Getting the sector right is a key determinant of portfolio returns. Along with regional allocation, it comes right after asset allocation and before security selection in a top-down process. There have been multiple studies in recent years showing that – alongside momentum – seasonality and earnings revisions indeed add value in sector allocation. We covered this some time ago1 by comparing individual factors such as momentum, fund flows and value, presenting a coherent framework on how to combine these elements in our Global Equity Sector Scorecard. We now expand on these considerations, examining common drivers of sector returns, and we highlight new approaches to sector allocation that can improve risk-adjusted returns.
In retrospect, 2018 was largely characterised by the return of volatility. This is hardly surprising given that in 2017, market volatility had been hovering at all-time lows, with stock markets persistently advancing, and global government yields declining in the aftermath of the Trump reflation trade until late in the year. We admit ongoing Brexit uncertainty and the escalation in trade relations associated with a drop in global trade growth has clearly played an important role as well, but the 2018 market environment still looks pretty much like a reversion to the mean.
While hedge fund and multi-asset fund managers might have been under pressure over that time, due to their pledge to deliver uncorrelated (or even positive) returns in all periods, stock pickers should be the beneficiaries of the broad-based increase in risk aversion. Chart 1 shows why: there is a strong link between cross-sectional volatility (e.g. sectors, regions) and the prevalent market uncertainty (here represented by the CBOE VIX Index), implying that rising uncertainty among market participants leads to greater return differences between sectors and vice versa. With volatility not expected to decline over the coming months due to late-cycle characteristics, the chance of earning alpha is increasing strongly. In plain language: good forecasting power will directly translate into (significantly) superior performance.
MSCI AC World sectors
The one-billion-dollar question is thus: how do we achieve greater returns from sector allocation? One approach would be to look at work previously done (as mentioned in the preamble). However, we would like to look at sector allocations from a different angle – we would like to understand whether there are common drivers and/or patterns in sector returns. This would provide more insight into how individual sectors behave with respect to each other and which factors dominate the overall direction. A principal component analysis (PCA) is an appropriate method to examine the behaviour of sector returns.
In our analysis, the set of variables is the individual sector returns (11 sectors means 11 variables). Focussing on the most important dimensions allows us to detect common patterns or features which all or only some of the sectors may have in common. In addition, there is a high chance of detecting contrasts and contrary behaviour which might be useful in determining how sectors perform in specific market phases.
From our PCA we learn that the first principal component accounts for about 80% of the total variance of sector returns (we maintain the suspense for a moment by not immediately revealing what this is!). The contribution of the other components declines quite sharply, with the second and the third component only responsible for 8% and 5% of the total variance.2 Due to the fact that the first three components represent almost 90% of the underlying sector variance, we place less focus on the analysis and interpretation of the remaining principal components (e.g. the fourth component only accounts for roughly 2% and interpretation appears difficult in light of the component’s positive relation to materials and real estate and a simultaneous negative relation to energy and healthcare).
In addition to the previous findings, it is quite interesting how individual sectors are related to the different principal components (often called factor loadings) which can be easily interpreted since a principal component can be described as a linear combination of the underlying variables (Chart 2). While the first principal component is positively related to each sector, the second component reflects a strong bias to largely defensive sectors3 (and a negative relation to the energy sector). Moreover, the third component loads positively on growth/cyclical stocks, while once again the energy sector has a negative sign. If we now produce the time series of the principal components by combining the loadings and the original time series of sector returns, we get a much deeper understanding of the first three components (Charts 3 and 4). Let’s end the suspense: as the factor loadings had already suggested, the first principal component perfectly describes the overall market development. Meanwhile, the second and third components reveal an important pattern in sector returns: after adjusting for the overall market direction, sector returns are primarily driven by the spread between cyclicals (with a slight growth tilt) and defensives. What’s more: the energy sector does not really fit into either basket as energy stocks may behave like cyclicals or defensive sectors, depending on the prevalent market environment and the direction of crude and gas prices (although one may claim this to be a good diversifying feature). The relatively high importance of the cyclical-defensive spread in the recent past also suggests that factor strategies such as value, momentum, low volatility, etc. are likely to have common features in certain periods which ultimately prevents them from being discovered in a principal component analysis. To summarise the previous section, the primary driver of sector returns (after the market beta) is investor risk appetite as reflected in the cyclical-defensive spread. While sector allocation on an individual sector level (e.g. overweight IT, underweight utilities) can be an important tool to generate alpha, the overall success of the strategy might be heavily dependent on the allocator’s ability to time the cyclical-defensive spread. The next logical step is to identify specific factors with the capability to forecast the relative attractiveness of cyclicals versus defensives and vice versa. As a matter of fact, we believe that a good starting point are macro surprises.
Economic data releases can be classified into six categories: housing and real estate, industrial sector, labour market, personal and household sector, retail sector, as well as survey and business cycle indicators (soft data). We obtain surprises in the data by comparing the actual release to the respective consensus forecast. For the purpose of timing sector allocation decisions, we are just interested in a binary outcome – whether aggregated surprises over a three-month period have been positive or negative. The big advantage of using binary macro surprises (as opposed to momentum or value figures) is that it does not require ranking single sectors according to their returns in a specific macro environment. We have shown above that the most important driver of sector returns (after the market itself) is the development of cyclical versus defensive stocks. Fortunately, this is highly correlated to the prevalent macro backdrop and, thereby, to macro surprises (Chart 5). The indicator is also robust: periods of prolonged cyclical underperformance have mostly been accompanied by disappointing macro data (as in 2008, 2011, 2014, and 2018). The relationship between the differentials in sector returns and macro surprises does not seem to have lost validity. We take advantage of this by employing a strategy that allocates between baskets of defensive and cyclical sectors – depending on the sign of macro surprises over the past three months. It outperforms quite nicely a benchmark with the same exposure as the strategy’s average exposure (Chart 6).
Sector allocation is an important source of alpha for a portfolio’s return. However, we show that the largest share of the return variation – after the market return itself – comes from the allocation between cyclical and defensive sectors. It might hence be sufficient to invest in a basket of cyclicals or defensives at the right time to reap most of the benefits from sector allocation. Macro surprises can help to time the switch from one to the other.
1 See 'Millennials and equity sector allocation', Thinking Ahead, July 2017.
2 We used monthly data from 2010 until 2019. A longer look-back period might be appropriate from a statistical perspective. However, one has to bear in mind that taking into account the period from 1998 to 2003 and 2006 to 2009 significantly change the final outcome given the bubbles seen in IT and telecom stocks at the end of last century and the massive exaggeration in the US real estate/housing market between 2005 and 2007. The results from a larger data sample would hence be significantly influenced by decisive differences between the first ten years of this century and the period after the global financial crisis as, for one thing, growth stocks currently do not have bubble-like valuations and, secondly, the characteristics of real estate stocks have dramatically changed from an economically-sensitive sector to a low-beta and interest-rate play. We therefore deem the period from 2010 on as the best proxy for the near future.
3 We partly follow the methodology of MSCI by classifying consumer discretionary, financials, industrials, IT and materials as cyclical sectors and consumer staples, healthcare, telecoms and utilities as defensive stocks. In addition, real estate is also treated as a defensive sector given its below-average beta to the market and negative correlation to interest rate changes.