Vitor Possebom
@vitorpossebom.bsky.social
Assistant Professor at Sao Paulo Schools of Economics (FGV). Ph.D. 2022 at Yale. Passionate about Econometrics, especially Causal Inference. (He/his)
https://sites.google.com/site/vitorapossebom/
https://sites.google.com/site/vitorapossebom/
All QMTE functions behave similarly. Although the level of the estimated QMTEs depends on the quantile, all functions are decreasing in the unobserved resistance to punishment.
November 27, 2023 at 12:47 PM
All QMTE functions behave similarly. Although the level of the estimated QMTEs depends on the quantile, all functions are decreasing in the unobserved resistance to punishment.
But things change a lot in the long-run! The DMTEs are decreasing. This indicates that defendants whom almost all judges would punish would recidivate faster because of the punishment, while defendants who would be punished only by tough judges would take longer
to recidivate.
to recidivate.
November 27, 2023 at 12:45 PM
But things change a lot in the long-run! The DMTEs are decreasing. This indicates that defendants whom almost all judges would punish would recidivate faster because of the punishment, while defendants who would be punished only by tough judges would take longer
to recidivate.
to recidivate.
The DMTEs are increasing in the short-run. This behavior indicates that defendants whom almost all judges would punish are less likely to recidivate, while defendants who would be punished only by tough judges are more likely to recidivate.
November 27, 2023 at 12:45 PM
The DMTEs are increasing in the short-run. This behavior indicates that defendants whom almost all judges would punish are less likely to recidivate, while defendants who would be punished only by tough judges are more likely to recidivate.
We use a conditional local IV estimand to identify the distributional treatment responses. After that, we can get the DMTE and the QMTE.
Our estimation procedure is a bit long, but super easy to implement! You only need to run a few logit models and take some averages!
Our estimation procedure is a bit long, but super easy to implement! You only need to run a few logit models and take some averages!
November 27, 2023 at 12:44 PM
We use a conditional local IV estimand to identify the distributional treatment responses. After that, we can get the DMTE and the QMTE.
Our estimation procedure is a bit long, but super easy to implement! You only need to run a few logit models and take some averages!
Our estimation procedure is a bit long, but super easy to implement! You only need to run a few logit models and take some averages!
We do so by looking at the Distributional MTE and the Quantile MTE of time-to-recidivism, a continuous variable.
We impose Random Assignment and Censoring. The last assumption imposes that the sentence date is independent of the defendant’s decision to commit another crime.
We impose Random Assignment and Censoring. The last assumption imposes that the sentence date is independent of the defendant’s decision to commit another crime.
November 27, 2023 at 12:43 PM
We do so by looking at the Distributional MTE and the Quantile MTE of time-to-recidivism, a continuous variable.
We impose Random Assignment and Censoring. The last assumption imposes that the sentence date is independent of the defendant’s decision to commit another crime.
We impose Random Assignment and Censoring. The last assumption imposes that the sentence date is independent of the defendant’s decision to commit another crime.
First, long-run recidivism matters in São Paulo, Brazil. A non-negligible share of recidivism events happens 6-8 years after the judge’s decision.
So analyzing the effect of judges’ decision on long-run recidivism may matter when discussing sentencing guidelines.
So analyzing the effect of judges’ decision on long-run recidivism may matter when discussing sentencing guidelines.
November 27, 2023 at 12:42 PM
First, long-run recidivism matters in São Paulo, Brazil. A non-negligible share of recidivism events happens 6-8 years after the judge’s decision.
So analyzing the effect of judges’ decision on long-run recidivism may matter when discussing sentencing guidelines.
So analyzing the effect of judges’ decision on long-run recidivism may matter when discussing sentencing guidelines.
Misdemeanor punishment (fines) can have heterogeneous effects on recidivism depending on the defendants’ types and time horizon.
We explore these sources of heterogeneity by looking at the marginal treatment effect of misdemeanor punishment on recidivism at many time horizons.
We explore these sources of heterogeneity by looking at the marginal treatment effect of misdemeanor punishment on recidivism at many time horizons.
November 27, 2023 at 12:41 PM
Misdemeanor punishment (fines) can have heterogeneous effects on recidivism depending on the defendants’ types and time horizon.
We explore these sources of heterogeneity by looking at the marginal treatment effect of misdemeanor punishment on recidivism at many time horizons.
We explore these sources of heterogeneity by looking at the marginal treatment effect of misdemeanor punishment on recidivism at many time horizons.
