A Closer Look at Student Performance in Upstate City Schools

Parents in central cities seeking good educations for their children face the disconcerting reality that relatively few city school children pass standardized tests required by New York State, suggesting that the schools are failing.  In Rochester, Buffalo, Syracuse and Schenectady, less than 20% of students passed state required English Language Arts and Mathematics exams in 2016 and 2018.

A considerable body of research shows that student performance is strongly related to socioeconomic status. Parents’ incomes and educational status predict almost three-quarters of the variation in performance between schools and school districts where parents are relatively affluent and well educated and those who are not.  In Buffalo, Syracuse, Utica and Schenectady, more than 80% of students were disadvantaged. In Rochester, more than 90% were.

Socioeconomic status and student performance on tests are strongly related to economic status once students become young adults. Only 23% of students from low-income families whose math scores were below average had above average socioeconomic status as young adults, according to a study by Anthony P. Carnevale, Megan L Fasules, Michael C. Quinn and Kathryn Peltier Campbell (“Born to Win, Schooled to Lose”). Eighty percent of students from high income families whose performance was above average on math tests in 10th grade had above average SES as adults.

But socioeconomic status (SES) does not tell the whole story. There are variations in performance that are not accounted for by economic and educational status. Adjusted for SES, students in Albany, Rochester and Syracuse performed below expectations.  In Rochester, only 9% of students in grades three through eight passed the state required exams in 2018. Based on the percentage of economically disadvantaged students and district educational levels, 17% were expected to pass.  In Syracuse, only 13% passed.  18% were expected to pass.  In Albany 18% passed, compared to 31% predicted by the two factors.

Although we might infer that unexplained performance differences are the result of differences in school quality, we cannot be certain that it is the cause. Because the available data has a limited number of variables, we cannot know why students in Albany, Rochester and Syracuse performed more poorly than expected. Poor performance could be the result of differences in classroom instruction, but it might also be the result of other factors, such as student differences that existed in pre-school years. We know that as early as third grade – the lowest grade level at which the tests are given – students in the poorly performing school districts did less well than would be predicted by their educational and economic statuses. This suggests that differences in early childhood might be important.

There are better measures of the effect of schools on student learning than comparing achievement on tests at a single point. One approach is to compare the educational growth of students in different schools and school districts. This approach reduces the effects of non-school related differences.

When student educational growth is compared different patterns appear. Differences between urban school districts are smaller using the educational growth measure than the SES/Education model. But differences within school districts are large in some cases. And charter schools are more successful in some cities than in others in providing better settings for student educational growth.

Charter Schools

Given the bleak academic performance of central city school students, it is not surprising that charter schools were established. The fundamental promise of these schools was to offer learning environments that would be more conducive to student success than existing district operated schools.  Supporters of the movement believed that charter schools, freed of bureaucratic rules and limits on school districts ability to incentivize teachers to help students perform better, could produce better results for city children.

Charter schools have been controversial.  Teachers unions have argued that they divert needed resources from existing public schools, and that they produce little real benefit for schools.  Assessed in aggregate across the nation, evaluations of them have been mixed, some showing benefits, others showing that students at charter schools do less well than at comparable schools operated by school districts.

Defenders point to consistent differences in charter school performance by state, reflecting differences in the way charter schools are authorized and evaluated.  For example, a recent study, “Charter School Performance in New York” issued in 2017 by the Center for Research on Educational Outcomes (CREDO) at Stanford University found that overall, students at charter schools performed significantly better than those at comparable school district operated schools.  CREDO compares the educational growth of students at charter and district operated schools to evaluate differences in outcomes.  The chart (reproduced above from the CREDO report) shows that overall, students at charter schools in New York State performed significantly better in mathematics than those in schools run by school districts, by about 1/10 standard deviation (one tenth of a standard deviation is equivalent to a move from the 50th percentile to the 54th percentile.)  In English, the difference amounts to a change from the 50th percentile to the 51st percentile.

The difference in performance in New York state was largely driven by charter school performance in New York City.  Upstate differences between charter school and district school performance were not statistically significant.  However, CREDO data shows that Uncommon Schools in Rochester was associated with a positive difference of 0.11 standard deviation in English compared with district operated schools – a difference of 4.37 percentiles.  In math, the difference was larger – 0.26 standard deviation, or 10.26 percentiles.

Not all charter schools outperformed their traditional public school counterparts. There were large variations in charter school performance.  Nearly half significantly outperformed district schools, but slightly more than half performed at levels that were not significantly different from schools operated by school districts or performed worse.. (p. 11).

New York State has recently begun publishing data on student growth at schools administering the state’s third through eighth grade exams. The state defines student growth as follows: “[A] Student Growth Percentile (SGP)…measure[s] a student’s improvement or growth relative to other students, considering the students’ prior academic histories…The SGP indicates whether a student grew more than or less than students with similar test histories in the state. New York defines Elementary/Middle-Level (EM) Growth as “three years of student-level growth in ELA and mathematics combined.”

Analyzing student progress to assess the quality of schools and school districts offers significant advantages over simple comparisons of student test scores. Student progress comparisons exclude the effect of differences in student performance that result from differences that are not related to school quality. Even so, there are weaknesses in the student growth measure unless some additional control measures are used. In New York’s case, three additional controls are included in constructing the measure: percentage of English Language Learners in the classroom, percentage of students with disabilities, and percentage of students in poverty.

Educational Growth in School Districts

Although Albany, Rochester and Syracuse performed relatively poorly compared with what the SES/education model predicted, the State’s student educational growth data did not show the same performance deficit. Large upstate cities except Binghamton showed student educational growth that was within two percentiles of the average for all school districts in the counties within which they were located. Utica did best, with the average student in the district being in the 52nd percentile in the state, 2.95 percentiles higher than the county average (49%).

In the upstate metropolitan counties studied, the districts that performed the best were a mix of large and small, suburban and rural places. In the best performing district – Whitney Point — the average student’s performance growth placed her or him in the 55th percentile.

The weakest performing districts were also a mix of district types. In the worst performing district, the average student’s performance growth placed him or her in the 41st percentile.

Although the gap between the best performing district and the worst is large – 15 percentiles – overall differences between districts in the upstate metropolitan counties studied were small – 64% of the districts were within plus or minus 2% of the average for all districts.

As a practical matter, although the data does not show that city school districts like Albany, Rochester and Syracuse are doing a particularly poor job of educating students, most in these districts perform poorly, and high percentages drop out before completing high school. And, as the following section shows, there are large variations in student educational growth among schools within the same school districts.

Educational Growth in Schools

The following section is based on student growth data published by New York state for 2017-2018. Charter schools from three counties – Albany (Albany), Monroe (Rochester) and Erie (Buffalo)– were included in this analysis because data was available for five or more charter schools in each of these counties.

This chart shows the average EM growth percentile at each school in the three cities compared with other schools. Three quarters of the schools had EM growth percentiles that were within a range of only six points – between the 46th and 52nd percentiles.

Charter school performance varied in the 2017-2018 New York State data. The best (Buffalo Academy of Science – 69th percentile) and worst performing school (Charter School of Inquiry in Buffalo – 39th percentile) were 30 percentiles apart in student growth. Similarly, there were large variations in district operated schools in each of the cities. For example, at the best performing district operated school (the Montessori School in Albany), average EM student growth was in the 60th percentile, while at the worst school (PS 82 in Buffalo), the average student was in the 38th growth percentile compared to students state-wide.

School sizes are relatively small. Random variations in performance could occur because of the small number of test subjects in each. To prevent misinterpretations of differences in school performance because of sample variability, I use the method applied in the CREDO study. For schools to be labeled better or worse performing than average, a statistical test was employed. For schools to be considered better or worse than average, the statistical test had to show with 95% confidence that the school’s student growth percentile was different from the average (49th percentile) of school districts in the areas examined.

NYSED data from 2017-2018 shows that in the three cities and counties studied, charter schools did perform better overall than district operated schools. 37% of charter schools had average EM growth percentiles that were significantly above average, compared with 23% of district operated schools. Even so, 63% of charter schools’ average EM growth percentiles were average or below average. For district operated schools, 77% were average or significantly below average. But there were large variations in the performance of district operated and charter schools in each of the three cities studied.

 Buffalo

Buffalo’s district operated schools had the highest percentage of significantly above average district operated schools and the lowest percentage of above average charter schools in the three cities.

For students in Buffalo, in many cases choosing to attend a charter school offers no real benefit. Thirteen district operated schools had EM growth percentiles that were significantly above average, compared with three charter schools. For students at average or above average district operated schools, there are few charter school alternatives where students on average show significantly higher educational growth. Eight district schools had growth percentiles that were significantly below average. Students at these schools might benefit from seeking to enroll in a charter school that had above average or average student growth scores.

Albany

The overall performance of district operated schools in Albany was significantly weaker than charter schools in the city and was the weakest of the three cities studied. Two of five charter schools had EM growth percentiles that were significantly above average, while three were average. Almost half (47%) of district operated schools had average growth percentiles that were significantly below average, while only 13% were significantly above average. There were exceptions, however. At the city’s Montessori Magnet School, the average EM growth percentile was 60%, while at William S. Hackett middle school, the EM growth profile was 55% — also above average.

Many Albany students in district operated schools might benefit by seeking to enroll in charter schools, particularly those attending the seven schools with below average growth percentages. For students at district operated schools with average EM growth percentiles, relatively little might be gained. For example, students at the Delaware Community school would be moving from a school with an average EM growth percentile of 51.8 to a charter school that would at best have an average growth percentile that is 3% higher.

