Empirical Evidence of ex post Moral Hazard in the Australian Market for Comprehensive Automobile Insurance

Transcription

1 Empirical Evidence of ex post Moral Hazard in the Australian Market for Comprehensive Automobile Insurance Rowell, D 1,3., Nghiem H.S 2., & Connelly L. 2,3 1. Australian Centre for Economic Research on Health (ACERH-UQ) 2. Centre of National Research on Disability and Rehabilitation Medicine (CONROD) 3. The School of Economics, University of Queensland

2 Outline 1. History of the term Moral hazard The economists perspective The Insurer s perspective 2. Differentiate ex post from ex ante moral hazard 3. Data Collection Description 4. Econometric models 5. Results 6. Conclusion

3 Introduction Costs of Australia RTC s (1996) $million % Human Costs Medical/ambulance/rehabilitation Long-term care Labour in the workplace Labour in the household Quality of life Legal Correctional services Workplace disruption Funeral 3 0 Coroner 1 0 Total Vehicle costs Repairs Unavailability of vehicles Towing Total General costs Travel delays Insurance administration Police Non-vehicle property damage Fire and emergency Total Overall total * Source Bureau of Transport Economics (2000, P. XI)

4 Contrasting Views of Moral Hazard Moral Hazard is Idiom An idiom A group of word established by usage as having a meaning not deducible from those individual words (OED) Not all teething problems require

5 Genesis of the idiom The phrase moral hazard is an first appeared in the Practice of Fire Underwriting in 1865 to describe the...the danger proceeding from motives to destroy property by fire, or permit its destruction. (Ducat 1865) Hazard In 1167 the word hazard entered the English language from France to describe a game of dice. In 1618, the word hazardous was first used to convey a sense of perilous (Harper 2001) Moral The word moral entered the English language via a Latin root mos to mean manner,custom,way Thus in the medieval period the word moral was not always used in the contemporary sense For example Moral scientists studied the behaviour of rational individuals to establish rules to guide behaviour (Daston 1983). Thus, the creation of moral hazard suited the times; from hazard with its moral overtones of danger; and from moral the moral scientists who made chaste use of the odds (Baker 1996).

6 Contrasting Views of Moral Hazard Moral Hazard: The Economists Perspective Focused on the role of incentives Insurance is good, but security as well as hazard may have an unfavourable effect upon industry Intensity of effort is diminished Carelessness is encouraged by insurance is the encouragement it gives to dishonesty (Haynes 1895) The problem of moral hazard in insurance has, in fact, little to do with morality but can be analysed with orthodox economic tools. (Pauly 1968)

7 Moral Hazard: Insurer s Perspective Definitions from early 20 th century Men who steal or lie [or] magnify a slight injury (McNeill 1900) Misrepresentation and negligence (Campbell 1902) Moral hazard was typically attributed to the (immoral) personal characteristics of individuals, authorities claimed that it was more likely among certain ethnic and social groups like Central Americans, Assyrians, drug addicts and homosexuals. (Rupprecht 1940; Remington and Hurren 1935; Shepherd and Webster 1957) Definitions from late 20 th century an imputed subjective characteristic of the insured that increases the probability of loss (Mehr and Cammack 1976) moral hazards, such as dishonesty, carelessness, and lack of concern from Practice of General Insurance (Hart et al )

8 Ex ante & Ex Post Moral Hazard 1. Ex ante moral hazard Occurs before the fact, insurance may increase the probability of an RTC 2. Ex post moral hazard Occurs after the fact, insurance may increase the cost of vehicle repair

9 Temporal Relationship Ex ante moral hazard Ex post moral hazard Insurance purchased RTC Claim lodged Liability accepted Repairs

10 1. Ex ante moral hazard Conceptual Models RTC=f(Insurance, ) 2. Ex post moral hazard RTC=f(Insurance, ) RTC=1

11 Hypothesis: Following an insured RTC both the vehicle owner and the smash repairer have an incentive to engage in behaviour that is not observable to the insurer, which can affect the total cost of the repair. Consumer-push The owner has an incentive to exaggerate the extent of the damage and claim for extra repairs. Producer-pull Ex post Moral Hazard Hypothesis The smash repairer has an incentive to charge a premium for insurance-jobs

