Schwartz (eds ) volatility; risk and uncertainty in financial markets (2011)

Inaccurate price discovery contributes to volatility, and good price discovery is difficult to achieve, especially when some investors’ are influenced by what they see other investors do

Zicklin School of Business Financial Markets Series

Robert A Schwartz, Editor

Baruch College/CUNY

Zicklin School of Business

New York, NY, USA

For other titles published in this series, go to

www.springer.com/series/7133

Other Books in the Series:

Technology and Regulation

Schwartz, Robert A., Byrne, John A.,

The New NASDAQ Marketplace

Schwartz, Robert A., Byme, John A., Colaninno, Antoinette:

Electronic vs Floor Based Trading

Coping with Institutional Order Flow

Schwartz, Robert A., Byrne, John A Colaninno, Antoinette:

A Trading Desk View of Market Qualily

Schwartz, Robert A., Byre, John A., Colaninno, Antoinette:

Call Auction Trading: New Answers to Old Questions

Schwartz, Robert A.: and Colaninno, Antoinette:

Regulation of Equity Markets

Schwartz, Robert A., Byrne, John A.,

Colaninno, Antoinette: Colaninno, Antoinette:

U.S

Schwartz, Robert A., Byrne, John A., Colaninno, Antoinette:

Schwartz, Robert A., Byrne, John A Colaninno, Antoinette:

Competition in a Consolidating Environment

Printed on acid-free paper

Springer New York Dordrecht Heidelberg London

All rights reserved This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis Use in connection with any form of information storage and retrieval, electronic adaptation, forbidden

The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights

Robert A Schwartz

Department of Finance

Zicklin School of Business

Baruch College, CUNY

USA

John Aidan Byrne Rockaway, NJ USA

Zicklin School of Business

Baruch College, CUNY

Contents

Preface xi

Chapter 3: What Is Happening With Financial Market Volatility and Why? 29

List of Participants

Robert Almgren New York University Adjunct Professor of Mathematical

Finance George Bodine General Motors Investment

Management

Director of Trading Harold Bradley Ewing Marion Kauffman

Foundation

Chief Investment Officer Erin Burnett CNBC Anchor and Reporter

Ian Domowitz Investment Technology Group Managing Director

Brendan Doran Chi-X Europe Limited Vice President, Business

Development Robert Engle Stern School of Business, NYU Michael Armellino Professor of

Finance, Nobel Laureate 2003 Reto Francioni Deutsche Börse AG CEO

Sandy Frucher The NASDAQ OMX Group Vice Chair

William Geyer JonesTrading Institutional

Services, LLC

CEO and President

Al Goll Commodity Futures Trading

Commission

Auditor Ken Hight Liquidnet, Inc Head of Global Equities

Brian Hyndman The NASDAQ OMX Group Senior Vice President

Tim Mahoney Bids Trading CEO

Terrence Martell Baruch College, CUNY Director, Weissman Center for

International Business, Saxe Distinguished Professor of Finance Albert J Menkveld VU University Amsterdam Associate Professor of Finance

Matt Moran Chicago Board Options

Exchange

Vice President

Deniz Ozenbas Montclair State University Associate Professor of Finance Michael Pagano Villanova University Professor of Finance

Richard Rosenblatt Rosenblatt Securities Inc CEO

James Ross NYSE Euronext Vice President NYSE Crossing Keith Ross PDQ Enterprises, LLC CEO

Asani Sarkar Federal Reserve Bank of New

York

Economist Stephen Sax FBN Securities Inc Vice President

Alec Schmidt ICAP Electronic Broking Senior Analyst

Robert Schwartz Baruch College, CUNY Speiser Professor of Finance Robert Shapiro Morgan Stanley Investment

Management

Executive Director Larry Tabb The Tabb Group Founder and CEO

Grant Vingoe Arnold and Porter Partner

Henri Waelbroeck Pipeline Trading Systems Vice President, Director of

Research Joseph Wald Knight Capital/EdgeTrade Managing Director Robert Wood University of Memphis Distinguished Professor of Finance Liuren Wu Baruch College, CUNY Associate Professor of Finance

Pipeline Trading Systems

Rosenblatt Securities Inc

The NASDAQ OMX Group

White Cap Trading LLC

Preface

Volatility and risk are of fundamental importance to the finance practitioners among us Indeed, volatility and risk are practically at the center of all our work Finance, as a subject, would not exist without them in our business school curriculum, nor in our academic research Simply put, finance would

be indistinguishable from deterministic economics For that matter, the presence of volatility and risk also bestows significant influence on the finance departments in banks and industrial firms

By the same token, terms like the following are part of our everyday financial parlance: Risk aversion, risk hedging, risk management, value at risk, risk measurement and risk premium In our industry, we have high-powered minds, high-powered valuation formulas, high-powered trading algorithms, and high-powered electronic technology to pull it all together And yet, today’s events show us what risk is really about, how at risk our financial markets truly are The events of the last several months also show

us how much we do not know

Let me contrast our group with an ant, yes, that little red or black creature that can crawl around and annoy us An ant has actually been classified as one of the dumbest creatures on earth! Yet, collectively, they are very intelligent Look at how are also highly intelligent These are the brilliant quants, financial engineers, entrepreneurs, academicians (if I may be so bold), the PhD’s in chemistry, physics, and mathematics, and so on and so forth Finance has attracted many of the brightest minds to its ranks Still, collectively, as we face these bear market conditions right now, we do not seem so smart As a group, we have just run into a startling, frightening hole Are we exactly the opposite of ants? How can we individually be so brilliant and, at the same time, collectively be so very dumb? Where are our intelligent answers and solutions for today’s challenging markets?

There is so much about volatility and risk that we do not understand Even more critically, there is a substantial amount of behavior about volatility and risk that we think we understand but, in truth, do not understand This kind of ignorance (mistakenly thinking we know our subject) can really come back and bite us

Risk isn’t the only contributor to volatility, and I believe we have lost sight of this Risk has a well-defined meaning to economists Risk exists when an outcome can be described as a draw from a probability distribution with known parameters Flip a fair coin and bet on the outcome: the chance

of heads equals 50%; the chance of tails equals 50% But beyond that we do not know the outcome until after we have flipped the coin That is risk, clear and simple In this type of scenario, we will perform a decent job of modeling risk given the probability distributions

However, along with risk, there is also uncertainty Here we do not know

the probability distributions In fact, we might not even know what all of the outcomes even are Uncertainty presents a huge challenge In my opinion,

we have not paid sufficient formal attention to uncertainty as a cause of volatility

Also high on the list of our ignorance is systemic risk and uncertainty In

free markets individual firms will fail Their demise may be understood in the light of Adam Smith’s invisible hand, or of Joseph Schumpeter’s creative destruction Systemic risk is another matter When a systemic breakdown occurs, it is the free market itself that has failed

High volatility has been with us for over a year now In my research, I have been focused on this topic for much longer, for many years Now, if you were to pick one word to describe our markets, what would that word be? My choice would be ‘volatility.’ So let’s go for it Let’s focus on this key property of a financial market I am not thinking of price fluctuations over lengthy, multi-year periods I do not have in mind risk and uncertainty about the more distant future I am thinking of the very appreciable volatility that we experience, day after day, on an intra-day basis In today’s turbulent environment, intra-day volatility is dramatic

We talk about Wall Street versus Main Street Financial markets are

absolutely essential for the smooth functioning of our broad economy There is, therefore, a huge connect between Wall Street and Main Street Financial capital enables firms to operate, just as oil enables physical capital, from bikes, to bulldozers, to airplanes, to run But the financial markets are also fragile We do not always think about it; and in ‘normal’ times we do not even see it But they are fragile This is especially so in today’s high frequency, electronic environment, given the large pools of capital that today can slosh anywhere around the world at a microsecond moment’s notice Take a look with a magnifying glass at the price movements, the swings that take place intra-day on a daily basis Price changes of one percent, two

percent or more are commonplace A one percent daily price move, annualized, translates into 250% We do not very often see annual swings of this magnitude In the opening and closing seconds and minutes of trading, intra-day price movements are even more accentuated How come? What explains it?

