Kristin Forbes, Joseph Gagnon, Christopher G. Collins 21 December 2021
The COVID-19 pandemic and corresponding policy responses have generated substantial uncertainty over the prognosis for inflation around the world. Some observers warned of higher inflation as economies were fueled by stimulus and rebounded from shutdowns (Blanchard 2021, Gagnon 2021, Goodhart and Pradhan 2021), while others were more sanguine (Ball et al. 2021, Brigone et al. 2021, Ha et al. 2021). After falling sharply in 2020, inflation over the last few months has regularly exceeded predictions from earlier this year.
Although many factors have undoubtedly contributed to these inflation surprises, these patterns support our new research on the Phillips curve that highlights the nonlinearity and global nature of inflation (Forbes et al. 2021). More specifically, two key findings of our recent research are that:
- when output exceeds potential, the upward pressure on prices (from reductions in slack) is far greater than any equivalent downward pressure (from increases in slack) when output is below potential; and
- global forces are an important factor behind inflationary pressures, even after taking into account domestic factors.
Although the Phillips curve is a simplification of many deeper causal relationships, a few modifications to the standard linear version show how it can be useful to understand some of the recent ‘puzzles’ in pandemic inflation.
The nonlinear Phillips curve
In Gagnon and Collins (2019), two of us argued that an environment of very low inflation bends the Phillips curve because of resistance of workers and firms to outright price and wage cuts. Indeed, the original Phillips curve from 1958 (Figure 1) was highly nonlinear and based on a period of very low inflation.
Figure 1 Original Phillips curve (1958), UK 1861-1931
Rate of change of money wage rates, percent per year
Source: Philipps (1958), retrieved from Wiley Online Library.
We showed that US inflation since the 1950s is well characterised by a Phillips curve that is linear when inflation is high and nonlinear when inflation is low – what we call the ‘low inflation bend’ model. In this model, changes in slack have little impact on inflation when there is spare capacity in the economy, but reductions in slack generate higher inflation when there is no spare capacity.
Our new research in Forbes et al. (2021) extends this model from just the US to a sample of 31 mostly advanced economies over the period from 1996 through 2017. We find that the earlier results for the US continue to apply to other countries and the low inflation bend model fits best, though other nonlinear models also fit the data better than a constant linear Phillips curve.
Figure 2 displays the resulting low-inflation-bend estimates of the Phillips curves for this larger sample of countries. The right-hand panel shows that the traditional Phillips curve relationship between slack and inflation holds when inflation is above 3% (with results similar if we use a threshold of 2–4%); in other words, inflation tends to pick up when slack falls in countries where inflation is already high. To be more concrete, a 1 percentage point reduction in slack corresponds to about a 0.3–0.4 percentage point increase in inflation.
Figure 2 The low inflation bend model
Source: Replicated from Forbes et al. (2021).
The left-hand panel, however, shows that this relationship disappears when inflation is low and slack exists. In these circumstances, there is essentially no downward pressure on inflation from changes in slack (beyond the normal dynamics that return inflation toward its long-run anchor). As slack is reduced and output returns above potential, however, the traditional Phillips curve relationship reasserts itself. The piecewise linear form of the curve is a tractable way to approximate the bending of the curve at high levels of slack. The curve may be interpreted as reflecting the additional forces operating on inflation beyond a benchmark gradual adjustment towards long-run expected inflation.
The role of global factors
While the traditional Phillips curve focuses on the relationship between domestic slack and inflation, Forbes (2020) shows that international influences can also significantly affect the slope of the Phillips curve – as well as having other meaningful effects on inflation. More specifically, global factors such as global commodity prices, global slack, exchange rates, and producer price competition can all significantly affect inflation, even after controlling for the standard domestic variables. These global variables are important not only when inflation is modelled using a Phillips curve framework, but also using alternate approaches (such as principal components or trend-cycle decomposition).
This research also shows that the role of these global factors increased over the decade before COVID. For example, the shared global component of CPI inflation more than doubled from about 27% from 1990–94 to 57% from 2015–17. Global slack, commodity prices, and supply chains had larger effects on domestic inflation than in earlier periods.
Does adjusting for these multifaceted global variables impact the nonlinearity of the Phillips curve? To test this, in Forbes et al. (2021) we incorporate the same set of five global variables into our nonlinear Phillips curve models. We find that the Phillips curve is still nonlinear and the low inflation bend model continues to provide the best fit for inflation in the sample of 31 countries, even after controlling for global factors. We also find that global factors continue to be important when allowing for nonlinearities in the relationship between slack and unemployment, and that the role of these global variables in explaining CPI inflation has increased over time.
These global factors and nonlinearities in the Phillips curve have played an important role in driving the sharp swings in inflation rates during COVID.