12 Australian Smash Repair Market 9.6 million vehicles (ABS 2003) Annual cost of repairs $3.89 billion (BTE 2000) Probability of a claim 12% (ISA 2009) Average size of claim $3100 (BTE 2000) 87% - 90% Vehicles comprehensively insured (BTE 2000)

13 Claims Data Proprietary Commercially sensitive Unavailable! Survey Data Data Collection In 1999 IMRAS Consulting conducted market research on consumer attitudes to the Australian smash repair industry 24,000 households randomly telephoned 4 States (NSW, Vic, Qld, WA) Metropolitan and rural 4,006 respondents (17%) 994 RTCs During 1997 to % annual probability of RTC

14 1. Dependent variable Cost of a repair 2. Explanatory variable Comprehensive insurance 3. Covariates Driver demographics Gender Vehicle characteristics Value ($) Make (26 makes) Age of vehicle (years) RTC characteristics Road services in attendance (EMT, Tow-truck, Police) Panels and parts damaged Data Categories

15 Road Traffic Crashes Over last 2 years ( ) did you have you had an RTC? n=994 Most recent RTC occurred in 1997 (n =148) Most recent RTC occurred in 1998 (n =293) Most recent RTC occurred in 1999 (n =473) Note: An additional 80 RTCs, which were not identified as most recent.

16 Descriptive Data Data Means Freq. % of RTCS Cost of repair (A$) $1,884 S.D. = $3,850 Comprehensively insured % Vehicle characteristics Value of vehicle (A$) $12,760 S.D. = $18,699 Age of vehicle (years) 8.4 S.D. = 6.4 Vehicle make Data Means Freq. % of RTCS Audi 3 0.3% Mitsubishi % BMW 6 0.6% Nissan % Chrysler/Jeep 5 0.5% Peugeot 4 0.4% Daewoo % Proton 2 0.2% Ford % Renault 2 0.2% Daihatsu % Rover 2 0.2% Holden/GMH % Saab 3 0.3% Honda % Seat 1 0.1% Hyundai % Subaru % Jaguar 1 0.1% Suzuki 7 0.7% Kia 1 0.1% Toyota % Land Rover 4 0.4% VW 9 0.9% Mazda % Volvo 7 0.7% Mercedes 4 0.4%

17 Data Severity of RTC Services attending RTC Means Freq. % of RTCS Ambulance % Tow Truck % Police % Towed away % Damaged parts Data Means Freq. % of RTCS No. of damaged parts 2.3 min/max (0-17) Front bonnet % Roof of car % Front bumpers * % Engine % Front lights * % Other * % Front windscreen * % Aerial * 4 0.4% Left side doors % Framework 8 0.8% Front wheel * % Bodywork % Rear wheel * % The whole rear section 1 0.1% Left side front panel/mudguard % The bull-bar * 6 0.6% Left side back panel/mudguard % Side windows * % Right side doors % It was burnt out/ written off % Right side front panel/mudguard % Mirrors * 4 0.4% Right side back panel/mudguard % Accessories * % Rear boot/hatch back % Door locks * % Rear bumper * % Seats * 3 0.3% Rear windscreen * % Muffler/ exhaust * 3 0.3% Rear lights * % Don't know * 2 0.2% Interior front dashboard * % Fuel tank 2 0.2% Interior other * % * Denotes a sub-set of vehicle-parts called Accessories

18 Three Econometric Models 1. Ex post moral hazard 2. Consumer-push 3. Producer-pull

19 Ex post Moral Hazard Cost = f (Insurance, M, N) OLS regression Cost is a continuous variable Insurance is a binary variable (=1 if insured; =0 if not) M is a vector of vehicle characteristics N is a vector of RTC severity controls

20 Dependent Variable: Cost Histogram of Costs Histogram of OLS Residuals

21 Shapiro Wilk Test for Normality A Shapiro Wilk test used to test if the error term is normally distributed. The null hypothesis, that the residuals are normally distributed, is not rejected if the p-value for test statistic W is greater than the selected level of statistical significance. W = 0.656, p-value < 0.01 Therefore the Shapiro-Wilk test for normality indicates that the null hypothesis should be rejected.

