As a parent in 94134, one dimension that feels missing here is school desirability. For many families, the main factor isn’t distance itself, but access to a school they feel good about within a reasonable commute, especially given that many families already drive their kids to school and San Francisco is geographically small.
In my experience, when parents choose schools outside their zip code, it’s usually driven by differences in academic outcomes, stability, or reputation, not because geography doesn’t matter.
The enrollment patterns described here seem more like a reflection of those differences than a problem caused by the assignment system alone.
Curious how the data looks when accounting for socioeconomic factors and school ratings, given that families face very different constraints in how much choice and mobility they have
I didn't want to present a comprehensive analysis here on how much distance vs. special academic programs, academic outcomes or anything else factors into school choice. I was just trying to illustrate how the pattern is in large part a result of "stepping stone" effects, so there is distance sensitivity but also a directional bias that creates the larger-scale pattern (many students from 94124 attend schools in 94134 or 94110, then many in those zip codes attend a school west or north of their zip code, and so on).
Also pertaining to your last point (socioeconomic factors and mobility), there is a group at Stanford that had access to the detailed application data from around 2018 to propose zone options, and as part of that they modeled individual choices. They do mention that “the CTIP1 status students’ choices were found to be more sensitive to distances” (Mentzer, K.L. “Student assignment design for the San Francisco Unified School District”. Ph.D. dissertation. Institute for Computational and Mathematical Engineering, Stanford University, p.36).
I think the narrative (not from your post, but other feedback I've seen) that the large-scale commute pattern is driven primarily by the CTIP1 tiebreaker is mostly wrong; as stated above, 94112 as no CTIP1 areas, 94134 is majority non-CTIP1 as well, yet plenty of students commute out. 94110 has more CTIP1 areas than those, but the presence of many citywide language programs that bring students in compensates for those commuting out to create a positive net commuters number.
A last note, even in a "neutral" situation (if school desirability didn't vary that much) you would expect a negative net commute from zip codes that have lots of students (e.g., 94112), especially if there is plenty of spare capacity in the system. What I mean is: let's say you have two neighborhoods A and B, and families from A like the schools in B as much as families in B like the schools in A, then if there are many more families in A and extra school capacity in B you'll have a net commute from A to B just from the differences in populations.
Thanks for the thoughtful response, I appreciate the additional context.
Your point about stepping-stone effects having a large impact on the broader enrollment patterns makes sense and I’ll take a look at the Stanford work you referenced.
Id be curious to see how citywide schools factor into this, and whether there are meaningful differences between K8 schools and citywide language programs in terms of how they affect commute patterns (not sure if such a dataset exists)
As a parent in 94134, one dimension that feels missing here is school desirability. For many families, the main factor isn’t distance itself, but access to a school they feel good about within a reasonable commute, especially given that many families already drive their kids to school and San Francisco is geographically small.
In my experience, when parents choose schools outside their zip code, it’s usually driven by differences in academic outcomes, stability, or reputation, not because geography doesn’t matter.
The enrollment patterns described here seem more like a reflection of those differences than a problem caused by the assignment system alone.
Curious how the data looks when accounting for socioeconomic factors and school ratings, given that families face very different constraints in how much choice and mobility they have
I should write another analysis on the more detailed school assignment results the district has been posting last Fall (https://www.sfusd.edu/schools/enroll/student-assignment-policy/annual-assignment-highlights).
I didn't want to present a comprehensive analysis here on how much distance vs. special academic programs, academic outcomes or anything else factors into school choice. I was just trying to illustrate how the pattern is in large part a result of "stepping stone" effects, so there is distance sensitivity but also a directional bias that creates the larger-scale pattern (many students from 94124 attend schools in 94134 or 94110, then many in those zip codes attend a school west or north of their zip code, and so on).
Also pertaining to your last point (socioeconomic factors and mobility), there is a group at Stanford that had access to the detailed application data from around 2018 to propose zone options, and as part of that they modeled individual choices. They do mention that “the CTIP1 status students’ choices were found to be more sensitive to distances” (Mentzer, K.L. “Student assignment design for the San Francisco Unified School District”. Ph.D. dissertation. Institute for Computational and Mathematical Engineering, Stanford University, p.36).
I think the narrative (not from your post, but other feedback I've seen) that the large-scale commute pattern is driven primarily by the CTIP1 tiebreaker is mostly wrong; as stated above, 94112 as no CTIP1 areas, 94134 is majority non-CTIP1 as well, yet plenty of students commute out. 94110 has more CTIP1 areas than those, but the presence of many citywide language programs that bring students in compensates for those commuting out to create a positive net commuters number.
A last note, even in a "neutral" situation (if school desirability didn't vary that much) you would expect a negative net commute from zip codes that have lots of students (e.g., 94112), especially if there is plenty of spare capacity in the system. What I mean is: let's say you have two neighborhoods A and B, and families from A like the schools in B as much as families in B like the schools in A, then if there are many more families in A and extra school capacity in B you'll have a net commute from A to B just from the differences in populations.
Thanks for the thoughtful response, I appreciate the additional context.
Your point about stepping-stone effects having a large impact on the broader enrollment patterns makes sense and I’ll take a look at the Stanford work you referenced.
Id be curious to see how citywide schools factor into this, and whether there are meaningful differences between K8 schools and citywide language programs in terms of how they affect commute patterns (not sure if such a dataset exists)