Academic evidence of accentuated daily and intra-day price volatility has accumulated over the years In a paper that I am currently completing with Mike Pagano and Lin Peng, we present evidence on volatility for a large sample of NASDAQ stocks for the year 2005.1 It was very striking that the three most volatile minutes in a trading day are the two minutes that follow the open, and the final minute that precedes the close What explains the accentuated intra-day price volatility? Why are the financial markets so fragile? I will briefly address two related items: price discovery and liquidity creation

I have been focusing on price discovery for many years Throughout, I have noted its importance in various publications and in my talks The fact

is security prices – the value of shares – are not found in the upstairs offices

of the stock analysts They are discovered in the marketplace

Share prices are not intrinsic values Share prices do not follow random

walks, and they are not simply and uniquely linked to ‘the fundamentals.’

How can they be when, in the face of enormously complex, imprecise, and incomplete information, investors form diverse expectations of future corporate performance? Thus, at any current moment, they evaluate shares differently And markets are not as informationally efficient as some of my colleagues would like to think I am not a proponent of the Efficient Markets Hypothesis (or EMH, as we like to say) I suggest that the word

‘efficient’ be replaced The proper adjective, in my opinion, is ‘humbling.’ The markets are indeed humbling

Inaccurate price discovery contributes to volatility, and good price discovery is difficult to achieve, especially when some investors’ are influenced by what they see other investors doing That is when we get information cascades That is when we get herding That is when volatility blows up When these things happen, a market can run into trouble

Arm-in-arm with price discovery is liquidity creation I have just completed a paper on this topic with Asani Sarkar and Nick Klagge, both from the New York Fed.2 In addressing the dynamic process of liquidity

creation, we consider something that we call the sidedness of markets

1 Pagano, M., Peng, L., and Schwartz, R., ‘The Quality of Price Formation at Market Openings and Closings: Evidence from the NASDAQ Stock Market.’

2 Klagge, N., Sarkar, A and Schwartz, R., ‘Liquidity Begets Liquidity: Implications for a Dark Pool Environment,’ Institutional Investor’s Guide to Global Liquidity, Winter 2009,

pp 15-20

Sidedness refers to the extent to which buyers and sellers are both actively present in a market, in roughly equal proportions, in brief periods of time (e.g., five minute intervals)

In previous work, Asani Sarkar and I have found that markets are generally two-sided, and that two-sidedness holds under a wide range of conditions.3 It holds for both NASDAQ and NYSE stocks; at market openings, mid-day, and at the close; on days with news and on days when there is no major news; and for both large orders and small orders We also observe that buyers and sellers tend to arrive in clusters, that within a day, two-sided trading bursts are commonly interspersed with periods of relative inactivity

But markets are not always two-sided At times, liquidity dries up on one side of the market and volatility spikes Information cascades and herding can take over, and a market can become one-sided Even if potential buyers and sellers are both in the offing, neither may be making their presence known And, when prices suddenly head south, one-sidedness is accentuated as buyers simply step aside Who wants to step up and try to catch the falling knife?

What are the conditions that lead to two-sidedness? What are the factors that trigger trade bursts? What causes a market to be one-sided? Illiquidity

is a cause of volatility and its counterpart, liquidity, does not just happen Liquidity creation is a process There is a great deal more that we need to learn about the process, about the dynamics of liquidity creation

As we all know, opacity is needed by the big players The large traders seek the protection of opacity by either going to a dark pool or, when going

to a more transparent limit order book market, by hiding their orders in a stream of retail flow by slicing and dicing them Nevertheless, there is post-trade reporting for all trades, and information can be gleaned on the general sidedness of markets

Opacity is one thing; fragmentation is another Whether liquidity pools are light or dark, fragmentation can disrupt the natural two-sidedness of markets Can connectivity between the dark pools that exist today in the U.S be effective? The real concern about the dark pools of today is not that they are dark; it is that connectivity may not be a viable substitute for consolidation

It is well known that order flow attracts order flow We have also seen that, over time, the equity markets have generally tended to consolidate Consolidation and two-sidedness are natural processes for an equity market They are the main dynamics that underlie liquidity creation However, modern technology facilitates the increased fragmentation of markets, and it

3 Sarkar, A and Schwartz, R., ‘Market Sidedness: Insights into Motives for Trade Initiation,’ Journal of Finance, February 2009, pp 375-423

supports the possibility of fragile, one-sided markets proliferating True, advanced technology also facilitates a greater integration of markets, but such liquidity aggregation may prove to be inadequate The extent to which the natural two-sidedness of markets stays resilient in the face of these developments remains to be seen, and the efficacy of liquidity creation hangs

in the balance

And then there is the temporal dimension of fragmentation I have, for a long time, been a proponent of electronic call auction trading I have long urged that calls be included in our predominantly continuous trading environment to open and to close markets A call is an explicit price discovery mechanism A call amasses liquidity at specific points in time A call delivers price improvement for participants who place aggressive limit orders, and this encourages them to, in fact, place aggressive limit orders The amassing of liquidity and the delivery of price improvement in call auction trading means that a call is more likely to deliver a two-sided market than its continuous market counterpart Mike Pagano, Lin Peng and I have done some analysis of NASDAQ’s new calls, and it appears that the calls have achieved volatility decreases that are both substantial and statistically significant.4

Another market structure feature that goes to the heart of the volatility issue is circuit breakers, or, as they are called in Germany, volatility interruptions In my opinion, volatility interruptions, which are brief, firm-specific trading halts, have some very desirable properties The interruptions are a check against order placement errors Most importantly, they also enable the market to switch from continuous trading to call auction trading;

in so doing, they sharpen the accuracy of price discovery

In addition to calls, circuit breakers, and volatility interruptions, there are other market structure solutions to the problem of extreme market

turbulence After the crash of ’87, I proposed the establishment of voluntary

stabilization funds that would buy and sell equity shares according to a strict and well-defined procedure A fund could be established by a listed company itself and run by a third-party fiduciary In a falling market, shares

of the company’s stock would be bought by the fund and, conversely, shares would be sold by the fund in a rising market The fund’s buy and sell orders would be submitted at pre-specified price points, in pre-specified amounts And, most importantly, these shares would be bought and sold in call auction trading only

This type of voluntary procedure would disrupt herding, it would bolster the two-sidedness of markets, and it would help to curb the bouts of sharply accentuated volatility that can occur at any time, and which have occurred in

4 Pagano, M., Peng, L., and Schwartz, R., ‘The Quality of Price Formation at Market Openings and Closings: Evidence from the NASDAQ Stock Market.’

