Equity Research - Goldman Sachs Asset Management

LIQUIDITY, VOLATILITY, FRAGILITY

ISSUE 68 | June 12, 2018 | 3:30PM EDTEquityResearchGlobal Macro Research

Goldman Sachs does and seeks to do business with companies covered in its research reports. As aresult, investors should be aware that the firm may have a conflict of interest that could affect theobjectivity of this report. Investors should consider this report as only a single factor in making theirinvestment decision. For Reg AC certification and other important disclosures, see the DisclosureAppendix, or go to www.gs.com/research/hedge.html. Analysts employed by non-US affiliates arenot registered/qualified as research analysts with FINRA in the U.S.

The Goldman Sachs Group, Inc.

From post-crisis “flash crashes” to the recent illiquid trading in Italian bonds, marketshave become familiar with a pattern: Liquidity seems to evaporate as volatility rises,arguably leading to price moves that seem excessive relative to their fundamentalcatalyst (if there is even one at all). Whether shifts in market structure and/orinvesting strategies could be generating this market fragility is Top of Mind. Weinterview GS and external experts on oft-blamed (but not necessarily guilty) reasonsfor such fragility, including high-frequency trading and exchange-traded products.Our interviewees’ perspectives differ, but they leave us with a key takeaway: Thefragmented structure of exchanges and market makers, coupled with investments

that demand near-instantaneous liquidity, suggests that periodic episodes of illiquidity are likely here to stay. Whetherone of these episodes will inadvertently lead to a more lasting market downturn is a risk worth watching.

I see a risk that markets are paying too little attentionto liquidity risk, much as they previously paid too littleattention to the risks posed by excess leverage.

- Charlie Himmelberg, Chief Markets Economist,Goldman Sachs

“If you’re a buy-side institutional investor comparingyour liquidity to, say, 15 years ago, you need to considernot only how much liquidity you were getting back then,but also what spreads you were paying and theinformation you were giving up.

- Doug Cifu, Co-Founder and CEO, Virtu Financial

“

INTERVIEWS WITH:

Doug Cifu, Co-Founder and CEO, Virtu Financial

Charlie Himmelberg, Chief Markets Economist, Goldman Sachs

Andrei Kirilenko, Director of the Centre for Global Finance andTechnology, Imperial College Business School

A DISCUSSION ON US MARKET STRUCTURE WITH GOLDMAN SACHS’:

Brian Levine, Co-Head, Global Equities Trading & Execution Services Raj Mahajan, Global Co-Head, Securities Systematic Solutions (S^3)Konstantin Shakhnovich, Global Co-Head, S^3

LOW VOL DOES NOT ALWAYS MEAN LOW RISKChristian Mueller-Glissmann and Alessio Rizzi, GS Multi-Asset Strategy Research

THE VIX SPIKE: CAUSES AND CONCERNSRocky Fishman, GS Equity Derivatives Research

WHAT’S INSIDE

of

Allison Nathan | [email protected]

TOPMIND

Marina Grushin | [email protected]

...AND MORE

David Groman | [email protected]

El

Goldman Sachs Global Investment Research 2

Top of Mind Issue 68

70

80

90

100

110

120

130

140

1988 1991 1994 1997 2000 2003 2006 2009 2012 2015 2018 2021

GS baselineIMFProposed fiscal measures

Macro news and views

US Japan Latest GS proprietary datapoints/major changes in views

• No major changes in views.

Datapoints/trends we’re focused on

• Several weeks of firm data pointing to a reacceleration ingrowth; our current activity indicator for May is now 3.8%.

• The stronger-than-expected May jobs report, bringing theunrounded unemployment rate to a 48-year low of 3.75%.

• Increased trade policy risks given the imposition of steel andaluminum tariffs and a negative turn in NAFTA negotiations;the tariffs so far should have only a modest macro effect.

Latest GS proprietary datapoints/major changes in views

• No major changes in views.


• The first negative GDP reading in nine quarters (-0.6% qoq ann.in Q1); however, we expect a return to positive territory in Q2.

• Falling unemployment and rising wages, though we believe it istoo soon to conclude that the labor market is overheating.

• A slight improvement in Japan’s Consumer Confidence Index inMay, marking the first month-over-month uptick in six months.

• A rebound in April retail sales following stagnation in Q1.

Flipping the switch US’ threatened vs. imposed tariffs by trading partner, $ bn

Heating up? Not necessarily Japanese unemployment rate, % yoy; wage growth, % mom (rhs)

Source: USITC, Goldman Sachs Global Investment Research. Source: MIC, MHLW, Goldman Sachs Global Investment Research.

Europe Emerging Markets (EM) Latest GS proprietary datapoints/major changes in views

• We lowered our Euro area GDP forecasts by 0.1pp to 2.2%for 2018 and 1.9% for 2019, given political uncertainty inItaly, a widening in sovereign spreads, and higher oil prices.


• The sizeable fiscal expansion proposed by Italy’s government.

• The ECB’s response to political developments and weakeractivity; we expect a “steady hand” approach and no changeto the Asset Purchase Programme at the June meeting.

• A larger-than-expected rebound in May Euro area core inflation.

Latest GS proprietary datapoints/major changes in views

• We added two 25bp hikes to our 2019 Reserve Bank of Indiaforecast on rising inflation expectations and capacity utilization;we now expect 100bp in cumulative hikes by end-2019.


• Generally healthy EM fundamentals in the face of a "stress test"of higher US rates, a stronger dollar, and higher oil prices.

• Argentina’s $50bn financing agreement with the IMF.

• Turkey’s rate hikes in May/June; we think rates are high enoughto slow the economy, though inflation may remain elevated.

A worrying expansion Italy government debt estimates for 2018-2021, % of GDP

Growth momentum still healthy GS EM Current Activity Indicators, % ann.

Source: Eurostat, Goldman Sachs Global Investment Research. Source: Goldman Sachs Global Investment Research.

0

50

100

150

200

250

300

Washingmachines

Solar cells andpanels

Steel andaluminum(Sec. 232)

China(Sec. 301)

Autos(Sec. 232)

NAFTAEUChinaRest of World

Threatened Imposed

NAFTAEU

ChinaRest of World

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

2.0

2.5

3.0

3.5

4.0

4.5

2014 2015 2016 2017 2018

Unemployment rate (lhs)

Monthly total wage (rhs)

Despite wage growth and falling unemployment, it's still too soon to conclude that the labor market is overheating

-10

-5

0

5

10

15

11 12 13 14 15 16 17 18

EM CAIIndia CAIRussia CAIBrazil CAIChina CAI

We provide a brief snapshot on the most important economies for the global markets

El



The market experienced yet another unnerving episode of illiquidity in the recent dramatic sell-off in Italian bonds. Unlike past “flash crashes”—think US equities in 2010, US Treasuries in 2014, or even the VIX spike earlier this year—this event clearly had a fundamental catalyst. Nevertheless, it marked the latest instance of liquidity seemingly evaporating as volatility rises, arguably leading to sharper price moves than warranted by the fundamental shock (if there is even one at all). Whether shifts in market structure and/or investing strategies could be generating this market fragility is Top of Mind.

So what are the oft-blamed (but not necessarily guilty) reasons for such fragility? Regulation seems to be front and center, including rules that have led to increased fragmentation of trading in US equities. Another possible driver is automation, in terms of both newer, “non-traditional” liquidity providers like high-frequency traders (HFTs) and the increasing popularity of algorithmic investing strategies. Finally, the rising use of Exchange Traded Products (ETPs) has also been thrown into the mix of potential culprits. (See pg. 18 for a glossary of relevant lingo.)

We first discuss these potential drivers of fragility with GS Chief Markets Economist Charlie Himmelberg. Of the above, he worries the most about the liquidity impact of HFTs—firms that typically post (and then cancel) orders to buy and sell in very quick succession. While Himmelberg believes HFTs have generally improved liquidity conditions during normal periods, he worries that their disadvantage in processing nuanced fundamental information may lead them to pull back in moments of uncertainty, triggering surprisingly large drops in liquidity that exacerbate price declines. In most cases so far, this type of liquidity collapse has proven short-lived. But he warns that if investors mistakenly attribute a technically-driven crash to worsening fundamentals, this could inadvertently reinforce market fears, potentially leading to a more severe downturn.

We continue the conversation with three experts from the Goldman Sachs Securities Division: Brian Levine, Co-Head of Global Equities Trading and Execution Services, and Raj Mahajan and Konstantin Shakhnovich, Global Co-Heads of Securities Systematic Solutions (S^3). They offer somewhat different perspectives on US market structure: Like Himmelberg, Levine worries about volatility generating price dislocations in a negative feedback loop in US equities markets. However, he doesn’t think such episodes have much to do with HFTs. Instead, he, Mahajan, and Shakhnovich largely attribute current fragility to the market’s complex structure. They point out that regulations like the 2005 Regulation National Market System (Reg NMS), which left exchanges competing mainly on price, may have unintentionally encouraged the fragmentation of execution venues and market makers while prioritizing speed over reliability and safety. The outcome is what Levine calls a genuine flash crash: When volatility rises, the market “breaks,” with participants facing inaccurate data and sharp price swings.

That said, Mahajan and Shakhnovich do flag one vulnerability they see among at least some HFTs—namely, low

capitalization relative to the value of open orders throughout the day. In their view, this can create operational risk—distinct from market risk—that could turn systemic in the event of a problem (e.g., a bug in an HFT’s system), given the large notional some HFTs trade and the mutualized central counterparty clearing risk in the equity markets.

Andrei Kirilenko, Director of the Centre for Global Finance and Technology at the Imperial College Business School, agrees that a technological glitch could be disruptive. However, he doesn’t believe that HFTs are necessarily a source of systemic or market risk; rather, he considers them highly skilled risk managers. In addition, his research suggests that HFTs were pretty much the only market participants that continued to supply liquidity throughout the 2010 equity flash crash, though they also added to volatility along the way.

With all of that in mind, we sit down with Doug Cifu, CEO of technology-driven market maker, Virtu, to learn how firms like his operate and what they make of liquidity conditions today. His key messages: Virtu—like market makers of all stripes—is just looking to earn the spread on a trade; the firm may adjust its liquidity provision in response to market conditions (again, like all other market participants) but it won’t shut off unless there are signs of a technological problem. Cifu also pushes back on the idea that volatility means markets are “malfunctioning.” Case in point: Markets worked properly the day of the VIX spike on February 5th (i.e., there were no post-trade reconciliation or settlement issues, etc.).

Of course, it isn’t only liquidity provision that has changed with technology and regulation. Liquidity consumption has evolved as well; hence the focus on the other possible sources of market fragility mentioned above—the rise of systematic trading and the proliferation of passive products like ETPs. For instance, GS Multi-Asset Strategists Christian Mueller-Glissmann and Alessio Rizzi note that some systematic strategies increase equity exposure when vol is low, potentially setting the stage for greater liquidity demand if/when vol rises. And while they (and Himmelberg) consider “plain vanilla” ETPs fairly benign, they caution that the more complex, levered variety can become “forced sellers” when prices decline.

Indeed, inverse and levered long VIX ETPs were at the heart of the February VIX event, as GS Equity Derivatives Strategist Rocky Fishman explains. He notes that those ETPs are unlikely to cause a similar episode anytime soon. But he also points out that growth in investments that demand near-instantaneous liquidity runs the risk of occasional liquidity mismatches. Our key takeaway from all of the above: In today’s highly fragmented market structure—so far untested by recession or crisis—whether episodes of illiquidity lead to a more lasting market downturn is a risk to watch.

Allison Nathan, Editor Email: [email protected] Tel: 212-357-7504 Goldman Sachs and Co. LLC

Liquidity, volatility, fragility

mailto:[email protected]

El



Charlie Himmelberg is Chief Markets Economist and Head of Global Markets Research at Goldman Sachs. Below, he raises the possibility that the growth of high-frequency trading (HFT) has increased the “fragility” of liquidity supply, leaving the market more prone to non-fundamentally driven price volatility as a result.

Allison Nathan: You’ve argued that today’s market structure shows signs of fragility. Why?

Charlie Himmelberg: Over the last several years, we’ve seen flash crashes happen in a number of major, liquid markets. It isn’t hard to find examples of price swings that seem excessive relative to the fundamental

news that triggered them, if there was even any fundamental news at all. Just last month, two-year Italian government bonds experienced a much bigger price decline than they had at any point during the 2011-2012 crisis when systemic risks to the Euro area were far more acute. And potential “cracks” in the market were also revealed during the VIX spike on February 5th.

I see two main takeaways from such events. First, technical dislocations are risky because they can fuel false fundamental narratives. On February 5th, traders familiar with the VIX could tell that the spike was a technical disruption and nothing more, yet many observers tried to attribute the move to the prior Friday’s slight upside surprise in average hourly earnings data. That narrative didn’t get much traction with market participants, but in a future dislocation under less benign macro conditions, it could, and might thus fuel even larger price declines.

The second point is that when market liquidity evaporates, selling pressure is amplified. And growing frequency of flash crashes during the post-crisis period provides some reason to worry that liquidity may have become less reliable—or more “fragile”—during market stress events. For over-the-counter (OTC) markets, I think it is widely understood that the quality of market liquidity has declined due to the new costs imposed on traditional broker-dealers by the post-crisis regulatory environment. For exchange-traded markets, by contrast, such claims are more controversial. But I see reasons to worry that the rapid growth of liquidity supply from newer market participants like HFTs has a “feast or famine” quality to it. In normal times, machines are more nimble than human traders, and liquidity is great. In distressed periods, by contrast, evidence suggests that HFTs may withdraw liquidity more aggressively than human traders, and the resulting reduction in liquidity can interact with selling pressures in a way that results in outsized price declines.

So that’s the fragility risk that I’ve been trying to explore. It feels particularly relevant in the current environment, where recession risk is low but valuations are high. The biggest risk to markets might therefore be markets themselves.

Allison Nathan: Let’s take a step back. How are you defining HFTs in this context?

Charlie Himmelberg: I define HFTs as algorithms that aim to profit by supplying liquidity, which in practice means posting a bid and an ask on both sides of the market. HFTs do this at

incredibly high speeds, responding in micro seconds to changes in order flow. By doing so, HFTs both facilitate price discovery and earn a tiny premium for their willingness to post two-sided prices.

