GOAL Post
The lowdown on the risk of an(other) equity drawdown
Global equities had a drawdown in August and have remained volatile since. Equity drawdowns are a common feature, but their size and frequency has varied over time - risk-adjusted returns were unusually high in the last 12 months.
While 'buy the dip' has been a successful strategy since the GFC, its success has been mixed over long horizons as often equity drawdown risk lingers. If an equity drawdown comes alongside a fast and broad de-risking across assets, i.e., our Risk Appetite Indicator (RAI) drops near -2 or below, or a material valuation de-rating, the asymmetry to 'buy the dip' is usually positive.
However, our RAI remains well above -2 and the valuation de-rating so far has been limited - thus, the asymmetry is not much better and macro momentum remains mixed. To better understand risks from here, we have developed a framework to assess the likelihood of equity drawdowns.
Strong price momentum as well as negative inflation levels and momentum (allowing for central bank easing) point to lower drawdown risk, while recent negative growth momentum and elevated Shiller P/Es point to elevated risk. The negative growth momentum suggests a larger probability of a >10% correction - recent higher equity volatility also points in this direction.
Combining all variables indicates that drawdown risk has picked up but remains relatively low at around 20% - historically, levels above 30% indicate a clear warning signal. In addition, valuations are a key driver of elevated drawdown risk - ignoring the absolute Shiller P/Es level points to lower risk.
With elevated equity valuations, mixed macro momentum and rising policy uncertainty, there is the risk of more equity drawdowns, in our view. As a result, we believe risk-adjusted returns for equities are likely to be lower into year-end. However, we think the risk of a bear market remains low with relatively low recession risk, helped by a healthy private sector and central bank easing.
While 60/40 portfolios have performed well since the summer, already dovish Fed pricing and upward pressure on term premia might mean a smaller bond buffer from here. Alternative safe havens such as Gold, Yen and CHF are likely to provide more diversification benefits. Higher vol of vol also increases the value of option hedges, in our view - we like selling upside on bonds to buy equity puts.
The lowdown on the risk of an(other) equity drawdown
Summertime sadness - global equities have a drawdown
Global equities have been on a volatile roundtrip this summer - at the beginning of August, >90% of global equity indices had at least a 5% drawdown and 40% had a correction of 10%. This followed the strong equity rally since Q3 last year with anchored volatility, resulting in strong risk-adjusted returns (the S&P 500 Sharpe ratio up until August 5th was above 3). As we wrote in our last asset allocation update, the ensuing bullish sentiment into the summer coupled with more negative macro momentum increased the risk of a setback. Most equity indices have recovered the equity drawdown in August and are near their all-time highs, even though equity volatility has resumed in September, in part due to continued growth concerns. For investors the question is now if the bull market resumes, with similarly high risk-adjusted returns, or if there is more bumpiness ahead and more risk of equity drawdowns.
Exhibit 1: In August 94% of equity indices had a drawdown of 5% or more
Proportion of 32 global equity indices in a 5% or 10% drawdown
Source: Datastream, Goldman Sachs Global Investment Research
No risk, no return - equity drawdowns are relatively common
Drawdowns are a common feature of equities but their size and frequency has varied over time. Smaller equity drawdowns are common, occurring roughly half the time (Exhibit 2). Large equity drawdowns are more rare and most of the time coincide with recessions. On average, in the last 100 years there has been an S&P 500 >20% drawdown nearly every 4-5 years but in recent years bear markets have been less frequent (Exhibit 3). The equity bull market in the 1990s was the longest, supported by structural tailwinds and the Tech Bubble. The post-GFC cycle was also unusually long, but there were three large S&P 500 drawdowns that came close to a bear market: (1) the Euro area crisis in 2011 with 19.4%, (2) the EM/oil crisis in 2015 with 14.2% and (3) the drawdown in Q4 2018 came very close with 19.8%.
