Estimating a conditional logit in willingness to pay (WTP) space

[Tom Ford]

Hello,
I am running a willingness to pay study, and I would like to calculate willingness to pay estimating using a conditional logit it in WTP space. Does anyone know if there is a specialized command that is able to do it?

I am able to calculate WTP in preference space, but I am struggling to find a way to calculate it in WTP space.

This is how I would do it in preference space:

* download dataset
use http://fmwww.bc.edu/repec/bocode/t/traindata.dta, clear
* estimate conditional logit in preference space
clogit y price contract local wknown, group(gid) 

Conditional (fixed-effects) logistic regression         Number of obs =  4,780
                                                        LR chi2(4)    = 288.28
                                                        Prob > chi2   = 0.0000
Log likelihood = -1512.4797                             Pseudo R2     = 0.0870

------------------------------------------------------------------------------
           y | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
       price |   .0671638   .0074337     9.04   0.000     .0525939    .0817336
    contract |   -.101055   .0150581    -6.71   0.000    -.1305683   -.0715417
       local |   1.097354   .0859804    12.76   0.000     .9288358    1.265873
      wknown |   .7221119   .0772121     9.35   0.000      .570779    .8734449
------------------------------------------------------------------------------


* calculate WTP
wtp price contract local wknown

       contract       local      wknown
wtp   1.5046055  -16.338488  -10.751511
 ll   .96752512   -20.49017   -13.94995
 ul   2.0416859  -12.186806  -7.5530713

I know that there is a specialized command that computes a mixed logit model in WTP space (by Arne Risa Hole ), but I cannot found the corresponding one for the simpler conditional logit model

gen mprice=-1*price
mixlogitwtp y, group(gid) id(pid) price(mprice) rand(contract local wknown) nrep(500)


Mixed logit model in WTP space                          Number of obs =  4,780
                                                        Wald chi2(4)  = 399.29
Log likelihood = -1403.9957                             Prob > chi2   = 0.0000

------------------------------------------------------------------------------
           y | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
Mean         |
    contract |     -2.351    .275658    -8.53   0.000    -2.891279    -1.81072
       local |   14.10279   1.680523     8.39   0.000     10.80903    17.39656
      wknown |   10.08263   1.559899     6.46   0.000     7.025284    13.13998
       price |  -2.458695   .1274281   -19.29   0.000     -2.70845   -2.208941
-------------+----------------------------------------------------------------
SD           |
    contract |   3.244338   .4760072     6.82   0.000     2.311381    4.177295
       local |    14.6302   2.159685     6.77   0.000     10.39729     18.8631
      wknown |   7.703539   1.304865     5.90   0.000      5.14605    10.26103
       price |   .7874773   .1222439     6.44   0.000     .5478836    1.027071
------------------------------------------------------------------------------
The sign of the estimated standard deviations is irrelevant: interpret them as

If anyone could help with this I would be extremely grateful

[Mikky Mouse]

Hi Tom, did you find a way out of this? I need similar insight on how you did it. Please, share if you have. Thanks

[Hong Il Yoo]

Mikky Mouse: You don't need a separate command to estimate the conditional logit model in the WTP space. In the conditional logit model case, the WTP space specification is a direct re-parameterisation of the preference space specification: the WTP space estimates are numerically identical to what you can obtain by applying -nlcom- to the preference model estimates. The distinction between the WTP space vs the preference space matters in the mixed logit case because going from one space to the other is not simply a matter of re-parameterisation, but also involves making a different set of distributional assumptions.

[Mikky Mouse]

@Hong Il Yoo, Thank you very much for your response. It was very much helpful.