Home » Research Blog » Macroeconomic cycles and asset class returns

Macroeconomic cycles and asset class returns

Jupyter Notebook

Indicators of growth and inflation cycles are plausible and successful predictors of asset class returns. For proof of concept, we propose a single balanced “cyclical strength score” based on point-in-time quantamental indicators of excess GDP growth, labor market tightening, and excess inflation. It has clear theoretical implications for all major asset markets, as rising operating rates and consumer price pressure raise real discount factors. Empirically, the cyclical strength score has displayed significant predictive power for equity, FX, and fixed income returns, as well as relative asset class positions. The direction of relationships has been in accordance with standard economic theory. Predictive power can be explained by rational inattention. Naïve PnLs based on cyclical strength scores have each produced long-term Sharpe ratios between 0.4 and 1 with little correlation with risk benchmarks. This suggests that a single indicator of cyclical economic strength can be the basis of a diversified portfolio.

The below post is based on proprietary research of Macrosynergy.

A Jupyter notebook for audit and replication of the research results can be downloaded here. The notebook 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.

This post ties in with this site’s summary of macro information inefficiency in financial markets.

Also, there is an academic research support program that sponsors data sets for relevant projects.

Defining a tradable concept of cyclical strength

Here cyclical economic strength refers to a state in which the economy is growing at a faster pace than it could sustain in the long run. This often occurs when short-term aggregate demand is exceeding the growth in productive capacity. It implies that operating rates increasing and that, all other things equal, the prospects for labor costs and prices are shifting upward. Such cyclical strength typically calls for monetary policy tightening, as downside risks for growth and inflation are diminishing or upside risks are building.

There are other definitions of cyclical strength in economics, but this one is particularly useful for building macro trading strategies for two reasons:

  • First, there is a range of economic indicators to track cyclical strength. For the below analysis, we focus on the three most popular ones: the estimated rate of GDP growth versus a “normal rate” (“excess GDP growth”), growing utilization of the labor force (“labor market tightening”), above-target consumer price growth (“excess inflation”).
  • Second, this definition of cyclical strength has clear directional implications for major asset markets. In a standard macroeconomic policy environment, growing operating rates and inflation raise real discount factors in the near term. Unexpected cyclical strengthening should reduce the value of equity and duration positions and increase the value of the local currency. Moreover, in the presence of some degree of rational inattention (view post here) with respect to economic trends, cyclical strength scores should have predictive power for subsequent asset class returns.

Specifying cyclical macro quantamental features

For a meaningful analysis of the impact of economic trends on market returns, we use indicators of the J.P. Morgan Macrosynergy Quantamental System (“JPMaQS”). Quantamental indicators are real-time information states of the market and the public with respect to an economic concept and, hence, are suitable for testing relations with subsequent returns and backtesting related trading strategies. For this post, we use quantamental indicators to replicate the market’s information state with respect to three aspects of cyclical strength:

  • Excess estimated GDP growth trend: For each day this is the latest estimated GDP growth trend (% over a year ago, 3-month moving average) minus a 5-year median of that country’s actual GDP growth rate. The historic median represents the growth rate that businesses and markets have grown used to. The GDP growth trend is estimated based on actual national accounts and monthly activity data, based on sets of regressions that replicate conventional charting methods in markets (view full documentation here). For subsequent aggregation and analysis, we then z-score the indicator (normalize volatility) around its zero value on an expanding out-of-sample basis using all cross sections for estimating the standard deviations.

  • Labor market tightening: This is a composite of three quantamental indicators that are jointly tracking the usage of the economy’s labor force. The first is employment growth relative to workforce growth, where the former is measured in % over a year ago and 3-month average and the latter is an estimate based on the latest available 5 years of workforce growth (view documentation here). The second sub-indicator measures changes in the unemployment rate over a year ago and over the last three months, both as a 3-month moving average (view documentation here). The third labor market indicator is the level of the unemployment rate versus a 10-year moving median, again as a 3-month moving average(view documentation here). All three indicators are z-scored, then combined with equal weights, and then the combination is again z-scored for subsequent analysis and aggregation.

  • Excess inflation: This is a composite of four different types of consumer inflation trends, namely headline and core CPI as both percent over a year ago rate and percent of the last 6 months over the previous six months seasonally and jump-adjusted and annualized (view documentation here). From all of these rates we subtract the effective inflation target of the central bank (view documentation here) and divide by the higher of the target or 2, to make target deviations in low- and high-inflation countries comparable. As with the other cyclical strength indicator the composite has been z-scored around its zero value on an expanding out-of-sample basis.