Exhibit 2: Equity drawdowns were relatively common across markets and large ones usually were around recessions
12m rolling drawdowns (data since 1973)
Source: Datastream, Haver Analytics, Goldman Sachs Global Investment Research
Exhibit 3: Bull markets have been longer since the 1990s
Source: Bloomberg, Goldman Sachs Global Investment Research
There have been roughly 22 S&P 500 bear markets (drawdown >20%) since 1928 - the average bear market peak to trough is around 40% and has lasted 16 months. Drawdowns of more than 10% (but less than 20%) are more common - the average correction peak to trough is 13% and has lasted 4 months. In general, bear markets and equity drawdowns have become somewhat less frequent since the 1990s (Exhibit 4) - in part this reflects longer business cycles and lower macro volatility, with the Great Moderation as well as more buffering from central banks owing to anchored inflation.
Exhibit 4: Bear markets and corrections have been less frequent since the 1990s
Occurrence of S&P 500 drawdowns within a 12-month window
Source: Bloomberg, Goldman Sachs Global Investment Research
Buy the dip or sell the rip - drawdowns can be a buying opportunity
Since the GFC, ‘buy the dip’ of the S&P 500 has been a successful strategy in response to any 10% drawdown or more. The average subsequent returns from simply buying the S&P 500 after a 10%+ dip were higher than the average since 2010 - hit ratios were also higher (Exhibit 5). However, 5-year returns suggest that by just buying the dips investors missed out on some very strong return periods without any drawdowns. Still, the success of 'buy the dip' also meant that strategies selling insurance on equities, such as put selling, had strong risk-adjusted returns.
Exhibit 5: Since 2010 buying the 10% dip has been very successful
Average return and hit ratio for positive returns for 'buy the 10%+ dip' strategy since 2010
Source: Bloomberg, Goldman Sachs Global Investment Research
However, a simple ‘buy the dip’ strategy for the S&P 500 since 1990, which includes the Tech Bubble burst and the GFC, has had worse near-term returns and hit ratios (Exhibit 6). Only a 'buy the dip' strategy for larger than 20% drawdowns has had average returns marginally better than unconditional ones but lower hit ratios - and not for longer time horizons (>2-year). The same is true for a 'buy the 20%+ dip' strategy since 1928 (Exhibit 7) - there is a small improvement in near-term returns but hit ratios only improve over very long horizons but average returns tend to be lower.
Exhibit 6: Since 1990, buying the 10% dip has been less successful
Average return and hit ratio for positive returns for 'buy the 10%+ dip' strategy since 1990
Source: Bloomberg, Goldman Sachs Global Investment Research
Exhibit 7: Since 1928, buying the dip has been less successful
Average return and hit ratio for positive returns for 'buy the 20%+ dip' strategy since 1928
Source: Bloomberg, Goldman Sachs Global Investment Research
Large equity drawdowns often create better asymmetry
Whether buying the dip is a good strategy depends on whether equity drawdown risk lingers. After an equity drawdown, the asymmetry is often better as valuations decline and sentiment and positioning become more bearish, but there is often risk of further deterioration of macro and market conditions. An equity drawdown can also result in tighter financial conditions, which can further weigh on macro momentum. Only after material declines in sentiment and positioning or valuations is the asymmetry to add risk often better, irrespective of subsequent macro conditions.
Sharp declines in our Risk Appetite Indicator indicate breadth in de-risking
Sharp declines in our Risk Appetite Indicator (RAI) historically have signalled a better asymmetry for risky assets. The hit ratio of positive S&P 500 returns in the subsequent 6-12m from levels below -2 is above 90%. With 27 risk premia and pair trades across assets it does capture the breadth of de-risking and, as it is based on 1-year rolling z-scores, it incorporates the speed of the sell-off, which increases the potential for an overshoot. As we wrote on August 6th, the RAI collapsed to one of its lowest levels since the 1990s intraday on August 5th, mostly due to the VIX spike, troughing at -2.6 before recovering to -1.3 by EOD - there was not the breadth that usually signals better asymmetry.
