lacea2024.knit

class: center, middle
# Schools unlike prisions: the effects of extending the school-day on crime.
### Francisco Cabrera-Hernandez `$^a$`  Bárbara Zárate-Tenorio `$^b$` 
#### `$^a$` CIDE `$^b$` OsloMet 
November, 2024
###LACEA Meeting, Montevideo 2024

---

## Motivation
- In 2021, **one in four adults** experienced at least one crime in Latin America (LAPOP, 2021).

- Social, public, and private costs of crime estimated in the region at around **3 percent of GDP** (Jaitman et al., 2017).

- Policy options range from coercive strategies to inequality reduction, and **human capital formation**.

- Education may impact through: **incapacitation**, higher opportunity costs of crime, and **better social bonds with peers/teachers**.

---
## This Paper

- **What are the effects of longer school days (grades 1-9) on minor and more serious crimes?**

- We use the **Mexican Full-Time Schools (FTS) program** and data on crime at the municipality level.

- We disentangle the effects of FTS from the **Secure Schools (SS) program**, a simultaneous intervention.

- We exploit the staggered implementation of FTS and SS (follower) interventions to provide new evidence on schooling and crime.

- We provide evidence for different crimes and in richer/poorer contexts.

---
## What do we learn?

###We provide short-term causal evidence on:

- Minor crimes **reduction in common robbery of 11% from the 2010/11 base**

- Results are **driven by middle-schools** (age 12 to 14) in poorer municipalities.

- No effects on other minor crimes nor on serious felonies.

- Effects **explained by the FTS program alone** and not by the SS nor their combination.

---

## What do we know on FTS?

- **Positive short-run academic achievement** *(Padilla-Romo, 2021; Cabrera-Hernandez, 2020; Aguero et al., 2021)*

- Effects on mothers and grandmothers' labour supply *(Padilla & Cabrera, 2019; Cabrera & Padilla 2020; Garganta and Zentner, 2020; Berthelon et. al. 2020)*

- **Reduces child labour** *(Kozhaya and Martinez-Flores, 2022)*

- Changes Mothers' fertility decisions, and **reduces teenage pregnancy** *(Berthelon & Kruger, 2011; Cabrera & Marquez, 2023)*

- **In the long-run FTS delays childbearing, increase educational attainment, improves test scores, preferences and non-cognitive abilities** *(Dominguez and Ruffini, 2021; Cabrera et. al. 2023)*.

---

## On Education and Crime?

- **Studies on the effects of education on crime are common in richer countries** *(e.g., Fischer and Argyle, 2018; Anderson, 2014; Gerlinger and Hipp, 2020; Akee et al., 2014)*

- Higher secondary school enrollment in Mexico relates to lower crime rates (1993-2007) *(Gleditsch et. al., 2022)*

- Berthelon & Kruger (2011) find that extending the school day in high school reduces youth crime in Chile.

- **We contribute with evidence on the incapacitation effect of schooling, for a middle-income country and on the externalities of FTS.**

---
## The Full-Time Schools programme

-   For grades 1 to 9, it extends school-day **from 4.5 hours to 8.  Aiming to improve learning**

-   Increased inputs, money, staff, teachers, and infrastructure (e.g. kitchens).

-   It started in 2007/8 in 500 schools, it reached **25,000 by 2018 or 60% of eligible.**

-   80% of municipalities are treated.

---
## FTS Time-Space Variation

---
## The Secure Schools programme

-  Aimed to **improve schools’ learning environment, security inside schools, and relations between children**.

- Provided guidelines and one course to staff focused on dealing with violence in schools, bullying and drug consumption.

- It started in 2007/8 and operated up to 2014/15 **reaching 13% of all schools.**

-  Schools in the top 150 municipalities with higher levels of crime (but **reached 613, or 25%**).

- Explained because SS schools followed FTS schools from 2009/10

---
## Secure Schools Space Variation (by FTS presence)

---
##Crime Context among the youth in Mexico

- Teenagers between 12 and 17 years represented *39% of the total convictions* (ENASJUP, 2022)

- 41% of convictions happen while enrolled in school.

- Among those younger than 22 years old, *robbery accounts for 21% of convictions*, followed by homicide (15%) and rape (9%)

---
## Data

-   We use administrative data on FTS and SS (SEP), and different types of crime (alleged offenses in courts) (SESNSP).

- We combine this at the municipal level  between 2010/11 to 2016/17.

- Data on crime  details **seven different categories of most common offenses** (97% of total crimes).

---
## Crimes

---
## Identification

-   Staggered implementation across municipalities.

`$$Y_{mt}=\nu_t+\theta_m+\beta_{1} FTS_{mt}+ \beta_{2} SS_{mt}+u_{mt}$$`

-   `$Y_{mt}$` is crimes per 100,000 of the population in municipality `$m$` in academic year `$t$`.

-   `$\nu_t$` are time fixed effects; `$\theta_{m}$` are municipality fixed effects;

-   `$FTS_{mt}$` is a variable equals to one if municipality `$m$` was ever exposed to full-time schooling;

-   `$SS_{mt}$` is a variable equal to one if municipality `$m$` was exposed to the secure schools program.

---
## Dynamic Effects

-   We allow our model to capture dynamic effects:

`$$Y_{mt}=\nu_t+\theta_{m}+\sum_{k=-4,k\neq -1}\delta_kFTS_{m,t-k}+\beta SS_{mt}+\lambda_{st}+u_{mt}$$`

-   `$FTS_{m,t-k}$`, indicates municipality degree of exposure, in academic year `$t$` to full-time schooling in `$k+1$` years.

-  `$k\geq 0$` is equal to one `$k+1$` years after opening its first FTS.

-  `$\delta_k$` is the intent-to-treat effect of the FTS program on crime under standard parallel trend assumptions.

- We add `$\delta_{st}$` trend as crime policies may be state dependent.

---
## Dynamic Effects

-   **We compute estimators robust to dynamic and heterogeneous effects  developed by Chaisemartin and D'Haultfoeuille (2020).**

-  Adequate as 48% of municipalities are ”always treated” by 2010/11.

- Goodman-Bacon (2021) decomposition: higher weight in always-treated versus timing groups.

- Never treated group is small. E.g. IW estimator (Sun and Abraham, 2020) rely on these.

---

## Average Results

---

## Average Results

---

## Dynamic Effects

---

## Dynamic Effects

---

## Average Results by SES

---

## Dynamic Effects by Education Level

---

##Correlated treatments

- **C&D (2022) discuss why under multiple correlated treatments, staggered estimators may be contaminated.**

- We estimate ”pure” effects of the FTS program in municipalities without SS schools (~60% of FTS-treated)

- **To compute the "pure" effect of SS we compare municipalities treated with SS in `$t$`:**

-  Where treated and control municipalities started with FTS in `$t-k-1$`.

-  We add the non-parametric trend of `$Y_{mt}$` in control and treated groups so that they depart from the same trend in `$t−1$`.

---
## Secure Schools follow FTS implementation.

---
## Average "Pure" effects

---

## Dynamic "Pure" effects

---
## Conclusions

-   FTS reduce common robbery in 11% of the base year.

-   Effects concentrate on low-SES municipalities and middle schools.

-   We offer evidence of the incapacitation channel of schooling (FTS)

-   *Social bonding* may not have a strong effect (SS)

-   Other long-term effects in the FTS literature may suggest that it could raise opportunity costs of crime.