Prev: How to run a loop?
Next: How to merge with duplicate observation on both sides
From: H S on 21 Mar 2010 18:17
Hi,

Sorry this is more of a statistical question .

I have data on the specific date when states adopt mandatory insurance
laws and I also have survey data on car dealerships. I want to
investigate the whether dealers profit increase after adoption of
mandatory insurnace state laws. The problem is that the survey data of
dealers are CENSORED at different CUTOFF values through out the years.
For example, in 1947 , all cardealers with more than 500 car sales are
reported in sample
in 1965, all cardealers with more than 3000 car sales are reported in
the sample.
in 1974, all cardealers with more than 7000 car sales are reported in
the sample.

I probably also want to put in YEAR-STATE dummies to capture fixed
effects of different years and states.

I wonder if my statisitcal analysis is valid? If there is a constant
cut-off value I would have used TOBIT. However, the situation is more
complex due to the fact that the cut-off value changes over time. I
could have normalized the cutoff to the maximum value but that means I
would have to drop a lot of data points which is already scarce. I
wonder what is the right statistical specification that would answer
my research question, which is the effect adoption of mandatory
insurnace state laws on dealer's profit.
 | 
Pages: 1
Prev: How to run a loop?
Next: How to merge with duplicate observation on both sides