full force since Labor Day 2008 My paper proposing this voluntary procedure was published 20 years ago.5 I still support the proposal today Dynamism and allocational efficiency are two powerfully positive attributes of a free market Instability is a free market’s Achilles heel In the last several months we have been hit by tidal waves of volatility Now fingers are being pointed at many factors, including the housing bubble, greed, hubris, accounting rule changes, the absence of certain short-selling restrictions, management failure, government failure, regulatory failure, and market structure failure In my opinion, regulatory intervention and market structure, stand out These two, if properly designed and implemented, could

do much to better stabilize our markets in a risky and uncertain world

In the final analysis, it is not a matter of free markets versus regulated markets Regulation is indeed needed But it must be appropriate A better understanding is required of the issues, concerns, and market failure realities upon which regulations should be based The sources of government failure must also be taken fully into account Excessive and ill-structured regulation can be extremely costly to financial markets in particular, and to society at large I hope that, after the dust has settled, we have achieved a stronger market structure, and a more appropriate regulatory structure But this much is certain: the financial turbulence of 2008 has provided us with an abundant amount of material to think about

Robert Schwartz

5 Schwartz, R., ‘A Proposal to Stabilize Stock Prices,’ Journal of Portfolio Management, Fall

1988, pp 4 - 11 Translated into Italian and published in Rivista Della Borsa, August

1989 Reprinted in Journal of Trading, Volume 4, Number 2, Spring 2009, pp 50-57

CHAPTER 1: INTRADAY VOLATILITY: THE EMPIRICAL EVIDENCE

Moderator: Asani Sarkar, Research Officer, Federal Reserve Bank of New York

Robert Almgren, Adjunct Professor of Mathematical Finance, New York University

Albert J Menkveld, Associate Professor of Finance, VU University Amsterdam

Liuren Wu, Associate Professor of Finance, Baruch College, CUNY

ASANI SARKER: We now have a new forecasting tool! It is in the very title of Bob Schwartz’s next conference (laughter) The forecast clearly worked well for this year’s conference because, of course, it is cleverly titled

‘volatility’ – and volatility, as you all well know, is major financial news today Volatility is significantly present in today’s challenging markets So,

if the title of next year’s conference is, say, ‘Negative Skewness of Returns,’ then we are really in for big trouble! However, if it is ‘Positive Skewness of Returns,’ then we can be very happy about the future (laughter)!

Our panelists today bring distinctive points of view from their own research on volatility and its impact on the marketplace Bob Schwartz talked earlier about Main Street and Wall Street One could argue that Main Street really only cares about macroeconomic volatility (such as the GDP growth) But for a finance academician, what is the relationship between financial market volatility and macroeconomic volatility? One of the important economic stylized facts of the last 20 years is that macroeconomic volatility has been falling on a secular basis In other words, volatility is secularly lower, even after adjusting for business cycles Macroeconomic volatility over the past 20 years is substantially lower than over the previous 30 years A manifestation of this is the shorter business cycles that we observe That is, the duration of the NBER-dated

1volatility of consumption, the volatility of inflation, and the volatility of

School of Business Financial Markets Series, DOI 10.1007/978-1-4419-1474-3_1,

R.A Schwartz et al (eds.), Volatility: Risk and Uncertainty in Financial Markets, Zicklin

© Springer Science+Business Media, LLC 2011

cycles are shorter.6 This is obviously good news for consumers, businesses and workers alike

At the same time, we do not observe a similar decline in financial market volatility For example, if you take the period 1995 to early 2000, macro-volatility was going down substantially, but the CBOE Volatility Index, or VIX7 (the equity market implied volatility) was actually rising during that same period This is a puzzle One would think that the factors that are causing macro-volatility to decline would have a similar affect on financial market volatility Yet those two measures of volatility seem to be moving in opposite directions Why is this the case? This is not very well understood at this point

One hypothesis is that financial innovation may be contributing to this The question is, why do we have output volatility? If you have a negative shock (for instance, an increase in the price of oil), then a firm has to adjust its capital structure, it has to decrease its leverage This is difficult to do because of various adjustment costs (for example, issuing equity financing is costly) Therefore, the firm cannot make the adjustment easily This leads

to a large decrease in investment and, therefore, a large decrease in output This is what triggers the output volatility

Financial innovation (by this academic researchers mean things like securitization and new forms of financing that essentially increase the financial flexibility of firms) decreases adjustment costs It also makes it easier for firms to adjust their capital structures to de-leverage during bad times Therefore, they do not need to reduce their investment during bad times, and so output does not go down as much It reduces output volatility But financial innovation also increases financial volatility Because the cost of financing is lower, the quantity of financing is more sensitive to asset pricing So you have large fluctuations in the issuance of debt and equity This leads, in turn, to greater financial volatility Interesting implications for the current environment follow from this development

One of the main factors in the current crisis is securitization and structured financing One conjecture might be that, because of all of the problems in this market, we should expect to see reduced securitization or reduced financial innovation, or at least an increased cost in financial

6 The National Bureau of Economic Research (NBER) is the major U.S non-profit economic research organization

7 VIX is an indicator of the expected future movement of the S&P over the coming 30 days, derived from the implied volatility premiums observed in the S&P index options market It

is often referred to as the ‘fear index’ because investors are prepared to pay higher premiums for option protection when more volatile markets are anticipated This means that, as the VIX rises higher, so does the expectation of more short-term risk by the markets

innovation What implication would this have for macroeconomic volatility and financial volatility?

Clearly, there is a lot we can focus on in this panel So, let’s get started

I will first ask each of the panelists to give a brief presentation Let’s begin with you Rob Almgren

ROBERT ALMGREN: I am speaking from the point of view of one of the ants that Bob Schwartz talked about I really do not have any opinions about whether volatility is good or bad, or about its causes I am speaking as

an agency algorithmic trader, working for a broker-dealer8 to execute transactions within a day, formerly in equities and now in futures Our concern is not whether volatility is good or bad, or whether or not the market should offer periodic crosses We only care about the execution of this transaction relative to a specific benchmark Volatility is simply a market property we have to measure, much as we measure the spread, the cumulative volume, or anything else

I will talk about the technology to do that, and, in particular, the importance of having an intra-day measure, real-time, of volatility You can characterize volatility across the term structure, across time, and across different products such as futures (By the way, there is relatively recent academic literature about this.) The obvious approach to measuring volatility

is to sample the price process at five-minute intervals, and then take the standard deviation That is your volatility But in each of those five-minute intervals, there may be hundreds or thousands of individual trades, or quote updates It is ridiculous to throw out that information in estimating the volatility

In particular, if you are trying to do something like steer an algorithm that adapts its execution to real-time variations in liquidity and volatility, then you want to know what the volatility was over the last minute, and what

it will likely be in the next minute Was it high, or was it low? Or, should

we speed up or slow down our algorithm? You cannot do that by averaging over time

Here are some exhibits that show what futures markets look like In Exhibit 1, we see the issues that we have to deal with in futures markets The first problem is that futures trade almost 24 hours a day This is a Euro-dollar future: the June 2008 contract that traded on April 1st 2008 Futures trade from about 5:00 p.m in Chicago to 4:00 p.m in Chicago time

8 Dr Almgren co-founded New York City-based Quantitative Brokers in 2008 and, at the firm, he oversees quantitative research and analysis of best execution algorithms and transactional cost measurement He also maintains his academic role in mathematical finance at New York University Quantitative Brokers describes itself as a fully independent and privately owned agency-only broker that specializes in execution algorithms for U.S interest rate futures

Exhibit 1 A Euro-dollar future: June 2008 Contract, Traded on April 1 st 2008

What we cannot see here is that, even though the market is open all night, there is very little trading activity in the night It is almost a 24-hour market, but most of the activity is during the day You want to measure volatility that somehow filters out the overnight

