Which Middle Schools Would You Close?

It's hard to reduce capacity by 4,000 seats

This is the second of a three-part series about SFUSD’s impending school closure announcement. I’ll look at some of the data that is publicly available about the middle schools, estimate how much of a reduction in capacity SFUSD needs, and show some alternative ways of reaching that target.

Basic Numbers

Here is a map showing middle school enrollment in SFUSD. You can mouseover to see the details of each school.

14% of SFUSD’s middle school students are in K-8 schools. For those K-8 schools, only the enrollment in grades 6-8 is shown above. Note that the largest K-8 school (Lawton) has fewer middle school students than the smallest pure middle school (Visitacion Valley).

Estimating Spare Capacity

SFUSD has calculated the capacity of each of its buildings by counting the number of classrooms, assuming that each class has 30 students, and that 85% of classes are in use at any one time. This quick-and-dirty approach sometimes results in capacity estimates that are lower than the number of students the school has enrolled in recent years. As I did for high schools, I’ve estimated the actual capacity of each school as the higher of SFUSD’s quick-and-dirty estimate and the highest enrollment in any year since 2000.¹

The chart below shows how the enrollment at each school compares with its capacity. None of the regular middle schools is operating at capacity but, in some cases, that’s because the district has capped their enrollment in order to spread students around the district. Those schools would be full if the district allowed them to be. The K-8 schools have comparatively little spare capacity because they have so few middle school classrooms².

If you were to draw a north-south line through the city that passes through Twin Peaks, most of the spare capacity would be found in the schools to the east of that line.

Estimating Overall Spare Capacity

SFUSD has said that it wants its schools to have 90-95% utilization. With 9,948 middle school students, that means it needs room for, in round numbers, between 10,500 and 11,000 students. It currently has room for 15,134 students. That implies it needs to reduce capacity by a little over 4,000 students.

Factors to Consider

I have complained before that many of the criteria SFUSD will use to make its decision are poorly defined and/or based on non-public information. That makes it impossible to predict the Composite Score of each group. Let’s have a look at some information that is available.

Facilities Costs

The Facilities Master Plan contains an estimate of the Facilities Condition Index and the total cost of performing the required maintenance and upkeep. Some school buildings are in much better shape than others. This is particularly evident when you consider the outlay on a per student basis.

The outlays per student were calculated using the current enrollment. If a school’s enrollment were to increase as a consequence of other schools closing, the cost per student of its upkeep would go down proportionately.

Academic Performance

The academic performance metric to be used by the district is defined as “state assessments of English Language Arts and Math performance and growth (data from California School Dashboard)”. Here’s what the dashboard shows for School A, which is an actual SFUSD middle school.

Students at school A scored an average of 30.2 points below standard in ELA, which was a 13 point improvement on the prior year. This combination put the school in the Yellow zone which is the middle of the five accountability tiers. The students scored an average of 61.3 points below standard in Math, which was 8.5 points worse than the prior year. This combination put the school in the Orange zone (i.e. the second-worst of the five tiers) for Math.

Students at School B, another SFUSD middle school, did much better:

For the Academic Performance metric, all these numbers have to be converted to a single number. The district hasn’t specified how this will be done so I’m going to use the simplest method I can think of: assign each of the accountability tiers a number from 1 to 5 (red=1; blue=5) and calculate the school’s score as the sum of its ELA and Math scores. So, school A’s score would be 3 (for the Yellow in ELA) plus 2 (for the Orange in Math) = 5 and school B’s score would be 5 (Blue in ELA) plus 4 (Green in Math) = 9. This process yields the following results:

This may appear to identify high performing schools but it doesn’t. It identifies schools with lots of high-performing students. That doesn’t mean the school is what made them high-performing³. Let me go in to a bit of detail to explain why.

