Statistical Work in Motor Insurance Technical Workshop, Sarajevo, 10 June 2015 Michael Theilmeier
German Insurance Association 2 Single records vs aggregated records (1) Business Statistics: Aggregated data are absolutely sufficient and much easier to handle. (2) Risk/Calculatory Statistics: Single information is - albeit not required - very advisable!
German Insurance Association 3 Risk statistics / calculatory statistics differentiated after relevant rating / risk criteria give detailed information on portfolio and claims situation and development by risk category net risk premium by risk category serve as a solid basis for the company s own tariffication
German Insurance Association 4 Why single records for risk and calculatory statistics? (1) With aggregated data by risk group, you have to stick to the given tariff structure. With single records modifications or refinements of the tariff structure can easily be done. (2) The data management of the insurance companies is based on single information on contracts and claims so that for them the data export should easily be done.
German Insurance Association 5 (3) Single information makes it easier to check the data Errors in the individual data records cannot be found unless separate records are used for the validation of data. (4) Statistical observations might be biased by strong random effects, e. g. due to large-scale claims or natural catastrophes. Removal of such unwelcome random effects is in an adequate way only possible on the basis of single data records.
German Insurance Association 6 Which items should be published per risk category? number of standardized contracts (one-year-contracts) number of claims claims expenditure claims frequency = number of claims per 1000 standardized contracts average amount of claims = claims expenditure / number of claims net risk premium = claims frequency * (loss costs) average amount of claims = claims expenditure / number of standardized contracts
German Insurance Association 7 annual average premium = net risk premium + risk margin + cost margin + profit margin + taxes
German Insurance Association 8 Collecting the data Exchange of information between Insurance Companies Vehicle Registration Office Institute of Risk (i.e. Insurance Association) Other Institutions Supervisory Authority Federal Statistical Office
German Insurance Association Some Data about the German Motor Market
10 German Insurance Association 10 German Motor insurance: premium income, 2014* Distribution of gross premiums written, in EUR billions Motor insurance overall: 24.3 billion EUR MTPL 59.8% 8.1 Comprehensive casco 33.3% 14.6 1.6 0.1 Partial Casco 6.5% 6.6% Passenger accident 0.4% * Preliminary result (status 26/02/2015) 04/06/2015
11 German Insurance Association 11 Development of German motor insurance contracts MTPL and comprehensive casco, standardized contracts, in millions 60 50 40 MTPL 59.3 30 20 10 Comprehensive Casco 26.4 0 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14* * Preliminary result (status 26/02/2015) 04/06/2015
12 German Insurance Association 12 Development of German average annual motor premium MTPL and comprehensive casco, in EUR 500 400 300 200 Comprehensive Casco MTPL 305 244 100 0 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14* * Preliminary result (status 26/02/2015) 04/06/2015
13 German Insurance Association 13 Development of German gross motor premium income MTPL and comprehensive casco, in EUR billions 16 14.6 12 MTPL 8 8.1 4 Comprehensive Casco 0 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14* * Preliminary result (status 26/02/2015) 04/06/2015
14 German Insurance Association 14 Underwriting results German motor insurance MTPL and comprehensive casco, in EUR millions 1,000 Comprehensive Casco 0-1,000 MTPL -2,000 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14* * Preliminary result (status 26/02/2015) 04/06/2015
15 German Insurance Association 15 Development of German motor claims frequency MTPL and comprehensive casco, permille 300 250 Comprehensive Casco 200 164 150 100 50 MTPL 59 0 79 81 83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 * Preliminary result (status 26/02/2015) 04/06/2015
16 German Insurance Association 16 Development of average amount of German motor claims MTPL and comprehensive casco, in EUR 4,000 3,575 3,000 2,000 MTPL 1,715 1,000 Comprehensive Casco 0 79 81 83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 * Preliminary result (status 26/02/2015) 04/06/2015
17 German Insurance Association 17 Development of average amount of German motor claims For bodily injury and material damage, MTPL, in EUR 16,000 15,967 12,000 8,000 bodily injury (incl. share of material damage) 4,000 material damage 2,435 0 79 81 83 85 87 89 91 93 95 97 99 01 03 05 07 09 11 13 * Preliminary result (status 26/02/2015) 04/06/2015
German Insurance Association Some selected Rating Factors
German Insurance Association 19 Rating factors in MTPL insurance/passenger cars Rating criteria used by nearly all market participants (standard criteria) Claim / No-claim class (bonus / malus) Type class Differentiated age of users Mileage per year in 1000 km Age of the motor vehicle when acquired Regional class Deductibles Group of users Profession (civil servant, farmer, other) Ownership of residential property
Schadenbedarfsindex German Insurance Association 20 MTPL KH-Versicherung Claim / No-claim Pkw: Schadenbedarfsindizes class (bonus /malus) normiert auf gew. Mittelwert 1 mit Jahreseinheitenverteilung 2008-2010 6 5 U 2012 U 99 4 3 2 1 0 S/SF-Stufe
German Insurance Association 21 Type class
German Insurance Association 22 22 Typklassenverzeichnis zum 01.07.2014, Auszug für Dacia Abkürzungen: WFS = Wegfahrsperre, Pb = Produktionsbeginn, Pe = Produktionsende; Klasse VK bzw. TK: Typklasse nach dem bis 2002 üblichen Vorgehen; Klasse VKN bzw. TKN: Typklasse nach dem aktuellen Vorgehen Unter Dacia werden in diesem Verzeichnis 70 Fahrzeugtypen geführt.
