The decline in market-based inflation measures has accelerated markedly since the outbreak of COVID-19. Inflation-linked swap (ILS) rates are useful for deriving market-based measures of inflation expectations. In an inflation-linked swap contract, one participant pays a fixed rate cash flow on a notional principal amount while the other participant pays a floating rate that is linked to an inflation index. Thus, the euro 5y5y ILS rate measures current five-year inflation expectations five years ahead in time by taking a long position in a ten-year inflation-linked swap contract and a short position in a five-year contract. This measure reached an historic low on 23 March 2020 of 0.71%. Although market-based inflation measures have rebounded somewhat following efforts by central banks and governments to stem the economic downturn, with a current level of 0.9% the market-based inflation measure is still well below the ECB’s target for medium-term inflation of below, but close to 2% (see Figure 1).

**Figure 1** Market-based inflation measures for the short and medium term

*Source*: DNB, Bloomberg. Last observation: 23-04-2020.

Markets perceive the outbreak of COVID-19 together with the unfolding oil price shock to be a persistent disinflationary shock, but are simultaneously grappling with the effects of the lockdown on the ability to accurately measure inflation. However, this crisis increases expectations of a long-lasting low-inflation environment, as reflected in a sharp price decline in long-term inflation-linked swaps rates (see Figure 2). Yet, the large uncertainty over the magnitude of disinflationary effects of the pandemic, as well as the market perception of increasingly constrained monetary policy, have increased fear among market participants of a low-inflation trap. Since mid-March, oil prices have somewhat stabilised and the Pandemic Emergency Purchase Programme (PEPP) has provided grounds for markets to foresee more flexibility in the conduct of monetary policy. Specifically, as stated by the ECB, “purchases will be conducted in a flexible manner allowing for fluctuations in the distribution of purchase flows over time, across asset classes and among jurisdictions”. Nonetheless, the current valuations of these long-dated inflation contracts are still in large contrast to January 2015 – just before the introduction of the Asset Purchase Programme (APP). At that point in time, the long end of the inflation swap curve still converged to the ECB’s inflation target.

**Figure 2** The Inflation swap curve has been especially hit in the longer end compared to previous shocks

*Source*:.DNB, Bloomberg Note: Shaded area reflects the ILS curve during the month march 2020, last observation: 23-04-2020.

## Decomposing inflation linked swap rates

Inflation-linked swap rates can be decomposed into an expectation component and an inflation risk premium. In order to read inflation expectations, we therefore need to filter out this inflation risk premium (IRP) from the price of the swap contracts. As neither the inflation expectations component nor the inflation risk premium is directly observable, they need to be estimated. We follow Camba-Mendez and Werner (2017), who propose an affine term structure model for deriving the inflation-linked swap curve.^{1}

Figure 3 shows the result of the decomposition of the 5y5y ILS into its expectation component and the inflation risk premium. The latter turned negative around the end of 2015 and has remained negative since then.^{2} Intuitively, the inflation risk premium can become negative when deflation risk rises. In case of deflation, a claim on a nominal asset becomes more valuable than a claim on an inflation protected asset. Therefore, investors in a nominal asset are willing to accept a lower, or even a negative, inflation risk premium (Campbell et al. 2017).

As a result of the COVID-19 outbreak, both inflation expectations and the IRP declined, which led to a sharp fall in the 5y5y ILS. Figure 4 splits the cumulated changes in the 5y5y ILS since April 2019 in the expectation component and the inflation risk premium. Two observations are noteworthy:

- First, the model attributes most of the decline – around 60% – to a decline in the IRP.
- Second, the more slowly moving inflation expectations component also fell strongly, from 1.6% in January 2020 to 1.3% in March 2020. Such a sharp decline has not been seen since the Global Financial Crisis.

**Figure 3** Decomposition of the 5y5y in an expectation component (EC) and the inflation risk premium (IRP)

*Source*: Authors calculations. Note: The decomposition of the 5y5y ILS is based on estimated model that follows the methodology outlined in Camba-Mendez and Werner (2017). Last observation: 24-03-2020.

**Figure 4** Cumulative changes in the 5y5y ILS since April 2019

*Source*: Authors calculations. Note: The decomposition of the 5y5y ILS into an expectation component (EC) and inflation risk premium (IRP) is based on estimated model that follows the methodology outlined in Camba-Mendez and Werner (2017). Last observation: 24-03-2020.

