Neighborhood Intervention, Crime and School Achievement

class: center, middle, inverse, title-slide

# Neighborhood Intervention, Crime and School Achievement
## Part I: Urban Renewal and Violent Crimes
### Marcelino Guerra
### March, 2022

---

# Recap

.pull-left[

* In the middle of 2014, The City of Fortaleza began an ongoing urban renewal project called "Areninhas." The intervention consists of synthetic football turf, sometimes with a playground and an outdoor gym. Besides, there is a substantial increase in street lighting

* The project targets highly vulnerable communities, and the City Hall aims to offer the population an urban amenity that enhances the well-being of residents, promoting physical activity and a sense of community

* The local government works together with residents to supervise and conserve the public equipment. Regularly, free football lessons are offered to the young locals, and amateur football championships are held on weekends

]

.pull-right[

As of now, there are 102 areninhas in Fortaleza and more than 150 in the rest of Ceara State. **The project aims to investigate the effects of this local amenity on crime and student achievement**.

![](figs/final2.png) 
.small[**Source: Fortaleza City Hall website and *O Povo* newspaper.**]

]

---
# Data I - Murders in Fortaleza

.pull-left[

* The data consists of 13,779 homicide occurrences recorded between **January 1st, 2011, and January 30th, 2020**. The final number of murder counts (with complete address and latitude/longitude) is 13,370

* Most occurrences have information about gender, age, schooling, criminal record, etc. Around 60% of the data also tells whether the murder happened indoors or on the streets

* The statistical group of the Secretariat of Public Security of Ceará (SSPDS-CE) is the primary data source. I got coordinates of occurrences with precise addresses using the Google Maps API

* To avoid too many zeros, homicides are aggregated to quarters and at the census tract level (this spatial aggregation might change). Murders are divided by the total population in each census tract. Hence, the outcome is homicide per 1,000 people.

]

.pull-right[

]

---
# Data II - Staggered Rollout of Areninhas

.pull-left[

<table class="table table" style="width: auto !important; margin-left: auto; margin-right: auto; margin-left: auto; margin-right: auto;">
 <thead>
  <tr>
   <th style="text-align:left;"> Year </th>
   <th style="text-align:right;"> Areninhas </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:left;"> 2014 </td>
   <td style="text-align:right;"> 1 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2015 </td>
   <td style="text-align:right;"> 2 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2016 </td>
   <td style="text-align:right;"> 18 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2017 </td>
   <td style="text-align:right;"> 2 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2018 </td>
   <td style="text-align:right;"> 2 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2019 </td>
   <td style="text-align:right;"> 17 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2020 </td>
   <td style="text-align:right;"> 31 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2021 </td>
   <td style="text-align:right;"> 15 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> 2022 </td>
   <td style="text-align:right;"> 14 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> Total </td>
   <td style="text-align:right;"> 102 </td>
  </tr>
</tbody>
<tfoot>
<tr>
<td style = 'padding: 0; border:0;' colspan='100%'><sup></sup> Source: Fortaleza's City Hall website. Note: There are two areninhas inside the same census tract, so the total number of city blocks is 101.</td>
</tr>
</tfoot>
</table>

]

.pull-right[

]

---
# Identification I: Treated vs Never Treated

.pull-left2[

* To construct the counterfactual scenario - i.e., what would have happened to the treated areas without the urban renewal -, the choice of the control group is crucial

* The Local Government did not choose the areas under intervention randomly: neighborhoods with very-low/low HDI were targeted

* The table shows that the treated areas and the rest of the city are fundamentally different in observables (and, most likely, unobservables)

]

.pull-right2[

<table>
 <thead>
   <tr>
   <td style="text-align:center;" colspan="5"> <b> t-test mean equality: Treated vs Never Treated </b> </td>
   </tr>
  <tr>
   <th style="text-align:left;">   </th>
   <th style="text-align:center;"> Treated </th>
   <th style="text-align:center;"> Never treated </th>
   <th style="text-align:center;"> Difference </th>

</tr>
 </thead>
<tbody>

<tr>
   <td style="text-align:left;"> Average Income (R$) </td>
   <td style="text-align:center;"> 1,907.982  </td>
   <td style="text-align:center;"> 2,489.519  </td>
   <td style="text-align:center;"> -581.53***  </td>
  
  </tr>
  <tr>
   <td style="text-align:left;"> More than 10 Minimum Wages   </td>
   <td style="text-align:center;"> 0.92% </td>
   <td style="text-align:center;"> 2.19% </td>
   <td style="text-align:center;"> -1.27%*** </td>
 