Exhibit 8: After large declines in risk appetite, there are usually fewer S&P 500 drawdowns
Risk appetite Indicator (RAI); Orange shading = subsequent S&P 500 drawdown >10%
Source: Datastream, Haver Analytics, Goldman Sachs Global Investment Research
Exhibit 9: From low levels of our RAI, the asymmetry to add risk improves
Average subsequent S&P 500 returns and hit ratios from low RAI levels (data since 1990)
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Since risky assets recovered in August and after the recent volatility, the RAI has now settled at -0.8, signalling less positive asymmetry on a stand-alone basis. From negative RAI levels above -2, there can be larger S&P 500 drawdowns in the next 6 months in case of weaker subsequent growth data (Exhibit 10). As we recently wrote, the fact that safe assets have not sold off materially indicates that there is less complacency, in our view. Still, we believe near-term equity drawdown risk will be conditioned by US and global growth data and if central bank easing helps further ease growth concerns.
Exhibit 10: Positive asymmetry for equity returns in the next 6 months from the current levels of our RAI improves if growth holds up
Data since 1991. CAI: Current Activity Indicator.
Source: Datastream, Haver Analytics, Goldman Sachs Global Investment Research
Material valuation de-rating improves asymmetry for equities
A large valuation de-rating can also result in a better asymmetry for forward equity returns. For example, a sharp valuation reset without large declines in earnings like in Q4 2018, one of the largest since the 1970s, indicates material risk to earnings is already discounted (Exhibit 11). Equity valuations tend de-rate ahead of profit recessions but tend to overshoot - in several cases, subsequent earnings declines were small. Valuation de-ratings >15% tend to improve average S&P 500 returns and hit ratios on a 12-month horizon (Exhibit 12).
Exhibit 11: During equity drawdowns, valuations usually de-rate
S&P 500 12-month forward P/E; Orange shading = S&P 500 subsequent drawdown >10%.
Source: Datastream, Goldman Sachs Global Investment Research
Exhibit 12: Large valuation de-ratings create positive asymmetry
Average subsequent S&P 500 returns and hit ratios after de-rating of 12-month forward P/E (data since 1990)
Source: Bloomberg, Goldman Sachs Global Investment Research
Near-term returns are again mixed outside of the largest valuation de-ratings. On a 6-month basis, there can still be large S&P 500 drawdowns (20%+) even after large valuation de-ratings. Only in case of a >20% de-rating and if no profit recession is expected, the asymmetry is quite positive. In the case of smaller de-ratings (10%-20%), investors can consider buying the dip if they expect earnings growth to remain positive. However, the recent valuation de-rating is small and earnings revisions have been flat to negative across markets YTD (outside of Japan).
Exhibit 13: Valuation de-ratings of more than 20% without a profit recession have been a buying opportunity - smaller de-ratings (10-20%) require positive earnings growth to mitigate negative tails
Data since 1990
Source: Datastream, Haver Analytics, Goldman Sachs Global Investment Research
The lowdown on the risk of an(other) equity drawdown
To better understand risks from here, we have developed a framework combining drivers of both the likelihood and the size of equity drawdowns. While smaller drawdowns can be difficult to predict, larger equity drawdowns are often conditioned by certain macro and market conditions. Equity drawdown risk is often due to a combination of poor asymmetry, either from the business cycle or the sentiment cycle, and negative macro or market momentum.
We use logit models to estimate equity drawdown risk over different investment horizons, which allows us to track the probability of equity drawdowns in real time (see Appendix for details). We include the sentiment and business cycle metrics from Global Strategy Paper: The Dynamic Balanced Bear - in addition we add equity valuation metrics as well as leading macro indicators and risk-clustering metrics, to increase the likelihood of anticipating equity drawdowns:
Business cycle scores, i.e., levels and momentum of the growth, inflation and policy scores.
Sentiment cycle metrics, i.e., Risk Appetite Indicator (RAI), S&P 500 realised volatility (1-month and 3-month) and price momentum metrics (6-month and 12-month trend).
