Calculating price elasticity of demand using probit model: margins, dy/ex or ey/ex?

[Karl Yesler]

Hi Statalist,

I'm trying to calculate the price elasticity of demand for a good in a panel dataset (i.e. I am interested in the % change in the probability of buying a good given a % change in price of the good).

My question is regarding implementing this in Stata: after running my -probit- say,

probit bought price, vce(cluster mem_id)

Which post-estimation command gives me the interpretation I want? (Again: % change in the probability of buying a good given a % change in price of the good)

margins dyex(price)

*OR

margins eyex(price)

And why?


As an aside, I know that when the outcome is not a binary probability, but a continuous variable (amount bought), I should use eyex but, the probability is confusing me...

reg amt price,  vce(cluster mem_id)

*This should be right for a continuous variable to get % change in amount bought given % change in price
margins eyex(price)

Many Thanks!

[Jorge Eduardo Perez Perez]

I am interested in the % change in the probability of buying a good given a % change in price of the good

Sounds like eyex, but i would argue that the semielasticity, dyex is easier to interpret, simply read as "I am y percentage points more likely to buy the good if the price decreases by x %"

[Karl Yesler]

Hi Jorge,

Thanks for your comment. I think that was why it was confusing to me because under "eyex" the interpretation is a percentage change in the probability but the percent change depends on what the average probability (the denominator) is. Whereas under "dyex", as you suggested, it is a simple change in the probability not a percentage change in the probability.

However, (I believe) the economics literature conventionally uses percentage change notation for definitions of price elasticity of demand, therefore I may need to use eyex in order to compare my estimates to others...