Abstract: This paper studies a unique panel dataset of transactions with
repeat customers of an insurer operating in a market in which insurers are
not required by law or contract to share information about their customers�
records. I use this dataset to test the asymmetric learning hypothesis that
sellers obtain over time private information that some of their repeat
customers have low risk, and that this learning enables sellers to make
higher profits in transactions with these repeat customers. Consistent with
this hypothesis, I find that the insurer in my dataset makes higher profits in
transactions with repeat customers and that these profits are driven by
transactions with repeat customers with good past claims history with the
insurer; that these higher profits result from repeat customers with good
claim history receiving a reduction in premiums that is lower than the
reduction in expected costs associated with such customers; and that
policyholders with bad claim history are more likely to flee their record by
switching to other insurers.