Valuations, both the S&P 500 Shiller P/E outright, vs. 10- and 20-year average and relative to their macro fair value (for details see Global Strategy Paper: The Strategic Balanced Bear).
Leading growth indicators such as Leading Economic Indicator (LEI)[1], ISM manufacturing survey and components and economic policy uncertainty.
Exhibit 14 shows the links between equity drawdown risk and changes in the different variables. Future changes in the business cycle scores (12m changes in growth, inflation, policy) are major drivers of equity drawdown risk, but that requires perfect foresight. Negative growth momentum tends to increase equity drawdown risk, especially in the near term - 6m changes and diffusion of the LEI work best. Positive price momentum tends to reduce drawdown risk - usually equities consolidate before a large drawdown. Higher equity volatility tends to increase near-term drawdown risk as risk tends to cluster (see also Global Strategy Paper: Balanced Bear Despair - Part 4), but it matters less for latent drawdown risk. However, changes in valuations matter more for the risk of latent, larger equity drawdowns. Finally, tighter policy tends to increase risk of equity drawdowns, both over 3m and 12m horizons.
Exhibit 14: Negative macro and market momentum matter most for near-term drawdown risk, valuations and the business cycle drive latent drawdown risk
Change in probability of a S&P 500 drawdown for a one standard deviation increase (univariate)
Source: Haver Analytics, Bloomberg, Goldman Sachs Global Investment Research
Exhibit 15 and Exhibit 16 show the current probability of a >20% drawdown in the next 12 months and >10% in the next 3 months, respectively, implied by the different variables - they are ranked by their historical track record of predicting drawdowns (the so-called AUC score, see Appendix for details). The recent strong price momentum as well as negative inflation levels and momentum (allowing for central bank easing) point to lower risk of a bear market, while negative growth momentum and elevated Shiller P/Es point to elevated risk. The negative growth momentum (LEI changes and diffusion, ISM new orders/inventories and changes in the unemployment rate) points to an even larger probability of a >10% correction - recently higher equity volatility also points in this direction. While valuations point to a higher risk of a correction, they do not have a strong track record in predicting them (lower AUC).
Exhibit 15: Weaker growth momentum, elevated valuations and tight policy drive higher risk of large drawdowns in the next 12 months, while negative inflation momentum and the positive trend reduce it
Implied probability of a 12m max drawdown > -20% ranked by univariate regression performance (AUC)
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Exhibit 16: Negative growth momentum coupled with increased volatility increases near-term drawdown risk, while the positive trend lowers it
Implied probability of a 3m max drawdown > -10% ranked by univariate regression performance (AUC)
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Combining all variables indicates that equity drawdown risk has picked up at c.20% but remains relatively low (Exhibit 17). The combined model struggles to fully anticipate equity drawdowns (R2 is only 0.2) - but, from levels above 30%, more than half of the times this was followed by either a correction larger than 10% or a drawdown of more than 20% (see Appendix for more details). And the average max drawdown in the next 2 years was very clearly above the unconditional levels if the probability was above 30%. Also, the signals for 3m corrections and 12m bear market risk are closely linked and sometimes complement each other - for example, during the Tech Bubble burst, the risk of a bear market picked up too early due to elevated valuations but the 3m correction signals helped time the start of the bear market.
Exhibit 17: Equity drawdown risk has picked up in the past few months but remains relatively low
Implied probability of S&P 500 drawdown based on multi-variate model (Orange/ light grey shading = S&P 500 subsequent drawdown >20%/>10%. Dashed line = unconditional probability)
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
A decomposition of the 12m drawdown probability reveals that a core driver of elevated risk are equity valuations and recent weaker leading growth indicators (Exhibit 18). However, the business cycle in aggregate and sentiment point to somewhat lower equity drawdown risk. However, the signal from equity valuations might be too bearish - while outright Shiller P/Es are elevated vs. the long-run history, they are less elevated compared to 10- and 20-year averages and a macro-implied fair value. As we wrote in Global Strategy Paper: The Strategic Balanced Bear, higher Shiller P/Es since the 1990s have at least in part been justified due to the uptrend in ROEs and structurally lower inflation. Excluding outright Shiller P/E from the models would result in lower equity drawdown risk from here (Exhibit 19). However, even without the bearish signal from valuations the risk of equity drawdowns has picked up, in particular with continued poor growth momentum (also outside the US).
