ADVANCED ECONOMETRICS I Theory (3/3) Instructor: Joaquim J. S. Ramalho E.mail: jjsro@iscte-iul.pt Personal Website: http://home.iscte-iul.pt/~jjsro Office: D5.10 Course Website: https://jjsramalho.wixsite.com/advecoi Fรฉnix: https://fenix.iscte-iul.pt/disciplinas/03089
3. Discrete Choice Models 3.2. Models for Ordered Choices Ordered choices: Values for the dependent variable: ๐ โ 0,1, โฆ , ๐ โ 1 Latent model: โ = ๐ฆ ๐ โฒ ๐พ + ๐ฃ ๐ ๐ ๐ โช ๐ฆ ๐ cannot include an intercept Individual behaviour observed only by intervals: โ โค ๐ฟ 0 0 if ๐ ๐ โ โค ๐ฟ ๐ , ๐ if ๐ฟ ๐โ1 < ๐ 1 โค ๐ โค ๐ โ 2 ๐ ๐ = เต ๐ โ > ๐ฟ ๐โ2 ๐ โ 1 if ๐ ๐ โช Example: โ is a latent measure of the health status โ ๐ ๐ โ ๐ ๐ is an observed health indicator: poor, satisfactory, good, excellent Assumption: the ๐ฟ ๐ โs are not known 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 2
3. Discrete Choice Models 3.2. Models for Ordered Choices Probabilities: Aim: โ in a given interval โช Modelling the probability of observing ๐ ๐ Each probability is based on the same ๐ป โ functions used with binary choices, being given by: โ โค ๐ฟ ๐ |๐ฆ ๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐ = ๐๐ ๐ฟ ๐โ1 < ๐ ๐ โ โค ๐ฟ ๐ |๐ฆ ๐ โ ๐๐ ๐ โ < ๐ฟ ๐โ1 |๐ฆ ๐ = ๐๐ ๐ ๐ ๐ โฒ ๐พ + ๐ฃ ๐ โค ๐ฟ ๐ |๐ฆ ๐ โ ๐๐ ๐ฆ ๐ โฒ ๐พ + ๐ฃ ๐ < ๐ฟ ๐โ1 |๐ฆ ๐ = ๐๐ ๐ฆ ๐ โฒ ๐พ|๐ฆ ๐ โ ๐๐ ๐ฃ ๐ < ๐ฟ ๐โ1 โ ๐ฆ ๐ โฒ ๐พ|๐ฆ ๐ = ๐๐ ๐ฃ ๐ โค ๐ฟ ๐ โ ๐ฆ ๐ โฒ ๐พ โ ๐ป ๐ฟ ๐โ1 โ ๐ฆ ๐ โฒ ๐พ = ๐ป ๐ฟ ๐ โ ๐ฆ ๐ Hence, the general case is: โฒ ๐พ ๐ป ๐ฟ 0 โ ๐ฆ ๐ if ๐ = 0 โฒ ๐พ โ ๐ป ๐ฟ ๐โ1 โ ๐ฆ ๐ โฒ ๐พ if 1 โค ๐ โค ๐ โ 2 ๐ป ๐ฟ ๐ โ ๐ฆ ๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐ = เต โฒ ๐พ 1 โ ๐ป ๐ฟ ๐โ2 โ ๐ฆ ๐ if ๐ = ๐ โ 1 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 3
3. Discrete Choice Models 3.2. Models for Ordered Choices Estimation: Parameters to be estimated: โช ๐พ โช ๐ฟ 0 , โฆ , ๐ฟ ๐โ2 Estimation method: โช Maximum likelihood Most common models: โช Ordered logit โช Ordered probit Stata ologit Y ๐ 1 โฆ ๐ ๐ oprobit Y ๐ 1 โฆ ๐ ๐ 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 4
3. Discrete Choice Models 3.2. Models for Ordered Choices Partial effects: Each ๐ ๐ affects ๐ probabilities: โ๐ ๐ = 1 โน โ๐๐ ๐ = ๐ ๐ โฒ ๐พ โ๐พ ๐ ๐ ๐ฟ 0 โ ๐ฆ ๐ if ๐ = 0 โฒ ๐พ โ ๐ ๐ฟ ๐โ1 โ ๐ฆ ๐ โฒ ๐พ ๐พ ๐ ๐ ๐ฟ ๐ โ ๐ฆ ๐ if 1 โค ๐ โค ๐ โ 2 = เต โฒ ๐พ ๐พ ๐ ๐ ๐ฟ ๐โ2 โ ๐ฆ ๐ if ๐ = ๐ โ 1 The sign of ๐พ ๐ is informative about the direction of โ๐๐ ๐ = 0 ๐ and โ๐๐ ๐ = ๐ โ 1 ๐ but not of the changes in the remaining probabilities 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 5
