Habitual and Occasional Lobbyers in the U.S. Steel Industry: An Em Algorithm Pooling Approach

Using U.S. steel firm data, we find that lobbying for import protection appears to be habit-forming. To identify heterogeneity in lobbying behavior among firms, we use an expectation-maximization algorithm to sort our firms into groups with different propensities to lobby and estimate the determinants of lobbying in each group. A two-pool model emerges: occasional lobbyers' lobbying depends on their market performance, and habitual lobbyers' lobbying only depends on past lobbying. The latter tends to be larger steel firms whose business is more focused in steel. Our evidence is consistent with dynamic economies of scale in protection seeking breeding protection-dependent firms.     

On a random coefficient probit model

Consider a panel-data and state-dependent binary model. The coefficient associated with the past status is assumed to he a normally distributed random variable. The estimation of unknown parameters is investigated for both fixed and random initials. Simulation results for the random initial conditions and empirical results for the U.S. steel industrypanel data are presented.     

A bivariate extension of the beta generated distribution derived from copulas

In this paper, we introduce a new class of bivariate distributions whose marginals are beta-generated distributions. Copulas are employed to construct this bivariate extension of the beta-generated distributions. It is shown that when Archimedean copulas and convex beta generators are used in generating bivariate distributions, the copulas of the resulting distributions also belong to the Archimedean family. The dependence of the proposed bivariate distributions is examined. Simulation results for beta generators and an application to financial risk management are presented.     

A Modification on the Friedman Test Statistic

A modification on the well-known, nonparametric Friedman test statistic is suggested in this article. Probability distributions of the suggested test statistic under the null hypothesis are tabulated for some small sample cases. In addition to an example, simulation results for various sample sizes are presented. The simulation indicates that the modified test statistic performs better than the Friedman test in detecting treatment effects of small differences especially when the sample size is small.