Estimating multiple hurdle models using Roodman cmp command in Stata

[Kaleb Jada]

Dear Statalist,

I am modeling a farmer decision problem that involve multiple stages(decision to engage in dairy production, decision to process milk into butter, decision to participate in butter market and decision as to how much to sell). I have already worked out the likelihood function but programming that into Stata proved to be beyond me. I know Roddman's cmp command handle a wide varieties of models but I could not figure out how to estimate hurdle models. Any help would be greatly appreciated!

Kind regards

[David Roodman]

Kaleb, is clearance of the first three hurdles only observed in certain cases, or in all cases?

If all the steps are fully observed, I suppose you would do something like:

cmp setup
cmp (engage_in_dairy = [xvars1]) (make_butter = [xvars2]) (quantity_sell = [xvars4]), ind($cmp_probit "engage_in_dairy*$cmp_probit" "make_butter*cond(quantity_sell, $cmp_cont, $cmp_left)")

This combines the final two stages into a single tobit equation.

--David

[Kaleb Jada]

David, Thank you for your help. I found your cmp command very flexible and lately I have been using it a lot.

1. To model cattle market participation with double sample selection
code:
cmp setup cmp (engage_in_cattle_fattening = [xvars1]) (particpate_cattlemarket = [xvars2]) (no_cattle_sold = [xvars3]), ind($cmp_probit "engage_in_cattle_fattening*$cmp_probit" "participate_cattlemarket*$cmp_cont") 2. Another model to estimate "trivariate probit" model with double sample selection like this

cmp setup cmp (took_credit = [xvars1]) (use_credit_onfarm = [xvars2]) (use_credit_on livestock = [xvars3]), ind($cmp_probit "took_credit*$cmp_probit" "use_credit_onfarm*$cmp_probit") Am I doing it right or should I re-read cmp documentation?

3.Also working on stochastic frontier analysis to analyze technical efficiency and identify source of inefficiency. But I would like to include selection equation. So how can I go about it using cmp?

Kind regards,
Kaleb

[Kaleb Jada]

David, Thank you for your help. I found your cmp command very flexible and lately I have been using it a lot.

1. To model cattle market participation with double sample selection

cmp setup
cmp (engage_in_cattle_fattening = [xvars1]) (particpate_cattlemarket = [xvars2]) (no_cattle_sold = [xvars3]), ind($cmp_probit "engage_in_cattle_fattening*$cmp_probit" "participate_cattlemarket*$cmp_cont") 

2. Another model to estimate "trivariate probit" model with double sample selection like this

cmp setup
cmp (took_credit = [xvars1]) (use_credit_onfarm = [xvars2]) (use_credit_on livestock = [xvars3]), ind($cmp_probit "took_credit*$cmp_probit" "use_credit_onfarm*$cmp_probit") 

Am I doing it right or should I re-read cmp documentation?

3.Also working on stochastic frontier analysis to analyze technical efficiency and identify source of inefficiency. But I would like to include selection equation. So how can I go about it using cmp?

Kind regards,
Kaleb

[David Roodman]

Kaleb, I'm not sure how you would do a stochastic frontier analysis. I haven't worked with those models, but I understand them to be mixed models in which the regime for each observation is itself modeled rather than known. cmp can't handle that...
--David