Research summary
The global inflation spike of 2021 to 2023 surprised central banks and financial markets worldwide. It has sparked many questions.
- What combination of events triggered the initial price rises?
- Why did these price increases set off a wave of large price increases in sector after sector that lasted for 3 years?
- And what can central bankers learn from this high inflation period following 3 decades when inflation was largely kept low?
This research examines the pattern of sector-specific price changes in New Zealand between 1984 and 2022 to understand why some events trigger an avalanche of price changes while others do not. It shows there are systematic differences in the ways that prices change in low and medium inflation environments. These changes lead to an intensification of inflationary forces whenever annual inflation creeps into or above the top half of the Reserve Bank’s target band of 1% to 3%, and an intensification of deflationary forces once inflation drops to the bottom half of the band.
The patterns can be detected by decomposing the headline inflation rate into a “trimmed mean” core group of sectors with moderate price increases, and upper and lower groups (or “trims”) with unusually large or unusually small price increases. These sectors can change every quarter. The research shows that when inflation increases, the size of the price changes in the upper group increases twice as fast as the price changes in the other groups, which accelerates inflation even further, making it harder for the central bank to keep inflation within the target band. Indeed, the large numbers of extremely large price increases occurring in 2021 and 2022 had not been seen since the 1980s, when inflation was similarly high. Conversely, when inflation is very low, there is less upward price flexibility but more downward price flexibility. In these circumstances, it becomes difficult to raise inflation to the middle of the target band.
The different pattern of price changes likely occurs because firms have different incentives to change prices when inflation is near zero compared to when it is higher. When inflation is very low, firms that want to reduce their relative prices to increase their sales must cut their actual prices. When inflation is moderate, firms that want to reduce their relative prices can simply wait till everyone else increases their prices. When inflation gets even higher, some firms that want to cut relative prices still need to increase their actual prices a little bit, as prices in other sectors have risen so much, while firms that want to increase relative prices raise prices by a lot. The detailed data patterns reveal that the price-setting behaviour of the ‘lower trim’ group is consistent with these incentives. These price dynamics help keep inflation very low when it is in the lower part of the target band but contribute to rapid acceleration when it rises.
These findings indicate that a central bank must be particularly vigilant to prevent inflation from increasing above target. They help explain why it was so difficult to raise inflation to the middle of the target band when inflation was very low in the 2010s. They also have another troubling implication. It appears that downwards price flexibility is minimised when the inflation rate is around 2% to 2.5%. If this is correct, choosing an inflation target at this level may maximise output and employment losses in the event of an economic downturn; if firms cannot cut prices flexibly, they may reduce labour or produce less.
Why was this research done?
This project was undertaken to find out whether we could better utilise the detailed information collected by Statistics New Zealand to identify transitory and persistent inflation, and to understand when the risk of inflation accelerating is particularly high. The project began by focussing on the reasons why the trimmed mean measure of core inflation lagged headline inflation as inflation accelerated in 2021 and 2022. From there the project explored the reasons why inflation dynamics are so different in high and low inflationary environments. It also explored whether the detailed price information available from Statistics New Zealand can be used to enhance inflation forecasts by better identifying the transitory and persistent components of quarterly price changes.
What data were used?
The paper uses detailed sector-specific price data to construct 4 composite series.
- The “core” trimmed mean inflation rate, or the inflation rate of the sectors with moderate price increases.
- The upper and lower trim inflation rates, which are the inflation rates of the goods and services with the highest or lowest price changes each quarter.
- The headline-trimmed mean inflation gap, which is the difference between the headline and trimmed mean inflation rates.
For the period 1999 to 2022, the composites were created from price index data on 109 different sectors that were obtained from the level 3 classification Consumer Price Index data published by Statistics New Zealand.
For the period 1983 to 1999, directly comparable data are no longer available from Statistics New Zealand. For this period, comparable composite series were created from detailed level 4 classification Consumer Price Index data obtained from the Reserve Bank archives covering approximately 300 sectors.
What are the key findings?
The first issue considered in the paper is the way the upper and lower trims, and the headline-trimmed mean inflation gap, vary with the inflation rate. These results indicate that increasingly large price increases in a small number of sectors contribute an increasingly large fraction of the headline inflation rate when the inflation rate rises.
The upper trim inflation rate increases by approximately 2 percentage points when the trimmed mean inflation rate increases by 1 percentage point, while the lower trim inflation rate increases by just under 1 percentage point. This means the trimmed mean core inflation estimate is downwardly biased as it misses the increasing prevalence of large price increases as inflation increases. Unless this bias is accounted for, the trimmed mean is an increasingly inaccurate indicator of underlying inflation when inflation increases.
Secondly, the paper investigates the extent that the trimmed mean estimates of the core inflation rate, combined with estimates of the upper and lower trims, can be used to identify transitory and persistent components of inflation. The paper finds that unusually large deviations in the upper and lower trim components are transitory and are rapidly reversed out of the headline inflation rate. This proves to be a useful technique to identify transitory price changes.
Thirdly, the paper examines the extent that the trimmed mean estimator can be used to improve forecasts of the headline inflation rate. It does this by creating synthetic forecasts of the headline inflation rate that could have been made in the past. The paper shows that when the trimmed mean information is incorporated into forecasts there is a modest but statistically significant improvement in forecast accuracy, particularly for short-term forecasts. This improvement mainly occurs because the trimmed mean series helps to identify transitory price changes.
Lastly, the paper investigates how the whole cross-sector distribution of price changes varies as the average inflation rate increases. The paper shows that the distribution of price changes with the inflation rate varies in a manner closely described by menu cost models of price-setting behaviour. The changing distribution helps explain why inflation is sometimes stuck at very low levels, and why outbreaks of high inflation persist.