3. Discrete Choice Models 3.3. Models for Multinomial Choices Multinomial choices: Values for the dependent variable: ๐ โ 0,1, โฆ , ๐ โ 1 Latent model: โช Each individual has a given utility associated with each alternative: โฒ ๐พ + ๐ฃ ๐๐ ๐ ๐๐ = ๐ฆ ๐๐ โช The selected alternative is the one that maximizes utility: ๐๐ ๐ ๐ = ๐|๐ฆ ๐ = ๐๐ ๐ ๐๐ = ๐๐๐ฆ ๐ ๐1 , โฆ ๐ ๐๐ |๐ฆ ๐ Main models: โช Multinomial Logit: ๐ ๐๐ ~ ๐ป๐ฃ๐๐๐๐ and ๐ ๐๐ independent โ๐ โช Multinomial Probit: ๐ ๐๐ ~ ๐๐๐ ๐๐๐ โช Nested Logit โช Random Parameters Logit 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 6
3. Discrete Choice Models 3.3. Models for Multinomial Choices Explanatory variables: ๐ฆ ๐๐ may include: โช ๐ฆ ๐๐ : variables that are different across individuals and alternatives โช ๐ฆ ๐ : variables that differ across alternatives but not individuals โช ๐ฆ ๐ : variables that differ across individuals but not alternatives Example: โช ๐ ๐ - selected means of transport to go to work โช ๐ฆ ๐๐ - time that each individual ๐ takes in going to work when using transport ๐ โช ๐ฆ ๐ - price of transport ๐ โช ๐ฆ ๐ - age of individual ๐ 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 7
3. Discrete Choice Models 3.3. Models for Multinomial Choices Multinomial logit: โฒ ๐พ+๐ฆ ๐ โฒ ๐พ ๐ ๐ ๐ฆ ๐๐ โฒ ๐พ + ๐ฆ ๐ โฒ ๐พ ๐ = ๐๐ ๐ ๐ = ๐|๐ฆ ๐๐ = ๐ป ๐ ๐ฆ ๐๐ โฒ ๐พ+๐ฆ ๐ โฒ ๐พ ๐ ๐โ1 ๐ ๐ฆ ๐๐ ฯ ๐=0 ๐พ ๐ has to be normalized, that is for one alternative (base outcome) its value is set to zero ๐พ cannot include a constant term Independence of Irrelevant Alternatives (IIA) โ the odds ratio between two alternatives does not depend on the remaining alternatives: โฒ ๐พ+๐ฆ ๐ โฒ ๐พ ๐ = ๐ ๐ฆ ๐๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐๐ โฒ ๐พ+๐ฆ ๐ โฒ ๐พ ๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐๐ ๐ ๐ฆ ๐๐ 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 8
3. Discrete Choice Models 3.3. Models for Multinomial Choices When all explanatory variables are of the type ๐ฆ ๐๐ and ๐ฆ ๐ , the choice between alternatives ๐ and ๐ is fully explained by diferences in the alternative characteristics: โฒ ๐พ = ๐ ๐ฆ ๐๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐๐ โฒ โ๐ฆ ๐๐ โฒ ๐พ โฒ ๐พ = ๐ ๐ฆ ๐๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐๐ ๐ ๐ฆ ๐๐ โช Is this case, the model is often called โ conditional logit โ Stata asclogit Y ๐ 1๐ โฆ, case( id ) alternatives( varname ) casevars( ๐ ๐ โฆ) basealternative( name ) When all explanatory variables are of the type ๐ฆ ๐ , the choice between alternatives ๐ e ๐ is fully explained by diferences between ๐พ ๐ e ๐พ ๐ : โฒ ๐พ ๐ = ๐ ๐ฆ ๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐ โฒ ๐พ ๐ โ๐พ ๐ โฒ ๐พ ๐ = ๐ ๐ฆ ๐ ๐๐ ๐ ๐ = ๐|๐ฆ ๐ ๐ ๐ฆ ๐ Stata mlogit Y ๐ 1 โฆ ๐ ๐ , baseoutcome( 0 ) 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 9
