480 likes | 570 Views
The Quantitative Side of Teacher Evaluation in New Jersey. Bruce D. Baker. Modern Teacher Evaluation Policies Making Certain Distinctions with Uncertain Information.
E N D
The Quantitative Side of Teacher Evaluation in New Jersey Bruce D. Baker
Modern Teacher Evaluation PoliciesMaking Certain Distinctions with Uncertain Information • First, the modern teacher evaluation template requires that objective measures of student achievement growth necessarily be considered in a weighting system of parallel components. • Placing the measures alongside one another in a weighting scheme assumes all measures in the scheme to be of equal validity and reliability but of varied importance (utility) – varied weight. • Second, modern teacher evaluation template requires that teachers be placed into effectiveness categories by assigning arbitrary numerical cutoffs to the aggregated weighted evaluation components. • That is, a teacher in the 25%ile or lower when combining all evaluation components might be assigned a rating of “ineffective,” whereas the teacher at the 26%ile might be labeled effective. • Third, the modern teacher evaluation template places inflexible timelines on the conditions for removal of tenure. • Typical legislation dictates that teacher tenure either can or must be revoked and the teacher dismissed after 2 consecutive years of being rated ineffective (where tenure can only be achieved after 3 consecutive years of being rate effective).
Due Process Concerns • Attribution/Validity • Face • There are many practical challenges including whether teachers should be held responsible for summer learning in an annual assessment model, or how to parse influence of teacher teams and/or teachers with assistants. • SGPs have their own face validity problem in the authors own words. One cannot reasonably evaluate someone on a measure not attributable to them. • Statistical • That there may be, and likely will be with an SGP, significant persistent bias – that is, other stuff affecting growth such as student assignment, classroom conditions, class sizes, etc. – which render the resulting estimate NOT attributable to the teacher. • Reliability • Lack of reliability of measures, jumping around from year to year, suggests also that the measures are not a valid representation of actual teacher quality. • Arbitrary/Unjustifiable Decisions • Cut-points imposed throughout the system make invalid assumptions regarding the statistics – that a 1 point differential is meaningful.
Legal Parallels • NJ Anti-Bullying Statute • Damned if you do, damned if you don’t! • Bullying statute forces district officials to make distinctions/apply labels which may not be justifiably applied (“bully” or “bullying behavior”) • Application of these labels leads to damages to the individual to whom they are applied (liberty interests/property interests) • Result? • Due process challenges • District backtracking (avoidance) • Measuring good teaching is no more precise an endeavor than measuring bullying! (and state mandates cannot make it so!)
TEACH NJ Issues Arbitrary Weightings Assumes equal Validity, but Varied Importance
TEACH NJ Issues Arbitrary Reduction/ Categorization of (forced) normally distributed data
Statistically inappropriate to declare +.49sd categorically different from +.50sd! [especially with noisy data]
Arbitrary, statistically unjustifiable collapsing to 4pt scale, then multiplied times arbitrary weight Assumes all parts equally valid & reliable, but of varied importance (wt)
At best, this applies to 20 to 30% of teachers • At the middle school level! (which would be the highest) • Requires differential contracts by staffing type • Some states/school districts “resolve” this problem by assigning all other teachers the average of those rated: • Significant “attribution” concern (due process) • Induces absurd practices • This problem undermines “reform” arguments that in cases of RIF, quality, not seniority should prevail because supposed “quality” measures only apply to those positions least likely to be reduced.
Usefulness of Teacher Effect Measures • Estimating teacher effects • Basic attribution problems • Seasonality • Spill-over • Stability Issues • Decomposing the Signal and the Noise • SGP and VAM • Attribution? • False signals in NJ SGP data • Debunking Disinformation
A little stat-geeking • What’s in a growth or VAM estimate? • The largest part is random noise… that is, if we look from year to year, across the same teachers, estimates jump around a lot, or vary a lot in “unexplained” and seemingly unpredictable ways. • The other two parts are: • False Signal, or predictable patterns that are predictable not as a function of anything the teacher is doing, but a function of other stuff outside the teacher’s control, that happens to have predictable influence • Student sort, classroom conditions, summer experiences, test form/scale and starting position of students on that scale. • True Signal, or that piece of the predictability of change in test score from time 1 to time 2 that might fairly be attributed to the role of the teacher in the classroom.
Distilling Signal from Noise Making high stakes personnel decisions on the basis of either Noise or False Signal is problematic! [& that may be the majority of the variation] Difficult if not implausible to accurately parse
SGPs & New Jersey • Student Growth Percentiles are not designed for inferring teacher influence on student outcomes. • Student Growth Percentiles do not (even try to) control for various factors outside of the teacher’s control. • Student Growth Percentiles are not backed by research on estimating teacher effectiveness. By contrast, research on SGPs has shown them to be poor at isolating teacher influence. • New Jersey’s Student Growth Percentile measures, at the school level, are significantly statistically biased with respect to student population characteristics and average performance level.
