School Vouchers Are Not a Cure For Segregation: Part IV, Milwaukee

Here are links to all five parts of the series:

Part I

Part II

Part III : New Orleans

Part IV: Milwaukee

PartV: Washington DC and Conclusion

* * *

Are America's private schools -- particularly the ones enrolling students using school vouchers -- "color blind"? Those who vouch for vouchers, like Kevin Chavous, seem to think so. But as I've shown in this series, America's private schools are highly segregated; that likely means school vouchers are expanding segregation, rather than ameliorating it.

To determine whether vouchers are "color blind," let's take a look at the oldest large-scale voucher program in the county: Milwaukee's Parental Choice Program (MPCP), now 25 years old. Just how big is MPCP?

It was the fall of 1990 when the national precedent-setting Milwaukee Parental Choice Program came to life, with participation of seven nonreligious private schools and 337 low-income students. There were a lot of limitations on who could take part. Total public spending in the first year: $733,800. 

It's a vastly different picture now. 

Many of the limitations are gone; an estimated 26,900 students who live in the city of Milwaukee are using vouchers to attend 117 private schools, the vast majority of them religious. Public spending for the current school year will exceed $190 million.[emphasis mine]

And what has that meant for educational achievement?

The overall success of vouchers has been far from what backers hoped for in 1990. Milwaukee's experiment was touted as a big breakthrough in urban education.

In reality, the outcomes for voucher schools overall are about the same as those for Milwaukee Public Schools, which is to say, they're not so good.

Building up quality doesn't seem to link to building up one sector over another.

Even the University of Arkansas' Department of Education Reform -- as reformy and choicy an outfit as you will ever find -- could only come up with the weakest of evidence in support of the MPCP:

When similar MPCP and MPS students are matched and tracked over four years, the achievement growth of MPCP students compared to MPS students is higher in reading but similar in math. The MPCP achievement advantage in reading is only conclusive in 2010-11, the year a high-stakes testing policy was added to the MPCP (Report #29).

When a snapshot of all MPCP students who took the state accountability test is compared to a snapshot of the performance of MPS students with similar income disadvantages, the MPCP students are performing at higher levels in the upper grades in reading and science but at lower levels in math at all grade levels examined and in reading and science in 4th grade (Report #32).

Ouch. And as I noted before, the National Education Policy Center has pointed out even this weak endorsement doesn't really hold up to the results of U-Ark's analysis. For example, attrition out of the MPCP is substantial:

Regarding the study comparing graduation and post-secondary enrollment and persistence rates of MPS and MPCP students (No. 30), Cobb raised methodological concerns.

By 12^th grade, he notes, roughly three out of four of the original 801 MPCP 9^th graders were no longer enrolled in a participating private school. The sample attrition “severely clouded” the inferences that could be legitimately drawn about MPCP’s real impact on graduation rates.

Additionally, only one of the findings in the study’s most carefully controlled analytic models was statistically significant by conventional measures. Both limitations, Cobb writes, prevent broad conclusions about whether MPCP really improves graduation and higher education continuation rates over MPS. [emphasis mine]

These weak results are supported by other research. In a much older study published by the National Bureau of Economic Research, there was no effect on reading scores, and the effect on math is, at best, a few percentage points.*

So it's very difficult to make a case, based on the data, that MPCP is doing much of anything to improve academic outcomes. Which may be why voucher supporters are turning their arguments away from academic outcomes and toward things like segregation. The question, however, is whether school vouchers are as ineffective at desegregating schools as they are at raising test scores.

Let's do some simple descriptive statistics to explore this question. We'll start with the public schools: what is the racial profile of Milwaukee County's publicly-funded schools, both charter and district but not private?

Again, what I've done here is create 10 "bins" across the horizontal axis; the height of each bin represents a percentage of the total student population. Each bin represents the percentage of a school's students who are black. In this histogram, for example, about 34 percent of Milwaukee County's students, enrolled in a publicly-funded school, attend a school whose student population is between 0 and 10 percent black. 19 percent attend a school that is 10-20 percent black, and so on.

The red line represents the percentage of children in the county under age 18 who are black: 34.7 percent (data from the Census Bureau's American Fact Finder, 2014 American Community Survey, 5-year estimate). Clearly, very few schools represent the racial profile of the entire county. Most Milwaukee County schools have far fewer black students, proportionally, than the entire area.

There are, however, many schools that enroll a student population that is almost entirely black (90 percent or higher, the bin furthest to the right). What we've got, then, is what we have throughout America: a highly segregated school system, reflecting highly segregated housing patterns throughout the region.

Now, one might argue that private schools could buck this trend, because they are not bound to enroll their students according to where those students live. In other words, a private school could, theoretically, draw students from both majority-white and majority-black neighborhoods to create a more integrated student population. So let's look at the NCES Private School Survey data and see whether the private schools are any less segregated.

In the public districts and charters, very few schools enrolled an integrated student population. But the trend toward segregation is even more pronounced in Milwaukee's private schools, even though they do not have to enroll students according to housing patterns.

If you compare the two histograms, you'll notice the bins furthest out -- representing schools with 1-10 percent black and 90-100 percent blacks student populations -- are even higher for the private schools than for the public schools. As bad as the segregation is in the public schools, it's arguably even worse in the private schools.

But hold on -- not all private schools are necessarily MPCP schools. Maybe the voucher schools are reversing the trend compared to the non-voucher schools. How can we check?

I've used this list of MPCP schools to create a database from the NCES data of just the vouchers schools. How do they compare to the public schools, and the total number of private schools?

Be careful reading this: Stata (my statistical program) changes the y-axis scale on each histogram. About 45 percent of the private schools in Milwaukee County have student populations that are less than 10 percent black. In the voucher schools, almost 30 percent are less than 10 percent black.

But in all the private schools, about 22 percent of the students attend schools that are 90 percent or more black; in the voucher schools only, about 28 percent of students are in similarly segregated schools.

In other words: In Milwaukee County, voucher students are more likely to be in a nearly all-black school than either public school students or the entire population of private school students.

There is very little evidence that Milwaukee's voucher program is helping to desegregate the area's schools. Why, then, should public money be used to help perpetuate a system that is anything but "color blind"?

I'll finish this up by looking at Washington, DC next.

Color blind?

* A running pet peeve of mine is how econometric research on the effects of education policy, such as  this study, will casually throw out the word "large" to describe an effect size without putting it into an actual educational context. Since when is a 5 to 6 percent change in a test score "large"? Especially when there is a high likelihood of unobserved variable bias in your model?

I get into this idea more here.