Marco Acosta

The Effects of Elite High School Eligibility on Testing Permalink

Evaluation of education policy often hinges on quantifying the value of better schooling. When school assignment is centralized and merit-based, students devote much of their effort to improving admission test performance, as test scores largely determine access to desirable schools. This study focuses on student effort and family resources devoted to elite school admission exam preparation as an approach to the value students and their families place on school quality. I exploit a policy rule that sets a middle school GPA threshold for students’ eligibility for elite high school admission in the Mexico City Metropolitan Area, and analyze the difference in test scores and the probability of attending private admission exam preparatory courses at the policy threshold using regression discontinuity estimates. I find that students who are eligible for elite schools, compared with those who are not, are 34% more likely to retake the high school admission exam, score 0.10 standard deviations higher, and are 15 percentage points more likely to take private preparatory courses.

Quantifying School Value Added in Double Shift Schooling Work in progress

Using individual-level panel data on test scores to control for student heterogeneity, I estimate the effects of afternoon school value added. I find that, on average, children attending the afternoon shift perform 5.7% and 6.6% standard deviations higher in their verbal and math test scores, respectively, compared to if they attended the morning session. The results are mainly driven by students in the lower quartile of the distribution who enroll in afternoon schools.

Slides

Recommended citation: Acosta, M. (2018). "Quantifying School Value Added in Double Shift Schooling". manuscript.

Anchoring of Inflation Expectations in Mexico Permalink

Published in Monetaria, 2017

This study analyzes short, medium and long run inflation expectations anchorage among professional forecasters from the private sector in Mexico before and after the financial crisis of 2008 by introducing a novel classification that catalogs to a large extent the econometric efforts that have been made for its measurement. The three dimensions covered by this classification are sensitivity, resilience and credibility. The results show that for the period evaluated after the 2008 financial crisis and as the horizon for which inflation forecasts are made increases, expectations are better anchored.

Recommended citation: Acosta, M. (2017). "Evaluación del anclaje de las expectativas de inflación en México; Monetaria . 39.1: 101-140

Structural Changes in the Inflation Persistence in Mexico Using the Quantile Regression Permalink

Published in El trimestre económico, 2018

It is well documented that inflation persistence in Mexico has experienced an unstable behavior through time at the conditional mean distribution. However, its behavior at conditional quantiles of the distribution have been not explored.This study determines the periods in which inflation persistence in Mexico presented structural changes in its conditional distribution using a quantile regression approach.The episodes found coincided with periods when Mexico’s economic policies underwent drastic changes that altered the price formation process. The evidence indicates that inflation shocks present an asymmetric behavior, while high magnitude negative shocks rapidly vanish, high magnitude positive shocks tend to be long lasting. Inflation converged to a stationary process in all its conditional quantiles under the inflation targeting regime. Besides, since 2009 the hypothesis that inflation adjusted for seasonal effects remains within the range variability of ± 1% point of the long-term inflation target of three percent cannot be statistically rejected.

Recommended citation: Acosta, M. (2018). " Un análisis de cambio estructural en la persistencia de la inflación en México usando la regresión cuantílica." El trimestre económico . 85(337), 169-193.

Machine Learning Core Inflation Permalink

Published in Economics Letters, 2018

In this article a novel methodology for building core inflation measures is proposed based on the k-means clustering machine learning algorithm. This new methodology is explored using Mexican CPI data in the spirit of getting a clear signal and having good predictions of the inflationary process based on selecting items with low volatility and assigning them to clusters. The results show that the core inflation built captures better the inflation signal and also outperforms the short-term inflation forecasts obtained by the trimmed means method and the core inflation excluding food and energy.

Recommended citation: Acosta, M. (2018). "Machine learning core inflation." Economics Letters . 169, 47-50.