MAIN HEADING: Ontario Private Passenger Automobile - Appendix B Independent Actuarial Analysis as of December 31, 2013

Final Release: November 7, 2014

Prepared by:

Pierre Lepage, FCAS, FCIA
(514) 840-2289
plepage@kpmg.ca

Anh Tu Le, FCAS, ACIA
(647) 777-5352
ale@kpmg.ca

kpmg.ca

This document has been prepared by KPMG for the Government of Ontario (Government). KPMG neither warrants nor represents that the information contained in this document is accurate, complete, sufficient or appropriate for use by any person or entity other than the Government or for any purpose other than that set out in the Automobile Insurance Transparency and Accountability Expert Report – 2014 Annual Report. This document may not be relied upon by any person or entity other than the Government, and KPMG hereby expressly disclaims any and all responsibility or liability to any person or entity other than the Government in connection with their use of this document.

Table of Contents

1 INTRODUCTION

2 KEY RESULTS AND FINDINGS

3 METHODOLOGY

4 EXHIBITS

1 INTRODUCTION

1.1 Identification of the Business under Review, Actuaries, and As of Date

As part of the Automobile Insurance Transparency and Accountability Expert Report – 2014 Annual Report (2014 Annual Report), the Ontario Ministry of Finance (MOF) requested KPMG LLP (KPMG) to prepare a quantitative analysis of the most recent data available, at the Ontario private passenger automobile (PPA) insurance industry level, to identify any decreases in PPA insurance claim costs since the major reforms introduced in September 2010 (Reforms) by the Government of Ontario (Government) and explore whether any of the identified decreases in PPA insurance claim costs have resulted in decreases in the average automobile insurance rates.

This Independent Actuarial Analysis focuses on the Ontario PPA claims experience at December 31, 2013 and was prepared by Mr. Pierre Lepage and Ms. Anh Tu Le. Mr. Lepage is a Fellow of the Canadian Institute of Actuaries (FCIA) and a Fellow of the Casualty Actuarial Society (FCAS). Ms. Le is an Associate of the Canadian Institute of Actuaries (ACIA) and an FCAS.

Contact details are:

Pierre Lepage
Partner, P&C Actuarial Services
KPMG Canada
Bay Adelaide Centre
333 Bay Street, Suite 4600
Toronto, ON M5H 2S5
Email: plepage@kpmg.ca
Tel: 514-840-2289
Fax: 416-777-8818
Anh Tu Le
Consulting Actuary, P&C Actuarial Services
KPMG Canada
Bay Adelaide Centre
333 Bay Street, Suite 4600
Toronto, ON M5H 2S5
Email: ale@kpmg.ca
Tel: 647-777-5352
Fax: 416-777-8818

1.2 Use and Distribution

This Independent Actuarial Analysis is part of the 2014 Annual Report, which was initiated as a transparency and accountability mechanism by the Government as part of the 2013 Ontario Budget1 (2013 Budget).

We understand that the 2014 Annual Report, including Appendix B - Independent Actuarial Analysis as of December 31, 2013 (Independent Actuarial Analysis), will be a publicly available document. We consent to such distribution on the condition that the Government will distribute the 2014 Annual Report in its entirety including all text and supporting appendices rather than any excerpts.

Any use or reliance on the 2014 Annual Report, including the Independent Actuarial Analysis, by any third party is done at their own risk. KPMG will not be liable for the consequences of any third party acting upon or relying upon any information or conclusions contained in this report.

This Independent Actuarial Analysis is prepared to provide a detailed documentation of the processes, methodology, and assumptions underlying the quantitative analysis of the most recent data available, at the Ontario private passenger automobile (PPA) insurance industry level. The exhibits attached in support of our findings and conclusions are an integral part of the Independent Actuarial Analysis. These sections are prepared so that our actuarial assumptions and judgments are documented. Judgments about the conclusions drawn in this appendix of the 2014 Annual Report should be made only after considering the Independent Actuarial Analysis in its entirety.

Our findings and conclusions are based on a number of assumptions. These assumptions, which are documented in subsequent sections of the Independent Actuarial Analysis, must be understood in order to place our conclusions in their appropriate context. In addition, our findings are subject to reliances and inherent limitations, which are also discussed in the Independent Actuarial Analysis.

1.3 Accepted Actuarial Practice

This Independent Actuarial Analysis of Ontario PPA experience is conducted in accordance with accepted actuarial practice in Canada.

1.4 Organization of the Independent Actuarial Analysis

The Independent Actuarial Analysis is organized in the following four major parts:

  • Introduction;
  • Key results and findings;
  • Methodology; and
  • Exhibits.

The exhibits are organized by sub-coverage and present the details of the selection of ultimate claims as well as the derivation of claim cost trends and impacts of the Reforms on claim cost levels (Reforms Impacts).

The estimates of ultimate claim costs contained in the Independent Actuarial Analysis are on a direct undiscounted basis. Reported claims include paid claims and provision for case estimates (also known as case outstanding). In the Independent Actuarial Analysis, the term IBNR is used in a broad, general sense, to represent development on case estimates (also referred to as supplemental or IBNER, incurred but not enough reported) and unreported claims (also referred to as pure IBNR).

1.5 Data Reliance

In preparing the Independent Actuarial Analysis, industry claim experience taken from the General Insurance Statistical Agency (GISA) exhibits related to All-Industry Ontario Automobile Insurance Private Passenger (excluding Farmers) Class of Business (PPA GISA Exhibits) compiled by the Insurance Bureau of Canada (IBC), on behalf of GISA, are the primary data sources.

Each of the insurers underwriting automobile insurance in Canada submit data under the Automobile Statistical Plan to IBC who then assembles the PPA GISA Exhibits. IBC performs numerous data edit checks, which are designed to promote data integrity. It is impractical for us to independently audit or verify the data due to the number of companies submitting data and our remoteness from the individual data elements. Therefore, we rely on the IBC/GISA data without independent audit or verification.

While we do not audit or independently verify the data provided by IBC on behalf of GISA, we do review the data for reasonableness and consistency. We also reconcile the data received electronically to the published reports. The underlying claim, expense, premium, and exposure data form the basis of our analysis and our findings and conclusions. Thus, the accuracy of this data is critical. The data contained in the 2013-2 PPA GISA Exhibits are deemed reliable and sufficient to perform the quantitative analysis requested by the MOF; and we use this data as published by GISA without modification.

In particular, we rely on the accident year claim development triangles and earned premium amounts published by GISA. All data are at December 31, 2013.

1.6 Reliance on Other Actuaries

Canadian actuarial standards of practice (SOP) are promulgated by the Actuarial Standards Board (ASB)2. Sections 1610.01 and 1610.02 of the SOP state: “The actuary may use and take responsibility for another person’s work if such actions are justified. If the actuary uses but does not take responsibility for another person’s work, then the actuary should so report.”

In valuing the ultimate claim costs for Ontario PPA at December 31, 2013, the work of Mr. Liam M. McFarlane, FCIA, FCAS, of Ernst & Young LLP (EY) was referenced for comparison purpose.

Mr. McFarlane has been engaged since 2009 to establish estimates of the all-industry Ontario PPA development factors for both the number and the amount of reported claims by accident half-year and full accident year on behalf of GISA. He has extensive experience with and knowledge of the Canadian automobile insurance market.

1.7 Conditions and Limitations (Limitations on Accuracy of Results)

1.7.1 Inherent Uncertainty

It must be understood that estimates of ultimate claim costs, claim cost trends, and Reforms Impacts are subject to large potential errors of estimation. The ultimate disposition of claims, whether incurred prior to or following the as of date, is subject to the outcome of events that have not yet occurred. Examples of these events include jury decisions, court interpretations, legislative changes, public attitudes, and social/economic conditions such as inflation. Any estimate of future costs underlying claim liabilities is subject to the inherent limitation on one’s ability to predict the aggregate course of future events. It should therefore be expected that the actual emergence of claims will vary, perhaps materially, from any estimate. The true ultimate claim costs for occurrence prior to December 31, 2013 will only be known after the passage of time – when the last claim has closed. Thus, no assurance can be given that the actual ultimate claim costs at December 31, 2013, for the Ontario PPA insurance industry as a whole, will not ultimately differ from the estimates contained herein.

In our judgment, we employ techniques and assumptions that are appropriate, and the conclusions presented herein are reasonable given the information currently available.

1.7.2 Extraordinary Future Emergence and Limitations of Ontario Private Passenger Automobile Historical Database

The Reforms are taken into account explicitly through the use of Reforms adjustment factors and judgmentally through our selections of actuarial projection factors (such as age-to-factors and trend factors) and methodologies for selecting ultimate claims for the most recent three accident years.

