Scenario analysis is essential in uncertain times, but it must be well-grounded, not just drawing an arbitrary line on a graph.
- You know times are uncertain when institutions start publishing scenarios alongside their normal forecasts. Two principles underlie good scenarios. First, the narrative underlying the scenario must hang together – no assuming people do things that are either individually irrational or infeasible. Second, the scenario must allow people to respond to the initial shock – avoid just drawing a price line on a graph and assuming people cannot adjust their behaviour.
- There is a tension, though, between the need to be realistic and a desire to highlight vulnerabilities or contemplate worst-case scenarios. There is a legitimate role for worst-case thinking, but it must be handled with care. Recall that even though our forecasts last year were less alarmist than many in the wake of Liberation Day, actual outcomes were even better than we had expected.
- The real value of scenarios is teasing out downstream effects and delayed responses. Sometimes the responses to temporary shocks can have lasting effects. From EV purchases to defence budgets, these could be especially important coming out of the current conflict. All the more reason to avoid just drawing a line and assuming that is the outcome.
In times as uncertain as these, you want to be able to articulate more than one possible future state of the world. That is where scenario analysis comes in. Indeed, when institutions start publishing scenarios, you can take it as a sign that things are unusually uncertain. The RBA used scenarios extensively during the pandemic. And while the IMF did not give scenarios any prominence in its previous full World Economic Outlook report in October, it did this week. Similarly, we have published several scenario iterations over the past six weeks and will provide an update later today in our Market Outlook report.
Using scenarios rather than simple uncertainty ranges or a description of risks is especially useful when the possible future states of the world are qualitatively different from the base case, and when the level of conviction about that base case is low. There is nothing wrong with a scenario that is only a little different from the base case, but it is not that interesting or informative.
Two key principles underlie good scenario analysis. First, the premise of the scenario needs to be well-grounded. The narrative motivating the initial impetus must hang together. It must be logically consistent and represent both the interests and the constraints of the actors within the system. In other words, scenarios should not assume that people do things that are either bonkers or infeasible. Individually rational behaviour that is harmful in aggregate, such as panic-buying toilet paper or conducting fire-sales of loss-making assets, should of course be allowed for, but make sure the action is indeed individually rational.
Second, the methods used in developing the scenario need to go beyond first-round thinking. You need to allow for other people to react. A good scenario cannot be just a one-off shock where nothing else changes. Sometimes it takes time for people to adjust to the new situation, so the reaction occurs with a lag. Teasing these reactions out is the value-add of the scenario.
There is a tension, though, between the need to be realistic and a desire to highlight vulnerabilities or contemplate worst-case scenarios. Sometimes a lack of realism is used to guard against a potential failure of imagination. There is a legitimate role for this approach when stress testing, for example. However, unrealistic assumptions in forecast-flavoured scenarios leave your forecasts open to misinterpretation. We have seen an example of this issue this week, where the IMF’s worst-case ‘severe’ scenario, which implies a global recession, has sparked unhelpful talk of recession here in Australia.
Part of the issue is that the IMF’s less-bad ‘adverse’ scenario (which assumes oil prices average USD100/bbl this year before moderating to USD75 next year) and worst-case ‘severe’ scenario (which assumes oil prices stay around USD125/bbl from the outbreak of the conflict all the way through to end-2027) are both ‘top-down’ scenarios that appear to start from an assumed path for oil prices. Unlike our own scenario work, they do not explicitly model loss of supply from the Middle East and work out what prices need to do to clear the global market. By effectively just drawing a line on a graph for oil prices, the scenarios assume away any scope for other producers to respond. This is not an issue in the near term, but the longer the assumption is maintained, the less realistic it is. High prices will spur non-OPEC producers such as the US and Canada to boost production. It takes time to get that expansion running, but eventually prices will start to ease.
Because it assumes high prices right through 2027, the ‘severe’ scenario is most challenged by this issue. This is not to say that the IMF should not have published that scenario, but it is important to understand the context and what it was trying to achieve in doing so.
When developing scenarios, you also need to avoid the trap of thinking only of downside risks. Part of the problem is that downside risks usually come from identifiable events. It is easy to construct a plausible narrative for a scenario starting from “this particular bad thing happened”. But as one of my old bosses used to remind us, sometimes you should also consider the risk everything just turns out a little bit better than expected.
As our April Market Outlook goes to publication later today, it is helpful to recall last April’s edition. The ‘Liberation Day’ tariffs had been announced a fortnight previously, and many voices in the market were predicting global or US recessions. At the time we took a more moderate view, based on the principles noted above. Firstly, continuing with very high tariff rates would have been an act of economic self-harm. While it was hard to bet on the Trump administration acting rationally in a context of ‘flood the zone’ headlines and intemperate social media posts, self-interest is still the best assumption. If there is nothing preventing someone from stopping punching themselves in the face, they will stop punching.
Secondly – and this was a key judgement in our forecast last year – other countries had agency and could respond. In particular, China had scope to stimulate, and this would also cushion growth in Australia. In that context, we note that the latest Chinese GDP growth has again surprised on the upside relative to market expectations.
It turns out that even we were too bearish a year ago: the global economy did much better than expected, and global trade kept expanding with barely a hiccup. Other factors, including a tech boom, turned out to be more important than expected. Trade patterns were also re-routed around the highest tariffs, and of course the US government de-escalated, as is proving to be its pattern.
These principles also hold for other kinds of scenarios. Whenever you read a prediction of doom concerning, say, adoption of a new technology, ask yourself what is preventing people from responding to ameliorate the bad thing. What prevents macroeconomic policymakers from easing, for example, or the tax system from redistributing unequal gains? While occasionally decision-makers decide that a bad thing is good actually, holding that position is itself fragile.
Where scenarios can really add value is when they highlight a lasting effect from a temporary shock. Whether it is the person who buys an EV in response to current high petrol prices and reduces their petrol consumption permanently, or the government that reassesses its defence spending, behavioural responses can change longer-term demand trends, and so the prices that prevail further out. Normalisation after a shock does not always take you back to where you were. That is another reason why it is so important to avoid just drawing a line on a graph and calling it a scenario, ignoring system-wide effects.