Rochester

Rochester’s district operated schools overall performed slightly better than Albany’s, with 29% having EM growth percentages that were significantly below average, compared with 47% in Albany. But, as in Albany, only 13% of district schools were significantly above average. As a group, charter schools in the Rochester area performed the best among those in the three cities, with 55% having significantly above average EM growth percentiles.

Six of nine schools with significantly above average EM growth percentiles were charter schools, compared to only three district operated schools. Among schools that had EM growth percentiles that were significantly below average, seven of nine were district operated schools.

Most schools operated by the Rochester school district had EM growth percentiles that were statistically average. For many of these students, seeking to transfer to a school ranking higher might not offer a significant advantage, given the amount of variation in school performance that could be associated with statistical sample “noise.” For students in below average schools operated by the Rochester School district, moving to a better performing school could provide greater educational opportunity.

Conclusions

Although differences in student growth between city school districts were not large, within school districts there were relatively large differences between the best and worst performing schools. In two of three cities – Albany and Rochester – charter schools as a group outperformed district operated schools. But in those cities some charter schools performed poorly, and some district operated schools performed well.

For students, avoiding schools whose performance is significantly below average could be beneficial to student growth. But, geographic accessibility and realities of competition for limited seats in better performing schools can make it difficult to get into them.

For policy makers there are several challenges. Attempts to turn around poorly performing schools do not have a promising track record. Brian Backstrom of the Rockefeller Institute at the State University of New York writes, Over the past half-century, billions of dollars have been spent across the nation on efforts to transform persistently low-performing public schools — most of them urban, most of them low-income, and most of them disproportionately enrolled with students of color — into models of success. It hasn’t worked.”

School turnaround efforts are often frustrated by forces that contribute to organizational inertia, ranging from unions that fear that members will lose their jobs to uneven implementation because of differences in senior and middle level managers’ commitment and ability and insufficient long term financial commitments.

Backstrom points out that the most effective approach to improving schools that perform poorly has been to close them. But doing so faces practical impediments – most notably that better alternatives must be available to displaced students if efforts are to succeed. This can be difficult, because existing, better performing schools usually have enrollment constraints that limit the number of additional students that they can accommodate.

Where city school districts have seen improved results, charter schools often play an important role. The challenge here is two-fold. One issue is that scalability is a problem. Evidence suggests that some charter school operators with proven track records – KIPP and Uncommon Schools are two examples – are more likely to provide statistically superior results than independently operated charter schools, but the successful organizations are constrained by their organizational capacity to grow while maintaining quality. Charter schools take time to establish and, in many cases, need substantial private financial resources to support their efforts. Quality control is a concern as well. Student performance at some charter schools is significantly worse than at most district operated schools. These schools do not add meaningful choice to those seeking greater educational growth. Here, the State Education department should be vigilant in monitoring charter school performance.

————————————————

An abridged version of this post appears in the Rochester Beacon.

 




Education, Economic Status and Student Performance in New York School Districts

Note:  A more recent, closer look at this subject may be found here:  https://policybynumbers.com/a-closer-look-at-student-performance-in-upstate-city-schools

In 1966, James S. Coleman and associates wrote the report, “Equality of Educational Opportunity,” for the United States Department of Health Education and Welfare, as required by the Civil Rights Act of 1964.  The report was commissioned to examine the causes of differences in educational outcomes experienced by minority group students, particularly African-American children compared to majority group whites.   Initially, the report’s sponsors had expected that differences in resources available to black, Hispanic and white students would explain the differences.

Coleman, however, found that another factor was far more important.  “The first finding is that the schools are remarkably similar in the way they relate to the achievement of their pupils when the socioeconomic background of the students is taken into account….When these factors are statistically controlled, however, it appears that differences between schools account for only a small fraction of differences in pupil achievement.” (pp. 21-22).  Coleman’s report has led to the notion that factors like parents’ education and incomes determine the performance students in schools.

But Coleman went on to do further research.  In 1982 he published, “Cognitive Outcomes in Public and Private Schools.”  Coleman found “strong evidence that there is, in vocabulary and mathematics, higher achievement for students in Catholic and private schools than in public; the results are less consistent in reading.” In large part, Coleman attributed this to his finding that “Private schools provide. a safer, more disciplined, and more ordered environment than do public schools.”  Coleman’s study is controversial. Critics have argued that his findings could be attributed to other factors, particularly the likelihood that self-selection may be an unmeasured but important factor in these schools’ outperformance.

In New York State, as in other places, students in central cities are much less likely to pass the state’s standardized tests than those in other, more affluent, locations.  One of the consequences of the long-term problem of high failure rates in these schools has been the call for reforms, such as the establishment of charter schools to provide alternatives – presumably better ones – to schools operated by public school districts.  Others have called for state takeovers of underperforming school systems.

What does the data show?  How strongly does socioeconomic status influence performance in school districts in New York State?  Do some districts outperform or underperform expectations?

Upstate Central City Student Performance

Few students in upstate central city school districts do well on New York’s standardized examinations. Fewer than one in five students in these cities passed the state’s English and mathematics examinations given between third and eighth grades in 2016 and 2018. Utica was the only large upstate city school district in which more than 20% of grade 3-8 students passed the state’s English Language Arts and Mathematics exams in 2016 and 2018. In Rochester, only 9% passed, while in Syracuse, 13% passed. In some selected nearby suburban districts, the picture was quite different – in most cases, more than 60% of students passed the state exams.

Economic Disadvantage and Parents’ Education as Predictors of Student Performance

Economic Disadvantage

New York State defines economically disadvantaged students and family as those who take part in assistance programs “such as the free or reduced-price lunch programs, Social Security Insurance (SSI), Food Stamps, Foster Care, Refugee Assistance (cash or medical assistance), Earned Income Tax Credit (EITC), Home Energy Assistance Program (HEAP), Safety Net Assistance (SNA), Bureau of Indian Affairs (BIA), or Family Assistance: Temporary Assistance for Needy Families (TANF).” When two years of data were combined to reduce random variation associated with small sample sizes, 72% of the variation in student performance was associated with the percentage of students defined as economically distressed. The chart below shows the strong relationship between the percentage of school district students passing New York’s Grades 3-8 ELA and Math exams and the percentage of disadvantaged students in school districts.

Education

We cannot directly measure the educational levels of parents in the school districts studied, but Census data permits us to examine the relationship between the percentage of adult district residents who have college degrees and student performance.   Separately, the percentage of college graduates is associated with 48% of the variation in student performance in school districts. The chart below shows the relationship.

Although economic disadvantage proved to be more closely related to student test outcomes at the school district level, the percentage of adults in a district with at least a four-year college degree increased the percentage of variation explained by a small amount – about 2%. Consequently, a linear model that includes both variables was created to understand the combined relationship between economic disadvantage, adult education levels and student performance. Including both variables in the model increased the percentage of variation explained by the model to 74% – nearly three-quarters.

Although the relationship between student performance and economic status and district educational levels is very strong, about one-quarter of the differing performance of school districts is not explained. One factor that could be associated with this unexplained variance is the instructional efficacy of schools. Note, however, that this possible explanation is inferential. Other unexamined factors could be in play as well.

Characteristics of School District Populations

In upstate central cities, high concentrations of disadvantaged students combine with adult populations that have relatively low educational levels. More than 85% of students in Rochester, Syracuse, Schenectady and Utica were economically disadvantaged in recent years. Only about one-quarter of adults were college graduates.

The characteristics of student bodies in large communities surrounding upstate cities are quite varied. In some communities less than 20% of students were economically disadvantaged. In almost all these communities, 50% or more of the students passed the state examination. In a few cases, more than 70% passed. In other communities outside upstate central cities, more than 50% of students were economically disadvantaged. In many cases, less than 40% of students passed the state’s ELA and Mathematics exams. But in only three cases (Lackawanna, Solvay and Watervliet) was the percentage of economically disadvantaged students as high as it was in the central cities. In these communities, as in upstate central cities, more than 70% of students failed the state exams.

As in the case of economic disadvantage, in most cases, cities had lower percentages of people with at least a four-year college degree than did most residents of communities outside them. But, the difference between the educational levels of city residents and residents of other communities was less clear-cut. In fact, about 10% of communities outside central cities had lower percentages of college graduates than did central city residents. And, in general students in school districts in those communities  performed better on state tests than central city residents. Overall, the relationship was not as strong as the relationship with economic disadvantage.

Which School Districts Outperformed and Underperformed?

Calls for reform of central city school districts, such as state school district takeovers and the charter school movement, rest on the notion that the poor performance of many students attending urban schools is at least in part caused by poor school management and instruction. How well do students at schools in central city school districts perform given the percentage of economically disadvantaged students and people without college degrees within them?

Overall, economic disadantage and education are very important predictors of school district performance.  But, there were significant differences in district performance after controlling for socioeconomic factors.  In Utica, student performance exceeded expected performance by 7% – 24% passing vs 17% predicted. Performance in Buffalo was close to model predictions – 19% passed vs 22% predicted. Schenectady’s performance was also close to model predictions – 17% passed, compared to the prediction of 20%. The Albany school district fell short of expectations by the largest amount.  Had Albany students performed as expected, nearly one-third (31%) would have passed.  Only 18% did. In Syracuse, 21% were predicted to pass but only 13% did.  In Rochester 17% were predicted to pass.  Only 9% did.

The next tables show the fifteen districts within the counties within which Albany, Buffalo, Rochester, Syracuse, Schenectady and Utica are located where student performance exceeded expectations and lagged expectations to the greatest degree.

