Here’s my quick response to the Brookings report released yesterday on the long term effects of vouchers on a randomized pool of participants in New York City.
Let’s say I conducted a study in which I rented a fleet of helicopters and used those helicopters to, on a daily basis, transport a group of randomly selected students from Camden, NJ to elite private day schools around NJ and Philadelphia. I then compared the college attendance patterns of the kids participating in the helicopter program to 100 other kids from Camden who also signed up for the program but were not selected and stayed in Camden public schools. It turns out that I find that the helicopter kids were more likely to attend college – therefore I conclude logically that “helicopters improve college attendance among poor, minority kids.” The simple policy solution then is to rent more helicopters and use them to send kids, well, wherever. After all, it’s the helicopters that matter????? Clearly, that would be a ridiculous assertion.
Similarly, the “major” findings of the new Brookings study were that, in particular, black participants in the voucher program seemed to have an increased likelihood of attending college. The study involved a randomized pool of individuals who qualified, applied and received the vouchers (and attended a private school) and those who qualified, applied and didn’t receive the vouchers.
The study purports to find [or at least the media spin on it] that “vouchers” as a treatment, worked especially for black students. I won’t spend my time quibbling over design and statistical issues here, because I think the simplest issue to address – the big one – is the definition of the treatment itself.
This is not necessarily a study of whether “vouchers” as a treatment affect long run outcomes. Just like my hypothetical above had little to do with helicopters! Rather, it is a study of using “vouchers” as a funding mechanism to place a relatively small sample of low income minority children into a set of schools with fewer low income and minority peers. Schools that happen to be private schools. So it’s not really a study of whether “private schools” are more (or less) effective either. As such, the study really has little or no policy implications for “voucher systems” themselves, or private schooling.
Personally, I was struck to find that the only reference to peer group or peer composition was in a single sentence at the end of the report – but this sentence really says it all:
To the extent that student learning is dependent on peer quality the impacts reported here could easily change.
Yeah… that’s no throwaway line here! In fact it has the potential to completely re-frame the entire paper.
So what is the treatment?
Well, the use of a “voucher” system to alter the educational setting for a group of kids is most certainly not the treatment. Voucher is merely the mechanism used here to achieve the treatment. It may be a policy mechanism that is useful under limited circumstances to achieve changes in educational setting. But the “voucher” is NOT the treatment.
And the use of vouchers in this narrow context may have few if any policy implications for new voucher programs in Indiana or Louisiana if they do not result in low income minority students being better integrated with higher income peers predisposed to have college aspirations.
The sector of schooling is most certainly not the treatment either – public or private school – catholic or non-religious school. While the sector of schooling is a variable in this analysis, it also may or may not have anything to do with the characteristics of the educational setting that most influenced college going behaviors. There’s a whole separate body of literature on that topic that is notably absent in this report. And it’s quite possible that we could find a policy mechanism which has nothing to do with either vouchers or private, or catholic schools which shifts more low income minority students into educational settings that promote college going behaviors.
So before we get in some huge tizzy about “vouchers” and “private school” effects, lets go back and define the treatment in this study for what it actually is and then try to figure out what it means in terms of effective policies for increasing college attendance among low income and minority students.
Tangent: I found this paragraph particularly interesting:
The voucher offer also has a much larger impact than does exposure to a more effective teacher. Elementary school teachers who are one standard deviation more effective than the average teacher are estimated to lift their students’ probability of going to college by 0.49 percentage points at age 20, relative to a mean of 38 percent, an increment of 1.25 percent (Chetty et al. 2011b). If one extrapolates that finding (as the researchers do not) to three years of effective teaching, the impact is 3.75 percent. The impacts identified here for African American students—an increase of 24 percent—are many times as large.
Basically, what this paragraph/extrapolation boils down to is the distinct possibility that the variations in setting (largely peer group) achieved in this voucher study yield what appear to be stronger effects than the measured (really noisy measured – which may matter here) variations in teacher effect in the Chetty study. In other words, quite possibly… peer composition is actually the strongest in-school effect on long term student outcomes!?? [note that this is speculative based on the somewhat questionable comparison made in the above paragraph].
Second Tangent: Quite honestly, the cost comparison comments in this paragraph are so shoddy, poorly documented, etc. that they do much to undermine the report and should be cut.
These impacts are somewhat larger than the long-term impacts of the much more costly class-size intervention in Tennessee. Dynarski et al. (2011) estimate that being assigned to a smaller class in the early elementary grades increased college enrollment rates among African Americans by 19 percent (5.8 percentage points on a base of 31 percent). Reduction of class size in Tennessee was estimated to cost $12,000 per student (Dynarski et al. 2011), whereas the social cost of the SCSF intervention was about $4,200 per student to the foundation and reduced costs to the taxpayer by reducing the number of students who would require instruction within the public sector. If the government had paid for the voucher, the expenditure could have taken the form of a simple transfer from the public sector to the private sector, which in the long run need not add to the per-pupil cost of education. In fact, it could decrease costs because Catholic schools spend less on average than public schools. Around the time of the SCSF evaluation, New York City public schools spent more than $5,000 per student, as compared to $2,400 at Catholic schools (Howell and Peterson 2006, 92).
First, this paragraph and the sources it cites provide little if any solid evidence regarding “costs” and “expenditures” citing only that the Archdiocese of NY at the time said so (p. ii) regarding Catholic school costs being about $2,400 per pupil (and ballparking out of nowhere the $5k public district figure). This paragraph also compares estimated expenditures on one strategy (class size reduction) in one context (TN) and point in time to partial subsidies on another strategy in another context at another point in time. A while back, I criticized another report, also by Matt Chingos (nothin’ personal, I generally like his work) in which he referred to Class Size Reduction as the “Most Expensive School Reform” without legitimately comparing the costs of CSR to any other reform. The above paragraph is strikingly similar in its gaping holes of logic and evidence. I could go on and on. The authors also make no attempt to provide reasonable assumptions & estimates for the full cost of operating and scaling up a large scale voucher system. As such, this stuff really has not place in this paper. For a more thorough discussion/analysis of public/private school spending, see:
Since the authors didn’t actually conduct any real analysis of schooling resources/finances, they really shouldn’t have gone there in their conclusions. This kind of back of the napkin, half-baked cost savings assertion really cheapens a study that does have some interesting findings to offer.