Benjamin Wong, Monash University and CAMA
15 February 2023
These are my personal views and should not be attributed to any other institutions or individuals.
iStock image
Figure 1 US year-on-year Personal Consumption Expenditure (PCE) inflation (percent)
Central banks have spent the last two years grappling with high inflation. Figure 1 presents U.S. year-on-year Personal Consumption Expenditure (PCE) inflation from 1959Q2 to 2022Q4 (currently the last quarterly inflation data point). While the inflation outcomes are arguably the worst in a generation, the consensus view is that inflation may have peaked in late 2022. One can visualise the peak inflation view from the slowing inflation at the end of the sample.
Figure 2 U.S. year-on-year Personal Consumption Expenditure (PCE), goods, and services inflation (percent)
A feature of the current inflation episode is the dynamics of goods inflation. Figure 2 presents year-on-year PCE inflation, as well as year-on-year inflation in both the goods and services sectors from 2020Q1 to 2022Q4. As is clear, inflation started picking up at the beginning of 2021. However, the extent of these increases in the goods sector is unprecedented. The supply-chain disruptions in 2021 have led to a large acceleration of goods inflation and arguably played a large role in the current inflation surge. The perception of a clear identifiable source of high inflation stemming from the supply-chain disruptions has led to a policy dilemma on how to best respond to inflation. The U.S. Federal Reserve, as well as central banks globally, hesitated to raise interest rates until late 2021. They communicated that much of the high inflation is transitory as there is little expectation that the supply-chain disruptions would continue indefinitely. It is also worth highlighting that with the widely accepted long and variable lags in the transmission of monetary policy, where interest rate decisions are only expected to affect the real economy and inflation in 12 to 18 months, not responding to inflation that is less persistent (or transitory) is also consistent with standard central banking doctrine.
Figure 2 also shows that while goods inflation has slowed down, services inflation has not. One way to view current developments is a recently developed model by Eo, Uzeda, and Wong (forthcoming), henceforth referred to as the EUW model. The EUW model estimates trend inflation in both the goods and services sectors. Trend inflation measures are designed to separate the persistent components of inflation (trend) from transitory noise and are so used to inform about the degree of underlying inflation. Central banks have a long tradition of appealing to concepts akin to trend inflation. For example, core inflation measures, such as excluding food and energy prices or other exclusion measures such as trimmed means or weighted medians, were developed to some extent for providing a read of the persistent component of inflation. While the EUW model is one of many models of trend inflation, the deliberate split between goods and services inflation does provide some natural structure to interpret the current inflation developments.
Figure 3 Estimated U.S. aggregate trend, trend goods, and trend services inflation. All inflation rates are expressed in annualised percent terms. The black dotted line represents the posterior median estimate with the red dotted lines representing 67% credible intervals. The shaded bars are NBER-dated recessions.
Figure 3 presents the estimated trends from the EUW model, with the sample extended from the analysis reported by Eo, Uzeda, and Wong (forthcoming) and Eo, Uzeda, and Wong (2022) to 2022Q4. The core insight from Eo, Uzeda, and Wong (forthcoming) is that since around the mid-1990s, the model has interpreted goods inflation as mainly transitory noise, and therefore all the variation in trend inflation can essentially be attributed to the services sector. This is evident from Figure 3 that goods inflation has tended to fluctuate (sometimes in a quite volatile fashion) around a stable trend in the goods sector in the later part of the sample. On the contrary, services inflation is often quite persistent, and trend services inflation tracks actual services inflation very closely. In other words, if monetary policy is concerned about responding to the persistent component, the EUW model would suggest that much of this persistent component comes from the services sector.
The main insights from the EUW model remain robust even when the sample is updated till 2022Q4. That is, the model still interprets goods inflation as largely transitory noise, albeit with a greater-than-usual degree of estimation uncertainty. The degree of estimation uncertainty is non-trivial. While the EUW model has a trend inflation estimate at slightly over 4% in 2022Q4, the 67% credible set ranges between 3 to just under 6%, which, as suggested by Figure 3, is historically large. It is also interesting to note that at the end of the sample, when PCE inflation is falling, trend inflation is still rising.
The EUW model is a state-space model. A feature of state-space models is that they produce both filtered and smoothed estimates. The former represents estimates conditioned on information up to a certain point in time whereas the latter represents estimates utilising information from the whole sample. Filtered estimates are also sometimes referred to as "real-time” estimates as they mimic the information set and uncertainty one would experience in, say, a policy environment. For example, a filtered estimate from 2021Q4 would only account for data up till 2021Q4, with the analyst not incorporating inflation developments in 2022. Comparing the filtered and smoothed estimate provides some perspective on how the EUW model viewed inflation as 2022 unfolded.
