class: inverse, center, middle <br> <br> <br> <h1 class="title-own">Zoned-In and Zoned-Out:<br/>Changes in School Attendance Zones over Time</h1> <br> <br> .small[Magdalena Bennett <br>*McCombs School of Business, The University of Texas at Austin* ] <br> .small[AEFP Conference<br>March 24th, 2023] --- # Attendance zones, new schools, and segregation - Schools can either **.darkorange[attract]** or **.darkorange[push away]** residents depending on socioeconomic characteristics .small[(Hasan & Kumar, 2019; Gibbons, Machin, & Silva 2013; Figlio & Lucas, 2004)] -- <br> <br> - Neighborhoods have important effects on **.darkorange[long-term outcomes]** .small[(Chetty et al. 2020)] -- <br> <br> - **.darkorange[Racial and socioeconomic disparities]** in the school system also have long-term effects on students .small[(Reardon, 2016; Billings et al., 2014)] -- .box-7Trans[New public schools opening → Changes in attendance zones] --- # This paper .pull-left[ - How do **.darkorange[changes in attendance zones (AZ)]** affect: - Zoned-in areas (i.e. neighborhoods)? - Zoned-out areas (i.e. neighborhoods and schools)? ] .pull-right[ .center[  ] ] --- # This paper .pull-left[ - How do **.darkorange[changes in attendance zones (AZ)]** affect: - Zoned-in areas (i.e. neighborhoods)? - Zoned-out areas (i.e. neighborhoods and schools)? - **.darkorange[New high schools]** in Texas that changed AZ `\(\rightarrow\)` mostly gentrified areas ] .pull-right[ .center[  ] ] --- # This paper .pull-left[ - How do **.darkorange[changes in attendance zones (AZ)]** affect: - Zoned-in areas (i.e. neighborhoods)? - Zoned-out areas (i.e. neighborhoods and schools)? - **.darkorange[New high schools]** in Texas that changed AZ `\(\rightarrow\)` mostly gentrified areas - **.darkorange[Outcomes of interest]**: - Differences in scores and score gaps between race/ethnicity - Differences in school composition - Differences in neighborhood composition ] .pull-right[ .center[  ] ] --- # What do we mean by zoned-in and zoned-out areas? -- .pull-left[ **Zoned-in area**: Attendance zone for a new school *S'* .center[ ] ] -- .pull-right[ **Zoned-out area**: New attendance zone for school *S* after the opening of schools *S'*. .center[ ] ] --- # What do we mean by zoned-in and zoned-out areas? .pull-left[ **Zoned-in area**: Attendance zone for a new school *S'* .center[ ] ] .pull-right[ **Spillover school**: School which had a part of their catchment area zoned-in to S'. .center[ ] ] --- # Data - **.darkorange[Common Core Data (CCD) [2005-2019]]**: Administrative data from NCES, including demographic and socioeconomic characteristics of schools. -- <br> <br> - **.darkorange[Texas Educacion Agency (TEA) data [2005-2019]]**: Performance data for schools over time. -- <br> <br> - **.darkorange[Attendance zones maps [2009-2017]]**: Geographic data for school boundaries over time from SABS, SABINS, and Maponics. -- <br> <br> - **.darkorange[Census and American Community Survey (ACS) data [2010-2019]]**: Demographic information at the census tract level -- <br> <br> - **.darkorange[Housing Prices data [2005-2019]]**: Information about housing prices over time from CoreLogic and Zillow *(coming soon)*. --- # Identification Strategy: An Augmented Synthetic Control Method - Use a **.darkorange[weighted average]** of similar districts/schools/AZ/neighborhoods to create a comparison group for affected areas. -- - Under **.darkorange[Augmented Synthetic Control Method]** (ASCM) (Ben-Michael et al., 2020) there is a correction for poor fit: `$$\hat{Y}^{aug}_{1T}(0) = \sum_{W_i=0}\gamma_iY_{iT}+\color{#db5f12}{(\hat{m}_{iT}(\mathbf{X_i})-\sum_{W_i=0}\gamma_i\hat{m}_{iT}(\mathbf{X_i}))}$$` .small[ - `\(m_{iT}\)`: Estimator for `\(Y_{iT}(0)\)` - Extrapolation for "bias correction". - If ridge regression is used `\(\rightarrow\)` penalization for extrapolation.] -- - Proposal of **.darkorange[sensitivity analysis to hidden bias]** .small[(Rosenbaum, 2002; Keele et al., 2019)]: - *How much should an unobserved confounder affect the probability of treatment (i.e. new school opening there vs in a control area) to explain away the results we find?* --- # Broader picture: What happens to districts? .pull-left[ - Compare districts with a new school between 2012 and 2016 vs districts with no new schools. - ASCM for different characteristics, adjusting for other baseline covariates (e.g. number of schools, enrollment, %FRPL, % race/ethnicity) ] .pull-right[ .center[ ] ] --- # Districts with new schools increase gap in Math <br> .center[ ] --- # ... no significant change in Reading <br> .center[ ] --- # What happens within districts? .pull-left[ - Identify **.darkorange[6 new high schools]** between 2012-2016 that change AZ. - Compare attendance zones *within districts* to create a counterfactual. ] .pull-right[ .center[ ] ] --- # What happens within districts? .pull-left[ - Identify **.darkorange[6 new high schools]** between 2012-2016 that change AZ. - Compare attendance zones *within districts* to create a counterfactual. <br> - **.darkorange[Important caveat]**: - Limited sample size `\(\rightarrow\)` under-powered. - Trends are suggestive. ] .pull-right[ .center[ ] ] --- # Zoned-in Areas: How do neighborhoods change? - No major changes in % white population (left) or % African American population (right) .pull-left[ .center[ ] ] .pull-right[ .center[ ] ] --- # Zoned-in Areas: How do neighborhoods change? - Suggestive increase in % of college educated population .center[ ] --- # Zoned-out Areas: How are schools affected? - **.darkorange[Demographics]**: Decreasing trend in white students enrollment (left) vs increasing trend in African American enrollment (right). .pull-left[ .center[ ] ] .pull-right[ .center[ ] ] --- # Zoned-out Areas: How are schools affected? - **.darkorange[Performance]**: Decreasing trend % of proficiency (sharper for African American students). .center[ ] --- # Zoned-out Areas: What about neighborhoods? - Increasing trend in % of African-American population (left) and decreasing trend in % college-educated population (right). .pull-left[ .center[ ] ] .pull-right[ .center[ ] ] --- # Wrapping up - Attendance zones have a huge role in **.darkorange[shaping neighborhoods]** and nearby areas. -- <br> <br> - Importance of understanding the effects of **.darkorange[new schools]** and their location and boundaries. -- <br> <br> - Effects of housing prices? Long-run outcomes? -- <br> <br> - **.darkorange[Next steps]**: - Include other states (e.g. CA) - Analyze housing prices over time. - Heterogeneity in effect for gentrified neighborhoods? --- class: inverse, center, middle <br> <br> <br> <h1 class="title-own">Zoned-In and Zoned-Out:<br/>Changes in School Attendance Zones over Time</h1> <br> <br> .small[Magdalena Bennett <br>*McCombs School of Business, The University of Texas at Austin* ] <br> .small[AEFP Conference<br>March 24th, 2023]