs construction workers and the lasting benefit can be measured, for example, by improved traffic, the lasting benefits of school technology may seem more subtle. We would argue, to the contrary, that a technology infrastructure for schools contains its own mechanism for accountability. The argument for school technology should drive home the notion that schools can be capable of determining whether the stimulus investment is having an impact on learning, discipline, graduation rates, and other measurable outcomes. Policy makers will not have to depend on promises of new forms of learning when they can put in place a technology infrastructure that provides school decision-makers with the information about whether the investment is making a difference. —DN