Given the data we saw above, it is perfectly reasonable to say that School B’s students are better than School A’s students at both ELA and Math. It is not correct to infer that school B is doing a better job than school A. It could that school A’s students left elementary school scoring 50 points below standard. If they’re now only 30 points below standard, school A has raised their average score relative to the standard by 20 points. That’s very good. Meanwhile, it could be that school B’s students averaged 50 points above standard in 5th grade. If they’re now scoring 60 points above standard, school B can take credit for raising them 10 points, which is good, but not as good as school A. You cannot assess the performance of a middle school using SBAC scores unless you adjust for the elementary school SBAC scores of its students.

The CDE actually has a way to do this, which it calls its growth model. Unfortunately, the growth model requires three years of data and the SBAC was not administered to everyone during the pandemic so the first official release of the growth data will not be until December 2024, long after SFUSD has to make its decisions about which schools to close. When the CDE published the growth model a couple of years ago, it released the growth data for the last pre-pandemic year, 2018-19, as a demonstration of concept. I’ve shown the actual data below for San Francisco’s middle schools.

The horizontal axis shows the percentage of students at the school who met or exceeded standards. At Everett (which was School A above), only of students 40% met or exceeded the standards in ELA while 28% did so in Math. Meanwhile, at AP Giannini (which was School B above), about 80% of students met or exceeded the standards in ELA and 70% met or exceeded the standards in Math. Just looking at those results would lead you to conclude that Giannini is much the better school.

When you adjust for their students’ scores from prior years, the story is different. AP Giannini’s growth score was around 110 in Math and 108 in ELA. The state average is set to 100 so this means AP Giannini’s students did 10 points better in Math than was expected and 8 points better than was expected in ELA. That’s a good result. But Everett was better. Its students did about 16 points better than expected in Math and 12 points better in ELA. Everett was, academically, the better school that year despite having fewer than 30% of its students meet standards in Math.

In the intervening years, Everett has been in the news for the wrong reasons so I’ll be shocked if it’s still so highly ranked when this year’s results are released in December. My point in introducing this old data is not to make the case for or against any particular school⁴ but to demonstrate that the data on the California School Dashboard, which is being used for the Academic Performance metric, can give a misleading impression of a school’s quality.

One more point on this growth stuff. Look at the schools in the bottom left where the growth score is below 80. If there are schools with scores in this range when the data is released in December, and they are not on the list of schools to close, the principals and staff need to be changed because they are actively damaging students.

Popularity

Here’s a slide from the SFUSD Middle School Fact Base. The schools with the most high-performing students, which are also the schools with the most Asian, White, and Two or More Race students, are also the most popular schools among applicants.

This means that popular schools attract students from across the city while unpopular ones have more local student populations. Here is where Lick students live. There are almost no students from the west side of the city.

Home residence of students at James Lick MS in 2023

Here is the equivalent chart for Roosevelt MS⁵. Although there are more students from near its location in the Inner Richmond than anywhere else, it still draws students from all across the city.

Home Residence of students at Roosevelt MS in 2023

An obvious conclusion is that if you’re directing students from a closing school to one of the schools at the top of the popularity list, you’ll encounter much less opposition than if you’re directing them to one of the schools at the bottom of the list.

Picking Schools To Close

If the district is to meet its building utilization goal, it has to reduce middle school capacity by about 4,000 students. Here are two quick-and-dumb ways of meeting that target that show the scale of the challenge:

1. Close the 3 smallest middle schools and all of the K-8 schools

2. Close the 3 biggest schools and 1 of the K-8 schools

One operational challenge with closing a middle school is that it’s rare to find another middle school with the space to incorporate all the displaced students. Consider Lick and Everett. They’re only a mile apart, have student populations that are both about 75% Latino, are both at just over 50% capacity, and are the bottom two schools in popularity among middle school applicants (only about 20% of those who would normally feed into these schools list them as their top choice and only just over 50% list them as any choice). They seem like ideal candidates to merge. The two schools have a combined 943 students. Everett’s building has space for 944 students. It’s a match made in heaven! Except that both buildings require a lot of work, Everett’s building is in worse condition than Lick’s, the district has said it doesn’t want to force students to move to a worse building, and Lick’s building has a capacity of only 816. It’s not so easy after all.

