GS Global Macroscope_watermark

paper, while the likelihood of negative returns rises as growth slows, it doesn’t

become high until GDP growth is below 1%. We should note that we assume in

Exhibit 17 that the market is pricing GDP growth 2 quarters before it happens.

Exhibit 17: At the current pace of GDP growth, the probability of negative returns is reasonably low

S&P 500 annual real total returns; 2Q lag for US GDP growth (since 1960)

Source: Datastream, Robert Shiller, Haver Analytics

Equally, without a recession, falling profits are also not very likely. As Exhibit 18 shows, all annual falls in profits (in Europe) have been in periods when either the domestic economies have been in recession or when the US has been in recession.

Exhibit 18: EPS rarely falls outside of recessions

SXXP annual realised earnings growth (plus our top-down strategy forecasts, light blue), grey shading indicates Euro area or UK recession, Orange indicators US-only recession

-10 -8

-11

-15

Source: FactSet, I/B/E/S, Datastream, Worldscope, Haver Analytics, Goldman Sachs Global Investment Research

But aren’t we more vulnerable now?

We see several areas where markets are potentially more vulnerable than in previous cycles:

1. Lack of policy flexibility

nThe US has already expanded fiscal policy, and its debt levels and budget deficit are rising, which could make it difficult to find room for significant easing.

nMeanwhile, the Euro area is in a much better position fiscally. According to IMF estimates, the Euro area budget will be in balance by 2022. Arguably, the Euro area countries would be able to stimulate their economies with a fiscal boost – but this would depend a great deal on the political mood in Germany, since it accounts for the bulk of the likely surplus.

nThere may be room for US interest rates to be cut in the next downturn but less so than in other downturns. Also, European interest rates may still be at or close to zero when the next US downturn hits. The same would be true for Japan.

It’s notable that, as rates have fallen – and in the case of the Euro area and Japan moved into negative territory – this has not pushed up equity valuations. Partly this is because negative rates have an adverse effect on Financials. It is also partly because low rates are a symptom of economic weakness, although we think the lack of policy room also has an impact – in Europe and Japan the gap between the dividend yield and 3-month interest remains wide, whereas it has fallen in the US (as rates have started to rise). So, even if we get a slowdown and no recession, the prospects of a coordinated and strong

policy boost globally is likely to be limited, suggesting any slowdown becomes more entrenched.

2. It’s the machines: change in the type of investing, with a large amount of investment in either passive or quant strategies (which can exacerbate market moves)

Our global asset allocation team finds that, while volatility has been relatively low in recent years, vol of vol has picked up since the 1990s. Also, longer periods of low volatility increase the risk of higher volatility in the medium term, as they can result in financial and market imbalances (see GOAL - Global Strategy Paper #28: A taste of the high (vol) life, 16 April 2018).

S&P 500 vol spikes have in recent years become both larger and faster. Exhibit 19 shows that the 10-year average volatility of S&P 500 1-month volatility has increased since the 2000s.The 5-year vol of vol has increased even more recently, indicating that this is not skewed by the global financial crisis (GFC).

The increase in vol of vol likely also reflects changes in the market microstructure, namely:

n Changes in bank regulation have reduced liquidity across assets.

n The rise of systematic investing, e.g., passive/smart beta strategies that reduce liquidity, can drive crowding.

n CTAs, risk parity and volatility target funds, which often invest very procyclically.

n Levered or liquidity-constrained exchange-traded products, e.g, on high yield credit or commodities.

n Short vol carry strategies, e.g., short VIX ETPs.

Exhibit 19: Vol of vol has trended up since the GFC

Volatility of S&P 500 1-month volatility (daily)

3. Low rates, concerns about growth have also created very clustered positions within equities

These clusters create risks that any correction driven initially by fundamentals could become more extreme in terms of market moves.

We can see this in two ways: (i) the performance of Growth versus Value companies has become extreme, as we show in Exhibit 20, and (ii) the valuation of high-growth companies has risen to an historically high premium versus low-growth stocks.

We see this concentration in the flows and positioning as well. ETFs now dominate investment flows into equity mutual funds globally (Exhibit 22). Most of these are index ETFs, such as SPX or SX5E, but some are sector ETFs, and for these, as we show in Exhibit 23, there has been a concentration of flows in recent years into certain sectors and far less so into others. Any major unwinding of these concentrated positions can add to market stress in the short term.

