Markus Gesmann, Raphael Rayees and Emily Clapham point the way towards a consistent framework for defining and measuring claims inflation
In the current time of globalisation, faced with questions over commodity supply, security and price volatility and potential fluctuations in currency rates, claims inflation constitutes a serious threat to both the profitability and the security of insurers worldwide. Despite this, there are a plethora of views on the extent, and even the existence, of claims inflation; this article aims to raise the profile of the issue and contribute to the debate.
First, an interesting fact: The Claims Inflation Working Party published its research report on "Claims Inflation - Uses and Abuses" [1] at GIRO in 2005. Eight years have passed since then, five of them in a downturn, and yet this document is still the first result that comes up today when googling 'Claims inflation in insurance'.
We put the question 'what is claims inflation' to several of our colleagues and practitioners of the Lloyd's Market. The majority of participants responded "somewhere between 3% to 5%", yet none had a definitive explanation of how to define or measure it. Despite this lack of consistency and industry engagement, inflation is regarded as a risk and a challenge for insurers, and they are certainly not alone.
The 2011 Lloyd's Risk Index [2], a survey of global corporate risk priorities and attitudes, listed inflation, together with changes in prices of material inputs, changes in legislation and currency fluctuations in the top 10 concerns of business leaders. Insurers also understand that claims inflation is an important metric, particularly for pricing long tail lines of business, as well as an influential factor in reserving, planning and capital setting.
With the obvious exception of motor insurance, high levels of claims inflation have not been a big issue for other lines of business in recent years. Unfortunately, in the current litigious and economic environment, inflation and claims inflation are only likely to head one way from today's levels - north. Any neglect of claims inflation by insurers, who face low interest rates and a softening market due to surplus capital, can result in a nasty surprise.
In a recent analysis, Milliman showed that an increase in claims inflation of 1% could increase liabilities disproportionally. As a rule of thumb, the authors approximated the impact of claims inflation on liabilities by multiplying the change in inflation with the number of payment years [3]. Hence, a change of claims inflation by 2% could have an impact of 16% on a book which takes 8 years to settle. We believe it is difficult to hedge this risk in today's environment, exemplified by the current dynamics brought about by the PPO claims awards.
The Milliman analysis shows how necessary it is to allow for claims inflation; but for many this is easier said than done. Measuring any type of inflation is complex and comes with a unique set of challenges; the recent discussions in the media around the differences between the Retail Price Index (RPI) and the Consumer Price Index (CPI) highlight this [4]. While subtle differences in coverage and calculation may appear small in the incremental data, they can have a material impact over a longer time period.
The difficulties of measuring inflation were seen again following the creation of the Eurozone currency union, which forced countries to agree upon a standard methodology to measure inflation centrally. This resulted in the Harmonised CPI (HCPI), which, despite its shortcomings, has set a standard to monitor inflation on a like for like basis.
This concept of measuring inflation on a like for like basis sounds quite natural; indeed the same methodology is used for adjusting inflation indices following consumer price changes and for constructing stock indices. However, when the Lloyd's market sought to establish a consistent framework to monitor rate movements of renewal business, an initial survey revealed that people, even within the same organisation, had different understanding of what relative price movements on the same risk could mean. Just like measuring inflation, the crucial aspect here was to agree a standard approach to ensure a like for like comparison year on year across syndicates and lines of business.
The decision taken by Lloyd's was to measure rate changes on a risk adjusted basis, which means underwriters have to estimate how much they could have charged a year ago for this year's policy on this year's terms & conditions and expected loss costs. The relative difference between these two prices is termed the risk adjusted rate change (RARC). Therefore RARC's are net of claims inflation and focus on the year on year impact in expected loss ratio.
Where to start?
Following our discussions with market practitioners, we feel that the industry needs a similar, consistent framework for defining claims inflation. As with price inflation, claims inflation comes with its own set of complexities; it is not measurable through direct observation, it can only be estimated using statistical techniques applied to historical data. In the face of globalisation, societal changes and technological changes, this historical data is no longer necessarily relevant to future trends. That being said, the Lloyd's rate monitoring approach, along with the methodology behind established indexes such as the HCPI, may still provide an insightful guide to achieving this framework.
