Measuring School Segregation in SFUSD

The trade-off between language programs and racial integration and why the lack of Latino and Black students at Lowell is not simply due to a surfeit of Asian students.

The previous post looked at the racial diversity of San Francisco’s schools, using the implicit assumption that the ideal was schools that matched the demographics of the city. This sentiment has a long history:

“[Lowell should be] reflective of all the city’s kids.”

Ramon Cortines, SFUSD Superintendent (1986-1992), as quoted in Class Action by Rand Quinn.

Of course, Cortines didn’t mean what he said. What he meant was that Lowell should be reflective of SFUSD’s kids, not the city’s kids. People who today criticize the composition of certain schools will sometimes speak of “San Francisco” or “the city” but will always mean the SFUSD average, not the city average. This post will adopt that mindset. Instead of comparing schools to a broad benchmark such as San Francisco’s Under 18 population, we’ll compare them to just the SFUSD population.

It is well known that SFUSD has proportionately fewer White students than the city as a whole because a substantial fraction of White students attend private schools. This otherwise very interesting article about the motivations for SFUSD’s new elementary school assignment system says that the districtwide student population is 32% Hispanic, 30% Asian, 14% White, and 7% Black. Those numbers are what you get when you do a quick lookup of Dataquest but they are the wrong ones to use here because:

• they include the numbers for charter schools, which should be excluded in a discussion about the level of integration of SFUSD’s schools or the SFUSD elementary school assignment system.

• they don’t adjust for the numbers of students whose race/ethnicity is Not Reported

Any comparison needs to be based on the mix of kids actually attending SFUSD schools. We also need to adjust for the changing composition of SFUSD’s schools. Asian students make up 42% of SFUSD’s high school population but only 31% of the elementary school population. White students are 12% of the high school population but 19% of the elementary school population. For this reason, each school will be compared to the average of its direct peers. Elementary (and K-8) schools will be compared to the combined population of all SFUSD K-5 grades. Middle (and K-8) schools will be compared to the combined population of all SFUSD 6-8 grades. High schools will be compared to the combined population of SFUSD’s fourteen comprehensive¹ high schools.

The following chart shows the demographic composition of each SFUSD school type and how it compares to the the raw numbers from Dataquest and the San Francisco Under 18 mix. The raw Dataquest numbers don’t add to 100% because 5% of students have Not Reported race/ethnicity. The San Francisco Under 18 numbers come from the Census Bureau which doesn’t break Filipino out as a separate group.

SFUSD’s data dashboard reports enrollment using more finely grained classifications (Asian is divided into Chinese, Korean, Japanese, Vietnamese, etc). I will make use of that detail when describing the impact of language programs on elementary school enrollment but didn’t want to use those non-standard categories as the base for the analysis.

As before, we will use divergence index scores to measure how much each school diverges from the relevant benchmark.

Elementary Schools

Here are the divergence index scores of the over 70 elementary schools, when compared to the new benchmark of the SFUSD elementary school population mix.

Switching to the SFUSD elementary school population mix as the benchmark instead of the San Francisco Under 18 population leads to some reshuffling of the order but doesn’t dramatically change the results. An integrated school is going to look integrated no matter what the algorithm for ranking them is. Redding and Sherman, which were 1st and 2nd when compared to the SF Under 18 population only slip a couple of places in the rankings. Lakeshore Elementary (near Lake Merced) emerges as the most integrated school, followed by McCoppin Elementary in Inner Richmond. At the other end, the three majority Black schools remain the least integrated.

The elementary schools have a median divergence score of 0.161, higher than the 0.071 for high schools and the 0.132 for middle schools. Some of this is because most elementary schools have attendance areas and thus reflect the characteristics of the neighborhoods from which they draw many of their students. But SFUSD’s elementary school language programs are a significant contributor to this segregation.

Here is the same divergence score data arranged in a histogram with each block representing a school and the colors indicating the language programs offered at the school.

Language Programs Hamper Integration

Schools with language programs in Spanish or Cantonese are more likely to have high divergence scores (i.e. to be less integrated) than those without language programs or with language programs in minority languages. Of the 25 least integrated schools, 20 offer language programs.

• 5 offer no language program (three are majority Black, two are majority White)

• 2 are newcomer schools, one for Latinos, one for Asians.

• 11 offer Spanish (5 Dual-Language Immersion; 6 Biliteracy)

• 7 offer Cantonese (1 Dual-Language Immersion; 6 Biliteracy)

• 2 offer Filipino².

