The Interaction Between Risk Classification and Adverse Selection: Evidence from California s Residential Earthquake Insurance Market

Transcription

1 The Interaction Between Risk Classification and Adverse Selection: Evidence from California s Residential Earthquake Insurance Market Xiao (Joyce) Lin, PhD candidate Actuarial Science, Risk Management and Insurance Wisconsin School of Business University of Wisconsin-Madison

2 Introduction I study the nature of selection issues and how they interact with risk classification in the context of California s residential earthquake insurance market In this market: No mandate on demand The dominant semi-public insurer (California Earthquake Authority) uses coarse geographic risk classification

3 Introduction I study the nature of selection issues and how they interact with risk classification in the context of California s residential earthquake insurance market In this market: No mandate on demand The dominant semi-public insurer (California Earthquake Authority) uses coarse geographic risk classification could lead to adverse selection against the CEA

4 Introduction I study the nature of selection issues and how they interact with risk classification in the context of California s residential earthquake insurance market In this market: No mandate on demand The dominant semi-public insurer (California Earthquake Authority) uses coarse geographic risk classification Coexistence of public and private underwriters

5 Introduction I study the nature of selection issues and how they interact with risk classification in the context of California s residential earthquake insurance market In this market: No mandate on demand The dominant semi-public insurer (California Earthquake Authority) uses coarse geographic risk classification Coexistence of public and private underwriters the adverse selection might be worsened by the finer risk classification in the private sector

6 Introduction I address three empirical questions: 1. How insurers classify earthquake risk geographically in California? 2. Does the limited risk classification by the CEA lead to adverse selection? 3. Are the private insurers picking up relatively lower-risk homeowners by using finer risk-classification schemes?

7 Literature Review Adverse selection and asymmetric information Akerlof (1970), Finkelstein & Porterba (2004), Cohen & Siegelman (2010) Catastrophe insurance demand analysis Palm (1992 and 1995), Kunreuther & Michel-Kerjan (2009), Athavale & Avila (2011) Public vs. Private insurers Czajkowski et al. (2012) The Equity-Efficiency Trade-off Picard (2008)

8 Background The California Earthquake Authority (CEA) CEA: Public-Private Partnership Publicly-managed, largely privately funded 20 participating insurers (19 at the time of study) representing 70% of the market The CEA policies: Eligibility Base rate: only varies by 19 territories A private fringe remains outside of the CEA

9 Data Unit of observation: zip-code level (1636 zip-codes in California) Insurance policy count data from the CDI (2009) Geological data on seismic risk from the USGS Census data from the U.S. Census Bureau Insurers rating schemes from public filings to the CDI

10 Data Unit of observation: zip-code level (1636 zip-codes in California) Insurance policy count data from the CDI (2009) Geological data on seismic risk from the USGS Census data from the U.S. Census Bureau Insurers rating schemes from public filings to the CDI

11 Data Insurance policy count data Table 3: Summary Statistics of Policy Counts and Take-up Rates by Zip-code Standard Mean Deviation Min* Median Max Overall Homeowners Policies 3,594 3, ,576 16,674 CEA Homeowners Policies 2,678 2, ,926 13,015 noncea Homeowners Policies 916 1, ,111 Overall Earthquake Policies ,947 CEA Earthquake Policies ,729 noncea Earthquake Policies ,268 Overall Earthquake Policy Take-up Rate 12.84% 10.97% 0.00% 9.86% 69.29% CEA Earthquake Policy Take-up Rate 12.38% 10.25% 0.00% 9.50% 54.55% noncea Earthquake Policy Take-up Rate 15.56% 21.70% 0.00% 9.09% 450%**

12 Data Geological data on Peak Ground Acceleration () The higher the value (from green to red), the higher the seismic risk.

13 CEA Rates (Based on 2007 rate manual) CEA territory rates are strongly related to average y= x R 2 = Policy-weighted

14 But there is substantial variation in underlying seismic risk within territory

15 Private insurers (e.g. GeoVera) price risks at a finer level Territory 27 Territory 7 Territory 22 Territory 6 GeoVera Rates Adj. R 2 = Adj. R 2 = Adj. R 2 = GeoVera Rates GeoVera Rates GeoVera Rates Adj. R 2 = Territory 12 Territory 26 Territory 20 Territory 11 GeoVera Rates Adj. R 2 = GeoVera Rates Adj. R 2 = Adj. R 2 = GeoVera Rates GeoVera Rates Adj. R 2 = Territory 23 Territory 5 Territory 8 Territory 15 GeoVera Rates Adj. R 2 = GeoVera Rates Adj. R 2 = GeoVera Rates Adj. R 2 = GeoVera Rates Adj. R 2 =

