Two applications of infrequency of purchase and double hurdle models to the study of time use /

Abstract

Using a household production framework, these articles examine two important components of people's time allocation. Data from the American Time Use Survey (ATUS) for 2003 -- 2010 are analyzed with infrequency of purchase and double hurdle models to account for the idiosyncrasies of the time diaries.

The first article investigates parents' time allocation between direct and indirect child care for producing human capital in their children. Sources of the "time gap" between men and women are identified by decomposition. Endogeneity and selection bias are managed simultaneously with a multi-step estimation process. I use multiple imputation to handle missing data and an inverse hyperbolic sine transformation to correct for heteroskedasticity and nonormality of residuals.

I find that mothers increase indirect child care time relative to direct child care as hourly earnings rise, evincing a substitution effect. This effect is stronger for whites, college graduates. and single parents. Greater amounts of both types of care are associated with higher incomes for mothers and fathers. Parent-students of both sexes devote less time to both types of care. More schooling is associated with sizable increases in both types of care for women. Longer work hours reduce child care time for men far more than women, suggesting that women reduce leisure or household work to preserve child care time. Rising earnings attenuates the time gap, while schooling increases it. Also, I find evidence of negative selection in reporting earnings for men in the ATUS.

Article two finds that Californians responded to the MINI pandemic of 2009 -- 2010 by reducing work time to avoid catching the disease, again using infrequency of purchase and double hurdle models. ATUS data are combined with official mortality reports and local newspaper article counts. Modeling separately by region, sex, and age, I find evidence that some workers responded to reports of the pandemic in the news media but not to actual changes in mortality, with the most consistent effects for younger females in southern California and the Bay Area. The relative performance of multiple imputation and inverse probability weight methods are examined, with MI showing some advantage in finding significant results.