The Governor’s Office of Student Achievement (GOSA) publishes mobility rates for Georgia public schools. The GOSA mobility rate, also known as the churn rate, counts the total number of student moves in or out of a school between October 1 and May 1, and divides by the total number of students enrolled on the first Tuesday in October.

The formula is significant because mobility rates can have large variation depending on the formula. The graph below compares the GOSA rate to an alternative rate, labeled the “new student rate”, defined as the percentage of students enrolled on May 1 who started at the school on or after October 1.

Notice the new student mobility rate is usually less than half the GOSA rate. This is because the GOSA rate counts both entries and exits, while the new-student rate only counts entries. Also, if a student moved into and then left a school, the GOSA rate would count that student twice. Both rates are useful, but the new student rate is more helpful context for state-testing results. Because state testing occurs in late April, the new student mobility rate is a close approximation of the percentage of tested students in a school who transferred into the school mid-year.

**Effect of Mobility**

To look at the effect of mobility, we look at the student growth percentiles of mobile versus non-mobile students. Student growth percentiles (SGP) are calculated by the state, and report a student’s percentile of test score growth compared to other students with similar prior test scores. As percentiles, SGP are reported on a hundred point scale.

APS students in grades 3-8 who changed schools during the 2017 school year had an SGP score 9.2 points lower in math, and 3.6 points lower in ELA than their non-mobile peers^{1}. However, mobile students might differ from non-mobile students in more ways than just their mobility. To improve the comparison between mobile and non-mobile students, we use linear regression to control for other student characteristics related to student growth.

The table below reports regression results of 2017 SGP score^{2} regressed on mobility, English learner status, direct certification for free lunch, special education status, and school fixed effects^{3}. After controlling for these other factors, mobile students have an SGP score 5.5 points lower in Math and 1.1 points lower in ELA than their non-mobile peers. The ELA difference is not statistically significant.

**Table 1: SGP Regression Results**

Mobility has a larger coefficient than direct certification and English Learner status in the math model, but this is because the dependant variable is test score *growth*. The other control variables are less temporal than mobility- a student who is direct certification this year tends to be direct certification the previous year. So the larger relationship between direct certification and achievement is mostly differenced out when measuring growth.

For context at the student level, a 5 point improvement in math SGP score typically translates to about five scale score points, which is 10% of the distance from developing to proficient achievement levels (475 to 525). At the school level, note that a school with a 20% new-student mobility rate would then have a Math SGP average that is one point lower due to mobile students (20% x 5 points).

The controls in the regression equation above are limited, and likely do not fully capture the difference in family circumstances between mobile and non-mobile students. For example, a student might be mobile because of a divorce or family seperation, which could have an effect on student achievement independent of mobility. So the reported SGP gaps are likely an over-estimate of the actual effects of mobility.

#### Footnotes

- This statistic compares students who moved into an APS school between September 30 and state testing in late April to APS students who did not change schools during that period.
- A more typical method would be to use student test score as the outcome variable instead of SGP, and control for prior test scores. This method would also result in much higher R-square values. However, using SGP allows us to include students who transferred in from out of district, which is a large percentage of mobile students.
- “School fixed effects” means that we are controlling for the average SGP effect of each individual school.