The list that follows is neither a prediction of what SFUSD will do nor a recommendation for what it should do. It’s simply an attempt to construct a list that achieves the capacity reduction target while living within the known political constraints.

Let’s go anti-clockwise around the city.

North

• A correspondent informed me that parents at Lilienthal have been told that the middle school there is closing and the students will be moved to Marina. I’m surprised that hasn’t made the news, if true, but let’s assume it is true. Marina underwent a partial modernization under the 2016 Facilities Bond so I can understand why the district might not want to close it. Lilienthal has the fewest disadvantaged students of any middle school so it’ll be interesting to see how many of those displaced students end up in Marina or in any other SFUSD school.

• If Marina stays, then Francisco goes. Marina has enough space for the displaced students.

Westside

• Presidio stays. It’s the second-largest school, is very popular with parents, and gets great results.

• Roosevelt goes. It’s a phenomenally popular school. As we saw above, 85% of applicants who have it as their feeder school list it as their top choice and 99% include it somewhere on their application, the highest percentage in the district. But equity considerations are probably going to demand that one of the westside schools gets the chop. Roosevelt had the maximum possible Academic Performance score but it is smallest of the big 4 westside schools so closing it will disrupt fewer students.

• Alice Fong Yu stays. It’s a 100% immersion school, so it’ll score the maximum possible on the Program Access criterion which is about 20% of the Composite Score. It’s also a very popular school among parents.

• Lawton stays. This is a closer call. On the one hand, it’s one of the most popular and high-achieving schools in the city. On the other hand, it’s located close to AP Giannini in the relatively prosperous Sunset, and is an English-only school with an average share of disadvantaged students so it will probably do poorly under the Equity criteria.

• Giannini stays. There’s no sound argument for closing the largest middle school.

• Hoover stays. It is several hundred under capacity so it could absorb some displaced Roosevelt students. If one of the big westside schools has to close, and it isn’t Roosevelt, Hoover is the other candidate. The size of the school means it would be even harder to accommodate the displaced students if Hoover is the one to close.

South

• Aptos stays. It’s been in the news in recent years, which is never a good thing for a school, but it’s the closest San Francisco has to a racially integrated middle school.

• Denman stays. There’s a $44m construction project underway with a second phase to follow.

• SF Community goes. It’s too small to be viable. The district has said it wants a minimum of two classes in each grade which means a middle school needs to have close to 200 students. SF Community is at 119.

• Brown stays. It only opened in 2015 and it’s in a brand new building. The district has tried very hard to make it successful, giving graduates guaranteed admission to Lowell, for example. Results have been mixed, and there's enough space at King for the entire student body, but it’s on an enrollment upswing the last two years so I’ll be surprised if the district pulls the plug.

• Visitacion Valley goes. It’s only at one-third of capacity and a majority of students who have it as their feeder school don’t even include it on their application. Students can all fit at Denman.

• King stays. It’s only at 52% of capacity and it’s hard to see how to raise that, assuming Brown stays. Visitacion Valley and SF Community students will probably go to Denman. King doesn’t have a Spanish program so couldn’t be a destination for Revere students. If you were to close King, there wouldn’t be enough capacity at Denman and Brown so where would those students go?

Center and East

• Bessie Carmichael stays. There is no middle school on Potrero Hill any more and the only real population growth that happened in the city in the 2010s was in the China Basin area. To shutter the closest middle school would be strange.

• Buena Vista Horace Mann stays. The district has embarked on a project to revamp the school. No way it closes.

• Revere goes. Revere’s general education program is too small to survive with only 13 kindergarten students. Its Spanish immersion program is doing better and the district has strongly hinted that it wants to consolidate language programs so it’s possible it could transform into a Spanish immersion K-5 school.

