On July 1, 2018, over 3000 government representatives were elected at the polls in one of the biggest general elections in Mexican history. The Mexican electoral authority (INE in Spanish) organized a system of quick counts to predict the results from some of these elections—namely, the Presidential election and nine governorship elections. For this purpose, nine academic statisticians were invited to form the so-called quick counts committee.
A quick count is a statistical method to estimate the results of an election, usually the day it takes place. However, unlike an exit poll—where data is obtained by asking people for which candidate they voted—a quick count is based on information reported directly from polling station officers, thereby making the data more reliable.
I worked with committee member Dr. Manuel Mendoza as his research assistant for this project, along with Gerardo Orantes. We were in charge of devising the sampling design for the state of Chiapas, as well as for estimating the results for the governorship elections in Chiapas, Guanajuato, and Morelos, and for estimating the Presidential election results (which was done by combining the results from multiple models, ours included).
The quick counts were a success, as the results from all elections—and specfically the Presidential election—were correctly estimated. Multiple reports, presentations, and other content were (and continue to be) derived from this project, and I created this page to compile the ones in which Dr. Mendoza, Gerardo, and I worked on.
I developed a dashboard (in Spanish) on Shiny apps that compiles the evolution of the estimations of all the members of the quick counts committee for the Presidential election. For our team, I also included biplots of the posterior distribution of the estimates.