  </tr>
  
  <tr>
   <td style="text-align:left;"> Up to 1 Minimum Wage  </td>
   <td style="text-align:center;"> 69.05% </td>
   <td style="text-align:center;"> 62.46%  </td>
   <td style="text-align:center;"> 6.58%*** </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> Population density </td>
   <td style="text-align:center;"> 11,097.98  </td>
   <td style="text-align:center;"> 18,796.28 </td>
   <td style="text-align:center;"> -7,699*** </td>
  </tr>
  <tr>
   <td style="text-align:left;"> Young 12-21 </td>
   <td style="text-align:center;"> 19.10% </td>
   <td style="text-align:center;"> 17.88% </td>
   <td style="text-align:center;"> 1.21%*** </td>
  </tr>
  <tr>
   <td style="text-align:left;"> Black and Brown 15-29 </td>
   <td style="text-align:center;"> 9.41% </td>
   <td style="text-align:center;"> 8.77% </td>
   <td style="text-align:center;"> 0.64%** </td>
  </tr>
  <tr>
   <td style="text-align:left;"> No pavement </td>
   <td style="text-align:center;"> 13.47% </td>
   <td style="text-align:center;"> 9.02% </td>
   <td style="text-align:center;"> 4.45%** </td>

</tr>
  
  <tr>
   <td style="text-align:left;"> No Street Lighting </td>
   <td style="text-align:center;"> 4.09% </td>
   <td style="text-align:center;"> 1.98% </td>
   <td style="text-align:center;"> 2.11%** </td>

</tr>

<tr>
   <td style="text-align:left;"> <b>Observations</b> </td>
   <td style="text-align:center;"> 101 </td>
   <td style="text-align:center;"> 2,919 </td>
   <td style="text-align:left;">  </td>

</tr>
    <tfoot><tr><td colspan="5"> <b>Note:</b> ***: 0.01, **: 0.05, *: 0.1 </td></tr></tfoot>
</tbody>
</table>

]

---

# Identification II

---

# Identification III: Treated before vs Treated after 2019

.pull-left2[

* The table shows that future treated regions (census tracts) are very similar to the ones treated earlier (until December 2018) on observables

* I conjecture that census tracts treated before 2019 and after 2019 are similar in observables and unobservables, and this choice of control group produces an apples-to-apples comparison

* The main identifying assumption is that, in the absence of the urban policy, crime in treated and control groups would have evolved in a parallel fashion

]

.pull-right2[

<table>
 <thead>
   <tr>
   <td style="text-align:center;" colspan="5"> <b> t-test mean equality: Treated (2014-2018) vs Future Treated </b> </td>
   </tr>
  <tr>
   <th style="text-align:left;">   </th>
   <th style="text-align:center;"> Treated before 2019 </th>
   <th style="text-align:center;"> Treated after 2019 </th>
   <th style="text-align:center;"> Difference </th>

</tr>
 </thead>
<tbody>

<tr>
   <td style="text-align:left;"> Average Income (R$) </td>
   <td style="text-align:center;"> 1,764.08  </td>
   <td style="text-align:center;"> 1,821.59  </td>
   <td style="text-align:center;"> -57.51 </td>
  
  </tr>
  <tr>
   <td style="text-align:left;"> More than 10 Minimum Wages   </td>
   <td style="text-align:center;"> 0.62% </td>
   <td style="text-align:center;"> 0.84% </td>
   <td style="text-align:center;"> -0.22% </td>
 
  </tr>
  
  <tr>
   <td style="text-align:left;"> Up to 1 Minimum Wage  </td>
   <td style="text-align:center;"> 70.23% </td>
   <td style="text-align:center;"> 70.80%  </td>
   <td style="text-align:center;"> -0.57% </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> Population density </td>
   <td style="text-align:center;"> 13,875.05  </td>
   <td style="text-align:center;"> 9,702.87 </td>
   <td style="text-align:center;"> 4,172.88 </td>
  </tr>
  <tr>
   <td style="text-align:left;"> Young 12-21 </td>
   <td style="text-align:center;"> 18.79% </td>
   <td style="text-align:center;"> 19.60% </td>
   <td style="text-align:center;"> -0.81% </td>
  </tr>
    <tr>
   <td style="text-align:left;"> Black and Brown 15-29 </td>
   <td style="text-align:center;"> 9.70% </td>
   <td style="text-align:center;"> 9.72% </td>
   <td style="text-align:center;"> 0.02% </td>
  </tr>
  <tr>
   <td style="text-align:left;"> No pavement </td>
   <td style="text-align:center;"> 10.71% </td>
   <td style="text-align:center;"> 18.72% </td>
   <td style="text-align:center;"> -8.01%* </td>

</tr>
  
  <tr>
   <td style="text-align:left;"> No Street Lighting </td>
   <td style="text-align:center;"> 3.76% </td>
   <td style="text-align:center;"> 4.49% </td>
   <td style="text-align:center;"> -0.73% </td>

