You may find all of my current and past academic research projects on this page. Please scroll below for abstracts of these papers or follow the links for more in-depth information. My academic work focuses primarily on deviations from rational expectations and the effects this may have on the business cycle. My professional research for the Cato Institute focuses on monetary policy. To access my work for the Cato Institute, please click here.
Investor Heuristics and Biases: Quantitative Effects on the U.S. Business Cycle
Job Market Paper
Abstract: This paper presents a medium-scale quantitative New-Keynesian DSGE model with financial intermediaries that incorporates a variety of behavioral modifications to investors. Investors in this model are subject to the anchoring and adjustment heuristic, endogenous confidence bias, and exogenous confidence bias. A Bayesian MCMC approach is utilized to estimate various iterations of this model, including shock process parameters, for the U.S. economy. The estimation aims to fit the model to six macroeconomic time series from 1988 through 2019, along with a measure of investors' expectations of future stock market performance. The estimated posterior means show that 18% or 31% of investor expectations arise from behavioral factors when modeled as exhibiting anchoring or confidence respectively. When exhibiting anchoring, investors are significantly pegged to 1-qtr and 1-year prior stock returns. In the model with endogenously propagated confidence, investors exhibit a confidence function that increases at a rate roughly halfway between a square and cubic root function. Both behavioral features are able to better fit the data as compared to the base model which includes only exogenous confidence shocks. In response to confidence shocks, the economy exhibits a dual-response fluctuation characterized by a large initial boom followed by a sustained recession. Models with behavioral features over-react to economic shocks compared to baseline, leading to more volatile business cycles.
Abstract: This paper presents a medium-scale quantitative New-Keynesian DSGE model with financial intermediaries that incorporates a variety of behavioral modifications to investors. Investors in this model are subject to the anchoring and adjustment heuristic, endogenous confidence bias, and exogenous confidence bias. A Bayesian MCMC approach is utilized to estimate various iterations of this model, including shock process parameters, for the U.S. economy. The estimation aims to fit the model to six macroeconomic time series from 1988 through 2019, along with a measure of investors' expectations of future stock market performance. The estimated posterior means show that 18% or 31% of investor expectations arise from behavioral factors when modeled as exhibiting anchoring or confidence respectively. When exhibiting anchoring, investors are significantly pegged to 1-qtr and 1-year prior stock returns. In the model with endogenously propagated confidence, investors exhibit a confidence function that increases at a rate roughly halfway between a square and cubic root function. Both behavioral features are able to better fit the data as compared to the base model which includes only exogenous confidence shocks. In response to confidence shocks, the economy exhibits a dual-response fluctuation characterized by a large initial boom followed by a sustained recession. Models with behavioral features over-react to economic shocks compared to baseline, leading to more volatile business cycles.
Effects of Investor Confidence Shocks on Business Cycles
Revise and Resubmit at Macroeconomic Dynamics.
Abstract: This paper extends a quantitative medium-scale New-Keynesian DSGE model with financial intermediaries to account for shocks to investor confidence. Shocks of this nature manifest themselves as per period changes to financial intermediaries' leverage ratios. A Bayesian MCMC approach is utilized to estimate the base and extended model, including shock process parameters, for the U.S. economy using five macroeconomic time series from 1984 through 2019. The estimation results suggest that confidence shocks have a large and sustained effect on the real economy. Overconfidence initially provides a boost to the economy but this effect subsides and then triggers a prolonged recession. In the base model, a decomposition of U.S. output growth into its constituent shocks shows that the effect of negative shocks to capital quality contributed significantly to the financial crisis of 2008. However, this effect is muted in the extended model suggesting that shocks to confidence were important contributors to the output gap during the Great Recession.
Abstract: This paper extends a quantitative medium-scale New-Keynesian DSGE model with financial intermediaries to account for shocks to investor confidence. Shocks of this nature manifest themselves as per period changes to financial intermediaries' leverage ratios. A Bayesian MCMC approach is utilized to estimate the base and extended model, including shock process parameters, for the U.S. economy using five macroeconomic time series from 1984 through 2019. The estimation results suggest that confidence shocks have a large and sustained effect on the real economy. Overconfidence initially provides a boost to the economy but this effect subsides and then triggers a prolonged recession. In the base model, a decomposition of U.S. output growth into its constituent shocks shows that the effect of negative shocks to capital quality contributed significantly to the financial crisis of 2008. However, this effect is muted in the extended model suggesting that shocks to confidence were important contributors to the output gap during the Great Recession.
The Empirical Effects of Cognitive Discounting on Government Spending Multipliers
with Yanyan Luo.
Work in progress.
Abstract: This paper aims to test the empirical effects of cognitive discounting on government spending multipliers in the U.S. To this end, the paper adds myopic agents to medium-scale New-Keynesian DSGE model with several frictions as well as Ricardian and non-Ricardian agents. In a calibrated framework, the results show that multipliers are intrinsically and non-linearly linked to the degree of myopia and the fraction of hand-to-mouth consumers at all horizons. The multiplier increases proportionally with myopia until the fraction of hand-to-mouth consumers crosses a certain threshold, following which the effect of myopia reverses. Additionally, the model is estimated using Bayesian MCMC techniques to fit several U.S. macro time series as well as forecasts of future government spending collected from the Survey of Professional Forecasters. The aim of this analysis is to determine the actual values of myopia and multipliers prevalent in the U.S. during its modern economic past.