The estimates of ultimate claim costs make no provision for potential future claims arising from causes not contained in the historical data. As our analysis is based entirely on the underlying claim data from Ontario PPA at the all-industry level, the projections will only include those categories of claims that are included within the historical experience. Thus, to the extent that certain claim types are either not present or have very limited frequency in the database, our future claim projections will not provide for such claims.

1.7.3 Rounding

All figures in the supporting exhibits are carried to a greater number of decimals than shown. Thus, totals and calculations may not agree due to rounding.

2 KEY RESULTS AND FINDINGS

2.1 Key Results and Findings

This section summarizes our estimates of ultimate claim costs, claim cost trends and Reforms Impacts for Ontario PPA based on GISA published data as of December 31, 2014. The underlying claim projection approach is documented in Section 3.3, while the derivation of claim cost trends and Reforms Impacts is presented in Sections 3.4. Our quantitative analysis is further detailed in exhibits contained in Section 4.

2.1.1 Identification of Sub-Coverages Analysed

Table 1 shows the injury and non-injury sub-coverages reviewed as part of the Independent Actuarial Analysis. Each sub-coverage listed in the table is analysed in a specific segment of Exhibit A. For ease of reference, these segments are identified in the third column of Table 1.

Table 1: Sub-Coverages
Coverage Group Sub-Coverage Segment
Injury Third Party Liability – Bodily injury (TPL-BI)
Accident Benefits – Disability Income (AB-DI)
Accident Benefits – Death Benefits (AB-DB)
Accident Benefits – Funeral (AB-F)
Accident Benefits – Medical & Rehabilitation (AB-MR)
Accident Benefits – Supplementary (AB-SUP)
Accident Benefits – Uninsured Automobile (AB-UA)
Underinsured Motorist (UM)
I
IV
V
VI
VII
VIII
XII
XIII
Non-injury Third Party Liability – Property Damage (TPL-PD)
Third Party Liability – Direct Compensation (TPL-DC)
Collision (COL)
Comprehensive (COM)
Specified Perils (SP)
II
III
IX
X
XI

2.1.2 Ultimate Claim Costs

We calculate the ultimate claim costs for Ontario PPA as of December 31, 2013 based on industry data as reported by GISA. The dataset is net of the same data exclusions stemming from GISA’s validation procedures.3 Section 3.3.1 describes the actuarial projection methods used for this analysis. Table 2 summarizes the overall results.

Table 2: Summary of Accident Year Direct Premium and Claim based on GISA Data
($m) Ontario PPA
Accident
Year
Earned
Premium
Incurred
Claim and
Expense4
Ultimate
Claim and
Expense5
Accident Year
Claim Ratio
2008 8,373 6,408 6,895 82%
2009 8,729 7,678 8,254 95%
2010 9,410 7,821 8,338 89%
2011 10,038 6,083 6,698 67%
2012 10,441 6,051 6,678 64%
2013 10,585 6,910 7,329 69%

The results presented in Table 2 indicate an improvement of 22 percentage points in the accident year claim ratio from 2010 to 2011. This improvement is explained by a decline in estimated ultimate claims and claim adjustment expenses of $1.6 billion, together with an increase in earned premiums of $0.6 billion from 2010 to 2011. Similarly, the results presented in Table 2 indicate that the accident year claim ratio improved by 25 percentage points from 2010 to 2012, due to estimated ultimate claims and claim adjustment expenses decreasing by $1.7 billion while earned premiums increased by $1.0 billion. The very preliminary estimate for the 2013 claim ratio still shows an improvement of 20 percentage points when compared to 2010 and a deterioration of 5 percentage points when compared t0 2012.

As the claims from a given accident year mature, the estimate of ultimate claim costs can normally be expected to become more certain, and the various actuarial projection methods will tend to converge to a narrower range of estimates. Section 3.3.3 highlights the variability of estimates of ultimate claim costs for TPL-BI, AB-DI and AB-MR for accident year 2011 and onward. Because of this variability, estimates of ultimate claims costs for these post-Reforms accident years continue to be very uncertain. As such, on-going monitoring of the estimates will be necessary in order to measure the performance of Ontario PPA insurance over this time period.

2.1.3 Claim Cost Trends and Reforms Impacts

For each automobile insurance sub-coverage listed in Table 1, we estimate the claim cost trends and the Reforms Impacts. Our analysis is based on the industry data reported by GISA as of December 31, 20136.

Section 3.2 summarizes the changes introduced by the Reforms. The following sub-coverages were most affected by the Reforms: TPL-BI, AB-DI and AB-MR. Other Accident Benefits sub-coverages such as AB-DB and AB-F were also affected but to a lesser degree. Table 3 summarizes our best estimates of the impact of Reforms on claim cost trends and claim cost levels based on the analysis and the methodologies described in Section 3.4.

Our analysis indicates that there is a strong inter-dependency between the estimated claim cost trends (both pre- and post-Reforms) and the impact of Reforms on claim cost levels. Consequently, estimates of claim cost trends, pre- and post-Reforms, as well as the estimates of Reforms Impacts should be extracted from the same modelling approach with the same underlying data and assumptions, and should be read and interpreted conjointly for a given coverage and sub-coverage. In other words, Reforms Impacts on claim cost levels shown in the last column of Table 3: Summary of Estimated Claim Cost Trends and Reforms Impacts should not be used in combination with claim cost trends stemming from another model or data source.

Table 3 : Summary of Estimated Claim Cost Trends and Reforms Impacts
Sub-Coverages Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Third Party Liability – Bodily Injury (TPL-BI) 5.3% 5.3% 0%
TPL-Property Damage (TPL-PD) 1.4% 1.4% 0%
TPL-Direct Compensation Property Damage (TPL-DC) 0.8% 0.8% 0%
Underinsured Motorist (UM) 4.0% 4.0% 0%
Total TPL+UM 3.9% 4.0% 0%
Accident Benefits – Medical and Rehabilitation (AB-MR) 17.0% 0.7% -48.1%
AB-Disability Income (AB-DI) 7.6% -0.2% -27.4%
AB-Funeral Expenses (AB-F) -6.2% -6.2% 0%
AB-Death Benefits (AB-DB) -8.4% -2.7% 0%
AB-Supplementary (SUP) 0.0% 0.0% 0%
Uninsured Automobile (AB-UA) 4.4% 4.4% 0%
Total AB+UA 14.9% 0.7% -43.2%
Total Compulsory Coverages 10.2% 2.7% -25.7%
Collision (COL) -1.4% -1.4% 0%
Comprehensive (COM) -3.2% -3.2% 0%
Total Physical Damages (PhysD) -1.9% -1.9% 0%
Total All Coverages 8.1%7 1.7%8 -21.6%9

Table 3 shows that:

  • The pre-Reforms claim cost trend was significant at 10.2% per annum for Total Compulsory Coverages.
  • The Reforms appear to have had a significant impact on claim cost trends; based on the first three years of data, the claim cost trend post-Reforms is estimated at 2.7% for Total Compulsory Coverages.
  • With the Reforms, the reduction of average claim cost level is estimated at 25.7% for Total Compulsory Coverages.

We note that any estimates or analysis of claim cost trends and Reforms Impacts are highly dependent on the dataset, as well as the modeling approach, length of data history, and assumptions. In Section 3.5, we perform sensitivity analyses on the data source, the length of historical period and the iterative model. The following charts illustrate the variability of estimates for pre- and post-Reforms trends and Reforms Impacts for the each of the three main coverages affected by the Reforms.

Pre-Reforms Trend Estimates
Post-Reforms Trend Estimates
Reforms Impact Estimates

The charts highlight the significant variability in pre- and post-Reforms claim cost trend estimates. The variability of the post-Reforms claim cost trends is not surprising since the estimates are only based on three years of data. These results should be considered to be early estimates based on the most recent data available. The demarcation between new emerging claim cost trends and volatility in claim costs is blurry. As such, the claim cost trends post-Reforms are subject to a higher degree of uncertainty. For instance, Exhibit A, Page 7a of the corresponding sections show that TPL-BI, AB-MR and AB-DI estimated claim costs have increased by 16.5%, 13.8% and 16.0% respectively between 2012 and 2013. Approximately 7% of the increase may be due to the harsh winter, as observed by the change in estimated TPL-DC frequency on Exhibit A, Section III, Page 7a. These recent claim cost movements also stress the importance of continuous monitoring of claim cost trends and Reforms Impacts. More importantly, it points to the necessity of determining and thoroughly documenting the estimates of pre- and post-Reforms trends as well as the Reforms Impacts that are specific for each insurer’s circumstances.

