Scenarios won’t transform economic forecasts

Thank you for commenting so expertly on the causes of the great inflation last week. I’ve collated some of the responses below, most of which concluded that the insides of the economic modelling sausage factory were pretty ugly. This week, I will take a look at economic scenarios. These promise to be the antidote to the forecasting difficulties central banks have experienced in recent years. You might notice I have some reservations. I’d love your thoughts. Email me: chris.giles@ft.com

Forecasts vs scenarios

A particular bugbear of mine occurs when officials paint some picture of the future and say: “This is a scenario not a forecast.” As I will explain shortly, this is a distinction without a difference in almost every case. Worse, the distinction generally has the specific purpose for the central banker of removing their accountability for errors of analysis or judgment. Ultimately, it makes public officials appear slippery and erodes trust in important economic institutions.

With views like these, you will understand that my ears pricked up when Bank of England governor Andrew Bailey gave evidence to UK parliamentarians last week. He said the former Federal Reserve chair Ben Bernanke, who is advising the bank on forecasting, was actively looking at a greater role for scenarios in future BoE analysis, policymaking and communication.

“If we were to use scenarios more — in other words, present scenarios around a forecast — would that help?”

Bailey posed a good question. A focus on scenarios has been gaining a growing following among central bankers along with some academics, particularly during the pandemic.

Before we can analyse the pros and cons, we need a brief detour into the difference between forecasts and scenarios. Central bankers often worry that the public believes their forecasts are unconditional and akin to predicting “the sun will rise in the east tomorrow”. In fact, all economic forecasts are conditional and depend on a series of assumptions relating to the path of interest rates, energy prices, the state of domestic and international politics and fiscal policy, among many other things.

The moment a forecast is conditional, however, it is just a scenario with a particular set of assumptions. There is no difference. Conceptually, it is also the same as seeing the future based on tea leaves, wrinkles on your palm or your star sign. These are also conditional forecasts (although with underlying models I don’t like much). Unlike central bankers, I think the public are quite good at understanding conditional forecasts and language has evolved to convey this meaning quite specifically. If I shout, “read the tea leaves”, you know I am am accusing you of failing to look at the evidence correctly.

Enough of that digression. There are some strong arguments in favour of using scenarios in addition to the usual forecasts. The first use case arises when you want to play the thought experiment, “what conditions would have to be in place to make something happen?” The Bank for International Settlements was particularly effective in 2022 on this, outlining what needed to happen for the world to get stuck in a high inflation equilibrium. For monetary policy, it is always worthwhile to think what would need to happen to end up in a high or low interest rate world, for example.

A second use for scenarios is to examine radically different conditioning assumptions. In the pandemic, it was helpful to look at the likely economic landscape under scenarios such as “a vaccine becomes available” and “new deadly Covid variants emerge”. These had starkly different policy prescriptions, as a working paper by Michael Bordo, Andrew Levin and Mickey Levy in June 2020 highlighted.

A third important use case arises from central banks’ delicate relationship with governments. Monetary policy officials generally have to base their main forecasts on existing fiscal policy, since anything else would suggest they were second guessing elected politicians. What should be possible occasionally, however, would be an alternative scenario based on a different fiscal policy. “What happens to inflation if the US government does not allow the 2017 tax cuts to expire in 2025?” That is something that could give an interesting and relevant answer. Alternatively, it might serve a purpose in showing the monetary policy consequences were trivial.

A fourth use case arises when considering alternative outcomes on a more routine basis, such as what the difference is between assuming energy prices follow the future’s market curve or something else. European Central Bank chief economist Philip Lane has said that different scenarios of the Ukraine war were helpful in framing ECB thinking in March 2022, for example.

But in the end I am still sceptical about the transformational value of building alternative scenarios around the main central bank forecasting scenarios — either in private or for public consumption.

I caught up with Professor Martin Weale, former BoE external monetary policy committee member, for a chat about this last week. He said it was very interesting to look at different scenarios and their implications for monetary policy. The problem when it came to setting policy, he added, was that you need to “know which scenario you are in” to be able to have a view on rates, so they don’t give you any more guidance than the normal forecasts.

