There are plausible relations between past and future short-term trends across and within financial markets. This is because market returns affect expected physical payoffs, risk premia, and the monetary policy outlook. However, the relations between past and future returns are unstable and often depend on the economic environment. As an example, this post shows that the impact of short-term commodity future trends on subsequent S&P500 future returns depends on the inflationary pressure in the U.S. economy. Empirical analysis suggests that macro-conditional trend signals outperform unconditional short-term trend signals regarding predictive power, accuracy and naïve PnL generations.
Please quote as “Sueppel, Ralph, ‘Conditional short-term trend signals,’ Macrosynergy research post, January 2025.”
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. Everyone with DataQuery access can download data, except for the latest months. Moreover, J.P. Morgan offers free trials on the complete dataset for institutional clients. An academic support program sponsors data sets for research projects.
The importance of macro conditions for short-term price signals
Short-term market price or return trends over a few days can predict future returns if the broader implications of market changes are priced sluggishly. There are two reasons for this:
- First, the development in one market segment may affect another through its influence on physical payoffs or risk premia. For example, commodity price returns are related to inflation and real interest rate expectations, affecting the discount factors for bonds and equity.
- Second, price trends within one market segment can incur payback or escalating dynamics, depending on how the broader economy and policymakers respond. For example, equity price increases often raise prospects for economic growth, incomes and corporate finances and, hence, lessen the need for monetary policy support for financial markets.
However, the relation between past and future price trends is typically neither simple nor stable but dependent on market conditions. One important such condition is the state of the economy:
- The effect of one market segment price trend on another often depends on the response of policymakers, governments, corporations and households and that response, in turn, depends on economic conditions. For example, the monetary policy response to commodity price increases depends on the distribution of inflation risks and the proximity of interest rates to the lower bound.
- Similarly, a market return trend within one market segment is more likely to reverse if it triggers economic headwinds and more likely to persist if it triggers economic tailwinds. In the case of equity markets, headwinds to price increases are more likely in an environment of higher inflation, and tailwinds are more likely in the presence of deflation risk, mainly because central banks are more motivated and able to adjust policy rates in the former case.
This connection between slower-moving economic developments and shorter-term price trends illustrates that macroeconomic information supports not just strategies in its own frequency domain but also higher-frequency trading signals.
In this post, we look at the relevance of short-term market price trends and macro conditions for the largest market in the world, namely U.S. equities, as represented by the Standard & Poor’s 500 index. We test two hypotheses:
- Returns in commodity futures markets should negatively predict future U.S. equity returns in an inflationary environment with an inflation-averse central bank since the latter will want to adjust expected real interest rates to new sources of price pressures. However, the predictive relation should turn positive when inflation is below target and deflation fears prevail since policy rates become constrained by the zero lower bound and increasing inflation expectations often reduce real interest rates.
- Returns in the equity market itself should negatively predict future S&P500 returns in an inflationary environment since rising stock prices further fuel economic growth and price pressures, which plausibly triggers a countervailing response by the central bank. No such policy response is required in an environment of deflation risk.
Short-term futures return trends
We calculate short-term trends of commodity and S&P 500 futures excess return indices. The indices are cumulative sums of generic returns without compounding effects. The returns have been taken from the J.P. Morgan Macrosynergy Quantamental System or “JPMaQS” (view documentation). We define the main short-term trend as the difference between a 3-day moving average and a 10-day moving average. For some robustness checks, we also look at the (more volatile) difference between a single-day level and a 5-day moving average. For commodity futures markets, we calculate return trends for three market segments:
- Agriculture and livestock (“food”): This group contains CBOT corn composite (COR), CBOT wheat composite (WHT), CBOT soybeans composite (SOY), NYBOT/ICE cotton #2 (CTN), NYBOT/ICE coffee ‘C’ Arabica (CFE), NYBOT/ICE raw cane sugar #11 (SGR), NYBOT/ICE frozen orange juice concentrate (NJO), CME live cattle composite (CAT), and CME lean hogs composite (HOG).