22 Results: Ex post Moral Hazard OLS Semi log Coefficient p-value Coefficient p-value Comprehensive insurance RTC severity Ambulance (0/1) Tow Truck (0/1) $ Police (0/1) < Towed away (0/1) Vehicle characteristics Vehicle value ($) E-05 <0.01 Vehicle value 2 ($) 1.54E-07 < E Vehicle Age Gender (=1 if male) Constant <0.01 R Note: 1. Unreported coefficients: Audi, BMW, Chrysler/Jeep, Daewoo, Daihatsu, Ford, Holden, Honda, Hyundai, Jaguar, Kia, Land Rover, Mazda, Mercedes, Mitsubishi, Nissan, Peugeot, Proton, Renault, Rover, Saab, Seat, Subaru, Suzuki, Toyota, VW & Volvo. 2. Ford is the omitted dummy variable

23 Consumer-push Damage = f (Insurance, M, N) Poisson Regression Damage is a number of parts repaired Insurance is a binary variable (=1 if insured; =0 if not) M is a vector of vehicle characteristics N is a vector of RTC severity controls

24 Results: Consumer-push Number of damaged Parts Number of damaged Accessories * Coefficient p-value Coefficient p-value Comprehensively insured Services Attending RTC Ambulance (0/1) Tow Truck (0/1) 0.38 < Police (0/1) Towed away (0/1) Vehicle Characteristics Vehicle value ($) 2.51E E Vehicle value 2 ($) 4.73E E Vehicle age (years) Gender (=1 if male) Constant 0.68 < E Note: 1) * Denotes Accessories : Front bumpers, Front lights, Front windscreen, Front wheel, Rear wheel, Rear bumper, Rear windscreen, Rear lights, Interior, Front dashboard, Interior other, Other, Aerial, Bull-bar, Side windows, Mirrors, Accessories, Door locks, Seats, Muffler, Unknown. 2) Coefficients for vehicle-make withheld 3) Dummy Ford omitted

25 Producer-pull Cost = f (Insurance, M, N, Damage) OLS Regression Cost is continuous variable Insurance is a binary variable (=1 if insured; =0 if not) M is a vector of vehicle characteristics N is a vector of RTC severity controls Damage is number of parts repaired

26 Results: Producer-pull OLS (R 2 =0.44) Semi log (R 2 =0.35) Coefficient p-value Coefficient p-value Comprehensive insurance RTC severity No of Parts < <0.01 Ambulance (0/1) Tow Truck (0/1) Police (0/1) Towed away (0/1) Vehicle characteristics Vehicle value ($) E-05 <0.01 Vehicle value 2 ($) 1.43E-07 < E Vehicle Age (years) Gender (=1 if male) Constant <0.01 Note: 1.Omitted coefficients: Audi, BMW, Chrysler/Jeep, Daewoo, Daihatsu, Holden, Honda, Hyundai, Jaguar, Kia, Land Rover, Mazda, Mercedes, Mitsubishi, Nissan, Peugeot, Proton, Renault, Rover, Saab, Seat, Subaru, Suzuki, Toyota, VW & Volvo. 2. Omitted dummy variable Ford

27 Conclusions Ex post moral hazard = $443 ($758) Not significant Note: 2011 estimates, adjusted for a CPI = 5%, are in parenthesis

28 Conclusion Subject to the proviso that the IMRAS data are representative of consumer experiences in the Australian smash repair industry and not confounded by significant reporting or measurement bias, these results imply the existence of an ex post moral hazard in the Australian market for comprehensive insurance. The dominant driver of cost is a producer-pull ($348) rather than a consumerpush effect, which appears to be negligible. Insurers who are interested in mitigating the non-trivial costs of ex post moral hazard may therefore wish to focus their over-sight on the behaviour of smash repairers rather than motorists.

29 Acknowledgements Australian Centre for Economic Research on Health (ACERH-UQ) PhD Scholarship Brian Gray Scholarship 2010 Provided by the Australian Prudential Regulatory Authority (APRA) and the Reserve Bank of Australia (RBA)

30 Bibliography Australian Bureau of Statistics Motor vehicle census ABS cat. no Australian Bureau of Statistics (ABS), Canberra. Bureau of Transport Economics Road Crash Costs in Australia BTE Report 102. Bureau of Transport Economics (BTE), Canberra.