</tr>

<tr>
   <td style="text-align:left;"> <b>Observations</b> </td>
   <td style="text-align:center;"> 25 </td>
   <td style="text-align:center;"> 54 </td>
   <td style="text-align:left;">  </td>

</tr>
    <tfoot><tr><td colspan="5"> <b>Note:</b> ***: 0.01, **: 0.05, *: 0.1 </td></tr></tfoot>
</tbody>
</table>

]

---

# Estimation Method

**TWFE**

`$$\text{Murder Rate}_{iq}=\lambda_{q}+\gamma_{i}+\beta Open_{iq}+\varepsilon_{iq}$$`

where `$\lambda_{q}$` is the time fixed effects (quarter-year), and `$\gamma_{i}$` refers to census tract fixed effects. `$\beta$` is the difference-in-differences estimate that captures the causal effect of interest: the extent to which the treated and non-treated city blocks differ in their homicide rates after the urban renewal policy

**TWFE with time-of-day heterogeneity**
  
To check whether this urban intervention have different effects within hours of day, I consider the following:

`$$\text{Murder Rate}_{iqt}=\lambda_{qt}+\gamma_{i}+\beta_{1}Open:Morning_{iqt}+\beta_{2}Open:Afternoon_{iqt}+\beta_{3}Open:Night_{iqt}+\varepsilon_{iqt}$$`

where `$\lambda_{qt}$` represents quarter-year-time-of-day fixed effects. The morning covers the hours from 8 to 11:59 am, afternoon 12:00 pm until 5:59 pm, and night 6:00 pm to 8:59 pm. The omitted category is 'closed,' covering the hours from 9:00 pm to 7:59 am.

---

# Direct Effect I

.pull-left2[

* The first column shows a naive comparison of murder rates between all treated census tracts and the rest of the city. The second column still uses full sample, but exploits the staggered rollout of the urban policy. This difference-in-differences estimate is negative but not statistically significant

* The other three columns use the restricted sample (treated before 2019 *vs* treated after 2019). Results are more precise and statistically significant: the urban renewal policy caused a decrease in murder rates around 47-61%

]

.pull-right2[
<font size="3.5" face="Helvetica" >
<table style="table-layout: fixed;">
 <thead>
   <tr>
   <td style="text-align:center;" colspan="6"> <b> Dependent variable: Murder Rate (per 1,000)/Murder counts (Poisson) </b> </td>
   </tr>
  <tr>
   <th style="text-align:left;">   </th>
   <th style="text-align:center;" colspan="2"> Full Sample  </th>
   <th style="text-align:center;"> Restricted Sample  </th>
   <th style="text-align:center;"> Restricted Sample (Poisson)  </th>
   <th style="text-align:center;"> Restricted Sample (w/o outlier)  </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:left;"> <b>Open </b> </td>
   <td style="text-align:center;">  </td>
   <td style="text-align:center;"> -0.204   </td>
   <td style="text-align:center;"> -0.166** </td>
   <td style="text-align:center;"> -0.513**  </td>
   <td style="text-align:center;"> -0.115**  </td>
  </tr>
  
  <tr>
   <td style="text-align:left;">           </td>
   <td style="text-align:center;"> </td>
   <td style="text-align:center;"> (0.224)</td>
   <td style="text-align:center;">(0.067) </td>
   <td style="text-align:center;"> (0.257) </td>
   <td style="text-align:center;"> (0.044) </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> <b>Areninha </b> </td>
   <td style="text-align:center;"> 0.039538 </td>
   <td style="text-align:center;">   </td>
   <td style="text-align:center;">   </td>
   <td style="text-align:center;">  </td>
   <td style="text-align:center;">  </td>

</tr>
  <tr>
   <td style="text-align:left;">  </td>
   <td style="text-align:center;"> (0.0346)</td>
   <td style="text-align:center;"> </td>
   <td style="text-align:center;">  </td>
   <td style="text-align:center;">  </td>
   <td style="text-align:center;">  </td>
  </tr>
  
  <tr>
   <td style="text-align:left;" > <b>Fixed-Effects:</b>  </td>
   <td style="text-align:left;" > None </td>
   <td style="text-align:left;" > Census tract and Quarter-Year </td>
   <td style="text-align:left;" > Census tract and Quarter-Year </td>
   <td style="text-align:left;" > Census tract and Quarter-Year </td>
   <td style="text-align:left;" > Census tract and Quarter-Year </td>
  </tr>
    <tr>
   <td style="text-align:center;" colspan="6">  </td>
  </tr>
  
