Kyle Kettler presents new working paper at the Econometric Society Latin American Workshop in Lima, Peru

Kyle Kettler speaking at the Latin American Workshop in Econometrics in Lima, Peru on August 22, 2018

Kyle Kettler presented a new collaboration with Prof. Matthew Harding of UCI and Prof. Jerry Hausman of MIT. The paper, titled “A Structural Approach to Dynamic Electricity Pricing and Consumer Welfare,” investigates time-dependent prices for electricity. The authors find that customers are better off under a price system where electricity is cheap in the morning and evening but much more expensive during the afternoon.

Utilities often charge higher rates during the afternoon and offer reduced prices during off-peak hours in an attempt to control how much power is required by the grid. A utility can charge a higher afternoon cost on hot days where electricity demand is likely to be high, or leave the price low on days where consumers don’t need to conserve. Using a machine learning algorithm to predict how consumers would have responded to changes in temperature under a more conventional flat rate, the authors model how different households responded to the changes in prices.

New paper on attrition bias in quantile panel models

New paper by Matthew Harding and Carlos Lamarche (University of Kentucky): “A Panel Quantile Approach to Attrition Bias in Big Data: Evidence from a Randomized Experiment”. The paper is available at https://arxiv.org/abs/1808.03364 and will be published in the Journal of Econometrics.

This paper introduces a quantile regression estimator for panel data models with individual hetero- geneity and attrition. The method is motivated by the fact that attrition bias is often encountered in Big Data applications. For example, many users sign-up for the latest program but few remain active users several months later, making the evaluation of such interventions inherently very chal- lenging. Building on earlier work by Hausman and Wise (1979), we provide a simple identification strategy that leads to a two-step estimation procedure. In the first step, the coefficients of interest in the selection equation are consistently estimated using parametric or nonparametric methods. In the second step, standard panel quantile methods are employed on a subset of weighted ob- servations. The estimator is computationally easy to implement in Big Data applications with a large number of subjects. We investigate the conditions under which the parameter estimator is asymptotically Gaussian and we carry out a series of Monte Carlo simulations to investigate the finite sample properties of the estimator. Lastly, using a simulation exercise, we apply the method to the evaluation of a recent Time-of-Day electricity pricing experiment inspired by the work of Aigner and Hausman (1980).

JEL: C21, C23, C25, C55.
Keywords: Attrition; Big Data; Quantile regression; Individual Effects; Time-of-Day Pricing

Internship @ Kinecta Federal Credit Union

Brandon Newberg is entering his senior year pursuing a Bachelor of Arts in Quantitative Economics at the UC Irvine School of Sciences. After witnessing the power of machine learning in an upper division Econometrics class taught by Dr. Matthew Harding, Brandon joined the Deep Data Lab at UCI to further explore the field.

Under the guidance of Dr. Harding in the Deep Data Lab, Brandon spent his junior year developing a solid foundation in machine learning by utilizing a variety of machine learning algorithms to tackle real-world problems such as predicting poverty for households. Now that Brandon had a firm introduction machine learning concepts, Dr. Harding helped to facilitate a summer internship for Brandon at a local credit union.

Throughout the summer of 2018, Brandon worked as a Business Intelligence intern for Kinecta Federal Credit Union. He was tasked with discovering if machine learning concepts could be applied to credit risk modeling to better facilitate target marketing campaigns for financial services.

To effectively tackle this project, Brandon needed to apply his knowledge of machine learning concepts to construct a credit risk model and translate the model’s output into actionable business decisions. Brandon began his internship by working closely with the database administrators to compile a data set with relevant predictive variables to form the basis of the credit risk model. He then utilized the R programming language to implement the machine learning algorithms such as Random Forests and Neural Networks to model the credit risk of bank members. Brandon finished the internship by developing a presentation that communicated the findings of his study to the senior executives.

This internship allowed Brandon to apply his coursework in Quantitative Economics and his research experience to real-world industry problems. Brandon will spend his senior year writing a thesis in applied machine learning and working as a research assistant in the Deep Data Lab to continue his exploration in the vast field of machine learning and artificial intelligence.