Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Page Not Found

Archive Layout with Content

Posts by Category

Posts by Collection

CV

CV

Markdown

Page not in menu

Page Archive

Portfolio

Research

Sitemap

Posts by Tags

Talk map

Talks and presentations

Teaching

Terms and Privacy Policy

Blog posts

Jupyter notebook markdown generator

Posts

AI-Predicted Variables in Statistical Inference

19 minute read

Published:

Predictions from AI models (text, image or segmentation classifiers) are increasingly used as data in the social sciences. This note reviews why they bias estimates and narrow standard errors, the six types of measurement error they can carry, and three methods to correct for them: prediction-powered inference, design-based supervised learning and multiple imputation.

portfolio

Portfolio item number 1

Short description of portfolio item number 1

Portfolio item number 2

Short description of portfolio item number 2

publications

The Heterogeneous Effect of Dropping Out for Higher Education Students: The French Case

Working paper · SSRN, 2024

French dropout policies focus on universities rather than vocational degrees. Using a causal random forest with distance to the nearest higher-education institution as an instrument, I find that 2SLS underestimates the employment and wage costs of dropping out, and that vocational dropouts face larger wage penalties than university dropouts.

Recommended citation: Tissandier Gaspard. (2024). The Heterogeneous Effect of Dropping Out for Higher Education Students: The French Case
Download Paper

Resources Misallocation and Predictive Policing: A Data Generation Process

Work in progress, 2024

Place-based predictive policing learns from crime data that partly depend on where officers were sent: crimes observed by the police rise with patrol presence, while complaints are, in principle, independent of it. This paper models the feedback loop between police allocation, observed crime, and complaints, and derives the conditions under which an area with less crime ends up observing more of it, and receiving more patrols. Depending on the initial distribution of crime and reporting rates, the system can be “designed to fail”, whether allocation relies on simple or machine-learning predictions.

Recommended citation: Tissandier Gaspard. (2024). Resources Misallocation and Predictive Policing: A Data Generation Process

PAVED With Good Intentions? An Evaluation of a French Police Predictive Policing System

Conditionally accepted · International Review of Law and Economics, 2025

From 2017 to 2019, the French Gendarmerie tested PAVED, a system predicting burglaries and vehicle thefts to guide patrols, in 11 of its 101 departments. Comparing these departments with several counterfactual groups, and with varying specifications, we find no detectable effect on burglaries and no robust effect on vehicle thefts.

Recommended citation: Lecorps, Yann and Tissandier Gaspard. (2025). PAVED With Good Intentions? An Evaluation of a French Police Predictive Policing System
Download Paper

Multilevel Modeling

Published · Oxford Research Encyclopedia of Criminology and Criminal Justice, 2025

An encyclopedia entry introducing multilevel models for criminology and criminal justice research. It explains why nested data, such as places within neighborhoods, require these models, and presents the main types: random intercept, random coefficient, fixed effect and growth models.

Recommended citation: Dugato, Marco and Tissandier, Gaspard. (2025). Multilevel Modeling. Oxford Research Encyclopedia of Criminology and Criminal Justice. doi:10.1093/acrefore/9780190264079.013.ORE_CRI-00816

From Data to Action: Benchmarking the Accuracy-Interpretability Trade-off of Machine Learning Algorithms for Crime Analysis

Working paper, 2026

Are complex machine-learning models worth their loss of interpretability for crime prediction? Benchmarking traditional, machine-learning and interpretable models on the prediction of five crime types in Newark, NJ, we find that interpretable models with only ten components perform nearly as well as the best boosting algorithms. We also document the “top-k trap”, a distortion affecting threshold-dependent accuracy metrics.

Recommended citation: Tissandier Gaspard and Gimenez-Santana Alejandro (2026). From Data to Action: Benchmarking the Accuracy-Interpretability Trade-off of Machine Learning Algorithms for Crime Analysis
Download Paper

Widening the Aperture of the Opportunity Lens: Engaging Local Stakeholders in the Interpretation of Crime Concentrations

Published · Crime Science, 2026

A content analysis of 212 Crime Science articles (2012–2025) shows that only 1.9% involved residents of the areas under study. Three case studies from the Newark Public Safety Collaborative show how community input changed the prevention strategies that data analysis alone would have suggested.

Recommended citation: La Vigne, N., Giménez-Santana, A., Tissandier, G., & Santos, A. (2026). Widening the aperture of the opportunity lens: engaging local stakeholders in the interpretation of crime concentrations. Crime Science. doi:10.1186/s40163-026-00285-x

Evaluating the Effect of Enhanced LED Street Lighting on Nighttime Violent Crime in Newark, NJ

Published · Journal of Quantitative Criminology, 2026

Between 2019 and 2021, Newark replaced about 1,500 high-pressure sodium streetlights with LED fixtures. Using a staggered difference-in-differences design, we find that nighttime outdoor violent crime fell by about 46% in the two quarters after replacement, mainly through fewer aggravated assaults and robberies. The effect then fades, suggesting an adaptation to such environmental changes.

Recommended citation: Gimenez-Santana, Alejandro; Tissandier, Gaspard; Zlaoui, Khalil; Santos, Adriana; Caplan, Joel; Kennedy, Les. (2026). Evaluating the Effect of Enhanced LED Street Lighting on Nighttime Violent Crime in Newark, NJ. Journal of Quantitative Criminology. doi:10.1007/s10940-026-09690-6
Download Paper | Download Slides

Do Public Art Murals Reduce Street Crime? Assessing the Hyperlocal Impact of Murals on Urban Crime

Work in progress, 2026

Using 130 murals installed in Newark between 2018 and 2025, we estimate the marginal effect of each additional mural on nearby crime with a difference-in-differences estimator designed for discrete, staggered treatments. Preliminary results show no effect on the mural’s own street segment, but each new mural is associated with fewer robberies and burglaries when adjacent segments are included.

Download Slides

talks

Talk 1 on Relevant Topic in Your Field

Published:

This is a description of your talk, which is a markdown file that can be all markdown-ified like any other post. Yay markdown!

Conference Proceeding talk 3 on Relevant Topic in Your Field

Published:

This is a description of your conference proceedings talk, note the different field in type. You can put anything in this field.

teaching

Artificial Intelligence in Criminal Justice

Undergraduate course, Rutgers University–Newark, School of Criminal Justice, 2024

An introduction to how Artificial Intelligence is used in the criminal justice system, from predictive policing to pre-trial risk assessment, combining the technical foundations of machine learning with case studies and a critical look at fairness. available on GitHub.