🚨New Working Paper🚨
Was Javert right to be suspicious? Unpacking treatment effect heterogeneity of alternative sentences on time-to-recidivism in Brazil
by Acerenza (scholar.google.com/citations?user…), Possebom and Sant’Anna (@pedrosantanna.bsky.social)
Paper: arxiv.org/abs/2311.13969
Was Javert right to be suspicious? Unpacking treatment effect heterogeneity of alternative sentences on time-to-recidivism in Brazil
by Acerenza (scholar.google.com/citations?user…), Possebom and Sant’Anna (@pedrosantanna.bsky.social)
Paper: arxiv.org/abs/2311.13969
November 27, 2023 at 12:41 PM
🚨New Working Paper🚨
Was Javert right to be suspicious? Unpacking treatment effect heterogeneity of alternative sentences on time-to-recidivism in Brazil
by Acerenza (scholar.google.com/citations?user…), Possebom and Sant’Anna (@pedrosantanna.bsky.social)
Paper: arxiv.org/abs/2311.13969
Was Javert right to be suspicious? Unpacking treatment effect heterogeneity of alternative sentences on time-to-recidivism in Brazil
by Acerenza (scholar.google.com/citations?user…), Possebom and Sant’Anna (@pedrosantanna.bsky.social)
Paper: arxiv.org/abs/2311.13969
Sometimes, I think that journals do not know that many of its readers have eyesight problems...
Elsevier's font size is so small and its line spacing is miniscule.
Elsevier's font size is so small and its line spacing is miniscule.
November 9, 2023 at 12:16 PM
Sometimes, I think that journals do not know that many of its readers have eyesight problems...
Elsevier's font size is so small and its line spacing is miniscule.
Elsevier's font size is so small and its line spacing is miniscule.
Probable Causation is an amazing podcast and this interview with Allison Stashko is awesome: open.spotify.com/episode/0rfq...
They discuss the effect of newly elected prosecutors and police killings in the US. It is based on this paper: drive.google.com/file/d/1LwnA....
They discuss the effect of newly elected prosecutors and police killings in the US. It is based on this paper: drive.google.com/file/d/1LwnA....
November 3, 2023 at 3:20 PM
Probable Causation is an amazing podcast and this interview with Allison Stashko is awesome: open.spotify.com/episode/0rfq...
They discuss the effect of newly elected prosecutors and police killings in the US. It is based on this paper: drive.google.com/file/d/1LwnA....
They discuss the effect of newly elected prosecutors and police killings in the US. It is based on this paper: drive.google.com/file/d/1LwnA....
One of my pastimes is to watch videos about writing fiction. Many things that work when writing fiction also work when writing academic papers!
I think this video on editing is quite useful: youtu.be/WLAmilJx3Us?....
E.g., accept that your manuscript will evolve a lot over time!
I think this video on editing is quite useful: youtu.be/WLAmilJx3Us?....
E.g., accept that your manuscript will evolve a lot over time!
November 2, 2023 at 2:21 PM
One of my pastimes is to watch videos about writing fiction. Many things that work when writing fiction also work when writing academic papers!
I think this video on editing is quite useful: youtu.be/WLAmilJx3Us?....
E.g., accept that your manuscript will evolve a lot over time!
I think this video on editing is quite useful: youtu.be/WLAmilJx3Us?....
E.g., accept that your manuscript will evolve a lot over time!
And both papers have super cool discussions on multi-valued treatment models with instruments. Both papers explain why this problem is much harder than the binary treatment problem! And both propose different solutions to it. I am still digesting them. But I learned a lot from both papers!
[end]
[end]
November 1, 2023 at 5:36 PM
And both papers have super cool discussions on multi-valued treatment models with instruments. Both papers explain why this problem is much harder than the binary treatment problem! And both propose different solutions to it. I am still digesting them. But I learned a lot from both papers!
[end]
[end]
So, increasing leniency in the conviction margin while decreasing it in the incarceration margin may benefit society and defendants.
But take this policy conclusion with a gigantic grain of salt. Recidivism, conviction and incarceration are very complicated topics with multiple dimensions.
+
But take this policy conclusion with a gigantic grain of salt. Recidivism, conviction and incarceration are very complicated topics with multiple dimensions.
+
November 1, 2023 at 5:33 PM
So, increasing leniency in the conviction margin while decreasing it in the incarceration margin may benefit society and defendants.
But take this policy conclusion with a gigantic grain of salt. Recidivism, conviction and incarceration are very complicated topics with multiple dimensions.
+
But take this policy conclusion with a gigantic grain of salt. Recidivism, conviction and incarceration are very complicated topics with multiple dimensions.
+
Econometrics Thread (#EconSky): "The Effect of Conviction and Incarceration on Recidivism"
Today, I want to talk about two recent working papers:
- Kamat, Norris and Pecenco (2023): bit.ly/3QjqdbX
- Humphries, Ouss, Stavreva, Stevenson and van Dijk (2023): bit.ly/3Mrrdtp
+
Today, I want to talk about two recent working papers:
- Kamat, Norris and Pecenco (2023): bit.ly/3QjqdbX
- Humphries, Ouss, Stavreva, Stevenson and van Dijk (2023): bit.ly/3Mrrdtp
+
November 1, 2023 at 5:22 PM
Econometrics Thread (#EconSky): "The Effect of Conviction and Incarceration on Recidivism"
Today, I want to talk about two recent working papers:
- Kamat, Norris and Pecenco (2023): bit.ly/3QjqdbX
- Humphries, Ouss, Stavreva, Stevenson and van Dijk (2023): bit.ly/3Mrrdtp
+
Today, I want to talk about two recent working papers:
- Kamat, Norris and Pecenco (2023): bit.ly/3QjqdbX
- Humphries, Ouss, Stavreva, Stevenson and van Dijk (2023): bit.ly/3Mrrdtp
+
Sometimes, academic life has its perks. Visiting the Federal University of Paraíba (sigaa.ufpb.br/sigaa/public...) was a wonderful experience. Their students and faculty are doing super cool research on relevant topics for Brazilian policy makers.