In addition, we also want to characterize what happened to the volatility just before the New York market’s open, and just after the New York market’s close, depending on the asset Stock futures, for example, very closely track the stock market over a period of time As shown in Exhibit 2, the data that you have to look at is very complicated if you are looking at it

in detail If you are trying to measure volatility minute by minute, there may

be several different bid and offer prices In equities, there are bids and offers from many exchanges In futures, there are bids and offers from direct quotes and implied quotes Direct quotes are entered specifically for the contract traded Implied quotes are generated as combinations of other quotes in the market, for example, an order filled as a calendar spread plus another contract Then you have trade prices that bounce between the bid and the offer The question is, exactly what volatility are we talking about? Remember, when you look at an overall picture like this, you are not seeing

a Brownian motion9 by any sort of measurable definition Nevertheless, you

9 Brownian motion (named after the Scottish botanist Robert Brown) is the seemingly random movement of particles suspended in a fluid (i.e a liquid or gas) or the mathematical model used to describe such random movements The mathematical model of Brownian motion

is a useful idealization of many real-world applications An often quoted example is stock market fluctuations

Close 4 PM Chicago (17:00 NY)

Open but slow overnight

Open 5 PM Chicago (18:00 NY)

EDT time on Tue 01 Apr 2008 20:00

still have to attach some number, which is the volatility, and you have to do

it from data like this

Exhibit 2 Research on Intra-Day Estimators

There are now fairly sophisticated ways to filter out the bid/offer noise One of the easy tricks is to use the bid and offer mid-point instead of the trade price With such a technique, you can update volatility minute by minute As shown in Exhibit 3, you can construct things like the term structure of volatility

Exhibit 3 Term Structure of Eurodollar Futures

EDT time on Tue 01 Apr 2008 11:00

This is volatility across different expirations Not surprisingly, the volatility varies by about a factor of two from short-term futures out to longer-term futures

SARKAR: Professor Menkveld?

ALBERT MENKVELD: I have two points First, I am trying to take a more generic approach to what we should make of daily price changes in our markets I will decompose volatility into two components: transitory volatility and permanent volatility It is important to measure both components of volatility to understand where the overall volatility is coming from Second, I want to make a point about algorithmic trading I come from Europe where we have been trading in electronic markets for some time now – at the German stock exchange (Deutsche Börse), Euronext, and the London Stock Exchange Once you are able to trade electronically, the big step is to design algorithms for electronic trading.10 We have looked at what this means for volatility and, in particular, what it means for liquidity

Is the way we are now trading good or bad?

When we teach our students in a first-year class about security prices,

we assume that prices follow random walks, which means that they are unpredictable in the short term Today’s price for a security is equal to yesterday’s price plus some innovation, some information that has been impounded into the price In this context, we do not worry about the friction that we find in actual trading processes, nor about how securities trade in the real world But we should worry about this When you consider the friction, you realize that you do not observe equilibrium, or efficient prices We see, instead, the prices we trade at, the transaction prices The mid-quotes, or observed prices, are related to the unobserved equilibrium prices, plus a deviation from that equilibrium price Sometimes we trade away from what

a market-clearing or efficient price would be

What is the reason? This has been discussed in microstructure literature for the past three decades One early suggestion came from the reality of our markets – that buyers and sellers do not arrive at the same time You might have buyers in the morning and sellers in the afternoon, or the next day They may all be willing to trade at the equilibrium price, but they just arrive

at different times If the sellers are coming in the morning and the buyers in

10 The introduction in late 2007 of the European Union’s Markets in Financial Instruments Directive (MiFID), which more closely linked the various market centers in Europe, further promoted the widespread usage of algorithmic trading across the continent According to Edhec Risk Advisory, 78 percent of European buy-side firms employed algorithms as early as October 2006 The ability to disseminate orders at rapid speed is one of the advantages of algorithmic, or rules-based computerized trading systems, said Keith Bear, co-author of a report on algorithmic trading for IBM Global Business Services, who noted that MiFID would actually make the need for fast and efficient trading systems more urgent

the afternoon, somebody is needed to match these order imbalances An intermediary must step in and absorb the buy-selling imbalance in the market As an intermediary, I am happy to take on that pile of securities sold to me But I will not be transacting at the equilibrium price I must be compensated for keeping that pile of securities until the afternoon, or the next day, until the buyers come So my bid, or the price that I am willing to trade with you, will be much lower than the equilibrium price

Risk bearing capital is like oil for our financial markets.11 We need it to match up buyers and sellers Even within the literature of the past 30 years,

we still do not have the econometric technology to measure the size of this risk

I analyzed the Dow Jones Index from yesterday to develop this idea I took the data returns in August which are shown in Exhibit 4

Exhibit 4 Volatility – Transitory versus Permanent

I then tried to decompose their size into two components If you look at transaction cost prices, you see the evidence and the presence of permanent volatility,12 that is, the presence of prices that are permanent in nature and

11 For an academic perspective, Menkveld recommends, What Happened to the Quants in August 2007? Evidence From Factors and Transaction Data Amir Khandani, Massachusetts Institute of Technology (MIT) and Andrew W Lo, MIT Sloan School of Management; National Bureau of Economic Research (NBER) October 24, 2008

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1288988

12 Menkveld, in a post-conference interview, distinguished between permanent and transitory volatility Transitory volatility, he noted, is associated with temporary price changes For example, say prices suddenly change 1 percent and then, a moment later, return and settle back at their original, starting prices, so that the net change is, in effect, zero, that illustrates the phenomena of temporary volatility Permanent volatility is clearly the

mt = mt-1 + wt (no-friction asset price dynamics)

pt = mt + Ɛt (transaction price)

Transitory (σƐ) vs Permanent (σw) Volatility

(e.g Grossman and Miller (1987))*

not temporary, like the price movements associated with transitory volatility

I calculated it in a very simple way Still, it conveys the idea of the two components It is 1.1 percent standard deviation of permanent volatility But

to emphasize the magnitude of the distance we are trading away from the equilibrium price, it is a substantial 60 basis points

Let’s look at the September data returns (Exhibit 4) In that month we see a lot more volatility Permanent volatility is up from 1.1 percent to 1.5 percent But look what is happening to transitory volatility, the size of the epsilon here That is 1.8 percent relative to 60 basis points – and that is a total of 180 basis points One interpretation is that risk-bearing capital at this time was very low Moving onto October in Exhibit 4, permanent volatility is huge, it is almost five percent And once again, there are the elevated levels of transitory volatility The point I wish to make is that transitory volatility must be understood, for it is a very important component

of overall volatility

I have a few words to say about algorithmic trading and electronic markets How does our transacting with algorithms in electronic markets change the price dynamics? Once again, I will demonstrate the thought with

a simple equation in Exhibit 5

Exhibit 5 Volatility – Algos Change Price Dynamics

The unobserved equilibrium price process is now defined in transaction time, and in between transactions there is information There are

opposite, occurring when prices do not systematically return to their original starting prices

mt = mt-1 + wt (no-friction asset price dynamics)

wt = ut + δqt (wt decomposition)

We add the signed transaction qt, i.e 1 for buy order and -1

for sell order Algorithms change the intraday price process, in particular

liquidity supply

• they reduce δ, increase σ u, quotes react to information w/o

trade arrival

• adverse selection risk reduces, ergo liquidity supply

improves, more opportunity for risk-sharing trades

• however, they need recurrent structure*

*Hendershott, T., Jones, C and Menkveld A (2008), Does Algorithmic Trading Improve Liquidity? Journal of Finance, forthcoming

econometric techniques to try and strip out the transitory volatility We decompose the ‘wt.’ One of the parts is public information arrival – information that everybody sees – and portfolio managers updating their quotes at the same time The other part is information in the order flow When there was a market buy, perhaps there was an informed trader I want

to update my estimate of the efficient price given that it was a market buy

So we have decomposed permanent volatility into those two components, public information arrival and private information in order flow

