Posts categorized "Statistics"

RiskMinds - VaR as simple as chartism?

Interesting panel debate at RiskMinds Wednesday morning, entitled "Sophisticated Complex Models vs. Crude Robust Risk Measures".

Riccardo Rebonato of RBS started off the debate in (untypically?) controversial style by saying that he thinks that the risk management models (mostly VaR) used in financial markets are peculiar. Peculiar in that coming from a physics background he is used to models that have "causal" links between inputs and outputs, whereas VaR is based simply on the P&L distribution of a portfolio i.e. all the information is contained in the data itself. Riccardo said the obvious analogy was with chartism, where decisions are made on the observed market data itself without any reference to external (exogenous) factors at all (perhaps he should have a discussion on endogenous risk with Jean-Phillippe Bouchard at Quant Invest). Riccardo suggested that in the range of models from those that are "over specified" with two many inputs to those in "reduced form", then VaR was far too much at the reduced form end.

In response to Riccardo's proposal that risk models should involve more causal ("factor") effects, Andreas Gottschling of Deutshe Bank countered with the quote from Harry S. Truman "Give me a one-handed economist! All my economists say, On the one hand on the other.". To which Riccardo acknowledged that maybe Economists and Econometrics were less suited to trading/analyst reports (e.g. give me a single view of what the prospects/returns will be) and more suited to risk management (e.g. give me a range of scenarios with supporting assumptions for each).

Chris Finger of RiskMetrics moved on to put forward an argument for standardisation of risk reporting, saying that it was impossible to say what methodology was behind the VaR numbers disclosed by major financial institutions. He proposed that risk reporting needed to be standardised and obligatory, but emphasised that risk management should not standardised. Paul Shotton of UBS agreed, saying that whilst micro-prudential risk of Pillar I had decreased risk on an individual institution level, it had increased systematic (macro) level risk and this was an area of failure for the regulators. On this the panel agreed, echoing a lot of what Avinash Persaud said in proposing the more diversity of risk management was highly desirable.

On standardisation, Riccardo noted that many banks had switched from using 10-day to adjusting up a 1-day VaR, and as a result presenting a less risky picture to analysts and regulators, regardless of how risky the "tail" of each institutions' P&L distribution is. Riccardo also proposed that there should be "constructive ambiguity" over what is asked of the banks by the regulators - put another way he suggested the regulators should come up with the "curriculum" for risk but not the "questions", as definitive questions encourage arbitrage.

Andreas then brought the debate back to its title, and put forward that maybe VaR should be replaced by simpler measures such as limits on notional traded. Paul suggested that VaR was only good for simpler products and portfolios, under "normal" market conditions. He said that he had been an advocate of more stress testing for a long time as a complimentary approach to VaR, but also combined with the simpler approach of limits.

It was an interesting debate, particularly with Riccardo's proposal on VaR being too simple a measure based on statistics, and wanting a more "causal" model to be developed. Using the example of June 2007, Riccardo said that everyone knew something big was about to happen but this was not reflected in VaR calculations since they are statistically based and inherently backwards-looking and not predictive. The lack of prediction is a very valid point, but putting forward a counter-view, then I get the argument about economists giving a range of outcomes, but surely these should be fed into the scenario engine rather than trying to develop econometric models of relationships between market variables. Econometric models are just as vunerable as any other to the mis-behaviour of markets (anyone seen a stable correlation lately?).

A few of the other risk managers there expressed other views, from the more buy-side folks who were more comfortable with factor-based modelling, to risk managers who said that VaR was already "structural" with explicit relationships between valuations and interest rate inputs for example. It would be good to understand more of Riccardo's ideas on this, since it appeals from making risk a more "forward-looking" process but I find it difficult to quite grasp what "causal" model you can have of markets that is itself robust to changes in market behaviour.