Inflation fell sharply in most countries during the first months of the pandemic as large sectors of the economy shut down and consumers sheltered at home and reduced their spending (see Figure 3). The collapse in global commodity prices was an important factor in this immediate fall in inflation, a global drag that increased as world slack spiked. Given the size of the economic contraction and increase in domestic slack in most countries, however, the reduction in inflation was smaller than would be expected based on a linear Phillips curve – even after controlling for these global factors. The large margin of domestic slack that opened up in most countries caused them to jump to the flat region of the Phillips curve, where any continued deterioration in activity had minimal additional impact on inflation.
Figure 3 Consumer price inflation, 2017–2021
Source: International Monetary Fund, World Economic Outlook, October.
The upswing in inflation this year was also widespread and reflected a mix of global factors and the nonlinearity in the Phillips curve. The rebound in economic activity increased demand for energy and other commodities that trade on global markets, generating an immediate spike in inflation. The slower recovery in supply chains, fed by strong consumer demand in the US and a series of supply shocks, further contributed to inflation around the world.
Even after accounting for these global factors, however, inflation has picked up more than expected in most countries. This likely reflects the low levels of slack today in many economies after accounting for supply-side changes from the pandemic. This has pushed countries back to the steeper section of the Phillips curve for the time being.
This lack of slack may not be immediately obvious as some conventional measures – such as the share of the working-age population that is employed – are still depressed relative to pre-COVID levels, suggesting that there is spare capacity in the US and other advanced economies. The pandemic, however, has reduced potential output in several ways: (1) some workers are afraid to return to workplaces that expose them to COVID; (2) one adult may need to remain at home to care for children given unreliable school schedules, or for elderly given health concerns; (3) some businesses and sectors have still not fully reopened due to continued restrictions and changes in work and spending patterns; and (4) government assistance and a corresponding buffer of savings have enabled workers to stay out of employment longer than usual. In addition, the pandemic has shifted consumer demand away from in-person services to goods, and away from urban centres towards less-populated areas, straining the capacity of the goods sector and certain localities more than would be expected for the same level of overall demand.
When these effects are combined with substantial fiscal and monetary stimulus, as has occurred in the US, economic output could exceed the existing potential supply. This would imply that the US (and other economies in a similar position) would be on the steep part of the Phillips curve where inflation is high – even after controlling for the temporary effects of global factors.
The prospects for inflation going forward are complicated. If COVID and its associated economic effects recede in coming months, supply should increase and demand should revert to a more normal pattern. This increase in potential supply could generate outright declines in the prices of some consumer goods and a rapid reversal of overall inflation. In addition, global factors will further contribute to lower inflation. This will initially occur as last year’s spike in global commodity prices rolls off the inflation indices, followed by a more gradual decline in price pressures as supply chain pressures abate.
This sanguine scenario, however, is not assured. If new variants (such as Omicron) generate renewed health concerns, workers could continue to remain out of the work force and slow the supply-side recovery. If governments and central banks are too slow in removing the stimulus that made sense when economies were locked down (or pass new fiscal packages), demand will continue to outpace supply, especially when any stimulus continues on top of what were already strong household balance sheets. In these scenarios, countries could remain on the steep section of the Phillips curve and, if inflation remains elevated for an extended period, could shift to the more linear version of the Phillips curve on the right of Figure 2.
On a more positive note, if governments understand the constraints of this nonlinear Phillips curve, they should understand what steps need to be taken to support the recovery while ensuring that inflation returns sustainably to target.
Ball, L, G Gopinath, D Leigh, P Mitra and A Spilimbergo (2021), “US Inflation: Set for Takeoff?”, VoxEU.org, 7 May.
Blanchard, O (2021), “US inflation is running high. What should we worry about now?”, Realtime Economic Issues Watch, Peterson Institute for International Economics, 11 November.
Brignone, D, A Dieppe and M Ricci (2021), “Quantifying the risks of persistently higher US inflation”, VoxEU.org, 1 November.
Forbes, K (2020), “Inflation Dynamics: Dead, Dormant, Or Determined Abroad?”, Brookings Papers on Economic Activity, Fall 2019 Meetings: 257-319.
Forbes, K, J Gagnon and C Collins (2021), “Low Inflation Bends the Phillips Curve around the World: Extended Results”, CEPR Discussion Paper 16583.
Gagnon, J (2021), “Inflation fears and the Biden stimulus: Look to the Korean War, not Vietnam“, Realtime Economic Issues Watch, Peterson Institute for International Economics, 25 February.
Gagnon, J and C Collins (2019), “Low Inflation Bends the Phillips Curve”, PIIE Working Paper No. 19-6, Peterson Institute for International Economics.
Goodhart, C and M Pradhan (2021), “What may happen when central banks wake up to more persistent inflation?”, VoxEU.org, 25 October.
Phillips, A W (1958), “The Relation between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957”, Economica 25: 283–99.
Ha, J, M A Kose and F Ohnsorge (2021), “Inflationary pressures: Likely temporary but challenging for policy design”, VoxEU.org, 14 July.