The performance of school districts outside central cities was quite varied. Among the fifteen school districts where student performance was the lowest compared to the model’s prediction, twelve of fifteen were outside central cities. Among those that exceeded expectations, all but one were outside central cities.

Three upstate cities – Albany, Rochester and Syracuse were in the group of poor performers. Among the school districts where students performed the best compared with the model’s prediction, only one – Utica – was a city school district.

Coleman’s finding continues to be true.  Socioeconomic factors explain most of the difference in performance between school districts.  But even after controlling for these factors, there are differences in performance between school districts.  Some are relatively large.

This data is not a causal analysis. We cannot draw conclusions as to the causes of differences in student performance in school districts compared with model predictions. Other unmeasured variables might explain the differences in performance. However, residents of school districts where performance is lower than predicted by the model might question whether their school district is performing its job effectively.

For residents of three large city school districts – Albany, Rochester and Syracuse, this question is particularly important because the percentage of students passing the state exams is lower than in almost all other school districts, and because controlling for known factors associated with poor performance, they did less well than expected.

(a version of this post also appeared in the Rochester Beacon).

 




Rochester’s Broken School System

Kent Gardner argues forcefully in the Rochester Beacon that Rochester’s school system is broken and in need of radical change. Gardner’s post highlights the efforts of a local organization, ROC the Future, to bring about reform of the city’s school system. Gardner is correct – students in the city’s schools do poorly compared to others in the state, and the state designated Distinguished Educator Jaime Aquino’s recent report shows a series of failures of leadership, communication and implementation in the school district.   Here I take a close look at the dimensions of student performance in Rochester compared with other districts in Monroe County and the state outside New York City and look at what my findings suggest for possible reforms.

The Effect of Economic Disadvantage on Performance

Economic disadvantage is the strongest predictor of student performance among the socio-economic variables available for review in the data available from the state test database. The percentage of economically disadvantaged students in school districts is associated with 70% of the differences in performance on the 2018 Grade 8 New York State English Language Arts exam among school districts outside New York City.  

New York State defines economically disadvantaged students and family as those who take part in assistance programs “such as the free or reduced-price lunch programs, Social Security Insurance (SSI), Food Stamps, Foster Care, Refugee Assistance (cash or medical assistance), Earned Income Tax Credit (EITC), Home Energy Assistance Program (HEAP), Safety Net Assistance (SNA), Bureau of Indian Affairs (BIA), or Family Assistance: Temporary Assistance for Needy Families (TANF).

Grade 3 Results 

Third graders in the Rochester school system had the lowest percentage – 17% – of students passing the English Language Arts Exam of school districts in New York State outside New York City.  Among large school districts, Rochester had the highest percent of third grade students were economically disadvantaged – 93%.  Students in Syracuse performed nearly as badly – 20% of Syracuse third grade students passed. Ninety percent of Syracuse third grade students were economically disadvantaged.

Grade 8 Results

Eighty-seven percent of eighth grade Rochester School District students were economically disadvantaged in 2018, among the highest percentage in the state, and only 11.4% of 8th graders received a passing grade on the test – the lowest among the school districts studied. For districts in Monroe County outside the City of Rochester, an average of 37% of students were economically disadvantaged, and 50% passed the eighth grade ELA exam, receiving a grade of 3 or 4.

Other upstate cities also had relatively high percentages of disadvantaged students and small percentages of students passing the state exam, but Rochester’s case was the most extreme.

Five of the nine upstate cities in the chart above had a lower 8th grade passing rate than would be expected based on the relationship between student performance and economic disadvantage.  But, even if the schools in these cities had performed as well as expected based on the model, student performance would still have been poor.  In the case of Rochester, 17% of eighth graders would have passed and not 11%.  For Syracuse, 19% would have passed, not 15%.

The strong association between economic disadvantage and poor student performance found in Rochester and other upstate cities reflects other studies that find that student achievement at the school district level is strongly related to family income. For example, a group of Stanford University researchers in a study reported on by the New York Times in “Money, Race and Success: How Your School District Compares” found that sixth grade students in the richest districts are four grade levels ahead of those in the poorest districts.

How the Performance of Rochester Schools Compares with Others in Monroe County

The performance of schools within school districts varied, but the overall relationship between economic disadvantage and student performance continued to be strong. In this case, test results from grades 3 and 4  were combined, as were results from grades seven and eight, because small numbers of students attend typical elementary schools. By combining grades, random sampling variation is decreased.

Third and Fourth Grade Results

Rochester School District third and fourth grade students were both less likely to pass the ELA exam than students in other Monroe County school districts, and more likely to be economically disadvantaged.  The data for grade three and four shows that overall, more than 70% of students at all Rochester City schools were disadvantaged. At three schools, all of the tested students were disadvantaged. The performance of third and fourth grade Rochester School District students varied, ranging from six percent passing at one elementary school to 37.5% at another.

Of 17 Rochester School district schools studied, only four had passing rates that equaled or exceeded the percentage that would be expected based on a linear model of the relationship between student performance and the percentage of disadvantaged students. Among those four schools, the amount by which performance exceeded expectations was not large. Four of the 17 were in the bottom 20% of schools based on performance when the percentage of disadvantaged students was considered.

As a group charter schools had a lower percentage of disadvantaged students (78%) in third and fourth grade than schools in the Rochester School District (91%).  Overall, performance of third and fourth grade students at charter schools was better than at schools in the Rochester School District when the concentration of poverty was considered.  Of the six charter schools examined, performance was near what would be expected based on the percentage of disadvantaged students at three schools. Performance at two charter schools relative to the percentage of disadvantaged students was in the top 3% of all schools.

Readers should not conclude from this finding that charter school students at some charter schools performed better than at most Rochester City schools solely because of charter school teaching methods. Other studies have shown that self-selection accounts for some of the better performance of some charter schools. Charter schools are likely to be seen as more challenging than public school alternatives and may attract parents and students seeking a more rigorous alternative. See this study for aa review the issue. The kind of analysis employed here cannot account for self-selection.

Seventh and Eighth Grade Results

Smaller percentages of eighth grade students in Rochester schools passed the 8th grade ELA exam than in other Monroe County school districts.  Between 0% and 24% of students passed the state’s ELA exam. At the same time, concentrations of economically disadvantaged students in Rochester were much higher than in other Monroe County school districts.  73% to 94% of eighth grade Rochester City School students were economically disadvantaged. Performance at five of the ten schools in the group was in the bottom 20% of schools based on the relationship between performance and economic disadvantage. Performance at the other five schools was near average.

As with third and fourth grades, a smaller percentage of seventh and eighth grade charter school students was economically disadvantaged (75%) than students in the Rochester School District (88%). Charter school performance was mixed, though overall it was significantly better than at Rochester School District schools.

Performance at one school was slightly below what was predicted.  At two charter schools, performance was average. Performance at two schools was in the top 20% of schools based on the model. Performance at the remaining charter schools was in the top two percent. Cautions again apply about the causes of the relatively good performance at two of the charter schools – self-selection may have influenced the outcomes.

Conclusions

Rochester’s school system is at a crossroads. Student performance in Rochester schools is poor, even accounting for effect of the high percentage of students who are disadvantaged. The recent report by distinguished educator Jaime Aquino documents shortcomings in leadership, communications and policy execution. These factors all call for substantial changes in current practice.

The data about charter schools is somewhat equivocal. While student performance at several charter schools was substantially better than at public schools with similar percentages of disadvantaged students, the performance of others was not above average. Because charter school students apply for admission, selection bias is a possible explanation of better performance.  It should also be noted that though most students at charter schools were economically disadvantaged, the concentration of disadvantaged students in Rochester School District schools was 13% higher than at charter schools as a group.

In the best performing schools in the region, more than 70% of students pass the state’s ELA exams, compared to 15% for third graders and 16% for eighth graders in Rochester City Schools. But, only 20% to 30% of students in the best performing districts were economically disadvantaged compared with about 90% in Rochester.

The pervasiveness of the relationship between economic insecurity and poor student performance points to the need reduce the concentration of poverty in the City of Rochester. As a city, Rochester is highly dependent on other levels of government – primarily the Federal government — for assistance in combatting poverty. The city/county anti-poverty initiative – RMAPI operates with involved entities on strategy development and coordination. It is important that RMAPI and its partners develop clear strategies implemented at a scale that is large enough to have meaningful impacts. Rochester’s Center for Governmental Research developed a series of options for policy initiatives for RMAPI in 2014, some of which the organization followed.

The most recent progress report posted on the RMAPI website describes its 2017 accomplishments. The most recent press release is dated January 2018. Although the organization made some programmatic initiatives in the 2015-2017 time-frame, its website lacks information on the current status of its efforts.

There are potential federal actions that could benefit low income families. Because poor families cannot invest in their children as much as wealthier parents, solutions like proposed refundable family tax credits could help the already existing Earned Income Tax Credit reduce income insecurity. Increasing food stamp benefits, particularly for families with children might be another approach. Because there is evidence that disadvantaged students perform better in schools that have lower concentrations of disadvantaged students, efforts to provide lower income families with access to housing in wealthier neighborhoods has been proposed as a remedy.

In part, the organizational failures in the Rochester City School District may reflect the difficult conditions within which they operate. But, students at city schools do not perform well, even considering the concentration of economically insecure students who attend them. Given the lasting consequences of student failure, community expectations demand better. The Distinguished Educator’s report demonstrates that the district must make a sharp course correction if it hopes to meet the needs of Rochester families.