What my definition doesn’t include are things like managed volatility strategies or other kinds of lower-frequency quant strategies that generate a profit for doing something other than supplying liquidity. These strategies are more accurately takers of liquidity, and while there is also a risk that some of them could drive outsized price declines via automatic selling, my focus is on liquidity supply. My concern is that under certain types of market distress, HFTs may be less reliable than human traders.

Allison Nathan: How does this HFT liquidity withdrawal play out, and what evidence is there for it happening?

Charlie Himmelberg: Academic research on flash crashes has demonstrated two things. For one, in many flash crashes, trade-level data show that HFTs pull back from the market. That hasn’t been the case in every flash crash, so I don’t want to over-generalize. But other research shows that HFTs tend to withdraw liquidity when they’re faced with complex fundamental news. The recent political headlines in Italy are a good example. In circumstances like these, HFTs appear to recognize that their algorithms cannot understand or process the information as well as human traders. This puts HFTs at a disadvantage since it exposes them to adverse selection by those better-informed buyers or sellers. In response, it only makes sense for the HFTs to widen out their bid-ask spreads, perhaps dramatically; in some cases they withdraw from the marketplace altogether. And while evidence shows that liquidity typically returns quickly, the logic of adverse selection obviously suggests the risk that it could take longer under conditions less benign than the current expansion.

When HFTs withdraw, it leaves a greater share of liquidity provision up to human traders, which would be fine if there were a deep bench of human traders with spare risk-taking capacity. But as I alluded to earlier, there isn’t. Over the past ten years, a variety of factors—regulatory constraints, low volatility, low returns, advances in technology—have contributed to the so-called “rise of the machines.” As a result, there is simply a smaller base of human traders who can step in when the HFTs pull back. And in the case of traditional broker-dealers, new regulations have put tighter limits on how much risk capital they can deploy when they do.

Allison Nathan: Aren’t you being unfair to machines? Don’t humans withdraw liquidity during downturns, too?

Charlie Himmelberg: Of course. These issues are not unique to HFTs. In the old days, traditional market makers would stop answering the phone or take bathroom breaks during scary market events. Liquidity always dries up during a big price decline and/or complex news headlines.

Interview with Charlie Himmelberg

El



Having said that, some research shows that HFTs withdraw liquidity more aggressively than human traders during certain types of market events. That may be because the HFTs are just faster at doing what any human trader would eventually do, too. But I also suspect it has to do with the fact that human traders have the informational advantage. Human traders also have ongoing relationships with their clients, and therefore face greater reputational consequences for failing to post a bid in an illiquid market than an anonymous machine; taking risk during periods of market illiquidity may be paid back with future trades in better times.

Allison Nathan: What makes you confident that human traders can understand complex information any better than sophisticated HFT algorithms?

Charlie Himmelberg: Well, I don’t think HFT algorithms are necessarily all that sophisticated; they’re mostly just fast. And under routine circumstances, the evidence suggests that this substitution of “speed for capital” allows machines to supply liquidity more cheaply than humans. But when the markets experience a big negative event and price discovery breaks down, I would argue that traditional market makers are more sophisticated in their ability to determine where prices will settle. They are able to determine when the markets have overshot that level, take principal risk accordingly, and thus stabilize price discovery. By contrast, HFT algorithms can process price and volume, but not value.

Allison Nathan: Are some markets more vulnerable to flash crashes than others?

Charlie Himmelberg: Investors typically associate liquidity problems with credit markets. I think they also assume that markets like S&P futures, Treasury futures, or FX futures are basically without liquidity issues. And generally speaking, I think that’s true. But these are also the markets where HFT has gained the greatest market share. So if you worry like I do that HFT has made liquidity supply more fragile, then you worry that the risk is highest where it is least expected.

Allison Nathan: Your focus is on HFTs. But lower-frequency quantitative strategies and passive investments more broadly often raise questions about crowding and other risks. Do you worry that these strategies could be contributing to market fragility?

Charlie Himmelberg: Well, there are certainly some well-known events where quantitative strategies became crowded and then abruptly reversed in a very short period of time. But any investing strategy can be prone to crowding, whether human or quant. The liquidity risk posed by any strategy is caused by the failure to account for liquidity conditions, including the risk that they could deteriorate meaningfully.

Allison Nathan: What about products like ETFs? Could they be a source of market fragility?

Charlie Himmelberg: I take a relatively benign view of ETFs. Some ETF critics question the idea that you can take an illiquid portfolio of underlying assets, put an ETF wrapper on it, and come up with something more liquid. To some, this seems like black magic. But to me, that’s how financial markets have always enhanced liquidity. The same thing happens routinely with corporate capital structures. Take a steel mill. Its underlying assets, the blast furnaces, are highly illiquid. Yet the

financial claims on the cash flows from those furnaces, in the form of debt and equity, trade with plenty of liquidity. Similarly, ETFs make it easier for a wider base of investors to trade the stream of cash flows from harder-to-trade portfolios of assets. Now, that’s somewhat disruptive, because in some markets it creates more price discovery than many investors—particularly buy-and-hold investors—are accustomed to. ETFs may also place more demands on market liquidity since that is part of their appeal to end users. But I think the basic product design is sound. In distressed markets the price of an ETF can deviate materially from the price of its underlying assets, but such pricing differentials are self-stabilizing, and I think the net effect of ETFs on market liquidity is additive.

All of that said, it’s important to distinguish between the vast majority of “plain vanilla” ETFs from levered or inverse ETFs. The latter make up only a narrow slice of the market, but they have more worrisome features to the extent they can become “forced sellers” in the face of price declines—as happened for certain VIX products in February.

Allison Nathan: How does this all fit into your “liquidity is the new leverage” narrative?

Charlie Himmelberg: That analogy is meant to invoke the potential unrecognized problems or imbalances that build up over the course of long expansions. Financial leverage was obviously the imbalance that built up during the pre-crisis period, but that has been contained in the current cycle. In this cycle, there have been dramatic shifts in the way that secondary markets source liquidity, but this market structure has not yet been stress-tested by a recession or major market event. I therefore see a risk that markets are paying too little attention to liquidity risk, much as they previously paid too little attention to the risks posed by excess leverage.

Allison Nathan: Does all of this amount to a tradeoff between liquidity risk and the efficiency gains associated with HFTs?

Charlie Himmelberg: Maybe. Some people see it as an acceptable tradeoff, since the markets work well in normal times, and most of the flash crashes have lasted less than a day. In fact, one could argue that HFTs are enabling a faster bounce back from flash crashes, in contrast to the 1987 crash, when it took a year for the markets to recover. But I don’t think we will be able to evaluate the tradeoff without another 10 or 20 years of experience with this new market architecture. The question is whether we're running a risk of frequent but ultimately benign flash crashes, or a risk that one of these flash crashes will interact negatively with other dynamics and cause a sustained price decline.

Allison Nathan: So how worried should we be?

Charlie Himmelberg: To be clear, I‘m just trying to flag a tail risk. But I see a real possibility that the rapid growth of HFT liquidity supply has increased market fragility. I think investors should factor this risk into their hedging strategies, perhaps diversifying into deeper out-of-the-money strikes in the more liquid markets. And as the recent price action in Italian bonds arguably illustrates, investors may also want to give more scrutiny to the role of fundamentals vs. technicals when interpreting price action.

El



Doug Cifu is the co-founder and Chief Executive Officer of Virtu Financial. Below, he argues that technology has helped democratize trading for all participants, driving down friction costs. The views stated herein are those of the interviewee and do not necessarily reflect those of Goldman Sachs.

Allison Nathan: How has the market-making landscape evolved, and where does Virtu fit into it?

Doug Cifu: Our history follows the evolution of the markets. My co-founder, Vinnie Viola, started out as a market maker in the pits of a commodity futures exchange. What he did back then—transfer risk

between buyers and sellers—is still what Virtu does today. We don’t take a directional view on an asset; we simply look to earn the spread on the trade. But instead of doing this by yelling at each other in trading pits or over the phone, we saw that trading could be done better through technology and automation, which could substantially reduce the friction costs incurred as natural buyers and sellers try to find each other. And, indeed, as the market has embraced technology, spreads have collapsed: Ten years ago, the bid-ask spreads in the FX and equities markets were 5x to 10x what participants receive today. At the same time, technology has enabled competition, and the number of platforms where market participants can transact has grown significantly. All of this has forced out less efficient players and created opportunities for firms like ours. As “non-traditional” market makers, we’re providing services that 20 years ago only the bank broker-dealers would provide.

Allison Nathan: Hasn’t regulation been an important driver of the evolution of market making?

Doug Cifu: That is a great point. Technology was necessary but insufficient; you had to have both changes in regulation and technological advancements, which happened to occur at around the same time.

On the regulatory front, pressure from buy-side and retail investors frustrated with the existing rules helped bring about Regulation Alternative Trading System (ATS) and the Order Handling Rules in the late 1990s, and then decimalization and Regulation National Market System (NMS) in the early/mid-2000s. Combined, these new rules helped create more transparent, competitive, and fair markets. With the traditional firms no longer able to control flows, companies like Virtu stepped into fill the vacuum in liquidity provisioning—again, adapting the age-old concept of earning a spread to a more technology-driven environment.

The result has been efficiency gains that benefit users. They’re the reason why you can make a trade online today at $4.95 that 20 years ago might have required a phone call and cost $300. Or, why, as an institutional investor, you can now look at pre-trade costs and decide to trade your own order or to bid it out. Historically, you only had one choice and you paid for it.

Allison Nathan: Despite the increased efficiency, a small number of market makers now handle the majority of volume in some markets. What should we make of that?

Doug Cifu: You are right; there are about seven or eight firms that compete for retail order flow in the US equity market right now. However, it is a very competitive market with objective measures of performance. The barriers to entry may be lower, but I actually think the barriers to successfully operating in this market may be higher. We have to stand ready to trade all NMS names, which is hard not only in terms of breadth, but also in terms of managing idiosyncratic risk. So there aren’t a ton of brokerage firms that choose to get into the game. Those that do participate have to compete to earn clients’ trades, and that competition has yielded enormous benefits to the retail investor, like low costs, immediate fills on orders, and price improvement of over $1bn last year.

Allison Nathan: So how much risk do you actually bear day-to-day?

Doug Cifu: Until our acquisition of KCG Holdings last year, we didn’t have direct customers. Instead, we provided a transparent, two-sided price to exchanges, alternative trading systems, banks, etc. In that business, our intraday and inter-day portfolio value at risk (VAR) was, and still is, essentially zero.

Since the KCG acquisition, we now have retail customers, and interact with order flow from firms like TD Ameritrade, Schwab, and Fidelity. We also deal with institutional investors in some capacities. By definition, this means we are taking on more risk and therefore need more capital. Providing a bid on a $100 million block of an ETF to a customer is obviously very different from quoting it on an exchange. That said, even in the retail business, the highest I’ve ever seen our intraday VAR is less than eight figures. The risk managers at Goldman Sachs would probably chuckle at that. But to me, it’s a big change. We are taking significant positions in single stocks and ETFs from our retail customers, and our risk management and hedging have evolved to reflect that. Additionally, through our integration of KCG, we are adapting and deploying short- and medium-term trading strategies to the Virtu framework as they are in-line with our market making objectives and improve internalization of the risk.

Allison Nathan: You mentioned the need for greater capitalization in light of greater market risk. Are non-traditional market makers sufficiently capitalized to address operational risk, given their larger role today?

Doug Cifu: When people think of traditional market makers, they often think of the Wall Street banks, which were—and still are—great at things like lending their balance sheet, financing complex over-the-counter (OTC) derivatives, and prime brokerage services. Unlike a bank, Virtu’s focus is wholly on providing the most competitive bid and offer in transparent, liquid, centrally cleared financial instruments. This focus gives us a high degree of capital efficiency that we pass on in the form of tighter spreads and higher capital returns to shareholders.

Interview with Doug Cifu

El



Allison Nathan: What do you see in periods of market stress, and how do you respond?

Doug Cifu: When the markets are stressed, our automated risk controls enable us to continue providing liquidity on the way down and on the way back up. Our models are tuned to keep us in the market, and, yes, some trades lose money. At one point on February 5th of this year, 80% of our orders were sell orders from the retail side. That desk lost an entire day’s P&L in about five minutes, but we kept trading the whole time. We can’t just say to our customers—especially our retail customers—“I’m not there.” So the idea that we would just shut off is wrong. In fact, during the Treasury flash crash in October 2014, it was Virtu and other nimbler participants who stayed in the market when others pulled out.

That said, as the markets move, we do make adjustments. We monitor our profitability and supply/demand imbalances in the markets, and adjust by increasing or decreasing the size and width of the passive liquidity we provide on the bid and offer. Our bid-offer has to reflect whatever price the market requires to absorb order flow at that moment. So if SPY spreads go from a penny wide—which they are 99.9% of the time—to a nickel wide, we are not going to price SPY at a penny, right? At the same time, if we are not providing the best price, we could miss a trade, which would end up going to one of our competitors. So there are times when we have to pretty much take it on the chin, especially on the retail side.

Allison Nathan: Have you ever had to withdraw from the market altogether?

Doug Cifu: The only time we’ll actually shut off is if we think there's a technology issue. If a strategy is losing or making an unexpected amount of money, our risk controls automatically “lock-down” the suspect strategy or symbol and alert our risk managers. Or if the prices we're calculating are very different from the market data, we may assume that something is broken and decide to stop trading that instrument to investigate the alerts. For example, on August 24, 2015, our risk controls triggered lock-downs in certain ETFs because we made five figures in a very short amount of time, when we would normally make only four figures for an entire day. The net asset value we were calculating on those ETFs looked very different from the market data. Those are the types of instances where our lock-downs will automatically halt our trading and market-making activity.

Allison Nathan: Do you see an underlying liquidity problem in the markets?

Doug Cifu: That depends. The vast majority of trading days are uneventful from a liquidity point of view. However, occasionally we see days when there will be less liquidity at the best bid-offer and through the various levels of the order book, and occasionally days when liquidity in a particular instrument will drop 30-40%. If the market is down significantly, that decline in liquidity shouldn’t come as a surprise. As spreads have narrowed, the risk-reward for market makers to provide meaningful size has changed. So if you’re a buy-side

institutional investor comparing your liquidity to, say, 15 years ago, you need to consider not only how much liquidity you were getting back then, but also what spreads you were paying and the information you were giving up as a result of having fewer, human outlets. Again, as a market maker, we are always two-sided, and we don’t shut off unless we suspect something is broken.