Section 3.4.2 of this Independent Actuarial Analysis mentions that a noticeable temporary surge in claim frequency appears to have affected the Ontario PPA insurance results between mid-2009 and the end of 2010. Possible explanation for such a surge could be linked to an acceleration in claims filing, treatment and procedure requests, and partial or full claim settlements in anticipation of the Reforms. Based on the available data, it would appear that the claim surge resorbed itself quickly after the adoption of the Reforms. This claim surge would not have been anticipated at the time the Reforms were devised. In this analysis, to avoid overestimating the impact of the Reforms on claim cost levels, the estimates were adjusted to exclude the effect of the temporary claim surge and the following self-correction. As such, to fully measure the change in claim costs since 2009, the adjusted Reforms Impacts would need to be combined with further frequency adjustments reflecting the temporary claim surge experienced in accident half-years 2009-12, 2010-06 and 2010-12. Based on the analysis performed, this frequency adjustment could possibly represent 8% to 10% of the AB-MR and AB-DI claim costs during that period.

3 METHODOLOGY

In Section 3 of the Independent Actuarial Assessment, we:

  • Discuss the data used for the quantitative analysis;
  • Summarize key coverage changes arising from the Reforms;
  • Summarize our analysis of estimated ultimate claim costs by sub-coverage. This summary includes a discussion of the methodologies used to estimate the ultimate claim costs;
  • Discuss the Reforms Impacts and post-Reforms trends by sub-coverage; and
  • Discuss the sensitivity of the estimates of trends and Reforms Impacts to different choices of data, modelling approach, and assumptions.

3.1 DATA

The SOP outline the responsibilities of the actuary with respect to data. According to Section 1530.02 of the SOP:

The work with respect to data consists of

  • identifying the data needed,
  • attempting to obtain them,
  • reviewing the data obtained, and
  • assessing sufficiency and reliability of the data obtained.

3.1.1 Identifying the Data Needed

To conduct the quantitative analysis for Ontario PPA experience at the industry level, we identify the following data requirements by coverage and sub-coverage:

  • Development triangles on direct basis for the following categories of claims (including ALAE):
    • Paid
    • Case estimates
    • Reported (sum of paid and case estimates)
    • Closed counts (including closed with payment and closed without payment)
    • Open (also referred to as pending) counts
    • Reported counts; and
  • Exposure information by accident year including earned premium and earned exposures.

3.1.2 Attempting to Obtain the Data

All of the data identified in Section 3.1.1 is available through the PPA Exhibits published by GISA.

3.1.3 Reviewing the Data Obtained

Each of the insurers underwriting automobile insurance in Canada submit data under the Automobile Statistical Plan to IBC who then assembles the data underlying the PPA GISA Exhibits. IBC performs numerous data edit checks, which are designed to promote data integrity. It is impractical for us to independently audit or verify the data due to the number of companies submitting data and our remoteness from the individual data elements. Therefore, we rely on the IBC/GISA data without independent audit or verification.

While we do not audit or independently verify the data provided by IBC on behalf of GISA, we do review the data for reasonableness and consistency. We also reconcile the data received electronically to the published reports. The underlying claim, claim adjustment expense, premium, and exposure data form the basis of our analysis and our findings and conclusions. Thus, the accuracy of this data is critical.

GISA bulletin 2014-61 titled 2013 Automobile Exhibit Introduction and Actual Loss Ratio Exhibit – Ontario outlines the validation and verification procedures as well as the consistency and reasonability checks that IBC performed, leading to some data exclusions. As mentioned in GISA bulletin 2014-61,

Every effort has been made to ensure the accuracy and completeness of the data. The responsibility for any errors or omissions in the data submitted under the statistical plan and presented in these exhibits remains with the reporting companies.10

The data contained in the 2013-2 PPA GISA Exhibits are deemed reliable and sufficient to perform the quantitative analysis requested by the MOF; and we use this data as published by GISA without modification.

3.1.4 Assessing Sufficiency and Reliability of the Data Obtained

As noted above, while we do not audit or independently verify the data provided by IBC on behalf of GISA, we do review the data for reasonableness and consistency. We also reconcile the data received electronically to the published reports.

The data contained in the 2013-2 PPA GISA Exhibits are deemed reliable and sufficient to perform the quantitative analysis requested by the MOF; and we use this data as published by GISA without modification.

3.1.5 GISA Data Limitations

GISA bulletin 2014-61 mentions:

The incurred claim amount factors, and the derived ultimate claim counts, for the most recent periods in the Accident Benefits (AB) coverage, are subject to a larger degree of uncertainty than usual. This uncertainty arises from many companies reporting large increases in case reserves in accident year 2009, but in greater part due to a significant change in the case reserving process of a major writer impacting the 2009-1 and subsequent diagonals. This change has had a material impact on the observed Private Passenger incurred claim amount factors for certain AB sub-coverage groups. As a result, the Private Passenger incurred claim amount factors were calculated on an all-industry basis excluding this major writer and applied to the all-industry data (including this major writer) for certain AB sub-coverage groups.11

For confidentiality reasons, GISA does not release the detailed data that would allow actuaries to independently assess the data exclusions and to reproduce GISA’s selection of development factors.

The published GISA data are not granular enough to allow similar exclusions. Thus reported claims, reported development factors, ultimate claim estimates, claim cost trends and Reforms Impacts, as outlined in Sections 3.3 and 0, are all derived using the all-industry basis.

3.1.6 Data Assumptions

According to the SOP, data assumptions are the assumptions, if any, needed to relieve insufficiency or unreliability in the obtainable data. There are no data assumptions required for the quantitative analysis presented in this Appendix, and we conclude that the data published by GISA is sufficiently reliable for the purpose of this analysis, notwithstanding the possible distortion due to significant changes in the case reserving process of a major writer.

3.2 Summary of Changes Introduced by the Reforms

On September 1, 2010, a new Statutory Accident Benefits Schedule (SABS) became effective in Ontario. Table 4 summarizes the key changes introduced by the Reforms:

Table 4: Coverage Changes Introduced with the Reforms
  Pre-Reforms12 Post-Reforms13
Medical Payments $100,000/person ($1 million if injury "catastrophic"), including rehabilitation, excluding health insurance and other medical plans; attendant care $72,000 ($1 million if injury "catastrophic"). Up to $3,500 for minor injury; up to $50,000/person for non-minor and non-catastrophic injury for up to 10 years; up to $1 million for catastrophic injury; attendant care up to $36,000 for non-minor and non-catastrophic injury up to 104 weeks; up to $1 million for catastrophic injury.
Funeral expense benefits: $7,843 (indexed) $6,000 (if optional indexation coverage is purchased, this amount may be higher).
Disability income benefits Income Replacement Benefit 80% of net wages up to $400/week, minimum $185/week; for 104 weeks maximum (longer if victim is unable to pursue any suitable occupation); capped at 12 weeks for Whiplash Associated Disorder I (WAD I) injuries and 16 weeks for WAD II injuries; seven-day wait.
Non-Earner Benefit (disabled unemployed persons, students enrolled in education full-time, or students who completed their education less than one year before the accident and are not employed) $185/week for 104 weeks; 26-week wait; indexed; limit two years; if student (as defined above) is still disabled after 104 weeks, Non-Earner Benefit is $320/week.
Income Replacement Benefit: 70% of gross wages to maximum $400/week, minimum $185/week for 104 weeks (longer if victim is unable to pursue any suitable occupation); nothing is payable for the first seven days of disability.
Non-earner Benefit (disabled unemployed persons, students enrolled in education full time, or students who completed their education less than one year before the accident and are not employed): $185/week for 104 weeks; 26-week wait; limit two years; if student (as defined above) is still disabled after 104 weeks, Non-earner Benefit is $320/week. Not available if the insured is eligible for, and elects to receive, the income replacement or caregiver benefit.
Death benefits Death within 180 days (or three years if continuously disabled prior to death); $65,360 minimum to spouse; $13,072 to surviving dependant; death of dependant $13,072 (indexed). Death within 180 days of accident (or three years if continuously disabled prior to death); $25,000 minimum to spouse, $10,000 to each surviving dependant, $10,000 to each parent/guardian (if optional indexation coverage is purchased, these amounts may be higher).
Right to sue for pain and suffering? Yes, if injury meets verbal threshold; subject to deductible. Lawsuit allowed only if injured person dies or sustains "permanent and serious" disfigurement and/or impairment of important physical, mental or psychological function. The court assesses damages and deducts $30,000 ($15,000 if Family Law Act claim). Yes, if injury meets severity test (called “threshold”), and subject to deductible. Lawsuit allowed only if injured person dies or sustains permanent and serious disfigurement and/or impairment of important physical, mental or psychological function. The court assesses damages and deducts $30,000 ($15,000 for a Family Law Act claim).
Right to sue for economic loss in excess of no-fault benefits? Yes. Injured person may sue for 80% of net income loss before trial, 100% of gross after trial; also for medical, rehabilitation and related costs when injury meets verbal threshold for pain and suffering claims. Yes. Income replacement award above no-fault benefit is based on net income after deductions for income tax, Canada Pension and Employment Insurance.
Injured person may sue for 70% of net income loss before trial, 100% of gross after trial; also for medical, rehabilitation and related costs when injury meets severity test for pain and suffering claims.

