Correlating Socioeconomic standing with API scores

Over on LeftInSF, Kim Knox takes a look at the correlation between the prevalence of socioeconomically disadvantaged students at a school and the schools API test scores. Her post is here. The discussion relates directly to the ongoing discussion about the merits of re-introducing race as a factor in school assignments.

Kim published a table showing the school's SD% and API scores and correctly concludes that the correlation between them is weak. Being a spreadsheet kind of geek, I pulled the data into Excel and looked at the correlation myself. The results are shown in this figure. Clearly, the API results are all over the place. The correlation is not only statistically weak, but fails to explain any significant portion of the variation among schools.

I took it a step further and looked at the schools that deviate furthest from the correlation. In theory, if there were a significant correlation, this should reveal the schools that have done the best or worst at overcoming their socio economic status. The "best" and "worst" are shown below. Noe that the variance from the API "predicted" by the regression is shown in the Var column.

The five "worst" schools:

SD%	School	API	Var
48%	Gloria R. Davis MS	561	-192
45%	John O'Connell HS	565	-191
52%	Bessie Carmichael ES	562	-188
53%	Mission HS	562	-187
60%	Everett MS	558	-184
24%	Metropolitan Arts and Technology Charter School	594	-181

And the five "best" schools are:

SD%	School	API	Var
32%	West Portal ES	900	132
45%	Lawton ES	891	135
80%	John Yehall Chin ES	873	149
55%	Ulloa ES	897	150
26%	Alice Fong Yu ES	944	171

One thing to note is that I excluded Newcomer and Lowell from the list. They are both outliers in different direction and they both serve a select community that invalidates any API regression.

The next thing that jumps out is that elementary schools are doing relatively well compared to the high schools and middle schools. The "best" list is all elementary schools while the "worst" is predominately HS and MS. The other compelling result is that schools with a high percentage of SD students that are also majority Asian dominate the "best" list. This confirms various other analysis of SFUSD test score results that have come before. Asian kids do very well, even if they come from poor households.

In spite of this weak correlation, I feel it does not justify using race in the enrollment process. Not only would SFUSD be stepping into a legal swamp, I just don't think it is the best way to serve the neediest AA and Latino students. If the goal is to help these students, the best place to aim is in the schools they attend.