Some thoughts from the regulator

Ian Laughlin, Executive Member - Finity Niche Insurer Conference, Sydney

Good morning – it’s good to be here again.

This is the Laughlin residence:

This is a neighbour’s place – 100 yards from where it started:

The year is 1936.

The location is the Waterford area about 30 kilometres south of Brisbane.

Here is a view of the same area from Google maps today:

You can see that the property damage today from a similar event would be far greater than in 1936.

These photos emerged recently when I was looking through some old albums of a recently deceased aunt, and they prompted me to find out a little more. As a first step, I went looking on the internet for cyclones in south east Queensland. What I found was quite interesting.

First, it seems the cause of the damage to the Laughlin residence was a tornado, which caused destruction over a wide area through to Wynnum, east of Brisbane!

Tornados are sometimes associated with cyclones, so I found out what I could about cyclones in the area. This shows cyclones in the 1935-36 season, and you can see there were a couple that passed near south-east Queensland, but they seemed to have been at different times to the tornado.

That then led me to wonder how frequent cyclones were in south-east Queensland. So I checked on the history of cyclones over the last 100 years or so.

Many of you will have seen this sort of map before, but it is interesting to note that cyclones in the south-east of Queensland do occur relatively frequently. Even northern NSW has been quite badly impacted by cyclones in the past.

This piece of simple analysis prompted me to think about insurers considering their catastrophe risk exposures and their reinsurance arrangements.

So more about that in a few minutes.

Here is what I will cover today:

LAGIC Latest

As you well know APRA’s revised capital requirements come into force from 1 January 2013.

The LAGIC project has been a very significant undertaking – for both industry and for APRA – extending over about three years. There have been four rounds of consultation, the last of which has only just been completed, as well as many other interactions with industry. This has proven to be well worthwhile, and we are very pleased with the support we were given by industry during the development of LAGIC. It has been generally constructive and undoubtedly helped produce a better outcome.

Last week, we issued the final versions of the new and revised prudential standards which implement LAGIC. Later this month all of the reporting standards will be released in final form.

So we and the industry are now very much focused on implementation, which in itself is a major undertaking. We believe industry is generally well prepared for the new regime. Most senior management and most boards are well engaged in the LAGIC rollout.

LAGIC introduces significant changes to reinsurance requirements, including the whole-of-portfolio approach and the horizontal requirement within the prescribed capital for insurance concentration risk. The latter is intended to address the risk of multiple large events within a year, and for many insurers is likely to involve different reinsurance arrangements to those used in the past. You will recall that we delayed the implementation of the horizontal requirement until 1 January 2014, to give insurers and reinsurers time to develop and put in place appropriate reinsurance arrangements. We are now concerned however that insurers have taken their eye off this particular ball, and are not making the most of the additional time to prepare properly for the introduction of the horizontal requirement. For example, the ICAAP must address the horizontal requirement from the commencement of LAGIC. This means we expect insurers to calculate their horizontal requirement at 1 January 2013, even though the horizontal requirement does not come into force until much later, so that they can ensure they are well prepared for its introduction.

This leads us nicely back to natural catastrophe reinsurance.

Natural Catastrophe Reinsurance

The industry has been a little more settled since we last spoke, in the sense that the weather and earthquake gods have been a bit kinder.

In another sense, there are fresh challenges as we seek to learn the lessons from the events of 2011.

At this conference last year, I set out some of those lessons, and today I will build on those comments in two particular areas: catastrophe modelling and the Reinsurance Arrangements Statement (ReAS).

Catastrophe Models

As you know, under LAGIC APRA sets a capital standard based on 1 in 200 probability of sufficiency, and the minimum catastrophe reinsurance arrangements are derived from this. APRA’s minimum requirement - and it is just a minimum - should be only one factor in considering appropriate levels of catastrophe reinsurance. The Board should consider whether 1 in 300, or 1 in 500, for example, would better reflect its risk appetite.

