SFUSD Elementary Class Sizes Are Too Small

And Class Size Variation Is Too High

School districts have a lot of freedom in how they spend their money. Some might choose to have smaller class sizes. Some might choose to have larger class sizes but to pay their teachers more. Others might choose to have larger class sizes but to employ more specialists who can provide one-on-one or small group assistance to struggling students. Districts are naturally going to end up making slightly different trade-offs between these spending priorities. To the extent that there is a wisdom of crowds, any district whose spending pattern deviates significantly from the average should do some soul-searching. Is it obtaining good results for its students or would it be better served by reallocating its money to one of the other priorities?

Some districts have a lot more revenue per student than others. A richer district will probably have more teachers, pay them better, and give them more in-class help. It will have smaller class sizes simply because it can afford to, not because it has favored that over other priorities.

This post is going to focus on elementary school class sizes. SFUSD is fortunate to receive more money per student than every other large district in the state. In terms of revenue per student, it trails only a few smaller Bay Area districts. Does this translate into smaller class sizes?

Methods

Dataquest shows enrollment by grade in each school but the CDE also keeps a record of the enrollment in every class in every public school in the state. I downloaded the most recently published data, which is from 2018-19. That means we miss out on the recent enrollment declines but it also means that the pandemic doesn’t provide an excuse for whatever we observe. I picked 1st grade as the point of comparison and focused only on regular classes (not PE or music).

My first thought was to calculate the average class size for every district but when I dug into the raw data, I found that was impractical. This is what the raw data shows for 1st grade classes in SFUSD.

There are a decent number of classes with just 1, 2, 3, or 4 students. These classes almost certainly represent special education students. For example, the raw data shows Alvarado ES had 1st grade class sizes that year of 22, 22, 20, 18, and 4 even though Alvarado’s SARC shows that it only had four 1st grade classes. The raw data shows that Sunnyside ES had class sizes of 22, 21, 18, 2, and 1 even though it has only three 1st grade classes. It would clearly be misleading to say that the average 1st grade class size at Alvarado is (22+ 22 + 20 + 18 + 4) / 5 = 17.2 and at Sunnyside is (22 + 21 + 18 + 2 + 1) / 5 = 12.8. We could adjust the denominator by excluding any class that is smaller than, say, 10 students (on the assumption that these tiny classes represent data artifacts rather than real classes). The averages for Alvarado and Sunnyside would then become 21.5 and 21.3, respectively.

We’re going to focus mainly on the distribution of class sizes in a district. This will enable us to ask questions such as:

• what percentage of students are in classes of 20 or fewer? Of 25 or fewer?

• what is the maximum number of students in a class?

• what is the class size of the median student (i.e. the class size such that 50% of students are in smaller classes and 50% are in larger classes)?

1st Grade Class Sizes

The standard argument for small class sizes is that the smaller the class size the more time a teacher can devote to each student and the more that student will learn. There is an enormous body of literature about the effect of class sizes. Here’s a decent summary of the evidence from the Brookings Institution.

Here’s how the SFUSD compares with ABC, Clovis, and Long Beach, three districts I’ve used previously as exemplars of high performing districts because each obtains much better results for its more disadvantaged students than SFUSD does.

SFUSD assigns no more than 22 kids to a kindergarten class so, as expected, nearly 100% of 1st graders are in classes of no more than 22 students. (It’s not exactly 100% because there were four classes of 23 scattered around the district). Even though 22 is the maximum, many classes have fewer than 22 students. About 40% of SFUSD 1st graders are in classes of 20 or fewer. In contrast, ABC, Clovis, and Long Beach have maximum class sizes of 28, 27, and 30 respectively and fewer than 25% of students are in classrooms of 22 or smaller. Clearly, what the teacher does in the classroom is far more important than the number of students in the classroom.

Here’s how the largest districts in other Bay Area counties compare with SFUSD.

Only Palo Alto has the same maximum class size as San Francisco. Most of the others are in the 24 to 28 range. There is a reason Palo Alto’s class sizes are so small. We’ve already mentioned that SFUSD gets substantially more than most districts in the state. But Palo Alto Unified is in a different league altogether. It receives so much local property tax revenue that it can afford to spend over $23,000 per student compared to SFUSD’s $18,000. In other words, the only district to match SFUSD’s elementary school class size has 25% more money than SFUSD does.

Here’s a histogram showing the maximum 1st grade class size in every district in the state that has more than 1000 first grade students. San Francisco is one of the two with a maximum of 23 (the other is Tracy’s Jefferson Elementary). And remember San Francisco would be alone at 22 if it weren’t for those four classes of 23.

The most common maximum class size among large districts is 26. The median is 27. It is clear that SFUSD is an outlier. It has Palo Alto level 1st grade class sizes without Palo Alto level income.

4th and 5th Grade

The focus on small class sizes, and the extra money to support it, has been at the K-3 level. Consider what happens when those students reach 4th and 5th grade. If a school has only one class per grade, the district is obliged to support that really small class size in grades 4 and 5 too. If a school has two classes per grade in K-3, the district is also on the hook to continue those small class sizes in grades 4-5 because no one wants a class of 40+. If a school has three classes per grade, then the school can save some money by dropping one of those classes for grades 4-5 at the cost of bringing the size up from 20 to 30. If a school has four classes per grade, consolidating into three for grades 4 and 5 brings the average up to only 27.

Of course, many SFUSD schools are under-enrolled. The class size differences are then magnified if a teacher is lost in grades 4 and 5. Three classes of 22 result in two classes of 33 but three classes of 18 result in two classes of 27. What was a difference of four is now a difference of six.

