Although public schools are not generally subject to direct competition for students, it is commonly thought that
they nonetheless face competition through parents' residential choice. Such competitive effects are likely to depend
on the relative proximity of school districts if it is less costly to move short distances than long, or if parents are
able to more easily send their children to nearby districts through open enrollment policies. Using panel data for
607 Ohio school districts from 1998 to 2007, I test for strategic interaction over teacher salaries and standardized
test scores. I present evidence that Ohio public school districts act to 'follow their neighbors'- that is, that they
attempt to exactly mirror changes in the inputs and outputs of nearby school districts and I show that this result
is robust to different definitions of 'neighbor.' I further show that conventional estimation of spatial autoregressive
models via Maximum Likelihood or via poorly-instrumented General Method of Moments may create large biases in
the estimated spatial autocorrelation coefficient. I suggest that this statistical phenomenon may explain some of the
differences in estimated magnitudes of school competition across the spatial literature