Advances and Applications in Statistics
© 2014 Pushpa Publishing House, Allahabad, India
Published Online: January 2015
Available online at http://www.pphmj.com/journals/adas.htm
Volume 43, Number 2, 2014, Pages 91-105
Received: April 8, 2014; Revised: September 23, 2014; Accepted: September 27, 2014
2010 Mathematics Subject Classification: 62J05, 62P25, 62-07.
Keywords and phrases: bivariate treatment effect, bivariate probit, multivariate Heckman
model, sample selection.
MEASURING BIVARIATE AVERAGE
TREATMENT EFFECT
Patrick Franco Alves and Gustavo T. L. da Costa
Department of Economics
Brasília University-UnB
Brasília, Brazil
e-mail: patrickfrancoalves@yahoo.com.br
Brazilian Institute of Geography and Statistics-IBGE
Rio de Janeiro, Brazil
e-mail: gustavo.costa@ibge.gov.br
Abstract
We present the methodology for the measurement of the average
treatment effects under a bivariate selection mechanism. The
formulation of the bivariate average treatment effect comes from the
multivariate sample-selection model, where the bivariate normal
distribution is necessary in order to derive the bivariate inverse Mills
ratio. Under this approach there are seven different average treatment
effects. An application case is done using a cross-section data for
Brazilian industrial firms. It is shown that this methodology can be
easily used in any bivariate self-selection mechanism case since there
is not an intricate computational solution for the problem.