Allison Nathan: Many commentators seemed to view the volatility on February 5th as evidence that something is wrong with the markets. It seems like you disagree?

Doug Cifu: You’re right. Are the markets perfect? No, but it’s generally agreed that they functioned properly on February 5th. Was there a lot of stock that got traded? Was it priced correctly for the orders that were in the book at that time? Did those orders get filled? Yes, yes, and yes. None of our national equity exchanges declared self-help, all of the major dark pools continued to function, and there were no post-trade reconciliation or settlement issues—as far as I’m aware. To me, we should celebrate that, instead of running around with our heads on fire as if the market was broken. Now, if people were unhappy because of how much the market fell, then they should have bought more stock that day. Nobody ever calls me to complain when the market climbs hundreds of points really quickly and experiences some liquidity bulges on the way up.

Allison Nathan: What problems—if any—do you see in today’s market structure?

Doug Cifu: There are 13 US equity exchanges and dozens of alternative trading systems or dark pools. Although Reg NMS had many good outcomes, like connecting the markets, it ultimately increased the complexity of the system. And because institutional investors still don’t want to send large orders to exchanges and thereby reveal their intentions to the market, Reg NMS ended up propagating other forms of off-exchange trading. I think it would help to have some revisions to Reg NMS that might lead to some consolidation and increase transparency around brokers’ routing decisions, while also ensuring that competition among trading venues remains robust.

Allison Nathan: How do you see market structure and the technology behind it evolving from here?

Doug Cifu: I think the markets will continue improving to the benefit of users and customers. Despite some observers’ worries about where market structure is going, I would humbly suggest that it is not so different from 30 years ago. It’s just more egalitarian and transparent. And as far as technology, the markets will continue to move at very rapid speeds; I suppose transacting at picoseconds is next. The time horizons will vary; mine might be seconds, hours, or days, while a portfolio manager’s might still be a few years. But at the end of the day, we will both continue to make informed judgments about market activity, even if we are using different tools to do it. And retail investors will remain better positioned from a cost and liquidity standpoint than before.

El



Algorithmic trading has grown across asset classes… Market share of algorithmic trading by asset class, %

…and regions (though the US still leads the way) Market share of algorithmic trading by region, %

Source: Aite Group, Goldman Sachs Global Investment Research. Source: Aite Group, Goldman Sachs Global Investment Research.

Algorithmic trading has become more popular across products within futures markets Market share of algorithmic trading in futures markets, %

Source: CFTC, Goldman Sachs Global Investment Research.

Market depth drops around fundamental news events… Number of contracts around news events

…as HFTs withdraw liquidity HFT share of market volume, %, around news events

Source: Eurex, Hautsch, Nikolaus and Noé, Michael and Zhang, S. Sarah (2017). Source: Eurex, Hautsch, Nikolaus and Noé, Michael and Zhang, S. Sarah (2017).

Futures

FX

Options

Fixed Income

Equities

0%

10%

20%

30%

40%

50%

60%

70%

2004 2006 2008 2010 2012 2014 2016

U.S.

Asia

LatAm

Europe

0%

10%

20%

30%

40%

50%

60%

70%

2004 2006 2008 2010 2012 2014 2016

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

PreciousBase

US TreasuryStirs

Deliverable SwapsG10

Emerging MarketsE MicrosUS Index

Select Sector IndexInternational Index

Refined ProductsNatural Gas

Crude OilBiofuelsLumber

LivestockGrainDairy

Commodity Index

Met

als

Rat

esFX

Equ

ities

Ene

rgy

Agr

icul

ture

2013-20142015-2016

50

70

90

110

130

150

170

190

210

230

250

-30 -15 +15 +30Time (minutes)

95% Confidence IntervalsMarket Depth of HFTs (average time)

Mean outside of macro events

News arrival

Total Market Depth

30%

35%

40%

45%

50%

55%

60%

65%

-30 -15 +15 +30Time (minutes)

95% Confidence IntervalParticipation Rate of HFTs (average time)

Mean outside of macro events

News arrival

Machines in the markets Special thanks to Charlie Himmelberg and James Weldon. For more, see Liquidity as the New Leverage: Will the Machines Amplify the Next Downturn?, 22 May 2018.

https://research.gs.com/content/research/en/reports/2018/05/22/93255efb-5656-401f-a85d-70e4d8e71769.html

El



Whi

le th

ere

are

man

y w

ays

to d

efin

e a

“fla

sh c

rash

,” s

uch

even

ts a

re o

ften

char

acte

rized

by

a la

rge

and

abru

pt m

ove

in p

rices

that

app

ears

exc

essi

ve r

elat

ive

to fu

ndam

enta

ls—

ofte

n w

ith n

o ob

viou

s ca

taly

st—

follo

wed

by

a sw

ift re

cove

ry. B

elow

, we

prov

ide

a sn

apsh

ot o

f fou

r pro

min

ent f

lash

cra

sh e

vent

s.

US Eq

uity F

lash C

rash (

May

6, 20

10)

Ch

ang

e to

tr

ou

gh

/pea

k*

Ap

pro

xim

ate

du

rati

on

of

even

t**

-6%

40 m

ins

A L

ondo

n-ba

sed

trad

er e

ngag

ed in

“sp

oofin

g” in

an

atte

mpt

to

man

ipul

ate

the

pric

e of

S&

P E

-min

i fut

ures

. La

ter

in t

he

sam

e da

y, a

mid

alre

ady

heig

hten

ed v

olat

ility

, a

larg

e E

-min

i sel

l ord

er t

rigge

red

a se

ll-of

f in

the

con

trac

ts;

the

algo

rithm

us

ed f

or t

he o

rder

was

bas

ed o

n m

arke

t vo

lum

e, r

esul

ting

in r

elat

ivel

y ra

pid

exec

utio

n an

d an

inte

nsifi

catio

n of

sel

ling

pres

sure

as

volu

mes

inc

reas

ed.

E-m

ini

and

equi

ty p

rices

dro

pped

sha

rply

and

liq

uidi

ty d

eter

iora

ted,

cre

atin

g pr

ice

dist

ortio

ns u

ntil

mar

ket p

artic

ipan

ts s

tepp

ed b

ack

in.

US 10

-Yea

r Trea

sury

“Flas

h Rall

y” (O

ctobe

r 15,

2014

)-3

1bp

1.6 ho

urs

A w

eake

r-tha

n-ex

pect

ed U

S re

tail

sale

s re

port

trig

gere

d an

initi

al d

eclin

e in

10-

year

Tre

asur

y yi

elds

. H

owev

er,

the

data

pr

int

was

soo

n fo

llow

ed b

y a

surg

e in

tra

ding

vol

umes

and

dw

indl

ing

orde

r bo

ok d

epth

, dr

ivin

g yi

elds

dow

n sh

arpl

y.

Sub

sequ

ent

rese

arch

fou

nd t

hat

high

-fre

quen

cy t

radi

ng (

HFT

) an

d “s

elf-

trad

ing”

—in

whi

ch t

he s

ame

entit

y ta

kes

both

si

des

in a

n ex

chan

ge—

appe

ared

to

be e

leva

ted

durin

g th

e ev

ent.

The

ral

ly u

ltim

atel

y re

vers

ed, c

oinc

idin

g w

ith a

num

ber

of la

rge

sell

orde

rs, b

ut li

quid

ity—

as m

easu

red

by m

arke

t de

pth—

rem

aine

d w

ell b

elow

long

-ter

m a

vera

ges.

Sterl

ing “F

lash E

vent”

(Octo

ber 7

, 201

6)-9

%

20 m

ins

Ste

rling

dep

reci

ated

ver

sus

the

US

dol

lar

in H

ong

Kon

g’s

early

mor

ning

. D

espi

te t

he l

ack

of a

cle

ar c

atal

yst,

the

cu

rren

cy’s

dec

line

seem

s to

hav

e ac

cele

rate

d as

inv

esto

rs a

ttem

pted

unw

ind

Ste

rling

hol

ding

s to

hed

ge o

ptio

ns

posi

tions

. In

tur

n, li

quid

ity f

ell s

harp

ly in

the

spo

t m

arke

ts,

and

futu

res

trad

ing

cam

e to

a t

empo

rary

hal

t. S

terli

ng s

oon

retr

aced

mos

t of

its

intr

aday

loss

es, t

houg

h bi

d-of

fer

spre

ads

rem

aine

d w

ide.

Whi

le t

he e

vent

’s c

ause

rem

ains

unc

erta

in,

obse

rver

s no

te t

hat

Ste

rling

may

hav

e be

en v

ulne

rabl

e du

e to

a s

hallo

w o

rder

boo

k an

d al

read

y ill

iqui

d tr

adin

g co

nditi

ons

give

n th

e tim

e of

day

and

the

occ

urre

nce

of b

ank

holid

ays

in C

hina

and

oth

er p

arts

of

Asi

a.

VIX

Spike

(Feb

ruary

5, 20

18)

+83

%

2 hou

rs Th

e S

&P

500

expe

rienc

ed it

s la

rges

t on

e-da

y de

clin

e in

mor

e th

an s

even

yea

rs, t

rigge

ring

a sh

arp

rise

in t

he V

IX a

nd V

IX

futu

res;

the

VIX

’s 2

0-po

int

spik

e on

Feb

ruar

y 5

repr

esen

ted

the

larg

est

one-

day

mov

e in

the

inde

x’s

hist

ory.

Thi

s se

t of

f a

feed

back

loo

p, c

ausi

ng i

ssue

rs o

f V

IX e

xcha

nge-

trad

ed p

rodu

cts

(ETP

s) t

o bu

y m

ore

VIX

fut

ures

con

trac

ts i

n or

der

to

mai

ntai

n th

e re

latio

nshi

p be

twee

n th

e va

lue

of t

heir

port

folio

s an

d th

e V

IX f

utur

es t

hey

trac

k. G

iven

the

ETP

s’ le

vera

ge

and

larg

e si

ze r

elat

ive

to t

he V

IX f

utur

es m

arke

t at

tha

t tim

e, t

heir

buyi

ng a

ctio

n dr

ove

the

pric

e of

VIX

fut

ures

—an

d le

vels

of

over

all v

olat

ility

—ev

en h

ighe

r (fo

r m

ore

deta

il, s

ee p

gs. 1

6-17

).

1050

1075

1100

1125

1150 1:

45 P

M2:

15 P

M2:

45 P

M3:

15 P

M3:

45 P

M

S&

P 5

00E

vent

w

indo

w

1.8

1.9

2.0

2.1

2.2 7

AM

9 A

M10

AM

US

10-

year

Tre

asu

ry, %

1.13

1.18

1.23

1.28 12

:00

AM

7:07

AM

7:24

AM

8:20

AM

GB

P/U

SD

1419242934

7 A

M10

AM

1 PM

4 PM

9 A

M12

PM

3 PM

Feb

/Mar

VIX

Futu

res

(wei

gh

ted

avg

.)

Feb

6Fe

b 5

Flash crash course

*Cha

nge

to tr

ough

/pea

k is

cal

cula

ted

betw

een

the

star

ting

poin

t an

d m

inim

um/m

axim

um in

the

sha

ded

even

t w

indo

w. T

reas

ury

calc

ulat

ion

refe

rs t

o th

e ch

ange

in y

ield

.**

App

roxi

mat

e du

ratio

n is

cal

cula

ted

as t

he t

ime

pass

ed in

the

sha

ded

even

t w

indo

w.

Sou

rces

: Gol

dman

Sac

hs G

loba

l Inv

estm

ent R

esea

rch;

Blo

ombe

rg; R

eute

rs; “

Find

ings

Reg

ardi

ng th

e M

arke

t Eve

nts

of M

ay 6

, 201

0,”

US

CFT

C a

nd U

S SE

C, 3

0 Se

ptem

ber 2

010;

“Jo

int

Staf

f Rep

ort:

The

U.S

. Tre

asur

y M

arke

t on

Oct

ober

15,

201

4,”

US

Dep

t. o

f the

Tre

asur

y, U

S C

FTC

, US

SEC

, FR

BN

Y, F

eder

al R

eser

ve B

oard

, 13

July

201

5; “

IMF

Wor

king

Pap

er: F

ragi

litie

s in

the

U.S

. Tre

asur

y M

arke

t: Le

sson

s fr

om th

e “F

lash

Ral

ly”

of O

ctob

er 1

5, 2

014,

” O

ctob

er 2

015;

“Th

e st

erlin

g ‘fl

ash

even

t’ o

f 7 O

ctob

er 2

016,

” B

ank

for

Inte

rnat

iona

l Set

tlem

ents

, Ja

nuar

y 20

17; “

VIX

Spik

e: Im

plic

atio

ns F

or C

redi

t, Eq

uity

and

Vol

atili

ty,”

Gol

dman

Sac

hs G

loba

l Inv

estm

ent R

esea

rch,

6 F

ebru

ary

2018

.

https://www.sec.gov/news/studies/2010/marketevents-report.pdf

https://www.treasury.gov/press-center/press-releases/Documents/Joint_Staff_Report_Treasury_10-15-2015.pdf

https://www.treasury.gov/press-center/press-releases/Documents/Joint_Staff_Report_Treasury_10-15-2015.pdf

https://www.imf.org/external/pubs/ft/wp/2015/wp15222.pdf

https://www.imf.org/external/pubs/ft/wp/2015/wp15222.pdf

https://www.bis.org/publ/mktc09.pdf

https://research.gs.com/content/research/en/reports/2018/02/06/fc554cee-afc0-4d83-b98f-1d0ba4b1a4fc.html

El



We sat down with Brian Levine, Goldman Sachs Co-Head of Global Equities Trading and Execution Services, and Raj Mahajan and Konstantin Shakhnovich, GS Global Co-Heads of Securities Systematic Solutions (S^3), to discuss the evolution of US market structure. The interviewees are employees of the Goldman Sachs Securities Division, not Goldman Sachs Research. The views stated herein are their own and may not necessarily reflect those of Goldman Sachs.

Brian Levine Raj Mahajan Konstantin Shakhnovich Allison Nathan: When it comes to the evolution of market structure and the electronification of trading specifically, what developments have been the most striking to you?