Exhibit 18: Valuations are contributing most to the recent increase in the risk of larger drawdowns
Shapley decomposition of probability of S&P 500 12m drawdown >20%
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Exhibit 19: Excluding valuations points to bear market risk in line with the unconditional probability
Dashed line = unconditional probability
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Asset allocation implications - better balance but value in alternative safe havens and convexity
We are mildly pro-risk for a 12m horizon but remain tactically Neutral in our asset allocation. The 'early' late cycle backdrop coupled with central bank cutting cycles should support risky assets and multi-asset portfolios over a 12-month horizon. In addition, there is potential for some growth acceleration, e.g., outside the US and in global manufacturing, as well as continued optionality on further economic benefits from AI this cycle. Near term, with elevated equity valuations, mixed macro momentum and rising policy uncertainty there is the risk of more equity drawdowns. As a result, risk-adjusted returns for equities are likely to be lower - this is consistent with higher equity drawdown risk, which points to lower equity returns and hit ratios near term (Exhibit 20). However, we think the risk of a bear market remains low with relatively low recession risk, helped by a healthy private sector and central bank easing.
Exhibit 20: Equity drawdown probabilities above 30% historically have resulted in lower risk-adjusted returns
Box highlights current implied-probability of S&P 500 max drawdown
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Better balance - 60/40 and risk parity portfolios had smaller drawdowns
Elevated equity drawdown risk points to the need for more risk management in multi-asset portfolios. With growth as the main driver of the 'risk off', 60/40 and risk parity portfolios have performed relatively well this summer - a US and European 60/40 portfolio had only a 4% peak-to-trough drawdown so far and again during the recent volatility they have been relatively unaffected (Exhibit 21). This is a major change relative to the last 2 years, when 60/40 portfolios had large drawdowns alongside equities with little buffer from bonds. In fact, risk parity strategies, which had one of their worst performances in a century in 2022, have performed particularly well YTD (Exhibit 22).
Exhibit 21: 60/40 portfolios have performed well in the summer
Relative total return performance indexed to 100
Source: Datastream, Goldman Sachs Global Investment Research
Exhibit 22: Risk parity strategies' performance has been particularly steady (outside Japan)
Weighted inversely by 3m realised volatility of equity and 10y bonds
Source: Datastream, Goldman Sachs Global Investment Research
Equity/bond correlations have shifted sharply negative in Q3, consistent with a shift from inflation to growth risks, similar to the US regional banks crisis early in 2023 (Exhibit 23). We think this time the negative equity/bond correlations are more likely to be sustained. Inflation has normalised materially since last year - from current levels of our inflation score, equity/bond correlations were usually more negative (Exhibit 24). Our rates team also has shown that bond markets are increasingly trading growth rather than inflation news. Since Q4 2023, our dynamic asset allocation (DAA) model also has suggested lower left tail risk for 60/40 portfolios, helped by low inflation scores and momentum levels but also already negative momentum in the policy score.
Exhibit 23: 3m rolling equity/bond correlation of weekly returns
Source: Datastream, Goldman Sachs Global Investment Research
Exhibit 24: Inflation normalisation supports more negative equity/bond correlations
Source: Bloomberg, Haver Analytics, Goldman Sachs Global Investment Research
Limited bond buffer - less negative equity beta for long-dated bonds
While bond yields are likely to decline in case of further growth shocks, already dovish Fed pricing and upward pressure on term premia might mean a smaller bond buffer from here. While bond yields have increased materially in the post COVID-19 recovery, they are still low compared to the start of previous S&P 500 bear markets - the average nominal yield was around 5.8%, real yields 2.3% and breakeven inflation around 3.5% (Exhibit 25). This points to a smaller buffer from bonds, especially if there were more reflationary policies following the US elections.