Obviously it is not a simple matter to assess the level of cover necessary to meet APRA’s standard and/or the board’s risk appetite, and models are used extensively for this purpose. All models have limitations. The output is dependent on the data provided, the assumptions in the model, the reliance on known history, which perils are modelled, the sophistication of the modelling engine etc. The relevance of the result of course is then subject to the vagaries of the real world (e.g. an unknown earthquake fault!). The major models undergo regular revision in light of experience, but they will continue to have deficiencies.

The net result is that at best a model gives an estimate and its results are subject to a significant margin of error.

This is very well illustrated on this slide reproduced from a Guy Carpenter paper, Managing Catastrophe Model Uncertainty (2011), which provides a very good explanation of the issue.

Note just how wide the band of uncertainty is in this illustrative example.

However, our reviews in the wake of last year's events suggest that management and boards may not understand the degree of uncertainty in the model output, and as a result they may be relying too heavily on these models in setting reinsurance cover.

More specifically we found in some cases:

• blind acceptance of model outputs - either internally or through reinsurance brokers - in the purchase of their catastrophe cover;
• use of multiple models to set their cover, but without clear rationale for the weightings chosen;
• a low level of understanding of the inputs and assumptions used and their effects, and what happens in the model "black box";
• a lack of questioning of model results; and
• a lack of board oversight and challenge in the use of the model.

With some insurers, there was recognition of model uncertainty, and a margin added to the model output to compensate. The margins varied greatly between these insurers. In some cases they were reasonably arbitrary; in others there was a considered approach to dealing with un-modelled risks, such as demand surge, etc.

It is worth noting here that one side-effect of the spate of natural disasters in 2011 was a number of revisions to models, demonstrating that they are "work-in-progress". For example, updates have been made for northern hemisphere windstorm, NZ and Japan earthquakes and locally, for storm, earthquake and bushfire perils.

We are urging management and boards to recognize the limitations of models, and to challenge and debate the model outputs, and in the process to recognize the inherent errors in these outputs. This is a matter of good governance. The sorts of things that should be considered include the following:

• the views of the model supplier as to any deficiencies and inherent inaccuracy in their model output, how up to date the model is etc.
• where different vendor models produce markedly different results, the reasons why;
• an assessment of the perils not adequately covered or not covered at all by the model;
• management's view of the quality of data provided as input to the model, including its granularity (e.g. zone, post code or street address);
• the appropriateness of assumptions and settings (e.g. construction types, subsoil composition) used in the model, and any weaknesses that they generate in the output;
• how well the model copes with back-testing against previous actual events;
• how any deficiencies identified are being addressed; and
• an appropriate allowance for model uncertainties.

A further challenge for management is to determine what extra information and analysis will help the board in determining final reinsurance cover. What can be done to bring the issue to life, so it not seen as simply an exercise for a computer program? What history (such as meteorological records) is available that might help? Has appropriate allowance been made for urban development and changes to topography or water courses? Have there been changes in weather patterns? What advice is available from reinsurers and/or brokers? What is the basis of their advice?

And then, what if? What if a cyclone hit Brisbane for example? What if an earthquake hit Perth? Scenario tests can be very helpful in addressing these sorts of questions. This includes assessing the impact of past events on the current environment to consider how the insurer would be affected. For example, a repeat of the 2011 events could be used to test revised reinsurance arrangements and ICAAP.

Or using my simple example, what impact would a repeat of the 1936 tornado have on the now highly-developed suburbs of Brisbane? One can imagine an event such as this generating interest from board members, and management using it as an opportunity to inform the board about the various types of windstorms, their frequency, and the risks they pose in south-east Queensland. The board might want to understand how the insurer’s reinsurance arrangements responded, and how in future they would respond to multiple such events (particularly in light of APRA’s horizontal requirement). The board also might want to consider this in the context of their risk appetite, and satisfy themselves that their ICAAP dealt adequately with the possibility of such events.

