While my work as a Data Scientist is largely focused on predictive modeling, my academic research focuses on causal inference.
I employ quasi-experimental designs to derive insights from non-experimental data.
Displaying Dislike Counts as a Means to Reduce Negative Online Comments
I measure the increase in negative comments on political videos around the time that YouTube removed the dislike count. I estimate the policy's causal effect using a regression discontinuity design.
Understanding VADER: Using LIME to Overcome the Accuracy-Interpretability Trade-off
LIME allows us to explain the predictions made by any kind of model. I review its application for tabular data, image recognition and natural language processing. I showcase its use with a popular NLP model.
The Redistributive Effects of Nonlinear Pricing in Mexico's Residential Electricity Market
I measure the effects of the Mexican domestic electricity market's pricing system. I find that the current system benefits households in the lowest income decile and moderates the electricity consumption of high-income households.
The Effect of Mobile Network Penetration on Protest Activity in Mexico
I show that the expansion of 3G and 4G mobile coverage has a positive causal impact on local protest activity. In particular, reaching full coverage increases local protest activity by 21%.
Fiscal Stimulus and Economic Growth During COVID-19
I estimate that the austerity measures implemented by the current Mexican administration are associated with an 8% decrease in 2020 GDP. Had the government spent a similar amount of resources as other comparable South American countries, Mexico's GDP could have grown by 0.8% in 2020.