Although the new standard auto insurance policies began to apply on or after September 1, 2010, some of the Reforms changes also applied to:

  • Auto insurance claims that occurred after November 1, 1996 and were open on or after September 1, 2010; and
  • Auto insurance policies that were in-force as of September 1, 2010, before their renewal on or after September 1, 2010.

In this Independent Actuarial Analysis, we consider the effect of the Reforms on the estimates of ultimate claim costs. We assumed that the Ontario automobile insurance industry incorporated changes into its case estimates where relevant, and we adjust actuarial assumptions for accident years 2010 through 2013 to explicitly reflect our expectations of the influence of the Reforms on recent and future claim experience in Ontario.

3.3 Ultimate Claim Costs

In Section 3.3 of this Independent Actuarial Analysis, we summarize our analysis of estimated ultimate claim costs by sub-coverage. We begin this summary with a discussion of the methodologies used to estimate the ultimate claim costs.

3.3.1 General Description of Methodologies

For both injury and non-injury coverage groups, ultimate claims are generally selected based on three traditional P&C actuarial methods: claim development, frequency-severity, and Bornhuetter-Ferguson. An important input to the Bornhuetter-Ferguson method is an estimate of expected claims, derived from the expected claims method using earned premium and an initial expected claim ratio. For the frequency-severity method, ultimate frequencies (i.e., ultimate counts per vehicle) are selected based on a comparison of projected ultimate counts to earned vehicles by accident year. Ultimate severities are selected based on a review of the development method applied to average paid values and to average incurred values. Selected severities for accident years 2010 through 2013 are explicitly adjusted to reflect the cost levels recognizing recent trends and the Reforms in Ontario. Below, we briefly describe each of these methods. For further information, the reader is referred to the Casualty Actuarial Society (CAS) text titled Estimating Unpaid Claims Using Basic Techniques, which was written by Jacqueline Friedland (Version 3), July 30, 2010, http://www.casact.org/pubs/Friedland_estimating.pdf

3.3.1.1 Development Method

The development method, also known as the chain ladder method, is one of the most frequently used methodologies for estimating unpaid claims. The distinguishing characteristic of the development method is that ultimate claims for each accident year are produced from recorded values assuming that future claims’ development is similar to prior years’ development. In this method, the actuary uses the development triangles to track the development history of a specific group of claims. The underlying assumption in the development technique is that claims recorded to date will continue to develop in a similar manner in the future – that the past is indicative of the future. That is, the development technique assumes that the relative change in a given year’s claims from one evaluation point to the next is similar to the relative change in prior years’ claims at similar evaluation points.

An implicit assumption in the development technique is that, for an immature accident year, the claims observed thus far tell you something about the claims yet to be observed. This is in contrast to the assumptions underlying the expected claims and Bornhuetter-Ferguson methods.

Other important assumptions of the development method include: consistent claim processing, a stable mix of types of claims, stable policy limits, and stable reinsurance/retention limits throughout the experience period.

3.3.1.2 Frequency/Severity Method

The frequency/severity method can be extremely valuable, especially when an organization is facing change arising from either internal or external forces. When using the frequency/severity method, the actuary is able to analyze the drivers of the ultimate claim values by examining the frequency and severity components separately and then multiplying the selected values to project the ultimate claim amounts. In the simplest form, the frequency/severity method multiplies the selected frequency rate by earned exposures in order to produce an estimate of ultimate counts; the ultimate counts are then multiplied by the selected severity value (i.e., the average cost per claim) to project the ultimate claims.

Two key assumptions of the frequency/severity method include: (a) claim counts are defined consistently over time, and (b) the mix by type of claim is reasonably consistent. Other important assumptions of the frequency/severity method are related to the adjustments used to restate historical experience at the cost level expected for future accident years. These are very significant assumptions for Ontario PPA, as historical experience is used to project ultimate claims for the most recent accident years, which are subject to a substantially different environment due to trend and Reforms.

3.3.1.3 Bornhuetter-Ferguson Method

Actuaries rely on the Bornhuetter-Ferguson method almost as often as they rely on the development method. The Bornhuetter-Ferguson method is essentially a blend of the development and expected claims methods (the frequency-severity method is used to determine the expected claims in this quantitative analysis). In the development method, we multiply actual claims by a cumulative claim development factor. This method can lead to erratic, unreliable projections when the cumulative development factor is large because a relatively small swing in reported claims or the reporting of an unusually large claim could result in a very large swing in projected ultimate claims. In the expected claims method, the unpaid claim estimate is equal to the difference between a predetermined estimate of expected claims and the actual payments. This has the advantage of stability, but it completely ignores actual results reported to date. The Bornhuetter-Ferguson method combines the two methods by splitting ultimate claims into two components: actual reported (or paid) claims and expected unreported (or unpaid) claims. As experience matures, more weight is given to the actual claims and the expected claims become gradually less important.

In the 1993 paper “Loss Development Using Credibility,”14 Eric Brosius described the Bornhuetter-Ferguson method as a credibility weighting between the development method and the expected claims method. In the development method, full credibility (i.e., Z = 1) is given to actual claim experience; and in the expected claims method, no credibility (i.e., Z = 0) is given to actual claims. In the Bornhuetter-Ferguson method, credibility is equal to the percentage of claims developed at a particular stage of maturity, which is a function of the cumulative claim development factor (i.e., Z = 1.00 / cumulative development factor). Therefore, more weight is given to the expected claims method in less mature years, and more weight is given to the development method in more mature years of the experience period.

The key assumption of the Bornhuetter-Ferguson method is that unreported (or unpaid) claims will develop based on expected claims. In other words, the claims reported to date contain no informational value as to the amount of claims yet-to-be reported. This is different from the development method where the primary assumption is that unreported (or unpaid) claims will develop based on reported (or paid) claims to date.

The reporting and payment patterns used in the Bornhuetter-Ferguson methods are the same as those selected in the development method. However, the application of the development factors differs between the two methods. It is also important to note that the expected claims used in the Bornhuetter-Ferguson method using reported claims are the same as those used in the Bornhuetter-Ferguson method using paid claims.

3.3.2 Key Assumptions of the Projection Methodologies

3.3.2.1 Development Patterns

The claim development analysis is conducted by sub-coverage based on accident half-year data at semi-annual valuations through December 31, 2013. For each sub-coverage, we typically calculate the following average age-to-age factors:

  • Simple average of the latest five, four and three years;
  • Medial average (i.e., average excluding high and low values) of the latest five, four and three years; and
  • Volume weighted average of the latest five, four and three years.

In selecting age-to-age factors for each age-to-age interval for each sub-coverage, we consider all of the following:

  • Variability in the individual age-to-age factors in both the short-term and the long-term;
  • Congruence of various average age-to-age factors, both short-term averages and long-term averages;
  • The number of recent age-to-age factors in each interval that are greater than or less than the various average values;
  • The influence of trends in severity and frequency of claims as well as the Reforms;
  • The influence of external factors in the historical claim experience (such as recent trends in referred mediations) as well as expectations for the effect of such actions in the future; and
  • The expected progression of the development patterns.

In our discussion of estimated ultimate claims, we document the selection of age-to-age factors for each sub-coverage.

Because of the highly leveraged nature of paid claims’ cumulative development factors, we do not use methods based on paid claim data for selecting ultimate claims for injury sub-coverages.

Nevertheless, the paid projection methods are included as a reasonableness check of the reported projection methods for the older, more mature accident years.

Large, highly leveraged claim development factors for the most recent three to five accident years lead to greater uncertainty in the projections and thus the inability to reliably use paid projection methods for selecting ultimate claims for the injury sub-coverages.

For all categories of data and sub-coverages, we rely on judgement in selecting factors for ages 144 months and beyond.

3.3.2.2 Initial Expected Claims

The initial expected claim is a critical component of the Bornhuetter-Ferguson method. We select initial expected claim for each line of business and each accident year based on a review ultimate claims derived from the frequency/severity method. This projection method includes adjustments for trends and Reforms Impacts. Section 3.4.2 describes the approach used to determine estimates of frequency and severity trends as well as estimates of Reforms Impacts for each sub-coverage.

3.3.3 Ultimate Claims by Sub-Coverage

The exhibits to this Independent Actuarial Analysis summarize key actuarial assumptions and selections, by sub-coverage and accident half-year. For each segment, we present:

  • The summary of ultimate claim estimates on Page 1;
  • The selected frequency and severity assumptions on Page 2;
  • The selected reported development factors on Page 3;
  • The selected paid development factors on Page 4; and
  • The selected count development factors on Page 5.

