Mike Gibilisco
@mgibilisco.bsky.social
Professor and political scientist at Caltech researching and teaching about conflict, political institutions, and connections between models and data.
michaelgibilisco.com
michaelgibilisco.com
Stand aside \bot here comes \swordcontradict
June 9, 2025 at 9:57 PM
Stand aside \bot here comes \swordcontradict
Even after 8 years, I don’t fully understand caltech’s skip day, but this is the first year with an inflatable dragon outside my office.
May 23, 2025 at 3:55 PM
Even after 8 years, I don’t fully understand caltech’s skip day, but this is the first year with an inflatable dragon outside my office.
First day of class for the spring quarter
April 1, 2025 at 4:37 PM
First day of class for the spring quarter
Our second change is we vary each groups’ incentives to compete. When Hamas has greater competitive incentives, both groups use more violence, which is an encouragement effect. When Fatah becomes more competitive, both groups use less violence, which is a discouragement effect.
March 14, 2025 at 3:38 PM
Our second change is we vary each groups’ incentives to compete. When Hamas has greater competitive incentives, both groups use more violence, which is an encouragement effect. When Fatah becomes more competitive, both groups use less violence, which is a discouragement effect.
Our first change is to set violence from one group to zero and see how the other responds: we find that Fatah always decreases its use of violence when Hamas never attacks, but Hamas actually increases its violence in periods where Fatah is very popular.
March 14, 2025 at 3:38 PM
Our first change is to set violence from one group to zero and see how the other responds: we find that Fatah always decreases its use of violence when Hamas never attacks, but Hamas actually increases its violence in periods where Fatah is very popular.
For model fit, we find that outbidding explains that data better than a no-competition model where groups do not value support. We use a nonnested model fit test to compare our outbidding model to a rival tit-for-tat model and find that outbidding also better explains groups' attacks.
March 14, 2025 at 3:38 PM
For model fit, we find that outbidding explains that data better than a no-competition model where groups do not value support. We use a nonnested model fit test to compare our outbidding model to a rival tit-for-tat model and find that outbidding also better explains groups' attacks.
For parameter estimates, we find that Fatah attacks are more effective at boosting public support than Hamas attacks, although Hamas has smaller attack costs and values its popularity more.
March 14, 2025 at 3:38 PM
For parameter estimates, we find that Fatah attacks are more effective at boosting public support than Hamas attacks, although Hamas has smaller attack costs and values its popularity more.
Our violence data are terrorist attacks from the GTD. We also use public opinion data from survey centers that track the support of groups over time by asking Palestinians what group they support/trust the most and what group would they vote for.
March 14, 2025 at 3:38 PM
Our violence data are terrorist attacks from the GTD. We also use public opinion data from survey centers that track the support of groups over time by asking Palestinians what group they support/trust the most and what group would they vote for.
Our model is a discrete-choice tug of war: groups decide whether to use violence to pull public opinion toward them, away from their rival. The key parameters are each group’s cost of violence, value of public support, effectiveness of moving support with violence.
March 14, 2025 at 3:38 PM
Our model is a discrete-choice tug of war: groups decide whether to use violence to pull public opinion toward them, away from their rival. The key parameters are each group’s cost of violence, value of public support, effectiveness of moving support with violence.
We played OuiSi in my writing seminar to practice justifying creative choices and then providing feedback on the choices. This is our creation.
February 7, 2025 at 8:03 PM
We played OuiSi in my writing seminar to practice justifying creative choices and then providing feedback on the choices. This is our creation.
I might not survive this conference
January 23, 2025 at 4:21 AM
I might not survive this conference
Is mdpi trolling me?
January 11, 2025 at 3:57 AM
Is mdpi trolling me?
The Ira Glass "taste gap" quote made it on my writing syllabus this semester, but I cannot remember what podcast I first heard it on. Was it This American Life?
January 6, 2025 at 5:11 PM
The Ira Glass "taste gap" quote made it on my writing syllabus this semester, but I cannot remember what podcast I first heard it on. Was it This American Life?
below the conventional threshold of 0.05
September 16, 2024 at 11:33 PM
below the conventional threshold of 0.05
They said it not me.
July 29, 2024 at 4:42 AM
They said it not me.
I found this gem and am excited to find areas of circles without using pi.
November 26, 2023 at 10:13 PM
I found this gem and am excited to find areas of circles without using pi.
The author uses this figure to illustrate what a strategy is, intuitive notions of complexity, technical definitions of complexity, experimental design and substantive results.
October 6, 2023 at 5:36 PM
The author uses this figure to illustrate what a strategy is, intuitive notions of complexity, technical definitions of complexity, experimental design and substantive results.
In our setting, misreporting is especially pernicious (not conditionally random): any variable that has a true causal effect also *directly* affects the misreporting probability. This is an equilibrium effect since the agent optimally conditions his misreporting strategy on behavior.
August 7, 2023 at 8:31 PM
In our setting, misreporting is especially pernicious (not conditionally random): any variable that has a true causal effect also *directly* affects the misreporting probability. This is an equilibrium effect since the agent optimally conditions his misreporting strategy on behavior.
Eg: consider the effects of economic opportunity on crime. Even with exogenous variation, misreported data can over- or underestimate treatment effects. The problem is enforcement and reporting are intertwined. Policies directly affecting one will indirectly affect the other.
August 7, 2023 at 8:30 PM
Eg: consider the effects of economic opportunity on crime. Even with exogenous variation, misreported data can over- or underestimate treatment effects. The problem is enforcement and reporting are intertwined. Policies directly affecting one will indirectly affect the other.
We also compute treatment effects (of changes in exogenous parameters) on both true and observed crime statistics. The difference between these two quantities is the effect of the treatment of equilibrium measurement error.
August 7, 2023 at 8:29 PM
We also compute treatment effects (of changes in exogenous parameters) on both true and observed crime statistics. The difference between these two quantities is the effect of the treatment of equilibrium measurement error.
We show that an increase in the agent’s costs of manipulating data can backfire and increase equilibrium rates of measurement. With higher costs, the agent has more incentives to choose appropriate behavior, which makes her reports more credible.
August 7, 2023 at 8:28 PM
We show that an increase in the agent’s costs of manipulating data can backfire and increase equilibrium rates of measurement. With higher costs, the agent has more incentives to choose appropriate behavior, which makes her reports more credible.
We model an enforcement interaction between a police agent and a potential criminal, which produces a crime statistic, which then has to be reported by the agent. The key tension is that the agent may lie because it wishes to signal it choose appropriate enforcement behavior.
August 7, 2023 at 8:27 PM
We model an enforcement interaction between a police agent and a potential criminal, which produces a crime statistic, which then has to be reported by the agent. The key tension is that the agent may lie because it wishes to signal it choose appropriate enforcement behavior.
Excited to post a new polisky working paper: https://michaelgibilisco.com/papers/draft_data_manipulation.pdf
Twitterless Carlo Horz and I study the production of crime data and the implications of strategic misreporting for researchers and policymakers.
Twitterless Carlo Horz and I study the production of crime data and the implications of strategic misreporting for researchers and policymakers.
August 7, 2023 at 8:24 PM
Excited to post a new polisky working paper: https://michaelgibilisco.com/papers/draft_data_manipulation.pdf
Twitterless Carlo Horz and I study the production of crime data and the implications of strategic misreporting for researchers and policymakers.
Twitterless Carlo Horz and I study the production of crime data and the implications of strategic misreporting for researchers and policymakers.