[Avtandil Abashishvili]

Dear Prof. Joao Santos Silva ,

Thank you very much for your help and prompt reply, I will run the command today. I assume it needs a lot of time as well.

I run ppml without product dummies and it works. I think in the both cases (poisson and ppml) problem arises when I include product fixed effects.

Best Regards,

Avto

[JIA JIA LIN]

Dear Joao, sorry to bother you. I am a student and I am writing my dissertation. I want to analyze internal migration in China by using the gravity model, there are some questions: (1) I just have one-year data with 900 observations, I am not sure if this kind of cross-section data could use ppml or not. (2) if I could use ppml, what I clustvar could use? (3) the dummy variables always dropped by stata, I do not know how to deal with it. (4) If my model fails to pass RESET test, if it means I need to change or add some other variables? Kind regards， LIN

[JIA JIA LIN]

Dear Joao, sorry to bother you. I am a student and I am writing my dissertation. I want to analyze internal migration in China by using the gravity model, there are some questions: (1) I just have one-year data with 900 observations, I am not sure if this kind of cross-section data could use ppml or not. (2) if I could use ppml, what I clustvar could use? (3) the dummy variables always dropped by stata, I do not know how to deal with it. (4) If my model fails to pass RESET test, if it means I need to change or add some other variables? Kind regards， LIN

[Joao Santos Silva]

Dear JIA JIA LIN,

1) Yes, you can use ppml
2) People often cluster by distance
3) I do not know what that dummy represents, but if Stata drops it that means it is collinear with something else and therefore not needed.
4) It does not mean that, but you can try that.

Best wishes,

Joao

[JIA JIA LIN]

Dear Prof. Joao Santos Silva,

Thank you very much for your help and reply, and these suggestions are really helpful for me.

Best Regards,
Lin

[JIA JIA LIN]

Dear Prof. Joao Santos Silva,
sorry to bother you.
I just have a question about PPML. Could the PPML used in other model except gravity model?
I want to analyze the determination of the rural-urban migration in China, and I think todaro model is better than gravity model to measure this question. In the todaro model, ， where m is the migration rate from rural to urban, and it equals to the ratio the number of migrants from rural to urban, and the labor force in rural. The rate of migration is a function related to the (1) the probability that an urban laborer could find a modern-sector job, which could be written as positive monotonic function of the current urban employment rate or a negative function of the urban unemployment rate, , which is the urban labor force and is the level of urban modern sector employment. (2) the urban-rural income differential, which can be expressed as where and are the average real incomes of individuals employed in the urban and the rural, and . (3) other factors, Z，like distance, personal contacts, urban amenities and so on which could be considered as the costs and benefits of migration in origin and destination.
I just want to know that could I use the PPML for todaro model?

with many thanks and regards,
JIA JIA LIN

[Joao Santos Silva]

Dear JIA JIA LIN,

PPML can be used to estimate any multiplicative model such as the gravity equation and the Cobb-Douglas production function. I am afraid I an not familiar with the details of the Todaro model and I cannot really read the expressions in your post, so I can comment on this particular case.

Best wishes,

Joao

[JIA JIA LIN]

Dear Prof. Joao Santos Silva,
I am really grateful that you answer my question and so sorry the problem of the expressions.

The Todaro model introduced that the migration rate is depend on the expected income differential, and the cost of migration. I will use the equation that the dependent variable is migration rate, and the independent variables includes the expected income differential, distance, urbanizaition rate and so on. In this position, I could use ppml，is it correct?

with many thanks and regards,
JIA JIA LIN

[Joao Santos Silva]

What is the functional form of the model?

[JIA JIA LIN]

Dear Prof. Joao Santos Silva,

_{m=f(q,p ,z), where m is the migration rate, q is the real incom differential, p is ths probability of individual to find a job in urban, and z include some cost variables like distance, et al.
then I write down the expression:m=a1q+a2p+a3z+e, where e is the error term.}

with many thanks and regards,
JIA JIA LIN

[Az Us]

Dear Prof. Santos Silva,

I am estimating the Gravity model for one country. Is it somehow different than estimating this model for several countries?

Is there an analogue of Haussman test for PPML estimator to understand random or fixed effect model should I go for?

kind regards,

Azamat

[Joao Santos Silva]

Dear JIA JIA LIN,

That is a linear model, do not estimate by PPML. PPML would be suitable for m = exp(a1q+a2p+a3z) + e, where q, p, and z could be in logs.

Best wishes,

Joao

[JIA JIA LIN]

Dear Prof. Joao Santos Silva,

_{Actually, the stadard todaro model is just m=f(q,p ,z), the equation m=a1q+a2p+a3z+e is what I think to use, maybe it is wrong. I think m=f(q,p ,z) is not the linear model and then i could use ppml, is it correct?}


with many thanks and regards,
JIA JIA LIN

[Joao Santos Silva]

Dear JIA JIA LIN,

You can use PPML is f(.) is the exponential function (which seems to make sense).

Best wishes,

Joao

[JIA JIA LIN]

Dear Prof. Joao Santos Silva, Thank you very much for your help and reply, and these suggestions are really helpful for me. with many thanks and regards, JIA JIA LIN