Konstantin Shakhnovich: It's amazing how quickly a market can go from almost entirely voice to entirely electronic. It took just seven years for the FX market to flip from roughly 85% voice / 15% electronic to essentially the other way around. It is also remarkable how difficult regulating market structure has been. For example, the unintended consequences of well-intentioned rules like Regulation National Market System (Reg NMS)—passed by the SEC in 2005—have been more pronounced than I think anyone expected.

Raj Mahajan: I agree. Reg NMS, among other things, aimed to ensure that participants in the US equity markets received the best price by forcing brokers and exchanges to route orders to the venue with the best quote—the National Best Bid and Offer or NBBO. This accelerated the electronification of the US equity exchanges. But it also introduced substantial complexity into the system. Identifying the best price is not so straightforward when there are about 40 different venues—13 US equity exchanges plus 20+ alternative trading systems in several different states. That, in turn, increased the importance of factors like the speed of exchanges’ data feeds.

Brian Levine: Indeed, since Reg NMS made price the predominant variable for choosing an execution venue, it effectively made speed in offering that best price the most important variable. That, coupled with the exchanges’ immunity from private lawsuits in the event of operational incidents, incentivized the exchanges to invest in shaving microseconds off of latency rather than improving their capacity, reliability, and safety. A highly visible example of this was the Facebook IPO in 2012, when technical difficulties on the NASDAQ exchange led to close to $1 billion worth of errors, with the exchange only liable for a tiny fraction of those losses. If exchanges had stronger incentives to enhance their stability and reliability—rather than just their speed—that would be healthier for markets.

KS: I would add that the emphasis on price and speed has lowered the barriers to entry, since exchanges and market makers now compete on those two variables. New players have proliferated as a result, further fragmenting the US equity markets and reinforcing the complexity Raj mentioned.

AN: Haven’t electronification and newer liquidity providers like high-frequency traders (HFTs) improved efficiency?

RM: The markets are certainly more efficient than they were when people in vests shouted across the floor of exchanges. Having more market makers competing for quotes is generally a positive thing, and it has pushed spreads tighter.

BL: That said, I think we have passed the point of diminishing marginal returns on the speed component of electronic trading. Does an “efficient” market really require that trades be made in 20 microseconds, which is how fast they are today? To put that in perspective, 20 microseconds is 20 millionths of a second; a millisecond is a thousandth of a second; the blink of an eye is 400 milliseconds. I also think whether the markets are more efficient depends on whom you ask. For small, short-term trades—the HFTs’ specialty—this is the case; SEC rules have been optimized for small trades with thin spreads. But for a big institution executing a large trade, very little improvement has been made; the estimated costs to execute large-sized orders have hardly improved over the past 10-15 years despite broad advancements in computer processing power.

KS: That’s true, but I don’t think there should have been an expectation for much improvement in these types of trades. If you are trying to sell a large chunk of a company as opposed to 100 shares, there is only so much technology can do.

AN: Is the concern that HFTs have introduced risk into the system justified?

RM: Yes, this concern is justified given that the underlying capitalization of at least some of these HFTs is very low. This is critical because the equity exchanges still have a mutualized central counterparty. But the amount of capital that these HFTs post on deposit to the National Securities Clearing Corporation (NSCC) is generally a small fraction of the notional they trade.

KS: I agree this is a serious concern, but it is important to understand why. There is an uninformed view out there that low capitalization is problematic because it prevents market makers from standing in the way of markets when they come under stress. But frankly, no sane market maker is ever going to actually stand there and lose a bunch of money to stabilize the market. Rather, low capitalization is a concern because of the operational risk it creates. While HFTs generally accrue little market risk, meaning they typically don’t warehouse much risk—if any—overnight, their operational risk scales to a fairly significant degree with how much they trade. Normally, everything is very tightly matched and risk-controlled. But if something goes wrong, like they have a bug in their system or their algo malfunctions, they could accumulate risk very quickly.

A discussion on market structure

El



AN: Could this translate into systemic risk?

KS: Absolutely. If one of these HFTs has a serious problem, the costs would be socialized to all the clearing members. It is not difficult to envision such a scenario affecting the capacity of the equity market to clear. It’s worth mentioning that this type of operational risk doesn’t exist in some other markets. For example, FX is effectively a bilateral market in which the brokers underwriting the risk basically acknowledge sufficient capital availability trade-by-trade. While the clearing house in the equity market also underwrites the risk, it has no standard protocol for ensuring that capital is available trade-by-trade, and its ability to shut off a trader is fairly limited.

BL: Part of the concern is that, in US equity markets, firms that clear and firms that execute are usually different. And firms that clear don’t have the ability to certify exchanges’ execution controls. A key issue is that exchanges today play a hybrid role between for-profit entities and self-regulating organizations, blurring the lines between for-profit functions and control functions. Establishing standard protocols for exchanges’ controls would really help reinforce the health of the system.

AN: What about the concern that HFTs cause or exacerbate flash crashes by reducing liquidity?

BL: If by flash crash you mean the market drops sharply but then snaps back, I think that concern is often overstated. Scarce liquidity in volatile markets is nothing new. In the days of manual markets, market makers just didn’t answer their phones. For example, in the ’87 crash, people were really crowded on one side of the trade and all ran for the exit at the same time. This isn’t much different than a hypothetical scenario today in which systematic strategies all try to exit a position at once. It might happen within a few seconds instead of the 20 minutes it took for someone to answer the phone. But I think there’s a benefit to the fact that the market readjusts quickly today. Not everyone agrees. Many investors feel uncomfortable with a stock falling to 70 from 80 in a very short period of time. But I don’t think there’s anything inherently wrong with that. Bottom line, I don’t blame HFTs for flash crashes. In the grand scheme of things, these players do little to improve or worsen the situation. Remember, they generally end the day flat—they’re not taking much market risk.

KS: I agree that HFTs have gotten too much attention in this regard. In general, I think people spend too much energy trying to define market makers and their role or obligation. In a trading environment, every player will always decide whether or not to participate, how much liquidity to offer, and at what price.

RM: I agree that it is almost impossible to define a market maker these days. Some exchanges offer incentives to trade a certain amount of volume or to post for a certain amount of time, but you can’t compel a participant to trade. In fact, studies of the Treasury flash crash in 2014 showed that the banks widened their quotes and offered less volume to a greater extent than HFTs.

Taking a step back, I’d say that these flash crashes are not very significant in the context of this new trading paradigm we’ve been discussing. Given the tremendous complexities in US equity market structure, for example, it’s actually pretty impressive how well the market functions.

AN: What does concern you?

BL: What’s more worrisome to me is a real flash crash, which I define as a situation when the market “breaks”: The data is wrong, everything trades at dislocated prices relative to the NBBO, and everyone—justifiably—widens their spreads. That happens almost every time there’s volatility, largely because message traffic increases dramatically. This is due to the fact that the opportunity set is greater and there’s no economic disincentive for sending messages to the market, so more electronic orders come in. This slows the system, widening spreads and generating price dislocations, which triggers even more orders and compounds the delays—a predicament that is only further exacerbated by the fragmentation of the equity markets. As this happens, stocks may trade outside of the NBBO briefly in millisecond or microsecond increments, constituting what I consider a genuine flash crash. All of this becomes a negative feedback loop that causes more volatility.

Interestingly, if you define a flash crash by the percentage of executions that took place outside the NBBO, one of the largest ones occurred in 2008 after the first TARP bill failed, according to internal analysis we did a few years ago. And the market didn’t snap back, with the SPX closing down 10% on the day and on its lows. I think that may have been why there wasn’t talk of a “flash crash” afterward, but clearly the market structurally failed pretty badly that day, too. This suggests to me that, in a situation with actual bad news, the current US market structure may not be able to handle it, and there could be a downward spiral.

AN: What could address these fragilities?

KS: At a high level, the system can be slowed to provide more time to digest information and potentially decrease message traffic. Batch auctions—batching orders/price updates, say, ten times per second—would basically limit the amount that any participant trades and provide time to determine whether a participant should be able to trade the next batch. That said, if there are no buyers, no market structure could resolve that.

BL: I’m a huge fan of batch auctions, if regulators are willing to explore a pretty significant overhaul of market structure. Unfortunately, most proposed fixes over the years have been after-the-fact, like circuit breakers. Rather than address market fragility with more refined circuit breakers, I’d rather cure the disease: If there are no buyers because the market broke, why not prevent that break? In my view, we have a good chance of achieving that if we explore incentives that would result in less message traffic, or if we increase exchange capacity. Anything we can do to reduce fragmentation would also help by lessening the burden of having to check with numerous venues to hit the NBBO.

RM: Those are useful policy prescriptions. But to me, the key is addressing capitalization. Capitalization should be a function of open orders in the market throughout the day. If capital requirements for HFTs are increased in this manner, HFTs will be more selective about which venues they quote on, which will alleviate some of the fragmentation. Furthermore, exchanges would have to compete for non-HFT or bank flow by offering incentives that may be more appealing to banks, especially incentives tied to stability and reliability. In the end, by addressing this ratio of open orders to capital, the market will naturally correct to make the overall ecosystem safer.

El



Andrei Kirilenko is the Director of the Centre for Global Finance and Technology at the Imperial College Business School. Previously, he was Professor of the Practice of Finance at the MIT Sloan School of Management. He also served as Chief Economist of the US Commodity Futures Trading Commission (2010-2012). He argues that common concerns around high-frequency traders (HFTs) aren’t always warranted, but that not all market participants can access the liquidity HFTs provide; and that more information about latency would improve price discovery and market resiliency. The views stated herein are those of the interviewee and do not necessarily reflect those of Goldman Sachs.

Marina Grushin: You have looked extensively at the impact of technology and automation on financial markets. How have they affected market structure, and specifically market making?

Andrei Kirilenko: Cheaper and faster computers, along with the development of financial algorithms,

have driven enormous structural change in the markets. One of the most notable changes has been the erosion of market makers’ specialized, niche position in the system. Their function remains the same: They facilitate continuity by buying and selling on demand, so that natural buyers and sellers do not have to arrive in the markets at the same time. However, market makers no longer have the monopoly they once did as, say, specialists on the New York Stock Exchange. Technology has enabled just about any market participant to automate a trading platform to buy and sell on demand. It has also enabled some proprietary traders—in particular HFTs—to provide what appears to be a variant of market-making services without necessarily having any obligation to be market makers.

Marina Grushin: What has your research shown about the costs and benefits of HFTs’ participation in the markets?

Andrei Kirilenko: HFTs make markets more continuous and offer market participants greater immediacy. They have also enabled smaller trades to take place within microseconds, at or very close to the best market price. Those are significant benefits. On the other hand, HFTs have made trading large volumes much more challenging. My empirical work shows that HFTs effectively act as a private matching engine—a sophisticated piece of computing technology that matches buy and sell orders—while embedding their own trading strategies into the logic of that engine. As a result, traders attempting to take a large directional position can no longer assume that their orders are entering a passive system. Rather, the system actively changes in response to—or in anticipation of—what they’re doing. If they want to buy, prices will start to rise. If they want to sell, prices will start to fall. In other words, there has been a large cost to investors who want to trade in size.

Marina Grushin: Can you elaborate on how HFTs have impacted market liquidity?

Andrei Kirilenko: High-speed, automated trading has challenged our traditional notion of liquidity. Liquidity was previously thought of as something passive, like a resting limit order: Once it’s placed, it sits there, ready to be picked up by someone who wants to consume that liquidity. In the presence

of HFTs, liquidity now reflects orders being submitted and canceled very quickly in addition to the orders passively awaiting a match. This means the market will give the appearance of liquidity, but if you try to trade against it, that liquidity may disappear like a phantom. Moreover, passive liquidity has become less transparent either by hiding inside “iceberg” orders on lit venues or by outright migrating into so-called dark pools, where participants are assured that price will not respond to their order flow.

Marina Grushin: Is this notion of “phantom” liquidity still as relevant as it was, say, five years ago? Haven’t investments in technology by other market participants reduced the speed advantage of HFTs?

Andrei Kirilenko: My research on latency indicates that fleeting liquidity definitely still exists. It’s true that high-speed trading has become commoditized. A fast market feed and colocation—whereby a trader’s hardware is housed close to an exchange’s matching engine—are not as expensive as they used to be. However, HFTs still have a large technological advantage. Let’s say your spending on trading technology gets you to the one-millisecond horizon, while the time it takes for an HFT to display and cancel an order is 30 microseconds. Moving from one millisecond to 30 microseconds is actually quite costly. You would need specialized hardware designed to run a particular protocol for a particular venue. But if you choose to remain at one millisecond, you won’t be able to touch that 30-microsecond liquidity. Needless to say, less technologically sophisticated players are even more limited in what liquidity they can “consume” at any given time.

Marina Grushin: How have you observed HFTs’ behavior change amid market stress, when liquidity is often scarce?

Andrei Kirilenko: I looked into this with a few other researchers in a study of the May 2010 US equity flash crash. We identified HFTs based on their trading dynamics—inventory balances and trading volume—and then analyzed how they and other market participants behaved during that event. We found that throughout the flash crash, HFTs did not change their trading strategies or their overall behavior. In my view, they continued to trade because they don’t particularly care about the price level; they only see one tick up or one tick down, and they basically make money on every tick.

In contrast, some other market participants—mainly traditional market makers who were providing passive limit orders before and after the event—did withdraw temporarily from the markets. Some of them did not come back until after the five-second trading pause kicked in at the very bottom of the flash crash. So at a certain point, the majority of trading activity was

Interview with Andrei Kirilenko

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1569067



El



happening only among HFTs, basically passing around a “hot potato.”

That said, there is evidence that HFTs exacerbated the sell-off and accelerated the price swings. Had they not been in the market, prices may not have moved as quickly or as far.

Marina Grushin: Market participants and the media have sometimes argued that HFTs increase market risk. What’s your view on this?

Andrei Kirilenko: I don’t believe HFTs pose a major risk to the markets. They are incredibly skilled at managing their inventory. Despite constantly buying and selling, they do not accumulate significant inventory. So their net open position at any time of the day—and therefore their market risk—is very small.

Marina Grushin: Do you think they are equipped to manage their operational risk—for instance, a technology glitch? Could that pose a systemic risk?

