Brian Dineen

Industrial Organization
Applied Econometrics
Economic History

Industrial Organization
Microeconomics
Economic History

Industrial Organization and Economic History (Oral), 2002
Microeconomic and Macroeconomic Theory (Written), 2001

Moral Hazard and Adverse Selection: Evidence from European Auto Insurance Markets

Professor Steve Berry
Professor Ben Polak
Professor Philip Haile
Dr. Michael Boozer

May 2005

M.Phil., Economics, Yale University, 2003
M.A., Economics, Yale University, 2002
B.A., History & Economics, University of Oxford, 1999 (Graduated first in degree program)

John F. Enders Fellowship, 2003
Yale Graduate Summer Fellowships, 2001 and 2002
Yale University Fellowship, 2000-2005
Hicks and Webb Medley Prizes, 1999, Graduated joint-first place in Economics Class
Gibbs Prize, 1999, Graduated first place in the History and Economics joint degree program
St. Anne’s College Academic Scholarships, 1997 and 1999

Intermediate Microeconomics, Fall 2004 and Spring 2003
Introduction to Finance, Summer 2004
Poverty under Post-Industrial Capitalism, Spring 2004
Econometrics and Data Analysis, Fall 2002

McDougal Coordinating Fellow, Yale University, 2002-2003 (selected by competitive interview)
McDougal Social Fellow, Yale University, 2001-2002 (selected by competitive interview)
Yale Business Management Study Group, 2004
Elected JCR (Undergraduate) President, St. Anne’s College, University of Oxford, 1997-1998

Do financial incentives, such as deductibles and no claims discounts, affect claim and accident rates in automobile insurance markets? My contribution to addressing this question exploits a unique feature of the UK auto insurance market to quantify the effect of financial incentives on driver behavior. In the UK most eligible drivers pay a small fee to "protect" their no claims discount. These policyholders can make a known number of claims within a given period before they suffer a very large discrete jump in their premium schedule. Under weak assumptions, namely convex utility cost of effort and concave utility, I show that a driver should increase costly claim prevention effort as the number of past claims increase. This result does not change when policyholders have the option of not reporting claims. In the UK, claims expire from a policyholder’s record after a predetermined number of years. When I incorporate this feature into the model, costly claim prevention effort still increases in the number of past claims, except in some extreme cases.

I test these predictions using proprietary data provided by one of the largest UK auto insurers and control for selection effects in two ways. Following Abbring, Chiappori and Pinquet (2003), I examine within-policyholder duration between claims. The large size of the dataset (approximately thirty thousand drivers have at least two accidents) is ideally suited to this test. However, this does not control for non-financial effects. For example a driver may change his behavior in response to a traumatic accident independent of the impact of financial incentives. I therefore propose a second method of controlling for selection that relies on certain types of claims being uninformative about the probability of other types of future claims. For example, a claim for damage incurred while the car was parked or a theft claim is uninformative about the policyholder’s driving ability. My initial results support Abbring, Chiappori and Pinquet’s finding that financial incentives do not affect policyholder driving behavior, although my results suggest that financial incentives do affect the decision to report claims. The results suggest that drivers change their driving behavior after accidents but only in response to non-financial considerations. These results have implications for regulatory authorities; they should focus on non-financial instruments in their efforts to lower the number of road accidents. Equally, my results have implications for insurers interested in developing innovative auto insurance contracts.

Using a second proprietary dataset, I examine patterns of asymmetric information in the Irish auto insurance market. This dataset contains all the policyholder characteristics observed by the insurance company and detailed data on claims. I find at most weak evidence that policyholders possess an information advantage over insurance companies. I find no evidence in the Irish market of the learning patterns found by Alma Cohen (2002) in the Israeli auto insurance market. The results suggest that information asymmetries may not be a large barrier to competition, even in markets with little information sharing between insurers.