Posted by Brian Sentance | 11 December 2009 | 4:52 pm

RiskMinds - The Failure of Risk Models

Avinash Persaud of Intelligence Capital gave the opening talk of the morning at RiskMinds (see first of set of posts from last year here) and put forward a lot of the very good ideas that he has contributed to in the recent Warwick Commission Report. Main points that Avinash made:

• Regulators were admirably quick in working out where past regulation had gone wrong in focussing too much on micro (individual institution) rather than macro (whole market)/systematic risk.
• The regulators then came out with promising papers on counter cyclical regulation and other positive ideas.
• These new ideas do not win votes however and do not satisfy the public's desire to punish someone - Avinash called this the "Bad Apple" policy, with "bad bankers, bad products, bad jurisdictions" being the perceived guilty parties.
• All past crises have resulted in demands for three things: i) more risk management; ii) more regulation; and iii) more transparency.
• These are fine as demands but evidently do not prevent financial crises.
• Avinash recalled his work back at JPMorgan in the early 90's when the 4:15 report was produced for Sam Weill, which eventually led to VAR reporting becoming widespread.
• He then fast forwarded to the Asian crisis of 97 where he saw the failings of VAR (or rather its widespread use) first hand with all players using VAR which when volatility increased caused an increase in VAR causing JPM (and all) to sell causing markets to fall, increasing vol causing more selling, increasing correlation and leading to what is called the "loss spiral".
• In light of the recent crisis, Avinash said the public perception is that bankers created a load of toxic bombs (products), through them at an unsuspecting public and ran away...
• ...and in his opinion the reality is that banks created a load of toxic bombs and ran straight towards them i.e. this was a failure of risk management where bankers did not understand the risks they were buying and selling.
• He then took us back to the 1950's and the formation of modern portfolio theory with Markowitz and Danzig working at the RAND Corporation.
• At that time banks and insurers were still separate, with FX and capital controls still in place meaning that not only could the "efficient frontier" of investment portfolios be observed but it could also be acted upon.
• Now everyone has the same information everyone can observe the efficient frontier of investment opportunities but cannot exploit or act upon it, since usually everyone moves in (the "herd") and the value observed is changed by this crowded participation in the market. Here he seems to be echoing a lot of what Bob Litterman said at QuantInvest last week over the "crowded trade" and that the barriers to market knowledge and our ability to act on this knowledge have been lowered forever.
• Avinash put forward that many of the models we use today assume the statistical independence of decision making process whereas the reality is that the market is homogenous (everyone is thinking/acting the same) and hence these models are invalid in this "crowded" context.
• In light of this, the problem of risk management is not about exogenous risk (risks from outside the market, from Black Swan events to normal distributions) but more about endogenous risk i.e. peoples behaviours upon seeing opportunities cause strategic risks. (Interesting given Jean-Phillippe Bouchard at QuantInvest commenting on what makes prices move). Put another way, behaviour is the issue not the financial instruments themselves.
• Avinash proposes that risk capacity (the ability of an institution to absorb a particular type of risk) shoudl be thought through more fully, with for example insurance and pension institutions with long-term liabilities having a much greater capacity to absorb liquidity risk than banks, and banks with short term funding being a better position to manage a loan book.
• He pointed out that regulation that uses market prices to protect us against movements in market prices is doomed to failure before it starts.
• Booms occur due to some perceived "paradigm shift" technolgy leading to dramatically improved risk/return ratios - he cited things such as cars, electricity, rail, dotcom and the mantra from those involved that "This time it is different..." (see "bubble" post from last year)
• Avinash thinks the regulators are significantly to blame for the last crisis since they themselves said the latest financial innovations in credit derivatives were making us safer through sharing out risk in the system.
• He said that there is no theory for making a complex system "safe" as a whole and that the regulators did not/do not "get" this idea.
• Diversity of approach and risks in a large systems (macro financial markets) is our only current defence and regulatory "best practice" has driven conformity not diversity in the market, making systemic risks higher not lower.
• So the regulators are themselves creating a homogenised market.
• In terms of solutions, he proposes that risk and audit committees need separating so that risk management does not become a "tick box" exercise.
• He further proposes that the risk management function is given some capital so that it can place hedges at a macro level for institution (i.e. looking at the resulting risk when divisional risks have been aggregated) - here is proposing moving to risk "management" as opposed to the much more common risk "reporting" found in many institutions.
• One risk management indicator idea he proposed was to put a portfolio management model together that was linked to VAR in order to see where the "herd" is moving to (e.g. low vol, high return Asian markets of the past etc) and to move or hedge against this.
• He is concerned that applying Basel II regulation to the Insurance industry with Solvency II will mean that all players will be dancing to same VAR tune which will introduce more risk as more institutions are forced to react in the same way to market movements and volatility.
• On the same lines, Credit Rating Agency regulation will create barriers to changes in ratings methodology in response to endogenous market risk, again meaning that everyone will be forced to behave and act in the same ways.
• He summarised that "endogenous risk" (movements in the market caused by the market) and not statistical distributions that are the key issue and diversity is the only solution.

