you can’t have a proportion as the dependent variable even though the same formulas and estimation techniques would be appropriate with a proportion. As Wooldridge notes, many Stata commands (logit, probit, hetprob) could analyze DVs that are proportions, but they impose the data constraint that the dependent variable must be coded as either 0 or 1, i. This handout will discuss a few different ways for analyzing such dependent variables: fractional response models (both heteroskedastic and non- heteroskedastic) and zero one-inflated beta models.
Baum (2008) gives as examples the share of consumers’ spending on food, the fraction of the vote for a candidate, or the fraction of days when air pollution is above acceptable levels in a city. Wooldridge (1996, 2011) gives the example of the proportion of employees that participate in a company’s pension plan. In many cases, the dependent variable of interest is a proportion, i. Since fracglm is still in beta form, there may be changes in the future. 299- stata-journal/article.html?article=st NOTE: Material in handout is current as of April 9, 2017. “How does one do regression when the dependent variable is a proportion?” ats.ucla/stat/stata/faq/proportion.htm “Stata Tip 63: Modeling Proportions” Kit Baum, The Stata Journal, Volume 8 Number 2: pp. “How do you fit a model when the dependent variable is a proportion?” stata/support/faqs/statistics/logit-transformation/. Wooldridge, Journal of Applied Econometrics, Vol. Analyzing Proportions: Fractional Response and Zero One Inflated Beta Models Richard Williams, University of Notre Dame, “Econometric Methods for Fractional Response Variables with an Application to 401 (K) Plan Participation Rates” Leslie E.