October 21, 2023 at 8:25 PM
Sometimes, academic life has its perks. Visiting the Federal University of Paraíba (sigaa.ufpb.br/sigaa/public...) was a wonderful experience. Their students and faculty are doing super cool research on relevant topics for Brazilian policy makers.
I also asked it to illustrate my paper on cherry picking with synthetic controls. @brunoferman.bsky.social, did you like it?
doi.org/10.1002/pam....
doi.org/10.1002/pam....
October 18, 2023 at 10:27 PM
I also asked it to illustrate my paper on cherry picking with synthetic controls. @brunoferman.bsky.social, did you like it?
doi.org/10.1002/pam....
doi.org/10.1002/pam....
I asked Bing’s Dall-E to illustrate my paper in a steam punk style!
Any thoughts on the results?
Any thoughts on the results?
October 18, 2023 at 10:21 PM
I asked Bing’s Dall-E to illustrate my paper in a steam punk style!
Any thoughts on the results?
Any thoughts on the results?
If you choose to use one or the other, you must correctly specify the chosen model. And that's risky.
Why not use both and hedge against misspecification in one of your models? Double robust methods do exactly this!
Why not use both and hedge against misspecification in one of your models? Double robust methods do exactly this!
October 15, 2023 at 11:53 PM
If you choose to use one or the other, you must correctly specify the chosen model. And that's risky.
Why not use both and hedge against misspecification in one of your models? Double robust methods do exactly this!
Why not use both and hedge against misspecification in one of your models? Double robust methods do exactly this!
My office during the morning!
October 13, 2023 at 1:08 PM
My office during the morning!
🚨Publication Alert🚨 "Crime and Mismeasured Punishment: Marginal Treatment Effect with Misclassification" by Vitor Possebom (2023, doi.org/10.1162/rest...)
I am so happy that my JMP was accepted at the REStat! I am thankful to so many people who helped me through this journey.
I am so happy that my JMP was accepted at the REStat! I am thankful to so many people who helped me through this journey.
October 12, 2023 at 4:10 PM
🚨Publication Alert🚨 "Crime and Mismeasured Punishment: Marginal Treatment Effect with Misclassification" by Vitor Possebom (2023, doi.org/10.1162/rest...)
I am so happy that my JMP was accepted at the REStat! I am thankful to so many people who helped me through this journey.
I am so happy that my JMP was accepted at the REStat! I am thankful to so many people who helped me through this journey.
Before we discuss estimation, let's discuss some empirical results. Carneiro, Heckman and Vytlacil (2011) estimate the marginal college premium. When V (U_S in the figure) is small, it means that the idiosyncratic benefit of attending college is large. (This bit is confusing.)
October 10, 2023 at 5:37 PM
Before we discuss estimation, let's discuss some empirical results. Carneiro, Heckman and Vytlacil (2011) estimate the marginal college premium. When V (U_S in the figure) is small, it means that the idiosyncratic benefit of attending college is large. (This bit is confusing.)
Now, how can we identify the MTE? If the instrument is continuous, we can use the LIV estimator in the lef-hand side of this equation to identify the MTE evaluated at V = p. This result is reason behind the MTE's name.
October 10, 2023 at 5:37 PM
Now, how can we identify the MTE? If the instrument is continuous, we can use the LIV estimator in the lef-hand side of this equation to identify the MTE evaluated at V = p. This result is reason behind the MTE's name.
In our example, it is the college premium for someone whose utility to attend college is equal to v. This parameter is beautiful because many famous treatment parameter can be written as weighted integrals of the MTE. For example, ATE, ATT and LATE are functions of the MTE.
October 10, 2023 at 5:35 PM
In our example, it is the college premium for someone whose utility to attend college is equal to v. This parameter is beautiful because many famous treatment parameter can be written as weighted integrals of the MTE. For example, ATE, ATT and LATE are functions of the MTE.
But what do we gain by being explicit about our selection model? We can define the marginal treatment effect, which, at this point, can be interpreted as the average treatment effect for an individual whose unobserved heterogeneity is equal to v.
October 10, 2023 at 5:35 PM
But what do we gain by being explicit about our selection model? We can define the marginal treatment effect, which, at this point, can be interpreted as the average treatment effect for an individual whose unobserved heterogeneity is equal to v.
Before defining the MTE, let's explain the underlying selection-into-treatment model. We have the usual potential outcomes, Y1 and Y0. Treatment (D) selection follows an index model, where Z is the instrument and V is some sort of unobserved heterogeneity.
October 10, 2023 at 5:33 PM
Before defining the MTE, let's explain the underlying selection-into-treatment model. We have the usual potential outcomes, Y1 and Y0. Treatment (D) selection follows an index model, where Z is the instrument and V is some sort of unobserved heterogeneity.