blog posts and news stories

How Efficacy Studies Can Help Decision-makers Decide if a Product is Likely to Work in Their Schools

We and our colleagues have been working on translating the results of rigorous studies of the impact of educational products, programs, and policies for people in school districts who are making the decisions whether to purchase or even just try out—pilot—the product. We are influenced by Stanford University Methodologist Lee Cronbach, especially his seminal book (1982) and article (1975) where he concludes “When we give proper weight to local conditions, any generalization is a working hypothesis, not a conclusion…positive results obtained with a new procedure for early education in one community warrant another community trying it. But instead of trusting that those results generalize, the next community needs its own local evaluation” (p. 125). In other words, we consider even the best designed experiment to be like a case study, as much about the local and moderating role of context, as about the treatment when interpreting the causal effect of the program.

Following the focus on context, we can consider characteristics of the people and of the institution where the experiment was conducted to be co-causes of the result that deserve full attention—even though, technically, only the treatment, which was randomly assigned was controlled. Here we argue that any generalization from a rigorous study, where the question is whether the product is likely to be worth trying in a new district, must consider the full context of the study.

Technically, in the language of evaluation research, these differences in who or where the product or “treatment” works are called “interaction effects” between the treatment and the characteristic of interest (e.g., subgroups of students by demographic category or achievement level, teachers with different skills, or bandwidth available in the building). The characteristic of interest can be called a “moderator”, since it changes, or moderates, the impact of the treatment. An interaction reveals if there is differential impact and whether a group with a particular characteristic is advantaged, disadvantaged, or unaffected by the product.

The rules set out by The Department of Education’s What Works Clearinghouse (WWC) focus on the validity of the experimental conclusion: Did the program work on average compared to a control group? Whether it works better for poor kids than for middle class kids, works better for uncertified teachers versus veteran teachers, increases or closes a gap between English learners and those who are proficient, are not part of the information provided in their reviews. But these differences are exactly what buyers need in order to understand whether the product is a good candidate for a population like theirs. If a program works substantially better for English proficient students than for English learners, and the purchasing school has largely the latter type of student, it is important that the school administrator know the context for the research and the result.

The accuracy of an experimental finding depends on it not being moderated by conditions. This is recognized with recent methods of generalization (Tipton, 2013) that essentially apply non-experimental adjustments to experimental results to make them more accurate and more relevant to specific local contexts.

Work by Jaciw (2016a, 2016b) takes this one step further.

First, he confirms the result that if the impact of the program is moderated, and if moderators are distributed differently between sites, then an experimental result from one site will yield a biased inference for another site. This would be the case, for example, if the impact of a program depends on individual socioeconomic status, and there is a difference between the study and inference sites in the proportion of individuals with low socioeconomic status. Conditions for this “external validity bias” are well understood, but the consequences are addressed much less often than the usual selection bias. Experiments can yield accurate results about the efficacy of a program for the sample studied, but that average may not apply either to a subgroup within the sample or to a population outside the study.

Second, he uses results from a multisite trial to show empirically that there is potential for significant bias when inferring experimental results from one subset of sites to other inference sites within the study; however, moderators can account for much of the variation in impact across sites. Average impact findings from experiments provide a summary of whether a program works, but leaves the consumer guessing about the boundary conditions for that effect—the limits beyond which the average effect ceases to apply. Cronbach was highly aware of this, titling a chapter in his 1982 book “The Limited Reach of Internal Validity”. Using terms like “unbiased” to describe impact findings from experiments is correct in a technical sense (i.e., the point estimate, on hypothetical repeated sampling, is centered on the true average effect for the sample studied), but it can impart an incorrect sense of the external validity of the result: that it applies beyond the instance of the study.

Implications of the work cited, are, first, that it is possible to unpack marginal impact estimates through subgroup and moderator analyses to arrive at more-accurate inferences for individuals. Second, that we should do so—why obscure differences by paying attention to only the grand mean impact estimate for the sample? And third, that we should be planful in deciding which subgroups to assess impacts for in the context of individual experiments.