- Crude oil, derivatives and natural gas (“energy”): The contracts are ICE Brent crude (BRT), NYMEX WTI light crude (WTI), NYMEX natural gas Henry Hub (NGS), NYMEX RBOB Gasoline (GSO), NYMEX Heating oil, and New York Harbor ULSD (HOL).
- Ferrous and non-ferrous metals (“metals”): This futures group includes LME aluminium (ALM), Comex copper (CPR), LME lead (LED), LME nickel (NIC), LME tin (TIN), LME zinc (ZNC), COMEX gold 100 Ounce (GLD), COMEX silver 5000 Ounce (SIV), NYMEX palladium (PAL), and NYMEX platinum (PLT).
We consider six types of potentially predictive return trends based on different baskets. Each contract has equal weight within these baskets, irrespective of volatility and the direct relation of the commodity to consumer price baskets. This means a preference for simplicity rather than optimization, which seems appropriate for delivering a proof of concept.
- The short-term agricultural commodity trend is an equally weighted linear composite of the return indices trends of all the contracts in the agriculture and livestock group.
- The short-term energy commodity price trend is an equally weighted linear composite of the return trends of all contracts in the energy group.
- The short-term metals commodity price trend is an equally weighted linear composite of the return trends of all contracts in the metals group.
- The short-term composite commodity price trend is an equally weighted linear composite of the return trends of the food, energy, and metals groups.
- The short-term equity price trend is simply the S&P500 futures return index trend.
- The aggregate commodities and equity price trend is an equally weighted linear composite of the short-term composite commodity and equity price trend.
All commodity futures trends have been calculated from 1995. If individual contracts are not available for the whole sample period, they are used from their inception date, and composite trends for prior dates are calculated for available contracts. All these short-term trends would call for frequent position changes, flipping from long to short roughly every 2 weeks.
Inflation condition factors and conditional futures trends
We use macro-quantamental indicators to track the inflation environment in the form of point-in-time information states in the U.S. Point-in-time data are a requirement for realistic backtests and are available from JPMaQS. In particular, we calculate four candidate inflation environment scores:
- Excess CPI inflation score: This is an average of four indicators of annualized consumer price inflation pressure relative to the Federal Reserve’s inflation target. The indicators are annual headline CPI inflation (documentation here), annual core PCE inflation (documentation here), annualized, seasonally adjusted and jump-adjusted core PCE growth of the past six months over the previous six months (documentation here), and two-year ahead CPI inflation expectations (documentation here).
For this and the other scores, the indicators are normalized, then averaged, and finally, the composite score is renormalized. Normalization here is always point-in-time and around the zero value. This means we divide by standard deviations of observations up to the end of the last full month before the observed value. To have more observations for normalization, we use a panel of both U.S. and euro area data rather than U.S. data alone. - Excess PPI inflation score: This is an average of four indicators of annualized output price growth in the U.S. economy minus the inflation target. The indicators are the standard annual producer price inflation, % over a year ago, 3-month moving average (documentation here), seasonally adjusted and annualized producer price inflation % 6 months over the previous six months (documentation here), economy-wide output prices, narrow estimate, % over a year ago, 3-month moving average (documentation here), and economy-wide output prices, broad estimate, % over a year ago, 3-month moving average (documentation here).
- Excess credit, sales and house price inflation score: This is an average of four non-CPI/PPI indicators related to inflationary pressure. The indicators are (1) private bank credit, jump-adjusted, % over a year ago (documentation here) minus estimated medium-term nominal GDP growth, (2) nominal retail sales, % over a year ago, 3-month moving average (documentation here) minus estimated medium-term nominal GDP growth, (3) an average of weekly and hourly earnings growth, % over a year ago, 3-month moving average (documentation here) minus estimated medium-term nominal GDP growth per worker, and (4) residential real estate price growth, % over a year ago, 3-month moving average (documentation here) minus the inflation target.
- Composite excess inflation score: This is simply the re-normalized average of the above three excess inflation scores.
Inflation pressure scores are slow-moving indicators with little volatility, changing signs on average every three years. Although the various scores are positively correlated, they display notable differences in dynamics.