Exhibit 25: Bonds able to buffer equities better from here with higher yields and lower inflation
US 10y nominal yields, real yields and breakeven inflation around historical equity bear markets
Source: Bloomberg, Haver Analytics, Goldman Sachs Global Investment Research
Some of the 'Fed put' has already been pulled forward. Current cumulative priced Fed cuts of 235 bps for the next 12 months are already close to the average during a US recession outside the 1970s (Exhibit 26). Of course, the Fed has hiked rates unusually aggressively, but current levels of the Fed funds rates are comparable to late-cycle backdrops ahead of the Tech Bubble burst and the GFC - and our economists have argued that the neutral rate is likely higher. In addition, there might be continued upward pressure on longer-dated bond yields from bond term premia due to bond supply and fiscal policy (Exhibit 27).
Exhibit 26: Markets are already pricing a large amount of Fed cuts in the next 12 months
Source: Haver Analytics, Goldman Sachs Global Investment Research
Exhibit 27: Yield curves might steepen more with upward pressure on bond term premia
Source: Datastream, Goldman Sachs Global Investment Research
Thus, even if equity/bond correlations remain negative, the sensitivity of bonds to equities is likely to be much lower during drawdowns. Since the late 1990s, for a 10% decline in equities bonds rallied 3-4% - but especially with dovish Fed pricing and more yield curve steepening, US 10-year bonds might rally less in 'risk off' from here. Similarly, equity/bond betas were less negative last cycle in Europe and Japan due to proximity to the zero lower bound before the COVID-19 crisis (Exhibit 28). This time, the reverse might be true - our rates strategy team thinks that with incoming data gradually bringing relief on the health of the US economy, longer-dated rate differentials between Europe and the US might widen. In part this is due to relatively hawkish ECB pricing, but also with less fiscal expansion ahead. As a result, European bonds could provide a better cost/risk reduction trade-off from here.
Exhibit 28: The bond buffer in Europe was already lower before the COVID-19 crisis
1-year rolling bond beta to equities (weekly returns)
Source: Datastream, Goldman Sachs Global Investment Research
Less diversification desperation with inflation normalisation
A smaller prospective bond buffer points to more value in alternative safe havens. With more negative equity/bond correlations there have been more multi-asset diversification opportunities, much in contrast to the last few years (Exhibit 29). As we wrote in our Balanced Bear Research, with positive equity/bond correlations in the last 2 years, a lot of traditional safe havens have also stopped working. However, in the recent drawdown, with more negative equity/bond correlations, safe havens such as the Yen and Swiss Franc performed very well again - Gold lagged, in part due to elevated positioning. Also, there were some benefits from international equity diversification with Europe outperforming. However, due to growth being the source of risk and micro headwinds, cyclical commodities like oil and copper have provided little diversification benefit.
Exhibit 29: Bonds have provided a relatively good buffer from equities in the recent S&P 500 drawdown
Source: Bloomberg, Haver Analytics, Datastream, Goldman Sachs Global Investment Research
Many safe assets faced headwinds from rising real yields in 2022 and in part in 2023, similar to the early 1980s. Both the S&P 500 and Gold became more negatively correlated with US 10-year TIPS yields in 2022 and, as a result, correlation between both increased (Exhibit 30) - in growth-driven 'risk off' episodes, we would expect to see more diversification benefits from allocations to Gold. Our commodities team is also bullish on Gold as it believes it is both a hedge for financial and geopolitical risks with added support from Fed rate cuts and EM central bank buying (our team maintains its 2025 target of $2,700/toz).