Local decision-makers’ primary concern should be with whether a program will work with their specific population, and to ask for causal evidence that considers local conditions through the moderating role of student, teacher, and school attributes. Looking at finer differences in impact may elicit criticism that it introduces another type of uncertainty—specifically from random sampling error—which may be minimal with gross impacts and large samples, but influential when looking at differences in impact with more and smaller samples. This is a fair criticism, but differential effects may be less susceptible to random perturbations (low power) than assumed, especially if subgroups are identified at individual levels in the context of cluster randomized trials (e.g., individual student-level SES, as opposed to school average SES) (Bloom, 2005; Jaciw, Lin, & Ma, 2016).

Bloom, H. S. (2005). Randomizing groups to evaluate place-based programs. In H. S. Bloom (Ed.), Learning more from social experiments. New York: Russell Sage Foundation.

Cronbach, L. J. (1975). Beyond the two disciplines of scientific psychology. American Psychologist, 116-127.

Cronbach, L. J. (1982). Designing evaluations of educational and social programs. San Francisco, CA: Jossey-Bass.

Jaciw, A. P. (2016). Applications of a within-study comparison approach for evaluating bias in generalized causal inferences from comparison group studies. Evaluation Review, (40)3, 241-276. Retrieved from

Jaciw, A. P. (2016). Assessing the accuracy of generalized inferences from comparison group studies using a within-study comparison approach: The methodology. Evaluation Review, (40)3, 199-240. Retrieved from

Jaciw, A., Lin, L., & Ma, B. (2016). An empirical study of design parameters for assessing differential impacts for students in group randomized trials. Evaluation Review. Retrieved from

Tipton, E. (2013). Improving generalizations from experiments using propensity score subclassification: Assumptions, properties, and contexts. Journal of Educational and Behavioral Statistics, 38, 239-266.


Spring 2018 Conference Season is Taking Shape

We’ll be on the road again this spring.


Andrew Jaciw and Denis Newman will be in Washington DC for the annual spring conference of the The Society for Research on Educational Effectiveness (SREE), the premier conference on rigorous research. Andrew Jaciw will present his paper: Leveraging Fidelity Data to Making Sense of Impact Results: Informing Practice through Research. His presentation will be a part of Session 2I: Research Methods - Post-Random Assignment Models: Fidelity, Attrition, Mediation & More from 8-10am on Thursday, March 1.


In March, Denis Newman will be attending SXSW EDU Conference & Festival in Austin, TX and presenting on a panel along with Malvika Bhagwat, Jason Palmer, and Karen Billings titled Can Evidence Even Keep Up with EdTech? This will address how researchers and companies can produce evidence that products work—in time for educators and administrators to make a knowledgeable buying decision under accelerating timelines.


Empirical staff will be presenting in 4 different sessions at the annual conference of the American Educational Research Association (AERA) in NYC in April, all under Division H (Research, Evaluation, and Assessment in Schools).

  1. For Quasi-experiments on Edtech Products, What Counts as Being Treated?
  2. Teacher evaluation rubric properties and associations with school characteristics: Evidence from the Texas evaluation system
  3. Indicators of Successful Teacher Recruitment and Retention in Oklahoma Rural Schools
  4. The Challenges and Successes of Conducting Large-scale Educational Research

In addition to these presentations, we are planning another of our celebrated receptions in NYC so stay tuned for details.


A panel on our Research Guidelines has been accepted at this major convention, considered the epicenter of edtech with thousands of users and 100s of companies, held this year in Chicago from June 24–27.


Getting Different Results from the Same Program in Different Contexts

The spring 2014 conference of the Society for Research in Educational Effectiveness (SREE) gave us much food for thought concerning the role of replication of experimental results in social science research. If two research teams get the same result from experiments on the same program, that gives us confidence that the original result was not a fluke or somehow biased.

But in his keynote, John Ioannidis of Stanford showed that even in medical research, where the context can be more tightly controlled, replication very often fails—researchers get different results. The original finding may have been biased, for example, through the tendency to suppress null findings where no positive effect was found and over-report large, but potentially spurious results. Replication of a result over the long run helps us to get past the biases. Though not as glamorous as discovery, replication is fundamental to science, and educational science is no exception.