Entertaining speaker with some interesting ideas that fly in the face of much of what is being done by the regulators today, and generally well received by many of the risk managers present. Behavioural finance and the "crowded trade" (i.e. everyone doing the same thing in the market causing movements within the market) seem to be key themes occuring in a lot of what academics and practitioners have said on risk management recently. Now what to do about it? Not sure that less (not more) regulation will find many fans at the moment...answers on a postcard please!

Posted by Brian Sentance | 8 December 2009 | 10:04 pm

Maths to Money - Quantitative Investment

I attended the Quant Invest 2009 event for the first time last week in Paris. The event is unsurprisingly about quantitative investment strategies, but with an institutional asset manager and hedge fund focus - so not so much about ultra-high frequency trading (although some present) but more about using quantitative techniques to manage medium/longer-term investment decisions and applied portfolio theory. A few highlights below that I found interesting:

• Pierre Guilleman of Swiss Life Asset Management gave an interesting 1/2 day workshop entitled "A random walk through models":
• He is a strong supporter of the need to understand more about the data and statistical assumptions upon which any quant investment model is based and how these fit with the desired investment objectives (similar to the Modeler's Manifesto)
• He made the point that good models can sometimes be almost annoyingly simple, and cited the example by a Professor Fair of Yale who had determined that US elections were predictable based on simple parameters such as past results, inflation and gdp and that policy did not seem to be a key factor at all - annoying for the politicians anyway! 
• Pierre seems very concerned that the Solvency II regulation applied to Life Institutions will negatively influence the investment policies of many institutions - applying sell-side risk measures like VAR to the insurance industry will drive a more short-term approach to investment. He strongly believes that VAR applied to his industry should have an expected return parameter introduced to fit with longer term investment horizons of 10 to 25 years.
• Bob Litterman of Goldman Sachs Asset Management opened the first "official" day of the conference:
• Bob put forward his "scientific" approach to investment modelling going through the stages of hypothesis, test and implement. He warned against overconfidence in investment (apparently 70% of us think we are "above average"...) and impulsiveness (quick impulsiveness test: "if a bat costs $1 more than the ball, and the bat and ball together cost $1.10 then how much does the bat cost?...") 
• He said that the failure of quantitative investment models in 2007 needed to be understood given the success of quant models over past decades. In particular he thought that quant investment became the "crowded trade" of 2007 with every hedge fund having a quant investment strategy. In terms of why this became a "crowded trade" Bob thinks that the barriers to entry into quant investment (particularly technology) have lowered significantly recently.  
• He noted that factor-based investment opportunities decay quicker than they used to due to increased competition - implying the need for a more dynamic and opportunistic investment approach.  
• GSAM are now looking at new markets and new investment instruments, trying to find areas of market disruption but without following what others are doing in the market.  
• He pointed out the conflict between investors wanting more transparency over what is done for them, against the need to be more proprietary about the investment models developed.  
• Next there was a talk on regulation from the French regulator that was dull, dull, dull both in terms of content and presentation style (when will regulators actually prepare well for the talks they give?)
• Panel debate was also pretty average, with the word "alpha" being used too much in my view - asset managers of a certain type seem to hide behind this word as an opaque "magic wand" to justify what they do.
• Jean-Phillippe Bouchard of Capital Fund Management did a great talk called "Why do Prices Move?". Some points from the talk:
• He started off with a reminder about the Efficient Markets Hypothesis (EMH) and how it says that crashes and market movements are caused by events outside (endogenous to) the market such as news, events etc.
• He then said this was not born out in the data, where extreme jumps in prices were only related to news only 5% of the time.
• Volatility looks like a long memory process with clustering of vol over time - similar to behaviour in complex systems
• The sign of order flow is predictable but the price movement is not, with only 1% of daily order volume accounting for price movements over 5%
• Even very liquid stocks have low immediate liquidity, meaning that price movements can play out over many hours and days as liquidity is sought to "play-out" some change in fundamental price levels.
• Joseph Masri of the Canadian Pension Plan Investment Board then did a good talk on Risk Management:
• Jo said that sell-side risk was easier to deal with in some ways since it involved fewer strategies in high volumes, and hence could be better resourced.
• Buy-side quantitative risk was harder due to its reliance onsell-side research and risk tools, the outsourcing of credit assessment to the credit rating agencies, the loss of Bear and Lehman's having caused the buy-side to have to do more risk management itself (and through third parties) rather than rely on the sell side risk management tools.
• He said that sell-side risk models are a good start for an asset manager, but need to be adapted to give both absolute and relative risk (to a benchmark fund for instance). All models are no substitute for risk governance.
• He described the cross over from risk methods: VAR, stress testing, factor-based and their applicability to market risk, credit and counterparty risk.
• Like Pierre he was not a fan of 1 or 10 day trading VAR being applied to investment managers since this risk measure was not suitable for long term investment in his view.
• On stress testing he said this needed to be top down (using historical events etc) as well as bottom up from knowing the detail of strategy/portfolio.
• In terms of challenges in risk management he said that VAR needed more stress testing to cope with the fat tails effect in markets, that liquidity risk both of counterparties and of illiquid products was vital and the importance of stress testing (he mentioned reverse stress testing) plus also the feedback (crowding effects) of having similar investment strategies to others in the market.
• Dale Gray of the IMF gave a very interesting talk on how he and Bob Merton have been applying the contingent claims model of a company (looking at equity in terms of option payoffs for shareholders and bondholders) to whole economies:
• He said that some of his work was being applied to produce a model for the pricing of the implicit guarantees offered by governments to banks
• He said these models were also applicable to macro-prudential risk
• Very interesting talk, and if he really has something of macro-level risk then this is great relative to the wooly approach by the regulators so far

