Posts Tagged ‘Heckman’


Missing Something?

February 11, 2010


Professor Matt Dull of Virginia Tech will present “Missing Something Important?: Using the Heckman Selection Model in Policy Research” on Wednesday, March 3rd from Noon – 1:30 in 309 Marvin Center. The Marvin Center is on the corner of 21st and H.

This talk will provide a nontechnical introduction to the use of the Heckman selection procedure to correct bias in the estimation of regression models due to nonrandom sample selection. Do you want to perform a regression analysis but worry your sample is biased due to missing observations that may be related to a variable of interest? The Heckman correction estimates a two stage model: first, a selection equation with a dichotomous dependent variable equaling 1 for observed and 0 for missing values; and second, an outcome equation predicting the model’s dependent variable. If correctly specified, the Heckman model produces unbiased parameter estimates and may even provide some useful information. Does your theory predict which cases may be missing? Drawing on two applications relevant to policy research – analysis of federal grant program applicants and analysis of survey data with a large number of “I don’t know” or “No-basis to judge” responses – I’ll discuss the Heckman technique as a potentially rich opportunity for (cautious) inference.

Matt Dull is Assistant Professor in the Center for Public Administration and Policy at Virginia Tech’s North Virginia campus. He received his Ph.D. in political science from the University of Wisconsin, Madison in 2006.