In contrast, of the 25 most integrated schools (i.e. the ones with the lowest divergence scores):

• 15 offer no language program

• 5 offer programs in minority languages (Japanese, Korean, Vietnamese, Arabic)

• 2 offer Mandarin immersion

• 2 (Spring Valley and Monroe) offer both Spanish and Cantonese programs, thereby ensuring a more balanced student body

• 1 (Garfield, near Coit Tower) offers Cantonese Dual-Language Immersion

Here’s a map that show the divergence of each elementary school and how it relates to the language programs offered at that school. The map can be explored here or by clicking on the map and opening the link.

Language programs increase divergence scores when they build on existing residential concentrations and reduce divergence scores when they bring in students with a different profile than the school’s attendance area.

• Schools with Spanish programs in attendance areas that are majority Latino (e.g. Bryant, Buena Vista/Horace Mann, Cesar Chavez, Cleveland ) or Cantonese programs near Chinatown or the outer Sunset (Lau, Chin, Ulloa) tend to have particularly high divergence scores.

• Four schools (Spring Valley on Nob Hill; Monroe, Hillcrest, and Taylor, neighboring schools in the the Excelsior and Portola districts) offer both Spanish and Cantonese programs, which ensures decent numbers of SFUSD’s two largest groups in each school. These four schools all have divergence scores below 0.140, putting them in the top 30.

• Starr King has a fairly low divergence score of 0.116, because its Mandarin immersion program attracts significant numbers of Asian and Two or more race kids to its location near the housing projects on the south side of Potrero Hill.

Language programs for languages other than Spanish and Cantonese don’t have as dramatic an effect on divergence scores because there are not enough native speakers in San Francisco for them to become the majority in any school. Nevertheless, these programs do produce their own concentrations even if that’s not reflected in the divergence scores.

• 48% (58 out of 122³) of all Korean kids in elementary school are enrolled at Claire Lilienthal which is the only school that offers a Korean Dual-Language Immersion program. Only 0.5% of the district’s elementary school kids are Korean so even though they are so concentrated in one school, they make up only 13% of the students at that school. Lilienthal is the sixth most integrated elementary school despite this concentration of Koreans because Korean kids are classified as Asian for standard education reporting purposes and Lilienthal has relatively few other Asian students.

• 52% (52 out of 100) of all Japanese elementary school students attend either Clarendon (32) or Rosa Parks (20), the two schools that offer Japanese FLES programs. Again, because they are only 0.4% of the district’s students, even though they are so concentrated at these schools they make up just 6% of the students at Clarendon and 5% at Rosa Parks. In fact, despite the Japanese students, these schools are both underweight in Asian students overall.

• 30% (183 out of 617) of all Filipino elementary school students attend one of the two schools to offer Filipino FLES programs (86 at Bessie Carmichael, 97 at Longfellow). They make up 21% of the students at Longfellow and 22% at Bessie Carmichael compared with only 4% citywide and are the main reason those schools’ divergence scores are so high. Nevertheless, since those schools have few Asian students, their divergence scores would be much lower if Filipinos were counted as Asian rather than being a separate group.

• 14% (48 out of 345) of all Middle Eastern/Arab kids in elementary school are enrolled at Redding ES, which is the only school in the district that offers an Arabic Foreign Language in Elementary School (FLES) option. They make up 19% of the students at the school compared with 1.5% citywide. This doesn’t distort the divergence index because Middle Eastern/Arab kids are classified as White for standard education reporting purposes and Redding is underweight in non-Arab White kids.

• 6% (25 out of 418) of all Vietnamese kids in elementary school are enrolled at Tenderloin ES which is the only school that offers a Vietnamese FLES program. They make up 9% of the school which otherwise has negligible numbers of Asian students.

It’s not just the schools with language programs that are affected. We don’t know (because SFUSD has not published this data anywhere I can find) whether in aggregate the enrollment in language programs comes close to the SFUSD average or skews one way or another. Having said that, I would be very surprised if White and Black students were not underrepresented in them. Which means that White and Black students are going to be overrepresented in the general education programs and in the schools that offer just general education programs.

There is thus a conflict between the desire to have racially integrated student bodies and the desire to offer students instruction in the language of their choice.

Middle Schools

The schools serving grades 6-8 (the thirteen pure middle schools and the eight K-8 schools) have a median divergence of 0.132 and a size-weighted mean of 0.126. Aptos is closest to the district average (0.006) and Buena Vista/Horace Mann furthest away (0.410).

The nine most integrated schools are all pure middle schools, meaning that most of the K-8 schools are all in the bottom half of the list. While the K-8 schools are less integrated, they do not share a common demographic profile. Bessie Carmichael has a large Filipino group. Alice Fong Yu and Lawton are majority Asian; SF Community is only 5% Asian. Buena Vista / Horace Mann and Paul Revere are majority Latino; Rooftop is 40% White and Claire Lilienthal has the largest percentage of two or more race students.

Here is the composition of their student bodies compared with the middle school average.

