Climate change analysis - Aviva on stress-testing for history's biggest 'What if?'

Ben Carr, analytics and capital modelling director at Aviva, reflects on the limitations of stress testing an investment portfolio for climate change, and the benefits of doing so nonetheless. Interview by Vincent Huck

How do you stress test a portfolio against climate risk? And how is that different from the more traditional financial stress tests?

In a traditional stress test framework, you would think about potential shocks to your balance sheet and their impact on the value of your assets and liabilities. There might be a fall in credit spreads or increases in interest rates, or something along those lines. There's a pretty direct link between those and the valuation models and approaches that you're using to value your balance sheet.

So what's different about climate stress testing? The first thing is that when you're talking about climate stress testing, you're really probably talking more about climate scenario analysis, rather than a simple stress test.

First, you develop a scenario - it could be anything between a 1.5 degree Paris-aligned scenario, to a hothouse scenario. Then, you look at, over an extended period of time, how do you expect the evolution of certain drivers in that scenario and how could they impact your business and your balance sheet?

One of the big challenges is that often the things that get defined in that scenario, like temperatures rising, sea levels rising, or increases in carbon prices, don't naturally link back to your valuation models.

So the first thing you need to do is think about how those types of drivers are going to impact the value of your assets and liabilities going forward, and then you can build models to consider the impact on their value, in those scenarios.

That is new and challenging, in particular there is lots of uncertainty around evolution of those drivers. So you need to make some assumptions about the evolution of these drivers, whilst at the same time being aware of the uncertainties associated with those, and then think about how you can link those back to the valuation of your assets and liabilities.

What other challenges and limitations are there?

The second challenge is that it's not really a straight profit-and-loss(P&L) shock. Unlike a stress on your credit spreads, which will have direct impact on your balance sheet, it is probably more helpful to think about these scenario analyses as providing a view on the impact on operating capital generation over an extended period. It's similar to economic growth scenario analysis, where you'll hear people say 'it could reduce GDP by 2%, 3% or 4%'. They don't mean by that, that GDP is going to fall by that amount. Rather, they mean that relative to a counterfactual scenario where these things don't happen or a central scenario, then you're going to have 2% or 3% less GDP growth than you would have done otherwise.

In the same way, climate scenario analysis is very much about what the impact is on operating capital generation, not what's the P&L shock to your business.

A third challenge related to this is that because climate risk happens over an extended period, it is actually not really about your current balance sheet. It's more about thinking of the impact over the scenario time horizon, taking into account the management actions that you have at your disposal, to adjust your business to the changing environment.

And this is probably the most interesting part of the analysis, not the pound-sterling impact, but what the implications of these different climate scenarios relatively speaking, [and] which ones are more, or less, detrimental to the business? And what actions are available to reduce my exposure in each scenario?

What has been the output of Aviva's own scenario analysis?

When we do our climate scenario analysis, what it tells us is that, the hothouse scenario is the scenario which is most damaging for our business [in the] long term. In particular, it negatively impacts long-term investment returns equities, corporate bonds, real estate and sovereigns. And, actually, there's relatively little as an insurer that you can do to hedge or avoid that risk, it's a systemic risk. And therefore, we are exposed to it if that risk materialises, because we always have to invest in assets to back our liabilities.

On the converse side, we see less risk in the 1.5 degree scenario. And, more importantly, we actually see potential opportunities in that scenario, if you're correctly positioned.

You could benefit from the transition if you are investing in low carbon infrastructure, if you're providing climate-conscious products to your customers, so they can support the transition. So, the analysis supports our view that we need to be advocating and strongly supporting and encouraging governments to take the actions they need to take, not to end up in a hothouse scenario. But we also need to be making sure that we're taking actions ourselves to be on the right side of the transition and take advantage of the opportunities that will arise.

It's not designed to be a precise modelling framework that enables you to make very granular investment decisions, for example, to actually invest in one asset versus another or one company versus another.

To really use it in that very granular way, you'd have to have a very clear view about what's currently priced into asset valuations. You've got to decide how much transition risk is currently priced into asset valuations? How much physical risk is currently priced into asset valuations? Those are very difficult questions to answer.

