Computing elasticities for dummy variables

[Penjani Kamanga]

Dear all

I am very new to STATA.

I am looking for help (command prompt or other ways) in computing elasticities for dummy variables. My model has 10 variables, 6 of which are factors. When I tried to compute using postestimation>margins,

. margins, predict(at means) eyex( gndhhh legmono weed fertility pesdisea otherfert labor expleg extens landtot ) continuous


I got this error:

factor variables not allowed in option eyex()
r(198);

Your help will be most appreciated.

Thanks

Penjani

[John Mullahy]

Penjani: An elasticity for a 0-1 dummy variable is not well defined. An elasticity is the (partial) derivative of the log-outcome with respect to the log-covariate or, in essence, the percentage/proportional change in the outcome due to a given percentage/proportional change in the covariate. With a 0-1 dummy variable, the latter change is undefined (i.e. moving from 0 to 1).

With a dummy variable, however, you can compute semi-elasticities that amount to the percent change in the outcome resulting from a change (not a percentage change) in the covariate. In margins this means using eydx(…) instead of eyex(…). In general this will work with factor RHS variables.

[Penjani Kamanga]

Dear John

Thanks so much for your advice and elaborate response!