Macro investing with style

Incorporating an assessment of the macro environment into asset allocation has become standard practice in the multi-asset industry. Essentially, it involves a ‘nowcasting’ stage where a portfolio manager identifies the current macro regime using a set of economic and financial indicators, and then turns to historical data to determine which assets fared better during similar times in the past. The main challenge of macro investing is identifying the economic regimes that explain the variation in asset returns.

Most commonly, macro regimes are states of the business cycle. Practitioners tend to distinguish four states when only focusing on economic growth: recovery, expansion, slowdown and recession¹. More complex approaches use inflation and monetary policy indicators to complement the growth-based metrics. 
 

The ground already covered

From the simple boom-and-bust approach² much ground has been covered in the business cycle analysis for portfolio management, with key contributions from research within investment management companies. Blitz and Van Vliet (2011)³ showed how traditional risk premia returns were vastly different depending on economic regimes. Illmanen et al. (2014)⁴ highlighted how the combination between inflation and growth regimes could further help explain the variation in returns. Finally, Blin et al. (2021) extended these results to the returns of alternative risk premia, shedding light on how they depend on growth, inflation and stress cycles.

The disappointing performance of equity factors led practitioners to explore the possibility of basing their timing on macro. Recently, we have witnessed multiple pieces of sell-side research advocating macro-driven allocation though mapping of factors onto stages of the business cycle. The links behind those relationships appear quite intuitive. For example, ‘safe’ quality factor outperforms in recessions, while ‘risky’ value factor benefits from economic recoveries. Nevertheless, evidence of the success of macro timing on equity factors is mixed, to say least. While some find modest support for improved performance⁵, others claim that business cycle indicators fail to explain the large variation in factor returns⁶.
 

Our approach

This paper describes how we implement macro investing in systematic equities as it differs from the mainstream approach. First, we look beyond traditional factors and consider timing the allocation between a wider range of equity styles, further differentiating them across sectors and regions. This flexibility allows us to expand the breadth of the strategy in order to potentially benefit performance. Additionally, we avoid modelling macro regimes explicitly, opting instead for a relevance-based approach⁷. In building our macro views, we incorporate the full track record of equity styles, weighting past returns according to their similarity to the current macro environment⁸. 

To characterise the macro environment, we picked seven indicators that reflect the state of key financial markets: equities, rates, oil, US dollar, credit, and the yield curve⁹. Since levels are meaningless without a context, we transformed these indices by ranking them over a moving window of three years. Each date in the history is therefore characterised by a state vector whose components range between 0 and 1. For example, the value of 1 for the equity component means that, as of today, the stock market is at its highest point over the three years.

To measure the similarity in the macro environment or macro distance between two dates, we compute the Euclidean distance between corresponding state vectors. To illustrate, Figure 1 shows the distances between 31 August 2023 and each year since 1991. Observations with short distances are more pertinent as they signal that market environments were similar. The similarity between 2005 and 2006 and the current market environment highlights its consistency with a late-cycle period.

Figure 1. Macro distance to 31 August 2023: consistent with late-cycle period 2005-2006
 

Source: LOIM, Bloomberg. The graph shows the minimum distance between 31 August 2023 vs the years 1991-2022. 

We adopt the definition of equity styles introduced in our paper on momentum strategy in equity styles¹⁰. For a given fundamental metric such as, for example, price-to-book, a given geographical region, and economic sector, we form a style portfolio. This portfolio consists only of stocks from the given sector and region: it is long the top 50% of stocks according to the chosen metrics, and short the bottom 50% of stocks¹¹. Crossing three regions (US, Europe and Japan), 11 GICS1¹² sectors and 48 fundamental metrics, we end up with 1584 single style portfolios, which we call simply styles. By construction, styles are sector and region neutral, meaning that their combination will yield a region and sector neutral portfolio as well.

The idea behind the relevance-based approach is to form convictions based on the average historical performance of assets weighted by relevance to the current macro environment. That is, past dates that are closer to today in macro terms receive higher weights. To specify the weighting scheme, we need to define an explicit function that maps macro distances onto weights. 

Instead of postulating an arbitrary weighting scheme, we took a different route. For each style we ran a time-series regression of its past returns on macro distances with the lag of one month:

where is the macro distance between the current date t  and the past date that is n months apart, and   is the historical average of those distances since the beginning of the observations.

