The way inflation is calculated could be flawed. Here's one potential solution

On the most recent edition of Livewire's Signal or Noise, we discussed the mixed messaging being emitted from the economic data. For instance, retail sales are still very strong and leading indicators like PMI surveys remain resilient despite the fact that since 2020, the cumulative inflation borne by Australian households is at about 17%. 

But what if I told you that the process for collecting and publishing inflation data is wrong? Not wrong in the sense that the figures are incorrect but rather that it's not being calculated in a way that suits the modern economy.

That's certainly the view of one of Livewire's readers - Steve Neubecker. Neubecker argues it's time for us to move to a trend-oriented method of calculating price increases and even submitted a piece to our inbox suggesting so. One Spanish academic agrees with Neubecker, recently coining a new term for this approach called "instantaneous inflation". But is the way we really measure data not fit for purpose anymore?

In this wire, I'll probe Neubecker's proposal and share some insights from two leading economists - Monash University's Dr Isaac Gross and Signal or Noise's very own Diana Mousina.

How is inflation currently calculated?

Before we go on, I think it's worth reminding you about how inflation is actually calculated in Australia - and some of the important caveats associated with it. 

• Inflation is calculated using a representative "basket" of goods and services. The idea is to collect a group of products that the "average Australian consumer" (whatever that means) purchases on a regular basis. 
• While not every household owns a car, cars are included in the CPI basket for simplicity's sake. The basket is updated every 12 months.
• Australian inflation only covers the eight state and territory capital cities. Or put another way, inflation is representative of the price increases being experienced in 2/3 of the population. 
• Inflation is a measure of price changes and not levels - and just as importantly, inflation is not a measure of the cost of living. That's a completely different index altogether.

Finally, and most importantly in the context of this piece, inflation is still measured quarterly in this country. We now have a monthly inflation indicator but it is just that - an indicator. The indicator only covers 2/3 of the entire CPI basket every month for the simple reason that some price changes take longer to affect consumers than others. 

The biggest problem with inflation

The biggest thorn in the ABS' side is that inflation is a lagging indicator - that is, it collates past data to show how much prices have increased across the economy over a certain period of time. 

This is the crux of Neubecker's problem: when inflation is going up as much it has been over the last two years, the current way of calculating inflation is not a suitable reflection of what the economy is actually going through. 

"The headline figures reported are actually the average percentage increase in prices since a year ago, which does not say much about what the inflation rate is right now," Neubecker wrote in a piece to us. 

"The reported inflation figures (averaged over the last 12 months) will therefore have approximately a 6 month lag in the timing, ie. the March 2023 official figure might actually be closer to what the real inflation rate was back in about September 2022," he added. 

Neubecker's proposed solution

Neubecker proposed to us a new idea, first authored by Dutch academic Jan Eeckhout. The concept of "instantaneous inflation". Put simply:

"The conventional measure of inflation lives in the past," Eeckhout wrote in his paper published earlier this year. "Because data is noisy, we cannot use the monthly inflation reading, but we can construct an average that puts higher weights on more recent observations," he added.

Essentially, the premise is that if you took out all the month-to-month noise (like sampling errors and data input issues), inflation arguably peaked in Australia six months before it showed up in the official data. 

If Neubecker is right, then the RBA has definitely hiked rates too far and too much - possibly as much as six meetings' worth of hikes. And remember, there are many leading and lagging indicators which now show the Australian economy is certainly slowing. The RBA's reasons for hiking at the June meeting were almost entirely centred around inflation. 

Dr Gross' view

Dr Isaac Gross is a lecturer at Monash University in Melbourne. Gross has a DPhil and a MPhil from Oxford University in Economics. From 2011 to 2013 he worked as an economist for the Reserve Bank of Australia and he is a former writer for The Economist.

Dr Gross says Neubecker's proposal is interesting but "flawed". Polynomial regressions can be used to make predictions about future data. But they can be affected by the most recent data point. If the most recent data point is very different from the other data points, it can make the predictions very inaccurate.

"That is exactly what happens in this application. Because the Q1 2023 inflation rate started to fall from the peak in Q4 this method extrapolates that change even further and predicts that the inflation rate is less than 4%, when the actual quarterly rate is 5.6%," Gross said.

In fact, Dr Gross argues this method adds noise and statistical bias rather than takes it away - making the Reserve Bank's job even harder than it already is!

Mousina's view

Diana Mousina is AMP's Deputy Chief Economist and the regular panellist on Livewire's economics show Signal or Noise. A graduate of UNSW, she began her career at the Commonwealth Bank before moving over to AMP. Along with her senior, Shane Oliver, they are two of the leading voices in the Australian economics community.

Mousina starts off by acknowledging Neubecker's main pain point. Inflation isn't moving fast enough, it is a lagging indicator, and it explains why many economists would rather look at the 3-month annualised data or the new monthly inflation indicator. But for those who want to use the latter as an indicator, Mousina has this piece of advice:

"The only issue with this is that it’s not seasonally adjusted, so there is still merit in looking at the year-over-year change which mostly strips out the seasonal effect," she said.

As for the timeliness of the data, Mousina understands where Neubecker is coming from but also notes that data collection is as important as calculation.

"Realistically, CPI data is only sampled monthly to quarterly. So the fastest accurate rate of change in price we can measure is the monthly change. Any attempt to calculate a “faster” moving measure requires some assumption," she said. 

Neubecker's proposed CREST method is not wrong, in AMP's view, but rather just an alternative way to look at the rate of change. But perhaps most importantly - Mousina has an important question for anyone who wants to propose a radically different way of calculating inflation such as this one.

"From an RBA point of view we think that this method is harder for people to understand. How can you explain this regression method to the everyday Australian as a basis for why the RBA made the decision it did? I think that would be a hard sell," she said.

You can read Steve's entire reasoning in the PDF below. We thank him very much for submitting such a thought provoking and interesting piece.