School Size, part 2
Introduction
Last week’s post showed that San Francisco has far more small schools than any other district except Oakland (which is planning to close or merge 11) and Marysville Joint Unified, the worst school district in California, and that most of the very small schools are elementary schools.
This post will look at the capacity of SFUSD’s schools and the enrollment trends to identify schools that would be candidates for closure. It will also cover the political considerations that might prevent any closures.
Planning for Growth, not Decline
Rather than planning for school closures, SFUSD is planning to build more schools. The district is building a new elementary school in Mission Bay projected to open in 2025 and it believes this will be the “first in a pipeline of new schools”. SF Planning believes that 11 new elementary schools will be needed by 2040 to accommodate population growth.
It should be noted that population growth doesn’t necessarily translate into school growth. Although San Francisco added nearly 70,000 people between 2010 and 2020, only 5700 of them were under 18. Of those 5700 extra kids, precisely none went to SFUSD: enrollment in SFUSD schools was flat between 2009-10 and 2019-20. Charter school enrollment grew by 1900 and private school enrollment by about 600. I don’t know where the other 3200 kids are. Surely they can’t all be under 5?
In a presentation about the planned new elementary school in Mission Bay, SFUSD included this very informative map which repays close study:
The areas in green are those where the schools have more spaces than there are students living in the neighborhood. The areas in red are those where the school capacity is below the student population of the area. The numbers show, for each of the current elementary school attendance areas, what is the capacity of the elementary schools in that area and how many TK-5 students there are living in that area. Essentially, if every student were to attend their local elementary school would those schools have enough capacity? For this purpose, the capacity of the citywide schools is added to the capacity of the elementary school attendance area in which it is located. So 1354 is not the capacity of Clarendon alone but the combined capacity of Clarendon and the three citywide schools located within the boundaries of its attendance area.
The central and northern areas have more capacity than they have students and the southeastern areas have less capacity than they have students. The implication of this is clear: even if all schools were equally good, and were seen to be equally good by everyone, it’s impossible to create an elementary school assignment process that does not involve a flow of students to the north and west.
But the situation is worse than that. The areas with the greatest school under-capacity are precisely the areas whose schools are under-enrolled. 920 students lived in the attendance area for Carver Elementary in Bayview which has a capacity for only 308. Even if Carver were the best and most popular school in the city, 600+ neighborhood kids would have to go elsewhere for elementary school. But Carver’s actual enrollment was only 151 in 2018-19 down from 391 in 1998-99. It was down to 111 in 2020. The people who live right next to Carver don’t want to send their kids there.
What to do? I can think of at least five strategies that could be employed for schools like Carver.
The trite answer is to say we’ll just make it better so that locals want to send their kids there. I don’t doubt the sincerity of SFUSD’s desire to improve schools like Carver but it’s clear it doesn’t know how. The district has tried pouring money into the school: it receives far more money per student than the average school. One calculation pegs its budget at $15,136 per pupil whereas near neighbor Bret Harte receives only $8035 per pupil. Whatever the problem is at Carver, it’s not money.
Switch to a neighborhood based system where students are automatically assigned to their neighborhood school and can only attend a different school if the neighborhood school is full. That would ensure those schools are full and it would minimize commute times but it would be open to obvious attack lines (“the district is forcing our kids to attend a bad school”). Many parents would be in the bizarre position of hoping to lose the lottery for their local school so that they would qualified to enter the lottery for any spare places that might exist at their preferred school.
Force students from outside the area to take the spots that locals don’t want to take. The new zone-based assignment system will do this.
Invite KIPP or some other charter school provider to take over the school. Given the record of their other schools in the city, KIPP could surely fill Carver to capacity with students drawn primarily from Bayview and neighboring areas. That might be in the interests of local children (who could attend a good school close to home) but it would be seen as an admission of defeat on the part of SFUSD.
Close the school. The existing students could all be accommodated at the nearby Drew ES (which has enough capacity for both). The problem with this approach is that, after a while, someone will come along and notice all the kids from Bayview commuting north and west for school and say: “it’s nuts that this area has so many kids but no school. If we open a school here, it’ll thrive.” At which point we’ll be back where we started.
SFUSD’s Overcapacity
If we add up the capacity numbers shown in the Mission Bay presentation, we find that the capacity of all existing elementary and K-8 schools is just shy of 29000 (actually 28950). SFUSD has not served more than 26,300 in twenty years.
In 2019-20 (before the pandemic-induced decline), the actual K-5 enrollment in SFUSD was 25800 (25797 to be precise), indicating capacity for an extra 3200 students. Given that San Francisco has 27 schools with fewer than 300 students, many way under 300, it could close ten of them and still have room for all its students.