In the authors words… Damian Betebenner: “Unfortunately Professor Baker conflates the data (i.e. the measure) with the use. A primary purpose in the development of the Colorado Growth Model (Student Growth Percentiles/SGPs) was to distinguish the measure from the use: To separate the description of student progress (the SGP) from the attribution of responsibility for that progress.” http://www.ednewscolorado.org/voices/student-growth-percentiles-and-shoe-leather
Playing Semantics… • When pressed on the point that GPs are not designed for attributing student gains to their teachers, those defending their use in teacher evaluation will often say… • “SGPs are a good measure of student growth, and shouldn’t teachers be accountable for student growth?” • Let’s be clear here, one cannot be accountable for something that is not rightly attributable to them!
The Bottom Line? • Policymakers seem to be moving forward on implementation of policies that display complete disregard for basic statistical principles – • that one simply cannot draw precise conclusions (and thus make definitive decisions) based on imprecise information. • Can’t draw a strict cut point through messy data. Same applies to high stakes cut scores for kids. • That one cannot make assertions about the accuracy of the position of any one point among thousands, based on the loose patterns we find in these types of data. • Good data informed decision making requires deep nuanced understanding of statistics, measures, what they mean… and most importantly WHAT THEY DON’T!(and can’t possibly)
Reasonable Alternatives? • To the extent these data can produce some true signal amidst the false signal and noise, central office data teams in large districts might be able to use several (not just one) rich, but varied models to screen for variations that warrant further exploration. • This screening approach, much like high-error-rate rapid diagnostics tests, might tell us where to focus some additional energy (that is, classroom and/or school observation). • We may then find that the signal was false, or that it really does tell us something either about how we’ve mismatched teachers and assignments, or the preparedness of some teachers. • But, the initial screening information should NEVER dictate the final decision (as it will under Toxic Trifecta models). • But, if we find that the data-driven analysis more often sends us down inefficient pathways, we might decide it’s just not worth it. But this cannot be achieved by centralized policy or through contractual agreements. Unfortunately current policies and recent contractual agreements prohibit thoughtful, efficient strategies! Screening Observation Validation [or NOT] & Questions?
9 to 15% (of those who were “good” or were “bad” in the previous year) move all the way from good to bad or bad to good. 20 to 35% who were “bad” stayed “bad” & 20 to 35% who were “good” stayed “good.” And this is between the two years that show the highest correlation for ELA.
For math, only about 7% of teachers jump all the way from being bad to good or good to bad (of those who were “good” or “bad” the previous year), and about 30 to 50% who were good remain good, or who were bad, remain bad.
But is the signal we find real or false? • Math – Likelihood of being labeled “good” • 15% less likely to be good in school with higher attendance rate • 7.3% less likely to be good for each 1 student increase in school average class size • 6.5% more likely to be good for each additional 1% proficient in Math • Math – Likelihood of being repeatedly labeled “good” • 19% less likely to be sequentially good in school with higher attendance rate (gr 4 to 8) • 6% less likely to be sequentially good in school with 1 additional student per class (gr 4 to 8) • 7.9% more likely to be sequentially good in school with 1% higher math proficiency rate. • Math Flip Side – Likelihood of being labeled “bad” • 14% more likely to be bad in school with higher attendance rate. • 7.9% more likely to be sequentially bad for each additional student in average class size (gr 4 to 8)
Figure 1 – Who is irreplaceable in 2006-07 after being irreplaceable in 2005-06? Awesome x 2 Awesomeness Important Tangent: Note how spreading data into percentiles makes pattern messier!
Figure 2 – Among those 2005-06 Irreplaceables, how do they reshuffle between 2006-07 & 2007-08? Awesome x 3
Figure 3 – How many of those teachers who were totally awesome in 2007-08 were still totally awesome in 2008-09? Awesome x 4? [but may have dropped out one prior year]
Figure 4 – How many of those teachers who were totally awesome in 2008-09 were still totally awesome in 2009-10?
Persistently Irreplaceable? Of the thousands of teachers for whom ratings exist for each year in NYC, there are 14 in math and 5 in ELA that stay in the top 20% for each year! Sure hope they don’t leave!
Is it really true that the most effective teachers are in the schools that already have high proficiency rates? Strong FALSE signal (bias)
Is it really true that the most effective teachers are in the schools that already have high proficiency rates? Strong FALSE signal (bias)
Is it really true that the most effective teachers are in the schools that serve the fewest minority students? Strong FALSE signal (bias)
Is it really true that the most effective teachers are in the schools that serve the fewest minority students? Strong FALSE signal (bias)
Okay… so is it really true that the most effective teachers are in the schools that serve the fewest non-proficient special education students? Significant FALSE signal (bias)
And what if the underlying measures are junk? Ceilings, Floors and Growth Possiblities?
