Inflation expectations wield great influence over fixed income returns. They determine the nominal yield required for a given equilibrium real interest rate, they influence inflation risk premia, and they shape the central bank’s course of action. There is no uniform inflation expectation metric than can be tracked in real-time. However, there are useful and complementary proxies, such as market-based breakeven inflation and economic data-based estimates. For trading strategies, these two can be combined. The advantage of breakeven rates is the real-time tracking of a broad range of influences. The advantages of economic data-based estimates are clarity, transparency, and precision of measurement. Changes in both inflation metrics help predict interest rate swap returns, but their combination is a better predictor than the individual series, emphasizing the complementarity of market and economic data.

The below post is based on proprietary research of Macrosynergy.

Some links for data access and documentation refer to the J.P. Morgan Markets website and require user credentials. Yet, the links are not necessary for understanding the content of the post.

This post ties in with this site’s summary on information efficiency.

### Breakeven inflation as a basis for expectations

Breakeven inflation rates are future inflation rates __implied in the prices of inflation-linked contracts__. There are two common types of such contracts.

**Inflation-linked bonds**are fixed interest securities whose payoff is indexed to a consumer price index. Breakeven inflation rates are calculated as the differences between nominal bond yields and real yields for inflation-linked bonds of roughly the same maturity.**Inflation swaps**are derivative contracts under which one party pays a fixed rate cash flow and the other pays a floating rate based on a consumer price index. The breakeven rate (or swap rate) for an inflation swap is the rate at which the fixed rate is equivalent to the floating rate making the swap price zero.

Breakeven inflation rates are a valuable but imperfect basis for estimating inflation expectations. On the positive side, they are __available in real-time and incorporate a broad range of inflation-relevant factors__. However, they are __also affected by market risk premia and the state of liquidity__ of the market. Historically, inflation risk has mostly commanded a positive risk premium, but after the great financial crisis in the 2010s and in times of deflation fears, premia have turned negative (view post here). Risk- and liquidity adjustment of breakeven inflation rates can enhance information value but is subject to interpretation and would require a separate post to explain.

In this post, we use prices of J.P. Morgan and breakeven rates as published in a reasonably clean format via the J.P. Morgan Macrosynergy Quantamental System (JPMaQS ). Breakeven inflation rates are often priced for 1, 2 or 5 years ahead. Meaningful bond-based breakeven rates are available for the following countries: Australia, Brazil, Chile, Colombia, Israel, Japan. Korea, New Zealand, South Africa, Sweden, Turkey, the United Kingdom and the United States (as well as some euro area countries). However, only the U.S. and the UK have more than 10 years of meaningful data. Swaps-based breakeven inflation rates have been priced for 10 years or more for five developed markets: Australia, the euro area, Japan, the UK and U.S.

The below panel shows 5-years ahead breakeven inflation rates for all countries for which these are available either based on bonds and swaps, as well as concurrent official or implied inflation targets. There are two important observations. First bonds- and swap-based inflation expectations can display notable and sustained differences, emphasizing the influence of idiosyncratic features to the respective markets. Second, longer-dated breakeven rates can deviate significantly from inflation targets, making them a valuable indicator for monetary policy.

### An economic data-based (formulaic) approach to inflation expectations

A complementary approach to market-implied breakeven rates is a formulaic estimate. The idea is to use publicly available information on inflation and combine them in a theoretically consistent and plausible way to proxy expectations, with or without parameter estimation. Indeed, many practitioners simply use the latest inflation as a proxy for future expectations. However, this seems unnecessarily restrictive and simplistic.

Here we use the Macrosynergy approach that has been applied for more than 10 years across developed and emerging countries. Again the data have been taken from JPMaQS in order to secure that CPI data are point in time and correctly reflect the historic information state of the market. The estimate __assumes that market participants form their inflation expectations based on the recent inflation rate (adjusted for jumps and outliers) and the effective inflation target__. For the 5-year forward horizon, the weight of recent inflation to the effective target is 1/5 to 4/5. The recent inflation rate is the average of the headline and core annual CPI inflation rate, with a 50% adjustment for jumps and outliers. The effective inflation target is the mean of the target range announced or implied by the authorities plus an adjusted for past “target misses”, which is the last 3 years’ average gap between actual inflation and the target means. Thus, this estimate of inflation expectations changes in response to three data-based changes:

- an update of recent headline or core CPI rates
- an update of the past three years’ performance of the targeted inflation rate relative to the inflation target
- a change in the inflation target itself.

Like breakeven inflation rates, formulaic expectations of the above type are a useful but imperfect basis for inflation expectation estimates. On the positive side, the formula is __completely transparent, makes sensible use of past inflation records, and describes a plausible way in which the public tends to form expectations __by remembering published inflation rates and official target commitments. However, the formula is __very limited in the variables it considers__, disregarding information of other prices, such as for commodities, other economic data, such as operating rates), exchange rates and other relevant news. These expectations are representative for the economy as a whole, as opposed to the financial market alone.