Also, economic models tend to bring the world back into some sort of order pretty quickly. A good illustration of this is to look at the scenarios facing the Fed in September 2008 in its “greenbook”, which is published five years in arrears. As the chart from Bordo, Levin and Levy below shows, none of the unemployment scenarios modelled by the Fed as Lehman Brothers collapsed was remotely serious enough, compared with the actual unemployment (red line) the US suffered after 2008.

My third reservation is that central banks are unlikely to use scenarios where they might be most valuable in examining potential fiscal policy moves and most likely to use them as a tool to minimise accountability for their analysis and decisions. Even if the world was different from the conditioning assumptions made in central bank forecasts, it is legitimate to question their actions and thinking.

A highly relevant scenario

I’ve lost count of the number of times I’ve heard central bankers say they must be cautious about inflation because energy prices might rise again. This is a reasonable scenario to ponder. It would be better, however, if European central bankers also questioned closely what happens if energy prices keep falling, as they have since mid-October. The falls in European natural gas futures are important and large as the chart shows. If you click on it and go online, you can toggle between UK and European wholesale prices.

Revisiting Bernanke and Blanchard

I received a large postbag regarding last week’s newsletter on the causes of the great inflation using the framework developed by Olivier Blanchard and Ben Bernanke. Some, such as Erik Nielsen, adviser to UniCredit, said the results showed the ECB could have been more cautious in raising rates since the initial inflation conditions were weak.

Others were less focused on the results than the estimation methodology. Paul Donovan at UBS worried that high vacancy to unemployment rates was not really demonstrating tight labour markets, but post-pandemic special factors. The model was potentially giving incorrect results, he said. Stefan Hofrichter at Allianz criticised the lack of fiscal and, particularly, monetary stimulus as a potential inflation driver in the modelling.

The most comprehensive response came from Marco Casiraghi and Krishna Guha at Evercore ISI, who have replicated the Bernanke Blanchard model for the US (having received the code from the authors) and played around with the assumptions. Their key conclusion is that the model is rather sensitive to the precise assumptions used. They argue that more plausible alternative specifications provide a very different policy conclusion. Instead of demand weakness and higher unemployment being necessary to bring inflation down, they find the opposite results are possible with a different specification of the same model.

Similar to Donovan, they worry that the vacancy rate is not the best measure of a tight labour market for the US and get very different results using the quits rate as shown in the following interactive chart. If you’re reading on email, click on the chart to view it interactively online.

What I’ve been reading and watching

• Advanced economy house prices are rising again, concludes my colleague Valentina Romei after analysing OECD data. Good news if you’re concerned about the fragility of the banking system. Bad news if you’re concerned about the affordability of housing

• Not surprising at all is the fact that central banks are making large losses arising from quantitative-easing programmes in a higher interest rate environment. Last week, we saw the ECB and Bundesbank reporting. These are genuine public finance losses for the consolidated public sector, reaped mostly by commercial banks, although most countries like to pretend otherwise

• Never one to avoid a scrap, Larry Summers of Harvard University has joined in the debate on why the US public appears so cross about a strong economy. People don’t like high borrowing costs, a working paper he has co-authored concludes

• Former BoE official David Bholat convincingly argues that the central bank digital currency blueprint in the UK creates “a product without an obvious market”. It won’t pay interest, prevent private-sector surveillance, be anonymous or safer than existing digital payment mechanisms

• In my column last week, I look at the real differences between the EU and US economies — rising relative employment offset by weakening productivity. Resolving this needs both structural reform and a supportive macro environment

A chart that matters

Don’t be surprised on Thursday if US inflation, as measured by the core personal consumption expenditure deflator (the one the Fed prioritises), is “surprising”. The consensus expectation is for the monthly increase in January to be 0.4 per cent, the same as the monthly rise in the consumer price index a couple of weeks ago. But while the PCE tends to rise a little less, the variability in outcomes relative to the CPI have become much greater since the pandemic. I’m not suggesting it will be much lower, but there have been many surprises recently.

Source: Economy - ft.com