าง 3. Discrete Choice Models 3.3. Models for Multinomial Choices Estimation: โช Maximum likelihood based on the following log-likelihood function: ๐ โฒ ๐พ + ๐ฆ ๐ โฒ ๐พ ๐ ๐๐ = เท ๐ ๐๐ ๐๐๐ ๐ป ๐ ๐ฆ ๐๐ ๐=1 โช ๐ ๐๐ = 1 if individual ๐ chooses alternative ๐ Partial effects: โช โ๐ ๐๐ = 1 โน โ โ๐๐ ๐ ๐ = ๐ ๐ = ๐พ ๐ ๐ป ๐ โ ๐ ๐๐ โ ๐ป ๐ โ โ ๐พ ๐ gives the sign of the partial effect โช โ๐ ๐ = 1 โน ๐โ1 ๐พ ๐ ๐ป ๐ โ โ โ๐๐ ๐ ๐พ = ฯ ๐=1 ๐ = ๐ ๐ = ๐ป ๐ โ ๐พ ๐ โ ๐พ , onde าง โ ๐พ ๐ gives the sign of the partial effect relative to the base alternative, not the sign of the overall effect 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 10
3. Discrete Choice Models 3.3. Models for Multinomial Choices Testing IIA โช Hausman test comparing: โ Full multinomial logit model โ Multinomial logit model excluding one or more alternatives โช If multinomial logit is the correct model, then both models produce consistent estimators (null hypothesis) โช If multinomial logit is not the correct model, then the results generated by both models will be different (alternative hypothesis) Stata mlogit Y ๐ 1 โฆ ๐ ๐ , baseoutcome( 0 ) (ou asclogit โฆ) estimates store Mod1 mlogit Y ๐ 1 โฆ ๐ ๐ if Y != 3 , baseoutcome( 0 ) (ou asclogit โฆ) estimates store Mod2 hausman Mod1 Mod2 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 11
3. Discrete Choice Models 3.3. Models for Multinomial Choices Multinomial probit: Not affected by the IIA property Very complex, requiring the computation of ๐ โ 1 integrals The version implemented in Stata assumes independent errors, which eliminates the only advantage of multinomial probit over multinomial logit Stata mprobit Y ๐ 1 โฆ ๐ ๐ , baseoutcome( 0 ) 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 12
3. Discrete Choice Models 3.3. Models for Multinomial Choices Nested logit: Not affected by the IIA property, grouping the choices in several sets in such a way that: โช Within each group, alternatives may be correlated โช Between groups, alternatives are independent Results from a sequential decision process โ example for a two-level process: Financing Bank debt Own funds Stock market Bank 1 Bank N Lisbon Frankfurt Stata โช Level 1 โ defining J groups, nlogit โฆ โช Level 2 โ defining ๐ ๐ choices in each group 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 13
3. Discrete Choice Models 3.3. Models for Multinomial Choices Random parameters logit: Latent model: โฒ ๐พ ๐ + ๐ฃ ๐๐ ๐ ๐๐ = ๐ฆ ๐๐ Most common assumption: ๐พ ๐ ~ ๐ ๐พ, ฮฃ ๐พ Not affected by the IIA property If ฮฃ ๐พ = 0 , it reduces to the Multinomial Logit model; hence, comparing the two models allows the IIA property to be tested 2020/2021 Joaquim J.S. Ramalho Advanced Econometrics I 14
Recommend
More recommend