Code - Nonlinear Decomposition Technique: I created a decomposition technique for logits, probits and other nonlinear models based on the linear Blinder-Oaxaca decomposition technique. The decomposition technique is useful for explaining the causes of racial, gender and other forms of inequality in economic, education, health and other outcomes. It has been widely used in thousands of studies across numerous fields and disciplines including economics, political science, psychology, sociology, social sciences, health, medicine, demography, environmental science, business and law. It has also been used to study differences, disparities or gaps in outcomes across many dimensions such as time periods, countries, states, and institutions.
Stata Command: fairlie (ssc install fairlie, replace)
Sample R Code
Sample SAS Code (original method)
Sample SAS Code (random ordering of variables)
Code - One-Observation per Peer Group (OOPG) Technique: We created a method for estimating peer effects for the outcome-on-outcome model and balance check model when urn sizes are fixed. Both models present challenges to estimation, and we created a straightforward estimation technique to address these challenges.
Sample Stata Code
Data - Comprehensive Startup Panel (CSP): I have been involved in several ongoing efforts to measure and create national datasets on entrepreneurship, startups, business ownership and self-employment. The most ambitious project is with the U.S. Census Bureau to capture information on the universe of U.S. startups including both employer and non-employer startups since 1996.
Data - Indicators of Early-Stage Entrepreneurship (KESE): I have been working with the Kauffman Foundation to create indicators of entrepreneurial activity since 2005. The KESE is a set of measures that represents new business creation in the United States, integrating several high-quality, timely sources of information on early-stage entrepreneurship.