Disability Income Data What do we need to capture and analyse now?

Transcription

1 Disability Income Data What do we need to capture and analyse now? Michael Dermody & Chris Scheuber 2015 KPMG Actuarial This presentation has been prepared for the Actuaries Institute 2015 Actuaries Summit. The Institute Council wishes it to be understood that opinions put forward herein are not necessarily those of the Institute and the Council is not responsible for those opinions.

2 Agenda 1. Current landscape 2. Insights from other regions & industries 3. Actions 4. Discussion

3 Industry Profitability Profit/(Loss) after tax $m Dec-10 Mar-11 Jun-11 Sep-11 Australian Industry Retail DI Experience Quarter to 31 December 2014 Dec-11 Mar-12 Jun-12 Sep-12 Dec-12 Mar-13 Jun-13 Sep-13 Dec-13 Mar-14 Jun-14 Sep-14 Dec-14 The Retail Disability Income loss (-$394M or 43% of gross revenue) related to both Ordinary and Superannuation business. The large reduction in 10 year bond rate in the December quarter would partly, but not fully, explain these losses. The remainder would likely be due to further reserve strengthening Super - Individual Disability Income Insurance Ordinary - Individual Disability Income Insurance Extract from KPMG Life Risk Profitability Update, December 2014 based on APRA Quarterly Life Insurance Performance Statistics

4 10 Industry s View on Maturity of Current State & Value from Development Do you think reserving and rating factors will be different in future from what they are now? Research, can t believe more insight wouldn t come out of that Systems development about improving understanding getting holistic view 2 - No Direct Life Offices Yes Reinsurers We started using external data sources we got value out of it and are going to continue with it.

5 DI Sources of Data Internal to Life company External to Life Co Point of Sale Policy Maintenance Claim Processing Agent Management Underwriting Administration Claims Admin Commission / Sales & Agent Management External to group - Popln Segment Medical Reports Advice Phone calls Paper Correspondence Medical Reports Rehab/ Investigator Reports Phone Calls Internal to group Banking General insurance Claim Mgt Aids

6 Current Landscape Existing Analysis Everyone Annual experience investigations: - One way analysis - Some two way analysis Some Annual experience investigations with GLM Some Early warning signals/lead indicators - deep dive when results poor Targeted actuarial analysis Discussion with frontline staff including UW, claims. Embryonic Group data analytics team Unstructured and structured Starting to consider Database with linked data across systems Individual claims file analysis

7 Current Landscape Level of analysis (ctd) No Yes - No Yes Do you capture data on all interactions with a customer (phone calls, etc)? Are you linking data across systems?

8 Current Landscape Level of analysis (ctd) No Yes Do you plan further development? Does it include specific system/data collection enhancements? - Mainly actuaries Specialist maths/statistics Other data/it specialists What capability / professionals have been involved? Data scientists? Statisticians?

9 Current Landscape Level of analysis (ctd) 8 Does your analysis leverage external data sources: - Data from other parts of the group outside life co; and / or - Data external to group? Yes - data elsewhere in group Yes - sourced data external to group No - but clear plans/begun exploring No - but No - and at this something like stage no to explore interest

10 DI vs Other Insurances/Banking Permanent Differences Volume Data Depth Free text, but Ethical/legal/cultural considerations Temporary Differences Data quality Monitoring impacts of change Data structure

11 DI Discrimination Commonwealth Legislation* Insurance/Super exemptions Age Discrimination Act 2004 Disability Discrimination Act 1992 Racial Discrimination Act 1975 Sex Discrimination Act 1984 Age/Health Status and Sex: Able to charge differential premiums on provided it is based on actuarial/statistical data and it is reasonable for the entity discriminating to rely on that data. Age/Health Status if no actuarial or statistical data is available/cannot reasonably be obtained then can charge different premiums where it is reasonable having regard to any other relevant factors. Racial discrimination not permitted. *There are also anti-discrimination laws in each state which prohibit discrimination on a number of other grounds eg sexual orientation.

12 Insights from Other Regions Current Analysis Consider external data (structured first then unstructured) credit scores, medical / prescription information, demographics, etc Unstructured analysis (no hypothesis leave analysis to find relationships) Text mining of claims notes, prescription data and treatment assists with understanding severity of claim Analysis by advisor helps to understand reasons for success in terms of volumes and profitability of business written Application to DI in Australia?? Likely relevant for DI in Australia investment required regarding text mining or capturing information on severity in codified field. Relevant for DI in Australia additional insight beyond analysis by channel and provides helpful feedback to the business

13 What do we do now? Leverage more existing internal data to do deeper analysis Build data advances into the system architecture Start collecting new data fields Use insights from analysis for decision making Overlay external data to analysis and decision making

14 Leverage more existing internal data to do deeper analysis How to do it Connect data from separate unconnected systems / sources Leverage existing unstructured data using text mining techniques Correct existing data errors Utilise more advanced techniques than one way factor analysis DI examples Claims to underwriting to admin Linkage to internal banking data Claims manager s and underwriter s notes Claim causes Occupation class Claim cause Generalised linear models CART models

15 Leverage more existing internal data to do deeper analysis How to do it Connect data from separate unconnected systems / sources Leverage existing unstructured data using text mining techniques Correct existing data errors Utilise more advanced techniques than one way factor analysis DI examples Claims to underwriting to admin Linkage to internal banking data Claims manager s and underwriter s notes Claim causes Occupation class Claim cause Generalised linear models CART models

16 Build data advances into the system architecture How to do it Automate links between systems Transform text fields into helpful dropdowns Build business rules to limit future data errors DI examples A policy and/or customer ID can be created which is looked up or pre-populated when entering data across all systems Transforming claim cause to a categorisation (ICD-10 and/or customised to include severity) Not allowing important fields to be left blank Ordering of dates for claims

17 DI Sources of Data Database / Master File Internal to life Co External to life Co Point of Sale Policy Maintenance Claim Processing Agent Management Underwriting Administration Claims Admin Commission / Sales & Agent Management External to group - Popln Segment Advice Phone calls Paper Correspondence Medical Reports Rehab/ Investigator Reports Phone Calls Internal to group Banking General insurance Claim Mgt Aids

18 Start collecting new data fields How to do it When applying for a policy Policy renewal and queries Claim inception and management DI examples BMI Credit information Validate existing fields provided at underwriting Current salary (even for agreed value) Number of dependants Offsets and different benefit types paid

19 Use insights from analysis for decision making How to do it Underwriting Claims Retention Strategy DI examples Use additional factors to differentiate pricing, bearing in mind legislative requirements Triaging of claims to focus effort better Providing expected durations for claims team provides understanding of fields not captured correctly Identifies more accurately lost value from lapses of individual policies Identifies policies more likely to lapse & more successful actions to retain them Identifies advisers who write most profitable business

20 Overlay external data to analysis and decision making How to do it Claims Strategy DI examples Social media to assist with investigations Credit scores by region Mosaic or similar which categorises individuals by address & contains population level information

21 Discussion