  <tr>
  
   <td style="text-align:left;"> <b>Mean of Outcome</b> </td>
   <td style="text-align:center;" colspan="2"> 0.1736  </td>
   <td style="text-align:center;" > 0.2719 </td>
    <td style="text-align:center;" >  </td>
     <td style="text-align:center;" > 0.2427 </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> <b>Observations</b> </td>
   <td style="text-align:center;" colspan="2"> 92,938   </td>
   <td style="text-align:center;"> 2,449 </td>
   <td style="text-align:center;"> 2,108  </td>
   <td style="text-align:center;"> 2,418   </td>

</tr>
   <tfoot><tr><td colspan="6">Standard-errors in parentheses and clustered at the Neighborhood level. The sample considers homicides occurrences between 8 am to 9 pm. The outlier is a treated census tract with only 38 people, hence the really high murder rate.  </td></tr></tfoot>
    <tfoot><tr><td colspan="6"> <b>Note:</b> ***: 0.01, **: 0.05, *: 0.1 </td></tr></tfoot>
</tbody>
</table>
]

---
# Direct Effect II

---

# Time-of-day Heterogeneity I

.pull-left2[

* Time-of-day heterogeneity analysis shows that the decrease in murder in treated census tracts is concentrated between 12 pm and 6 pm, supporting the "more eyes on the streets" thesis

* Pupils that study during the afternoon and commute from home to school around 1:00 pm and from school to home between 5:00-5:30 pm would benefit the most

* Although negative, estimates related to crime occurrences in the morning and night are not statistically significant. I will probably get more precise results using all violent crimes instead of only murder

]

.pull-right2[
<font size="3.5" face="Helvetica" >
<table style="table-layout: fixed;">
 <thead>
   <tr>
   <td style="text-align:center;" colspan="4"> <b> Dependent variable: Murder Rate (per 1,000)/Murder counts (Poisson) </b> </td>
   </tr>
  <tr>
   <th style="text-align:left;">   </th>
   <th style="text-align:center;"> Restricted Sample  </th>
   <th style="text-align:center;"> Restricted Sample (Poisson)  </th>
   <th style="text-align:center;"> Restricted Sample (w/o outlier)  </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:left;"> <b>Open:Morning </b> </td>
   <td style="text-align:center;"> -0.103   </td>
   <td style="text-align:center;"> -1.242**  </td>
   <td style="text-align:center;"> -0.034  </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> <b> </b> </td>
   <td style="text-align:center;"> (0.074)   </td>
   <td style="text-align:center;"> (0.611) </td>
   <td style="text-align:center;"> (0.028) </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> <b> Open:Afternoon </b>  </td>
   <td style="text-align:center;"> -0.150*</td>
   <td style="text-align:center;"> -0.952*** </td>
   <td style="text-align:center;"> -0.080** </td>

</tr>
  
  <tr>
   <td style="text-align:left;"> <b> </b> </td>
   <td style="text-align:center;"> (0.080)   </td>
   <td style="text-align:center;"> (0.337) </td>
   <td style="text-align:center;"> (0.039) </td>
  </tr>
  
  <tr>
   <td style="text-align:left;"> <b>Open:Night </b> </td>
   <td style="text-align:center;"> -0.097 </td>
   <td style="text-align:center;"> -0.536  </td>
   <td style="text-align:center;"> -0.028  </td>

</tr>

<tr>
   <td style="text-align:left;" > <b>Fixed-Effects:</b>  </td>
   <td style="text-align:left;" > Census tract and Quarter-Year-Time-of-day </td>
   <td style="text-align:left;" > Census tract and Quarter-Year-Time-of-day </td>
   <td style="text-align:left;" > Census tract and Quarter-Year-Time-of-day</td>
  </tr>

<tr>
  
   <td style="text-align:left;"> <b>Mean of Outcome</b> </td>
   <td style="text-align:center;" > 0.066  </td>
   <td style="text-align:center;" >        </td>
    <td style="text-align:center;" > 0.061 </td>

</tr>
  
  <tr>
   <td style="text-align:left;"> <b>Observations</b> </td>
   <td style="text-align:center;"> 12,524 </td>
   <td style="text-align:center;"> 11,232  </td>
   <td style="text-align:center;"> 12,400   </td>

</tr>
   <tfoot><tr><td colspan="4">Standard-errors in parentheses and clustered at the Neighborhood level. The sample considers all homicides occurrences. The omitted category for time-of-day is when the equipment is closed: between 9 pm and 8 am. </td></tr></tfoot>
    <tfoot><tr><td colspan="4"> <b>Note:</b> ***: 0.01, **: 0.05, *: 0.1 </td></tr></tfoot>
</tbody>
</table>

]

---
# Time-of-day Heterogeneity II

---
# Spatial Displacement: Census tracts vs Grids vs Street-level