Phi Beta Cons

Has NYC Discovered the Trick for Closing the Racial Achievement Gap?

That question has important ramifications for college admissions and affirmative-action policies. The schools claim the answer is yes, but as Eduwonkette has argued, and as yours truly will further argue in the ridiculously long post after the jump, that doesn’t appear to be the case.

The schools measure the gap in terms of proficiency rates, but one can adjust these rates by changing the test’s difficulty, or by concentrating on students just below the “proficiency” cutoff — both of these steps can close the gap statistically without closing the gap in terms of actual achievement. Eduwonkette has demonstrated that no such gap-closing occurred on the NAEP test.
The schools refused to give her the state tests’ “scale score” data — that is, data adjusted to be comparable from year to year — because she’s anonymous. A few other bloggers and I requested it and got it, and she has some fresh thoughts here.
(The state and NAEP tests are two different things. The state test, given to comply with No Child Left Behind, is the one used to assess “proficiency,” though in addition to the cutoff, normal scores are given.)
There are three problems with the data. One, until 2006, the state only administered tests in grades 4 and 8; the city administered (slightly different) tests to the missed grades down to 3. Two, even in grades 4 and 8, the state changed the way it calculated scale scores at this time. Three, in 2007, the city started making “English-language learners” take the language test after one year in the U.S. (the previous rule was three), so in the last two years, the white-Hispanic gap might be artificially large relative to the previous years.
Eduwonkette limits her discussion to the years 2003 and 2008, and grades 4 and 8, giving her four data points for each test (language and math). She finds no closing of gaps, except in eighth-grade language, and claims that this occurred because whites and Asians lost ground, not because blacks and Hispanics caught up. The scale scores were calculated differently in the two years, though, so I don’t think one can say with confidence whether a given group’s scores went up or down in an absolute sense.
Because we have standard-deviation numbers, however, we can say how groups changed relative to each other. I think her other conclusion are solid, though with only four data points it’s always possible that one is an outlier (say, a given racial group scored abnormally high in one year, but it wasn’t part of a trend).
I ran some numbers myself. I tried to use more data points — since the city and state tests measure the same things (language and math), albeit differently, we can use the standard-deviation numbers to get a decent grip on the whole system’s work on the gap. It’s not perfect, but it’s a way to find out whether Eduwonkette’s more selective data points were abnormal, and it’s certainly better than using the proficiency rate.
I calculated the white-Hispanic and white-black gaps in terms of standard deviations from the white mean, for all grades and both tests. I averaged all the grades’ gaps in each year. Here’s a graph of my results (ELA is the language test):

It does look like the black-white language gap has been closing a little bit since 2005, but there’s nothing all that dramatic going on anywhere else. To condense the data even further, I averaged together the two gaps and both tests. Here’s the resulting graph:

This looks quite a bit more encouraging, but look how tight the left-hand scale is. After 2002 (a year when blacks and Hispanics both scored particularly low on math relative to whites), the average black or Hispanic student never scored more than .74 standard deviations below the average white, and this year, barely got within .7 SDs. At a rate of a little more than .01 SDs per year — assuming this trend between 2006 and 2008 isn’t due to random fluctuations, assuming it will continue, assuming it is due to NYC’s policies, and assuming that other districts can implement similar policies to similar effect — there’s no way affirmative action will be unnecessary by the Supreme Court’s 25-year Grutter deadline. At least not if we define “unnecessary,” as liberals do, as “unnecessary because all groups of people perform equally on every measure of every ability.”
(UPDATE: Also, it looks like the drop is mainly due to the aforementioned drop in the black-white language gap. It fell about .1 SD over these three years, which, when averaged in with three statistics that didn’t really move, will pretty much account for the .03 drop in the overall gap. Then again, until the change in ELL scores in 2007, Hispanics had scored slightly better than blacks in language, but thereafter scored significantly worse. This means that the gap could have closed in 2007 and 2008 a little more than it looks like.)
I would like to see if there are specific improvements in individual grades — I’m too lazy to do too much more of this, so you can get the city-provided spreadsheet here, and my rearranged and pared-down spreadsheet here. Let me know of any interesting results, or e-mail if you want me to send you an Excel document, at phibetacons[at]nationalreview[dot]com. Also let me know if you find any errors in my work; I checked back over the spreadsheets and formulas, but accidents happen (note: already fixed a weird glitch with the charts).

Most Popular

Elections

Biden Is Still Underperforming Hillary

On October 20, 2016: Hillary Clinton was up 6.8 in the RealClearPolitics average in Pennsylvania. Today, Biden is up 3.8. (that race only tightened to 2.1 in the last week of 2016.) Clinton was up twelve points in Michigan on that day. Biden is up 7.5 right now. Clinton was up 6.5 in Wisconsin. Biden is ... Read More
Elections

Biden Is Still Underperforming Hillary

On October 20, 2016: Hillary Clinton was up 6.8 in the RealClearPolitics average in Pennsylvania. Today, Biden is up 3.8. (that race only tightened to 2.1 in the last week of 2016.) Clinton was up twelve points in Michigan on that day. Biden is up 7.5 right now. Clinton was up 6.5 in Wisconsin. Biden is ... Read More

Trump: No

Editor’s Note: The following is one of three essays, each from a different perspective, in the latest edition of National Review on the question of whether to vote for President Trump. The views below reflect those of the individual author, not of the NR editorial board as a whole. The other two essays can be ... Read More

Trump: No

Editor’s Note: The following is one of three essays, each from a different perspective, in the latest edition of National Review on the question of whether to vote for President Trump. The views below reflect those of the individual author, not of the NR editorial board as a whole. The other two essays can be ... Read More

Trump: Yes

Editor’s Note: The following is one of three essays, each from a different perspective, in the latest edition of National Review on the question of whether to vote for President Trump. The views below reflect those of the individual author, not of the NR editorial board as a whole. The other two essays can be ... Read More

Trump: Yes

Editor’s Note: The following is one of three essays, each from a different perspective, in the latest edition of National Review on the question of whether to vote for President Trump. The views below reflect those of the individual author, not of the NR editorial board as a whole. The other two essays can be ... Read More
Elections

How Trump Might Be Winning

I’m far too dumb to be able to shed any light on polls, but I do know something about celebrity and I think I can guarantee this: If President Trump wins re-election, Robert Cahaly is going to become very famous very quickly. Who is Robert Cahaly? The chief pollster for the Trafalgar Group, the only major ... Read More
Elections

How Trump Might Be Winning

I’m far too dumb to be able to shed any light on polls, but I do know something about celebrity and I think I can guarantee this: If President Trump wins re-election, Robert Cahaly is going to become very famous very quickly. Who is Robert Cahaly? The chief pollster for the Trafalgar Group, the only major ... Read More