Asian students make up 38% of all middle schoolers but the five most integrated schools and 9 of the top 11 are all more than 40% Asian. In contrast, the six least integrated schools all have student bodies that are less than 10% Asian. Language programs are part of the reason. Three of the four most segregated are located in or near the Mission and have Spanish language programs (but no Chinese language programs). This causes them to be between 70% and 90% Latino.

The scores of Visitacion Valley and Bessie Carmichael illustrate some of the nuances of divergence scores. Visitacion Valley has practically no White kids (0.3%) which is pretty remarkable: in no other school is one of the six largest ethnic groups just missing. But its divergence score is not higher than 0.167 because no one group dominates the school: it is overweight in Latino, Pacific Islander, Filipino, and African American students. Meanwhile, Bessie Carmichael offers a Filipino language program, as a consequence of which it is 29% Filipino when the middle school average is 4%. This causes it to have a high divergence score of 0.243, worse than 13 other middle schools. But Bessie Carmichael has comparatively few (9%) Asian students. If Filipinos were included in the Asian group instead of being a separate group, the school would move from 14th to 2nd most integrated. Denman (13% Filipino) would move from 7th to 3rd for the same reason.

High Schools

Here is how the composition of the fourteen high schools compares with the high school average.

Compared to the SFUSD high school average, the fourteen high schools have a median divergence score of 0.071. Lincoln has the lowest divergence score at only 0.012, just ahead of Galileo, Wallenberg, and Balboa. Lowell ranks 6th but its divergence score of 0.062 is still below the median (and the size-weighted average which is also 0.071). SOTA is the school most affected by the change of benchmark. It has the lowest divergence score when measured against the Under 18 population of the city but the 9th (out of 14) lowest score when measured against the SFUSD high school average. The five schools less integrated than it are all majority Latino.

A lot of attention is made to the composition of Lowell in particular. A commonly held perception is that it is dominated by Asians.

“Lowell High School is a public school. It is not just for one ethnic group to be their private school on taxpayer dollars”.

Virginia Marshall, President emeritus of The SF Alliance of Black School Educators, and member of the 2021-22 Equity Action Audit Committee of Lowell.

It is indeed majority Asian but SFUSD high schools are 42% Asian so it’s not surprising that some schools are majority Asian, particularly ones in areas with large Asian populations. Compared with the other westside high schools, Lowell has a lower percentage of Asian students (53%) than Washington (59%) and only a slightly higher percentage than Lincoln (51%)⁴. Even Galileo in North Beach (56%) has a higher percentage of Asian students than Lowell.

What makes Lowell stand out is not that it has the third highest percentage of Asian students but that it also has the second highest percentage of White students, the second highest percentage of two or more race students, and the third highest percentage (after Burton and Balboa) of Filipino students. No other school is overweight in all these groups. It is the combined extra numbers from all of these groups, rather than simply the number of Asians, that crowd out the Latino and Black students.

Underrepresented Minorities

There is a way to make Lowell have a higher divergence score. Instead of treating each racial group as separate, we could just divide the population in two, creating an Underrepresented Minorities (URM) group that includes Latino, Black, Pacific Islander and Native American students (all the groups in which Lowell is underweight) and an Others group comprised of Asian, White, Two or more race, and Filipino students. URM students comprise 37% of all SFUSD high schoolers and Lowell has the lowest percentage of URM students, 15%. Evaluated in this way, Lowell ranks 9th out of 14. The schools that rank below are the five with majority Latino populations.

Incidentally, if we do use the percentage of Underrepresented Minorities as our benchmark, SOTA ranks 3rd (just behind Balboa and Lincoln). SOTA’s divergence from this benchmark is low because it is underweight in Asian students to compensate for being overweight in White students and and it is overweight in Black students to compensate for being underweight in Latino students. There’s no measure of racial diversity which ranks both Lowell and SOTA in the bottom half of all high schools.

1. “Comprehensive” high schools are those that an incoming 9th grader can sign up for. There are in addition three small alternative or continuation high schools that kids at risk of not graduating might transfer to.

2. The numbers don’t add up because Longfellow offers both Filipino and Spanish.

3. These numbers come the SFUSD Data Dashboard for 2021-22.

4. These numbers are higher than you might see elsewhere because I exclude students whose race/ethnicity is Not Reported from the percentage calculations. Here’s a simple analogy to illustrate why. Imagine you know that there are 100 people in a room and that you know 40 are aged under 18 and 40 are aged over 18. You don’t know anything about the ages of the remaining 20. Saying that 40% are under 18 and 40% are over 18 is wrong because those other 20 must be either under 18 or over 18. Saying 50% are under 18 and 50% are over 18 may not be exactly right but it is the best estimate you can make given the information available.