{smcl}
{hline}
help for {hi: decomlogit}
{hline}

{title:Discrimination and detailed decomposition in a logit model}

{p 8 15}
{cmd:decomlogit} {it:depvar varlist} [{cmd:if}] [{cmd:in}], {cmd:disc}({it:varname}) [{cmdab:d:etail}]


{p}
{it:depvar}:{space 10} Dependent binary indicator variable in the logit model.

{p}
{it:varlist}:{space 9} Indicator and continuous independent variables. 
{break}{space 17} Note, should not include the discrimination variable.

{p}
{cmd:disc}({it:varname}):{space 3} Discrimination variable, e.g. gender.

{title:Description}

{p}
{cmd:decomlogit} decomposes differences in probabilities across e.g. gender in a logit model into differences due to dicrimination and qualifications.

{p}
It applies the method described in Nielsen (1998).

{title:Options}

{p}
{cmd:detail} reports the specific contribution from each continuous variable for each type. If not specified only the accumulated contribution from all continuous variables is reported.

{title:Saved results}

{p}
{cmd:decomlogit} saves the following in {cmd:e()}:

{p}
{cmd:e(DQT)}{space 2} Vector containing difference due to discrimination, D, qualification, Q, and, total difference.

{p}
{cmd:e(D)}{space 4} Matrix containing specific contribution from indicator and continuous variables for each type. The last row contains totals.


{title:Method}
The method used to separate the contribution from the discrete and the continuous variables is based on total derivatives. 
The sample is split into a number of different types on the basis of the categories of the indicator variables.

{p}
Standard errors are calculated by the delta method.

{title:Example}

{p 2}
{cmd: . decomlogit emp married child experience, disc(female)}


{title:References}

{p}
Nielsen, H. S. (1998), Discrimination and Detailed Decomposition in a Logit Model. {it:Economic Letters} 61: 115-120.
 
{title:Authors}

{p}
Mikkel N. Hermansen

{p}
Helena Skyt Nielsen, University of Aarhus. Contact: hnielsen@econ.au.dk
{p_end}