We do not repeat all of these details in the text description that follows. Instead, for each sub-coverage, we summarize at a high level the following:

  • Methodology for selecting ultimate claims;
  • Special considerations in the selection of age-to-age factors; and
  • Other special considerations.
3.3.3.1 Third Party Liability – Bodily injury

For TPL-BI, shown in Exhibit A, Segment I, Page 3, we generally select the reported claim development factors based on the average of the latest six accident half-years (simple, medial or volume weighted). The intent of selecting such shorter term averages is to be more responsive to the most recent experience, post-Reforms, and to minimize the influence of some unusually large age-to-age factors in the experience period.

We judgmentally select a tail factor of 1.002 to reflect the magnitude of the reported claims still unpaid after 144 months.

For the frequency-severity model, presented in Exhibit A, Segment I, Page 6, we select a severity trend factor of -1.2% per annum and Reforms Impacts factor of 0%. As shown in Exhibit A, Segment I, Page 2, the selected 2013-06 and 2013-12 ultimate severity of $138,494 and $149,111 are based on the average of five corresponding accident half-years from 2008 through 2012. The volatility of the frequency is smoothed using the long-term relationship between TPL-BI frequency and TPL-DC frequency. The resulting selected TPL-BI frequency varies between 0.22% and 0.24% in the latest 6 accident half-years.

As shown in Exhibit A, Segment I, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2011-06 onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $3.6 billion.

Figure 1 : TPL-BI - Estimates of Ultimate Claims ($’000)  for AY 2004 to 2013

Figure 1 : TPL-BI - Estimates of Ultimate Claims ($’000) for AY 2004 to 2013

In Figure 1 , the blue line represents the estimates of ultimate claims resulting from our selected methods for TPL-BI by accident half-year. The red line corresponds to the ultimate claim estimates resulting from GISA selected claim development factors. A comparison of these two lines indicates that our estimates of ultimate claims are within reasonable range of GISA’s estimates for TPL-BI. Furthermore, the area shaded in green illustrates the underlying variability of the ultimate claim estimates as delimited by the minimum and maximum values of all the actuarial projection methodologies used to estimate the TPL-BI ultimate claims in Exhibit A, Segment I, Page 1. As the gap between the lower and upper bounds widens, the variability, and therefore the uncertainty associated with a particular selection of ultimate claim amount, increases. As such, it can be observed from Figure 1: TPL-BI - Estimates of Ultimate Claims ($’000) for AY 2004 to 2013 that accident half-years 2009-12 and onward continue to show important levels of uncertainty.

3.3.3.2 Third Party Liability – Property Damage

For TPL-PD, shown in Exhibit A, Segment II, Page 3, we generally select the same reported claim development factors as GISA, with the exception of 36-42 and 42-48 where we smoothed the development factors.

For the frequency-severity model, presented in Exhibit A, Segment II, Page 6 we select the severity trend factor of 2.8% per annum. As shown in Exhibit A, Segment II, Page 2, the selected 2013-06 and 2013-12 ultimate severity of $4,840 and $5,092 are based on the average of five corresponding accident half-years from 2008 through 2012. The resulting selected TPL-PD frequency varies between 0.14% and 0.16% in the last 16 accident half-years.

As shown in Exhibit A, Segment II, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2013-06 and 2013-12. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $20.3 million.

3.3.3.3 Third Party Liability – Direct Compensation

TPL-DC claims tend to have a very short settlement period. As shown in Exhibit A, Segment III, Page 1, the ultimate claims resulting from Reported Development method were selected for all accident half-years. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $14.9 million, which approximately corresponds to the claim volume being reported in less than one week.

3.3.3.4 Accident Benefits – Disability Income

For AB-DI, shown in Exhibit A, Segment IV, Page 3, we select the reported claim development factors based on the simple average of the latest four accident half-years from development periods 6-12 to 48-54 inclusively. The medial of the latest eight accident half-years is generally selected for the later periods with a few smoothing points. The intent of such shorter term average is to be more responsive to the most recent experience, post-Reforms, and minimize the influence of some unusually large age-to-age factors in the experience period.

We judgmentally select the tail factor of 1.021 to reflect the nature of the claims, the magnitude of the reported claims still unpaid after 144 months and the scale of development in the periods preceding 144 months.

For the frequency-severity model, presented in Exhibit A, Segment IV, Page 6, we select a pre-Reforms severity trend factor of 7.6% per annum, post-Reforms severity trend factor of -0.2% and a Reforms Impact factor of -27.4%. As shown in Exhibit A, Segment IV, Page 2, the selected 2013-12 ultimate severity of $27,717 is based on the medial average of ten accident half-years from 2008 through 2012. The volatility of the frequency is smoothed using long-term relationship between AB-DI frequency and TPL-DC frequency. The resulting selected AB-DI frequency varies between 0.21% and 0.23% in the latest 6 accident half-years.

As shown in Exhibit A, Segment IV, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2011-06 and onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $0.6 billion.

Figure 2 : AB-DI - Estimates of Ultimate  Claims ($’000) for AY 2004 to 2013

Figure 2 : AB-DI - Estimates of Ultimate Claims ($’000) for AY 2004 to 2013

In Figure 2 , the blue line represents the estimates of ultimate claims resulting from our selected methods for AB-DI by accident half-year. The red line corresponds to the ultimate claim estimates resulting from GISA selected claim development factors. A comparison of these two lines indicates that our estimates of ultimate claims are within reasonable range of GISA’s estimates for AB-DI. Furthermore, the area shaded in green illustrates the underlying variability of the ultimate claim estimates as delimited by the minimum and maximum values of all the actuarial projection methodologies used to estimate the AB-DI ultimate claims in Exhibit A, Segment IV, Page 1. As the gap between the lower and upper bounds widens, the variability, and therefore the uncertainty associated with a particular selection of ultimate claim amount, increases. As such, it can be observed from Figure 2 that accident half-years 2011-06 and onward continue to show important levels of uncertainty.

3.3.3.5 Accident Benefits – Death Benefits

AB-DB claims tend to have a relatively quick reporting and predictable development. The severity volatility is narrow, leading to reasonably predictive average reported claim on an all-industry basis. As shown in Exhibit A, Segment V, Page 1, the ultimate claims resulting from Reported Development method were selected for all accident half-years. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to -$0.5 million.

3.3.3.6 Accident Benefits – Funeral

AB-F claims tend to have a relatively quick reporting and predictable development. The severity volatility is narrow, leading to reasonably predictive average reported claim on an all-industry basis.

As shown in Exhibit A, Segment VI, Page 1, the ultimate claims resulting from Reported Development method were selected for all accident half-years. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to -$0.1 million.

3.3.3.7 Accident Benefits – Medical & Rehabilitation

For AB-MR, shown in Exhibit A, Segment VII, Page 3, we select the reported claim development factors based on the simple average of the latest three accident half-years from development periods 6-12 to 36-42 inclusively. The volume-weighted average of the latest six accident half-years is generally selected for the later periods with a few smoothing points. The intent of such short term average is to be more responsive to the most recent experience, post-Reforms, and minimize the influence of some unusually large age-to-age factors in the experience period.

We judgmentally select the tail factor of 1.012 to reflect the nature of the claims, the magnitude of the reported claims still unpaid after 144 months and the scale of development in the periods preceding 144 months.

For the frequency-severity model, presented in Exhibit A, Segment VII, Page 6, we select a severity pre-Reforms trend factor of 17.0% per annum, a severity post-Reforms trend factor of 0.7% and Reforms Impacts factor of -48.1%. As shown in Exhibit A, Segment VII, Page 2, the selected 2013-12 ultimate severity of $31,009 is based on the medial average of ten accident half-years from 2008 through 2012. The volatility of the frequency is smoothed using long-term relationship between AB-MR frequency and TPL-DC frequency. The resulting selected AB-MR frequency varies between 0.70% and 0.76% in the latest 6 accident half-years.

As shown in Exhibit A, Segment VII, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2011-06 and onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $1.9 billion.

Figure 3 : AB-MR - Estimates of Ultimate  Claims ($’000) for AY 2004 to 2013

Figure 3 : AB-MR - Estimates of Ultimate Claims ($’000) for AY 2004 to 2013

In Figure 3 , the blue line represents the estimates of ultimate claims resulting from our selected methods for AB-MR by accident half-year. The red line corresponds to the ultimate claim estimates resulting from GISA selected claim development factors. A comparison of these two lines indicates that our estimates of ultimate claims are within reasonable range of GISA’s estimates for AB-MR. Furthermore, the area shaded in green illustrates the underlying variability of the ultimate claim estimates as delimited by the minimum and maximum values of all the actuarial projection methodologies used to estimate the AB-MR ultimate claims in Exhibit A, Segment VII, Page 1. As the gap between the lower and upper bounds widens, the variability, and therefore the uncertainty associated with a particular selection of ultimate claim amount, increases. As such, it can be observed from Figure 3 that accident half-years 2011-06 and onward continue to show important levels of uncertainty.