A conditional market trend here is defined as the original market trend times a winsorized inflation pressure score. Winsorization here means that the absolute values of the score cannot exceed two standard deviations to de-emphasize outliers. If price pressure is positive, the conditional market trend signal predicts subsequent equity returns negatively, in proportion to the product of the trend and the absolute pressure score. If price pressure is negative, the conditional market trend signal predicts subsequent equity returns positively, in proportion to the product of the trend and the absolute pressure score. For example, if there is very high inflation pressure and commodity and equity prices have been soaring, the signal calls for particularly strong payback.
This means that there are notable differences between unconditional futures trends and conditional trends, as illustrated in the chart below. The difference shows in both size and direction, albeit in “normal times” of slightly elevated inflation, the two have been positively correlated.
Predicting weekly S&P500 returns
We first assess the predictive power of conditional and unconditional futures return trends at the end of a week for the subsequent week’s S&P500 returns. The analysis thus assumes that positions in the S&P500 mini are taken at the end of the latest close of the relevant commodity futures market, which is 5 pm EST. We ask two questions:
- Do (negative) short-term futures trends predict subsequent S&P500 returns?
- Does the predictive power improve if the trend is conditioned on the inflation environment?
In the above sections, we have presented a range plausible range of future trends and inflation condition factors in this post. Economic theory does not provide clear guidance on these is preferable. Hence, we take an average signal based on all combinations of trends and macro factors.
The below scatters show weekly-frequency correlations between (negative) futures trend signals and subsequent returns. In all cases, the predictive relationship has been positive. For the case of the main trend definition (3-day versus 10-day moving averages) all relations are highly significant. However, conditioning the futures trend signals on the inflation state increases forward correlation and statistical significance. Double-checking the results based on the 1-day versus 5-day moving averages confirms the forward correlation and shows that it improves markedly through macro conditioning.
The benefit of macro conditioning has been even more conspicuous with respect to weekly accuracy, i.e., the ratio of correctly predicted return directions, and balanced accuracy, i.e., the average of correctly predicted negative and positive return directions. Unconditional futures return trends struggle to take the 50% hurdle, while conditional futures return trends all have been around 52%.
The PnL value of a simple macro-conditional price signal
Finally, we investigate if conditional futures trend signals can add material economic value by managing exposure to the S&P500. The analysis is based on a “naïve” profit and loss series. Those are calculated by taking positions in the form of one unit of expected volatility of the S&P500 per unit of normalized signal. Positions are adjusted weekly using the previous week’s final signal value. The trading signals are capped at a maximum of four standard deviations as a risk limit. All PnLs are scaled to 10% annualized standard deviations.
Both unconditional and conditional futures trend signals would have generated modest risk-adjusted excess returns in the long run, albeit with great seasonality. Macro-conditioning mattered. The 30-year Sharpe ratio of the unconditional futures trend signal was near 0.35, while the Sharpe ratio of the conditional signals was almost 0.5. The Sortino ratios were 0.5 and 0.7, respectively. Importantly, the correlation of the PnL based on the unconditional trend signal with the S&P500 is less than 5%, suggesting that most of the value generation has been additive to overall U.S. equity exposure. The seasonality of the PnL arises from the concentration of larger signals in periods of very high or very low inflation pressure.
Risk-adjusted returns of a strategy with a single signal type and single high liquid contract are not normally very high. There is a trade-off between performance ratios and the scalability of absolute returns. In the case of the S&P500, the emphasis is on scalability.
Another way of looking at value generation is to consider two volatility-targeted long-only S&P500 portfolios over the past 30 years. The first one maintains a constant risk-adjusted long and simply adjusts positions each week to pursue a 10% volatility target. The second one manages the risk-adjusted long exposure by overlaying a winsorized conditional trend signal, such that at all times, the exposure varies between zero and double the average long exposure in volatility terms. This simple management of leverage alone would have increased the long-term Sharpe ratio of S&P500 exposure from 0.6 to nearly 0.8 and the Sortino ratio from above 0.8 to above 1.