At the same time, as long as the concentration of poverty in Rochester continues to be very high, efforts to substantially improve student performance at city schools are likely to be only marginally successful.  The performance of students in Rochester City school reflects the nexus of income inequality and organizational failures. Ameliorating the problem will require a multi-pronged approach that addresses both problems.




School Segregation is Increasing in New York’s Cities and Suburbs

Recent articles in the New York Times and The Nation have focused on efforts to resegregate schools in the South, by carving new predominantly white school districts out of larger county-wide school districts that are predominantly black and Hispanic.  The articles examined a recent federal court decision that permitted the creation of the Gardendale School District near Birmingham, Alabama.  The new district is 75% white, in a county school district that has a majority of black and Hispanic students.

In 1954, the United States Supreme Court, in Brown vs. Board of Education, outlawed the creation of segregated school systems by law.  While first efforts to combat segregation focused on legally created barriers to integration in the South, later, courts ordered busing to combat segregation in northern school districts, like Boston.  These efforts were met with fierce resistance from parents who did not want their children to be bused to schools that had large minority student populations outside their neighborhoods .

Resistance to school integration has been has been widespread.  While legally created separate schools in the same school system for white and black students have been eliminated, opposition to efforts to combat segregation based on residential patterns has been widespread and largely successful.  Today, the schools attended by black and Hispanic students typically have far higher concentrations of minority students than those attended by white students.  While segregation in the South was the result of laws that created separate school systems for white and black students, today much of the segregation results from the concentration of black and Hispanic students in cities with majority black and Hispanic populations.

In an earlier post, I examined the growth of segregation of black and Hispanic students in metropolitan areas in New York State.  In this post, I compare the concentration of black and Hispanic students with white students in schools in cities and suburbs in New York metropolitan areas.

Changes in School Enrollment 

In upstate metropolitan areas, and in the suburbs in the New York metropolitan area, enrollments of black and Hispanic students have increased substantially between 1990-91 and 2014-15 – by more than 50,000 upstate and by 140,000 in the New York suburbs.

  • Black student enrollments increased in upstate metropolitan areas grew by 23,000, while Hispanic enrollments grew by 32,000.
  • In Westchester, Orange and Rockland counties in the New York metropolitan area, black student enrollments grew by 15,000 and Hispanic enrollments grew by 45,000.
  • In New York City, black student enrollments decreased by 113,000 while Hispanic enrollments increased by 68,000.

White student enrollments decreased significantly both upstate (by 125,000) and in the New York metropolitan area (by 113,000).  Nationally, enrollments of black and Hispanic students increased by 7.4 million, between 1994 and 2014 (1990 data is not available) while white student enrollments decreased by 4.2 million.

Overall, school enrollments increased in New York City and its suburbs between 1990-91 and 2014-15, while they decreased in upstate metropolitan areas. Nationally, enrollments increased 12.6% between 1995 and 2014.

In percentage terms, school enrollments nationally were 49.2% white and 42.1% black and Hispanic in 2014-15.

  • Upstate metropolitan areas (66.3% white) and New York City suburbs (55.2% white) had higher percentages of white student enrollments than the nation, while New York City had higher percentages of black and Hispanic students.
  • National level data for 1990 showed a student population of 27.2% black and hispanic students, and 69.4% white students.

By 2014-15 the composition of student populations in schools had changed significantly from the 1990’s, nationally, in upstate New York metropolitan areas and in the New York metropolitan area, with large increases in the percentage of black and Hispanic students. New York City was the only exception – black and Hispanic students decreased as a percentage of the total.

Increasing Minority Student Concentrations in City Schools

In cities in upstate metropolitan areas, black and Hispanic student populations grew substantially as a percentage of the total – by nearly 25% on average.  Black and Hispanic student populations as a percentage of the total grew in suburbs as well, but the growth was much smaller – only 6.4% on average.  In the Orange-Rockland-Westchester portion of the New  York City metropolitan area, the growth of black and Hispanic students as a percentage of the total was about equal in cities and suburbs – 16% on average.

Most upstate cities have student populations that are majority black and Hispanic, while most suburban areas in upstate metropolitan areas have student bodies that are less than 10% black and hispanic.  On average, the gap in black and hispanic student percentages between upstate cities and suburbs grew from 44% to 63%.

Schools attended by Typical Black and Hispanic Students Differ from those attended by Typical White Students

This section compares the racial and ethnic composition of schools attended by typical black and Hispanic students with those attended by white students in 2014-15.  It does so by finding the percentage of black/Hispanic students at schools for a median student in each racial/ethnic group.  Computing the median involves sorting all the students in a group (black/Hispanic or white) in a metropolitan area by the percentage of minority students in the schools that they attend, and finding the percentage of black/Hispanic students in the school attended by a student who is at the exact middle of the sort.  Half of the white or Hispanic/black students would be attending schools with an equal or higher percentage of Hispanic/black students, while half would have an equal or lower percentage.

The data shows that in both cities and suburbs upstate, black and Hispanic students typically attend schools with higher concentrations of black and Hispanic students than do white students.

  • For example, in the Buffalo-Niagara Falls MSA, black and Hispanic students living in cities typically attend schools where 78% of the students are black or Hispanic.
  • White students in those cities typically attend schools whose student bodies are 46% black – a difference of 32%.
  • In other upstate Metropolitan areas, the concentration of black and Hispanic students in city schools ranges from no higher in the city of Binghamton to 20% higher in Utica-Rome.

Within suburban school districts in New York’s metropolitan areas, black and Hispanic students typically attend schools that have higher percentages of black and Hispanic students.

  • In the Rochester metropolitan area, black and Hispanic students living outside Rochester typically attend schools with 24% black and Hispanic students, while white students typically attend schools with 9% black and Hispanic students.
  • In other upstate metropolitan areas, the differences ranged from 2% to 11%.

Since most black and Hispanic students in metropolitan areas live in cities, while most white students live outside them, it is useful to compare the percentage of black and Hispanic students in schools typically attended by black and Hispanic students in cities with the percentage of black and Hispanic students in schools attended by typical white students outside cities.  Here, the contrast is stronger.

  • For example, In the Albany-Schenectady-Troy metropolitan area, a typical black or Hispanic student living in a city would attended a school that had 67.5% black and Hispanic students.
  • In contrast, typical white students living outside Albany, Schenectady and Troy attended schools that had 5.1% black and Hispanic students, a difference of 62.4%.

Differences were large in other upstate metropolitan areas, as well.  The difference in the percentage of black and Hispanic students in city schools attended by typical black and Hispanic students and schools outside cities attended by typical white students was 80.6% in the Rochester MSA, and 73% in the Buffalo-Niagara Falls MSA.

Conclusions

Since the 1954 Brown vs. Board of Education Supreme Court decision, it has been illegal to maintain separate schools for minority students and white students in a school district.  But, efforts to create racial balance in schools in cities and metropolitan areas in New York state and elsewhere have largely been unsuccessful.

In fact, the data shows that over the past 25 years, changes in living patterns have seen large increases in black and Hispanic populations in central cities in New York state, but relatively little change in areas outside them.  As a result, because school districts in New York State often follow city and town boundaries, black and Hispanic students are increasingly concentrated in city schools.

  • School districts in cities in upstate metropolitan areas have seen substantial increases in the percentage of students who are black and Hispanic – from 47.6% to 72.4% between 1990-91 and 2014-15.
  • In contrast outside Upstate cities, the average percentage of black and Hispanic students only grew from 2.8% to 9.2%.

The increasing concentration of black and Hispanic students within cities is not the full explanation of their increasing segregation.  Within cities and outside them, black and Hispanic students are likely to attend schools with higher percentages of black and Hispanic students than are whites.  This is largely the result of residential housing segregation within communities in our metropolitan areas.  As a result, the difference between the racial and ethnic composition of schools typically attended by black and Hispanic students and white students has grown larger – in five of seven metropolitan areas typical black and Hispanic students attended schools that had 65% or more black and Hispanic students, while in five of seven metropolitan areas typical white students attended schools whose populations had 5.1% or less black and Hispanic students.

The growth of racial segregation in New York schools is paralleled by its growth nationwide:  The U. S. General Accounting Office found in 2016 that “Over time, there has been a large increase in schools that are the most isolated by poverty and race. From school years 2000-01 to 2013-14 (most recent data available), both the percentage of K-12 public schools that were high poverty and comprised of mostly Black or Hispanic students (H/PBH) and the students attending these schools grew significantly. In these schools 75 to 100 percent of the students were eligible for free or reduced-price lunch, and 75 to 100 percent of the students were Black or Hispanic.”

Although there are significant potential benefits from schools that are more representative of the diversity of the population as a whole, the barriers to change are substantial.  While New York state has not seen the creation of white enclave school districts carved out of larger majority minority districts, the existing structure of local school districts has a similar effect.

There is no silver bullet that will remedy the growth of segregated schools in New York state, or elsewhere.  Remedies tried in the past, like school busing, have been very unpopular, and have generally failed.  Historically, federal housing policies in the 20th century supported racial segregation.  Similarly, suburban zoning laws and resistance to low and moderate income multi-family housing continue to play a role in preventing minority residents from living in them.  In the current political environment, with an administration in Washington, D. C. that is not supportive of federal intervention to promote integration, segregation in our schools is likely to continue to increase.

_________________________________________________________

Note:  For the Orange-Rockland-Westchester portion of the New York City Metropolitan Area, cities are:  Mount Vernon, New Rochelle, White Plains and Yonkers.