The composite cyclical strength score is an equally weighted average of the above three z-scored aspects of that strength. Whenever not all components of the scores are available an average of the remaining z-scores is formed. Moreover, we calculate relative z-scores for each currency area, except for the U.S. and the euro area. These relative scores are needed as signals for FX-related strategies. They are calculated by subtracting from the score of the reference country the score of the currency area that is its natural benchmark, which is mostly the U.S. but for most European countries it is the euro area, and for the UK, Turkey, and Russia, it is a basket of USD and EUR.

Cyclical strength and asset class returns since 2000

Cyclical strength and equity index future returns

We can investigate the relationship between the cyclical strength of the economy and subsequent equity index futures returns for 18 developed and emerging economies’ currency areas (AUD, BRL, CAD, CHF, EUR, GBP, JPY, KRW, MXN, MYR, PLN, SEK, SGD, THB, TRY, TWD, USD, ZAR) since 2000. For reference of the currency symbols see the annex at the end of the post. Generic futures return data are taken from JPMaQS and for cross-country comparability we use volatility-target versions, i.e. returns for positions that have an expected annualized 10% return volatility (view documentation here).

On balance, mainstream economic theory suggests that cyclical strength, as opposed to sustained inflation-free growth, should be negatively related to local equity index futures returns. This is because high operating rates and inflation require tighter monetary policy and increase labor cost pressure. In conjunction with rational inattention of parts of the market, related point-in-time information should have predictive power.

Indeed, there has been a highly significant negative relation between the broad composite cyclical strength indicator and subsequent future returns across the panel. The probability of this relationship being not a chance event in the sample is more than 98% at a monthly frequency and over 99% probability at a quarterly frequency based on the Macrosynergy panel test. For details on the test methodology, one can view a post here. Similarly, Pearson and Kendall correlation coefficients of the pooled dataset have been negative, with near 100% significance.

Monthly balanced accuracy, the average ratio of correct prediction of positive and negative returns based on the (negative of) cyclical strength has been 52.6%. The cyclical strength signal’s positive return predictions have been correct in 61% of all months, and its negative predictions in 43% of all months. Predictive power has been quite seasonal, with positive correlation and above-50% accuracy in 54% and 58% of all years respectively.

Across cycle indicator components, excess inflation posted higher predictive accuracy and return correlation than excess growth and labor market tightening, but all three showed above 51% accuracy and a highly significant (negative) correlation with subsequent equity returns.

We calculate naïve PnLs based on standard rules, used in our previous posts.

  1. Positions are taken based on cyclical strength scores, in units of vol-targeted positions.
  2. The z-scores are winsorized at 3 standard deviations to reduce the impact of data outliers.
  3. Positions are rebalanced monthly with a one-day slippage for trading.
  4. The long-term volatility of the PnL for positions across all currency areas has been set to 10% annualized.

It is important to note that this PnL calculation method does not consider transaction costs or realistic risk management rules.

Naïve PnL generation has been positive, but highly seasonal, naturally producing most value around times of recessions and recoveries. The long-term Sharpe ratio of the naïve PnL based on the cyclical strength score has been 0.75 from 2000 to 2023 (April). PnL generation has been largely uncorrelated with returns of long-only equity benchmarks (such as the S&P 500) and, hence, has been additive to standard long equity market exposure.

Cyclical strength and FX forward returns

We can investigate the relationship between the cyclical strength of the economy and subsequent 1-month FX forward returns for 27 “smaller” currency areas (‘AUD, BRL, CAD, CHF, CLP, COP, CZK, GBP, HUF, ILS, JPY, KRW, MXN, MYR, NOK, NZD, PEN, PHP, PLN, RON, RUB, SEK, SGD, THB, TRY, TWD, ZAR’) since 2000. FX forward returns data are taken from JPMaQS and for cross-country comparability we use (10%) volatility-target versions (view documentation here). Periods of illiquidity or non-convertibility have been blacklisted. The exchange rates underlying the forward contracts are based on the reference currency against its “dominant benchmark”, which for most cases is the U.S. dollar. However, for some European currencies, the euro is more appropriate and for three currencies (GBP, RUB, and TRY) both the euro and the dollar serve as benchmarks.

The analysis of FX returns focuses on relative cyclical strength, which compares the cyclical strength score of the reference currency against the cyclical strength in the benchmark currency area. According to standard economic reasoning, currency areas with positive relative scores are more likely to experience relative monetary tightening and positive forward returns. Conversely, currency areas with negative relative scores are more likely to experience negative returns, all other things being equal. Since relative cyclical strength across countries is not easy to follow for investors, it is likely that there is some degree of rational inattention and that systematic relative strength scores have some predictive power.

