Research

Your abstract goes here.

Your abstract goes here.

I study the efficiency of firms' production decisions when these decisions generate costs for consumers with moral concerns. Specifically, I study firms' choices between ethical and unethical production regimes when compared to an efficient benchmark. Firms only observe consumers' behavioral responses to moral concern, i.e., the change in their demand, while the efficient benchmark would value the level of moral harm itself. Two economies that differ in their level of moral harm faced by consumers may induce the same regime choice by firms even when the efficient benchmark would differ. I show that this leads to firms generically making inefficient regime choices when consumers display intermediate levels of moral concern, while they choose efficiently for high and low levels of moral concern. When moral concern rises, demand can become less elastic, letting firms generate greater markups in the unethical regime, which causes firms to persist with unethical production even when inefficient. In contrast, if the fall in demand due to increased moral concern is high enough, firms may choose to produce ethically even when the cost saving from unethical production makes switching to ethical production inefficient. A corrective Pigouvian tax or subsidy can restore efficiency in principle, but its rate cannot be identified from market data alone.

Existing research in differential privacy, whose applications have exploded across functional areas in the last few years, describes an intrinsic trade-off between the privacy of a dataset and its utility for analytics. Resolving this trade-off critically impacts potential applications of differential privacy to protect privacy in datasets even while enabling analytics using them. In contrast to the existing literature, this paper shows how differential privacy can be employed to precisely—not approximately—retrieve the analytics on the original dataset. We examine, conceptually and empirically, the impact of noise addition on the quality of data analytics. We show that the accuracy of analytics following noise addition increases with the privacy budget and the variance of the independent variable. Also, the accuracy of analytics following noise addition increases disproportionately with an increase in the privacy budget when the variance of the independent variable is greater. Using actual data to which we add Laplace noise, we provide evidence supporting these two predictions. We then demonstrate our central thesis that, once the privacy budget employed for differential privacy is declared and certain conditions for noise addition are satisfied, the slope parameters in the original dataset can be accurately retrieved using the estimates in the modified dataset of the variance of the independent variable and the slope parameter. Thus, differential privacy can enable robust privacy as well as precise data analytics.

This paper analyzes techniques to enable differential privacy by adding Laplace noise to healthcare data. First, as healthcare data contain natural constraints for data to take only integral values, we show that drawing only integral values does not provide differential privacy. In contrast, rounding randomly drawn values to the nearest integer provides differential privacy. Second, when a variable is constructed using two other variables, noise must be added to only one of them. Third, if the constructed variable is a fraction, then noise must be added to its constituent private variables, and not to the fraction directly. Fourth, the accuracy of analytics following noise addition increases with the privacy budget, ϵ, and the variance of the independent variable. Finally, the accuracy of analytics following noise addition increases disproportionately with an increase in the privacy budget when the variance of the independent variable is greater. Using actual healthcare data, we provide evidence supporting the two predictions on the accuracy of data analytics. Crucially, to enable accuracy of data analytics with differential privacy, we derive a relationship to extract the slope parameter in the original dataset using the slope parameter in the noisy dataset.