๐ ๐๐ฑ๐ญ๐๐ง๐๐๐ ๐๐๐๐๐ฅ๐ข๐ง๐: ๐๐ ๐๐๐ซ๐๐ก ๐๐๐๐
PhD students, do not miss this opportunity to attend the ENS Lyon Summer School in Empirical Research Methods.
@ensdelyon.bsky.social @vincentbagilet.bsky.social @cedricchambru.bsky.social @kenza-elass.bsky.social @elisamougin.bsky.social
#Economics
It provides ready to use tools to analyze and visualize these weights and observations contributing to identification.
The package is still under development, please reach out if you encounter issues using it, Iโll do my best to improve it.
The associated paper: ssrn.com/abstract=6071752
ididvar computes identifying variation weights as the leverage of each observation but after having partialled out the controls (including those identification-related) and FE.
Some observations do not contribute to identification and dropping them would not change the estimate.
Visualization of the set of observations contributing to identification in an example regression
Visualization of the set of observations contributing to identification in an example regression
Visualization of the distribution of identifying variation weights in an example regression
In my updated paper, I introduce an R package, ididvar, that provides tools to easily identify the identifying variation in a regression: vincentbagilet.github.io/ididvar/โจโจ
More information can be found on the package and project websites: vincentbagilet.github.io/causal_exaggeration/ididvar.html
A systematic reporting of power calculations in observational studies would help gauge the risk of falling into this low power trap.
The trade-off is mediated by the variation used for identification. I provide tools to explore it, including an R package, ididvar: vincentbagilet.github.io/ididvar/
If fixed effects or controls absorb a lot of the variation in X (more than in Y), the resulting identifying variation might be limited, leading to imprecision and exaggeration.
When leveraging exogenous shocks, the variation sometimes only comes from a limited number of treated observations. Power can thus be low and estimates inflated.
IV only uses part of the variation in the treatment, the portion explained by the instrument. When the "strength" of the instrument is low, the IV is imprecise and can induce exaggeration.
A "naive" OLS can, on average, produce significant estimates that are closer to the true effect than the IV.
RDD discards variation by only considering observations within the bandwidth. It decreases the effective sample size.
On average significant estimates may never get close to the true effect.
Consequential exaggeration has been documented in econ (often by a factor 2 to 4). Its two ingredients, present in the lit, are selection on significance and lack of statistical power.
I highlight a mechanism explaining exaggeration despite our extensive use of convincing causal inference methods.
Causal id strat reduce power because they throw out part of the variation. Using too little variation can create an exaggeration bias, even if this variation is exogenous.
I characterize this trade-off theoretically and through realistic MC simulations, and explore its prevalence in the literature.
When power is low, the distribution of estimates is spread out. Only estimates that are 1.96 sd away from 0 are statistically significant.
With low statistical power, significant estimates are always exaggerated.
The credibility revolution has made causal inference methods ubiquitous in economics. Yet there is widespread evidence of selection on significance and associated biases in the literature. I show that these two phenomena interact to reduce the reliability of published estimates: while causal identification strategies alleviate bias from confounders, by restricting the variation used for identification they reduce statistical power and can exacerbate another bias--exaggeration. I characterize this confounding-exaggeration trade-off theoretically and through realistic Monte Carlo simulations, and explore its prevalence in the literature. In realistic settings, exaggeration can exceed the confounding bias these methods aim to eliminate. Finally, I propose practical solutions to navigate this trade-off, including a tool to identify the variation and observations actually driving identification in an applied causal study.
I updated my paper showing that causal inference methods induce a trade of between confoundings and exaggerating true effect sizes.
By limiting the variation used for identification, they reduce statistical power and precision and can exacerbate another bias, exaggeration.
ssrn.com/abstract=607...
Join us at CERGIC ! ๐ผ
We offer a ๐๐๐ผ-๐๐ฒ๐ฎ๐ฟ ๐ฝ๐ผ๐๐๐ฑ๐ผ๐ฐ ๐ฝ๐ผ๐๐ถ๐๐ถ๐ผ๐ป in Applied Economics and Earth Observation Data. The position will start on September 1st, 2026. The successful candidate will be based at Ecole Normale Superieure de Lyon.
For more information: www.cergic-lyon.fr/postdoctoral...
๐จCall for application for the ENS Lyon Summer School in Empirical Research Methods ๐จ
The event is hosted by the cergic.bsky.social from June 30 to July 2, 2026.
Audience: PhD students in economics
๐
Deadline: 1 March 2026
#Econsky #Economics
Flyer program Summer School 2026 at ENS de Lyon contact: econ.summer.school@ens-lyon.fr
We are delighted to announce the second edition of our 3-day Summer School in Empirical Research Methods at @cergic.bsky.social
The event will take place in Lyon from June 30 to July 2, 2026.
Program and additional information at:
www.cergic-lyon.fr/summer-school
๐จ New working paper! We show that women in France emit 26% less COโ than men from food & transport.
๐ฅFood + ๐ Transport =50% of individual carbon footprint โ we show this isnโt just about biology or labor market differences.
A thread๐๐งต(1/12)
Link to wp: www.lse.ac.uk/granthaminst...
Don't forget, on Monday May 26 and Tuesday May 27, the "Workshop in Gender Economics" will be held at @ensdelyon.bsky.social.
Please find the updated program below ๐