• Lick and Everett merge. I know I pointed out some problems with the idea but it’s too natural a fit. Lick and Everett have over 800 empty spaces between them. The district wants no more than about 1,050 empty spaces across all schools⁶. If you want to keep both Lick and Everett open, you have to either fill every other school close to its capacity or you have to raise the enrollment at those schools. The results of the current feeder system indicate that this will be challenging: when ranking schools by the percentage of feeder families that want to attend them, these are the two bottom ranking schools.

• Rooftop stays. This was originally on my chopping block because the middle school building has the worst Facilities Condition Index score of any SFUSD building. But the actual cost is somehow a mere $8m⁷ and we might be glad of the space given that Lick into Everett is a tight squeeze.

Assessment

Closing or merging Lilienthal, Francisco, Roosevelt, Visitacion Valley, SF Community, Revere, and Lick would reduce the capacity by 4,185, pretty much exactly the required level of reduction. 24% of all students would be affected, including 21% of Asian students, 28% of Latino students, 25% of White students, and 26% of Black students. The most controversial part of the list is the inclusion of Roosevelt. Where would the displaced students go? If we’re closing Francisco, there’s no room at Marina. The remaining westside schools could accommodate many but not all. Some would have to go as far as Aptos. One of the other flaws is that it leaves King as a large half-empty school with no prospect of increasing its enrollment.

Alternatives

Let’s try some alternatives to Roosevelt. Hoover is so big that, if we simply closed it instead of Roosevelt, we’d have reduced the overall capacity by too much. We could give a reprieve to a couple of the K-8 schools that were previously on the list (e.g. Lilienthal, and SF Community) or to Lick, the smallest of the middle schools on the list. The challenge then becomes getting displaced families to send their children to Lick.

All the schools that I grouped under the South heading had 2,806 students. Aptos, Denman, and Brown have a combined capacity of 3,061 students. If we were to close King instead of Roosevelt, no-one would be forced to commute outside that area. Since King is smaller than Roosevelt, this alternative would disrupt fewer students in total but the equity optics would worsen: 31% of Latino students would be affected compared to only 15% of Asian students.

Another alternative is to close Aptos instead of Roosevelt. In this case, the remaining three schools in the South (i.e. Denman, King, and Brown) only have room for 2,576. Some of the displaced Aptos students would have to move to Hoover, which would be closer to home for some of them. We’d affect 26% of all students, including 22% of Asian and 33% of Latino students.

What would you do?

1. Only Hoover, Presidio, Roosevelt, and Giannini had their capacities changed in this way. The biggest changes were at Hoover (1,097 to 1,303) and Presidio (1,046 to 1,219). In total, the changes added about 500 students to the capacity.

2. SFUSD’s method for calculating the capacity of K-8 schools was not consistent with its method for middle schools because it assumed 33 students in middle school grades rather than 30. I produced my own estimate based on the number of classrooms. A full K-8 school has three classrooms in each grade from K-3 and two classrooms in each grade from 4-8 (because the maximum class size rises from 22 to 33). That would translate into a total of 3*4 + 2*5 = 22 classrooms, of which six are allocated to grades 6-8. I estimated the middle school capacity of K-8 schools as number of classrooms * 6/22 * 30 students per class * 85% utilization.

3. If all you knew about these schools was the share of students who are either Asian, White, or of Two or more races, that one piece of data would predict about 74% of the variation in academic scores (i.e. the R-squared of the regression in 0.74, which is outrageously high given all the other things that are supposed to affect educational outcomes). If you took the same students and put them in a different school, they might perform just as well.

4. That’s why I didn't label any of the other schools or let you mouseover the marks to see the school names. Schools that did really badly in 2019 might do much better in 2024 and vice versa. We should judge them on how they are now, not how they were back then.

5. Similar charts for all schools can be found here.

6. I don’t understand how the FCI can be so high but the dollar cost so low.

7. A 90% building utilization goal for 9,948 students implies space for 9948/0.9 = 11,053 students. 11,053 - 9,948 = 1,055.