Indeed, empirical evidence since 2000 shows a highly significant positive relation between relative cyclical strength scores and subsequent monthly or quarterly FX returns. The Macrosynergy panel test, Pearson correlation, and Kendall correlation all show a positive relation with almost 100% statistical probability that this did not occur by chance.

Monthly balanced accuracy has been 53.3%, with the ratio of correct positive return predictions at 58% and the ratio of correct negative predictions at 48%. Positive prediction results have been less seasonal than in the case of equity index futures returns, with above-50% balanced accuracy positive correlation in three-quarters of all sample years and over 80% of all currency markets. All three subcomponents of cyclical strength showed highly significant positive correlation and above-50% balanced accuracy. The labor market subcomponent posted the strongest predictive power.

The naïve PnL across all currencies from 2000 to 2023 (April) has been strong and positive most of the time. The long-term naïve Sharpe ratio would have been near 1 since 2000, with just a 7% correlation with the S&P500. The cyclical strength signal would have avoided some drawdowns of a long-only strategy in small countries’ FX. However, it would not have avoided a drawdown during the great financial crisis of 2008.

Cyclical strength and FX versus equity returns

We can analyze the relationship between cyclical strength and relative positions in FX forwards versus local-currency equity index futures for 17 currency areas that have both liquid FX and equity futures markets (‘AUD, BRL, CAD, CHF, EUR, GBP, JPY, KRW, MXN, MYR, PLN, SEK, SGD, THB, TRY, TWD, ZAR’). To focus on the actual relative performance both lags of the trade are vol targeted at 10%.

The economic reasoning of the above sections applies, and we would expect cyclical strength in a currency area to favor FX performance over local-currency equity performance. Also, we test this hypothesis for both outright and relative cyclical strength indicators.

Both outright and relative cyclical strength have displayed a positive and highly significant relationship with subsequent monthly and quarterly FX versus equity returns, by all metrics. The predictive power of the relative cyclical strength signal has been slightly stronger in terms of correlation and significance, but slightly weaker in terms of directional prediction accuracy.

Monthly balanced accuracy has been 52.3% for the outright signal, and 51.3% for the relative signal. Across sub-components, the inflation and labor market scores have displayed on balance a bit higher accuracy and forward return correlation.

PnLs for both outright and relative cyclical strength signals have been economically significant, with a long-term Sharpe ratio of 0.6 and 0.8 respectively. Importantly, correlations of the naïve strategies with the broad equity market have been near zero. This differentiates them from a strategy that is always long equity versus FX. Equity typically carries higher risk premia in the spirit of the capital asset pricing theory and would have outperformed FX positions consistently over the last 10 years. However, the cyclical strength signals would also have generated a strong PnL from 2000 to 2012, when the equity long failed to perform.

Cyclical strength and interest rate swap returns

In principle, we can investigate the relation between cyclical strength and subsequent 5-year interest rate swap returns for 25 developed and emerging economies’ currency areas (‘AUD, CAD, CHF, CLP, COP, CZK, EUR, GBP, HUF, ILS, JPY, KRW, MXN, MYR, NOK, NZD, PLN, RUB, SEK, SGD, THB, TRY, TWD, USD, ZAR’). Not all countries have been available since 2000, however, and some markets have suffered from occasional illiquidity and had to be temporarily blacklisted. Also, for some countries, cross-currency swaps have been used. Generic IRS return data are taken from JPMaQS and for cross-country comparability we use volatility-target versions, i.e. returns for positions that have an expected annualized 10% return volatility (view documentation here).

In principle the general hypothesis is simple: greater cyclical strength calls for higher real and nominal interest rates. In conjunction with some rational inattention in markets, a high strength score bodes for weaker or negative duration returns.

Indeed, the correlation between cyclical strength and subsequent IRS returns has been negative. However, the significance of that relationship across the panel is questionable. True, standard Pearson or Kendall correlation based on pooled data looks highly significant. However, the Macrosynergy panel test, which adjusts for pseudo-replication of observations in countries that experience similar features and target variation, assigns only 74% to the probability that the relationship did not appear by chance. This reflects the correlation of cyclical strength across economies, leading to pseudo-replication of experiences in the panel, and the strong dependence of most countries’ IRS returns on developments in the U.S. and – to a lesser extent – in the euro area.

Indeed, the negative relation is clearer and stronger for the U.S. and the euro area alone, with greater significance, even though it is based on a fraction of about 10% of the global data.

For the whole panel, the monthly balanced accuracy of 5-year IRS return predictions has been 52.2%, with a 58% hit ratio of positive predictions and a 47% hit ratio of negative predictions. However, predictive quality has been highly seasonal with above-50% accuracy in only 62% of all sample years and positive correlation in less than half of the years.