3.3.3.8 Accident Benefits – Supplementary

For AB-SUP, shown in Exhibit A, Segment VIII, Page 3, we select the same reported claim development factors as GISA.

For the frequency-severity model, presented in Exhibit A, Segment VIII, Page 6, we select a severity trend factor of 0% per annum. As shown in Exhibit A, Segment VIII, Page 2, the selected 2013-12 ultimate severity of $42,944 is based on the medial average of ten accident half-years from 2008 through 2012. The AB-SUP frequency is extremely low and we judgmentally select a long-term average of three claim counts per accident half-year.

As shown in Exhibit A, Segment VIII, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2009-06 and onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $0.5 million.

3.3.3.9 Collision Including the All Perils Component

For collision, shown in Exhibit A, Segment IX, Page 3, we select the same reported claim development factors as GISA. Collision claims tend to have a relatively quick reporting and predictable development. As shown in Exhibit A, Segment IX, Page 1, the selected ultimate claims for prior periods are stemming from the Reported Development method for all accident half-years. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to -$4.1 million.

3.3.3.10 Comprehensive Including the All Perils Component

Comprehensive claims tend to have a relatively quick reporting and predictable development. As shown in Exhibit A, Segment X, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for accident year 2013. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $14.8 million, which corresponds to the claim volume being reported in approximately 2 weeks.

3.3.3.11 Specified Perils

Specified perils claims experience a low claim volume. We select the same reported claim development factors as GISA. As shown in Exhibit A, Segment XI, Page 1, the ultimate claims resulting from Reported Development method were selected for all accident half-years. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $10 thousand.

3.3.3.12 Accident Benefits – Uninsured Automobile

For AB-UA, shown in Exhibit A, Segment XII, Page 3, we select the same reported claim development factors as GISA.

For the frequency-severity model, presented in Exhibit A, Segment XII, Page 6, we select a severity trend factor of 12.0% per annum. The selected 2013-12 ultimate severity of $81,135 is based on the medial of eight accident half-years from 2008 through 2012. The AB-UA frequency varies between 0.02% and 0.03% since 2008.

As shown in Exhibit A, Segment XII, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2011-06 and onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $228.0 million.

3.3.3.13 Underinsured Motorist

For UM, shown in Exhibit A, Segment XIII, Page 3, we generally select the reported claim development factors based on the volume-weighted average of the latest six accident half-years with some smoothing after 60 months of development. The intent of such short term average is is to be more responsive to the most recent experience, post-Reforms, and minimize the influence of some unusually large age-to-age factors in the experience period.

We judgmentally select the tail factor of 1.004 to reflect the magnitude of the reported claims still unpaid after 144 months.

For the frequency-severity model, presented in Exhibit A, Segment XIII, Page 6, we select a severity trend factor of 14.0% per annum prior to 2007 and -7.3% afterwards. As shown in Exhibit A, Segment XIII, Page 2, the selected 2013-12 ultimate severity of $185,630 is based on the average of ten accident half-years from 2008-2012. The AB-UA frequency is steadily increasing since 2008 from 0.003% and 0.005%.

As shown in Exhibit A, Segment XIII, Page 1, the ultimate claims resulting from the Bornhuetter-Ferguson Reported method were selected for 2011-06 and onward. The selected ultimate claims for prior periods are stemming from the Reported Development method. The total selected IBNR for accident half-years 2002-06 through 2013-12 amounts to $108.9 million.

3.4 Trends and Reforms Impacts

Section 3.4 discusses the Reforms Impacts and post-Reforms trends by sub-coverage.

3.4.1 General Description of Methodologies

Exhibit A, Pages 6 through 8 of each Segment contain our PPA claim cost trend analysis based on the overall industry experience. We conduct the trend analysis separately for each of the coverages and sub-coverages previously identified. We select different trend factors for pre- and post-Reforms periods for AB-MR, AB-DI, and AB-DB sub-coverages, and select the same trend factors for pre- and post-Reforms periods for all other coverages and sub-coverages.

For this analysis, we rely on ultimate claims and counts stemming from the reported development method to derive the severity and frequency. We do not incorporate a credibility procedure in the trend analysis. It is important to note that the purpose of the analysis is to determine an estimates of the claim cost trends and Reforms Impacts.

We follow a similar approach for the trend analysis of both severity and frequency. Although our exhibits include the claim cost, we do not select claim cost trends directly. We include additional factors such as Reforms Impacts, seasonality, and/or post-Reforms trend if these factors appear significant based on statistical tests. In the top section of Exhibit A, Page 6a, we present graphs comparing actual values to the fitted values. In the middle section of Page 6a, we present tables with the following:

  • Accident half-year ending (2002-06 through 2013-12);
  • Adjustment factor;
  • Adjusted value (claim cost, severity or frequency);
  • Fitted value (claim cost, severity or frequency);
  • Year variable (1 through 24);
  • Indicator variable 1 representing seasonality;
  • Indicator variable 2 representing the impact of Bill 198;
  • Indicator variable 3 representing the Reforms Impacts; and
  • Indicator variable 4 representing the accident half-year variable of post-Reforms trend.

Finally, the bottom portion of Page 6a summarizes the indicated trend rate based on the regression analysis and numerous statistical values derived from the selected regression equation. These values indicate the strength of the regression equation in fitting the historical data and projecting future trend rates. We also show our selected trend rate (highlighted by a box).

For legislation that occurs part way through an accident half-year period, we use a partial indicator variable. For the Reforms, the partial indicator variables are calculated as

ln [6m/(4+2m)] / ln [m]; where m is the estimated product change conversion factor, determined on an iterative basis.

3.4.2 Trends and Reforms Impacts By Sub-Coverage

3.4.2.1 Third Party Liability – Bodily injury

For TPL-BI, shown in Exhibit A, Segment I, Page 6a, we incorporate seasonality and the impact of Bill 198, starting in accident half-year 2003-12, into the regression analysis. We excluded the data points corresponding to 2009-12, 2010-06 and 2010-12, which reflect a temporary claim surge immediately preceding the implementation of the Reforms. Apart from the temporary claim surge, the Reforms do not appear to have affected the claim cost trend or the claim cost level as originally anticipated by the Financial Services Commission of Ontario (FSCO) in its Five-Year Review Reform Simplified Filing Guidelines for Proposed Revisions to Private Passenger Automobile Insurance Rates and Risk Classification Systems, Appendix E – Benchmarks, issued in March 2010 (FSCO 2010 Benchmarks). The R-squared of the regression analysis is high at 0.91 for frequency and at 0.90 for severity. It results in a claim cost trend of 5.3% per annum.

It would appear that the temporary claim surge possibly increased the 2009-12, 2010-06 and 2010-12 average claim cost by 2% above the pre-Reforms trend and quickly reversed itself after the adoption of the Reforms.

3.4.2.2 Third Party Liability – Property Damage

For TPL-PD, shown in Exhibit A, Segment II, Page 6a, we did not incorporate seasonality into the regression analysis. The R-squared of the regression analysis is fair at 0.74 for frequency, and high at 0.87 for severity. It results in a claim cost trend of 1.4% per annum.

3.4.2.3 Third Party Liability – Direct Compensation

For TPL-DC, shown in Exhibit A, Segment III, Page 6a, we incorporate seasonality into the regression analysis. We excluded the data points prior 2003-12, as they reflect an older claim environment. The R-squared of the regression analysis is low at 0.44 for frequency, which could be explained by the ultimate counts per vehicle remaining relatively flat over the historical period. For severity, the R-squared of the regression analysis is high at 0.82. It results in a claim cost trend of 0.8% per annum.

3.4.2.4 Accident Benefits – Disability Income

For AB-DI, shown in Exhibit A, Segment IV, Page 6a, we incorporate seasonality, the Reforms Impacts and the post-Reforms trend variable into the regression analysis for severity. We excluded the data points prior 2004-06 for both frequency and severity, as they reflect an older claim environment. The R-squared of the regression analysis for frequency is very low at 0.11, which could be explained by the ultimate counts per vehicle remaining relatively flat over the historical period apart for the temporary claim surge immediately preceding the implementation of the Reforms. The R-squared of the severity regression analysis is high at 0.95. It results in a claim cost trend of 7.6% per annum pre-Reforms, -0.2% post-Reforms, and an estimate of Reforms Impacts of -27.4%.

It would appear that the temporary claim surge possibly increased the 2009-12, 2010-06 and 2010-12 average claim cost by 8.5% above the pre-Reforms trend and quickly reversed itself after the adoption of the Reforms.