Andrei Kirilenko: There do seem to be some concerns about this given HFTs’ dominance in certain markets. But I would point out that HFTs are typically trading with their own money. If something goes wrong, they’re the ones that will get run over. It is hard to envision taxpayers bailing them out, so moral hazard doesn’t appear to be an issue. Therefore, to me, HFTs are taking a calculated risk in deciding how and where to operate. The fact that they are not present in all markets suggests that they restrict themselves to areas where they have sufficient capitalization, a strong understanding of market dynamics, and a technological advantage. And if one of them goes down, it will simply be replaced.

Marina Grushin: What does worry you?

Andrei Kirilenko: I would be more concerned in a scenario where HFTs withdraw intermediation from the markets on a large scale. That would indicate that we have a very serious problem on our hands. HFTs’ technology enables them to operate in front of the rest of the market, or very close to the front. So if they stop trading, it would signal some very large source of uncertainty. For instance, one thing that would certainly rattle HFTs—along with other market participants—is any sort of issue with the data feed.

More broadly, I would note that the much greater complexity of the entire marketplace today creates new risks. Instead of isolated, passive platforms, we have an actively evolving ecosystem of trading venues that constantly respond to different signals. With this greater complexity, the potential for technological malfunctions or sudden events such as flash crashes has also increased. You can imagine an announcement or regulatory move of some kind leading to a large liquidity event, which then gets exacerbated by the presence of HFTs.

Marina Grushin: What do you make of the flash crashes in recent years? How worried should we be about them?

Andrei Kirilenko: Based on my own observations and regulators’ statements, flash crashes seem to be a fixture of the markets. They can affect everything from a single stock to a massive price discovery vehicle like the S&P 500. My takeaway from these episodes is that, while they see prices move quickly, they often cause information to become available

with a greater delay, both through the market feed and traders’ own communication platforms. In other words, prices speed up, but the time to react to them slows down. Along the way, it appears that algorithms can cause prices to move in cascades, contributing to a reallocation of liquidity.

While these events do happen periodically, I would argue that we have increasingly effective tools to deal with them. In the past, the most common safeguard was the circuit breaker, which would kick in—quite disruptively—once prices had already moved. Today, we have various pre-trade safeguards that can help keep prices from moving. So-called throttles, for example, can limit volumes if an algorithm appears likely to result in extreme price moves. And market participants cannot connect to trading platforms unless their algorithms are tested under adverse market conditions. In short, while there are risks in automated trading systems that we don’t fully understand, there are also continuous efforts to address them.

Marina Grushin: Overall, what do you see as the biggest shortcoming in market structure today?

Andrei Kirilenko: To me, the most important problem is technological in nature. Automated trading relies on messages passing through complex pieces of technology that can influence how quickly, and in what sequence, orders end up in the matching engine. Never mind that switches can burn down, fiber optic cables need to be replaced, etc.; two orders that arrive at an exchange at the same time can actually end up in the matching engine at different times simply due to a router configuration. This creates significant technological uncertainty for market participants, who have little to no transparency on the latency of their trades. And, as the press has documented, some traders have also exploited this for profit.

Marina Grushin: As a former regulator, what policy prescriptions would you recommend to address this?

Andrei Kirilenko: I would advocate for greater transparency on latency information available to the markets. Market participants know and calculate their own latencies, but they don’t know them for the market as a whole. Their algorithms react to time, yet they don’t get to see it directly in their market data feed. My recommendation would be to add some notion of time to the prices and quantities that are already submitted to the feed. The exchanges collect this information already, and I think regulators should engage in more active discussions with them about releasing it.

If we took this step, market participants would be able to optimize their algorithms for latency across platforms. They could learn to look through the transactions that take place beyond the speed at which they can actually trade—essentially tuning out the noise and focusing on the liquidity they can really touch. Think of it as a central limit order book for just the one-millisecond latency horizon. That book might look much sparser than the overall market, and it might show a bid-ask spread of three ticks as opposed to one. But for market participants who can trade in one millisecond, that book would be a more accurate reflection of the market they are facing. This could potentially help curb price reactions and thereby make markets more resilient. To me, that’s long overdue.

El



Christian Mueller-Glissmann and Alessio Rizzi see shifts in market structure muddying the links between fundamentals, volatility, and risk

Volatility is often used as a measure of risk. When volatility is driven by fundamentals, this makes sense: As the fundamental backdrop worsens, volatility, and in turn, risk, increases, and vice versa. However, low volatility does not necessarily mean low risk, as periods of low vol can actually mask latent risks that owe to excessive risk-taking, especially if this search for yield occurs in illiquid assets. Low volatility can result in investors underestimating liquidity risk. On the other hand, episodes of high volatility may not always reflect the fundamental backdrop, as the S&P 500 vol spike earlier this year demonstrated. Changes in systematic investment strategies and market microstructure since the Global Financial Crisis (GFC) have the potential to exacerbate these disconnects, muddying the relationship between fundamentals, volatility, and risk.

Fundamentals disconnect

Equity volatility is normally driven by fundamentals. It tends to be countercyclical: As the business cycle matures, inflation tends to pick up and recession risk increases, both of which can drive higher volatility. Conversely, low-vol periods typically coincide with a generally supportive macro backdrop, which tends to support equity returns. However, in recent years, equity market volatility has disconnected more frequently from fundamentals. While macro volatility (of GDP and inflation) has generally declined since the mid-1980s when central banks started aggressively targeting inflation (ushering in the “Great Moderation”), realized S&P 500 volatility has actually trended upward since the 1990s. In addition, S&P 500 vol spikes have become both larger and faster in recent years—occasionally without a clear and material fundamental catalyst. As a result, vol of vol is approaching levels similar to 1987, even as the global economy has continued to expand since the GFC and the equity bull market is closing in on the longest on record (which was in the 1990s).

Low vol does not always mean low risk

This disconnect implies that volatility can send the wrong signal to investors. Indeed, low volatility does not always mean low risk. In recent decades, a lack of inflation has allowed central banks to err on the side of caution and buffer the business cycle, enabling longer expansions and longer periods of low volatility. But longer low-vol periods can drive excessive risk-taking and releveraging (a phenomenon dubbed the “volatility paradox”), which can eventually lead to a volatile unwind of imbalances, as happened during the GFC. In fact, the external shocks that interrupted the two low-vol regimes since the GFC—namely, the Euro area crisis and the EM/oil crisis—may have helped prevent a build-up of larger imbalances during the current cycle.

Today, the potential for excessive risk-taking by investors may be especially pronounced given the rise of systematic, pro-

cyclical investing strategies tied to volatility. These include short volatility and option selling strategies; Commodity Trading Advisors (CTAs, $350bn in AUM according to BarclayHedge), which effectively buy assets with positive momentum and therefore tend to invest in equities as they trend higher during low-vol regimes; and risk parity and volatility targeting funds (combined AUM close to $1tn1), which might also increase exposure to equities amid low vol. Such rule-based strategies may contribute to low-vol regimes, but they can also have quick, simultaneous unwinds during periods of high volatility (although of course risk allocation rules differ somewhat across strategies). And in cases of extreme positioning, reversals may not require a large fundamental shock. As a result, they have the potential to generate higher volatility outside recessions compared to the past.

Technical difficulties

A related concern is that shifts in market microstructure have reduced market liquidity, and could thus exacerbate volatility in periods of market stress; put differently, low volatility can mask illiquidity risks. Rising volatility can then sharply reduce liquidity; for example in S&P 500 futures, bid-ask spreads and order book sizes have become very sensitive to the level of volatility. Relevant shifts in market structure and the overall market landscape include: (1) changes in bank regulation since the GFC; (2) the growth of high-frequency traders, which some argue could exacerbate liquidity scarcity during “flash crash” episodes (see pg. 8); (3) the rise of passive/smart beta strategies, which can drive crowding; (4) growth in ETPs that are levered, inverse (e.g., the short VIX ETPs involved in the vol spike earlier this year—see pg. 16), and/or on illiquid underliers.

Overall, we see increased risk that volatility disconnects again from fundamentals, and that markets experience more sharp and unpredictable “risk-off” periods. This is especially true as the macro backdrop has become less strong, with the growth/inflation mix worsening year-to-date. Therefore, while we continue to recommend a pro-risk asset allocation (overweight equities and commodities, underweight bonds), we are also overweight cash near term. This should help investors weather more fundamentally-driven volatility (i.e., due to further worsening of the growth/inflation mix and elevated policy uncertainty) and more technically-driven episodes of market turbulence. To hedge against smaller, technical corrections, shorter-dated put spreads, e.g., S&P 500 1-month 97-93% put spreads, appear attractive—put/call skew iselevated and has trended up since the GFC due to changes inUS bank regulation.

Christian Mueller-Glissmann, Sr. Multi-Asset Strategist Email: [email protected] Goldman Sachs International Tel: +44-20-7774-1714

Alessio Rizzi, Multi-Asset Strategist Email: [email protected] Goldman Sachs International Tel: +44-20-7552-3976

For details see: Global Strategy Paper: Taste of the High (Vol) Life, 16 April 2018.

1 Estimating the total size of risk parity, vol target funds, and similar funds that are implicitly or explicitly “short volatility” is difficult due to lack of public data. Recent academic research has come to similar numbers: see Bhansali & Harris (2018).

Low vol does not always mean low risk

https://research.gs.com/content/research/en/reports/2017/06/20/811348e7-1049-402f-b43b-9b64a28dc7e5.sitePilot.html

El



Lessons from volatility

Market vol has disconnected from macro vol GDP and inflation volatility (5-year, lhs); S&P 500 1-month realized vol (5-year average, rhs)

Vol of vol has trended up since the GFC Daily volatility of percent changes in S&P 500 1-month realized volatility (%)

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

Liquidity is procyclical... S&P 500 realized volatility (lhs); daily bid-ask spread for S&P 500 futures, 1st/2nd contract (%, rhs)

...and can be linked to volatility Avg. daily bid-ask sizes on top of S&P 500 e-mini futures book, 1st/ 2nd contract ($mm, lhs); S&P 500 realized vol (1m avg, rhs, inverted)

Source: Bloomberg, Goldman Sachs Global Investment Research. Source: Bloomberg, Goldman Sachs Global Investment Research.

S&P 500 volatility spikes are not limited to recessions and bear markets S&P 500 1-month realized volatility (grey shading = bear market, dark blue shading = US recession, light blue shading = both)

Source: NBER, Bloomberg, Goldman Sachs Global Investment Research.

5

10

15

20

25

30

35

40

0

2

4

6

8

10

12

14

28 38 48 58 68 78 88 98 08 18

GDP volatility (5-year)

Inflation volatility (5-year)

S&P 500 1-month volatility (5-year average, RHS)

While macro vol has been low, market vol

has been on an upward trend

70%

90%

110%

130%

150%

170%

38 48 58 68 78 88 98 08 18

1-year5-year10-year Vol of vol is

approaching 1987-like levels

0.0%

0.3%

0.6%

0.9%

1.2%

1.5%

0

5

10

15

20

25

30

35

40

45

50

07 08 09 10 11 12 13 14 15 16 17 18

S&P 500 1-month volatilityBid-Ask (avg 1m, rhs)Bid-Ask (avg 1m, rhs)

0

5

10

15

20

25

30

35

40

45

500

10

20

30

40

50

60

70

80

90

05 07 09 11 13 15 17

Avg Top of Book Size - $mm daily (rolling 1m avg) Vol 1m (RHS, inverted)

0

10

20

30

40

50

60

70

80

90

28 32 36 40 44 48 52 56 60 64 68 72 76 80 84 88 92 96 00 04 08 12 16

Industrial production

plungeWW IIbegins

Post war inventory

crisis

Start of ColdWar

Suez Crisis

Eisenhower recession

Cuban MissileCrisis

John F. Kennedy

assassinated

Escalation Cold War

Great Depression

Vietnam war &

monetary tightening

First Oil

price shock

Iran oil production

cut

Volcker monetary tightening

Black Monday

Invasion of Kuwait

Global Financial

Crisis

9/11Asian

financial crisis

LTCM/Russian Crisis

US debt downgrade &

Euro area crisis

EM/oilcrisis

Tech bubble burst

May flash crash

Special thanks to Christian Mueller-Glissmann, Alessio Rizzi, and the multi-asset strategy research team. For more, see accompanying article, “Low vol does not always mean low risk” on pg. 14 as well as Global Strategy Paper: Taste of the High (Vol) Life, 16 April 2018.

https://research.gs.com/content/research/en/reports/2017/06/20/811348e7-1049-402f-b43b-9b64a28dc7e5.sitePilot.html

El



Rocky Fishman argues that the precise drivers of the February VIX spike are unlikely to repeat themselves, but that other aspects of market structure raise the risk of volatile price action

Q: The VIX jumped 20 points in a few hours on February 5th. What happened?

A: The S&P 500 fell sharply in the afternoon of February 5th, with SPX futures finishing the day down over 5%—the largest one-day decline since 2011. The VIX’s move was even more extreme: the 20-point spike toward its closing level represented the largest one-day change in the VIX ever. Known rebalancing procedures of short-term VIX futures-based products contributed materially to the equity market selloff, and particularly to the outsized move in volatility metrics, in our view. The lack of a clear fundamental or economic catalyst for a selloff of this magnitude also suggests that this event was primarily technical in nature.

The (VIX) spike of all spikes One-month weighted average of 1st and 2nd VIX futures, intraday on 5-Feb-18

Source: Bloomberg, Goldman Sachs Global Investment Research.

Q: What are VIX and volatility derivatives, and why do investors trade them in the first place?

A: Investors trade volatility derivatives, including VIX futures and options, for three primary reasons: (1) to hedge portfolios, because volatility typically rises when risky assets across asset classes are falling; (2) to collect yield by selling typically-overpriced options, because implied volatility often exceeds subsequent realized volatility; and (3) to manage option positions’ risk, because options’ value is in part a function of volatility levels. The VIX itself cannot be traded, but VIX futures and options are linked with the level of the VIX at a single future moment in time and provide a liquid, listed marketplace for investors to trade volatility risk.

Q: What are VIX exchange-traded products, and why were they so impactful on February 5th?

