Rising inflation is a natural headwind for equity markets in economies with inflation-targeting central banks. As consumer prices accelerate, expected monetary policy rates and discount factors tend to increase more than dividend growth. Over the past three and a half decades, there has been a strong negative correlation between changes in reported inflation and simultaneous global equity futures returns. A similar negative relationship is evident between seasonally adjusted CPI trends and concurrent returns.
Furthermore, inflation dynamics have shown predictive power. Short-term changes and trends in consumer price growth have proven to be valuable early warning signals for serious market downturns and leading indicators of recoveries. Also, inflation-sensitive strategies have performed on par with long-only portfolios during stable periods. Overall, they enhanced risk-adjusted returns by improving the timing of equity risk exposure.
The post below is based on Macrosynergy’s proprietary research. It extends and supersedes a previous post on inflation as an equity trading signal.
Please quote as “Sueppel, Elena and Sueppel, Ralph, ‘Inflation and equity markets,’ Macrosynergy research post, September 2024.”
A Jupyter notebook that allows audit and replication of the research results can be downloaded here. The notebook’s operation requires access to J.P. Morgan DataQuery to download data from JPMaQS, a premium service of quantamental indicators. J.P. Morgan offers free trials for institutional clients.
Also, an academic research support program sponsors data sets for relevant projects.
This post ties in with this site’s summary of macro trends and systematic strategies.
Theory and evidence
Academic research broadly supports a negative relationship between inflation and equity market returns in modern developed economies. The relation arises from the positive effect of current inflation on future expected inflation and two types of negative consequences of higher expected inflation on equity valuation.
- The real discount factor effect: In modern inflation-targeting regimes, there is a positive relation between expected inflation and real interest rates. Higher inflation calls for tighter monetary policies and higher risk premia for holding long-maturity nominal debt. Higher real interest rates, in turn, increase the real discount rate for future dividends. Put differently, the ratio of nominal dividend growth to nominal discount rate decreases, reducing the present market value of future cash flows. This relation depends on the central bank’s commitment to fighting high inflation and deflation. The IMF paper of Zhang (2022) concludes that “Classical theories of monetary economics predict that real stock returns are negatively correlated with inflation when monetary policy is countercyclical… If a country’s monetary authority conducts a more countercyclical monetary policy, the stock return-inflation relation becomes more negative.”
- Sticky cash flow expectations: Rising CPI inflation expectations typically fail to push up predictions of dividend growth proportionally. As described by Katz, Lustig, and Nielsen (2016): “The nominal risk-free rate is more sensitive than the cash flows to changes in expected inflation. This ‘discounting effect’ arises from the stickiness of cash flows… Price stickiness implies nominal cash flow growth is sticky in the short run [due maybe to longer-term contracts], and so expected nominal cash-flow growth changes less than one-for-one with changes in expected inflation.” The effect is confirmed in Knox and Timmer (2024), who conclude that “in response to higher-than-expected inflation, investors expect firms’ nominal cash flows to remain stagnant while discount rates increase, resulting in lower stock prices.”
Empirical results echo these effects. Jung and Pyun (2022), for example, document “robust empirical evidence [of] a significant … negative relationship between inflation and equity prices… in the post-WWII U.S. data.”. This does not yet mean that inflation data are also useful predictors of equity returns. After all, CPI inflation rates are released timely and widely followed by market participants. However, less conspicuous indicators, such as short-term changes in global inflation or short-term seasonally adjusted trends in global CPI, may not always be fully priced due to rational inattention (view post here).
Macro-quantamental data for an empirical analysis
This post focuses on the short-term dynamics of CPI growth for a set of 17 global economies that feature liquidly traded equity index futures and make up about two-thirds of global GDP. The countries and local-currency index futures used for this post are in alphabetical order of the currency symbol:
- Australia (AUD): Standard and Poor’s / Australian Stock Exchange 200
- Brazil (BRL): Brazil Bovespa
- Canada (CAD): Standard and Poor’s / Toronto Stock Exchange 60 Index
- Switzerland (CHF): Swiss Market (SMI)
- Euro area (EUR): EURO STOXX 50
- UK (GBP): FTSE 100
- India (INR): CNX Nifty (50)
- Japan (JPY): Nikkei 225 Stock Average
- South Korea (KRW): Korea Stock Exchange KOSPI 20
- Mexico (MXN): Mexico IPC (Bolsa)
- Malaysia (MYR): FTSE Bursa Malaysia KLCI
- Sweden (SEK): OMX Stockholm 30 (OMXS30)
- Singapore (SGD): MSCI Singapore (Free)
- Thailand (THB): Bangkok S.E.T. 50
- Taiwan (TWD): Taiwan Stock Exchange Weighed TAIEX
- U.S. (USD: Standard and Poor’s 500 Composite
- South Africa (ZAR): FTSE / JSE Top 40
To validate predictive power in a meaningful way, we use point-in-time information states of the J.P. Morgan Macrosynergy Quantamental System (JPMaQS) for two types of categories:
- Change in the most widely watched CPI and core CPI inflation (% over a year ago) measure of the currency area over the past three months (view documentation here and here).