More Regional Diversity but a Larger Racial/Ethnic Divide in New York Schools

This post examines changes in the ethnic and racial compositions of kindergarten through twelfth grade schools in New York State metropolitan areas over the past 25 years.  During that period, the student population, like the general population has become more diverse, with the percentage of students identified as white decreasing, while minority group members, particularly Hispanics, have become a larger proportion of the total population.  But, has the increased diversity of the overall student population been reflected in individual schools?  The data show that although New York’s metropolitan regions are more diverse than they were in the past, the gap in racial and ethnic composition in our schools is larger.

New York Metropolitan Areas and the Nation as a Whole are Racially and Ethnically more Diverse than in 1990

Between 1990 and 2015, the percentage of the national population that identified itself as white decreased from more than 75% to 62%, while Hispanics increased from 9% to 17%.  The number of people who identified as Asian increased from 2.8% to 5.1%. (Source: U.S. Census Bureau, 2011-2015 American Community Survey 5-Year Estimates, 1990 Decennial Census.) Upstate metropolitan areas are much less diverse than the nation as a whole.  Smaller metropolitan areas had very small minority group populations – the Binghamton metropolitan area was 87% white in 2015, while the Utica area was 86% white. Larger upstate metropolitan areas were also less diverse than the nation in 2015.  Buffalo was 78.5% white, while Rochester was 77.4% white.  In contrast, in the New York City metropolitan area, white residents were less than half of the population – 44.4% – while Hispanics were nearly one-quarter of the total.

Student Populations in the United States and in New York Metropolitan Areas are more Diverse than in 1990

The kindergarten through grade twelve population has shown a greater demographic shift than the population as a whole at the national level between the 1990-1991 school year and 2014-2015.  The percentage of students who identified as white decreased by 17.3%, from 69.4% to 52.1%.  The Hispanic population increased by 13.1%, while the percentage of students identified as Asian increased from 3.2% to 5.1%. Source:  National Center for Educational Statistics – Elementary and Secondary Information System.

Compared to the nearly even split between white and minority group students nationally, upstate metropolitan areas are relatively less diverse – particularly smaller metropolitan areas, like Binghamton, Utica and Syracuse.  But even these areas have seen significant increases in the percentage of students who are black, Hispanic and Asian since 1990.

  • Small metropolitan areas, like Binghamton and Utica-Rome, were about 80% white in 2014-2015, and had black and hispanic populations that each comprised less than 10% of the total.  In 1990-1991, these metropolitan areas were more than 90% white.
  • Medium sized metropolitan areas had k-12 student populations that were between 65% (Buffalo) and 75% (Syracuse) white in 2014-2015.  In 1990-91 the percentage of white students was 15% to 20% higher in each of these metropolitan areas.  In these metropolitan areas, the percentage of students who identified as black and Hispanic increased by about 10%.
  • In New York City, in 1990-91, the school population was 18.9% white in 1990-91, 37.6% black and 34.5% Hispanic.  By 2014-15 the white population had decreased to 15.5% of the total, while the black population decreased to 25.1%. The Hispanic population in New York schools increased to 40.9%, while the Asian population grew from 7.8% to 18.1%.
  • Suburban areas around New York City on Long Island, and Westchester, Orange and Rockland Counties had large decreases in the percentage of white students in school populations – more than 20%.  These areas saw large increases in Hispanic students, who increased by 15% to 17% as a percentage of school populations.

In the next section, I look at the question of whether the increased ethnic and racial diversity of student populations in metropolitan areas is associated with decreased racial and ethnic segregation in schools.

The Racial and Ethnic Composition of Schools attended by Whites, Blacks and Hispanics Differed more in 2015 than in 1990

This section compares the racial and ethnic composition of schools attended by typical black and Hispanic students with those attended by white students in 1990 and 2015.  It does so by finding the percentage of black/Hispanic students at a school for a median student in each racial/ethnic group.  Computing the median involves sorting all the students in a group (black/Hispanic or white) in a metropolitan area by the percentage of minority students in the schools that they attend, and finding the percentage of black/Hispanic students in the school attended by a student who is at the exact middle of the sort.  Half of the white or Hispanic/black students would be attending schools with an equal or higher percentage of Hispanic/black students, while half would have an equal or lower percentage.

The difference between the percentage of minority students in schools attended by typical white students compared to typical black/Hispanic students increased between 1990-1991 and 2014-2015, with one exception (New York City).  Despite the increase of Hispanic and black students from less than 10% (18% in Rochester) to between 20% and 30% of the student population in upstate metropolitan areas, typical white students attend schools with Hispanic/black populations that make up about 5% of the total.  In contrast, black and hispanic students typically attended schools in 2015 where black and Hispanic students make up more than 50% of the population.

Also of concern is the fact that the racial/ethnic gap between schools that typical white and Hispanic/black students attend increased between 1990-1991 and 2014-2015.

 

  • In the Albany-Schenectady-Troy metropolitan area, in 1990-1991, typical white students would have attended schools with 2% black and Hispanic students. Typical black and Hispanic students attended schools with 38% black students in that year.
  • In 2014-2015 typical white students attended schools with 5.8% black/Hispanic students, while typical black and Hispanic students attended schools with 55% black/Hispanic students.
  • The gap between the schools attended by typical white and typical black/Hispanic students increased from 36% to 49%.

Other upstate metropolitan areas had increases in the gap between white and black/Hispanic percentages as well:

  • In the Binghamton MSA, in 1990-1991 median white students attended schools that were 1.6% black/Hispanic while median black and Hispanic students attended schools that were 10.4% black/Hispanic.  In 2015, whites typically attended schools that were 4.7% black/Hispanic, while blacks and Hispanics typically attended schools that were 31.4% black/Hispanic.  The gap increased from 8.8% to 26.7%
  • In Buffalo-Niagara Falls, in 1990-91 median white students attended schools that were 1.4% black/Hispanic. Typical black/Hispanic students attended schools that were 54.9% black/Hispanic. In 2014-15 typical white students attended schools that were 5.3% black/Hispanic, while median black and Hispanic students attended schools that were 63% black/Hispanic. The gap increased from 53.3% to 57.7%
  • In the Rochester MSA in 1990-1991, typical white students attended schools that were 4.2% black/Hispanic, while black and Hispanic pupils attended schools that were 67.3% black/Hispanic.  In 2014-2015 the comparable numbers were 9.2% and 82.7%.  The gap increased from 63.1% to 73.5%
  • In Syracuse in 1990-1991 median white students attended schools that were 1.6% black/Hispanic, while typical black and Hispanic students attended schools that were 43.4% black/Hispanic.  In 2014-15 the percentages were 5.5% and 52.7%.  The gap increased from 41.8% to 47.2%
  • In Utica-Rome, in 1990-1991, typical white students attended schools that were 1.1% black/Hispanic, while median black and Hispanic students attended schools that were 23.3% black/Hispanic.  In 2014-15 the numbers were 3.3% for white students and 61.7% for black and Hispanic students.  The gap increased from 22.2% to 58.4%

Suburban counties in the New York City Metropolitan area saw similar changes:

  • In Nassau and Suffolk Counties in 1990-1991, typical white students attended schools at which black and Hispanic students made up 5.8% of the population.  In 1990-91, black and Hispanic students attended schools which were 50% black/Hispanic.
  • In 2014-15, white students typically attended schools that had 14.5% black/Hispanic students.  Typical Black/Hispanic students schools that were 64.8% black/Hispanic
  • The racial/ethnic gap increased by 6.1%.
  • In Westchester, Orange and Rockland Counties, in 1980-81, typical white students attended schools that were 7.1% black/Hispanic, while typical black/Hispanic students attended schools that were 49.9% black/Hispanic.
  • In 2014-2015, in those counties, typical white students attended schools that were 19.4% black/Hispanic, while typical black/Hispanic students attended schools that were 71.6% black/Hispanic.
  • The racial/ethnic gap in Westchester, Orange and Rockland counties increased by 9.4%.

New York City differed from other locations in New York state in that the differences in the racial/ethnic composition of schools attended by white students and black/Hispanic students has stayed constant at about 60% in 1990-91 and 2014-15.

Conclusions

Although our nation is often characterized as a melting pot that has absorbed many ethnicities and races, black/Hispanic students increasingly attend schools that are primarily black/Hispanic, while white students continue to attend schools that are overwhelmingly white.  Over the period between 1990-1991 and 2014-15, the gap between the schools typically attended by whites and those attended by blacks and Hispanics increased – in schools attended by median black and Hispanic students, the percentage of black and Hispanic students increased more than it did in schools attended by median white students.

This is important because black and Hispanic residents and whites have relatively little interaction in their neighborhoods, and in the schools that their children attend.  Cities and the schools within them continue to see large decreases in white populations.  City minority populations are stable or increasing, while suburban communities remain overwhelmingly white.  Housing segregation is reinforced by the fact that average incomes of blacks and Hispanics are substantially lower than white residents, so that relatively high suburban housing costs are a barrier to suburban housing choices.

Economic and residential separation is strongly rooted in New York and the United States. Causes include discriminatory government housing policies and non-governmental practices that denied black people access to home ownership in the growing suburbs of the post-World War II era,  historically lower average levels of minority educational attainment, and lower black and Hispanic incomes at the same levels of educational attainment as whites.

The concentration of minority students in schools with high percentages of minority students has negative consequences for minority students, largely because these schools typically have high concentrations of low-income students.  Economically disadvantaged students who attend schools with relatively few disadvantaged students do better on evaluations of student performance than those who attend schools with high concentrations of disadvantaged students (see this, and this).   The fact that black and hispanic students increasingly attend schools with high concentrations of minority students should be of concern.