For the moment when we do our scenario analysis, we use a counter-factual assumption – i.e. that nothing is priced from a transition- or physical risk perspective, and that helps us to understand the relative size of the risks in different scenarios which is the purpose of the analysis.

Why does one need a scenario to tell you that a hothouse scenario is more detrimental than a 1.5 degree scenario?

There is definitely a value in doing it, in the sense that you set out narratives around these scenarios. It helps people to understand how those scenarios could impact them, and the sorts of risks and opportunities there are, and therefore helps to start the process of thinking about how it could impact their business, and what actions they might take.

You don't need a quantitative stress test to start to have that conversation, , it's not so much about numbers, it's more around being systematic in the analysis.

If you really want to identify specific areas, specific products, specific actions that you might want to take, then having a more systematic approach and really thinking through how a driver is going to impact a business, ranking the risks associated with those drivers for different asset classes and different counterparties, can add a lot of value.

You need to understand the limitations though. And you shouldn't over rely on the outputs. It is also really important to communicate those limitations clearly when presenting the results of any analysis.

But putting a number on something does have the advantage of giving people an opportunity to challenge and disagree with it and that is often the best way to ensure you improve and refine the analysis over time.

Ulitmately, more quantitative, systematic approaches, will be required if these risks are really going to be embeded into financial services, strategic thinking and decision making.

When insurers do their climate stress tests, it can never be on the entire portfolio. It's usually on corporates, equity and real estate. Isn't that a limitation? How accurate is a 'final number', when it's only covering a small part of the portfolio?

We do scenario analysis on our shareholder funds at moment, and we do incorporate equities, corporate bonds, sovereigns, and real estate, the main assets we are exposed to. And we do include general insurance liabilities and our life insurance liabilities.

We are trying to capture the whole book, and I agree to make it meaningful, you need to do all of that. It is work in progress, I wouldn't say it's perfect. We don't capture all risks associated with all asset classes at the moment. There are still gaps, and it's important to look at the results understanding what those gaps are. There's also a question at the moment with some of the less liquid assets and how to bring them into this analysis, because you often haven't necessarily got the data that you need to incorporate them properly.

How regularly should one stress-test a portfolio against climate risk?

We currently run the scenario analysis quarterly and disclose the results annually. Given the long-term nature of the risk I don't think that more frequently than quarterly is needed. If you are setting targets, clearly you want to perform the analysis at a frequency that makes sense to check that you are still on track to meet the target.

So once you've done the analysis, how does that then fit into management actions?

We recently announced our 2040 net-zero strategy, and that was certainly informed by the output of the analysis that we have run for the last three years.

Also as part of the Prudential Regulation Authority supervisory statement 3/19, we're having to embed climate-related risks and opportunities into our risk management framework. So, we're certainly using the metrics that we've developed for the TCFD, to build that into our risk appetite framework.

These have also informed our commitments, like the ones we've taken through our Net-Zero Asset Owner Alliance membership, to reduce our carbon intensity by 25% over the next five years. We did a lot of analysis to understand what the implications were, how much decarbonisation we could expect from counterparts themselves, what additional levers we need to pull to meet the target - other than through engagement, companies reducing their overall emissions, and the opportunity to reinvest assets when they mature.

All of these metrics are really starting to be embedded in the way that we're running the business. And we've recently included these metrics in our remuneration for top management.

At the level of your portfolio, what are the 'management actions'?

The first management action is engagement with investee companies, that is the first lever. The next step is when we reinvest maturing assets we've got a choice to invest in low-carbon infrastructure, real estate, green assets, green bonds, sustainable bonds, and we've set targets around that, too.

Then beyond that, you get to the 'sector-leader' type approach, which is looking at who the people are within a particular sector who are making strides, and switching your investments to those. Changing your strategic asset allocation doesn't really solve the problem, because from a 'real economy' perspective, if we switch [out of high emitting sectors], whilst it reduces our carbon footprint, it doesn't necessarily reduce the global carbon footprint, and you lose the ability to influence.