I am not going to propose which schools to close or merge because that’s a multi-factor decision and I don’t have enough data. For example, in selecting which schools to close or consolidate, it’s necessary to take the capacity of the schools into account. Suppose a district needs to close one of its two schools. School A has only filled 100 of its 400 spaces while school B has filled all its 250 spaces. Even though school B is obviously more successful, it is the one that would have to close because it doesn’t have the space to incorporate school A’s students.
The chart from the Mission Bay presentation doesn’t given the capacity of all elementary schools, just those with no citywide school in their attendance areas. As a sanity check, I compared the declared capacity with the actual enrollment history for the schools with the smallest capacities.
Malcolm X: Capacity 176. I suspect this capacity number is so low because it was at the time sharing its campus with KIPP Mission Bay Elementary, a charter school. Malcolm X had 445 students in 1997. In 2005-06, when it was considered for closure but reprieved, it had 164 students and this was supposed to be 45% of capacity which would imply a capacity of 364.
McCoppin: Capacity 242. It had 278 as recently as 2013 and 353 in 1998.
Milk: Capacity 248. Milk’s largest enrollment was 253 in 2011.
New Traditions: Capacity 263. Peak enrollment was 260 in 2015
Sheridan: Capacity 264. It was 360 in 2000 but hasn’t exceeded 264 since 2002.
Bryant: Capacity 271. This school hasn’t come close to this level since moving to its current location.
Peabody: Capacity 277. It has not exceeded this capacity since 1994.
Yick Wo: Capacity 281. Never been at capacity.
Garfield: Capacity 285. Last exceeded this capacity in 1995
Harte: Capacity 294. Was at 412 in 2001.
Redding: Capacity 297. Was at 331 in 2009 and 381 in 1998.
Carver: Capacity 308. Was at 391 in 1998 and last exceeded the supposed capacity in 2002.
El Dorado: Capacity 308. Was at 325 in 2004 and 349 in 2001.
Parker: Capacity 308. Was at 336 in 2004.
Why were several of the schools listed here well above their stated capacities in the past? It is unclear how the capacity numbers were produced and what class size limits they assume. It is possible that class size reduction policies introduced since the 1990s might have reduced the capacity of schools. But, if that’s true, I can’t explain why some of the larger and more popular schools don’t seem to have been affected. For example:
Jefferson: Capacity 502. Enrollment 513 in 1998, 509 in 2020
Ulloa: Capacity 528. Enrollment 510 in 1998, 528 in 2020
West Portal: Capacity 593. Enrollment: 569 in 1998, 575 in 2020
Enrollment Trends
Elementary Schools
Here is a graph showing K-5 public school enrollment in San Francisco since 2000. Note that the vertical axis starts at 20,000 to emphasize the year-to-year fluctuations. There was a precipitous drop in enrollment in the early 2000s, in response to which SFUSD decided in 2005-06 to close or merge a number of schools. This didn’t reduce capacity all that much because the 500+ capacity Feinstein Elementary was opening in 2006.
Enrollment numbers then rose by a couple of thousand over the next five years. The Chinese Immersion School at DeAvila opened in 2009 on the site of the DeAvila Elementary school that had closed at the end of the 2005-06 school year. SF Public Montessori opened in 2010. Total public school enrollment was pretty much flat from 2012-2019 but SFUSD lost about 500 students to charter schools. Then came the pandemic. SFUSD elementary schools lost another 500 in 2020. Official 2021 figures have not been published but SFUSD’s dashboard shows K-5 enrollment below 25000, a drop of another 300.
School-Level Enrollment Changes
It’s very hard to measure school quality. Test scores are an insufficient indicator without very careful adjustment for school composition. A school with lots of college graduate parents is going to have higher test scores than a school where most of the parents don’t speak English and haven’t attended college, even if the teachers in the second school are better. An indirect approach to measuring school quality is to look at enrollment trends. In districts with neighborhood schools, declining enrollment might reflect changes in the demographics of a school’s catchment area. But in San Francisco, where everyone can apply to every school, declining enrollment indicates that parents don’t believe the school delivers a good education.
Here is the 2020 enrollment compared the average over the ten-year period from 2011 to 2020. The size of the disc represents the magnitude of the change in terms of the number of students gained or lost. Green discs represent schools whose 2020 enrollment was higher than its 10-year average. Red discs represent schools whose 2020 enrollment was lower than their 10-year averages. Since the aggregate 2020 enrollment was 25,215, lower than the 25,870 average over the ten years, a typical school’s enrollment will also be below its 10-year average. As we can see, the enrollment losses have not been evenly distributed.