Now what has happened with algorithmic trading in our markets? One idea is that this is liquidity being demanded by people who have better information using algorithms Algorithms are different from human traders

in their capacity to process all this mountain of information So it is probably the best-informed order flow The probability of informed order flow goes up, and so more of this price discovery is coming from the order flow

The other interpretation is that perhaps algorithms are operating on the liquidity supply side Some of these firms – proprietary trading funds, hedge funds – are shipping in limit orders They are updating based on all the information in the market before you can hit them on their limit orders, given that they have this very good information processing power If you follow that line of thought, you can actually find that more of the information is revealed through quote updates rather than order flow

I have collaborated on this research with Terrence Hendershott of Berkeley and Charles Jones of Columbia We tried to decompose that innovation, and we tried to find empirically what is happening with algorithms in our markets We looked at a couple of algorithms along the way Apparently, the hedge funds are acting as electronic market makers, using algorithms to shoot their inventories through the market They are providing liquidity and, in so doing, make our markets more efficient So the first component you see goes up, and the second component goes down That is what we find empirically for 800 stocks on the NYSE for a period from 2001 to 2005.13

LIUREN WU: I was going to talk about intra-day volatility as well, but Bob Schwartz asked me to talk about something related to the current economic crisis This crisis affects my thinking more about long-term issues than short-term issues Why do we care about volatility? I will not go as far

as Asani Sarkar who talked about consumption volatility, but still it is a longer-term kind of issue

A central theme for finance research is to understand the trade-off between the risk you take and the compensation you get in return For every

13 This talk is based on the following manuscript: Terrence Hendershott and Albert J Menkveld, ‘Price Pressures,’ working paper VU University Amsterdam

actual unit of risk we want an actual unit of return We call that unit the risk premium But, unfortunately, finance academia has failed miserably so far

in explaining any kind of risk premium When the theory cannot explain the evidence, we call the evidence a puzzle I once said that where there is a risk there is a risk premium puzzle I will review some of the puzzles and discuss what the economic crisis tells us about them

The inherent difficulties are highlighted by the fact that we cannot explain any of the risk premium puzzles One difficulty is how to measure the risk As Bob said earlier, there is a risk measure from a model, and there

is also uncertainly about the model Both the risk and the uncertainty should require compensation, but neither is easy to estimate Even if we know how

to measure risk and can do it accurately, there are different types of risks

We must consider compensation differently for each of them For example, intra-day volatility – the daily fluctuations that Robert Almgren just spoke about in the futures environment – are totally different animals from the losses and risks that are large and rare, the types of risks that occur with super-low frequency Once they occur, they have super large impacts When

we consider compensations for these different kinds of risks, we should treat them differently

Let’s go back to risk premium puzzles The first one in the traditional finance literature is the equity risk premium puzzle This says that, from our view, buying stocks gives us too high an average return compared to buying treasuries, even after we adjust for the risk difference The question is whether we have adjusted enough for risk For example, if we take the S&P

500 index, on average we see a risk premium of about four to six percent per year It is just the mean number, that mean is very hard to estimate, and it changes a lot Historically, volatility is about 10 to 20 percent If we use the Sharpe ratio to gauge the profitability per unit risk, the Sharpe ratio is around 0.4

I revisited the calculation yesterday I downloaded about 20 years of S&P returns and calculated realized variances using a monthly horizon The low number I got is about five percent The high numbers before this year were 40 to 50 percent During the past month the number reached about 80 percent The highest number I got is 83 percent The question is: If a four percent risk premium is too high for a 10 percent risk, is it still too high for

an 83 percent risk?

The other problem is that the high volatilities are often realized when the index is going downhill For example, because of the current crisis, returns over the past four or five years are completely wiped out What this tells us

is we have a risk measure in volatility But we do not really know what it is and it does change greatly over time The high number was 50 percent before this year, but now we get 80 percent So, not only can the risk change, but we do not know where it will go

In the over-the-counter derivative market, we have a contract called a variance swap contract A variance swap is a forward contract If you sign

it today, at maturity you will receive the difference between a fixed number, which we call a variance swap rate, and the realized variance Therefore, by signing this contract on, say, the S&P 500 index, you can eliminate your uncertainty in the underlying variance of the index return Suppose your whole purpose is to receive the equity risk premium without being worried about how the index variance varies over time You can long the equity index together with the index variance swap to remove the uncertainty on the variance

But the presence of this contract has caused another problem for academia The variance swap contracts are quoted at such high variance swap rates that shorting variance through the contracts generates a huge positive premium If you think that the equity risk premium is high, selling variance generates a much higher premium I noted earlier that the Sharpe ratio for the long equity index is about 0.4 But the Sharpe ratio for shorting

a variance swap contract on the index can go as high as one to three, depending on how you measure it We can always argue about whether the Sharpe ratio is the right measure, but if you are using this measure, the variance risk premium is indeed very high

The variance swap contract is popular in the industry One reason is that people can hedge it reasonably easily You can replicate closely the contract using a portfolio of vanilla options across different strikes at the same maturity The weighting is proportional to one over the strike squared The CBOE has the VIX index On September 2003, the CBOE revised the VIX

to make it approximate the 30-day variance swap I have published a paper

in the Journal of Derivatives titled, ‘A Tale of Two Indices.’ It describes the

differences and the economic meanings of the CBOE’s old and the revised VIX index.14

newly-I did another experiment with the Vnewly-IX index newly-I downloaded the Vnewly-IX data and compared them with ex-post, one-month realized variance From

1990 to 2007, if I sell 100 million notional of the variance swap, on average

I can make 1.4 million The maximum profit I get is 16 million for 100 million notional The maximum loss is 12.7 million If I calculate the Sharpe ratio, the number is very high I should decide to short the contract given the high Sharpe ratio The question is how much I should sell?

Assume that I own 100 million, and I want to make sure that I do not lose my shirt I want to have at least 10 million left even in the worst case Historically, from 1990 to 2007, the worst case is a 12.7 million loss for each 100 million investment So I can lever it up to seven Seven times 12.7

14 A Tale of Two Indices, Peter Carr, and Liuren Wu, Journal of Derivatives, 2006, 13(3),

13-29 http://faculty.baruch.cuny.edu/lwu/papers/vix.pdf

gives about a 90 million loss, so I will still have about 10 million left Over the past month or so these investments are mainly losses, and the largest loss reached 58 million per 100 million investments Well, I levered it seven times Seven times 58 is approximately 400 million, so I would have lost all

my money Not only that, I probably would have lost your money as well The experiment shows you that with this kind of risk, normally you make positive returns; you always get a premium, and then suddenly you lose a lot

The variance swap contract is essentially a portfolio of options The key premium is not coming from everywhere; it is just coming from the low strike end, the put options Before this year, the major premium is from selling puts to insure against market crashes There is this well documented volatility skew on the equity index options market It says that the cost of buying an out-of-the-money put, which is insurance against a market crash,

is much more expensive than buying an out-of-the-money call, which is like

a lottery ticket on the index Accordingly, selling the insurance against a market crash makes lots of money