Misrepresentations • NJ Commissioner Christopher Cerf explained: • “You are looking at the progress students make and that fully takes into account socio-economic status,” Cerf said. “By focusing on the starting point, it equalizes for things like special education and poverty and so on.”[17] (emphasis added) • Why this statement is untrue: • First, comparisons of individual students don’t actually explain what happens when a group of students is aggregated to their teacher and the teacher is assigned the median student’s growth score to represent his/her effectiveness, where teachers don’t all have an evenly distributed mix of kids who started at similar points (to other teachers). So, in one sense, this statement doesn’t even address the issue. • Second, this statement is simply factually incorrect, even regarding the individual student. The statement is not supported by research on estimating teacher effects which largely finds that sufficiently precise student, classroom and school level factors do relate to variations not only in initial performance level but also in performance gains. [17]http://www.wnyc.org/articles/new-jersey-news/2013/mar/18/everything-you-need-know-about-students-baked-their-test-scores-new-jersy-education-officials-say/
Further research on this point… • Two recent working papers compare SGP and VAM estimates for teacher and school evaluation and both raise concerns about the face validity and statistical properties of SGPs. • Goldhaber and Walch (2012) conclude: “For the purpose of starting conversations about student achievement, SGPs might be a useful tool, but one might wish to use a different methodology for rewarding teacher performance or making high-stakes teacher selection decisions” (p. 30).[6] • Ehlert and colleagues (2012) note: “Although SGPs are currently employed for this purpose by several states, we argue that they (a) cannot be used for causal inference (nor were they designed to be used as such) and (b) are the least successful of the three models [Student Growth Percentiles, One-Step VAM & Two-Step VAM] in leveling the playing field across schools” (p. 23).[7] [6]Goldhaber, D., & Walch, J. (2012). Does the model matter? Exploring the relationship between different student achievement-based teacher assessments. University of Washington at Bothell, Center for Education Data & Research. CEDR Working Paper 2012-6. [7]Ehlert, M., Koedel, C., &Parsons, E., & Podgursky, M. (2012). Selecting growth measures for school and teacher evaluations. National Center for Analysis of Longitudinal Data in Education Research (CALDAR). Working Paper #80.
Misrepresentations • “The Christie administration cites its own research to back up its plans, the most favored being the recent Measures of Effective Teaching (MET) project funded by the Gates Foundation, which tracked 3,000 teachers over three years and found that student achievement measures in general are a critical component in determining a teacher’s effectiveness.”[23] • The Gates Foundation MET project did not study the use of Student Growth Percentile Models. Rather, the Gates Foundation MET project studied the use of value-added models, applying those models under the direction of leading researchers in the field, testing their effects on fall to spring gains, and on alternative forms of assessments. Even with these more thoroughly vetted value-added models, the Gates MET project uncovered, though largely ignored, numerous serious concerns regarding the use of value-added metrics. External reviewers of the Gates MET project reports pointed out that while the MET researchers maintained their support for the method, the actual findings of their report cast serious doubt on its usefulness.[24] • [23]http://www.njspotlight.com/stories/13/03/18/fine-print-overview-of-measures-for-tracking-growth/ • [24] Rothstein, J. (2011). Review of “Learning About Teaching: Initial Findings from the Measures of Effective Teaching Project.” Boulder, CO: National Education Policy Center. Retrieved [date] from http://nepc.colorado.edu/thinktank/review-learning-about-teaching. [accessed 2-may-13]
Misrepresentations • But… even though these ratings look unstable from year to year, they are about as stable as baseball batting averages from year to year, and clearly batting average is “good” statistic for making baseball decisions? • Not so, say the baseball stat geeks: • Not surprisingly, Batting Average comes in at about the same consistency for hitters as ERA for pitchers. One reason why BA is so inconsistent is that it is highly correlated to Batting Average on Balls in Play (BABIP)–.79–and BABIP only has a year-to-year correlation of .35. • Descriptive statistics like OBP and SLG fare much better, both coming in at .62 and .63 respectively. When many argue that OBP is a better statistic than BA it is for a number of reasons, but one is that it’s more reliable in terms of identifying a hitter’s true skill since it correlates more year-to-year. http://www.beyondtheboxscore.com/2011/9/1/2393318/what-hitting-metrics-are-consistent-year-to-year Put simply, VAM estimates ARE about as useful as batting average – NOT VERY!
About the Chetty Study… • While the Chetty, Friedman and Rockoffstudies suggest that variation many, many years ago, absent high stakes assessment, in NYC, across classrooms of kids, are associated with small wage differences of those kids at age 30 (thus arguing that teaching quality – as measured by variation in classroom level student gains), this study has no direct implications for what might work in hiring, retaining and compensating teachers. • The presence of variation across thousands of teachers, even if loosely correlated with other stuff, provides little basis for identifying any one single teacher as persistently good or bad.