The schools in the center and west of the city have experienced modest changes. For example, Clarendon’s enrollment over the decade ranged between 541 and 586 and its 2020 enrollment of 546 was 14 below its ten-year average of 560.1.
The heaviest declines were experienced by schools south and east of 280. For example, El Dorado Elementary’s enrollment of 143 was less than half the 308 it was earlier in the decade and a massive 83.5 below its ten-year average of 226.5. Nearby Bret Harte on the other hand, was at a 10-year high of 239 students, 35 above its 10-year average of 204. As shown in this earlier post, the under 18 population south and east of 280 fell during the decade and this probably affected the enrollment of the schools there.
The North-East corner of the city (Nob Hill, Chinatown, North Beach) also experienced significant enrollment changes. The largest elementary school in the city, Gordon Lau Elementary was a full 68 students ahead of its 10-year average but the neighboring schools (Spring Valley, Jean Parker, Sherman, Yick Wo, Garfield) were all at their 10-year lows.
Small School Enrollment Trends
Here’s another map, but just focusing on the schools with fewer than 300 students, the level below which Oakland judged that elementary schools couldn’t be staffed efficiently. That omits some schools that have experienced large enrollment losses but which are still above that threshold. For example, Visitacion Valley Elementary and Guadalupe Elementary were each down about 80 from their 10-year averages and about 150 from their 10-year highs but they still had 313 and 349 students, respectively, so they don’t appear on this map.
The disc size now represents the actual 2020 enrollment. Notice that there are practically no small schools west of Twin Peaks. Also, quite a few of the schools show up in green, indicating that they were above their 10-year averages. The North Beach / Nob Hill area stands out. There are seven small schools in that area and none of them is thriving.
There are seven schools in the city with 152 or fewer students. Two are the newcomer schools: Ed Lee, near Chinatown and Mission Education Center in Noe Valley. A third, Cobb, is in Lower Pacific Heights. The other four (El Dorado, Carver, Drew, and Malcolm X) are close to each other in the Bayview / Visitacion Valley area.
Middle Schools
Now let’s turn to middle schools. Here’s how grade 6-8 enrollment (i.e. including both K-8 schools and pure middle schools) has changed since 2000.
Middle school enrollment is much easier to forecast than elementary school enrollment: grade 6 enrollment should be the grade 1 enrollment from five years before with some adjustment for the net attrition SFUSD experiences at every grade level (i.e. grade x is always smaller than the previous year’s grade x-1, at least until 9th grade). When the district decided in 2005 to close a couple of middle schools, it would have been able to forecast the enrollment decline that we see continuing through 2011.
Although elementary school enrollment increased in the late 2000s, many of those additional students chose to attend charter schools at the middle school level rather than district schools. Enrollment in 2020 was only fractionally above the 2011 low.
Here is the picture for individual schools, comparing their 2020-21 enrollment with the ten-year average to get a sense of which schools are doing better than others.
A few takeaways:
The biggest middle schools, such as AP Giannini and Hoover, are barely visible in the picture because they are close to their long-term averages.
Denman and Everett are both well above their 10-year averages. Denman is close to its 10-year peak and nearly 100 higher than its 10-year average.
Everett has fluctuated enormously, from a low of 344 to a high of 709 but its current 622 is still a long way above its 10-year average of 550.
Lick has also fluctuated a lot. It peaked at 655 in 2016 but then lost 200 students in just five years including a drop of 100 in just one year.
The three southeastern middle schools: Visitacion Valley, King, and Brown, are a combined 180 under their 10-year averages. Brown’s enrollment was once 382 but is now only half that at 190.
High Schools
The high school picture looks different. We see the enrollment decline following through from elementary and middle school but the rebound that should have occurred starting around 2014 was a damp squib. (Again, the y-axis has been truncated to emphasize the year-to-year change). Enrollment in 2020-21 was less than 300 above the 2015 nadir and basically at the 10-year average.
It’s impossible to explain why the rebound didn’t happen without access to longitudinal student records. I believe SFUSD’s policies on repeating grades have changed over time and this may have an effect. Also, I would have thought that the increased graduation rate of which SFUSD boasts would require higher enrollment since you can’t graduate without first being enrolled.
Historically, SFUSD’s 9th grade classes have been several hundred larger than the previous year’s 8th grade classes because there weren’t as many spots available in private high schools compared to private middle schools. That bump didn’t happen in 2020.
This is the picture when we look at individual schools.
At the elementary and middle school levels, the schools in the southeast were losing students. Instead of losing enrollment, the biggest high school in the southeast, Burton, is at its 10-year high while Marshall is also on an upswing. Galileo, in North Beach, is furthest below its 10-year average in numerical terms but that school is so large that this is still a small drop in percentage terms.