- The short-term excess growth trend in the most widely watched CPI and core CPI indicators, measured as % change of the past three months over the previous three months, seasonally- and jump-adjusted, at an annualized rate (view documentation here and here). Excess here means over and above an effective or estimated inflation target (view documentation here).
For some countries, we use early estimates of official monthly CPI data in accordance with market conventions. These are the UK, Japan, Singapore, the U.S. (for core PCE), and South Africa.
As a matter of principle, the quantamental format matches measurements with the time at which they are the latest instances published to financial markets. JPMaQS uses business daily timestamps that conceptually refer to the end of the New York trading day. These are called real-time dates. For any given real-time date, an indicator is calculated based on the full information state, typically by one or more time series. This allows realistic backtesting. Information state-contingent time series are called vintages. This data format implies that a transformation (such as % change) of a quantamental indicator is different from a quantamental indicator of a transformation. The former operates on the first dimension (real-time dates), while the latter operates on the second dimension (observation dates). CPI inflation changes here refer to the latest reported changes in inflation.
Information states of most headline inflation metrics are available back to 1990, with Brazil and the euro area being the only exceptions. Information states of core inflation available to the early 1990s for most countries, but some have started late publishing an official core CPI metric.
To derive meaningful information states, back to 1990 for most countries, we use a technique that could be called “hierarchical merging”. This means that if a quantamental indicator is not available before a certain historical date, we use the nearest substitute, i.e., the indicator that would have been used by markets instead or that is the best estimator for the unavailable data. Thus, for headline inflation, we use the “early estimates” series as far back as available and the regular releases before that date. For core inflation, we use, in hierarchical order, early estimates of core CPI, regular core CPI, early estimates of CPI, and regular CPI. Moreover, since most countries do not have official or estimated inflation targets back to 1990, we “backfill” missing early years with the first available inflation target. This implies greater inaccuracies of excess inflation estimates in the 1990s.
The panel below shows the information states of three-month headline and core inflation changes. They post ample fluctuations, with headline CPI inflation changes typically posting greater amplitudes. By the standards of macroeconomic indicators, they are volatile. Related trading signals would flip their sign several times a year.
The characteristics of short-term excess CPI trends are similar to those of inflation changes. In the below panel, we plotted timelines of the 3-month over 3-month changes for headline CPIs together with standard annual inflation rates to illustrate the excess volatility. It is this volatility that discourages the broader public and many market participants from regularly monitoring short-term CPI trends. Most fluctuations are inconsequential. However, in times of shifts in the inflation environment, they pick up the new trend a bit earlier than annual rates.
Since equity markets are internationally connected and correlated, this post focuses on global aggregates of short-term CPI dynamics. We calculate two weighted baskets of information states: one for all 17 economies (GLB) and one excluding the United States (GLX). The basket weights are concurrent information states of nominal GDPs in USD, meaning the U.S. has the biggest weight, about 40%.
The charts below show information states of short-term excess headline inflation for the global baskets, together with the standard annual excess inflation rates for illustration. These short-term inflation trends are volatile but often indicative of the subsequent direction of annual inflation rates.
In the sections below, we analyze the relationship between global short-term inflation changes and CPI trends on the one hand and returns on global baskets of local-currency equity index futures on the other. For the equity futures, “global” means an equally weighted basket of tradable and recorded countries for any date. Equal weights are the best way to reduce the influence of other country-specific factors and focus the analysis on global inflation dynamics. Again, we create one basket for all markets and one without the U.S.
At each date in the past, the basket uses all tradable futures, for which JPMaQs provides generic return data (view documentation here). Indeed, the basket started with only four countries in the early 1990s, expanded to 12 countries by the end of the 1990s, and reached its fully diversified state of 17 countries with the inclusion of Thailand in 2006.