Language immersion programs compound the problem. With the exception of schools like Alice Fong Yu (Cantonese) and Marshall (Spanish) and the Chinese Immersion School at DeAvila, schools that have language programs also have general education (GE) programs. For class size planning purposes, the language and GE programs are effectively separate schools because the kids can’t be combined in a single classroom. Lau ES and Taylor ES have five full classes in kindergarten but those classes are divided among two separate programs at Lau and three at Taylor. Taylor is effectively not a five class school but three separate schools each with one or two classes per grade.

One solution is to mix 4th and 5th graders in the same class. Glass-half-full people will see this as a less expensive way to get a smaller class size. Glass-half-empty people will bemoan the inability to fund an extra teacher that would obviate the need for the mixed-grade classroom. Across the state, 8% of 4th and 5th graders are in mixed-grade classrooms. In SFUSD, that number is 17%. About 40% of SFUSD’s elementary schools have at least one 4th/5th mixed-grade classroom.

The following chart shows the average class size in each school. The schools labeled in red have at least one mixed age classroom. The variation in class size is striking. Thirteen schools had average class sizes of more than 30 while another 13 had average class sizes under 20.

Let’s consider the schools that have four total 4th and 5th grade classes. The highest average class size in the city belongs to Sunset ES. It has three fully enrolled classes of 22 (or even 23) in grades K-3. But they lose a teacher for grades 4 and 5 so those kids end up in classes of 32 or 33. Lawton K-8, Sloat ES, Sunnyside ES, and Yu K-8 are all in the same situation.

Further down is Sutro ES. Sutro had only two classes in K-3 from which 41 students moved to grade 4. That justified two 4th grade classes and it also had two classes for its 52 5th graders. Tenderloin ES has a slightly higher average than Sutro but got there very differently. It had 56 students in 2nd grade (split across 4 classes so an average of 14!), 60 in 3rd grade (3 classes so an average of 20), 49 in 4th, and 45 in 5th.

Most of the schools below Sutro have language programs and GE programs. The kids in the two programs can’t be combined in a single class so the district ends up with one small class in each program in each year. For example, Muir ES had classes of 19 and 20 in 4th grade and 12 and 14 in 5th grade. Because one was general education and the other was Spanish Biliteracy, they didn’t have the option of combining the 5th grade class into one.

El Dorado ES and Carver ES don’t have language programs. They just don’t have enough students. El Dorado had separate classes of 12 and 15 in 4th grade at a time when other schools had over 30 per class. For Carver, the numbers were 13 and 16.

Here’s another way of looking at the exact same data, this time with the number of students on the y-axis¹. This makes the unfairness even more evident. Among schools with around 60 4th and 5th graders, why does Sheridan have one fewer teacher than Garfield and two fewer than El Dorado? Similarly, some schools with 120+ students in 4th and 5th grade are served by only 4 teachers while others with fewer students have five or even six teachers. It is hard to justify such disparity.

Compared to Other Districts

Here is how SFUSD’s class size for 4th/5th grades compares with other districts.

Just as before, SFUSD has significantly smaller class sizes than the high performing districts I follow. SFUSD’s curve has a different shape than the others. The other districts have very few really small classes. 75% of their kids are in classes that are within 5 or 6 of the district’s max. For example, Long Beach’s maximum class size is 35 but 75% are in classes of 31-35. For Clovis, 75% are between 33 and 37 and for ABC 75% are between 28 and 33. SFUSD’s curve is practically a straight line. To span 75% of students, our range needs to be 21-33. Nearly one quarter of students are in classes of 20 or fewer.

Notice again the contrast between SFUSD’s shape and every other district. Palo Alto can afford to have a maximum class size of 24 but they have a smaller percentage of students in classes of 20 or fewer than SFUSD whose maximum is closer to 33.

Every district treats its 4th and 5th graders more equitably than San Francisco does. San Rafael, Berkeley, and Fremont all have maximum class sizes of 29-30, lower than San Francisco, but San Francisco has a higher proportion of students in classes of 24 or fewer and a much higher proportion in classes bigger than 30.

Conclusion

San Francisco is an outlier in two ways.

1. Its K-3 class size is far smaller than comparison districts. It matches that of much richer Palo Alto.

2. The variation in class sizes across schools within the district is enormous. No other district shows such variation. This starts in K-3 and then gets worse in grades 4-5. Some students continue to get Palo Alto or better class sizes but many others are in much larger classes.

To be clear, SFUSD did not set out deliberately to create this disparity. The disparity arose out of other decisions. It operates too many schools. It operates too many classrooms for the number of students, meaning that some classes are inevitably going to be smaller than others. It combines general education and language programs in the same school. And then it sets a K-3 class size that it cannot afford. The result of these combined decisions is what last week’s post showed. SFUSD has to take money from its high schools to pay for the commitments it has made at the elementary school level.

One thing I’ll be looking for when the Elementary School Assignment team moves on from discussing Theories of Action and starts performing actions is how they deal with language programs. They have already said that every zone will have access to each type of language program. I’ll be looking to see if they stop mixing language programs and general education programs in the same school. If they do, it’ll be a sign that they’re taking the budget crisis seriously.

1. yes, I’m aware that the independent variable, the number of students, should properly be on the x-axis, and the dependent variable, the number of classes for those students, should be on the y-axis. But then you wouldn’t be able to read the labels and I decided legibility was more important.