Home » Research Blog » Equity trend following and macro headwinds

Equity trend following and macro headwinds

Jupyter Notebook

Market price trends often foster economic trends that eventually oppose them. Theory and empirical evidence support this phenomenon for equity markets and suggest that macro headwind (or tailwind) indicators are powerful modifiers of trend following strategies. As a simple example, we calculate a macro support factor for equity index futures in the eight largest developed markets based on labor markets, inflation, and equity carry. This factor is used to modify standard trend following signals. The modification increases the predictive power of the trend signal and roughly doubles the risk-adjusted return of a stylized global trend following strategy since 2000.

The below post is based on proprietary research of Macrosynergy Ltd. It ties in with this site’s summary of the importance of macro trends.

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

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

A reminder of the macro headwind theory

Here macro headwinds are defined as economic developments that undermine prevailing price trends in financial markets. Conversely, macro tailwinds are developments that support current market trends. The macro headwind theory emphasizes that market trends themselves can over time invite countervailing macroeconomic developments. For example, declines in bond yields may boost credit and spending growth, which subsequently puts upward pressure on interest rates. Also, currency appreciation can undermine the competitiveness of a country, as discussed in a previous post here. The present post illustrates that macro headwinds also matter in equity markets and that tracking them helps enhance standard trend following strategies.

The benefits of integrating market trends and macro headwinds

Trend following is a strategy that relies on the persistent direction of market prices as a signal. It assumes that today’s trend can serve as a positive predictor for tomorrow’s returns. Trend following is considered a valid strategy type in the equity space because asset owners are susceptible to behavioral biases and rational inattention, both of which plausibly result in sustained trends. Academic research has highlighted the “disposition effect,” whereby investors tend to sell assets that have generated profits and hold onto those that have incurred losses. Additionally, the high information costs and limited attention further rationalize investors’ inclination to mimic the decisions of others with an information advantage.

Equity market capitalization in developed markets typically falls between 60% to 200% of GDP for most developed countries. Consequently, equity price trends influence the macroeconomy. When stock prices increase, they contribute to the growth of household wealth and create favorable conditions for corporate investment. This leads to a rise in aggregate demand for goods, tighter labor markets, and potentially even inflationary pressure. Therefore, considering the market’s information state concerning relevant macro trends can enhance the ability to predict the sustainability of market trends. In other words, incorporating macroeconomic information alongside price trends can provide a more well-informed signal.

On the other hand, integrating price trends into macro signals can be advantageous for risk management in pure macro strategies. By including price trends, trend following strategies offer a built-in stop-loss mechanism. If price trends escalate contrary to the macro trend signal or deviate from their predicted direction for extended periods, the stop-loss mechanism helps manage and mitigate risks.

A simple plausible macro headwind for equities

Drawing from conventional macroeconomic theory, we construct a straightforward macro support score specifically designed for local-currency equity markets in developed countries. This support score can then be applied in the context of market price trends as a headwind (or tailwind) indicator.

The hypothesis is that an equity-supportive macroenvironment is characterized by a slack labor market, low inflation, and high earnings and dividend yields relative to real funding costs (“equity carry”). The constituents and composite scores have been calculated in the simplest plausible terms and do not involve any feedback on their predictive power of formal optimization. We disregard in this post other macro influences, such as the real effective exchange rate, business sentiment, and real economic growth. A wider set of macro trends may improve the support factor but is not required for a simple proof of concept.

We consider macroeconomic information for eight developed countries with liquid local-currency equity index futures, to which we refer by their currency symbols in the below analysis: Australia (AUD), Canada (CAD), Switzerland (CHF), the euro area (EUR), UK (GBP), Japan (JPY), Sweden (SEK), and the U.S. (USD). For a meaningful macro support state calculation, 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 concerning an economic concept and, hence, are suitable for backtesting the impact of macro information on trend following strategies. For most of the below concepts, we use daily quantamental information states from 2000 to mid-June 2023.

  • Labor market slack: We describe the state of the labor market as an equally weighted average of three constituents. The first is the employment growth (% over a year, 3-month moving average, or quarterly average) relative to the trend growth in the workforce and put in negative terms (view documentation of underlying data here and here). The second constituent is the unemployment rate relative to its previous 5-year moving median (documentation here). And the third constituent is the growth rate of the main wage metric (% over a year, 3-month moving average or quarterly average) over and above labor productivity growth and the inflation target (see documentation of these data here and here). The constituents are z-scored around their natural neutral level, with standard deviation estimated on a rolling out-of-sample basis and a limit of 2 (to de-emphasize outliers). The three metrics can display very different patterns but are each a valid aspect of labor market slack.

  • Inflation shortfall: This is the negative difference between core CPI inflation and the local inflation target. Core inflation mainly excludes food and energy items (view documentation here) and is particularly relevant for CPI trends and monetary policy trends. We use two versions of core inflation metrics, the standard percent change over a year ago and a statistically more attractive percent change of the latest six months over the previous six months, seasonally adjusted and annualized. The two metrics are scored and then combined in equal terms.

  • Excess real equity carry: This is the estimated carry of the country’s main equity index. It is calculated as the difference between (1) the average of expected forward dividend and earnings yields and (2) the main local-currency real short-term interest rate. We use two versions of this carry from the JPMaQS database (view documentation here). One is simply the percent annualized carry on the notional of the contract. The other is the percent annualized carry on the risk capital of a position scaled to a 10% volatility target. The neutral level for both has been set to a carry-to-volatility ratio (“Carry Sharpe”) of 0.3. Real carry is attractive if the Sharpe ratio resulting from the carry return alone is at least 0.3. Again, the two versions are z-scored and simply combined with equal weights as for the other indicators. The real equity carry is a natural complement to the inflation score: high inflation may bode for tighter monetary policy in greater interest rate risk, but can also push up the equity carry, if real interest rates decline or become negative.

