COVID-19 inflation was a supply shock

We use data from S&P Global on purchasing managers' indices (PMIs) for 45 economies. These are monthly opinion surveys on supplier delivery times, input prices (what companies pay suppliers) and output prices (what companies charge customers), as well as new orders and inventories. In our first post, we used this data to discuss the massive scale of supply chain disruptions during COVID-19, comparable in magnitude to what Japan experienced after the Fukushima nuclear disaster in 2011, except that COVID-19 supply disruptions in the U.S. lasted much longer. In our second post, we looked at corporate margins, which rose sharply after the spike in delivery delays and have yet to normalize. We believe this increase in margins reflects increased efforts by companies to conserve inventory following the shock of empty shelves at the height of COVID-19, an explanation for margin expansion that is clear and perhaps more plausible than “greedflation.” This paper examines the extent to which the inflation shock caused by COVID-19 is due to supply chain disruptions, including through higher margins, and draws conclusions for the future pace of disinflation.

Figure 3. The supply shock: firms’ production and input prices

We now go through the key potential drivers of the COVID-19 inflation spike. Figure 3 shows the PMIs' production and input prices, which we converted to z-scores by downgrading and scaling by their historical standard deviation (a z-score of one can be interpreted as a value one standard deviation above the pre-COVID average). This chart shows that production prices have risen more than input prices, an increase that correlates with the COVID-19 inflation spike. Figure 4 uses the difference between our production and input price z-scores as a proxy for firms' margins, a novel measure of supply chain disruptions that we have not seen in other related work. Figure 4 shows that delivery times have normalized to pre-COVID levels in late 2022, while our proxy for margins has fallen much more slowly and is only now slipping into negative territory (a sign that margins are falling). The preferred measure of labor market tightness is the job openings-to-unemployment ratio (see Figure 5). This has risen sharply during the pandemic, suggesting that the high number of job openings relative to the number of unemployed may have contributed to rising inflation. Finally, Figure 6 shows two other potential drivers: oil prices, which rose sharply as the global economy recovered from COVID-19, and movements in the broad, trade-weighted dollar, which could also affect inflation.

Figure 5. The labour market: JOLTS data on job vacancies

A greater role for supply chains in the Covid-19 inflation surge

We use an econometric model to link annual core PCE (personal consumption expenditures) inflation to measures of supply chain disruptions, labor market tightness, oil prices, and the trade-weighted dollar. We use three measures of supply chain disruptions: (i) PMI shipping delays, which are widely used in the literature; (ii) our margin proxy; and (iii) new firm orders minus inventories, constructing the latter in a similar way to our margin proxy (the difference between the z-scores for new orders and inventories). Our measure of labor market tightness is the ratio of job openings to unemployed. In addition to oil prices and the broad dollar, we consider long-term inflation expectations, for which we use the Federal Reserve Bank of Philadelphia's expected inflation rate over the next ten years.

Figure 7. Drivers of PCE core inflation, baseline model

Figure 7 breaks down the drivers of core PCE inflation in our base model, which uses only delivery times as an indicator of supply chain disruptions. Figure 8 shows the same for our extended model, where – in addition to delivery times – we also use margins and new orders minus inventories as an indicator of supply chain problems. In both cases, the black line represents year-over-year core PCE inflation, from which we have subtracted the constant in our regression and the contribution of 10-year inflation expectations. The bars show the contributions of different drivers. This leads to three conclusions. First, regardless of which model we use, supply disruptions explain the majority of COVID-era inflation. If we use only lead times, supply chains explain 58% of year-on-year inflation in Q4 2021, while this figure rises to 79% in the case of our expanded list of supply chain variables (these shares are 25% and 62% in Q4 2022, respectively). Second, margins are by far the most important supply chain variable and play no role in the vacancies-unemployment metric. The share of inflation in Q4 2021 attributable to lead times rises from 58% in our simple model to 24% when we include margins. Labor market tightness—and thus excessive policy stimulus—appears to have played a minor role. Third, margins are significantly lagging the normalization of lead times. In fact, this delay is so important that the overall supply chain effect in Q4 2023 is positive at 13% in our extended model, while that figure is -27% in the model that only considers delivery times. In fact, margins are just now turning negative in the latest data, meaning this could be a major source of disinflation in the future.

Our finding that margins play a key role – and crowd out lead times – is a key result. The fact that lead times no longer drive inflation in late 2022 leads Bernanke and Blanchard to conclude that demand must be the explanation for the still elevated inflation of recent times. Our result provides an alternative explanation. Warehouse managers may be slow to normalize margins because the shock of empty shelves at the height of COVID-19 was so severe. So high inflation after lead times normalize could still have supply disruptions as the main cause.

The importance of margins as a lagged transmission channel also has important implications for the Fed. We attribute much of the decline in inflation since the peak of the pandemic to supply-related factors rather than demand, which – by extension – means that the Fed's rate hikes may have played only a modest role in lowering inflation. If true, it suggests that the Fed may have tightened monetary policy too much in 2022 and should now move to easing, perhaps in rapid succession.

Of course, our analysis is subject to important caveats. One problem that hampers all work in this area is that the COVID-19 episode is unique, meaning that our results may not generalize beyond the sample and into the future. This is a problem that Bergholt and co-authors highlight. A further complication arises from the fact that extended lead times and increased margins inevitably have a demand component. Our reading of the existing literature suggests that variations related to supply chain variables are largely attributed to “supply,” but this may underestimate the role of demand. Finally, it is possible that tight labor markets have a disproportionate effect on inflation, an argument for a nonlinear Phillips curve as proposed by Gagnon and Collins. Bernanke and Blanchard find no evidence of such nonlinearity during COVID-19, but this again raises the possibility that the role of demand may be underestimated.