As we have written before, in 2022 the Dollar became a core diversifier to protect 60/40 portfolios from US rate shocks and positive equity/bond correlations - in fact, the Dollar/S&P 500 correlation remained deeply negative in recent years, especially vs. procyclical FX (Exhibit 31). However, both Yen and CHF rallied sharply during the recent 'risk off', indicating a potential correlation shift. Our FX team has recently updated its FX hedge framework - short USD/JPY, USD/CHF and long USD/MXN remain its favoured recession hedges but believes this can be costly due to carry; shorts in AUD or GBP are also reactive to recessionary shocks, but are less dependent on oil price moves and low carry.
Exhibit 30: Gold became more correlated with equities in 2022 as both suffered from higher real yields
1-year rolling correlation (weekly returns)
Source: Datastream, Goldman Sachs Global Investment Research
Exhibit 31: Yen and Swiss Franc have become less 'risk off', the Dollar more so
1-year rolling correlation with S&P 500 (weekly returns)
Source: Datastream, Goldman Sachs Global Investment Research
Besides adding 'risk reducing' assets to a balanced portfolio, investors could also reduce risk within equities – after relatively poor performance since the COVID-19 crisis, we see more value in defensive equities (Exhibit 32). Most of those have underperformed global and US equities due to a larger weight in capital-heavy, value sectors that suffered more from rising rates due to more leverage - as a result, they often trade at large valuation discounts to the market. This has been particularly true for both real estate and infrastructure stocks, which saw additional headwinds from excess supply in commercial real estate and regulation. With more negative equity/bond correlations during 'risk off', we think these areas would see more relief from lower rates and can outperform again. Regional indices such as FTSE 100 and SMI (in USD) can provide broad exposure to these themes of high shareholder yield and low volatility, respectively. Our European strategy team has also highlighted its preference for UK stocks, helped by discounted valuations and potential policy tailwinds.
Exhibit 32: Defensive areas within equities have de-rated since the COVID-19 crisis
Total return performance (rebased to 100. Orange shade: S&P 500 drawdown > -10%)
Source: Datastream, Haver Analytics, Goldman Sachs Global Investment Research
Convexity comeback - growing value of options, prefer risky asset hedges
Higher equity drawdown risk due to growth shocks also increases the value of option hedges. During the August 5th drawdown, the VIX spiked to 65, its highest level outside the COVID-19 crisis and the GFC. While technical factors related to positioning and carry trades exacerbated the moves, it also indicates a potential regime shift in equity risk. In the last 2 years, vol of vol was relatively low and equity volatility has been relatively capped (Exhibit 33) - this was in large part because rate shocks were the main driver of equity drawdowns, which means slower drawdowns due to valuation de-rating with more dispersion within equities. Equity drawdowns due to growth shocks tend to be faster with higher correlations across stocks, which results in larger vol spikes and more vol of vol. This makes option hedges more valuable as portfolio overlays again, in our view - indeed, the VVIX has trended up since the beginning of August, indicating a convexity comeback (Exhibit 34).
Exhibit 33: Vol of vol has started to pick up from low levels in the post COVID-19 recovery
Volatility of S&P 500 1-month volatility (daily returns)
Source: Bloomberg, Goldman Sachs Global Investment Research
Exhibit 34: Implied vol of vol indicates more negative convexity in equities
VVIX (implied volatility of the VIX)
Source: Bloomberg, Goldman Sachs Global Investment Research
With the handover from inflation to growth risks, there should also be a continued increase of equity volatility relative to rates volatility (Exhibit 35). As we recently wrote, our model linking the relative implied volatility of rates (MOVE index) and equities (VIX index) to fundamental macro variables (financial conditions, economic growth, macro and policy uncertainty, dispersion in rates forecasts) still points to a continued reset in rates vs. equity vol. In addition, the lack of bond buffer (especially for longer-dated yields) means downside convexity on rates might be lower outside extreme negative growth scenarios, which would also materially weigh on equities. We continue to like selling upside on longer-dated US bonds (e.g., TLT calls) to partially fund the cost of buying equity puts - having said that, with the recent increases in implied equity vol we would wait for a reset to get a better entry point (Exhibit 36).