## Assessing model uncertainty

Given that the inflation risk premium is unobserved and must be estimated, it is inevitably subject to model uncertainty. To assess the latter, Figure 5 plots the IRP resulting from six different alternative models as a min-max range together with the IRP resulting from our benchmark model. Although the dynamics seem broadly in line for all these models, large differences exist with respect to the level of IRP and concomitantly, the inflation expectations component. The fact that our baseline model by and large reflects the lower bound of the range in Figure 5 suggests that in other models the inflation expectations component fell more strongly. While the baseline model estimates the expectations component to stand at 1.33% in March 2020, other models see it as low as 0.46%. This, however, implies an even wider gap between survey-based inflation measures and market-based measures. For instance, according to the ECB’s Survey of Professional Forecasters, expected long-term euro area inflation stood at 1.7% in the first quarter of 2020 (which we interpret as close to, but below 2%).

**Figure 5** Model range of IRP estimates

*Source*: Authors calculations. Note: The dark-blue line represents our benchmark model which follows the methodology outlined in Camba-Mendez and Werner (2017). The min-max range comes collects other IRP estimates from a setof standard term-structure models, which include variants of a Dynamic Nelson Siegel Model, the Adrian, Crump and Moench (2013) model and an arbitrage-free model from Nyholm (2018), all to which we applied the ILS curve. Last observation: 24-03-2020.

## The distribution of risks surrounding lower inflation expectations is narrower than five years ago

Complementary to the estimate of inflation expectations from inflation-linked swaps, inflation options can give an indication of the distribution of risk surrounding market-based inflation expectations. By deriving the option-implied probability distribution from inflation options with different exercise levels, we can calculate the probability investors give to a certain inflation outcome (Breeden and Litzenberger 1978 Smith 2012).

This is reflected in Figure 6, which shows that the distribution surrounding inflation expectations has shifted to the left since January 2020. Moreover, the distribution surrounding inflation expectations is narrower compared to January 2015, which implies investors are relatively certain this low-inflation environment will last for the coming five years. This finding is in line with the IRP found earlier, as this reflects the low compensation that investors demand for inflation uncertainty.

**Figure 6** Option-implied distribution for 5-year inflation options

*Source*: Authors calculations. Note: Last observation 23-04-2020.

## From an inflation bet to a deflation hedge

The decline in market-based inflation expectations has both a structural component (because the market expectations have been trending downward since 2015) and a temporal component (because COVID-19 has aggravated the decline). The decline in market-based inflation expectations suggests that market participants have adjusted their expectations to a low-inflation environment. They gauge that the inflation risk is more to the downside (deflation) than to the upside (high inflation). For investors, this implies that nominal bonds are no longer an ‘inflation bet’ but rather a ‘deflation hedge’.

*Authors’ note: The views presented are those of the authors and do not reflect the official position of De Nederlandsche Bank or the Eurosystem.*

## References

Adrian, T, R K Crump and E Moench (2013), “Pricing the term structure with linear regressions”, *Journal of Financial Economics* 110(1): 110-138.

Breeden, D T and R H Litzenberger (1978), “Prices of state-contingent claims implicit in option prices”, *Journal of Business*, 621-651.

Camba-Mendez, G and T Werner (2017), “The inflation risk premium in the post-Lehman period”, ECB Working Paper 2033.

Campbell, J Y, A Sunderam and L M Viceira (2017), “Inflation bets or deflation hedges? The changing risks of nominal bonds”, *Critical Finance Review* 6: 263-301.

Nyholm, K (2018), “A flexible short-rate based four factor arbitrage-free term structure model with an explicit monetary policy rule”, unpublished manuscript.

Smith, T (2012), “Option-implied probability distributions for future inflation”, *Bank of England Quarterly Bulletin*, Q3.

## Endnotes

^{1} We estimate the model on the sample April 2005 until December 2014 and additionally calibrate the factor means to their average between April 2005 and December 2009.

^{2} A negative sign can occur if there is a shift of the balance in the risk of inflation. For instance, if market participants see the risk of stagflation, the IRP is positive as investors like to hedge for high inflation that will ‘eat up’ real returns in the economic downturn. In contrast, the expectation of a deflationary recession could lead to a negative IRP as nominal bonds perform well in case of deflation (Camba-Mendez and Werner 2017).