A: Exchange-traded funds and exchange-traded notes (which we collectively refer to as exchange-traded products, or ETPs) were first launched on VIX futures strategies in 2009. The first

ETP provided long exposure to VIX futures, and subsequent ETPs also included short exposure to VIX futures and levered long exposure to VIX futures. In the early years of the VIX ETP market, the products were predominantly used to establish long volatility exposure, but by early this year strong performance of short ETPs and some early 2018 inflows had led the short volatility ETPs to collectively outsize the long ETPs.

Like levered equity ETFs, the inverse and levered long VIX ETPs are designed to return a multiple (-1x or 2x respectively) of a VIX futures strategy’s return on a one-day basis only. Every day, the issuers of these ETPs re-size the derivatives portfolio used to back each product to maintain a defined multiple of the product’s closing value. When VIX futures rise, both inverse product issuers and levered long product issuers are economically driven to buy VIX futures: the inverse product issuers to reduce a short position that has become too large relative to the product’s AUM, and the levered issuers to supplement a long position that has not grown as quickly as the product itself. We estimate that February 5th’s spike in volatility left issuers with over 200,000 VIX futures to buy toward the end of the trading session—an outsized amount (around 50% of the open interest of the first two VIX futures) that led to sharp moves in all measures of implied volatility.

Q: Was the short volatility position of the VIX ETPs the key reason they were impactful on February 5th?

A: No. The fundamental characteristic of the VIX ETPs that contributed to February 5th’s volatility spike was their daily leverage, not their net short volatility positioning. Both inverse and levered long ETPs contributed to the late afternoon volatility because both buy VIX futures near the close when volatility is rising. These products’ ability to move markets to the extent they did was driven by two characteristics: (1) they have a real-time, end-of-day, feedback loop, in which the final price of VIX futures is an input to the number of VIX futures they need to be long/short; and (2) they were too large a percentage of the VIX futures market they participated in (AUM growth resulted in part because the short ETPs do not distribute profits).

Both long and short products were significant contributors VIX ETPs: VIX futures to buy on a 5-point VIX futures spike, thousands of contracts

Source: Reuters, Goldman Sachs Global Investment Research.

10

15

20

25

30

11:30 12:00 12:30 13:00 13:30 14:00 14:30 15:00 15:30 16:00

2013-17 Closing Peak (27.7 on 01-Sep-15)

2-Feb Close (15.2)

2017 Median (12.6)

Spike intensified toward the close

0

50

100

150

200

2016Median

2017Median

Entering5-Feb Spike

Now(29-May)

Short ETPs

Long ETPs

ETPs’ current impact is small

Entering5-Feb Spike

The VIX spike: causes and concerns

https://research.gs.com/content/research/en/reports/2018/02/07/fa6f6c94-d012-4383-8585-338a81342060.html


El



Q: Could a similar VIX ETP-driven event happen again?

A: Not for now, and not likely anytime soon. Even if a similar equity selloff were to occur today, the combination of the much-diminished size of short VIX ETPs ($600mn AUM now vs. $4bn at their peak) and changes to product structure (reducing the leverage of two key products) would leave VIX ETP issuers with a small fraction of February 5th’s futures-to-buy.

More broadly, though, the large amount of assets invested in products and strategies that demand near-instantaneous liquidity could continue to create occasional mismatches with markets that can see diminished liquidity in moments of stress. Liquidity has its own supply-demand dynamics, and when demand (systematic strategies’ need to trade) exceeds supply (market makers’ willingness to facilitate trades), the cost of liquidity (market impact of a given trade) will rise. Individual trades moving markets more than they once would have, or more than warranted by fundamentals, means higher volatility.

Q: How has the role of volatility itself evolved in portfolio management, and how might volatility be connected with market fragility?

A: Implied volatility was originally a metric used to observe and contextualize option prices when the Black-Scholes formula was published (connecting the price of options with implied volatility) and the VIX index was launched (linking expected volatility to the price of a portfolio of options). The role of volatility has widened since then. The growth in variance swap trading, and later VIX derivative trading, turned volatility into an active, traded asset, with more short-dated volatility risk traded via VIX derivatives than via the SPX options they were designed to track. Short-dated implied volatility is now primarily driven by investors’ views about volatility, in our view. In the post-GFC period, volatility has gained traction as an input to portfolio construction processes of systematic investment strategies. Managed volatility funds, for example, do not trade derivatives at all, but rather use volatility to determine how large an allocation to equities they hold at a given moment. As a result, volatility can drive semi-automatic equity selling, which in turn can boost expected volatility in times of market stress.

Fast growth in volatility trading Open interest of first two monthly VIX futures, thousands of contracts, two-month rolling average

Source: Reuters, Goldman Sachs Global Investment Research.

Q: Did February’s spike in volatility influence underlying equity market prices?

A: In our view, yes. At a very direct, in-the-moment, level, it is likely that some of those market participants who had sold VIX futures to ETP issuers sought to crudely neutralize the market risk in these new short volatility positions by selling S&P 500 futures (on the assumption that VIX futures are negatively correlated with the SPX). Less directly, long-only asset allocation strategies that use volatility as an input may have been driven to reduce equity positioning, as described above. Even if those strategies would reduce equity positions slowly, investors worried about upcoming systematic strategy selling may have chosen to reduce their own equity positions. At the most indirect level, the rise in volatility damaged investor confidence and also influenced various funds’ risk metrics. In any case, the equity market selloff reverberated broadly: even Asian equity markets fell sharply when they opened on February 6th.

Vol still elevated SPX quarterly realized volatility, %; dotted lines indicate 5/50/85th percentiles, last 50Y

Source: Bloomberg, Goldman Sachs Global Investment Research.

Q: Volatility is still elevated, four months after February’s selloff. Does that mean we have entered a high-volatility environment?

A: In our view, no. While S&P 500 volatility has been above average for most of 2018, it has not been extreme. Moreover, it has recently drifted back down below historical averages, moving closer in line with the low levels of volatility seen in other markets, including equity index volatility outside of the US and interest rate volatility. With near-term recession risk low and upside remaining for equities, we expect the VIX to remain below average. That said, weakening economic conditions and consistently elevated realized volatility both justify a VIX above its 2017 range. The VIX is therefore unlikely to re-test last year’s historically unusual lows in the near future.

Rocky Fishman, Equity Derivatives Strategist Email: [email protected] Goldman Sachs and Co. LLC Tel: 212-902-3396

For more, see VIX: Q&A on the Trading Dynamics of ETPs, 7 February 2018.

0

100

200

300

400

500

2010 2011 2012 2013 2014 2015 2016 2017 2018

Open interest in VIX futures grew quickly prior to 2018

8.5%6.8% 7.3% 7.1%

5.7%

19.8%

13.8%

0

5

10

15

20

2016Q4

2017Q1

2017Q2

2017Q3

2017Q4

2018Q1

2018Q2

50th percentile

85th percentile

5th percentile

Months after February’s VIX spike, volatility has been above average in Q2

https://research.gs.com/content/research/en/reports/2018/02/23/df976f4f-5d38-4380-b9ef-5f2011af138a.html

https://research.gs.com/content/research/en/reports/2018/05/09/f20d5ce8-0ccd-45be-8496-f041745e9449.html

https://research.gs.com/content/research/en/reports/2018/05/09/f20d5ce8-0ccd-45be-8496-f041745e9449.html

https://research.gs.com/content/research/en/reports/2018/05/11/4d3e0930-ed85-45f2-899c-31898fe3f616.html


El



Algorithmic trading (sometimes described as systematic or “quant” trading): The use of computers and/or mathematical models to perform various types of trading activity—e.g., make decisions, submit trades, manage orders—according to specific criteria. Algorithmic trading is employed for everything from index fund rebalancing and factor investing to breaking up large orders according to volume or time.

Alternative Trading System (ATS): A non-exchange trading venue. Examples include some electronic communication networks and dark pools.

Bid-ask spread: The difference between the price at which someone is willing to buy an asset (the bid) and the price at which someone is willing to sell the same asset (the ask or the offer). “Being” or “posting” the bid (offer) refers to being a willing buyer (seller) of an asset.

Broker-dealer: A trading intermediary that acts both as an agent, or broker (matching buyers and sellers on commission), and as a principal, or dealer (buying/selling for its own account).

Central limit order book (CLOB): A platform that automatically, and usually anonymously, matches buyers and sellers. The CLOB also lists the quantity of securities being bid/ offered at a given price, providing a display of market depth.

Centralized clearing: A system in which transactions clear through a single counterparty (rather than through various bilateral counterparties).

Commodity trading advisors (CTAs): Funds that use futures and other derivatives to achieve their investment objectives. CTAs initially operated predominantly in commodities markets but currently invest across multiple asset classes. While CTAs differ widely, the majority engage in systematic, trend-following strategies with the aim of capturing directional market moves.

Dark pools: Alternative trading systems in which bid and offer prices are not made public until after a trade is executed, with the intention of anonymously facilitating large orders and limiting their impact on market price.

Exchange-traded products (ETPs): Investment vehicles that track an index or basket of assets and trade on an exchange at market-determined prices. There are two primary types of ETPs: exchange-traded funds (ETFs) and exchange-traded notes (ETNs). ETFs share some investment features of mutual funds, with investors buying ETF shares to own a proportional interest in the pooled assets. ETNs are unsecured unsubordinated debt obligations, structured as a promise to pay a pattern of returns based on the return of the stated index.

High-frequency trading (HFT): Computer-driven trading that takes place with extremely low latency (i.e., in microseconds or milliseconds). HFT firms employ advanced technology to enhance speed, including computer systems located on or near the premises of exchange servers (“colocation”). HFT involves generating and rapidly canceling numerous orders as algorithms react to market conditions.

Liquidity risk premium: The additional return that compensates investors for the risk of holding a less liquid security (i.e., one that cannot be as readily converted into cash).

Managed volatility funds: Funds that explicitly target a level (or range) of realized portfolio volatility, and thus dynamically change their asset allocation—usually within days—in response to changing market conditions. A typical fund contains a portfolio of equities and bonds, and will at times add a short equity futures position to reduce volatility. The largest managed volatility products do not employ material amounts of leverage, and are most commonly found in products that aim to give investors steady cash flows and upside exposure to equities.

Market liquidity: The ability to buy or sell an asset without significantly impacting the overall market price.

Market maker: A market participant prepared to buy or sell securities on its own account at bid or offer prices quoted for a guaranteed number of shares. Examples of “traditional” market makers include specialized traders at exchanges and broker-dealers. Recent years have seen the growth of newer, “non-traditional” market makers (e.g., high-frequency traders) that apply technology to profit from bid-ask spreads while maintaining minimal intraday and inter-day risk.

Over-the-counter (OTC): Transactions that take place bilaterally rather than on an exchange. Post-crisis financial reforms have included efforts to move OTC derivatives toward centralized clearing models. OTC markets are sometimes referred to as dealer markets.

Passive investing: Investing in vehicles that track the returns of a market index or basket of assets, such as index mutual funds and ETFs.

Price dislocation: A large mispricing of assets relative to their fundamental value.

Regulation National Market System (Reg NMS): Rules passed by the US Securities and Exchange Commission (SEC) in 2005 intended to “modernize and strengthen” the regulation of US equity markets. Reg NMS established the Order Protection Rule, which aims to ensure that every market participant receives the best price by requiring brokers and exchanges to route orders to the venue with the lowest quote (i.e., the National Best Bid and Offer or NBBO). Other issues covered by Reg NMS include market access and the dissemination of market data.

Risk parity funds: Funds that aim to dynamically balance the risk of different asset classes in a portfolio. This differs from a 60/40 stock-bond allocation, in which risks are skewed towards equities. Risk parity funds may lever up bond holdings to achieve a higher level of risk that is more comparable to equities. Many risk parity funds also have a volatility target. These funds typically use long-term targets of asset class volatility and correlation, and adjust their allocations more slowly in response to volatility than managed volatility funds.

Short volatility strategies: Investments designed to profit from options’ being overpriced relative to their expected value. Short vol strategies generate income either by directly selling options or by selling volatility derivatives that are economically linked to portfolios of options. Inverse VIX ETPs, which track the returns from selling 1-2m VIX futures, are one example.

Market structure and liquidity lingo

Source: Congressional Research Service, US SEC, Financial Times Lexicon, Investment Company Institute, Goldman Sachs Global Investment Research.

El



Rev

isio

n N

otes

GS

Con

sG

SC

ons

GS

Con

sG

SC

ons

GS

Con

sG

SC

ons

2018

2019

2018

2019

GLO

BAL

4.1

3.8

4.0

3.7

--

--

--

--

--

--

US

2.9

2.8

2.2

2.4

1.15

1.18

1.25

1.24

2725

2830

2875

2870

2.25 to 2.50

3.25 to 3.50

3.25

3.60

On

May

30,

we

low

ered

our

3m

/6m

/12m

EU

R/U

SD

fore

cast

s to

1.1

5/1.

20/1

.25

from

1.2

0/1.

27/1

.30

prev

ious

ly g

iven

pol

itica

l dev

elop

men

ts in

Ital

y.

EUR

O A

REA

2.2

2.3

1.9

1.9

1.15

1.18

1.25

1.24

3550

3620

3700

3685

0.00

0.00

--

On

April

27,

we

low

ered

our

3m

/6m

/12m

Eur

o S

toxx

50

targ

ets

to 3

550/

3600

/370

0 fro

m 3

650/

3700

/395

0 pr

evio

usly

, ref

lect

ing

a sm

all s

low

dow

n in

Eur

o ar

ea g

row

th. O

n M

ay 3

0, w

e lo

wer

ed o

ur 3

m/6

m/1

2m

EU

R/U

SD

fore

cast

s to

1.1

5/1.

20/1

.25

from

1.2

0/1.

27/1

.30

prev

ious

ly g

iven

pol

itica

l dev

elop

men

ts in

Ital

y. O

n Ju

ne 8

, we

low

ered

our

201

8 an

d 20

19 E

uro

area

GD

P g

row

th fo

reca

sts

to 2

.2%

and

1.9

%, r

espe

ctiv

ely,

on

polit

ical

unc

erta

inty

in It

aly,

a w

iden

ing

of s

over

eign

spr

eads

, and

hig

her o

il pr

ices

.

GER

MAN

Y2.

12.

31.

92.

01.

151.

181.

251.

24-

1335

0-

1360

0-

-1.

001.

