| Answers to many empirical economic questions typically involve quantifying the relationship between a set of explanatory variables and an outcome of interest. Such analyses provide useful statistics (e.g., marginal effects, treatment effects) or allow for meaningful predictions. Depending on the question, economists may rely on econometric models to provide additional information, namely the entire density of a random variable, in order to use different moments of the distribution or accurately capture tails of the distribution. This study examines the ability of several different econometric models to explain the distribution of an outcome. Using a Monte Carlo experiment, I evaluate different economic approaches that are frequently used by economists to deal with distributions that are positive, skewed and long tailed. Each econometric model is then evaluated for its performance in estimating the expected outcome and fitting the distribution particularly in the tails. The distribution of future medical care expenditure is used throughout the paper to exemplify how the shape of the distribution can affect optimal behavior, such as the purchase of health insurance, when future medical care expenses are uncertain. |