There were some other good talks from Danielle Bernardi on Behavioural Finance, Martin Martens on Fixed Income Quant Investment, Vassilios Papathanakos on Stochastic Portfolio Theory (seemed to be a "holy grail" of investment model, giving good returns even in the crisis - begs the question why he is telling everyone about it?), Claudio Albanese on unified derivative pricing/calibration across all markets (again another "holy grail" worth more investigation) and Terry Lyons on speeding up monte carlo simulations.

Overall a good conference although the quality of the asset managers present seemed very digital from those who really seemed to know what they talking about to those who plainly did not (in my limited view!). Along this line of thought, I think it be good to test whether there is an inverse relationship between the quality of the asset manager and the amount of times they use the word "alpha" to explain what they are doing...

Posted by Brian Sentance | 5 December 2009 | 2:08 pm

It's in the news...

I went along to the Forum on News Analytics over in Canary Wharf on Monday evening, organised by Professor Gautam Mitra from OptiRisk / Carisma at Brunel University. We seem to be in the early days of transforming news articles into quantifiable/machine-readable data so that it can be processed automatically/systematically in trading and risk management. It was a good event with both vendors and practititioners attending so was reasonably balanced between vendor hype and the current state of market practice.

As background on what is meant by news analytics data, then for example you might count the number of news articles about a particular company and look at whether the quantity of news articles might be a predictor of some change in the company's stock price or volatility. Moving on from this simple approach (assuming that you are clever enough to be certain about what news is about what company), then you can then move towards assessing whether the news is negative, neutral or positive in sentiment about a company/stock.

The context here is about having the capability to automatically process/analyse any kind of text-based news story, not just those from research analysts that might be nicely tagged with such quantifiers of sentiment (see http://www.rixml.org/ on xml standards for analyst data). The way in which the meaning of the text is "quantified" uses some form of Natural Language Processing.

The event started with a brief talk by Dan di Bartolemeo of Northfield Information Services. I hadn't heard of him or his company before (maybe I should pay more attention!) but he seemed a very solid speaker with strong academic and practical background in investment management and modelling. He referenced a few academic papers (available via their web site) on news analytics, and how news analytics and implied volatility could provide better estimates of future volatility than implied volatility alone. He also made some good points about how investment "models" are calibrated to history and how such models need to adapt to "today" - he put it as "how are things different now from the past?" and put forward the idea of a framework for assessing and potentially modifying a model to respond to the "now" situation. He also suggested that the market can react very differently to "expected news" (having a range of investment "what ifs" planned for a known earnings announcement) as opposed to unexpected information (we are back into the realms of the Black Swan and the ultimate in uncertainty wisdom from Donald Runsfeld)

Armando Gonzalez of RavenPack then began by explaining how RavenPack had become involved in applying text analysis to finance (it seems the subject has its origins, like a lot of things, in the military). RavenPack seem to be highest profile quantified news vendor at the moment, and whilst Armando is obviously biassed towards pushing the concept that money can be made by adding quantified news data to trading models, he said that not many firms are as yet systematically processing news and most people are relying upon manual interpretation of the news they buy/use. Some of the studies Ravenpack have on market news and prices are very interesting, showing how a news event can take up to 20 mins before the market settles on a new "fair" price level for a stock. Additionally, and maybe an interesting reflection on human behaviour, was that in bull markets there are usually twice as many positive stories about companies than negative, but strikingly in a bear market there was still almost equal amounts of positive and negative news - so humans are basically optimists! (or delusional, or just plain greedy...take your pick!)