40

On

May

30,

we

low

ered

our

3m

/6m

/12m

EU

R/U

SD

fore

cast

s to

1.1

5/1.

20/1

.25

from

1.2

0/1.

27/1

.30

prev

ious

ly g

iven

pol

itica

l dev

elop

men

ts in

Ital

y. O

n Ju

ne 8

, we

low

ered

our

201

8 an

d 20

19 G

erm

an G

DP

gr

owth

fore

cast

s to

2.1

% a

nd 1

.9%

, res

pect

ivel

y, fr

om 2

.3%

and

2.0

% p

revi

ousl

y on

wea

ker-

than

-exp

ecte

d ac

tivity

and

hig

her o

il pr

ices

.

JAPA

N1.

11.

21.

31.

011

010

911

510

717

00-

1950

--0

.10

-0.1

00.

100.

40

CH

INA

6.6

6.5

6.1

6.3

6.30

6.39

6.20

6.36

--

107

-3.

002.

75-

-

BR

AZIL

2.0

2.5

3.1

2.9

3.50

3.65

3.10

3.50

--

--

6.25

8.00

--

On

May

25,

we

rais

ed o

ur 3

m U

SD

/BR

L fo

reca

sts

to 3

.50

from

3.3

0 pr

evio

usly

, ref

lect

ing

grea

ter U

SD

st

reng

th a

nd E

uro

area

wea

knes

s.

Com

mod

ities

GS

Con

sG

SC

ons

GS

Con

sG

SC

ons

GS

Con

sG

SC

ons

GS

Con

sG

SC

ons

82.5

7475

7273

00-

8000

-13

50-

1450

-40

0-

375

-O

n M

ay 1

, we

rais

ed o

ur 3

m/6

m/1

2m C

BO

T C

orn

fore

cast

s to

400

/375

/375

cen

ts/b

u fro

m 3

50/3

50/3

50

prev

ious

ly a

s hi

gher

spo

t ene

rgy

pric

es, a

sof

ter U

SD

, and

low

er U

S a

crea

ge w

ill lik

ely

put g

radu

al u

pwar

d pr

essu

re o

n pr

ices

.

Not

e: R

ecen

t rev

isio

ns m

arke

d in

red;

GD

P c

onse

nsus

is B

loom

berg

; all

othe

r con

sens

us is

Reu

ters

; com

mod

ity 1

2-m

o co

nsen

sus

is R

eute

rs fo

r 201

8 av

erag

e.+ C

NY

dai

ly fi

xS

ourc

e: B

loom

berg

, Tho

mso

n R

eute

rs, G

oldm

an S

achs

Glo

bal I

nves

tmen

t Res

earc

h.

Fed

Fund

s

Bre

nt c

rude

oil

($/b

bl)

Cop

per (

$/m

t)G

old

($/to

z)C

orn

(cen

t/bu)

EUR

/$EU

R/$

DAX

30

DAX

30

$/C

NY+

$/C

NY+

MXC

NM

XCN

$/JP

Y$/

JPY

TOPI

XTO

PIX

3-m

th12

-mth

3-m

th12

-mth

3-m

th3-

mth

12-m

th12

-mth

$/B

RL

$/B

RL

BO

VESP

AB

OVE

SPA

SP50

0SP

500

EUR

/$EU

R/$

Euro

Sto

xx 5

0Eu

ro S

toxx

50

EUR

/$EU

R/$

MR

O R

ate

7-D

ay R

epo

Polic

y D

epos

it

GD

P G

row

th (%

yoy

)FX

Equi

tyR

ates

(% e

op)

2018

2019

3-m

th12

-mth

3-m

th12

-mth

Polic

y*10

-yr

Summary of our key forecasts

El



Current Activity Indicator (CAI)

Our CAI measures the growth signal in a broad range of weekly and monthly indicators, offering an alternative to Gross Domestic Product (GDP). GDP is a useful but imperfect guide to current activity. In most countries, it is only available quarterly and is released with a substantial delay, and its initial estimates are often heavily revised. GDP also ignores important measures of real activity, such as employment and the purchasing managers’ indexes (PMIs). All of these problems reduce the effectiveness of GDP for investment and policy decisions. Our CAIs aim to address GDP’s shortcomings and provide a more timely read on the pace of growth. We currently calculate CAIs for the US, Euro area, Japan, UK, and 29 other countries. For more, see Global Economics Analyst: Trackin’ All Over the World – Our New Global CAI, 25 February 2017.

Financial Conditions Index (FCI)

Financial conditions are important because shifts in monetary policy do not tell the whole story. Our FCIs attempt to measure the direct and indirect effects of monetary policy on economic activity. We feel they provide a better gauge of the overall financial climate because they include variables that directly affect spending on domestically produced goods and services. Each FCI is calculated as a weighted average of a policy rate, a long-term riskless bond yield, a corporate credit spread, an equity price variable, and a trade-weighted exchange rate; in the Euro area we also include a sovereign credit spread. The weights mirror the effects of the financial variables on real GDP growth in our models over a one-year horizon.

Global Leading Indicator (GLI)

Our GLIs provide a more timely reading on the state of the global industrial cycle than the existing alternatives, and in a way that is largely independent of market variables. Global cyclical swings are important to a huge range of asset classes; as a result, we have come to rely on this consistent leading measure of the global cycle. Over the past few years, our GLI has provided early signals on turning points in the global cycle on a number of occasions and has helped confirm or deny the direction in which markets were heading. Our GLI currently includes the following components: Consumer Confidence aggregate, Japan IP inventory/sales ratio, Korea exports, S&P GS Industrial Metals Index, US Initial jobless claims, Belgian and Netherlands manufacturing surveys, Global PMI, GS Australian and Canadian dollar trade weighted index aggregate, Global new orders less inventories, Baltic Dry Index.

Goldman Sachs Analyst Index (GSAI)

Our US GSAI is based on a monthly survey of Goldman Sachs equity analysts to obtain their assessments of business conditions in the industries they follow. The results provide timely “bottom-up” information about US economic activity to supplement and cross-check our analysis of “top-down” data. Based on their responses, we create a diffusion index for economic activity comparable to the ISM’s indexes for activity in the manufacturing and nonmanufacturing sectors.

Macro-Data Assessment Platform (MAP)

Our MAP scores facilitate rapid interpretation of new data releases. In essence, MAP combines into one simple measure the importance of a specific data release (i.e., its historical correlation with GDP) and the degree of surprise relative to the consensus forecast. We put a sign on the degree of surprise, so that an underperformance will be characterized with a negative number and an outperformance with a positive number. We rank each of these two components on a scale from 0 to 5, and the MAP score will be the product of the two, i.e., from -25 to +25. The idea is that when data are released, the assessment we make will include a MAP score of, for example, +20 (5;+4)—which would indicate that the data has a very high correlation to GDP (the ‘5’) and that it came out well above consensus expectations (the ‘+4’)—for a total MAP value of ‘+20.’ We currently employ MAP for US, EMEA and Asia data releases.

Real-Time Inflation and Activity Framework (RETINA)

RETINA provides a comprehensive econometric methodology able to filter incoming information from the most up-to-date high frequency variables in order to track real GDP growth in the Euro area. Along with a GDP tracker, RETINA also captures the interrelated mechanisms of the area-wide pricing chain, providing a short-term view on inflation dynamics.

Glossary of GS proprietary indices

El



Issue 67 Regulating Big Tech April 26, 2018

Issue 52 OPEC and Oil Opportunities November 22, 2016

Issue 66 Trade Wars 2.0 March 28, 2018

Issue 51 US Presidential Prospects October 18, 2016

Issue 65 Has a Bond Bear Market Begun? February 28, 2018

Issue 50 Central Bank Choices and Challenges October 6, 2016

Issue 64 Is Bitcoin a (Bursting) Bubble? February 5, 2018

Issue 49 Trade Trends September 29, 2016

Special Issue 2017 Update, and a Peek at 2018 December 14, 2017

Issue 48 Breaking Down “Brexit Means Brexit” June 23, 2016

Issue 63 Late-Cycle for Longer? November 9, 2017

Issue 47 Political Uncertainty June 23, 2016

Issue 62 China’s Big Reshuffle October 12, 2017

Issue 46 Factoring in Financial Conditions May 26, 2016

Issue 61 Fiscal Agenda in Focus October 5, 2017

Issue 45 Brazil at a Crossroads April 27, 2016

Issue 60 The Rundown on Runoff September 11, 2017

Issue 44 Policy Stimulus: Running on Empty? March 23, 2016

Issue 59 Regulatory Rollback July 26, 2017

Issue 43 Credit Tremors March 2, 2016

Issue 58 The Fed’s Dual Dilemma June 21, 2017

Issue 42 China Ripple Effects February 9, 2016

Issue 57 Geopolitical Risks May 16, 2017


Issue 56 Animal Spirits, Growth, and Markets April 17, 2017

Issue 41 Liftoff December 10, 2015

Issue 55 European Elections: More Surprises Ahead? March 14, 2017

Issue 40 Reserve Reversal Rundown November 3, 2015

Issue 54 Trade Wars February 6, 2017

Issue 39 Picking Apart the Productivity Paradox October 5, 2015


Issue 38 What’s Going on in China? September 24, 2015

Issue 53 The Return of Reflation December 7, 2016

Issue 37 A Look at Liquidity August 2, 2015

Source of photos: www.istockphoto.com, www.shutterstock.com, US Department of State/Wikimedia Commons/Public Domain.

Top of Mind archive: click to access

https://research.gs.com/content/research/en/reports/2018/04/26/f0da881b-9273-4120-b74e-a161eb25d25e.pdf



https://360.gs.com/gs/portal/?st=1&action=action.binary&d=22823448&fn=/document.pdf


https://research.gs.com/content/research/en/reports/2018/03/28/fe2f67f9-ae9e-4db9-ba3f-14835f611db2.pdf





https://research.gs.com/content/research/en/reports/2018/02/28/b4445366-6f6c-4caf-99d6-bb7cfd9c9be8.pdf



https://360.gs.com/gs/portal/?st=1&action=action.binary&d=22548793&fn=/document.pdfhttps://360.gs.com/gs/portal/?st=1&action=action.binary&d=22505284&fn=/document.pdf


https://research.gs.com/content/research/en/reports/2018/02/05/42d255b3-256a-48ba-a3d0-da81b3eeb75a.pdf





https://research.gs.com/content/research/en/reports/2017/12/14/06a49172-d4ae-46b6-9339-c4f67d3efb0f.pdf



https://360.gs.com/gir/portal/?st=1&action=action.binary&d=22137354&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21929128&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21776501&fn=/document.pdf

https://360.gs.com/gir/portal/?st=1&action=action.binary&d=22137354&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21929128&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21776501&fn=/document.pdf

https://research.gs.com/content/research/en/reports/2017/11/09/4984370c-77ff-4eae-ae65-4d55fc859aef.pdf



https://360.gs.com/gir/portal/?st=1&action=action.binary&d=21929128&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21776501&fn=/document.pdf

https://360.gs.com/gir/portal/?st=1&action=action.binary&d=21929128&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=21776501&fn=/document.pdf

https://research.gs.com/content/research/en/reports/2017/10/12/e01e956e-01fa-494f-9cd2-ff17b3a7a01b.pdf



https://360.gs.com/gir/portal/?st=1&action=action.binary&d=21776501&fn=/document.pdf


https://360.gs.com/gir/portal/?action=action.doc&d=c6b94c6d6c244b6ea842aae25001bbb8





https://research.gs.com/content/research/en/reports/2017/09/11/282dc2c0-26f1-4580-88b1-1bea9faab93f.pdf





https://research.gs.com/content/research/en/reports/2017/07/26/2aae389e-f9a1-4681-8b6f-d86abb1038d9.pdf













































https://360.gs.com/gs/portal/?st=1&action=action.binary&d=22548793&fn=/document.pdfhttps://360.gs.com/gs/portal/?st=1&action=action.binary&d=22505284&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=22137354&fn=/document.pdf


https://360.gs.com/gs/portal/?st=1&action=action.binary&d=22505284&fn=/document.pdfhttps://360.gs.com/gir/portal/?st=1&action=action.binary&d=22137354&fn=/document.pdf


























El



Disclosure Appendix Reg AC We, Allison Nathan, Marina Grushin, David Groman, and Charlie Himmelberg hereby certify that all of the views expressed in this report accurately reflect our personal views, which have not been influenced by considerations of the firm's business or client relationships. We, Rocky Fishman, Christian Mueller-Glissmann, and Alessio Rizzi hereby certify that all of the views expressed in this report accurately reflect our personal views about the subject company or companies and its or their securities. We also certify that no part of our compensation was, is or will be, directly or indirectly, related to the specific recommendations or views expressed in this report. Unless otherwise stated, the individuals listed on the cover page of this report are analysts in Goldman Sachs' Global Investment Research division.

Distribution of ratings/ investment banking relationships

Disclosures required by United States laws and regulations See company-specific regulatory disclosures above for any of the following disclosures required as to companies referred to in this report: manager or co-manager in a pending transaction; 1% or other ownership; compensation for certain services; types of client relationships; managed/co-managed public offerings in prior periods; directorships; for equity securities, market making and/or specialist role. Goldman Sachs trades or may trade as a principal in debt securities (or in related derivatives) of issuers discussed in this report.

The following are additional required disclosures: Ownership and material conflicts of interest: Goldman Sachs policy prohibits its analysts, professionals reporting to analysts and members of their households from owning securities of any company in the analyst's area of coverage. Analyst compensation: Analysts are paid in part based on the profitability of Goldman Sachs, which includes investment banking revenues. Analyst as officer or director: Goldman Sachs policy generally prohibits its analysts, persons reporting to analysts or members of their households from serving as an officer, director or advisor of any company in the analyst's area of coverage. Non-U.S. Analysts: Non-U.S. analysts may not be associated persons of Goldman Sachs & Co. LLC and therefore may not be subject to FINRA Rule 2241 or FINRA Rule 2242 restrictions on communications with subject company, public appearances and trading securities held by the analysts.

