When inflation models go wrong

At the annual European Central Bank forum in the beautiful resort of Sintra in Portugal, ECB president Christine Lagarde came close to ruling out back-to-back rate cuts, saying the bank would “take time” to assess the landscape before making a decision.

Jay Powell, her counterpart from the Federal Reserve, was more upbeat, saying the US central bank now saw inflation “resuming its disinflationary trend”, although he was in no hurry to act.

The point of the ECB forum is not only to hear from central bankers assessing the evidence as they see it. These conferences also bring challenge from academics and those in financial markets to enrich thinking and policy.

They are valuable exercises. At least, they are as long as the papers presented are high quality.

The sniff test

In academia, economic models are required to pass lots of statistical tests before publication. Sadly, the battery of robustness does not generally include “the sniff test”. This is a simple test that is passed if informed observers think the results are plausible and reasonable. If the results fail the test, they are either revolutionary and everyone should take note or the model is seriously flawed.

At Sintra, some papers failed the sniff test.

One of the keynote papers, presented by Giorgio Primiceri of Northwestern university, sought to explain the recent inflationary episode. As regular readers will know, this sort of study is now a growth industry in academic circles, with contradictory results already produced by Ben Bernanke and Olivier Blanchard, the IMF and the Bank for International Settlements among others.

The main result here was that demand drove inflation’s rise and fall in the US and even more so in the Eurozone. Supply shocks since the pandemic were almost irrelevant. The most important chart showing the dominance of yellow bars representing a demand shock is below.

It was brave to present this result at an ECB forum because the central bank’s own research shows supply shocks were three times more important than demand shocks in the post-pandemic inflation episode.

But it was also surprising to go big on this result. Our eyes and the news told us that natural gas pipelines to Europe were blown up in 2022, there was the mother of all energy crises that required governments to intervene in markets and the wholesale gas price surged to 10 times its normal level that summer. All of this screamed supply, supply, supply.

So, the sniff test rejects the result above. Either something revolutionary has happened or the model that generated it was wrong.

Given other models produce the exact opposite result, I am going to downplay the revolutionary possibility and concentrate on what might have gone wrong in the economic analysis.

There are two possibilities. First, the source of the problem could be that the model distinguished between supply and demand shocks very crudely and incorrectly.

It assumed that if both prices and output went up, there was a positive demand shock. On the other hand, if prices rose, while output declined, there was a negative supply shock.

What we know is that in the aftermath of the pandemic output recovered strongly, there was high inflation and economies were reopening after Covid. To me that was a number of supply shocks in different directions with the possibility of some excess demand. Reopening from the pandemic was not a massive demand shock, but the model would be in danger of categorising it as such.

Second, the model was estimated using data in a period without much inflation, so will find that central banks are generally able to keep inflation close to 2 per cent and do not accommodate supply shocks. By construction, therefore, that means it will have a tendency to label inflation episodes as demand shocks.

Many others last week made similar points. Former ECB vice-president Vitor Constâncio said on X that the paper used “a simple inadequate model to reach the (wrong) conclusion”. Kamil Kovar, associate director at Moody’s Analytics, wrote a thread on X saying the paper was “an example of what I don’t like about current macroeconometrics”, and that the authors should think before presenting results like that.

A second paper at the forum, by Kristin Forbes of MIT and Jongrim Ha and M. Ayhan Kose of the World Bank, contained great analysis of past interest rate cycles. Sadly, it too used the same technique to split demand from supply shocks, got the same implausible result and also failed the sniff test.

Let’s hope the ECB takes these papers seriously but not literally. The good news is that all the signs from policymakers in Sintra, such as Lagarde in her opening speech, suggest they are using the sniff test appropriately.

The scale test

It is a fundamental duty of economists to be able to distinguish big from small. Some of the smartest members of the profession sometimes find this rule difficult to follow.

There was a really interesting session at Sintra on geopolitical shocks and inflation. A paper by Matteo Iacoviello of the Federal Reserve contained a wonderful data set measuring geopolitical stress from news reports (including those in the Financial Times) since 1950. I am pleased to say the key chart passed the sniff test.

Iacoviello then used this data as an explanation of inflation in an economic model and concluded that the Russian invasion of Ukraine reduced global GDP by about 1 per cent and raised global inflation by a maximum of 1 percentage point in 2022.

The problem here is not the result that geopolitical tension raises inflation, but its scale. Global inflation, measured by the IMF, was five percentage points higher in 2022 than its average this century.

A much more powerful conclusion, therefore, would have been that Russia’s invasion of Ukraine was much more inflationary than we would expect from the normal consequences of similar geopolitics.

That is a more interesting result. It should give us hope that in a more geopolitically uncertain world ahead, we are not likely to have repeats of the 2022 inflation experience unless they are centred on one of the world’s food and energy exporting regions.

What I’ve been reading and watching

• The Bank of England and Banque de France have refrained from dishing out advice to their newly elected members of parliament. But a new finance minister in Ireland has not been met with the same restraint. Gabriel Makhlouf, Ireland’s central bank governor, said the government risked “making the inflation problem worse by overspending”.

• Mohamed El-Erian calls on the Fed to make the Jackson Hole conference next month more relevant with a focus, among other things, on the Fed’s forecasting errors and its “out of date” monetary policy framework.

• Over at Free Lunch, Martin Sandbu expertly examines some of the recent analyses of inflation and, like me, finds they fall short. His article is what you should read if you favour “high pressure” economies and the view that monetary policy was not very important in the current disinflation

• Consumer goods companies are losing their pricing power (after raising prices a lot in recent years). Good.

A chart that matters

The UK’s new chancellor, Rachel Reeves, used a speech on Monday to sweep away many obstacles to UK infrastructure and house building. Answering questions, she said she had no plans to change the way that the Bank of England paid interest on reserves built up under quantitative easing, despite these costing twice as much in the UK as elsewhere.

Reeves has a difficult job. Just how difficult is set out in this wonderful chart compiled by my colleague Valentina Romei. Nothing in the UK economy is as favourable as it was when Tony Blair became prime minister in 1997. She will have to hope there is therefore much more to improve.