Once the linear relationship above is estimated, we use it to forecast the next-month style return. To do this we set n=0 in the formula above noting that D_t.t= 0 :

The expected return consists of two components. The first one is the intercept of regression , which, by construction, is equal to the average historical return of the style. This component can be interpreted as a ‘strategic’ forecast as it is not expected to change much from one date to another¹³. This is a typical quant signal based on a long track record of the underlying factor. 

The second term  is a ‘tactical’ forecast that is driven by the current macro environment¹⁴. Styles with higher betas are those that performed better in more distant macro regimes, hence, they are expected to underperform in the current one. This explains the negative sign. It is important to distinguish tactical and strategic components to appreciate the value-added contribution of macro timing. 

Using the procedure described in the previous section, we designed two strategies. One strategy builds on the strategic forecast of style returns, while the other incorporates only the tactical one. When running regressions, we excluded the dates within a 12-month window to avoid the interaction with the momentum in styles. Indeed, the most recent periods are naturally close to today in macro terms, therefore, recent style returns will mechanically receive higher weights in the forecast, and the performance will be enhanced by the momentum effect¹⁵.

Figure 2.  Backtest timing of equity styles: strategic vs tactical 2022-2002 
 

Source: LOIM, Bloomberg. Combo is an equal-weight combination of strategic and tactical portfolios. For illustrative purposes only.


Figure 2 shows the backtest of the two strategies with monthly rebalancing frequency as well as their equal-weight combination. The historical observations used to compute expected returns start from the end of 1992¹⁶, and the number of observations expands as we progress in time. The backtest starts from the end of 2022 when 10 years of data become available. Both strategic and tactical portfolios were built using a mean-variance optimisation to form a portfolio of styles maintaining an ex-ante volatility of 1% p.a. 

We observe that the strategic allocation that is solely based on historical style returns generates a positive performance. Most importantly, the tactical allocation adds value too. It is interesting to note that the two strategies diversify each other well. During the periods when the strategic allocation was detrimental, the tactical positioning was beneficial and vice versa.
 

Thus far we have demonstrated that macro investing can be successfully applied to time a wide set of equity styles. This result begs the question: what happens if we use the same approach to time few ‘traditional’ factors? Apart from obvious practical implications, we are curious to see how the use of a broader set of styles helps improve the performance.

To answer this question, we formed five factor portfolios (value, quality, momentum, low beta, and small size) using our proprietary methodology on stocks in the MSCI World index in a sector and regional neutral way. We then applied the macro timing methodology to the five factor portfolios.

Figure 3. Backtest timing of traditional factors: strategic vs tactical 2022-2002
 

Source: LOIM, Bloomberg. Combo is an equal-weight combination of strategic and tactical portfolios. For illustrative purposes only.

Figure 3 shows the performance of the three strategies defined in the previous section. The performance of the strategic portfolio reminds us of the behaviour of a typical multi-factor portfolio. While historically this portfolio offered an attractive profile, it added little value over last several years. The resemblance is not surprising given that the strategic portfolio is close to a static one where allocations are determined by long-term factor returns. 

The tactical portfolio seems to provide some diversification. It is less stable in the early 2000s, likely due to the limited history used to compute expected return. However, over the last decade, it offered a valuable addition to the strategic allocation. Overall, the combination of the two portfolios exhibits a smoother performance, though it is not as attractive as the strategy based on a wider set of styles. 

To illustrate how the factor allocation is formed, figure 4 provides a snapshot of factor weights in the three portfolios as of the end of August 2023. We observe that while strategically we tend to favour quality¹⁷, tactically, we prefer allocating more to momentum today. It is also worth noting that the current tactical overlay drives the weight of small size into a negative territory, favouring a large cap bias.

Figure 4. Snapshot of factor weights as of 31 August 2023
 

Source: LOIM. Factor weights in a portfolio with 1% volatility p.a. Combo is an average between strategic and tactical portfolios. For illustrative purposes only.

Macro investing has been traditionally implemented in multi-asset portfolios. However, there is a growing interest in its application to alternative risk premia. In this note, we demonstrated that this investment approach could add value in systematic equity as well. We outlined our implementation of macro for the timing of equity styles, where it offers a clear enhancement over a static solution based on long-term returns. This macro strategy represents a building block of our book of systematic equity alpha, along with style momentum and machine learning strategies.