Work in progress.
Abstract: This paper aims to test the empirical effects of cognitive discounting on government spending multipliers in the U.S. To this end, the paper adds myopic agents to medium-scale New-Keynesian DSGE model with several frictions as well as Ricardian and non-Ricardian agents. In a calibrated framework, the results show that multipliers are intrinsically and non-linearly linked to the degree of myopia and the fraction of hand-to-mouth consumers at all horizons. The multiplier increases proportionally with myopia until the fraction of hand-to-mouth consumers crosses a certain threshold, following which the effect of myopia reverses. Additionally, the model is estimated using Bayesian MCMC techniques to fit several U.S. macro time series as well as forecasts of future government spending collected from the Survey of Professional Forecasters. The aim of this analysis is to determine the actual values of myopia and multipliers prevalent in the U.S. during its modern economic past.
Myopia, Hand-to-Mouth Agents, and Determinacy
with Yanyan Luo.
Work in progress.
Abstract: Models with myopic agents are able to generate large government spending multipliers. However, this investigation has been limited to simple models of the aggregate economy. In this paper, myopic agents a’la Gabaix (2020) are introduced to a benchmark monetary DSGE model with both Ricardian and non-Ricardian agents that is commonly used to study the effects of fiscal intermediation. While myopic optimizing agents are able to boost government spending multipliers, the model is indeterminate for reasonable cognitive discounting values. A deeper analysis of the model’s determinacy regions uncovers a trade-off between active monetary policy and myopia: a determinate model must exhibit severe cognitive discounting at the cost of the Taylor Principle or vice-versa.
Work in progress.
Abstract: Models with myopic agents are able to generate large government spending multipliers. However, this investigation has been limited to simple models of the aggregate economy. In this paper, myopic agents a’la Gabaix (2020) are introduced to a benchmark monetary DSGE model with both Ricardian and non-Ricardian agents that is commonly used to study the effects of fiscal intermediation. While myopic optimizing agents are able to boost government spending multipliers, the model is indeterminate for reasonable cognitive discounting values. A deeper analysis of the model’s determinacy regions uncovers a trade-off between active monetary policy and myopia: a determinate model must exhibit severe cognitive discounting at the cost of the Taylor Principle or vice-versa.
Beyond Beta: An Analysis of Alternative Risk Measures
Undergraduate Senior Thesis, Honors Distinction.
Abstract: Investors require a return from investing in stock securities that adequately compensate the investors for the risk level assumed. Therefore, any calculation of expected returns from a stock requires knowledge of the risk of the security. While there is no strong consensus on an ideal risk measure, traditionally risk has been conceptualized as volatility and is measured by the ß of the stock or portfolio. This paper hypothesizes that alternative risk measures such as higher order moments, size, leverage, and price-to-book value add explanatory power to the ß when predicting stock returns. Empirical analysis is conducted using both regression and portfolio methodologies and data collected on over 300 NYSE companies. The results demonstrate a clear lack of statistical significance of alternative risk measures in explaining returns and show that the relationship between returns and ß for the time period 2003 – 2014 is negative. Additional testing is conducted by analyzing the impact of the financial crisis on the results and by changing market indices, neither of which significantly change the results obtained. This paper also builds a theoretical framework that may be used to model stock prices using a martingale process.
Abstract: Investors require a return from investing in stock securities that adequately compensate the investors for the risk level assumed. Therefore, any calculation of expected returns from a stock requires knowledge of the risk of the security. While there is no strong consensus on an ideal risk measure, traditionally risk has been conceptualized as volatility and is measured by the ß of the stock or portfolio. This paper hypothesizes that alternative risk measures such as higher order moments, size, leverage, and price-to-book value add explanatory power to the ß when predicting stock returns. Empirical analysis is conducted using both regression and portfolio methodologies and data collected on over 300 NYSE companies. The results demonstrate a clear lack of statistical significance of alternative risk measures in explaining returns and show that the relationship between returns and ß for the time period 2003 – 2014 is negative. Additional testing is conducted by analyzing the impact of the financial crisis on the results and by changing market indices, neither of which significantly change the results obtained. This paper also builds a theoretical framework that may be used to model stock prices using a martingale process.
Sales Forecasting using Regression and Artificial Neural Networks
with G.H. Nguyen, R.F. Snyder, R.D. Pasteur, and R.D. Wooster.
Published in the Proceedings of MCURCSM 2013.
Abstract: The goal of this paper is to incorporate regression techniques and artificial neural network (ANN) models to predict industry sales, which exhibit a seasonal pattern, by using both historical sales and non-seasonal economic indicators. Both short-term and long-term predictive models were constructed, ranging from one-quarter predictions to twenty-quarter predictions. The step-by-step process was as follows: deseasonalize the data set choose the relevant economic indicators using various statistical techniques, make predictions with ANNs, reseasonalize the predictions, and compute the errors of the predictions.
Published in the Proceedings of MCURCSM 2013.
Abstract: The goal of this paper is to incorporate regression techniques and artificial neural network (ANN) models to predict industry sales, which exhibit a seasonal pattern, by using both historical sales and non-seasonal economic indicators. Both short-term and long-term predictive models were constructed, ranging from one-quarter predictions to twenty-quarter predictions. The step-by-step process was as follows: deseasonalize the data set choose the relevant economic indicators using various statistical techniques, make predictions with ANNs, reseasonalize the predictions, and compute the errors of the predictions.