There could be useful discussions with the board about the insurer’s catastrophe model, how well calibrated it was for such an event, and whether the model settings and data input in hindsight had been appropriate. This might lead to questions about the relationship with the model supplier, and how the supplier reviews such an event and whether it might result in refinements to the model. Similarly, the board would very likely be interested in the reinsurer’s response, and whether the experience might lead to a reassessment of risk in south-east Queensland.

My point here is that it is important to engage the board in these sorts of discussions to deepen its understanding, and to help it better do its job.

APRA too needs to better develop its understanding of models and their weaknesses (including why different models produce different results for the same perils), and of current practices. We are reviewing processes in a number of insurers and conducting discussions with reinsurers, model vendors and reinsurance brokers to better understand how the models are used, the governance arrangements in place, what improvements could be made etc.. We are also interested in a more detailed understanding of whether and how modelling error is addressed.

For more information on the use of models, see the draft Prudential Practice Guide GPG116 – Insurance Concentration Risk, released in September.

Reinsurance Arrangements Statement (ReAS)

Our review of reinsurance arrangements statements post last year's natural disasters indicated that while some are of quite good quality, many ReASs need significant improvement. Areas of concern include the omission of critical details such as the number of reinstatement covers, and unclear explanations of reinsurance programs. In some cases, there were inconsistencies between the narrative descriptions and diagrams.

We also found cases where there was a lack of information, explanation and/or worked examples showing how the reinsurance would respond, particularly to multiple events in the one year. It was also not always clear how the various reinsurance covers complemented each other to provide comprehensive protection.

That is not a good report card, and clearly improvement is needed for many insurers.

APRA is currently in the process of speaking to entities where it found any of these issues, to ensure they are adequately addressed.

As part of this process, and a more intensive review in 2013, we will be looking to ensure that:

• there are no gaps in cover, or if there are gaps they are known and understood and properly managed - for example with capital;
• reinsurance programs adequately respond to differing levels of claims, and to multiple events - the latter will include consideration of the LAGIC horizontal requirement , sideways cover, and retention protection;
• the ReAS is a clear and accurate implementation of the reinsurance management strategy (ReMS);
• gross exposures, probable maximum loss calculations and hazards covered are well considered and justified;
• there are appropriate governance arrangements in place; and
• there is consistency between the ICAAP, the ReMS and ReAS;

APRA will provide feedback to individual insurers and the industry on the outcomes of our reviews.

Note that whilst our focus has been on natural catastrophe events, our reviews will include non-property and niche business (medical indemnity, liability classes, LMI etc.)

An unsatisfactory outcome for an insurer will result in close supervisory attention, including the possibility of a Pillar 2 adjustment to capital requirements.

Note that APRA is not an adviser. The quality and effectiveness of the reinsurance arrangements and the ReAS are the responsibility of management and board. We suggest that good use is made of actuaries and auditors (both internal and external) to provide critical independent review of the ReAS and its use.

Indeed, we expect that the Appointed Actuary will review the ReAS, ICAAP, strategic plan and business plan for consistency and gaps, as part of his or her assessment of the suitability and adequacy of the reinsurance arrangements. We would also expect any issues to be addressed in the FCR.

Internally, APRA will be conducting training for our front-line staff to ensure the reviews are effective and efficient. We will also start work on a prudential practice guide on the ReAS. We will be seeking to standardise the collection of major ReAS components (treaty type, classes, limit, excess, premium, reinsurer) to complement schematics and narrative.

Reinsurance counterparty risk

In 2010/11, APRA undertook a data collection to assess whether the failure or significant downgrade of a major reinsurer was a material risk for the industry. Pleasingly, the information provided showed that the industry is well diversified in terms of reinsurance counterparties from a geographical standpoint and across APRA and non- APRA-authorised reinsurers. In excess of 90% of the reinsurers utilised are rated APRA grade 3 (S&P A-rated) or higher. As part of the LAGIC reporting package APRA consulted on proposals to include reinsurance counterparty information in the regular data collection. We plan to finalise our approach to ongoing counterparty risk data collection in 2013.