The Politics
We’ve looked at evidence that SFUSD has too many small schools and at evidence that it has enough excess capacity to be able to close some of them. Now let’s look at why it may not do that.
The History
As mentioned above, the last time SFUSD closed schools was 2005-06. This article explains the decision and the politics. In a sign of how little has changed, the article was written by Heather Knight and quotes board member Mark Sanchez. Two elementary schools (Cabrillo and Swett) were closed at the time, as were some middle schools. Several other elementary schools were considered for closure but reprieved. Here is how they have fared since:
Rosa Parks has the largest enrollment (422), but it’s down from its 10-year average of 441. It has fluctuated quite a bit with a low of 391 and a high of 487.
Jose Ortega is thriving at 386 students, compared to a 10-year average of 372.
McKinley is at 344, close to its 10-year low.
Peabody is small (273) but stable (10-year range 254-277)
Starr King and Daniel Webster were going to be merged. Both are now thriving. Webster at its 10-year high and Starr King is just off its 10-year high.
New Traditions was going to be merged with Grattan. New Traditions is at 253 but is stable: its 10-year range is a narrow 234-260. Grattan is larger and even more stable (10-year range 384-397).
Sheridan is very small (211) but quite stable (10-year range 202-229)
Malcolm X has never thrived (10-year range 85-112). It had 164 when it was reprieved in 2005-06.
In other words, some would still be on the chopping block today while others definitely wouldn’t be.
The Money and Race Angles
As the map above showed, most of the uneconomic schools (i.e. those with fewer than 300 students) lie in the east of the city. In an earlier post, we saw that it was possible to draw a roughly north-south line through the city such that all the schools with below average numbers of poor kids (i.e. kids eligible for free or reduced-price meals) were to the west and all the schools with above average numbers of poor kids were to the east.
Here’s a chart showing the size of each school and the percentage of students who are eligible for free and reduced-price meals (FRPM eligible), the standard measure of poverty in education. Above 300 students, there’s no real correlation between school size and poverty rate. But the vast majority of schools with fewer than 300 students are high poverty. 19 are > 60% FRPM-eligible compared with only 5 which are < 40% FRPM-eligible.
Any attempt to close uneconomic schools will therefore disproportionately affect students from poorer families. That by itself would be politically controversial.
And then there’s the race angle. The average Asian elementary school student attends a school of 479 students but the average Black elementary school student attends a school of only 356 students. The numbers for White and Latino students (426 and 404 respectively) lie between those two. (In other words, the student groups that do better on standardized tests are on average in larger schools which implies that the educational advantage conferred by smaller schools, if any, must be comparatively small.)
Black and Latino students are 12% and 36% of the students at schools with fewer than 300 students compared with 5% and 29% of the students at schools with more than 300 students. The most uneconomic schools, those with fewer than 200 students, are 34% Black and 31% Latino and include all three majority Black schools. Any attempt to close smaller schools will inevitably impact those communities more and will be portrayed as an attack on those communities.
The Real Estate Angle
The district owns its properties, and doesn’t charge its own schools rent, so neither the schools nor the district face economic pressure to use school space efficiently.
The board members boast of their dislike for charter schools but SFUSD is obliged by law to find room for them but can and does charge them rent. If SFUSD closes a school, that building becomes available to charter schools. Indeed, most of the school sites closed by SFUSD in the 2000s are used today by charter schools:
Creative Arts Charter School and Gateway Middle School share the campus of Golden Gate Elementary (1601 Turk St)
KIPP SF Bay Academy and Gateway High share the campus of Benjamin Franklin Middle (1430 Scott St)
KIPP SF College Preparatory School is on the site of the former Gloria R Davis Middle School (1195 Hudson St in Bayview)
Mission Prep (1050 York St) is located where Bryant Elementary used to be until Bryant moved to occupy the former site of Buena Vista Elementary after it merged with Horace Mann Middle School.
City Arts and Tech, a charter high school, and Jordan School for Equity, a SFUSD high school, share the campus of the former Luther Burbank Middle School (325 La Grande Ave)
Other school sites are used by SFUSD for its own purposes. For example, Cabrillo Elementary (735 24th Ave, in the Richmond) is used by SFUSD for professional development. John Swett Elementary (727 Golden Gate Ave in the Western Addition) is now SF County Civic Center Secondary. Whereas the elementary school had around 300 students, the secondary school had 64 in 2020.
Money that a neutral observer would think is wasted by operating too many schools or using space inefficiently can be justified as a necessary expenditure for the goal of restricting charter schools. Think of SFUSD as a landlord with a low property tax base who can afford to hold a property off the market instead of renting it to people he doesn’t like.