In the course of the conference, I was reminded that the challenge to conducting replication work is, in a sense, compounded in social science research. “Effect heterogeneity”—finding different results in different contexts—is common for many legitimate reasons. For instance, experimental controls seldom get placebos. They receive the program already in place, often referred to as “business as usual,” and this can vary across experiments of the same intervention and contribute to different results. Also, experiments of the same program carried out in different contexts are likely to be adapted given demands or affordances of the situation, and flexible implementation may lead to different results. The challenge is to disentangle differences in effects that give insight into how programs are adapted in response to conditions, from bias in results that John Ioannidis considered. In other fields (e.g., the “hard sciences”), less context dependency and more-robust effects may make it easier to diagnose when variation in findings is illegitimate. In education, this may be more challenging and reminds me why educational research is in many ways the ‘hardest science’ of all, as David Berliner has emphasized in the past.

Once separated from distortions of bias and properly differentiated from the usual kind of “noise” or random error, differences in effects can actually be leveraged to better understand how and for whom programs work. Building systematic differences in conditions into our research designs can be revealing. Such efforts should, however, be considered with the role of replication in mind—an approach to research that purposively builds in heterogeneity, in a sense, seeks to find where impacts don’t replicate, but for good reason. Non-reproducibility in this case is not haphazard, it is purposive.

What are some approaches to leveraging and understanding effect heterogeneity? We envision randomized trials where heterogeneity is built into the design by comparing different versions of a program or implementing in diverse settings across which program effects are hypothesized to vary. A planning phase of an RCT would allow discussions with experts and stakeholders about potential drivers of heterogeneity. Pertinent questions to address during this period include: what are the attributes of participants and settings across which we expect effects to vary and why? Under which conditions and how do we expect program implementation to change? Hypothesizing which factors will moderate effects before the experiment is conducted would add credibility to results if they corroborate the theory. A thoughtful approach of this sort can be contrasted with the usual approach whereby differential effects of program are explored as afterthoughts, with the results carrying little weight.

Building in conditions for understanding effect heterogeneity will have implications for experimental design. Increasing variation in outcomes affects statistical power and the sensitivity of designs to detect effects. We will need a better understanding of the parameters affecting precision of estimates. At Empirical, we have started using results from several of our experiments to explore parameters affecting sensitivity of tests for detecting differential impact. For example, we have been documenting the variation across schools in differences in performance depending on student characteristics such as individual SES, gender, and LEP status. This variation determines how precisely we are able to estimate the average difference between student subgroups in the impact of a program.

Some may feel that introducing heterogeneity to better understand conditions for observing program effects is going down a slippery slope. Their thinking is that it is better to focus on program impacts averaged across the study population and to replicate those effects across conditions; and that building sources of variation into the design may lead to loose interpretations and loss of rigor in design and analysis. We appreciate the cautionary element of this position. However, we believe that a systematic study of how a program interacts with conditions can be done in a disciplined way without giving up the usual strategies for ensuring the validity of results.

We are excited about the possibility that education research is entering a period of disciplined scientific inquiry to better understand how differences in students, contexts, and programs interact, with the hope that the resulting work will lead to greater opportunity and better fit of program solutions to individuals.


State Reports Show Almost All Teachers Are Effective or Highly So. Is This Good News?

The New York Times recently picked up a story, originally reported in Education Week two months ago, that school systems using formal methods for classroom observation as part of their educator evaluations are giving all but a very small percent of teachers high ratings—a phenomenon commonly known as the “widget effect.” The Times quotes Russ Whitehurst as suggesting that “It would be an unusual profession that at least 5 percent are not deemed ineffective.”

Responding to the story in her blog, Diane Ravitch calls it “unintentionally hilarious,” portraying the so-called reformers as upset that their own expensive evaluation methods are finding that most teachers are good at what they do. In closing, she asks, “Where did all those ineffective teachers go?”