Placement risk

The catastrophe events in Australia and New Zealand in 2011 led to some reinsurers reducing or restructuring the cover they were willing to provide in the region, or requiring significant price increases for some types of cover. Our concerns about the placement risk this engendered have eased considerably as the market has responded quite well. Again, we urge insurers to consider their horizontal requirement needs well as early as possible to ensure that they encounter no reinsurance placement problems leading up to the start of 2014.

ICAAP – a reminder

The Prudential Capital Requirement (PCR on the slide) is the minimum capital requirement under LAGIC (and is made up of two components, Pillar 1 and Pillar 2).

The PCR is a "hard floor" which must not be breached.

If capital does fall below PCR, the insurer needs to correct the position promptly to avoid APRA taking serious measures and potentially taking control out of the hands of the board and management.

Thus the insurer is expected to target a level of capital well in excess of PCR consistent with its risk appetite, and to set a process for managing deviations from that target under what we call the Internal Capital Adequacy Assessment Process, or ICAAP.

You will see on the diagram that APRA’s level of supervision will become more intense as the capital position approaches PCR.

As its name suggests, ICAAP is a process, but of course we are interested in both the process and its effective management, and ultimately the outcomes of the process.

ICAAP was explained in our various consultation papers, and ultimately in the final prudential standards made in May 2012. To provide further help, APRA has just issued a Prudential Practice Guide (PPG). I urge you all to read that guidance.

I don’t want to repeat what has been published in the prudential standards and practice guide, but as a reminder, at a minimum the ICAAP must include the items on this slide:

APRA expects that the ICAAP will be developed by the insurer’s senior management with input from relevant business areas and experts. However, the ICAAP is ultimately the responsibility of the Board. The Board should be actively engaged in the development of the insurer’s ICAAP and its implementation, and must approve the ICAAP.

There are two documents that need to be produced - the ICAAP summary statement and the ICAAP report.

The ICAAP summary statement is a high level document describing and summarising the capital assessment and management processes of the insurer. This document is expected to be developed and available to APRA from 1 January 2013.

The ICAAP report is an annual report that will summarise the operation of the ICAAP over the last year and give an overview of expected capital management actions and issues for the coming period. It should include details of actual capital levels and capital management actions during the previous period, and projected capital levels and planned capital management actions. It should also summarise any changes to capital targets or to other aspects of the ICAAP.

We expect the ICAAP, the ICAAP Summary and the ICAAP Report to be powerful and useful tools for the board and that they will be integral to the management of the business. APRA too, will pay close attention to them.

The main point that I want to leave you with is that the board and management have responsibility for developing an ICAAP, and for setting target capital and managing the capital position around that target in accordance with their ICAAP. This includes the use of reinsurance.

It is critically important that this is done well.

Stress Testing

Let me make a few comments on the role of stress testing in developing ICAAP and managing capital levels.

APRA expects stress testing to be used both in setting the insurer’s risk appetite and in the insurer’s ongoing capital management. It therefore should be an integral part of the insurer's ICAAP.

Rigorous, forward-looking stress testing supports capital and reinsurance management by assessing the impact of severe events, including a series of compounding events or changes in market conditions that could adversely impact the insurer.

Stress testing is done in various ways, but one approach involves a plausible but very adverse scenario and an assessment of the impact of that scenario.

I mentioned earlier that insurers should make greater use of scenario testing as part of the development of their ReMS.

Stress testing and scenario testing are powerful tools. They help APRA and management and board to understand the capacity of the organization to cope with the specified stress or scenario, and they also provide insights into vulnerabilities of the insurer. From this, lessons can be learnt that might lead to changes in strategy or operations or to capital management.

There is also a very useful concept called reverse stress testing. This is a little different, in that it seeks to determine a scenario that for example, would cause the company to breach its regulatory capital requirements. This can give fresh insights into vulnerabilities, in a way that will be well-understood by the board.

We want to encourage management and boards to consider stress testing as one of the more important weapons in their ICAAP armoury, and to use it accordingly.

Closing Remarks

Thanks for your time.

I hope this has given you a feel for likely areas of attention as we move into the implementation mode for LAGIC.