Overall, we expect to see less of a binary factor driven market in the future; the persistent support from QE and falling bond yields that supported these trends in the past are fading somewhat. Instead, we see a market that is more eclectic in terms of sector and factor leadership. But this also is consistent with a flatter profile of overall market returns, in our view.

The shape and profile of bear markets in the past

Bear markets historically have come in different shapes and sizes. No two bear markets are completely alike and the conditions that trigger them vary. In our analysis we have defined a bear market as a nominal price decline of 20% or more; while the average bear market has experienced declines of 35% over a period of 2 years, the standard deviation is very large indeed. In previous work (Share Despair, December 2002), we split historical bear markets into three categories, taking data for the US back to the 19th century (see Exhibit 25). The classification is not scientific. For the earlier bear markets in particular, we have made an assessment about the triggers.

Defining different types of bear markets

In general, we can split bear markets into three categories:

1.Cyclical bear markets - typically a function of rising interest rates, impending recessions and falls in profits. They are a function of the economic cycle.

2.Event-driven bear markets - triggered by a one-off ‘shock’ that does not lead to a domestic recession (such as a war, oil price shock, EM crisis or technical market dislocation).

3.Structural bear market - triggered by structural imbalances and financial bubbles. Very often there is a ‘price’ shock such as deflation that follows.

Exhibit 24: Characteristics of a bear market

Exhibit 25: US bear markets since the 1800s

S - Structural bear market, E - Event-driven bear market, C - Cyclical bear market

Source: GFD, Datastream, Bloomberg, Goldman Sachs Global Investment Research

By splitting bear markets into these groups we find that:

nCyclical and ‘event-driven’ bear markets generally see price falls of around 30%, while structural ones see much larger falls, of around 50%.

nEvent-driven bear markets tend to be the shortest, lasting an average of 8 months, cyclical bear markets last an average of 26 months and structural bear markets last an average of three and a half years.

nEvent-driven and cyclical bear markets tend to revert to their previous market highs after around 1 year, while structural bear markets take an average of 10 years to return to previous highs.

It is worth noting that these data are in nominal terms, whereas in reality the bear markets of the 1970s were more pronounced given the very high inflation at that time.

Exhibit 26: Structural bear markets tend to be deeper, longer...

Average drawdown of US bear markets since the 1800s

-30

-40

-50

-60

Decline (%)

-70

Source: Bloomberg, Goldman Sachs Global Investment Research

Exhibit 27: ... and take longer to recover

Average recovery of US bear markets since the 1800s

-30

-40

-50

-60

Decline (%)

-70

Source: Bloomberg, Goldman Sachs Global Investment Research

How much of a bear market is driven by earnings?

Exhibit 28 shows bear markets since the 1960s and the change in EPS during and around each bear market. On average, since 1960 EPS has fallen by just 6% during bear markets, but this is distorted by two things:

1.EPS does not typically fall (or falls very little) in event-driven bear markets – these are all about de-risking and hence a decline in valuation; they are not directly cyclically driven.

2.The actual period of decline in EPS does not coincide exactly with the dates of the bear market in prices; nor would we expect it to, given that we would expect equity investors to try to anticipate the cycle.

Removing event-driven bear markets gives an average EPS decline during these bear markets of 17%. Removing event-driven bear markets and looking at the entire decline around the bear market (taking into account that the precise timing of the EPS decline differs in each cycle) gives an average EPS fall of 38%.This fall is similar to the average price fall in bear markets since the 1960s (excluding event-driven bear markets) of 40%.

Exhibit 28: EPS around bear markets - EPS falls as much as prices in Cyclical & Structural Bear markets but the timing is different

S - Structural bear market, E - Event-driven bear market, C - Cyclical bear market

* Change in EPS during the actual Bear market

** This includes the entire period of EPS decline around each Bear market from peak to trough in EPS

Source: Bloomberg, Shiller, Goldman Sachs Global Investment Research

This suggests that bear markets (excluding event-driven) are largely EPS-driven, although they may appear to be P/E-driven at first, given that valuations may decline before EPS. Can we say anything on the timing of earnings declines? This is tricky. The average (ex event-driven) is for EPS declines to lag the start of the bear market by 5 months. Or, put another way, prices start to fall 5 months before EPS does. But the range is very wide. In 1973 it took 20 months, whereas in 1990 the decline in EPS started 13 months before the bear market. The most recent non-event-driven bear market, in October 2007, saw EPS start to decline 4 months before prices started to fall. We have described this in the past as an end of cycle ‘optimism phase’, where the market continues to rally contrary to the evidence that earnings have already peaked.

Some key drivers around previous bear markets