Figure 4 Estimated U.S. aggregate trend, trend goods, and trend services inflation expressed in annualised percent terms. The black marked line represents the posterior median smoothed estimate of trend inflation. The red line represents the filtered posterior median estimate of trend inflation, and the blue dotted lines represent 67% filtered credible intervals.
Figure 4 presents the filtered estimates of aggregate trend inflation, trend goods, and trend services inflation from 2021Q4 to 2022Q4. Both the posterior median, which represents the point estimate, and the 67% credible interval of the filtered estimates are presented. For comparison, the smoothed estimates of the posterior median are also presented. Viewing the filtered estimates provides a slightly different perspective on the estimation uncertainty. While the degree of estimation uncertainty is still historically high at the end of the sample, as suggested by Figure 1, this has reduced somewhat towards the end of 2022. Focusing on the degree of filtered uncertainty for trend goods inflation, this source of uncertainty became extremely large, especially in 2022Q1 and 2022Q2. These quarters in early 2022 coincided with the outbreak of the Russia-Ukraine war and the exacerbation of some of the supply-chain disruptions in early 2022. While the EUW model still regards goods inflation as transitory, the incoming data in early 2022 being partly at odds with this view is reflected in a large degree of estimation uncertainty for the level of trend goods inflation. With the more recent data showing a decline in goods inflation, the estimation uncertainty of trend goods inflation reduced substantially by 2022Q4, reflecting a reinforcement of the model's view that the increases in 2022 were transitory.
Perhaps more concerningly, trend services inflation has continued to increase, and the uncertainty around trend services inflation actually reduced as we went through 2022. Recall that the interpretation of the EUW model is that much of the variation in aggregation trend inflation can be traced to trend services inflation. That trend services inflation continues to increase has led to the (smoothed) estimates of aggregate trend inflation estimates increasing through to the end of 2022. Another way of stating the above is that the model is more confident of the view that trend services inflation has increased.
There are two key messages from the above exercise that are worth emphasising. First, with the fall in goods inflation, estimation uncertainty about the aggregate level of trend inflation has reduced towards the end of 2022. However, this reduced uncertainty has clustered at a higher estimate of aggregate trend inflation of over 4%. In other words, if we were less certain a year ago, we are now more certain that a reasonably large component of high inflation is, and continues to be, persistent, and can thus be expected to stick around into the near future. Given services inflation is the persistent component of inflation from the EUW model, without a large and sustained fall in services inflation, it is challenging to reconcile how trend, and therefore future, inflation would fall back into the range that we are comfortable with.
Second, while commentators suggest good news from inflation possibly peaking late last year, this may represent a glass-half-full interpretation. While inflation has declined in recent months, these come mainly from goods inflation falling, which one can partly attribute to the supply-chain disruptions resolving. In other words, goods inflation may well have turned out to be transitory, as first communicated in 2021, albeit lasting a bit longer than expected. However, while inflation may have peaked, it is not obvious from the EUW model that trend inflation has peaked. While the filtered estimate has come down from 2022Q3 to 2022Q4, the smoothed estimate has been continuously increasing since 2021Q4. In other words, even if inflation has been declining in recent months, the component of inflation that is perhaps the most challenging to bring down may still be rising and potentially a long away from the stated 2% inflation target.
While this post is very much U.S.-centric, given the global nature of the supply-chain disruptions and the similarities with other economies, the key point should resonate more broadly. Moreover, Eo, Uzeda, and Wong (forthcoming) also show that their results carry over to small open economies such as Australia and Canada, so the commentary contained in this post may also generalise more widely. If the recent falls in inflation are mostly due to factors that are out of the direct control of monetary policy, such as goods inflation falling due to supply chains normalising, 2023 and 2024 will be in the 12 to 18 month window where the effects of the steep rate hikes in late 2021 and 2022 should filter through to bring down inflation if the conventional wisdom of monetary policy operating with long and variable lags holds. If so, one should be hopeful that there will be corresponding reductions in services inflation and therefore bringing down the persistent component of inflation. One should nonetheless remain vigilant for future data releases to reconcile with this hopeful view. In this regard, the recent communication that the Federal Reserve is focusing on core services, albeit excluding housing, should be seen as a welcomed development. While inflation may well have peaked, it would appear premature at this point to claim the end is in sight. Just as a distinguished statesman once said, “Now this is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning”, we may still have some way to go.
References
Eo, Y., Uzeda, L., Wong, B forthcoming. "Understanding trend inflation through the lens of the goods and services sectors," Journal of Applied Econometrics
Eo, Y., Uzeda, L., Wong, B (2022) Goods inflation is likely transitory, but upside risks to longer-term inflation remain VoxEU.org, 29 April.