But if you think that selling this insurance against a market crash makes

a lot of money and you want to do that, you should also take a look at the corporate credit market If you think that selling put options generates a high-risk premium, selling credit insurance historically generates an even higher premium There are several studies that compare the put premium with the credit premium The credit risk premium is larger In the industry, a simple, direct way to expose yourself to credit risk to gain the credit risk premium, is through short positions on the credit swap contracts (CDS) The buy-side of the CDS contracts pays a pre-specified quarterly premium until a default event, or the expiry of the contract, whichever is earlier If default happens before the expiry, the buy-side stops paying the premium It can receive par value from the sell-side on the corporate bond

of the company under default If this bond is worth 40 cents on the dollar, the buy-side essentially receives a 60-cent compensation To be on the short side of this contract is historically profitable, more profitable than selling puts

If you put all these different risks and risk premiums together, you find a ranking between them Long equity index is profitable, but it is not as profitable as short variance Short variance is profitable, but not as profitable

as selling far out-of-money puts Selling put options is not as profitable as selling credit insurance But you know the rest of the story already The whole market, whether you call it Wall Street or Main Street, is being dragged down, in part, by these credit contracts

When we review these different kinds of risk, we find that the standard deviation or the Sharpe ratio is a highly inappropriate risk-return tradeoff measure, especially for risks that are rare in frequency but huge in impact

In these cases, if you do not lever, the absolute premium is still small We

are talking about basis points, even if the Sharpe ratio is very high But if

you lever it up, once that big rare event happens, you will be wiped out and,

as in the current situation, the whole country will be dragged down with you

From an academic perspective, this experience tells us that when we talk

about risk premiums, about whether there is a puzzle or not, we should

identify exactly the kind of risk that we are talking about

SARKER: Now we are open for questions from the audience

UNIDENTIFIED SPEAKER [From the Floor]: My first question is for

Albert Menkveld What is the relationship between permanent volatility and

transitory volatility? My second question is for Liuren Wu People are

talking about generating negative skewness in the portfolio That can have a

higher Sharpe ratio when they report to investors What is the implication,

and how do we adjust the risk to measure the performance of the hedge

funds or the mutual funds?

MENKVELD: The academic answer is that transitory and permanent

volatility relate in a positive way If I push up permanent volatility I just

make the cost of carrying inventory higher Think about what these

intermediaries do, matching buyers and sellers who arrive at different times

The intermediaries hold suboptimal inventories They accept risk that others

can diversify away from They are compensated for taking on this risk That

price risk is increased if we increase permanent volatility So they will

demand larger returns, and the epsilons will grow as well In that sense, the

permanent volatility has pushed up the transitory volatility That is the

academic answer

There is an institutional perspective here We had a talk yesterday at

NYU by Andy Lo at MIT that explains it He presented an interesting paper

describing a scenario where investors, leveraged money managers, for

example, must satisfy a margin call on their positions Consequently, they

must use more of their capital, which means there is less capital available in

the market to provide, in effect, liquidity or market making.15 When money

and capital are withdrawn from the market in this fashion, there is then less

risk-bearing capacity to match up buyers and sellers at different points in

time That increases transitory volatility

WU: Regarding your second question about the negative skewness, if

the Sharpe ratio is the right measure, the most profitable strategy would be to

sell insurance on something that is very rare That is because the Sharpe ratio

15 See, What Happened to the Quants in August 2007? Evidence From Factors and

Transaction Data Amir Khandani, Massachusetts Institute of Technology (MIT) and

Andrew W Lo, MIT Sloan School of Management; National Bureau of Economic

Research (NBER) October 24, 2008 ( http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1288988 ).

would be infinity before that thing happens, but that thing happens very rarely So you receive a premium every month and there is no standard deviation until that thing happens Then you are wiped out Economically,

we do have that adjustment in the very fundamental utility function definition, where most of the utilities we specify satisfy what is called an Inada16 condition That says that your consumption cannot be zero At consumption zero, the marginal utility of getting a little bit more is infinite

If you use this kind of utility function to guide your investment decision, you will never lever up to the point where you can lose everything You will always have something left

But with insurance contracts, to make sure that you do not have a chance

of losing everything, you will have to have a very high reserve, which prevents you from levering up Yet, without leverage, the absolute dollar amount from the insurance premium is actually very low An alternative is to look at the historical minimum and regard that as the maximum loss you can incur, and to lever up based on this maximum loss estimate

ALMGREN: One comment in terms of the negative skew enhancing Sharpe ratios

Here is a fairly straightforward statistical fact If you are measuring mean and variance over finite time intervals, you can improve your mean, variance trade-off by capping your winners and letting your losers run If somebody

is measuring you every week on your mean return and your variance of returns, you can play this game, and it skews the distribution down It boils down to the fact that the Sharpe ratio is based on the mean and variance Alexander Schied and his collaborators show that if you switch to a utility function this behavior goes away This mathematical fact has implications for investment performance evaluation If you measure variance and volatility across intervals that are too long – say, weekly instead of daily – then optimal strategies can have undesirable properties.17

UNIDENTIFIED SPEAKER [From the Floor]: When the market makes the kind of moves it has been making, there is a lot of forced selling You have hedge funds that have to liquidate positions CEOs are meeting margin

16 The Inada condition in the utility function, notes Wu, shows that the marginal utility at zero consumption (the first derivative of the utility function at zero) is infinite, since a person would presumably do anything to get ‘some’ minimum consumption, while marginal utility at infinite consumption is zero In other words, if a consumer had an infinite amount

of food already, this consumer doesn’t presumably care about obtaining more food In dollar terms, the utility of one extra dollar for billionaires is essentially zero But for a person who is impoverished and hungry, this situation can be the cause to ‘kill for,’ Wu descriptively explains

17 See, Optimal Basket Liquidation for CARA Investors is Deterministic ( http://papers.ssrn.com/sol3/papers.cfm?abstract_id=150027 7) See also, Publications by Alexander Schied ( http://www.alexschied.de/publications.html )

calls in their own company stock And in terms of the variance, I know that insurance companies have to buy the three-month variance when risk and volatility are at a certain level What is your opinion on that, and how do you think that affects volatility, maybe in excess of the level where it is supposed to be?

MENKVELD: That is why I find it valuable to think about the market’s capacity to absorb temporal imbalances in order flow If there is no money because people have to meet all kinds of capital requirements, transitory volatility is driven up Here is an example: In one week in August, 2007 a lot of hedge funds lost a lot of money because of margin calls We saw that liquidity supply dried up quickly because there was less capital available to absorb the temporal order flow imbalances You might call it excessive volatility I say it was transitory volatility that was pushed up

STEPHEN SAX, (FBN Securities) [From the Floor]: My question is for Albert Menkveld on the study of efficient markets using algos You discovered that the market was much more efficient with hedge funds using algos There are two sides to this: the players who use the algos such as hedge funds, and many of the broker-dealers; and then there are the players who don’t use algos and are being disadvantaged, players such as some investors and broker-dealers Who wins? Does anybody come out a loser, or does everybody win? In markets, historically, for every winner there is an offsetting loser The bottom line is that if these algorithmic tools did not work and make net money no one would use them

MENKVELD: I did not have time to develop that thought We have a small section on this in the paper There is an interesting route that leads you

to a situation where everybody is winning, that is the liquidity route What happens, we find, is that you as a supplier of liquidity (of limit orders that are shot into the market by algos that are improving liquidity supply) are one component of liquidity You need to be compensated for the losses you have

by trading against informed investors It is labeled adverse selection risk in this literature If that is lower because you are quicker to process the information when supplying liquidity, then bid ask spreads can tighten That

is what we find at the NYSE when the algos entered the market in 2004.18

18 By summer 2004, the New York Stock Exchange was preparing for its ‘hybrid’ market that would, as the exchange later noted, make ‘speed and execution certainty available to a wider variety of orders’ as well as ‘provide an opportunity for price improvement [in an auction market environment] for those who desire it.’ See, ‘The 'Hybrid' Approach: A Review of The NYSE's Market Structure Proposal.’ (New York Stock Exchange Inc.), published December 09, 2004 The hybrid market, which was eventually implemented, was outlined at a time when the SEC was prodding exchanges to operate ‘fast’ markets in its proposed Regulation NMS, and as electronic and algorithmic trading had changed the face of the industry