We’re a research company working actively on teacher evaluation, so we’re interested in these kinds of questions. Should state-of-the-art observation protocols have found more teachers in the “needs improvement” category or at least 5% labeled “ineffective”? We present here an informal analysis meant to get an approximate answer, but based on data that was collected in a very rigorous manner. As one of the partners in the Gates Foundation’s Measures of Effective Teaching (MET) project, Empirical Education has access to a large dataset available for this examination, including videotaped lessons for almost 2,000 teachers coded according to a number of popular observational frameworks. Since the MET raters were trained intensively using methods approved by the protocol developers and had no acquaintance or supervisory relationship with the teachers in the videos, there is reason to think that the results show the kind of distribution intended by the developers of the observation methods. We can then compare the results in this controlled environment to the results referred to in the EdWeek and Times articles, which were based on reporting by state agencies. We used a simple (but reasonable) way of calculating the distribution of teachers in the MET data according to the categories in one popular protocol and compared it to the results reported by one of the states for a district known to have trained principals and other observers in the same protocol. We show the results here. The light bars show the distribution of the ratings in the MET data. We can see that a small percentage are rated “highly effective” and an equally small percentage “unsatisfactory.” So although the number doesn’t come up to the percent suggested by Russ Whitehurst, this well-developed method finds only 2% of a large sample of teachers to be in the bottom category. About 63% are considered “effective”, while a third are given a “needs improvement” rating. The dark bars are the ratings given by the school district using the same protocol. This shows a distribution typical of what EdWeek and the Times reported, where 97% are rated as “highly effective” or “effective.” It is interesting that the school district and MET research both found a very small percentage of unsatisfactory teachers.

Where we find a big difference is in the fact that the research program deemed only a small number of teachers to be exceptional while the school system used that category much more liberally. The other major difference is in the “needs improvement” category. When the observational protocol is used as designed, a solid number of teachers are viewed as doing OK but potentially doing much better. Both in research and in practice, the observational protocol divides most teachers between two categories. In the research setting, the distinction is between teachers who are effective and those who need improvement. In practice, users of the same protocol distinguish between effective and highly effective teachers. Both identify a small percent as unsatisfactory.

Our analysis suggests two problems with the use of the protocol in practice: first, the process does not provide feedback to teachers who are developing their skills, and, second, it does not distinguish between very good teachers and truly exceptional ones. We can imagine all sorts of practical pressures that, for the evaluators (principals, coaches and other administrators) decrease the value of identifying teachers who are less than fully effective and can benefit from developing specific skills. For example, unless all the evaluators in a district simultaneously agree to implement more stringent evaluations, then teachers in the schools where such evaluations are implemented will be disadvantaged. It will help to also have consistent training and calibration for the evaluators as well as accountability, which can be done with a fairly straightforward examination of the distribution of ratings.

Although this was a very informal analysis with a number of areas where we approximated results, we think we can conclude that Russ Whitehurst probably overstated the estimate of ineffective teachers but Diane Ravitch probably understated the estimate of teachers who could use some help and guidance in getting better at what they do.

Postscript. Because we are researchers and not committed to the validity of the observational methods, we need to state that we don’t know the extent to which the teachers labeled ineffective are generally less capable of raising student achievement. But researchers are notorious for ending all our reports with “more research is needed!”


Making Vendor Research More Credible

The latest evidence that research can be both rigorous and relevant was the subject of an announcement that the Software and Information Industry Association (SIIA) made last month about their new guidelines for conducting effectiveness research. The document is aimed at SIIA members, most of whom are executives of education software and technology companies and not necessarily schooled in research methodology. The main goal in publishing the guidelines is to improve the quality—and therefore the credibility—of research sponsored by the industry. The document provides SIIA members with things to keep in mind when contracting for research or using research in marketing materials. The document also has value for educators, especially those responsible for purchasing decisions. That’s an important point that I’ll get back to.

One thing to make clear in this blog entry is that while your humble blogger (DN) is given credit as the author, the Guidelines actually came from a working group of SIIA members who put in many months of brainstorming, discussion, and review. DN’s primary contribution was just to organize the ideas, ensure they were technically accurate, and put them into easy to understand language.

Here’s a taste of some of the ideas contained in the 22 guidelines:

  • With a few exceptions, all research should be reported regardless of the result. Cherry picking just the studies with strong positive results distorts the facts and in the long run hurts credibility. One lesson that might be taken from this is that conducting several small studies may be preferable to trying to prove a product effective (or not) in a single study.

  • Always provide a link to the full report. Too often in marketing materials (including those of advocacy groups, not just publishers) a fact such as “8th grade math achievement increased from 31% in 2004 to 63% in 2005,” is offered with no citation. In this specific case, the fact was widely cited but after considerable digging could be traced back to a report described by the project director as “anecdotal”.

  • Be sure to take implementation into account. In education, all instructional programs require setting up complex systems of teacher-student interaction, which can vary in numerous ways. Issues of how research can support the process and what to do with inadequate or outright failed implementation must be understood by researchers and consumers of research.

