Flávio Riva

Hi, my name is Flávio Riva. Welcome to my website! My main interests are:

using causal inference tools to study human capital formation policies;
learning how policy-makers extract meaning from diagrams—such as problem trees, theories of change, and monitoring systems—that represent the policies they implement.

Doctoral Degree, Public Administration and Government — EAESP-FGV, 2022

B.A. and an M.A. in Economics — EESP-FGV, 2016

B.A. in Sociology/Anthropology — FFLCH-USP 2013

Here are my Google Scholar and LinkedIn. You can reach me at flavio.riva@imdsbrasil.org or flaviorussoriva@gmail.com.

Academic research

Artificial Intelligence, Writing Achievement, and Pedagogy: Evidence from High Schools in Brazil

with Bruno Ferman and Lycia Lima

October 2025, Submitted

Other materials: J-PAL Evidence to Policy | J-PAL Summary | AEA RCT Registry

Media Coverage: Nexo | Valor Econômico | O Globo | Porvir

Educational technologies with different mixes of inputs from artificial and human intelligence had similar effects on writing scores in a large-scale field experiment. Teachers reacted to the new technologies by providing more of the inputs they use to foster the development of complex skills: individualized support.

Jovem Aprendiz Program, Employability, and Formal Labor Income: An Analysis of the First Generations Exposed to the Apprenticeship Law of 2000

with Raphael Corbi, Sergio Guimarães, Matheus Leal and Paulo Tafner

Imds Technical Note, 2025, non peer reviewed

Local labor markets in Brazil were differentially affected by a law that incentivized the adoption of formal apprenticeship contracts for teenagers. Leveraging age-eligibility rules in a difference-in-differences framework, we find that apprenticeships increase employment and earnings among individuals aged 25 to 29.

Probability of Causation with Sample Selection: A Reanalysis of the Impacts of Jóvenes en Acción on Formality

with Vitor Possebom

Journal of Business & Economic Statistics, 2024

Other materials: Arxiv ungated version.

The probability of causation is partially identified for individuals who are always observed regardless of treatment status, and sharp bounds on this parameter can be built under three increasingly restrictive sets of assumptions.

Technical Education, Non-Cognitive Skills and Labor Market Outcomes: Experimental Evidence from Brazil

with Juliana Camargo, Lycia Lima, and Andre Portela Souza

IZA Journal of Labor Economics, 2021

In progress

Parenting Interventions and Learning in Middle Childhood: Evidence from Brazil

with Carolina Melo

Effects of Childhood Access to Primary Health Care on Learning: Evidence from Brazil's Family Health Program

with Leticia Nunes and Pedro H. Chaves Maia

Formal Dual Apprenticeships and Teenagers' School-to-Work Transition: Evidence from Brazil's Jovem Aprendiz

with Raphael Corbi and Matheus Leal

Delivering Health at School and Learning: Evidence from Public Schools in Brazil

with Lia Braga and Rudi Rocha