The below charts of cumulative basket returns (not including compounding effects) show consistent positive drifts over the past 35 years, with recurrent episodes of setbacks. The strong trend plausibly reflects favorable macro reforms, de-regulation, and disinflation or low inflation of the past decades.
Global inflation changes and equity future returns
Concurrent relations
Following the academic research, we expect years of rising inflation to be more negative for equity markets than years of falling inflation. To check this, we plot average annual information states of 3-month changes in global inflation against concurrent annual basket returns since 1990 and run a simple linear regression.
Indeed, the relationship between reported inflation changes and global futures returns has been negative, with a Pearson correlation coefficient of 35% and near 100% significance. Six of the seven worst years for global equity futures (2008, 2000, 1990, 2011, 2018, and 2022) were years of rising inflation.
Predictive relations
Next, we test whether information states of global inflation changes also had predictive power for subsequent global futures returns. For this purpose, we look at the relationship between end-of-month information on inflation changes and the next month’s future returns.
Indeed, the predictive relation for both core and headline inflation has been negative and highly significant. The negative relation has been even stronger when excluding the U.S., which supports the view that U.S. data watching is more efficient than in other markets.
The negative predictive relation has not been concentrated on a specific decade. It was highly significant for both the first and second halves of the sample.
The statistics confirm the value of inflation changes as a global equity performance predictor. Non-parametric forward correlation has also been negative, with a 99% probability of significance. Monthly accuracy, the ratio of correctly predicted equity market direction, has been 55% for headline inflation changes and 57% for core inflation changes. Balanced accuracy ratios, i.e., average correctly predicted negative and positive market returns, have also been 55-56%.
Finally, we illustrate the value of inflation changes for equity market timing through their application to a naïve PnL with weekly rebalancing, following standard rules of Macrosynergy posts. To this end, we normalize monthly inflation changes on a rolling basis, dividing by past standard deviations and winsorizing (i.e., capping and flooring) the resulting scores at three standard deviations to contain potential extreme outliers. We also add 0.5 standard deviations to the signal to get a long bias that makes the strategy comparable to a long-only portfolio. Then, we use the signal at the end of each week to take a commensurate position at the beginning of the next week, allowing for one-day slippage for trading. The resulting PnL is naïve because it disregards transaction costs, which depend on position sizes and risk management rules. Moreover, for graphical representation, we show timelines of the PnLs that scaled to 10% annualized volatility.
The chart below compares the value generation of a long-only portfolio of the (up to) 17 equity index futures with one that has been managed based on inflation changes. The important result is that inflation change signals have mitigated or avoided all major drawdowns of the equity markets since 1990 without a significant performance loss in other periods. Simply put, inflation changes have been a most useful warning signal in the inflation targeting era. The inflation change signal has also been useful in predicting significant post-crash recoveries.
In the long run, the Sharpe ratio of the inflation-managed strategies for the global baskets has been around 0.8, versus less than 0.6 for the long-only portfolio. The Sortino ratio would have been lifted even more from 0.8 to 1.1-1.3. That difference is notable for a single standard factor. The advantage of inflation change-based signals for the basket that excludes the U.S. has been even greater, with Sharpe ratios reaching 0.9 and the Sortino ratios 1.2-1.3.
Global excess inflation and equity futures returns
Concurrent relations
As for inflation changes, there has been a highly significant concurrent relationship between annual averages of global excess inflation and concurrent annual returns on a global basket of local-currency equity index futures. The correlation has been negative, with a Pearson coefficient of over 40%.
Predictive relations
Predictive relations at a monthly frequency have likewise been negative and significant. However, they have been less strong than for inflation changes. The relationship was weakest for excess core inflation in non-U.S. markets.
Predictive monthly accuracy and balanced accuracy have been similar to that of inflation changes. However, naïve PnL generation has been a bit softer, and the outperformance of risk-adjusted long-only returns only really kicked in the 2010s as a gradual albeit consistent effect. This reflects the problem of data quality and roughness of inflation target estimates in the 1990s. However, it may also reflect that CPI trends are a little slower in providing information on changes in inflation than annual inflation changes.
Altogether, the consideration of global short-term inflation trends would have lifted the Sharpe ratio of an equity futures portfolio from under 0.6 to roughly 0.7. The Sortino ratio would have increased from 0.8 to around 1.
The improvement of the Sharpe ratio would have been larger for a basket excluding the U.S., lifting the long-term Sharpe ratio to 0.8 and the Sortino to 1.2, possibly reflecting the lower information efficiency of non-U.S. markets.