There are a few available sources on the subject to consult. The paper of the Claims Inflation Working Party from 2005 still provides valuable insight; it outlines the key drivers of claims inflation and provides an overview of methods to estimate claims inflation. Towers Watson publishes a US claims cost index, which is based on the seminal work by Norton Masterson of the 1960s [6, 7]. The index is based on a selection of inflation factors for the US market and faces the same challenges as any other index; coverage and calculation. However, it also provides a blueprint to create an inflation index based on macro-economic data, which is tailored to specific portfolios. The European Statistical Official provides a wealth of data and the inflation dashboard [8] allows users to extract the inflation factors most relevant to a business.
Of course, none of these macro-economic metrics measure either change in frequency of claims or social inflation. The RAND Corporation published a detailed review of the dramatic increase in claims frequency and severity of medical malpractice claims in the US in the early 1970s [9].Its model suggested that the single most powerful predictor of claims frequency and severity is urbanisation. Note also that claims inflation can vary by the size of claims and its impact can be amplified in excess of loss layers.
Consider the expected volatility of expected claims inflation, noting that inflation is more likely to go up than down, and use stress testing to establish what effects a spike or a shift in inflation levels would have on both profitability and solvency levels.
Historical loss triangles contain implicit information on claims inflation. Consulting the papers of Barnett & Zehnwirth [10] and Christofides [11] offers ideas on how changes in the payment year trends, reflecting claims inflation, can be modelled. With modern statistical software, it is now relatively straightforward to implement these models; as demonstrated in a blog post by Gesmann [12] using R earlier this year.
Indeed, a better understanding about how to extract historical claims inflation from historical data can provide a good starting point for measuring and mitigating claims inflation, and form a basis for future strategies to navigate it.
Engaging with colleagues, particularly those in the actuarial, underwriting and claims functions can help to establish a consistent framework which will work across different lines of business. So, too, will considering which data and at what level of granularity should be captured, how assumptions on claims inflation could be back-tested, or indeed in the future be compared against actual experience.
This brings us back to the first principles: Define the use cases for claims inflation, and acknowledge that assumptions and time horizons may differ between pricing, reserving, planning and capital modelling.
By doing so, we may be in a much better position to answer that elusive question: "What is claims inflation?" Only then we can consider how to measure, monitor, manage and potentially mitigate inflationary effects.
The authors would like to thank Tom Bolt, Henry Johnson, James Orr and Sush Amar for their valuable feedback and comments. All views and opinions outlined are the authors' own.
References:
[1] Simon Sheaf, Simon Brickman and Will Forster. Claims Inflation - Uses and Abuses. GIRO Conference, 2005.
[2] Lloyd's Risk Index, Lloyd's, 2011, http://www.lloyds.com/news-and-insight/risk-insight/reports
[3] Derek Jones and Thomas Ryan, Milliman. Estimating the impact of claim inflation on self-insured liabilities, July 2010, http://insight.milliman.com/article.php?cntid=7305
[4] Jill Leyland, RPI versus CPI - The Definitive Account, Significance, Royal Statistical Society, http://www.significancemagazine.org/details/webexclusive/1314363/RPI-versus-CPI--The-Definitive-Account.html
[5] Lloyd's. Performance Management Data Return. 2009 - 2013, http://lloyds.com/pmdr
[6] Jeremy P. Pecora and Emily M. Thompson, Towers Watson Claims Cost Index, Emphasis, December 2012, http://www.towerswatson.com/en/Insights/Newsletters/Global/emphasis/2012/Towers-Watson-Claim-Cost-Index
[7] Norton Masterson. Economic Factors in Liability and Property Insurance Claims Costs. Casualty Actuarial Society, 1981. http://casact.net/pubs/proceed/proceed68/68061.pdf
[8] Eurostats. The Harmonised Index of Consumer Prices (HICP), European Commission, http://epp.eurostat.ec.europa.eu/inflation_dashboard/
[9] PM Danzón. The Frequency and Severity of Medical Malpractice Claims. RAND Corporation. 1983. www.rand.org/pubs/reports/2006/R2870.pdf
[10] Glen Barnett and Ben Zehnwirth. Best estimates for reserves. Proceedings of the CAS, LXXXVII(167), November 2000. http://www.casact.org/pubs/proceed/proceed00/00245.pdf
[11] Stavros Christofides. Regression models based on log-incremental payments. Claims Reserving Manual. Volume 2 D5. September 1997, http://www.actuaries.org.uk/sites/all/files/documents/pdf/crm2-D5.pdf