Because white students in New York typically attend schools that have few black and Hispanics attending them, they do not gain the benefit of exposure to other cultures. Researchers have found that diverse schools promote reductions in levels of racial and ethnic prejudice, stereotyping and fears of “others” and that all students, including white students attending diverse schools, show “higher achievement in mathematics, science, language and reading.”

My next post will compare changes in minority and white student populations in cities and suburbs in New York state.




New York’s Dysfunctional School Spending Patterns

For many years, government spending in New York State has far exceeded the national average. State and local governments in New York had the second highest per capita spending in the nation in 2013.[1]

screen-shot-2016-12-08-at-8-52-07-am

Local government spending contributes significantly to New York’s high spending levels. Local government spending in New York averages $9,800 per person – the highest spending level in the nation.

screen-shot-2016-12-08-at-8-53-54-amSchools contribute the largest amount of local spending in New York.  They account for 49% of all local spending, and 61% of local taxes.  The contribution of schools to taxes is greater than to spending because a substantial portion of county spending is for social service programs funded by the State government.

New York school spending per student is far higher than the average for the nation – 87% higher.  Spending is particularly high in the suburbs surrounding New York City.  Because of the large regional disparities in school spending levels in New York State, they are likely to exacerbate educational inequality.

School Spending In New York State

screen-shot-2016-12-08-at-8-56-48-amCompared to other states, New York State spending per pupil is very high – 87% higher than the national average, and is significantly higher than the second state — Alaska (67% higher).[2] Even compared to neighboring states in the Northeast and Midwest, New York school spending is an outlier. New York’s spending per student is 16% higher than the states with the next highest expenditures – Connecticut and New Jersey and is much higher than the average for the northeast, excluding New York State – $20,610, vs. $15,639 – a difference of 32%.  The difference between New York and the average per pupil operating spending in the Midwestern states in the group was even larger – $20,610 vs. $11,556 – a difference of 78%.

This post will examine regional variations in school spending in New York state and in neighboring states, and regional variations within New York and neighboring states.

Household Income and School Spending

screen-shot-2016-12-08-at-8-58-18-amOperating spending per student is related to median household income at the state level, with income explaining about 40% of the variation in state average per pupil operating expenditures.  Schools in states with median household incomes below the national median of $51,939[3] spent $9,549 per student on average in 2014, compared with $14,268 for those above the average.  But, New York’s school spending is much higher than would be predicted from its residents’ median income.  New York’s median household income was $58,687, 13 percent higher than the national median.  The State’s average operating spending, $20,610, was 87% higher than the national average.

Regional Variations in Education Spending

screen-shot-2016-12-08-at-9-00-49-amAbout two-thirds of New York State’s 19,500,000 residents lived in the New York metropolitan area in 2014.  In that area, there is a sharp income division between the 8.4 million residents of New York City, whose median household income is $52,737 and the 4.9 million residents of suburbs in New York State around the city, which average $89,047.  The remaining 6.3 million New York residents who live outside the New York City metropolitan area have median household incomes that are like that in New York City and in the nation, about $53,000 on average.

Per pupil spending for current operations in New York State largely reflects regional household income differences.  Per pupil spending is much higher in the New York City suburbs than elsewhere, averaging $23,680 in 2014.  Spending in New York City was $18,579, while the average for upstate metropolitan counties was $16,846.

School spending in New York City suburban counties is much higher than for the rest of the state, and far exceeds per pupil expenditures measured at the state level outside New York State.   Expenditures in New York City, and in areas outside the New York City metropolitan area are lower, but are like those in the states with the highest spending levels in the region – Connecticut and New Jersey, and significantly higher than the Northeast median of $15,639 and the Midwest rust belt median of $11,556.

Cities and Suburbs – Income Disparities
screen-shot-2016-12-08-at-9-04-33-amTo understand spending differences between regions within New York State and areas outside the State, it is useful to look at per pupil spending within the major metropolitan areas in neighboring states, and outside those areas.  This is so because median incomes in metropolitan areas like New York City, Boston, and Philadelphia are significantly higher than in the rest of the states where they are located.  By comparing spending within and outside those areas, we can get a better understanding of how school spending differs in comparable areas within New York State and outside it.

Each of the four Northeastern states examined for regional disparities showed similar patterns of median household incomes.  Central City incomes were near or below state averages, as were incomes in areas outside major metropolitan areas.  Suburbs in major metropolitan areas showed very high incomes compared to state averages and central cities.

In each of the three major metropolitan areas studied median household incomes in central cities were between 32% and 52% less than in the suburban communities around them.  Philadelphia’s median household income is very low – $37,460, which is 51% less than median household income of the surrounding suburbs.  New York City’s median household income was 41% less than the surrounding suburbs in New York State.

Regional Spending Patterns
screen-shot-2016-12-08-at-9-10-37-amThe median household income for upstate metropolitan counties was like that in areas of Pennsylvania outside the Philadelphia metropolitan area, and below those in Massachusetts and New Jersey outside major metropolitan areas.  Per student operating expenditures in upstate metropolitan counties was higher than in neighboring states, averaging $16,846 compared with 12,641 in Pennsylvania and $13,883 in Massachusetts.

Median household income in suburbs around New York city was higher ($89,000) than in New Jersey ($71,000 around Philadelphia, and $75,000 around New York City), Pennsylvania ($76,915) and Massachusetts ($80,000).  New York City suburbs’ school spending per pupil was much higher than similar areas in nearby states – averaging $23,680 compared with $17,000 in New Jersey, $13,000 in Pennsylvania and $14,810 in Massachusetts.

New York City’s median household income was near that of Boston – $52,737 vs. $54,485, but per pupil expenditures in Boston were significantly higher than in New York city – $21,567 in Boston vs. $18,579 in New York City.  Philadelphia’s median household income was much lower than New York’s or Boston’s – $37,460, and per pupil expenditures were also relatively low for the region – $13,013.

The Impact of Disadvantaged Students on City Schools

Academic studies have consistently demonstrated that the cost of educating economically disadvantaged students is substantially higher than those from more affluent backgrounds.  For example, William Duncombe and John Yinger[4] of the Maxwell School at Syracuse University estimated the additional cost of estimating poor students and those with limited English proficiency to be 111% to 215%.

screen-shot-2016-12-08-at-9-11-45-amSince city school districts have much higher percentages of economically disadvantaged students than the suburban schools around them, to successfully educate their student bodies, per pupil spending should be higher in the cities than in the suburbs.  And, in Massachusetts, that is the case – Boston schools spend 48% more per pupil than the suburban schools around them.  But in New York State, the reverse is true – suburban schools spend 22% more per student than New York City schools.

Suburban areas around New York City, Philadelphia and Boston have significantly higher median incomes than those in the remainder of the states within which they are located.  In Massachusetts and Pennsylvania, per capita operating spending differs little from the suburbs of major cities to areas outside those metropolitan areas.  But in New York State, New York City metropolitan suburban school spending per student is far higher than in the remainder of the state.  Overall, there is substantially more regional spending inequality between suburban areas and New York City and the rest of the state than in similar areas in other states.

Implications

Overall, school operating spending per student in New York State is substantially higher than in neighboring Northeast states, and in the rust belt states of the Midwest.  Not surprisingly, the contrasts between New York’s high school spending levels are greater compared to schools in places like Ohio, Michigan, Indiana and Illinois than they are with New York’s immediate neighbors, like Massachusetts, Connecticut, Pennsylvania and New Jersey. When the State is broken down into regions – New York City, the suburbs in the New York Metropolitan area, and the rest of the state, only New York City shows spending levels that aren’t out of line with comparable regional entities.

The high spending levels found in New York State and its regions cannot be explained by differences in median household incomes, which are small compared with locations outside the state.   New York’s per student operating costs are 32% higher than the average of other Northeast states, and 78% higher than Midwest rust belt state.  Yet, New York’s median household income is significantly lower than some of its neighboring states, including New Jersey, where the median household income is 23% higher than in New York and in Massachusetts, where the median household income is 16% higher.

At the same time, New York has larger, regional variations in per pupil school spending that appear to exacerbate educational inequalities.  While the costs of meeting educational needs of students in districts with high concentrations of disadvantaged students are substantially higher than those in places with few of these students, wealthy New York City suburbs spend far more per student than New York City or other parts of the state. These large differences are not found in other nearby states.

[1] State and Local Finance Initiative – Data Query System:  http://slfdqs.taxpolicycenter.org/pages.cfm

[2] Source: U. S. Census Bureau:

http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid=SSF_2014_00A08&prodType=table

[3] In 2014.

[4] William D. Duncombe and John Yinger, “How Much More Does a Disadvantaged Student Cost?”  (2004) Center for Policy Research, Maxwell School of Citizenship and Public Affairs, Syracuse University. Paper 103.




More on Race, Income and Student Achievement

A few months ago, I wrote about the link between economic disadvantage and poor student performance.  I looked at the performance of students on the State’s annual student assessment for grades 3 to 8, and found that the percentage of economically disadvantaged students in schools and school districts accounted for about three quarters of the difference in performance.

performance vs disadvantaged

The data showed that a ten percent increase in students who were economically disadvantaged was associated with a six percent decrease in performance on the statewide assessment.

The New York Times, on April 29, published an article “Money, Race and Success: How Your School District Compares” that reports on a study that was based on national level data from the National Assessment of Educational Progress.  The article points out that:

“We’ve long known of the persistent and troublesome academic gap between white students and their black and Hispanic peers in public schools.

We’ve long understood the primary reason, too: A higher proportion of black and Hispanic children come from poor families. A new analysis of reading and math test score data from across the country confirms just how much socioeconomic conditions matter.