Tariff structures and methodologies Move of Objective Criteria for Motor Tariffs: From: Value (MOD), Kw, ccm (MTPL) To: Type class of vehicle 23
Vehicle Value, Average Premium and Loss Costs correlate 160.000 Average Premium vs. Loss Cost per Vehicle, UY 2012 140.000 120.000 100.000 80.000 60.000 40.000 20.000 0 200.000 300.000 500.000 750.000 1.000.000 1.250.000 1.500.000 1.800.000 2.500.000 3.000.000 4.000.000 5.000.000 more Vehicle Value Average Premium Loss Cost per Vehicle 24
Vehicle Value, Average Premium correlate. but not Loss Costs Average Premium vs. Loss Cost per Vehicle, UY 2013 3.500 3.000 2.500 2.000 1.500 1.000 500 0 5.000 15.000 30.000 40.000 60.000 80.000 100.000 140.000 170.000 200.000 250.000 400.000 more Vehicle Value Average Premium Loss Cost per Vehicle Presentation for CAP, 1 April 2008 25
German Insurance Association 26 Differentiated age of users
German Insurance Association 27 [Presentation Title] [Name Proprietary of Presenter and Confidential ] [DD.MM.YYYY] General Reinsurance AG
German Insurance Association 28 Mileage per year in 1000 km
German Insurance Association 29 Age of the motor vehicle when acquired
German Insurance Association 30 MTPL Regional class and population density Klassen* Bremen Kiel Hamburg Hannover Schwerin Rostock Magdeburg Berlin Cottbus 1 (33) 2 (64) 3 (43) 4 (37) 5 (31) 6 (49) 7 (45) 8 (33) 9 (35) 10 (23) 11 (17) 12 (9) Bremen Kiel Hamburg Hannover Schwerin Rostock Magdeburg Berlin Cottbus Aachen Düsseldorf Köln Kassel Erfurt Leipzig Dresden Aachen Düsseldorf Köln Kassel Erfurt Leipzig Dresden Frankfurt am Main Würzburg Nürnberg Karlsruhe Stuttgart München Freiburg * Klassengrenzen nach unverbindlicher Bekanntgabe des GDV Einwohnerdichte bis 50 (17) > 50 bis 100 (43) > 100 bis 250 (63) > 250 bis 500 (66) > 500 bis 1000 (51) > 1000 bis 1500 (31) > 1500 bis 2000 (32) > 2000 bis 2300 (33) > 2300 bis 2600 (32) > 2600 bis 2800 (26) > 2800 bis 2900 (21) > 3000 (3) Frankfurt am Main Würzburg Karlsruhe Stuttgart Freiburg Nürnberg München
German Insurance Association 31 Deductible; difference compared to 150
German Insurance Association 32 Group of users
German Insurance Association 33 Profession (civil servant, farmer, other)
German Insurance Association 34 Ownership of residential property
German Insurance Association 35 Additional rating criteria Bonus / Malus rules in respect of second cars Categories of occupation groups in the case of non-civil-servants Loyalty discount Young family
German Insurance Association 36 CAT events in MOD Hail as a significant peril Sophisticated modelling quite challenging Market modell for Germany with limited (but realistic) parameters developed by GDV Basis: 152 events (1984 2013) Spreadsheet for company modell supplied
German Insurance Association 37 CAT events in MOD; market modell data results Average loss costs per vehicle year 1984-2013: around 16 1 in 200 years: 84; VaR 1 in 50 years: 55 (TVaR: 79)
German Insurance Association 38 CAT events in MOD; market modell data results 2013 event(s) Andreas and Bernd: 20.20 = about 1 in 18 years Munich 1984: 49.80 at 2013 prices; modell says about 1 in 120 years; but different loss adjusting today
German Insurance Association 39 Cummulation Losses (mainly hail) 30%-500%
Wilhelmstraße 43 / 43 G, D-10117 Berlin Postfach 08 02 64, D-10002 Berlin Tel.: +49 30 2020-5000 Fax: +49 30 2020-6000 51, rue Montoyer B-1000 Brüssel Tel.: +32 2 28247-30 Fax: +32 2 28247-39 www.gdv.de @gdv_de
German Insurance Association Appendix
German Insurance Association 42 1. Standardisation requirements for a uniform set of data (1) Central issue of the statistics manual (2) Each insurance company might use a different IT infrastructure but should deliver its data set in a fixed mandatory format (3) The structure of the uniform set of data is primarily based on the given tariff but should be open for extensions and refinements and other official sources (driving licence, registration-document, data from the vehicle registration office,...)
