Close

Complete Captcha

• 1 month - 13 posts - Latest - RSS
• Thread: 1 Goods vs 0 No Goods

1. Economist
   646a
   

   I want to to test whether there is a gender gap in the likelihood of a claim. I have micro data with millions of individuals, their characteristics, and whether they filled for a claim in the last 12 months. The data is from several insurance companies.
   I was thinking of running a logit regression as follow:

   claim=Logit(b0+b1female +B2'X2+B3'X3)

   Where X2 is a vector of individual characteristics, and X3 is a vector of firm characteristics.

   My questions:
   1. Do I need to control for other variables, such as time trends, macro variables, number of cars in the county etc.
   2. Do I have an endogeniety bias? I was thinking that people that like to take risks might be correlated with one of the variables in X2
   3. Any paper out there you would recommend reading?

   1 month ago # QUOTE 0 Volod 0 Vlad !

2. Economist
   1829
   

   You should take care to control for people who make more than one claim. They’re going to be weird.

   1 month ago # QUOTE 0 Volod 0 Vlad !

3. Economist
   646a
   

   Maybe then to use count data analysis

   1 month ago # QUOTE 0 Volod 0 Vlad !

4. Economist
   1829
   

   Maybe then to use count data analysis

   Industry uses zero inflated poisson to model hazard. I don’t know why that and not something else

   1 month ago # QUOTE 0 Volod 0 Vlad !

5. Economist
   5953
   

   I want to to test whether there is a gender gap in the likelihood of a claim. I have micro data with millions of individuals, their characteristics, and whether they filled for a claim in the last 12 months. The data is from several insurance companies.
   I was thinking of running a logit regression as follow:
   claim=Logit(b0+b1female +B2'X2+B3'X3)
   Where X2 is a vector of individual characteristics, and X3 is a vector of firm characteristics.

   If you have micro-data at the individual level, why for control for firm characteristics which don't vary per individual? It's like economists never heard of multilevel models.

   1 month ago # QUOTE 1 Volod 0 Vlad !

6. Economist
   0c6c
   

   I want to to test whether there is a gender gap in the likelihood of a claim. I have micro data with millions of individuals, their characteristics, and whether they filled for a claim in the last 12 months. The data is from several insurance companies.
   I was thinking of running a logit regression as follow:
   claim=Logit(b0+b1female +B2'X2+B3'X3)
   Where X2 is a vector of individual characteristics, and X3 is a vector of firm characteristics.

   If you have micro-data at the individual level, why for control for firm characteristics which don't vary per individual? It's like economists never heard of multilevel models.

   People change auto insurance cc companies every six months.

   Did you even think about that?
   What a noob

   1 month ago # QUOTE 0 Volod 0 Vlad !

7. Economist
   c3d3
   

   I want to to test whether there is a gender gap in the likelihood of a claim. I have micro data with millions of individuals, their characteristics, and whether they filled for a claim in the last 12 months. The data is from several insurance companies.
   I was thinking of running a logit regression as follow:
   claim=Logit(b0+b1female +B2'X2+B3'X3)
   Where X2 is a vector of individual characteristics, and X3 is a vector of firm characteristics.

   If you have micro-data at the individual level, why for control for firm characteristics which don't vary per individual? It's like economists never heard of multilevel models.

   The regression compares the effect of the companies characteristics on the likelihood of a claim. It's not the same individual over time, but instead is a comparison in the risk of a clam given company A vs company B.

   1 month ago # QUOTE 0 Volod 0 Vlad !

8. Economist
   5953
   

   The regression compares the effect of the companies characteristics on the likelihood of a claim. It's not the same individual over time, but instead is a comparison in the risk of a clam given company A vs company B.

   OP has to clarify whether his data is longitudinal.

   1 month ago # QUOTE 0 Volod 0 Vlad !

9. Economist
   d89e
   

   I want to to test whether there is a gender gap in the likelihood of a claim. I have micro data with millions of individuals, their characteristics, and whether they filled for a claim in the last 12 months. The data is from several insurance companies.
   I was thinking of running a logit regression as follow:
   claim=Logit(b0+b1female +B2'X2+B3'X3)
   Where X2 is a vector of individual characteristics, and X3 is a vector of firm characteristics.

   If you have micro-data at the individual level, why for control for firm characteristics which don't vary per individual? It's like economists never heard of multilevel models.

   People change auto insurance cc companies every six months.
   Did you even think about that?
   What a noob

   Who does this besides bad drivers?

   1 month ago # QUOTE 0 Volod 0 Vlad !

10. Economist
    646a
    

    The data is longitudinal. I find that adding macro variables for example, can also be significant. That is, if there are many cars on the road, the likelihood of a car accident is increasing. I was hoping to find a "bible" article that will outline the issues and the solutions.

    1 month ago # QUOTE 0 Volod 0 Vlad !

11. Economist
    5953
    

    The data is longitudinal.

    There you go. Now brush up your metrics game. There's 100x more interesting models to build with this data than throwing it all on one pseudo cross-sectional pile and estimate a puny logit.

    1 month ago # QUOTE 0 Volod 0 Vlad !

12. Economist
    646a
    

    Like what? I am asking for a specific model that is preferred!

    1 month ago # QUOTE 0 Volod 0 Vlad !

13. Economist
    5953
    

    Like what? I am asking for a specific model that is preferred!

    https://doi.org/10.1017/asb.2021.39

    1 month ago # QUOTE 0 Volod 0 Vlad !

Reply

Post

Send Post »

Markup: a blockquote code em strong ul ol li.