12-17-2015 04:17 PM
Has anyone worked on Price Elasticity model at customer level? In Insurance Industry, insurance policies from a insurer that have been given a renewal offer in the period of 2 years are taken. Target Variable: (Binary) whether a customer renew an existing insurance contract or policy? I have read somewhere they generally use the treatment - the rate change: percentage change in
premium from the current to the new rate, categorized into ordered values. My question - How price elasticity would be calculated give it's a logistic regression model (binary)? How can model help in finding the customers which are more sensitive toward policy premium changes?
Any help would be highly appreciated!
12-21-2015 11:09 AM - edited 12-21-2015 03:56 PM
I never really thought about this before. Here's an idea for price elasticity:
a) compute the parameters of your logitstic regression.
For each person in your sample:
b) predict the probability that someone will renew at Price = Price based on that person's characteristics .This is quantity B
c) predict the probability that someone will renew at Price = Price - 0.5% based on that person's characteristics .This is quantity C.
d) predict the probability that someone will renew at Price = Price + 0.5% based on that person's characteristics .This is quantity D.
e)The quantity demanded for that person increases by (C-D)/ B percent for each 1% price drop.
Average the amount obtained in e) over your whole population to get average price elasticity.
------------- regarding the "finding which customers are more sensible to price changes", you would need to interact the "price" variable with other variables.
For example, you could have "price" variable and "price * binary male=1" variable If the coefficient for "price*male" is greater than 0, then men are more sensitive to price than women.