Knowledge that Transforms

Decades of campaigns have cautioned households in Bangladesh about waterborne contaminants such as arsenic. In addition to switching water sources, mothers can protect young children from contaminated water by breastfeeding longer. We exploit time series variation in whether children were born before or after a nationwide information campaign and geographic variation in exposure to arsenic. We find that mothers breast-feed children longer in response to the campaign, especially when they have less access to uncontaminated wells, and that infants are more likely to be exclusively breast-fed. We find consistent evidence of lower mortality rates and diarrheal incidence for infants.

In the past twenty years, the share of top incomes in Switzerland has risen, while exhibiting large variations. Switzerland is similar to European countries for the top 1% but closer to the United States for higher top income groups. With the synthetic control method, we close a time gap in the tax data, exploiting the fact that Swiss cantons changed their tax system at different points in time. Using social security data, which cover all top labor incomes, we document the growing importance of labor compared to capital incomes among top income earners in Switzerland.

Business decision makers are increasingly using predictive social media analytic tools in forecasting exercises but ignoring potential model uncertainty. Using data on the universe of Twitter messages, we calculate the sentiment regarding each film to understand whether these opinions affect box office opening and DVD retail sales. Our results contrasting eleven different econometric strategies including penalization methods indicate that accounting for model uncertainty can lead to large gains in forecast accuracy. While penalization methods do not outperform model averaging on forecast accuracy, evidence indicates they perform equivalently at the variable selection stage. Finally, incorporating social media data greatly improves forecast accuracy.

We derive the exact expression for the weights of the Hodrick-Prescott (HP) filter in a finite sample without making any assumptions about the statistical properties of the time series. We use the results to give insights into the properties of the HP filter and to build a fast algorithm with computational improvements by a factor of up to three times in samples typical in economics.

This paper introduces a new estimator for the fixed-effects ordered logit model. The proposed method has two advantages over existing estimators. First, it estimates the differences in the cut points along with the regression coefficient, leading to provide bounds on partial effects. Second, the proposed estimator for the regression coefficient is more efficient. I use the fact that the ordered logit model with J outcomes and T observations can be converted to a binary choice logit model in (J - 1)T ways. As an empirical illustration, I examine the income-health gradient for children using the Medical Expenditure Panel Survey.

Missing data are a common challenge facing empirical researchers. This paper presents a general GMM framework and estimator for dealing with missing values of an explanatory variable in linear regression analysis. The GMM estimator is efficient under assumptions needed for consistency of linear-imputation methods. The estimator, which also allows for a specification test of the missingness assumptions, is compared to existing linear imputation, complete data, and dummy variable methods commonly used in empirical research. The dummy variable method is generally inconsistent even when data are missing completely at random, and the dummy variable method, when consistent, can be less efficient than the complete data method.

We document that rotation group bias—the tendency for the unemployment rate to vary systematically by month in sample—in the Current Population Survey (CPS) has worsened over time. Estimated unemployment rates for earlier rotation groups have grown sharply relative to later rotation groups; both should be nationally representative samples. This bias increased discretely after the 1994 CPS redesign, and rising nonresponse rates are likely a significant contributor. Survey nonresponse increased after the redesign, mirroring the evolution of rotation group bias. Consistent with this explanation, rotation group bias for households that responded in all eight interviews remained stable over time.

We study a field experiment at a large technology company. Employees were encouraged to submit ideas on process and product improvements. The company randomly assigned nineteen teams into treatment and control groups. Treatment team employees received rewards if their ideas were approved. Nothing changed for control team employees. Our main finding is that rewards substantially increased the quality of ideas. Rewards increased participation in the suggestion system but decreased ideas per participating employee, with no net effect on the quantity of ideas. Broader participation persisted after the reward was discontinued, suggesting habituation. We find no evidence for motivational crowding out.

We provide the first evaluation of the effect of the U.S. minimum legal drinking age (MLDA) on nonfatal injuries. Using administrative records from several states and a regression discontinuity approach, we document that inpatient hospital admissions and emergency department (ED) visits increase by 8.4 and 71.3 per 10,000 person-years, respectively, at age 21. These effects are due mainly to an increase in the rate at which young men experience accidental injuries, alcohol overdoses, and injuries inflicted by others. Our results suggest that the literature’s disproportionate focus on mortality leads to a significant underestimation of the benefits of tighter alcohol control.

We assess whether welfare reform affects earnings only through mean impacts that are constant within but vary across subgroups. This is important because researchers interested in treatment effect heterogeneity typically focus on estimating mean impacts that only vary across subgroups. Using a novel approach to simulating treatment group earnings under the constant mean impacts within subgroup model, we find this model does a poor job of capturing treatment effect heterogeneity for Connecticut's Jobs First welfare reform experiment. Notably, ignoring within-group heterogeneitywould lead one to miss evidence that treatment effects are consistent with basic labor supply theory.