  • Watch out for the control condition. In education there are no placebos. In almost all cases we are comparing a new program to whatever is in place. Depending on how well the existing program works, the program being evaluated may appear to have an impact or not. This calls for careful consideration of where to test a product and understandable concern by educators as to how well a particular product tested in another district will perform against what is already in place in their district.

The Guidelines are not just aimed at industry. SIIA believes that as decision-makers at schools begin to see a commitment to providing stronger research, their trust in the results will increase. It is also in the educators’ interest to review the guidelines because they provide a reference point for what actionable research should look like. Ultimately, the Guidelines provide educators with help in conducting their own research, whether it is on their own or in partnership with the education technology providers.


Software Industry Sets High Standards for Product Evaluation Research

The Software & Information Industry Association (SIIA) announced the release of their new report, authored by our very own Dr. Denis Newman under the direction of the SIIA Education Division’s Research & Evaluation Working Group, the guidelines provide practical considerations and share best practices of product evaluation design, conduct, and reporting. Written primarily for publishers and developers of education technology, the guidelines reflect the high standards necessary to carry out rigorous, unbiased effectiveness research. Reviewers of the guidelines included Larry Hedges with Northwestern University, Robert Slavin with Johns Hopkins University, and Talbot Bielefeldt with the International Society for Technology in Education (ISTE). A delegation of software publishers presented the Guidelines May 17 at the US Department of Education to John Q. Easton (Director of IES) and Karen Cator (Director of the Office of Education Technology). The document is now available to the public at the link above.


Research: From NCLB to Obama’s Blueprint for ESEA

We can finally put “Scientifically Based Research” to rest. The term that appeared more than 100 times in NCLB appears zero times in the Obama administration’s Blueprint for Reform, which is the document outlining its approach to the reauthorization of ESEA. The term was always an awkward neologism, coined presumably to avoid simply saying “scientific research.” It also allowed NCLB to contain an explicit definition to be enforced—a definition stipulating not just any scientific activities, but research aimed at coming to causal conclusions about the effectiveness of some product, policy, or laboratory procedure.

A side effect of the SBR focus has been the growth of a compliance mentality among both school systems and publishers. Schools needed some assurance that a product was backed by SBR before they would spend money, while textbooks were ranked in terms of the number of SBR-proven elements they contained.

Some have wondered if the scarcity of the word “research” in the new Blueprint might signal a retreat from scientific rigor and the use of research in educational decisions (see, for example, Debra Viadero’s blog). Although the approach is indeed different, the new focus makes a stronger case for research and extends its scope into decisions at all levels.

The Blueprint shifts the focus to effectiveness. The terms “effective” or “effectiveness” appear about 95 times in the document. “Evidence” appears 18 times. And the compliance mentality is specifically called out as something to eliminate.

“We will ask policymakers and educators at all levels to carefully analyze the impact of their policies, practices, and systems on student outcomes. … And across programs, we will focus less on compliance and more on enabling effective local strategies to flourish.” (p. 35)

Instead of the stiff definition of SBR, we now have a call to “policymakers and educators at all levels to carefully analyze the impact of their policies, practices, and systems on student outcomes.” Thus we have a new definition for what’s expected: carefully analyzing impact. The call does not go out to researchers per se, but to policymakers and educators at all levels. This is not a directive from the federal government to comply with the conclusions of scientists funded to conduct SBR. Instead, scientific research is everybody’s business now.

Carefully analyzing the impact of practices on student outcomes is scientific research. For example, conducting research carefully requires making sure the right comparisons are made. A study that is biased by comparing two groups with very different motivations or resources is not a careful analysis of impact. A study that simply compares the averages of two groups without any statistical calculations can mistakenly identify a difference when there is none, or vice versa. A study that takes no measure of how schools or teachers used a new practice—or that uses tests of student outcomes that don’t measure what is important—can’t be considered a careful analysis of impact. Building the capacity to use adequate study design and statistical analysis will have to be on the agenda of the ESEA if the Blueprint is followed.

Far from reducing the role of research in the U.S. education system, the Blueprint for ESEA actually advocates a radical expansion. The word “research” is used only a few times, and “science” is used only in the context of STEM education. Nonetheless, the call for widespread careful analysis of the evidence of effective practices that impact student achievement broadens the scope of research, turning all policymakers and educators into practitioners of science.