Children in the school districts with the highest concentrations of poverty score an average of more than four grade levels below children in the richest districts.”

The data for the analysis comes from “The Geography of Racial/Ethnic Test Score Gaps” by Sean F. Reardon, Demetra Kalogrides and Kenneth Shores of Stanford.

In “School Segregation and Student Performance Gaps,” Reardon argues:

“Although it is clear that racial segregation is linked to academic achievement gaps, the mechanisms underlying this link have been debated since Coleman published his eponymous 1966 report. In this paper, I examine sixteen distinct measures of segregation to determine which is most strongly associated with academic achievement gaps. I find very clear evidence that one aspect of segregation in particular—the disparity in average school poverty rates between white and black students’ schools—is consistently the single most powerful correlate of achievement gaps, a pattern that holds in both bivariate and multivariate analyses.”

My own research showed that upstate cities have increasingly high poverty rates – in Syracuse and Rochester, for example, more than 50% of children under 18 live in poverty, and that minority group members living in upstate cities have far lower median incomes than white city residents or of white suburban residents.

Median Family Income – 2014
Upstate Cities
  Black Hispanic White
Albany $39,077 $29,268 $84,422
Buffalo $29,155 $21,803 $55,516
Rochester $28,752 $23,717 $56,178
Utica $22,975 $18,149 $51,043
Syracuse $27,902 $23,438 $57,246
Troy $21,563 $20,061 $60,843
Schenectady $27,338 $25,111 $62,818
Median Family Income – 2014
Outside Cities
  Black Hispanic White
Albany $67,400 $78,594 $91,693
Buffalo $39,001 $44,463 $77,996
Rochester $44,716 $44,179 $81,432
Utica $50,785 $34,792 $70,457
Syracuse $48,187 $57,778 $80,714
Troy $47,521 $67,381 $84,992
Schenectady $65,062 $52,505 $88,674

Median family incomes for blacks in most upstate cities were between $20,000 and $30,000 in 2013, while white families in suburbs around most upstate cities had median incomes of between $80,000 and $90,000.

The increasing separation of residents of upstate and suburbs by income is one of the most dramatic changes in the past half century in the way upstate residents live.  In 1969, 14.2% of the residents of the City of Rochester, and 14.1% of the residents of Syracuse lived in poverty, compared with 11.1% for New York State as a whole.  By 2013, 33.9% of Rochester residents lived in poverty, while in Syracuse, 36.5% lived in poverty.  The poverty rate for New  York State was 15.6%.  So, New York State saw in increase in the percentage of residents living in poverty of 4.5%, while residents of Rochester had an increase of 19.7%, and 25.4%.

Percentage of Residents Living in Poverty
1969 1989 1999 2013
Albany 14.2%  N/A 21.5% 25.3%
Buffalo 15.2% 25.6% 26.6% 31.4%
Rochester 12.4% 23.5% 25.9% 33.9%
Syracuse 14.1% 22.7% 27.3% 36.5%
Schenectady N/A N/A 20.8% 24.8%
Troy N/A N/A 19.1% 27.3%
Utica N/A N/A 24.5% 31.7%
New York State 11.1% 13.0% 14.6% 15.6%

(Data for cities with populations of less than 100,000 were not available for years before 1999).

Recent studies, like the national study by Reardon and other Stanford researchers, show strong associations between the concentration of black and hispanic residents in areas with high concentrations of poverty and poor student performance.

Unless we address the factors that lead to the growth of concentrated poverty in upstate central cities, and the continued separation of white, hispanic and black residents in upstate metropolitan areas, how can we effectively combat the persistent economic differences between black, Hispanic and white residents of upstate metropolitan areas, and the cycle of intergenerational poverty?




Failing Schools – Bill Hammond’s follow up discussion

Bill Hammond’s piece may be found here:

Bill Hammond: Why New York’s ‘Failing Schools’ Fail — and How We Can Turn the Tide




Should Teachers be Evaluated by Student Performance on Standardized Tests?

In January 2015, Governor Cuomo proposed changing the state’s teacher evaluation system to increase reliance on measures of student progress on statewide standardized tests, using a so called “Value Added Model.” In his 2015 State of the State address, he said:

“Now 38% of high schools students are college ready. 38%. 98.7% of high school teachers are rated effective. How can that be? How can 38% of the students be ready, but 98% of the teachers effective? 31% of third to eight graders are proficient in English, but 99% of the teachers are rated effective. 35% of third to eighth graders are proficient in math but 98% of the math teachers are rated effective. Who are we kidding, my friends? The problem is clear and the solution is clear. We need real, accurate, fair teacher evaluations.

We asked the State Department of Education for their ideas and they gave us their feedback and we accept their recommendation. To reduce the over-testing of students we will eliminate local exams and base 50% of the evaluation on state exams. Second, the other 50% of the evaluations should be limited to independent classroom observations. Teachers may not be rated effective or highly effective unless they are effective in the test and the observation categories. We will stop local score inflation, which is resulted in virtually all teachers being rated by setting scoring bans in the state law.”

The proposal was very unpopular with teachers and the unions that represent them.  Their opposition led to a boycott of the testing by significant numbers of students in many school districts.  Because of the controversy, the Board of Regents, apparently at the Governor’s behest, changed the rules to delay implementation of the rules for four years.

Some critics of the change in direction have argued that the Governor “caved in” to the unions.  For example, the New York Post, in an editorial on November 29th, titled “Did the teachers unions just break Andrew Cuomo” said:

For years, Cuomo has been hitting his head against the wall on getting a real state teacher-rating system. Every time he seems to make progress, it’s followed by delays, postponements and revisions that ensure nothing meaningful happens.

Now he’s reportedly set to give up — abandoning the effort to use student scores on state tests to help judge teacher performance. If so, teachers will be judged subjectively, probably by their own peers. Count on every teacher to rank as just peachy — and incompetents to keep on “teaching.”

Just as the teachers unions have demanded all along.

All of us want our children to have good teachers.  As a child, while most of my teachers were competent, I was was taught by a few individuals who had no business being teachers.  One of my teachers used test questions found in the text books that we used, and permitted us to use the answer keys in the back of the books to find the correct answers.  A music teacher gave my class “study hall” on a number of occasions, and put his head on his desk, to sleep.  That kind of “teaching” cheats children, by denying them the opportunity to learn.

But, is the use of student progress on standardized tests an accurate way to measure teacher effectiveness?  Unfortunately, the answer is no.

There are two basic statistical problems involved in the use of student performance on standardized tests to measure teacher performance.  The first involves the question of whether the students in a given classroom are representative of the entire student population in a school district.   Unless students are assigned in a random fashion across the district, variations in student abilities could affect their performance in systematic ways that are unrelated to the effectiveness of teachers.  The American Statistical Association points out:

VAM [the test based teacher evaluation method] scores are calculated from classroom-level hererogeneity that is not explained by the background variables in the regression model. Those classroom-level differences may be due in part to other factors that are not included in the model (for example, class size, teaching “high-need” students, or having students who receive extracurricular tutoring). The validity of the VAM scores as a measure of teacher contributions depends on how well the particular regression model adopted adjusts for other factors that might systematically affect, or bias, a teacher’s VAM score.

The form of the model may lead to biased VAM scores for some teachers. For example,gifted” students or those with disabilities may exhibit smaller gains in test scores if the model does not accurately account for their status.

Similarly, the Educational Testing Service, the developers of the College Board exams and others, says:

The fundamental concern is that, if making causal attributions is the goal, then no statistical model, however complex, and no method of analysis, however sophisticated, can fully compensate for the lack of randomization. The problem is that, in the absence of randomization, it is hard to discount alternative explanations for the results that are found. (This explains why many consider randomized experiments the gold standard in scientific work.

 Specifically, teacher effects based on statistical estimates may actually represent the combined contributions of many factors in addition to the real teacher contribution we are after. Thus the estimate could be fundamentally off target.

 Further, it is usually difficult to determine how off target an estimate is. Clearly, substantial discrepancies would seriously undermine the utility of inferences made on the basis of the analysis.

The second statistical problem stems from the small number of students that most teachers work with.  For example, elementary school teachers, with classes of twenty or thirty students, see only a small sample of all the students in a school district.  Because of sample variability, those small samples are unlikely to be accurately represent typical students in a school district.  Consider the idea of forecasting the result of an election from a sample of 25 voters – the likelihood of an accurate result is small.  For that reason, researchers seek large sample sizes to ensure accuracy.  ETS describes the problem this way:

With a relatively small number of students contributing to the estimated effect for a particular teacher, the averaging power of randomization can’t work for all teachers in a given year. Suppose, for example, that there are a small number of truly disruptive students in a cohort. While all teachers may have an equal chance of finding one (or more) of those students in their class each year, only a few actually will — with potentially deleterious impact on the academic growth of the class in that year. The bottom line is that even if teachers and students come together in more or less random ways, estimated teacher effects can be quite variable from year to year.

Teacher performance is important.  As parents, we want our children to have every opportunity to succeed  Incompetent teachers can limit that opportunity, so it is important that the people who teach our children are capable of teaching effectively.  Teacher evaluation is an important way in which administrators can help ensure that teachers are competent.  But the mindless use of unreliable teacher evaluation methods cannot ensure teacher competency.

So, while it is easy to characterize the Board of Regents decision to postpone implementation of the proposal as a political decision that reflects the power of teacher unions, in fact, the decision reflects the reality that the use of student performance on standardized tests as the primary way to evaluate teachers is not a good way to measure their effectiveness.

 




Can Charter Schools break the Poverty-Poor Student Performance Link?