3.4.2.5 Accident Benefits – Death Benefits

For AB-DB, shown in Exhibit A, Segment V, Page 6a, we incorporate seasonality, the impact of Bill 198 and the post-Reforms trend variable into the regression analysis for frequency. We also incorporate the impact of Bill 198 and the post-Reforms trend variable into the regression analysis for severity. The R-squared of the frequency regression analysis is high at 0.91. The R-squared of the regression analysis for severity is low at 0.54. It results in a claim cost trend of -8.4% per annum pre-Reforms, and -2.7% post-Reforms.

3.4.2.6 Accident Benefits – Funeral

For AB-F, shown in Exhibit A, Segment VI, Page 6a, we incorporate seasonality into the regression analysis. We excluded the severity data points corresponding to 2004-12 and 2005-06, as they appear to be statistical outliers. The R-squared of the frequency regression analysis is high at 0.87, while the R-squared of the severity regression analysis is low at 0.55, reflecting the low credibility stemming from the low claim volume. It results in a claim cost trend of -6.2% per annum.

3.4.2.7 Accident Benefits – Medical & Rehabilitation

For AB-MR, shown in Exhibit A, Segment VII, Page 6a, we incorporate seasonality, the impact of Bill 198, the Reforms Impacts and the post-Reforms trend variable into the regression analysis for severity. We incorporate the impact of Bill 198 into the regression analysis for frequency. We excluded the data points corresponding to 2009-12, 2010-06 and 2010-12, which reflect a temporary claim surge immediately preceding the implementation of the Reforms. The R-squared of the regression analysis is very high at 0.89 for frequency and at 0.97 for severity.

It results in a claim cost trend of 17.0% per annum pre-Reforms, 0.7% post-Reforms, and an estimate of Reforms Impacts of -48.1%.

The temporary claim surge possibly increased the 2009-12, 2010-06 and 2010-12 average claim cost by 9.8% above the pre-Reforms trend and quickly reversed itself after the adoption of the Reforms.

3.4.2.8 Accident Benefits – Supplementary

There is extremely limited data for this sub-coverage. We judgementally assume 0% trend for AB-SUP.

3.4.2.9 Collision

For COL, shown in Exhibit A, Segment IX, Page 6a, we incorporate seasonality into the regression analysis for severity only. The R-squared of the regression analysis is high at 0.87 for frequency and at 0.93 for severity. It results in a claim cost trend of -1.4% per annum.

3.4.2.10 Comprehensive

For COM, shown in Exhibit A, Segment X, Page 6a, we incorporate seasonality into the regression analysis for severity only. We excluded the data points prior to 2005-06, as they reflect an older claim environment for frequency and severity. The R-squared of the frequency regression analysis is very low at 0.18 which could be explained by the ultimate counts per vehicle remaining relatively flat over the historical period. The R-squared of the severity regression analysis is fair at 0.71. It results in a claim cost trend of -3.2% per annum.

3.4.2.11 Specified Perils

Due to the low claim volume, we select to use the comprehensive trend for both frequency and severity.

3.5 Sensitivities

Our analysis indicates that there is a strong inter-dependency between the estimated claim cost trends (both pre- and post-Reforms) and the Reforms Impacts. Consequently estimates of claim cost trends, pre- and post-Reforms, as well as the estimates of Reforms Impacts should be extracted from the same modelling approach, with the same underlying data and assumptions, and should be read and interpreted conjointly for a given coverage and sub-coverage.

Section 3.5 examines the effect of different choices of data, modelling approach, and assumptions on the estimates of claim cost trends and Reforms Impacts. The focus of this section is on TPL-BI, AB-DI and AB-MR. It is important to recognize that the purpose of this section is not to discredit or favour a certain set of estimates, but rather to demonstrate the variability of claim cost trend and Reforms Impacts estimates.

3.5.1 Exclusion of a Major Writer from GISA Data

As mentioned in Section 3.1.5, significant change in the case reserving process of a major writer impacted the 2009-1 and subsequent diagonals and it led to its exclusion from GISA’s selection of claim development factors, FSCO’s 2014 Benchmarks15 of trends and Reforms Impacts, but not from the published data available to insurers. The published GISA data are not granular enough to allow similar exclusions. Thus reported claims, reported development factors, ultimate claim estimates, claim cost trends and Reforms Impacts, as outlined in Sections 3.3 and 0, are all derived using the all-industry data as published by GISA.

For the purpose of the 2014 Annual Report, FSCO agreed to confidentially share some portions of their review of the GISA data (excluding the experience of the major writer) regarding ultimate claim cost, frequency and severity. Detailed data are not available as part of the published GISA data to independently compare the exclusion, the selections of claim development factors or the assumptions underlying the trend fitting performed by FSCO with those presented in the 2014 Annual Report. This section compares our results based on the all-industry data published by GISA with FSCO’s results based on industry data net of the major writer exclusion.

3.5.1.1 Third Party Liability – Bodily injury

In the following graphs, the dark blue line represents the indicated measures underlying the models set out in Section 3.4.2. The green line represents FSCO’s adjusted measures. The line in turquoise represents FSCO’s fitted line. For sensitivity testing purposes, we apply the same set of independent variables as detailed in Section 3.4.2 to FSCO’s adjusted data (the green line) and estimate a claim cost trend of 5.1% per annum. Similarly, we refit FSCO’s fitted data (the turquoise line) with the same set of independent variables and estimate a claim cost trend of 4.4%, which ties back to FSCO’s 2014 Benchmarks16.

Indicated vs. Fitted Loss Cost
Indicated vs. Fitted Severity
Indicated vs. Fitted Frequency
3.5.1.2 Accident Benefits – Disability Income

Similar to the analysis performed for TPL-BI, for sensitivity testing purposes, we apply the same set of independent variables as detailed in Section 3.4.2 to FSCO’s adjusted data (the green line) and estimate a pre-Reforms claim cost trend of 6.7% per annum, a post-Reforms claim cost trend of -1.8%, and derive an estimate of Reforms Impacts of -22.3%. Similarly, we refit FSCO’s fitted data (the turquoise line) with the same set of independent variables and estimate a pre-Reforms claim cost trend of 4.9% per annum, a post-Reforms claim cost trend of -1.9%, and derive an estimate of Reforms Impacts of -19.7%. FSCO’s 2014 Benchmarks17 state that its pre-Reforms claim cost trend is estimated at 5.1% per annum, post-Reforms claim cost trend at 4.4%, and its estimate of Reforms Impact is at -33%.We have not attempted to reproduce FSCO’s methodology as outlined the Exhibit 5 of FSCO’s Technical Notes18. Further testing would be required to bridge the observed differences.

Indicated vs. Fitted Loss Cost
Indicated vs. Fitted Severity
Indicated vs. Fitted Frequency
3.5.1.3 Accident Benefits – Medical & Rehabilitation

Similar to AB-DI, we apply the same set of independent variables as detailed in Section 3.4.2 to FSCO’s adjusted data (the green line) and estimate a pre-Reforms claim cost trend of 15.7% per annum, a post-Reforms claim cost trend of 0.0% and derive an estimate of Reforms Impacts of -44.0%. Similarly, we refit FSCO’s fitted data (the turquoise line) with the same set of independent variables and estimate a pre-Reforms claim cost trend of 16.2% per annum, a post-Reforms claim cost trend of 0.1%, and derive an estimate of Reforms Impacts of -44.2%. FSCO’s 2014 Benchmarks19 state that its pre-Reforms claim cost trend is estimated between 6.7% and 8.3% per annum, post-Reforms claim cost trend 3.6% and 6.7%, and its estimate of Reforms Impacts is at -49%. We have not attempted to reproduce FSCO’s methodology as outlined the Exhibit 5 of FSCO’s Technical Notes20. Further testing would be required to bridge the observed differences.

Indicated vs. Fitted Loss Cost
Indicated vs. Fitted Severity
Indicated vs. Fitted Frequency

3.5.2 Length of Historical Periods

The number of accident half-years included in the historical periods underlying the selected trend model is a critical choice that has significant impact on the resulting claim cost trends and Reforms Impacts estimates. In selecting the length of the historical period, actuaries seek to achieve a fine balance between reactivity and over-fitting. This section compares claim cost trends and Reforms Impacts estimates using the same data as in Section 3.4.2, but using shorter historical periods.