Mark Vreijling of Semlab followed Armando and suggested that a lot of their sales prospects understandably desire "proof" of the benefits of adding quantified news to trading, but this was a little ironic since most financial institutions have been paying to receive "raw" news for years, presumably because they perceive beneift from it. Mark also mentioned that the application of quantified news to risk management was a new but growing area for him and his colleagues.

Gurvinder Brar of Macquarie then went into some of the practicallities of quantifying and using news in automated trading. He suggested that you need to understand what is really "news" (containing information on something that has just happened) and what is merely an news "article" (like a "feature" in a magazine etc). Assessing relevance of news was also difficult and he added that setting a hierarchy of what kind of events are important to your trading was a key step in dealing with news data. Fundamentally he suggested that why wait for five days for analysts to publish their assessment of a market or company-specific event when you could react to the event in near real-time.

The event then went into "panel" mode where the following points came out:

• Dan thought that a real challenge was integrating quantified news with all of the other relevant datasets (market data, but also reference data etc)
• Armando picked up on Dan's point by giving the example news about Gillette which at one point was about Gillette the company but then on acquisition became news about the Gillette "brand" which became a part of Proctor and Gamble.
• Dan said that a key problem with processing news was also understanding what news was simply ignored by the news wires i.e. we know what is being talked about, but what could have been talked about, why was it ignored and is it (even so) relevant to trading?
• Mark and Armando said that the "context" for the news story was vital and that market expectations can turn many "negative" news stories into positive outcomes for trading e.g. the market likes bad news when it is not as "bad" as everyone thought.
• Dan made a very interesting point about trading in terms of categorising trades as "want to" trades and "have to" trades. He gave the example of a trade being observed that seemingly has no news associated/prompting it - so does this mean the trade is occuring because somebody "has to" make the trade (a fund facing an welcome client redemption for example?) or because there has been some information leak to a market participant and such a participant "wants to" make a trade before the news becomes available to the market as a whole.
• I think all of the panel members then collectively hesitated before answering the next question from the audience, with Microsoft having one of their "text search" R&D team (think Bing...) asking about news categorisation and quantification.
• Dan also mentioned something that I have only recently become more aware of, which is that apart from major markets in the US, most exchanges world-wide do not publish whether a trade was a "buy" or "sell" trade (they just publish the price and transaction size). Obviously knowing the direction of the trade would be useful to any trading model, and Dan referred to this as wanting to know the "signed volume".
• A member of the audience then asked whether most quantified news had been based on just the English language and the concensus was that most was based on English, but Natural Language Processing can be trained in other languages relatively easily. A few members of the panel pointed out that all languages change, even English, requiring constant retraining, and also that certain languages, countries and cultures added further complication to the recognition process.
• The next question asked was whether the panel could outline the major areas that quantified news is applied in - the answer included intraday (but not quite real-time) trading, algorithmic execution, lower frequency portofolio rebalancing and in compliance/risk/market abuse detection.
• A good debate ensued about whether "news" was provided by the official newswires or by the web itself. The panel (and audience) concensus seemed to favour the premise the news wires are the source of news and the web is a reflection/regurgitation of this news. That said, Gurvinder of Macquarie gave the nice counter example of the analysts/news wires not making much of the new Apple iPod, when looking at the web it was possible to see that the public were in contrast very enthusiastic about it.

Overall an interesting event. I think the application of "quantified news" to risk management is interesting - maths and financial theory is very interesting but markets are driven by people's behaviour and if "quantified news" can help us understand this better it has to help in avoiding (some!) of the future problems to be faced in the market.

Posted by Brian Sentance | 12 November 2009 | 12:06 am

Tibco buys Insightful...

...it must be summer (maybe not in the UK according to the weather?), seems like I missed this but Tibco has just finalised its purchase of Insightful, the makers of the S-Plus statistical package. A release from Insightful explaining the deal can be found by clicking here. Not something that strikes me as an immediate "that makes obvious sense" but not a negative either, so let's see...

Posted by Brian Sentance | 10 September 2008 | 7:02 pm