Additional disclosures required under the laws and regulations of jurisdictions other than the United States

The following disclosures are those required by the jurisdiction indicated, except to the extent already made above pursuant to United States laws and regulations. Australia: Goldman Sachs Australia Pty Ltd and its affiliates are not authorised deposit-taking institutions (as that term is defined in the Banking Act 1959 (Cth)) in Australia and do not provide banking services, nor carry on a banking business, in Australia. This research, and any access to it, is intended only for "wholesale clients" within the meaning of the Australian Corporations Act, unless otherwise agreed by Goldman Sachs. In producing research reports, members of the Global Investment Research Division of Goldman Sachs Australia may attend site visits and other meetings hosted by the companies and other entities which are the subject of its research reports. In some instances the costs of such site visits or meetings may be met in part or in whole by the issuers concerned if Goldman Sachs Australia considers it is appropriate and reasonable in the specific circumstances relating to the site visit or meeting. To the extent that the contents of this document contains any financial product advice, it is general advice only and has been prepared by Goldman Sachs without taking into account a client's objectives, financial situation or needs. A client should, before acting on any such advice, consider the appropriateness of the advice having regard to the client's own objectives, financial situation and needs. Brazil: Disclosure information in relation to CVM Instruction 483 is available at http://www.gs.com/worldwide/brazil/area/gir/index.html. Where applicable, the Brazil-registered analyst primarily responsible for the content of this research report, as defined in Article 16 of CVM Instruction 483, is the first author named at the beginning of this report, unless indicated otherwise at the end of the text. Canada: Goldman Sachs Canada Inc. is an affiliate of The Goldman Sachs Group Inc. and therefore is included in the company specific disclosures relating to Goldman Sachs (as defined above). Goldman Sachs Canada Inc. has approved of, and agreed to take responsibility for, this research report in Canada if and to the extent that Goldman Sachs Canada Inc. disseminates this research report to its clients. Hong Kong: Further information on the securities of covered companies referred to in this research may be obtained on request from Goldman Sachs (Asia) L.L.C. India: Further information on the subject company or companies referred to in this research may be obtained from Goldman Sachs (India) Securities Private Limited, Research Analyst - SEBI Registration Number INH000001493, 951-A, Rational House, Appasaheb Marathe Marg, Prabhadevi, Mumbai 400 025, India, Corporate Identity Number U74140MH2006FTC160634, Phone +91 22 6616 9000, Fax +91 22 6616 9001. Goldman Sachs may beneficially own 1% or more of the securities (as such term is defined in clause 2 (h) the Indian Securities Contracts (Regulation) Act, 1956) of the subject company or companies referred to in this research report. Japan: See below. Korea: Further information on the subject company or companies referred to in this research may be obtained from Goldman Sachs (Asia) L.L.C., Seoul Branch. New Zealand: Goldman Sachs New Zealand Limited and its affiliates are neither "registered banks" nor "deposit takers" (as defined in the Reserve Bank of New Zealand Act 1989) in New Zealand. This research, and any access to it, is intended for "wholesale clients" (as defined in the Financial Advisers Act 2008) unless otherwise agreed by Goldman Sachs. Russia: Research reports distributed in the Russian Federation are not advertising as defined in the Russian legislation, but are information and analysis not having product promotion as their main purpose and do not provide appraisal within the meaning of the

El



Russian legislation on appraisal activity. Singapore: Further information on the covered companies referred to in this research may be obtained from Goldman Sachs (Singapore) Pte. (Company Number: 198602165W). Taiwan: This material is for reference only and must not be reprinted without permission. Investors should carefully consider their own investment risk. Investment results are the responsibility of the individual investor. United Kingdom: Persons who would be categorized as retail clients in the United Kingdom, as such term is defined in the rules of the Financial Conduct Authority, should read this research in conjunction with prior Goldman Sachs research on the covered companies referred to herein and should refer to the risk warnings that have been sent to them by Goldman Sachs International. A copy of these risks warnings, and a glossary of certain financial terms used in this report, are available from Goldman Sachs International on request.

European Union: Disclosure information in relation to Article 4 (1) (d) and Article 6 (2) of the European Commission Directive 2003/125/EC is available at http://www.gs.com/disclosures/europeanpolicy.html which states the European Policy for Managing Conflicts of Interest in Connection with Investment Research.

Japan: Goldman Sachs Japan Co., Ltd. is a Financial Instrument Dealer registered with the Kanto Financial Bureau under registration number Kinsho 69, and a member of Japan Securities Dealers Association, Financial Futures Association of Japan and Type II Financial Instruments Firms Association. Sales and purchase of equities are subject to commission pre-determined with clients plus consumption tax. See company-specific disclosures as to any applicable disclosures required by Japanese stock exchanges, the Japanese Securities Dealers Association or the Japanese Securities Finance Company.

Ratings, coverage groups and views and related definitions Buy (B), Neutral (N), Sell (S)-Analysts recommend stocks as Buys or Sells for inclusion on various regional Investment Lists. Being assigned a Buy or Sell on an Investment List is determined by a stock's total return potential relative to its coverage. Any stock not assigned as a Buy or a Sell on an Investment List with an active rating (i.e., a stock that is not Rating Suspended, Not Rated, Coverage Suspended or Not Covered), is deemed Neutral. Each regional Investment Review Committee manages various regional Investment Lists to a global guideline of 25%-35% of stocks as Buy and 10%-15% of stocks as Sell; however, the distribution of Buys and Sells in any particular analyst's coverage group may vary as determined by the regional Investment Review Committee. Additionally, each Investment Review Committee manages Regional Conviction lists, which represent investment recommendations focused on the size of the total return potential and/or the likelihood of the realization of the return across their respective areas of coverage. The addition or removal of stocks from such Conviction lists do not represent a change in the analysts' investment rating for such stocks.

Total return potential represents the upside or downside differential between the current share price and the price target, including all paid or anticipated dividends, expected during the time horizon associated with the price target. Price targets are required for all covered stocks. The total return potential, price target and associated time horizon are stated in each report adding or reiterating an Investment List membership.

Coverage groups and views: A list of all stocks in each coverage group is available by primary analyst, stock and coverage group at http://www.gs.com/research/hedge.html. The analyst assigns one of the following coverage views which represents the analyst's investment outlook on the coverage group relative to the group's historical fundamentals and/or valuation. Attractive (A). The investment outlook over the following 12 months is favorable relative to the coverage group's historical fundamentals and/or valuation. Neutral (N). The investment outlook over the following 12 months is neutral relative to the coverage group's historical fundamentals and/or valuation. Cautious (C). The investment outlook over the following 12 months is unfavorable relative to the coverage group's historical fundamentals and/or valuation.

Not Rated (NR). The investment rating and target price have been removed pursuant to Goldman Sachs policy when Goldman Sachs is acting in an advisory capacity in a merger or strategic transaction involving this company and in certain other circumstances. Rating Suspended (RS). Goldman Sachs Research has suspended the investment rating and price target for this stock, because there is not a sufficient fundamental basis for determining, or there are legal, regulatory or policy constraints around publishing, an investment rating or target. The previous investment rating and price target, if any, are no longer in effect for this stock and should not be relied upon. Coverage Suspended (CS). Goldman Sachs has suspended coverage of this company. Not Covered (NC). Goldman Sachs does not cover this company. Not Available or Not Applicable (NA). The information is not available for display or is not applicable. Not Meaningful (NM). The information is not meaningful and is therefore excluded.

Global product; distributing entities The Global Investment Research Division of Goldman Sachs produces and distributes research products for clients of Goldman Sachs on a The Global Investment Research Division of Goldman Sachs produces and distributes research products for clients of Goldman Sachs on a global basis. Analysts based in Goldman Sachs offices around the world produce equity research on industries and companies, and research on macroeconomics, currencies, commodities and portfolio strategy. This research is disseminated in Australia by Goldman Sachs Australia Pty Ltd (ABN 21 006 797 897); in Brazil by Goldman Sachs do Brasil Corretora de Títulos e Valores Mobiliários S.A.; Ombudsman Goldman Sachs Brazil: 0800 727 5764 and / or [email protected]. Available Weekdays (except holidays), from 9am to 6pm. Ouvidoria Goldman Sachs Brasil: 0800 727 5764 e/ou [email protected]. Horário de funcionamento: segunda-feira à sexta-feira (exceto feriados), das 9h às 18h; in Canada by either Goldman Sachs Canada Inc. or Goldman Sachs & Co. LLC; in Hong Kong by Goldman Sachs (Asia) L.L.C.; in India by Goldman Sachs (India) Securities Private Ltd.; in Japan by Goldman Sachs Japan Co., Ltd.; in the Republic of Korea by Goldman Sachs (Asia) L.L.C., Seoul Branch; in New Zealand by Goldman Sachs New Zealand Limited; in Russia by OOO Goldman Sachs; in Singapore by Goldman Sachs (Singapore) Pte. (Company Number: 198602165W); and in the United States of America by Goldman Sachs & Co. LLC. Goldman Sachs International has approved this research in connection with its distribution in the United Kingdom and European Union.

European Union: Goldman Sachs International authorised by the Prudential Regulation Authority and regulated by the Financial Conduct Authority and the Prudential Regulation Authority, has approved this research in connection with its distribution in the European Union and United Kingdom; Goldman Sachs AG and Goldman Sachs International Zweigniederlassung Frankfurt, regulated by the Bundesanstalt für Finanzdienstleistungsaufsicht, may also distribute research in Germany.

El



General disclosures This research is for our clients only. Other than disclosures relating to Goldman Sachs, this research is based on current public information that we consider reliable, but we do not represent it is accurate or complete, and it should not be relied on as such. The information, opinions, estimates and forecasts contained herein are as of the date hereof and are subject to change without prior notification. We seek to update our research as appropriate, but various regulations may prevent us from doing so. Other than certain industry reports published on a periodic basis, the large majority of reports are published at irregular intervals as appropriate in the analyst's judgment. Goldman Sachs conducts a global full-service, integrated investment banking, investment management, and brokerage business. We have investment banking and other business relationships with a substantial percentage of the companies covered by our Global Investment Research Division. Goldman Sachs & Co. LLC, the United States broker dealer, is a member of SIPC (http://www.sipc.org). Our salespeople, traders, and other professionals may provide oral or written market commentary or trading strategies to our clients and principal trading desks that reflect opinions that are contrary to the opinions expressed in this research. Our asset management area, principal trading desks and investing businesses may make investment decisions that are inconsistent with the recommendations or views expressed in this research. The analysts named in this report may have from time to time discussed with our clients, including Goldman Sachs salespersons and traders, or may discuss in this report, trading strategies that reference catalysts or events that may have a near-term impact on the market price of the equity securities discussed in this report, which impact may be directionally counter to the analysts' published price target expectations for such stocks. Any such trading strategies are distinct from and do not affect the analysts' fundamental equity rating for such stocks, which rating reflects a stock's return potential relative to its coverage group as described herein. We and our affiliates, officers, directors, and employees, excluding equity and credit analysts, will from time to time have long or short positions in, act as principal in, and buy or sell, the securities or derivatives, if any, referred to in this research. The views attributed to third party presenters at Goldman Sachs arranged conferences, including individuals from other parts of Goldman Sachs, do not necessarily reflect those of Global Investment Research and are not an official view of Goldman Sachs. Any third party referenced herein, including any salespeople, traders and other professionals or members of their household, may have positions in the products mentioned that are inconsistent with the views expressed by analysts named in this report. This research is not an offer to sell or the solicitation of an offer to buy any security in any jurisdiction where such an offer or solicitation would be illegal. It does not constitute a personal recommendation or take into account the particular investment objectives, financial situations, or needs of individual clients. Clients should consider whether any advice or recommendation in this research is suitable for their particular circumstances and, if appropriate, seek professional advice, including tax advice. The price and value of investments referred to in this research and the income from them may fluctuate. Past performance is not a guide to future performance, future returns are not guaranteed, and a loss of original capital may occur. Fluctuations in exchange rates could have adverse effects on the value or price of, or income derived from, certain investments. Certain transactions, including those involving futures, options, and other derivatives, give rise to substantial risk and are not suitable for all investors. Investors should review current options disclosure documents which are available from Goldman Sachs sales representatives or at http://www.theocc.com/about/publications/character-risks.jsp. Transaction costs may be significant in option strategies calling for multiple purchase and sales of options such as spreads. Supporting documentation will be supplied upon request. Differing Levels of Service provided by Global Investment Research: The level and types of services provided to you by the Global Investment Research division of GS may vary as compared to that provided to internal and other external clients of GS, depending on various factors including your individual preferences as to the frequency and manner of receiving communication, your risk profile and investment focus and perspective (e.g., marketwide, sector specific, long term, short term), the size and scope of your overall client relationship with GS, and legal and regulatory constraints. As an example, certain clients may request to receive notifications when research on specific securities is published, and certain clients may request that specific data underlying analysts' fundamental analysis available on our internal client websites be delivered to them electronically through data feeds or otherwise. No change to an analyst's fundamental research views (e.g., ratings, price targets, or material changes to earnings estimates for equity securities), will be communicated to any client prior to inclusion of such information in a research report broadly disseminated through electronic publication to our internal client websites or through other means, as necessary, to all clients who are entitled to receive such reports. All research reports are disseminated and available to all clients simultaneously through electronic publication to our internal client websites. Not all research content is redistributed to our clients or available to third-party aggregators, nor is Goldman Sachs responsible for the redistribution of our research by third party aggregators. For research, models or other data related to one or more securities, markets or asset classes (including related services) that may be available to you, please contact your GS representative or go to http://360.gs.com. Disclosure information is also available at http://www.gs.com/research/hedge.html or from Research Compliance, 200 West Street, New York, NY 10282. This material has been prepared by Goldman Sachs Global Investment Research and is not a product of GSAM. It is financial research prepared in compliance with applicable provisions of law designed to promote the independence of financial analysis and is subject to a prohibition on trading following the distribution of financial research. The views and opinions expressed may differ from those of GSAM or other departments or divisions of Goldman Sachs and its affiliates. This information discusses general market activity, industry or sector trends, or other broad-based economic, market or political conditions and should not be construed as investment advice. Investors are urged to consult with their financial advisors before buying or selling any securities. This information may not be current and Goldman Sachs Global Investment Research has no obligation to provide any updates or change.

© 2018 Goldman Sachs. No part of this material may be (i) copied, photocopied or duplicated in any form by any means or (ii) redistributed without the prior written consent of The Goldman Sachs Group, Inc.

Equity Research - Goldman Sachs Asset Management

Documents

Transcript of Equity Research - Goldman Sachs Asset Management