Working paper
"Forecasting Unemployment using Deep Neural Networks" Results are available upon request.
Most recently, machine learning methods have demonstrated accurate performance in different domains, including economics. Current methods for forecasting macroeconomic indicators are dominated by statistical methods that require a large amount of data and some prior knowledge of the data-generating process, which may not be available. In this study, I explore machine learning methods to predict U.S. civilian unemployment across four forecast horizons using models built on three neural network architectures- Long Short-Term Memory (LSTM), Encoder Decoder, and Multi-Layer Perceptron (MLP). Using various performance metrics, results show that all the models perform well in predicting the unemployment rate at short-term horizons. However, only the Encoder-Decoder model performs well at the medium-term horizon. The predictiveness of the models suggests that the models were able to learn the economic relationship between the past and future values of the unemployment rate. The findings in this study imply that models built using machine learning methods, such as deep neural networks may have predictive powers as those built using statistical methods.
"Effect of the Earned Income Tax Credit on Occupational Skill Mismatch" Job Market Paper
By exploring several expansions of the Earned Income tax Credit (EITC), this paper provides an intent-to-treat estimate of job match quality response to wage subsidies. As a conceptual framework, I develop a simple job search model with wage subsidies, which predicts that increases in the subsidy benefit increase the marginal opportunity cost of search, increasing the search cost. The increase in the cost of search reduces the net-of-cost benefit of holding out or waiting for better job opportunities, creating incentives for job seekers to lower their reservation wages, consequently reducing the potential of forming a better job match. The underlying hypothesis of this study based on the predictions of this search model is that the EITC may have an unintended consequence of creating worse job matches in an initial job taken after reentering the labor market. I define the quality of a job match as the difference between the set of skills required by an occupation and the set of abilities a worker possesses for learning those skills. Using a simulated instrument approach and data from the 1979 National Longitudinal Survey of Youth and the Occupational Information Network, results reveal that the EITC increases skill mismatch in an initial job, driven by workers being overqualified for their jobs. Coefficient estimates show that much of the effect is concentrated among single women with some level of college education. Evidence also reveals that mismatch lowers the starting wage of workers. These findings suggest that the effort of policymakers to reduce welfare dependency by using wage subsidies to promote employment may lead to unintended consequences such as poor match quality and lower starting wages.
Works in progress
"The Earned Income Tax Credit and Temporal Flexibility"
Traditionally, women are known to value and effectively use flexibility in the workplace. But in theory, flexibility is associated with a price, and the credit benefit from the Earned Income tax Credit (EITC) may affect the price of flexibility faced by workers eligible for the EITC benefit and, consequently, the demand for flexibility. Additionally, the income effects from the EITC may affect the demand for flexibility if flexibility is a normal good. Workplace flexibility has been operationalized empirically using different measures. In this paper, I follow Goldin(2014) and measure workplace flexibility using occupation characteristics that reflect several aspects of a job, such as the time pressure associated with the job, the need for workers to be around at particular times, the flexibility of the occupation with regard to scheduling, the groups, and workers the employee must regularly keep in touch with, and the degree to which the worker has close substitutes. This measure of flexibility is referred to as temporal flexibility. Using the public version of the National Longitudinal Survey of Youth 1979 cohort (NLSY79) and the O*NET database, I find evidence of a positive relationship between the EITC generosity and the demand for job flexibility.
"Effect of the Earned Income Tax Credit on Job Satisfaction and Quit Behavior"
In this study, I first investigate whether there is any relationship between the EITC and job satisfaction. The benefit incentives from the EITC may affect the quality of job matches workers form in the labor market, the flexibility workers demand in their jobs, and the wages workers earn on a job. Evidence from the literature shows that these factors are associated with the satisfaction workers report on their jobs. I exploit the EITC expansions in the 1980s and 1990s to provide evidence on whether the increasing generosity of the EITC has affected the job satisfaction of non-college-educated single women, using the 1980 to 2006 waves of the NLYS79 cohort to obtain individual-level information on measures of job satisfaction and detailed labor market information, and estimating an ordered logit model. Preliminary results using the federal variations in the EITC show that increases in the EITC increase the probability of reporting being dissatisfied with one's job but reduce the probability of reporting being very satisfied. Second, I examine whether the generosity of the EITC affects the relationship between job satisfaction and job quits. Among the EITC population, quitting a job could potentially mean exiting the labor market since intermittent labor force participation is often observed among women. The wage subsidy from the EITC implies that when a worker quits a job, she does not only lose the wage income from the job but the subsidy benefit. The increased opportunity cost of exiting the labor market may discourage quitting a job to exit the labor market for a given level of job satisfaction. Moreover, the EITC may increase the probability of quitting a job by encouraging job-to-job transitions if the credit benefit affects match quality and match quality is associated with job-to-job transitions. I examine whether the generosity of the EITC affects the relationship between job satisfaction and job quits using a logit model, exploring the federal and state-level variation in the EITC from expansions in the EITC over more than two decades. Results from preliminary analysis using the federal variation in the EITC show that increases in the EITC generosity reduce the probability of quitting one's job, even among workers who report being dissatisfied with their jobs.
"Are Formal and Informal Loans Substitutes or Complements?: The Case of Household Enterprise Credit Access in Ghana" with Danny Turkson
The number one challenge facing Household Enterprises (HEs) in Ghana is limited access to loans. Many of these entrepreneurs are considered not creditworthy by formal financial institutions because of their inability to satisfy rigid banking terms and conditions, hence leaving them with limited options of borrowing from an informal source which mostly comes at a high cost. The purpose of this study is to identify the factors that influence HEs’ access to formal and informal loans, and whether demand for formal and informal loans are substitutes or complements.
Using Ghana’s 2013 informal enterprise survey of 729 household enterprises and ordinary probit and bivariate probit models, results for the formal financial market show that the entrepreneur’s experience, enterprise age, enterprise size, and possession of assets were the significant determinants, indicating a positive relationship except enterprise age which indicated a negative relationship. Also, the determinants of access to informal loans were the entrepreneur’s gender and household size. We use a bivariate probit regression to estimate the relationship between demand for formal and informal loans. Our findings indicate that formal and informal loans are imperfect substitutes, with the probability of substituting formal and informal loans being 19.5%.
Our paper provides empirical evidence that outlines the determinants of loan accessibility by HEs across both the formal and informal credit markets and the link between credit demand across both formal and informal financial sectors.