You can’t model what you don’t understand

Central banks and economists worldwide failed to forecast the path of inflation after Covid. Many see the remedy as better economic modelling, but you can’t model what you don’t understand.

The current inflation experience, hopefully nearly over, is an opportunity for a better understanding. We need to start this re-examination now, while the evidence is still fresh.

Imperfect understanding isn’t surprising. For most economies, the last period of sustained inflation was fifty years ago: the stagflation of the 1970s. Since then, much has changed.

At the macro level, a key change is the near-universal success of three decades of inflation targeting. Central banks built low-inflation credibility.

This influences price-setters. When firms came to reset their prices in the disruptive demand/supply imbalances of the post-Covid period, they not only considered their own market power, but what was “normal”. Even if firms had market power to raise prices by more, the inflation target was the starting point: how much ahead of this figure could they go while still retaining customers, reputation and social license?

In the chronic inflationary environment of the 1970s, it had become normal to build automatic inflation-based increases into contracts — especially rental contracts and wage agreements (COLAs). These automatic responses produced round-robins of inflationary momentum – in the jargon, inflation had persistence.

To overcome such persistence, it was necessary to have a sharp contraction of the economy, until slack in the economy asserted strong discipline on price-setters. Above-normal unemployment was necessary. Hence US Fed chair Volcker’s 1979 shock saw the Fed funds rate go to 17 per cent and unemployment to nearly 11 per cent. In Australia, the 1990 “recession we had to have” was similar in pain and outcome.

Inflation targeting was the policy response to this painful period, with central banks committing to keep inflation low. Even the US, which didn’t introduce a formal target until 2012, made its determination clear. After the Volcker Shock, who could backslide?

This is the first difference between today and the 1970s: even though Covid (and then the Ukraine war) shocked current inflation sharply upwards, inflation expectations, so far, remain anchored by the three decades of successful inflation targeting. Longer-term bond yields indicate market expectations of a return to target. COLAs aren’t reappearing in wage and rental contracts.

As well, price-setting at the micro-level has changed.

There is far more price differentiation and variation, with prices often now depending on time and place. Price setters create product differentiation through targeted advertising. Short-term price adjustments (“specials”) are common, loyalty schemes abound, honeymoon offers play on customer inertia, and goods are bundled together to make it harder for consumers to assess value.

At the same time, the web revolution has created more opportunities for buyers to shop around. Amazon (and the gig supply revolution) has enlarged sources of purchase. Competition has belatedly come to taxis, electricity and telecommunications. International competition is sharper. More inputs are provided by competitive tender.

Companies are bigger, but competition is enforced by regulation and surveillance to identify collusion.

There are more government-provided services (education, health, infrastructure) with stable prices. More goods are provided free to the consumer – social media, consumer credit.

Perhaps most important, there is more wage flexibility: automatic inflation-related setting is rare.

In short, pricing behaviour has changed and inflation persistence has surely weakened. First-round supply-side increases remain once-off.

Some basic truths remain. If the economy runs too hot, price setters (firms and labour) find it easier to raise prices. When “full employment” is reached, the Phillips curve will assert itself, just as it did in the 1970s. Inflation targeting probably flattened the curve, but when the economy runs hot, prices and wages will tend to rise.

Critically, if this happens for too long, inflation expectations won’t stay anchored.

Without an anchor for inflation expectations, the 1970s inflation could end only by crunching the economy, as Volcker did. In Australia, Paul Keating’s economic adviser said he “heard the economy snap”.

Central banks around the world have treated the current episode differently. Rather than crunch the economy, they have endeavoured to remove the heat, and rely on anchored inflation expectations to transition to a smooth landing at the target.

Certainly, they have raised rates from the abnormally low levels of the post-GFC decade. But in inflation-adjusted terms, policy rates are barely positive. Policy-makers know that if they are too eager to get back to significantly positive real rates, they may crunch the economy. If they are too slow, however, inflation expectations may not remain passively anchored.

The RBA is prepared to wait another two years – until end-2025 – to get back to target. The next year or two will determine whether this experiment succeeds.

The current fashion in economics uses the immense data sources generated by today’s economy to evaluate “natural experiments” – comparisons of detailed economic behaviour which don’t lend themselves to conventional modelling.

The post-Covid inflationary experience provides a perfect natural experiment.

While the data are still fresh, let’s get the RBA, the ABS, and the ACCC to work together, gathering the micro data for the past three years to understand how modern-day price setting works.

The RBA would learn more about persistence. The ABS might identify gaps in statistical coverage.

The ACCC would know where its interventions would have most impact on inflation. And price setters would understand the limits of their power.

Stephen Grenville is a former Reserve Bank deputy governor and a nonresident fellow at the Lowy Institute.