Exhibit 35: Equity volatility increased more than rates volatility during the recent volatility spike, but resettled again afterwards
Source: Bloomberg, Goldman Sachs Global Investment Research
Exhibit 36: Selling TLT calls can reduce the cost of buying equity puts
Cost as % of spot of selling a 3m 110% call on US 20y+ bonds (TLT) and buying a 3m 95% put on S&P 500
Source: Goldman Sachs FICC and Equities, Goldman Sachs Global Investment Research
Until a larger volatility reset, with current higher implied volatility, there is more value in risk reversals - 'risk off' skew in selective equities and FX as well as Gold remains relatively low (Exhibit 37). Limited upside to equities into year-end coupled with more downside convexity points to more value in being long downside skew, in our view. Exhibit 38 ranks assets which have a high beta to our RAI PC1 'Global growth' factor relative to their implied vol - commodity-related and European equity puts, including on DAX, and credit still screen relatively well. Credit spreads have lagged during the recent risk off - we still like HYG puts, especially in case Fed easing disappoints relative to expectations. In fact HYG remains near all-time highs and CDX HY spreads are near their tight, decoupling from recent VIX moves.
Exhibit 37: Options on US bonds, safe haven FX are pricing more risk-off asymmetry, less in the dollar, US equity, gold
5-year percentile of 3-month 25-delta implied vol and normalised risk-off/risk-on skew
Source: Goldman Sachs FICC and Equities, Goldman Sachs Global Investment Research
Exhibit 38: Vol on risky assets, e.g., DAX, CDS, looks attractive to hedge a growth shock
5y beta to PC1 “Global growth” / 3m 25-delta put implied vol. * call implied vol, - beta to PC1. Multivariate regression on PC1 and PC2
Source: Bloomberg, Goldman Sachs FICC and Equities, Goldman Sachs Global Investment Research
Appendix: Modelling equity drawdown risk
We use univariate and multivariate logit models to link the different macro and market variables to equity drawdown risk. We estimate both the probability of a >10% correction over the next 3 months and a >20% drawdown/bear market over the next 12 months. Exhibit 39 shows the t-stats and the pseudo R2 for the univariate logit models estimated across indicators. The low R2 indicates the difficulty of predicting equity drawdowns, at least based on individual variables alone. The precision measures the fraction of times when a high level of the probability has been followed by an equity drawdown – i.e., a higher precision means that the variable gives fewer false signals. However, the recall is the fraction of drawdowns which have been anticipated correctly.
Exhibit 39: T-stat and goodness-of-fit measures for the univariate logit models estimated on the maximum available history since 1950
Note: AUC is the area under the Receiver Operating Characteristic curve and measures the trade-off between the precision and the recall of each variable. The higher the AUC, the better the signal from that variable, with an AUC of 0.5 associated with an uncorrelated variable.
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
In order to combine variables in another model we run a multi-variate logit model on the variables grouped by type: business cycle scores, sentiment cycle indicators, valuations and leading growth indicators. The multi-variate model results in higher R2, closer to 0.2, as we show above (see Exhibit 17). Exhibit 40 shows that above probabilities of 30%, the precision of the model increases - this is particularly the case above 45%, which is when subsequently much more often than not a large equity drawdown followed.
Exhibit 40: Equity drawdown probabilities above 30% tend to have good predictive power for subsequent drawdowns
Source: Haver Analytics, Datastream, Goldman Sachs Global Investment Research
- 1 ^ In-line with the Conference board's 3Ds rule, we use the 6-month change in the LEI and LEI dispersion (for details see https://www.conference-board.org/topics/us-leading-indicators).
Investors should consider this report as only a single factor in making their investment decision. For Reg AC certification and other important disclosures, see the Disclosure Appendix, or go to www.gs.com/research/hedge.html.
Gallery
View all 40 exhibitsThis report is intended for distribution to GS institutional clients only.