The above aggregate scores are re-normalized – merely to equalize standard deviations – and linearly combined to an aggregate macro support score. A high value of that score means favorable macro conditions for the local equity market and a low score means unfavorable conditions.

The below chart panel shows the three macro scores since 2000. They have very different patterns and dynamics. The labor market slack score moves the most gradually, while the inflation shortfall tends to have more pronounced mini cycles. The equity carry metric displays most short-term volatility due to the impact of price volatility on one of its constituents.

Equity carry based on reliable sets of earnings and dividend predictions for the equity index is only available from 2004 for countries outside the U.S. and in these cases the macro support score only uses the labor market slack and inflation shortfall. The time series patterns of aggregate scores are similar across countries but far from uniform.

Standard and modified trend signal

We are looking at futures of the following equity indices (view documentation of return calculation here):

  • AUD: Standard and Poor’s / Australian Stock Exchange 200
  • CAD: Standard and Poor’s / Toronto Stock Exchange 60 Index
  • CHF: Swiss Market (SMI)
  • GBP: FTSE 100
  • JPY: Nikkei 225 Stock Average
  • SEK: OMX Stockholm 30 (OMXS30)
  • USD: Standard and Poor’s 500 Composite

As for previous trend following posts, we defined as a trend the most quoted standard of the difference between the 50-day and 200-day moving averages of the equity index return indices, which are plotted in the below panel. On average, the resulting trend signals change direction roughly once or twice a year.

To account for macro head- and tailwinds we modify the (normalized) trend signal by the macro support score. Modification means that we multiply the original trend signal with a coefficient that can take values between 0 and 2, i.e., can reduce or double the original trend signal depending on the agreement of the macro situation with the price trend. The modification coefficient is a logistic (sigmoid) function of macro support z- score.

Specifically, the adjustment implements the following equation:

modified_trend = ((1 - sign(trend)) + sign(trend) * coef) * trend


coef = 2/(1 + exp(-2 * macrosupport_score))

where trend means the original trend, sign() returns the sign of its element as 1/-1, and macrosupport_score denotes the composite macro support score.

Here, applying the macro support score depends on the sign of the concurrent trend signal: if the trend signal is positive macro support enhances it, and macro resistance reduces it. The logistic function translates the macro support score such that for a value of zero, it is 1. For values of -1 and 1, it is 0.25 and 1.75, respectively, and for its minimum and maximum of -3 and 3, it is 0 and 2, respectively.

The below chart panel shows the time series of standard and modified trend signals. Naturally, modification never changes the sign of the signal and the direction of the market position. However, it does alter the position sizes and the risk profile of the strategy over time. For example, the strong macro support for local equity markets in the 2010s magnified risk-taking in positive trends and reduced risk-taking in negative trends. Similarly, the adverse macro trends in some countries in the 2000s resulted in greater emphasis on short positions before the great financial crisis.

Empirical evidence of the macro headwind consideration

We check if modifying the standard trend following signal by macro headwinds and tailwinds affects the predictive power and stylized PnL generation.

We must consider some structural features of these data to assess the significance of predictive power across eight countries. Equity market performance is highly correlated across developed markets. There is also a positive correlation of macro support factors across countries. This reflects the market and economic structure. Companies have an international presence and depend on economic and financial conditions in other countries as much as on those in their home country. Generally, there is natural “leakage” of one country’s macro information state onto the performance of equity markets in others. In particular, smaller countries’ markets strongly depend on the conditions in large economies.

Therefore, treating the experiences of local equity markets as independent datasets would lead to pseudo-replication and obviously overstate the significance of any findings. Neither can we simply apply standard panel analysis because we cannot rely on cross-country differences as plausible predictors for markets. For each equity market, it is not primarily local conditions that determine price development but some version of global conditions. Hence, to assess the predictive power of standard and modified trend scores, we do not look at markets separately but at global developed market aggregates. Aggregation is done by linear combination using the USD GDP weights of countries (view documentation here). Thus, large countries, such as the U.S. and the euro area have larger weights in the aggregate of return targets and signal features of the analysis.

Standard correlation analysis shows that global standard trend scores at a monthly frequency have been positively and significantly related to subsequent global monthly equity index future returns.

Modifying the trend signals by macro support scores enhances correlation and significance notably according to both Pearson correlation (parametric) metrics and Kendal correlation (non-parametric) statistics.

Next we calculate naïve PnLs based on standard rules used in Macrosynergy posts. This means that positions are calculated based on standard and modified trend scores across the eight developed equity market index futures. Positions are rebalanced monthly based on signals available at the end of each month with a one-day slippage added for trading during the first day of the next month. This simplistic rebalancing is a standard for testing the economic value of factors and is not optimal. Transaction costs are disregarded at this stage because they depend heavily on the value of assets under strategy management. The long-term volatility of the PnL for positions across all currency areas has been set to 10% annualized for ease of presentation.

The main finding is that macro-modification would have roughly doubled risk-adjusted returns of a directional equity trend following strategy since 2000. The long-term naïve Sharpe ratio of the modified trend signal would have been 0.55 with a negative overall correlation to the S&P500, compared to a Sharpe of 0.27 for the standard trend following strategy. The outperformance developed through four distinct phases episodes since 2004.

The outperformance of the modified trend signal is not very sensitive to the choice of the modifier from within the scope of the economic measures discussed above. Thus, all three constituents of the macro support score would have enhanced trend following turns significantly, albeit individually not quite as much as the composite score. Long-term Sharpe ratios of the individual scores have been between 0.44 and 0.49.



Related articles