Saturday, February 16, 2013
Auditorium/Exhibit Hall C (Hynes Convention Center)
The American Time Use Survey (ATUS) is an annual survey that records demographic information and the minutes per day that residents of the United States spend on a variety of activities. Motivated by the recent recessions, we choose to explore the relationship between employment status and time use. For several decades, there was no meaningful gender difference in unemployment rate in the United States. However in August 2009, the difference grew to be 2.7% higher for males, the largest in the post-war era. We aim to explore the implications of this change by developing an approach of predicting ATUS subjects by employment status and gender from their time use data. We begin by employing variable selection (association analysis using heat-maps) and dimension reduction methods (principal component analysis - PCA). We then use the obtained results to create two logistic regression models to classify the population into four classes. We further use analysis of variance (ANOVA) and appropriate post-hoc tests to provide an overview of the specific differences between the four classes. Our research reveals that gender and unemployment status can be accurately predicted from time use data.