A long-term “naïve” PnL calculation based on the cyclical strength signal alone does not show consistent value generation. Positive returns were only accumulated since 2014, and even during this period, there was a large drawdown during the COVID pandemic. The empirical Sharpe ratio of 0.41 was only accomplished thanks to a broad global payer position of the strategy during the phase of yield increases after the pandemic.

Cyclical strength and curve flattening returns

We define a curve flattening trade as a receiver position in a 5-year interest rate swap and a payer position in a 2-year IRS, with the overall trade being volatility neutral, i.e. both legs having the same expected return standard deviation based on a standard exponential lookback window with an 11-day half-life. This is roughly equivalent to a difference of 5-year and 2-year vol-targeted IRS returns as available on JPMaQS. The sample for this analysis is the same 25 markets that were used for the directional IRS return analysis.

As long as monetary policy regimes are credible, and markets are rationally inattentive there is a strong case for cyclical strength predicting positive flattening returns and cyclical weakness negative flattening returns. That is cause monetary policy typically tightens into cyclical strength, raising near-term policy rate expectations, while controlling long-term inflation and managing a return of the economy to a steady state.

Empirically, there is strong evidence for the positive relation between cyclical strength and subsequent returns on curve-flattening positions since 2000. Significance has been 99.6% or higher at a monthly or quarterly frequency according to the Macrosynergy panel test, the Pearson correlation of the data pool, and the Kendall correlation of the pool.

The monthly balanced accuracy of the cyclical strength composite score with respect to subsequent curve flattening returns has been 52.9%, with 56% correct positive directional predictions and 50% correct negative directional predictions. Predictive power has been quite seasonal, with positive correlation and above 50% accuracy in 58-66% of all sample years. All components of the cyclical strength score displayed predictive power, but the growth sub-score produced the highest correlation and balanced accuracy.

A naïve PnL with curve flattening and steepening positions based on the cyclical strength composite indicator would have produced a high long-term Sharpe ratio of 0.93 with no equity correlation and negative U.S. fixed come return correlation. However, naïve value generation was highly seasonal and concentrated in the 2020s.

Cyclical strength and FX versus IRS returns

We investigate the relation between relative cyclical strength and volatility-weighted positions in local currency FX forwards and 5-year IRS payers (short duration) across 23 currencies (AUD, CAD, CHF, CLP, COP, CZK, GBP, HUF, ILS, JPY, KRW, MXN, MYR, NOK, NZD, PLN, RUB, SEK, SGD, THB, TRY, TWD, ZAR).

Mainstream economic theory suggests that stronger growth and inflationary pressure support local currency strength and fixed-income weakness and – in conjunction with some rational inattention – should help predict the returns on long FX versus short-duration positions.

Indeed, also this hypothesis is backed up by empirical evidence. There has been a highly significant positive relation between relative cyclical strength scores and subsequent FX versus duration returns at both the monthly and quarterly frequency. The statistical probability that this relation has not been due to chance is near 100% according to the Macrosynergy panel test, the pooled Pearson correlation p-value, and the pooled Kendall correlation p-value.

The monthly balanced accuracy of predictions based on the relative cyclical strength score has been 51.9%. Across the sub-scores the labor market tightening score posted the highest balanced accuracy with 52.4%.

A naïve PnL of vol-targeted FX versus duration positions based on the cyclical strength score alone would have produced a respectable long-term Sharpe ratio of 0.63 since 2000 with just 5% correlation with the S&P500. The PnL has been seasonal, albeit not highly concentrated on a single episode. 


Annex: Currency symbols

The mean of the currency symbols is as follows: AUD (Australian dollar), BRL (Brazilian real), CAD (Canadian dollar), CHF (Swiss franc), CLP (Chilean peso), CNY (Chinese yuan renminbi), COP (Colombian peso), CZK (Czech Republic koruna), DEM (German mark), ESP (Spanish peseta), EUR (Euro), FRF (French franc), GBP (British pound), HKD (Hong Kong dollar), HUF (Hungarian forint), IDR (Indonesian rupiah), ITL (Italian lira), JPY (Japanese yen), KRW (Korean won), MXN (Mexican peso), MYR (Malaysian ringgit), NLG (Dutch guilder), NOK (Norwegian krone), NZD (New Zealand dollar), PEN (Peruvian sol), PHP (Phillipine peso), PLN (Polish zloty), RON (Romanian leu), RUB (Russian ruble), SEK (Swedish krona), SGD (Singaporean dollar), THB (Thai baht), TRY (Turkish lira), TWD (Taiwanese dollar), USD (U.S. dollar), ZAR (South African rand).


Related articles