There was a friction at the NYSE that was reduced; the electronic door was opened up, which is what we key in on in the study If that is the case,

if transaction costs are down, then there is more opportunity for everybody

in the economy to share risks by trading because of the simple fact that it is cheaper to trade

DANIEL NACHTMAN (Bank of America) [From the Floor]: Lately

we are seeing volatility at the close becoming extreme, and, at the same time, algos are being used very heavily at the close What is the empirical evidence on the algos driving up volatility? I would almost assume that is happening

MENKVELD: We do not find too much of what I label permanent volatility So, it is not as if volatility is up or down, but we see how volatility is impounded into prices And that is now happening in the quoting process, rather than in the fact that trades arrive We just need to wait for the trade to find out what the value of the security is If there are a lot of market buys with the probability of some informed trading, it probably

is a good sign

These algorithms exploit a history of similar market conditions In order

to run the algorithm, you need to understand the correlations between the prices in the different securities In some sense you need to have a history of similar market conditions to measure those correlations and use them If that history is not there, you cannot run the algorithm successfully Perhaps they are switched off If so, we rely on humans again to put the prices in the market, and then things might actually be worse We do not say anything in the paper about this It is beyond the scope of the paper

ALMGREN: I do not have specific data because I have not been in the markets for a couple of months But I have pretty good speculation on what

is happening The close is a special time Everybody wants to get the close price It is a sort of mark to market event, and they think about it all night The better your technology is, the closer you can get to target the close We used to talk to options traders who would try to hedge over the last hour with 15-minute VWAP orders We built an algo that tries to hedge into the close, literally in the last couple of minutes, to nail the correct delta for the option

It is an inevitable fact that the better your technology, the more that late-day volatility will get pushed into the last 10 minutes, into the last five minutes, into the last minute This may be good, or this may be bad Regardless, I do not think that any of us can stop it

ALEC SCHMIDT (ICAP Electronic Broking) [From the Floor]: I have

a question for Robert Almgren If I understand your presentation correctly, you calculate intra-day volatility using tick data

ALMGREN: Absolutely

SCHMIDT [From the Floor]: But then you would have to incorporate the transitory volatility that Albert Menkveld was telling us about So the question is, are you doing that?

ALMGREN: I did not completely understand Albert’s distinction between permanent and transitory volatility I know that when you are trading a stock over one day, there is sort of a moot distinction What we are trying to do is anticipate a price motion over the next couple of minutes so that we can do things like calibrate how far away to place a limit order You are just trying to set a scale, and you want to know if it is moving a lot, or a little, over the last couple of minutes relative to the day It is an approximate measure But you need some measure of how much it is likely to move in the next couple of minutes relative to different things You need to do that using tick data It is not an after-the-fact historical study, it is a ‘how should we steer our algo right now’ issue

UNIDENTIFIED SPEAKER [From the Floor]: Mr Almgren, the 1987 Crash was blamed in part on the use of program stop losses19– the precursor perhaps to algo trading Once Capitol Hill is finished with the investment bankers, do you expect that they will take a regulatory bite out of the algo industry?

ALMGREN: No, I think that algos are entirely good for everybody (laughter) And luckily we have empirical evidence to back that up Bad algos can be destabilizing in the markets, but what is an algorithm? An algorithm is your own trading strategy, which you have taken the trouble to specify precisely Instead of watching the screen and saying, ‘Oh now I will

do this or that,’ you have written a program If you do it wrong there can be bad feedback effects But that was also true 20 years ago Today, the algorithms are a lot better

As to your comment about the liquidity providers and suppliers not meeting at the same time in the market, algos can help to smooth that out Some of the liquidity is resting inside the systems of the broker-dealers They are not taking the trade, but they are holding it and they are waiting for counter-parties to appear The client puts the order in in the morning telling

19 Program trading was blamed by many early commentators for the crash This assisted trading included index arbitrage and portfolio insurance The former aims to make profits on discrepancies between markets, by simultaneously buying in one and shorting a position of the same size in a similar type of market Portfolio insurance involves the sale

computer-of stock index futures to safeguard against the value computer-of a stock portfolio declining

the trader to deal with it over the day, to do whatever he or she wants during the course of the day Then the trader has the option to hold that order and wait for the counterparties

SARKER: The clock is ticking and we have to end with that one Thank you everyone for a fine session

CHAPTER 2: OPENING ADDRESS: RETO

FRANCIONI

Reto Francioni, CEO, Deutsche Börse AG

We have just witnessed a most enlightening discussion on intra-day volatility It shows that the topic of this year’s Financial Markets Conference has been aptly chosen Volatility is a topic that very closely concerns not only traders, but also us as operators of regulated markets We have all been given a reminder of this by some truly dramatic trading days in the course of this year In the second half of January, and again this month, the international equity markets experienced extremely large turnovers and considerable index movements, pushing the workload of our trading systems

to their limits

My opening address is organized around three major theses Firstly, in recent years, exchange trading has undergone some major structural changes: trading has become more international, more competitive, and much faster None of these macro trends, however, has a one-way relation to volatility Each of them has the potential to either reduce or enhance volatility Secondly, empirical evidence on Xetra, Deutsche Börse’s electronic cash market, seems to support the view that there is no definite long-term trend in the development of volatility At the same time, however,

it makes clear that volatility comes in seasonal peaks, and is a feature of markets we have to cope with in the long run: We have to live and therefore

to deal with volatility We are currently witnessing such a peak of a particularly high magnitude – and we have seen similar ones before in the past decade And thirdly, exchanges’ market design can play an important role in dealing with volatility Volatility interruptions may in this context turn out to be a relevant alternative to fully-fledged circuit breakers, since they do not interrupt the process of price discovery This underlines the significance of the efforts undertaken by exchanges to improve market design and facilitate safe and orderly trading

19

R.A Schwartz et al (eds.), Volatility: Risk and Uncertainty in Financial Markets, Zicklin

© Springer Science+Business Media, LLC 2011

School of Business Financial Markets Series, DOI 10.1007/978-1-4419-1474-3_2,

Regarding thesis number one: Volatility in January and again in recent weeks was, of course, exceptional due to external shocks However, there are also a number of structural changes in the securities industry that may in the long run have an effect on market volatility The trends I am referring to are, firstly, an increase in cross-border trading, secondly, an increase in competition between exchanges and alternative trading platforms, and, thirdly, an increase in algorithmic trading

Firstly, cross-border securities trading has grown massively In recent years, the major market participants worldwide have increasingly taken a global, cross-asset class perspective On the exchange side, this development has been mirrored by two trends:

ƒ On the one hand, regulatory change, especially in the EU, has

further harmonized market regulation, and has thus eased border market access On a global scale, the dialogue on mutual recognition, especially between Europe and the US, will eventually make cross-border trading easier as well Once issues of supervisory cooperation and investor protection have been sorted out, we will see trans-Atlantic and trans-Pacific trading taking place – not only via subsidiaries, but through direct remote access

cross-ƒ On the other hand, a new wave of cross-border mergers of exchange operators has led to the emergence of a new group of global players competing on an international scale Operators based in the US have entered the Western European market through major mergers On the derivatives market, Deutsche Börse and Swiss Exchange

subsidiary Eurex is now unified with the US options exchange ISE here in New York, led by David Krell And the wave of acquisitions and cooperations is spilling over to markets in Eastern Europe, the Middle East, and Asia With the further development of markets in these regions, especially in Asia, I am sure we will see cross-border acquisitions emerging from major players there as well

The emergence of cross-border trading, if it creates larger liquidity pools, may have a dampening effect on volatility However, we should keep

in mind that so far this kind of cross-border consolidation mainly takes place

on the level of exchange operators Stock markets themselves are still largely national affairs, licensed and supervised by national authorities In addition, the increasing interconnectedness of market operators may increase the speed by which trends in one market spread to others by further removing friction for exchange users pursuing their global trading strategies As a consequence, volatility on the markets as a whole may increase

A second trend worth pointing out here is an increase in competition for exchanges, mainly due to the emergence of new alternative trading

platforms In the US, this trend goes back to the 1980s, when the first electronic communication networks and alternative trading systems were established In Europe, where fully electronic exchange trading was introduced earlier than in the US and by the incumbent exchanges, such alternative platforms used to have little success and were usually dismantled only a few months after they had been established However, recently they seem to be gaining wider acceptance Instrumental for their emergence was the coming into force of the EU Markets in Financial Instruments Directive

or MiFID in November last year It has introduced the legal form of called multilateral trading platforms as a more lightly regulated alternative to exchanges

so-At Deutsche Börse, we are monitoring this new competition closely The largest MTF, Chi-X, so far has a market share in DAX 30 stocks of around ten percent Volumes are, however, very volatile, and market depth is low

We also observe a strong positive correlation between trading volumes on Xetra and Chi-X We believe that MTFs attract liquidity not from exchanges, but from OTC trading and previously internalized business, by targeting specific customer groups Further to that, their activity results in increased arbitrage activities between platforms, including alternative platforms and incumbent exchanges, and therefore creates new trading volume for the whole market

The effects such new markets may have on market stability are difficult

to judge On the one hand, they may contribute to a fragmentation of liquidity and thus increase volatility in single stocks On the other hand, they may lead to an increase in arbitrage activities, and attract order flow that had

so far been confined to the formerly opaque OTC market In any case, I think the current situation underlines the role of exchanges as anchors of stability, providing access to liquidity, full transparency, stable trading systems, and clearing houses with central counterparties as functioning risk-management systems In this context, I feel tempted to add that Deutsche Börse subsidiary Eurex Clearing is the only clearinghouse worldwide that is able to perform event-driven real-time risk management intraday

Thirdly, algorithmic trading: At Deutsche Börse, the percentage of algo trading in overall trading volume has persistently increased in recent years, and has reached levels of above 40 percent now Similar developments have taken place in exchange trading elsewhere as well As a consequence, trading has become much faster, and the average size of orders has decreased Arbitrage activities have been taken to a new level For exchanges, the additional liquidity provided by algo trades is of course a welcome development However, it also presents us with a major technological challenge: The requirements regarding system latency have massively increased

This is the main reason why exchanges worldwide have been updating the performance of their electronic trading systems On Xetra, for instance,

we have shortened the average order round trip time to almost one tenth of the levels of November 2006 The new dimension of speed has also made physical proximity an issue again Algo traders at Deutsche Börse increasingly opt for using our co-location service in Frankfurt, close to the backend server of the trading system, in order to further decrease round trip times to an average of 7 milliseconds The minimum we can reach has even come down to some 2 milliseconds

I think it makes an interesting research topic to judge whether algorithmic trading increases or decreases volatility On the one hand, it may amplify the swings of business cycles if it is behaving pro-cyclically On the other hand, it adds to liquidity and should therefore have a dampening effect

on volatility in single stocks In addition, it monitors the market for signs of mis-pricing and by exploiting them for its trading re-aligns prices This should also decrease volatility in single stocks

Looking at empirical evidence, for example, during the very active and volatile trading days in January, we analyzed how algorithmic trading developed, and we observed that there was a strong correlation: selling activities by algorithmic traders coincided with a DAX decline and buying activities by algorithmic traders coincided with a DAX recovery I am only talking of a strong correlation here, not of cause and effect – we have not studied the direction of the causal relation yet In any case, the algos seem to have got their timing right, and come out of these days of market turmoil not with losses, but with gains

As I said, I do not see any clear-cut and unambiguous causalities between the structural changes just outlined, and the direction volatility may take I am confident, however, that the discussions we have been and will be having today will provide us with new insights concerning the causal links between such macro trends in global securities trading, and the volatility of markets In any case, we will need to take a close look at the empirical evidence available in order to arrive at meaningful results This brings me to

my second thesis: Empirical evidence on Xetra, Deutsche Börse’s electronic cash market, seems to support the view that there is no definite long-term trend in the development of volatility Volatility is a feature of markets we have to cope with in the long run

At Deutsche Börse, we have had more than ten years of experience with electronic trading on the cash market side This month in fact marks the tenth anniversary of Xetra as a fully fledged electronic trading system because it reached complete functionality with Release 3 in October 1998 I would like to take this opportunity to share some of this experience with you My observations are of course only preliminary and intuitive But maybe one of you feels encouraged to analyze these data in greater depth

As you all know, volatility can be measured by two methods: firstly,

by some statistical indicator for fluctuations, such as the standard deviation; and secondly, by calculating the implied volatility inherent in option prices

The former is a historical indicator, and measures actual price movements, the latter is forward-looking, and reflects the expectations of market participants At Deutsche Börse, we compute the Volatility-DAX or VDAX, which presents implied volatility in an index form, and for which we have a time series available that goes back to 1992, as shown in Exhibit 6

Exhibit 6 Volatility in DAX stocks traded at Deutsche Börse: 1992-2008

Looking at the VDAX in this time period, I would like to make three observations:

Firstly, a comparison of the former with the latter half of this time period suggests that the overall level of volatility seems to have increased slightly Average volatility in the first half of this period stood at 20.5, and in the second half it reached 24.3 However, this is mainly due to an increase in short periods of exceptional volatility – it seems that there is no steady upward trend At the same time, volatility has not declined to the levels we had experienced during the early 1990s, when, between 1992 and 1996, its average value had been 15.5 In other words: volatility is a feature of markets we have to live with – and we have to deal with

Secondly, volatility comes in peaks and is induced by external shocks However, in the short run, it does not seem to persist After VDAX had reached the level of 63 in October 2002, it came down again to levels that were markedly lower than in the five years before: All through 2005 and

2006, VDAX remained in the range between 10 and 20, whereas the period between 1999 and 2001 had been characterized by levels between 20 and 30 And thirdly, current levels of volatility, which at Deutsche Börse reached their all-time high of 64 on 10 October, are not unprecedented in terms of their general magnitude We have seen levels of around or above 60 already twice during the past ten years By saying this I do not mean to play down the severity of the current financial crisis After all, the events that had led to such peaks in volatility were all very drastic external shocks to the world

VDAX NEW: Performance Jan 1992 - Oct 2008

Jan 96

Jan 97

Jan 98

Jan 99

Jan 00

Jan 01

Jan 02

Jan 03

Jan 04

Jan 05

Jan 06

Jan 07

Jan 08

08 10 1998

VDAX NEW:57.84

07 10 2002 VDAX NEW:62.63

10 10 2008 VDAX NEW:64.19