Data are available for 30 currency areas on JPMaQS. The below shows the formulaic expectations in comparison to the breakeven rates. __The formulaic estimates display much fewer and milder high-frequency changes__. However, they also feature sudden jumps and drops when the official inflation target is changed, as has been conspicuous in Japan in 2013 and Korea in 2016. In the U.S., the only country for which there is breakeven and formulaic history since 2000, the bonds-based-breakeven rates have fluctuated around the formulaic expectations, reinforcing the suggestion that the former is suitable for tracking shorter-term changes and the latter for longer-term adjustments.

### Two types of inflation expectations and IRS returns

Most economists would agree that __rising inflation expectations cause – all other things equal – rising high-grade fixed income yields and tend, by themselves, to lead to negative fixed receiver returns__. Rising inflation expectations are a headwind for rates receivers or long-duration positions for three reasons:

__Higher inflation expectations require higher nominal yields to establish a given level of real (inflation-adjusted) interest rates__. In theory macroeconomic equilibria are established through real, not nominal interest rates. Real rates govern intertemporal substitution, real financing costs, real asset values, and so forth. This means that if all other conditions for an equilibrium are unchanged a 1% higher inflation expectation translates into a 1% higher nominal yield.__Higher inflation expectations also increase inflation risk premia__. This reflects two simple effects. First, higher inflation rates lead to higher inflation variance versus expectations. Second, high inflation often coincides with positive directional premia for inflation risk, as an unexpected increase in inflation pushes up real interest rates and compresses equity and other asset values. By contrast, countries in deflation, with policy rates near the zero lower bound, can command negative premia for inflation since upside surprises reduce real interest rates and support equity and other asset markets.__Higher inflation expectations mostly require central banks to tighten financial conditions__, in particular, if expectations rise above official inflation targets. A more hawkish tone or actual monetary tightening often translates into rising real interest rates.

In frictionless markets, with zero information costs, the relation between inflation expectations changes and interest rate swap returns should be strictly concurrent. However__, in the presence of trading costs and macro information costs as well as large pools of passive or trend-based macro investments is more plausible that markets do not continuously, immediately, and fully account for inflation expectations__. Hence, we hypothesize that changes in such expectations over the past month will predict interest rate swap returns over the next month.

In the empirical analysis, we focus on returns for longer-dated (5-year) swaps in the two largest markets, the U.S. and the euro area, because these are plausibly the only fixed income markets that are mostly affected by their own inflation expectations. Smaller countries’ fixed income returns depend to a large extent on developments in the dollar and – in Europe – the euro market. The targets are 5-year fixed receiver returns with a constant annualized volatility target of 10%, mimicking simple risk management and avoiding correlations to be dominant by more volatile periods or markets.

Also, we calculate changes in both types of inflation expectation metrics relative to the local inflation target, as changes over the past 21 trading days (approximately a trading month) in three-month moving averages (to mitigate the effects of outliers in the swap markets). The below panel shows that __expectations changes deliver a shorter-term tactical trading signal__. Also, swap breakeven inflation changes are much larger and more variable than those of formulaic expectations and – on many occasions – very different in pattern. This reflects that breakeven inflation expectations respond to all sorts of high-frequency information, including those that have more to do with the market structure than the inflation outlook. Changes in formulaic expectations only occur if additional actual CPI data have been released or the inflation target has changed.

Based on the overall panel for U.S. and euro area data from 2005 to 2022 __monthly correlation between 5-year breakeven inflation changes and subsequent returns has been negative, as expected, and significantly so__. Also, accuracy, ratio of correctly predicted return direction, and balanced accuracy, the average ratio of correctly predicted positive and negative return directions of monthly returns, have both been a respectable 53.2%.

Formulaic expectation changes posted a slightly stronger and more significant relationship with subsequent 5-year IRS receiver returns. Also, accuracy and balanced accuracy of predicting return directions has been a bit higher at 53.6% and 53.7% respectively.

### Composite expectation changes and IRS returns

The changes in breakeven inflation and formulaic inflation expectations have very different characteristics and complementary information value. Therefore, we would expect that a trading signal that combines both can generate more value than the two metrics individually.

Empirically, the commonality of the two types of inflation change signals has been limited. This is no surprise as breakeven inflations move in accordance with a large range of factors, while formulaic changes mainly respond to CPI releases and their impact on near-term inflation expectations as well as on the credibility of the central bank. The correlation between the two types of changes has been positive, but with a correlation coefficient of no more than 15%.

The simplest way of combining the two types of expectations is to normalize them and add them up. Here, we calculate z-scores around the neutral levels based on expanding windows of standard deviation estimates for both (to avoid look-ahead bias in scaling), contain both at a maximum of 2.5 standard deviations (to contain outliers) and then simply sum them. The resulting composite score is noticeable different from the individual score.

The negative correlation of the composite score with subsequent 5-year IRS returns on a monthly frequency has been near 19% and the probability of significance near 100%, both stronger than the individual correlation statistics.

The combination of the two metrics also results in higher monthly return prediction accuracy and balanced accuracy of 56.8% and 56.9% respectively for 2005-2022, pointing to a signal that would produce significant trading value even on its own. The combined score would have been right in 59.2% of its positive monthly return predictions and 54.5% of its negative monthly return predictions, suggesting that the signal is valuable in states of both rising and falling yields.