German Insurance Association 43 (4) The structure of the uniform data set should be well considered, e. g. it should give enough leeway for modifications and refinements without noteworthy effort. include a lot of data fields which can be used by the companies for their own individual purposes. In Germany the GDV standard record format is not only used by the association s member companies but also by others as brokers and consultancy firms.
German Insurance Association 44 (5) The data on contracts on the one hand and on claims on the other hand are usually handled separately within the companies (frequently claims management is outsourced). That is why it might be helpful to distinguish between data records on contracts and data records on claims
German Insurance Association 45 (6) Standard data record on contracts should include a code number for the insurance company a code number for the insured person (both information are necessary to match the information on the contract with the corresponding information on claims) the date of start and expiration of the contract (to take into account how long the policyholder was insured within a year to calculate standardised contracts) all information on relevant tariff and risk criteria
German Insurance Association 46 (7) Standard data records on claims should include a code number for the insurance company a code number for the insured person (to match contract and claims data) all information on relevant tariff and risk criteria all information on claims date when and where the claim has occured character of claim (material damage, bodily injury,...) claims payment claims provisions...
German Insurance Association 47 (8) The structure of both sort of records should be similar and consistent to each other. (9) There should be sufficient data fields in the standard data record for each tariff criteria, e. g. one data field for the distinction between the different sort of vehicles (passenger cars, lorries, busses etc.) and additive data fields for further characteristics (lorries less than 10 tons/more than 10 tons, busses less than 20 seats/more than 20 seats etc.)
German Insurance Association 48 (10) It is better to have detailed than aggregated information, e. g. : It is better to record the postal code or the district where the policyholder is registered than to record only the regional class! On basis of these information modifications or refinements respectively can easily be done. The same is true for the tariff criteria kw. Better to record the actual kw power than only the kw class to be armed for the case of modifications.
German Insurance Association 49 (11) Exception: bonus / malus system Particularly for bonus/malus recording the complete claims history of each policyholder might be very elaborate and costly. Much easier it is to record the bonus/malus class only. Provided each company obeys the rules of the bonus/malus system this information is absolutely sufficient. (12) Last but not least: A statistics manual should define each data field as accurate as possible. E. g. what is exactly meant by claims payment and claims provision? What should be included?
German Insurance Association 50 2 Requirements on data management Database containing summary information about each dataset reported from insurance companies: delivery date reported insurance year reported classes of insurance insurance company (code number) number of records, claims contact person(s) in insurance company remarks: errors and implausibilities
German Insurance Association 51 Database including the codes of the statistics manual, their groupings and relations to each other, e. g. regarding: sort of vehicle bonus-malus classes kw-classes postal code/district etc.
German Insurance Association 52 3. Checking and evaluating the data Step 1: Checking the data (e. g. by means of a delivery programme service) within the insurance company
German Insurance Association 53 Step 2: Checking the data within the Institute of Risk by means of the same delivery programme service regarding - formal errors, e. g. if in a data field a code is reported which is not existent or is not defined - can-be errors (implausibilities), e. g. if the claims payment is untypically high or low ( enquire call with the regarding company)
German Insurance Association 54 Step 3: In case of too many errors in the reported data set the reporting company should be requested to correct the data and re-deliver them to the Institute of Risk. It is crucial to eliminate errors in the data set already in the company s own data management system. Otherwise one runs the risk that the same errors will be carried forward from year to year. Moreover the insurance company itself should have an interest in a valid data basis (for calculation purposes).
German Insurance Association 55 Step 4: The whole process of compiling market statistics should be understood as a process of controlling. Are the resulting market statistics feasible and adequate for their purposes? The results should be evaluated from different angles, e. g. with the focus on different risk/rating criteria.
German Insurance Association 56 Step 5: In MTPL it could be considered to let the supervisory authority take an appropriate part in the process of evaluating the statistics as an additional controlling instance.
German Insurance Association 57 Step 6: The insurance industry which has to deliver the data should be involved in the process of assessing the market statistics. That could be done within a committee which includes experts and representatives of the insurance companies, actuaries from the Institute of Risk and maybe also representatives of the supervisory authority.