MeasureResults® to be Launched at CoSN 2010

Empirical Education will launch its web-based educational research solution, MeasureResults on March 1 at the Consortium for School Networking conference in Washington, DC. MeasureResults is a suite of online tools that makes rigorous research designs and statistical processes accessible to school systems and educational publishers who want to evaluate the effectiveness of products and services aimed at improving student performance.

“MeasureResults will change the way that school districts and product developers conduct rigorous evaluations,” said Denis Newman, Empirical Education President. “Instead of hiring outside evaluators or onsite research experts or statisticians, MeasureResults allows school district personnel to design a study, collect data, and review reports in our user-friendly online platform.”

MeasureResults grew out of a federally funded research project to develop a low-cost method for schools to conduct their own research. The product was developed for commercial distribution under a Small Business Innovation Research grant from the U.S. Department of Education. By moving the educational research processes online, MeasureResults makes school-run evaluations more efficient and less expensive.


Stimulating Innovation and Evidence

After a massive infusion of stimulus money into K-12 technology through the Title IID “Enhancing Education Through Technology” (EETT) grants, known also as “ed-tech” grants, the administration is planning to cut funding for the program in future budgets.

Well, they’re not exactly “cutting” funding for technology, but consolidating the dedicated technology funding stream into a larger enterprise, awkwardly named the “Effective Teaching and Learning for a Complete Education” program. For advocates of educational technology, here’s why this may not be so much a blow as a challenge and an opportunity.

Consider the approach stated at the White House “fact sheet”:

“The Department of Education funds dozens of programs that narrowly limit what states, districts, and schools can do with funds. Some of these programs have little evidence of success, while others are demonstrably failing to improve student achievement. The President’s Budget eliminates six discretionary programs and consolidates 38 K-12 programs into 11 new programs that emphasize using competition to allocate funds, giving communities more choices around activities, and using rigorous evidence to fund what works…Finally, the Budget dedicates funds for the rigorous evaluation of education programs so that we can scale up what works and eliminate what does not.”

From this, technology advocates might worry that policy is being guided by the findings of “no discernable impact” from a number of federally funded technology evaluations (including the evaluation mandated by the EETT legislation itself).

But this is not the case. The White House declares, “The President strongly believes that technology, when used creatively and effectively, can transform education and training in the same way that it has transformed the private sector.”

The administration is not moving away from the use of computers, electronic whiteboards, data systems, Internet connections, web resources, instructional software, and so on in education. Rather, the intention is that these tools are integrated, where appropriate and effective, into all of the other programs.

This does put technology funding on a very different footing. It is no longer in its own category. Where school administrators are considering funding from the “Effective Teaching and Learning for a Complete Education” program, they may place a technology option up against an approach to lower class size, a professional development program, or other innovations that may integrate technologies as a small piece of an overall intervention. Districts would no longer write proposals to EETT to obtain financial support to invest in technology solutions. Technology vendors will increasingly be competing for the attention of school district decision-makers on the basis of the comparative effectiveness of their solution—not just in comparison to other technologies but in comparison to other innovative solutions. The administration has clearly signaled that innovative and effective technologies will be looked upon favorably. It has also signaled that effectiveness is the key criterion.

As an Empirical Education team prepares for a visit to Washington DC for the conference of the Consortium for School Networking and the Software and Information Industry Association’s EdTech Government Forum, (we are active members in both organizations), we have to consider our message to the education technology vendors and school system technology advocates. (Coincidentally, we will also be presenting research at the annual conference of the Society for Research on Educational Effectiveness, also held in DC that week). As a research company we are constrained from taking an advocacy role—in principle we have to maintain that the effectiveness of any intervention is an empirical issue. But we do see the infusion of short term stimulus funding into educational technology through the EETT program as an opportunity for schools and publishers. Working jointly to gather the evidence from the technologies put in place this year and next will put schools and publishers in a strong position to advocate for continued investment in the technologies that prove effective.

While it may have seemed so in 1993 when the U.S. Department of Education’s Office of Educational Technology was first established, technology can no longer be considered inherently innovative. The proposed federal budget is asking educators and developers to innovate to find effective technology applications. The stimulus package is giving the short term impetus to get the evidence in place.