[Carlo Lazzaro]

Rodrigo:
1) dependent variable: it is more effiecient to group all the dummies in una unique, four-level, categorical variable
2) if the level of the dependednt variable are ranked, go -ologit-; it they are not ranked, go -mlogit-.
3) the predictor is correctly coded; However., it always strikes me seeing a simple (ie, whit an unique predictor only) regression model, as the risk of misspecification is usually high (and, if so, the outcome is not reliable).

[Rodrigo Abed]

Great Carlo, thanks for your help! Yes, I agree with comment number 3. I guess that, since I am only interested in checking if there is a significant difference with in the dependent variable based on the type of ownership, I did not include other variables. But it wouldn't hurt to see how the significance level changes if I were to add more predictors.