3.5.2.1 Third Party Liability – Bodily injury

The first row of the following table displays estimates of the claim cost trends and Reforms Impacts based on the original historical periods included in Section 3.4.2. The second row shows an alternate scenario, which relies on a shorter historical period of 14 accident half-years, excluding the claim surge that preceded the implementation of the Reforms. As can be observed in the table below, the alternate scenario leads to indications that are significantly different than those resulting from the model presented in Section 3.4.2. While the alternate scenario indicates a counter-intuitive pre-Reforms trend estimate of -1.1%, its estimate of Reforms Impacts is at 10.2%. It hints toward the potential for claim transfer from Accident Benefits, which FSCO 2010 Benchmarks originally estimated in its Reforms Impacts of 26%.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 5.3% 5.3% 0%
Latest 14 accident half-years, excl. 2009-12, 2010-06 & 2010-12 -1.1% 8.0% 10.2%
3.5.2.2 Accident Benefits – Disability Income

Similar to TPL-BI, the first row of the following table displays estimates of the AB-DI trends and Reforms Impacts based on the original historical periods included in Section 3.4.2. The second row shows an alternate scenario, which relies on a shorter period of 14 accident half-years. As can be observed in the table below, the alternate scenario indicates trend and Reforms Impacts estimates that are within a reasonable range from those estimated using the original model presented in Section 3.4.2 with the claim cost trend at 6.5% per annum pre-Reforms, -0.2% post-Reforms, and an estimate of Reforms Impacts of -25.9%.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 7.6% -0.2% -27.4%
Latest 14 accident half-years 6.5% -0.2% -25.9%
3.5.2.3 Accident Benefits – Medical & Rehabilitation

Similar to TPL-BI, the first row of the following table displays estimates of the AB-MR trends and Reforms Impacts based on the original historical periods included in Section 3.4.2. The second row shows an alternate scenario, which relies on a shorter period of 14 accident half-years, excluding the claim surge that preceded the implementation of the Reforms. As can be observed in the table below, the alternate scenario indicates trend and Reforms Impacts estimates that are within a reasonable range from those estimated in the original model presented in Section 3.4.2 with the claim cost trend at 20.1% per annum pre-Reforms, 0.9% post-Reforms, and an estimate of Reforms Impacts of -50.6%.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 17.0% 0.7% -48.1%
Latest 14 accident half-years, excl. 2009-12, 2010-06 & 2010-12 20.1% 0.9% -50.6%

3.5.3 Iterative Model

The source of the ultimate frequencies, severities and claim costs included in the trend models is a critical choice that can significantly affect the resulting claim cost trends and Reforms Impacts estimates. The ultimate claim amounts and ultimate claim counts used in the trending models presented in Section 3.4.2 stem from the development method. The resulting trends and Reforms Impacts feed into the frequency-severity method, which in turn is an input to the Bornhuetter-Ferguson method. As described in Section 3.3.3, the results from Bornhuetter-Ferguson method are often selected as the “new” ultimates for the most recent accident half-years.

The models presented in Section 3.4.2 can be re-used iteratively by feeding the “new” ultimates to determine a subsequent set of claim cost trends and Reforms Impacts estimates.

Section 3.5.3 compares claim cost trends and Reforms Impacts estimates from the original iteration with the results from the 2nd iteration.

3.5.3.1 Third Party Liability – Bodily injury

The first row of the following table displays estimates of the TPL-BI claim cost trends and Reforms Impacts based on the original iteration relying on for the models presented in Section 3.4.2. The second row shows an alternate scenario using the results of the 2nd iteration. The alternate scenario indicates claim cost trends and Reforms Impacts estimates that are within a reasonable range from those estimated in the original model presented in Section 3.4.2 with the claim cost trend at 5.6% per annum.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 5.3% 5.3% 0%
2nd Iteration 5.6% 5.6% 0%
3.5.3.2 Accident Benefits – Disability Income

Similar to TPL-BI, the first row of the following table displays the estimates of the AB-DI claim cost trends and Reforms Impacts based on the original iteration relying on for the models presented in Section 3.4.2. The second row shows an alternate scenario using the results of the 2nd iteration. The alternate scenario indicates claim cost trends and Reforms Impacts estimates that are within a reasonable range from those estimated in the original model presented in Section 3.4.2 with the claim cost trend at 7.6% per annum pre-Reforms, and an estimate of Reforms Impacts of -25.1%. Based on a comparison between the two models presented in the table, the -2.9% post-Reforms trend estimate might appear to be optimistic.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 7.6% -0.2% -27.4%
2nd Iteration 7.6% -2.9% -25.1%
3.5.3.3 Accident Benefits – Medical & Rehabilitation

Similar to TPL-BI, the first row of the following table displays the estimates of the AB-MR claim cost trends and Reforms Impacts based on the original iteration relying on for the models presented in Section 3.4.2. The second row shows an alternate scenario using the results of the 2nd iteration. The alternate scenario indicates claim cost trends and Reforms Impacts estimates that are within a reasonable range from those estimated in the original model presented in Section 3.4.2 with the claim cost trend at 17.0% per annum pre-Reforms, 0.0% post-Reforms and an estimate of Reforms Impacts of -47.3%.

Historical Periods Included Trends Pre-2010 Reforms Trends Post-2010 Reforms Reforms Impacts on Claim Cost Levels
Original model as set in Section 3.4.2 17.0% 0.7% -48.1%
2nd Iteration 17.0% 0.0% -47.3%

4 EXHIBITS

The following is an index to all Segments in Exhibit A:

Segment I - Third Party Liability – Bodily injury (TPL-BI)
Segment II - Third Party Liability – Property Damage (TPL-PD)
Segment III - Third Party Liability – Direct Compensation (TPL-DC)
Segment IV - Accident Benefits – Disability Income (AB-DI)
Segment V - Accident Benefits – Death Benefits (AB-DB)
Segment VI - Accident Benefits – Funeral (AB-F)
Segment VII - Accident Benefits – Medical & Rehabilitation (AB-MR)
Segment VIII - Accident Benefits – Supplementary (AB-SUP)
Segment IX - Collision (COL)
Segment X - Comprehensive (COM)
Segment XI - Specified Perils (SP)
Segment XII - Accident Benefits – Uninsured Automobile (AB-UA)
Segment XIII - Underinsured Motorist (UM)

These exhibits are available in the PDF version of this report. If you require an accessible version of these tables, please contact FinanceCommunications.fin@ontario.ca.

© 2014 KPMG LLP, a Canadian limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative (“KPMG International”), a Swiss entity.  All rights reserved.

The KPMG name, logo and “cutting through complexity” are registered trademarks or trademarks of KPMG International.

[1] The Government of Ontario, Ministry of Finance, “2013 Ontario Budget Chapter IV: Tax, Pension and Financial Services”, last modified May 2, 2013. Accessed on September 20, 2014, www.fin.gov.on.ca/en/budget/ontariobudgets/2013/ch4.html#ch4c.

[2] The ASB was established by the Canadian Institute of Actuaries (CIA) as an independent body; the mission of the ASB is to develop, establish, and maintain Standards of Practice governing actuarial practice in Canada. Throughout this Independent Actuarial Analysis, we use the abbreviation “SOP” to refer to Canadian actuarial Standards of Practice promulgated by the ASB.

[3] I.e., the Independent Actuarial Analysis is based on data from the PPA GISA Exhibits.

[4] Expenses include claim adjustment expenses (CAE) only.

[5] Expenses include claim adjustment expenses (CAE) only.

[6] The dataset used is net of the same data exclusions stemming from GISA’s validation procedures (i.e., the GISA data that we used in this report is consistent with the data that GISA publishes externally.)

[7] Weighted based on 2009 ultimate claim costs.

[8] Weighted based on 2013 ultimate claim costs.

[9] Weighted based on 2009 ultimate claim costs that have been trended by one year.

[10] GISA, “2013 Automobile Exhibit Introduction and Actual Loss Ratio Exhibit – Ontario”, bulletin 2014-61, July 11, 2014.

[11] GISA, “2013 Automobile Exhibit Introduction and Actual Loss Ratio Exhibit – Ontario”, bulletin 2014-61, July 11, 2014.

[12] IBC, “Facts of the General Insurance Industry in Canada - 2009”, Facts 34th Edition, 2009. ISSN 1197 3404

[13] IBC, “Facts of the Property & Casualty Insurance Industry in Canada – 2013”, Facts 35th Edition, 2013. ISSN 1197 3404

[14] CAS Study Note, 1993

[15] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 2 – Benchmark Assumptions for Private Passenger Automobile Insurance Filings for Reference Purposes, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf

[16] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 2 – Benchmark Assumptions for Private Passenger Automobile Insurance Filings for Reference Purposes, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf

[17] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 2 – Benchmark Assumptions for Private Passenger Automobile Insurance Filings for Reference Purposes, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf

[18] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 5 – Analysis of Reform Cost Adjustment Factors and Loss Trends, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf

[19] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 2 – Benchmark Assumptions for Private Passenger Automobile Insurance Filings for Reference Purposes, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf

[20] FSCO, Technical Notes for Automobile Insurance Rate and Risk Classification Filings, Exhibit 5 – Analysis of Reform Cost Adjustment Factors and Loss Trends, October 2014. Accessed November 6, 2014. http://www.fsco.gov.on.ca/en/auto/filing-guidelines/Documents/Technical-Notes.pdf