16 Demand for CEA earthquake insurance a map of zip-code level CEA take-up rates

17 Is the CEA take-up rate positively related to risk level () within a rating territory (adverse selection)? Territory 27 Territory 7 Territory 22 Territory 6 CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = Territory 12 Territory 26 Territory 20 Territory 11 CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = Territory 23 Territory 5 Territory 8 Territory 15 CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 = CEA Take-up Rates y = x Adj. R 2 =

18 CEA take-up rate is positively related to risk level, all else equal Regression Framework: CEA Earthquake Insurance Demand = β 0 + β 1 Objective Risk + β 2 Ln Home Value + β 3 Ln Income + β 4 Demographic characteristics + β 5 Territory + ε Use the take-up rate of CEA policies as the dependent variable Observations are 1636 zip-codes in California. Only the cross-sectional data in 2009 is used.

19 CEA take-up rate is positively related to risk level, all else equal (1) (2) (3) (4) (5) Variable Estimate Std. Error Estimate Std. Error Estimate Std. Error Estimate Std. Error Estimate Std. Error Objective Risk Measure () *** *** *** *** *** Log (Median Value of Owner-Occupied Homes) *** *** *** *** Log (Median Household Income) *** *** *** Population with At Least College Degree (%) *** *** *** Median Household Size *** *** Gender (female%) *** *** Median Age Percentage of Black or African American Percentage of Asian Percentage of other races *** Household with Children under 18 Years Old Log (Population Per Square Mile) *** Territory *** *** *** *** *** Territory *** ** *** *** ***.. (Intercept) *** *** *** ** *** Observations Adjusted R-squared Note: The observations are 1636 zip-codes in California. All regressions are estimated using weighted least square method, with the number of homeowners' policies in each zip-code being the weight

20 Is the adverse selection problem against the CEA worsened by private insurers? Private sector uses finer risk classification schemes To compare the demand for public (CEA) vs. private: CEA EQ share = CEA s earthquake market share Territory average of the CEA s earthquake market share CEA EQ share = β 0 + β 1 Objective Risk + β 2 CEA HO share + β 3 Ln Home Value + β 4 Ln Income + β 5 Demographic characteristics + β 6 Territory + ε

21 Is the adverse selection problem against the CEA worsened by private insurers? Private sector uses finer risk classification schemes To compare the demand for public (CEA) vs. private CEA EQ share = CEA s earthquake market share Territory average of the CEA s earthquake market share CEA EQ share = β 0 + β 1 Objective Risk + β 2 CEA HO share + β 3 Ln Home Value + β 4 Ln Income + β 5 Demographic characteristics + β 6 Territory + ε

22 Is the adverse selection problem against the CEA worsened by private insurers? Private sector uses finer risk classification schemes To compare the demand for public (CEA) vs. private CEA EQ share = CEA s earthquake market share Territory average of the CEA s earthquake market share CEA EQ share = β 0 + β 1 Objective Risk + β 2 CEA HO share + β 3 Ln Home Value + β 4 Ln Income + β 5 Demographic characteristics + β 6 Territory + ε

23 Is there a positive relationship between variation in CEA s earthquake market share and underlying risk? Territory 27 Territory 7 Territory 22 Territory 6 y = x*** Adj. R 2 = y = x Adj. R 2 = y = x* Adj. R 2 = y = x*** Adj. R 2 = CEA EQ share CEA EQ share CEA EQ share CEA EQ share Territory 12 Territory 26 Territory 20 Territory 11 y = x** Adj. R 2 = y = x Adj. R 2 = y = x*** Adj. R 2 = y = x* Adj. R 2 = CEA EQ share CEA EQ share CEA EQ share CEA EQ share Territory 23 Territory 5 Territory 8 Territory 15 y = x Adj. R 2 = y = x Adj. R 2 = y = x* Adj. R 2 = y = x Adj. R 2 = CEA EQ share CEA EQ share CEA EQ share CEA EQ share

24 Conclusion and Discussion CEA offers highly cross-subsidized rates (while GeoVera uses much finer classification schemes) Strong evidence of adverse selection against the CEA Slight evidence of private insurers cherry-picking the CEA -- Why?: other possible drivers of demand; different marketing tactics

25 Thank you! Questions and Suggestions?