In an earlier post, I argued that school based solutions to the problem of the poor performance of students in central city schools were not likely to succeed because they ignored the impact of the concentration of disadvantaged students on student achievement.  The data showed that 79% of the variation in performance in school performance in upstate New York metropolitan areas was related to the concentration of economically disadvantaged students within them.

Discussions about the benefits of charter schools tend to be heated – inflamed by ideological differences.  But whatever one’s feelings are about the virtues of preserving public education, or of competition in improving educational opportunity, before making judgements, we should examine the available data about their effectiveness.

At the outset, it should be noted that evaluating the true impact of charter schools is difficult.  Ideally, the performance of charter and public schools should be compared by selecting and assigning students at random and following their progress over a period of years.  But, in reality, students in charter schools are not selected at random, and matched samples of public school students are not available for comparison. Published analyses on the subject have pointed out the need to adjust performance comparisons of students at public and charter schools for selection bias, because charter school students are to a large degree self-selected.

Where competent analyses comparing charter and public schools have been done, the findings have been mixed. One review of the available studies concluded:

“Taken in the aggregate, the empirical evidence to date leads one to conclude that we do not have definitive knowledge about the impacts of public charter schools on students and schools. But in reviewing the existing evidence, one is also struck by the fact that the impacts of charter schools appear to be very contextual. Some public charter schools are better than others. Some are very successful in meeting student needs, and others are not very successful…. Consequently, the impacts of public charter schools should not be painted with one broad brush stroke. Each should be judged on its own evidence and performance.”

Other studies  have found significant advantages for charter schools in central cities. Atila Abdulkadiroglu, Joshua Angrist, Susan Dynarski, Thomas J. Kane and Parag Pathak, in “Accountability and Flexibility in Public Schools: Evidence from Boston’s Charters and Pilots” found:

“A consistent pattern has emerged from this research. In urban areas, where students are overwhelmingly low-achieving, poor and nonwhite, charter schools tend to do better than other public schools in improving student achievement. By contrast, outside of urban areas, where students tend to be white and middle class, charters do no better and sometimes do worse than public schools.”

My research is based on a reanalysis of state education data on the performance of students on the 2015 Statewide Student Assessment.  It cannot provide a controlled analysis of the performance of charter school students, compared with those in public schools.  For that reason, the data available to me cannot produce conclusive evidence about the effectiveness of charter schools.

Because publicly available data is cross-sectional, it provides information about the performance of students at a given point in time, but unlike longitudinal studies, it does not directly measure their gains over a year or years.  For that reason, when a  cross-sectional study finds out-performance, or under-performance, there is the danger of making an attribution error, because we don’t know whether the out-performance or under-performance was a characteristic of the student population that was unrelated to the effectiveness of the schools being evaluated.  For example, the students at out-performing schools might have characteristics related to their selection that would predispose them to perform better than other students.

With those limitations in mind, it is worth looking at the New York State Education Department data on student performance from the 2015 Statewide Student Assessment, controlling for the concentration of poverty in schools, to see whether students at charter schools do significantly better than those at public schools with similar concentrations of disadvantaged students.  The chart below shows the performance of students in public and charter schools in all counties in metropolitan areas, except for the City of New York:

Public Charter Outside NYC

Note that data was available for only 33 charter schools outside New York City, so conclusions from this group of schools must be regarded as tentative.  Still, a few things stand out.  First, the performance of charter schools was quite varied – several charter schools were among the worst performers compared to schools with similar concentrations of disadvantaged students, while a number of others, particularly those with high concentrations of disadvantaged students performed better.  Second, for charter schools, unlike public schools, student performance was not related to the concentration of poverty.

As a group, students at charter schools did slightly better than at public schools with the same concentrations of disadvantaged students. However, the fact that 24% (8 of 33) schools exceeded the percent of students predicted to pass by 20% or more, based on the concentration of poor students, is significant.  Only 1.9% of public schools outside New York City had student performance reaching that level.   And, as Abdulkadiroglu, et. al. found, the benefit from charter schools was most significant for students in schools with high concentrations of poor students.

The performance of the better charter schools in urban counties outside New York City was significantly better than average schools with high concentrations of disadvantaged students, but not as good as at schools with few poor students.  Most of the better performing charter schools had about 40% of students passing the Statewide Assessment, compared with as many as 60% in schools with few disadvantaged students.

School Performance in New York City

The concentration of disadvantaged students in New York City schools is associated with 52% of the variation in student performance between the schools.  Compared to public schools in urban counties outside New York City economic disadvantage is a less powerful predictor of student performance in City schools – 52% vs. 79%.  For charter schools, the relationship between the concentration of poverty and student performance was very weak – explaining only 8% of the difference in student performance.  As with other counties, the performance of charter schools was quite heterogeneous. Students at charter schools in New York City as a group did better than those at public schools with similar concentrations of disadvantaged students.  At the same time, a number of Charter schools performed less well than the average of public schools with the same concentration of poor students.

The weaker relationship between the concentration of poverty and student performance in New York City schools appears to be in part a consequence of the city’s policy of creating specialized schools with selective admission criteria.  For example, the Medgar Evers College Preparatory School includes questions about student performance on the Statewide assessment in its application form.  Another example is the TAG Young Scholars School, which describes its admission policy this way:  “Prospective students must be tested by The New York City Department of Education to determine whether they qualify for a seat in one of the City’s Gifted and Talented programs.” Note that while charter schools often use lotteries to select students, they are not permitted to use test performance as a selection criterion.

These selective public schools raise the issue of causal attribution, since unlike schools that do not choose students based on test scores, it is likely that student bodies enter the selective public schools at higher levels of performance than students at other public and charter schools, and that their better performance may primarily be a result of selection criteria, rather than teaching at the schools.

Public Charter NYC

Some charter schools and public schools in New York City did as well as schools with low percentages of disadvantaged students.  Some of the best performing public schools with high concentrations of disadvantaged students use test performance as one criterion for admission.  Since charter schools are not permitted to exclusively serve high performing populations, the performance of the best charter schools is more remarkable.  At 34 of 148 (23%) of charter schools, 20% or more students than were expected to pass based on the concentration of disadvantaged students passed the statewide assessment. Among public schools in New York City, including those that have selective admissions, 8.9% of schools exceeded their predicted performance level by 20% or more.

While this data cannot prove that the excellent performance of some charter schools was the result of the schools themselves, rather than some other factor, it is consistent with studies that have shown charter schools to be advantageous for disadvantaged students in central cities.

Implications

Much of the discussion about the performance of schools, and how to improve outcomes, has focused on the common core and its testing requirements.  The purpose of these requirements was to provide a universal set of assessment tools that would provide comparable data about student progress across systems.

The results of the testing have been disappointing to many, since, as the figures above show, large percentages of students did not achieve passing grades.  For example, Governor Cuomo’s 2015 The State of New York’s Failing Schools report stated, “It is incongruous that 99% of teachers were rated effective, while only 35.8 percent of our students are proficient in math and 31.4 percent in English language arts. How can so many of our teachers be succeeding when so many of our students are struggling?”

Governor Cuomo’s proposal to improve student performance included the creation of a teacher evaluation system that relied more heavily (50%) on the performance of students in standardized tests, a process to make it easier to remove substandard teachers, and a process to place under-performing schools in receivership.  Several of the proposals have problems.  Teacher evaluation systems that rely heavily on the progress of students on standardized tests suffer from statistical defects that result in low reliability of results – a subject for a future blog post.  The process for identifying under-performing schools does not effectively identify schools that are under-performing relative to the concentration of students in poverty within them.

Most significantly, by focusing almost exclusively on accountability for under performing teachers and schools, the proposal does not offer a strategy for overall improvement of New York’s schools.  Accountability focused methods focus on remedying or removing the worst five or ten percent of schools and teachers in the system, but do nothing to help the great majority achieve better results.

If New York’s education system is to make strides in improving student outcomes, it must encourage schools and teachers to adopt known classroom teaching strategies and effective curriculum choices that have the potential to improve overall outcomes.  Since a significant number of charter schools have achieved excellent student outcomes, it would be helpful if the strategies they use could be considered for adoption in schools that do not perform well.  The state should focus on finding ways to encourage the use of effective strategies, by disseminating information and incentivizing their adoption.

Considerable research has been done on the strategies employed by effective charter schools in improving student performance.  For example, “Getting Beneath the Veil of Effective Schools: Evidence from New York City,” by Will Dobbie and Roland G. Fryer of Harvard University found that:  “traditionally collected input measures – class size, per pupil expenditure, the fraction of teachers with no certification, and the fraction of teachers with an advanced degree – are not correlated with school effectiveness.  In stark contrast…an index of five policies…explains approximately 45% of the variation in school effectiveness.”  They are consistent with the approaches used by “no excuses” model charter schools that emphasize selective teacher hiring, extensive teacher feedback, increased instructional time, and a focus on discipline and academic achievement.

For most schools in cities with high concentrations of disadvantaged students in central cities, academic performance remains poor. In some of these schools less than 10% of students received passing grades on the statewide assessment, and the overwhelming majority of schools with concentrations of disadvantaged students of 90% or more had less than 20% of students passing.

But almost one quarter of charter schools and a few public schools have broken the link between poverty and poor school performance.  At these schools, more than 40% of students passed the statewide assessment, despite very high concentrations of poverty within them.

Accountability based approaches aimed at weeding out ineffective teachers, or taking control of schools from boards of education will benefit only a small minority of students statewide.  Instead, we should focus on making use of what works in improving student performance at the best charter schools, encouraging poor performing schools to adopt effective techniques.