Is the common view of financial risk completely wrong?
Helo everyone. I'm a physicist looking to enter the HF space and I'm confused about how the notion of financial risk is measured and dealt with in this industry.
From what I see, finacial professionals usually associate the "riskyness" of an asset by how much its value "fluctuates" relative to the market. And furthermore, the larger this variance is, the faster the asset should appreciate relative to the rest of the market, with the ratio of the return of the asset relative to the market being given by the so-called "beta" of the CAPM. There is also the "alpha", which gives you the "expected return" on the asset. As I understand it, this allows the engineering of (allegedly ) low risk portfolios through strategies such as gathering many low alpha, low beta assets and using leverage to increase the return of the portfolio without increasing the beta (thus supposedly keeping the risk down). So, if I understand this correctly, when most people talk about the financial risk, they actually just mean volatility.
But how in the world is volatility an appropriate proxy of risk? Sure, it's convenient, but I find it incredibly hard to believe that such a measure can be of any long-term use. This measure of risk assumes that the historically-measured beta and alpha of an asset are a good predictor for the future beta and alpha, which seems absurd given how often the markets boom and crash in completely unpredictable ways. For instance, I'm sure the volatility in the market value of Lehman Brothers and Bear Sterns must've been relatively low in the years leading up to the 2007 financial crisis, yet certainly the actual risk of these two assets was enormous (but unknown, to most).
Surely, a more sensible way to think about risk must be in terms of what one knows and doesn't know about the asset in question. I.e., the real risk associated with an asset is all the unknowns. E.g. incorrect assumptions, the possibility of various unforeseen events and so on. Of course these things are difficult to quantify, but it must be better to deal with a non-quantitative measure of risk than dealing with a quantitative measure of risk that's incorrect. Right?
Or perhaps I'm misunderstanding something?
Bump & TLDR: Typically in the financial industry, risk is seen to be equivalent to volatility. Is this a fair measure of risk?
NO
may expand later but already written a lot today (not on WSO). my elevator pitch thoughts are
risk is a loss from which you cannot recover
I recommend nassim taleb and howard marks for how to think about this
the way many financial metrics are calculated are on shit math that may or may not happen in the real life and should be viewed VERY skeptically
the combination of relying on shit assumptions, hubris, and sometimes leverage has led to untold numbers of blowups and even more one trick ponies (or lottery ticket winners who continue to collect new aum...PAULSON)
What you're saying makes perfect sense, and thanks for the author reccommendations. I'm glad to hear that many in the industry have a nuanced understanding of risk that goes beyond simple financial metrics. Would it be possible for you to name the institutional investors (hedge funds, prop traders, etc.) that rely on fake math and rubbish assumptions, and those that don't?
Seth Klarman said it best:
"I find it preposterous that a single number reflecting past price fluctuations could be thought to completely describe the risk in a security. Beta views risk solely from the perspective of market prices, failing to take into consideration specific business fundamentals or economic developments.
The price level is also ignored, as if IBM selling at 50 dollars per share would not be a lower-risk investment than the same IBM at 100 dollars per share. Beta fails to allow for the influence that investors themselves can exert on the riskiness of their holdings through such efforts as proxy contests, shareholder resolutions, communications with management, or the ultimate purchase of sufficient stock to gain corporate control and with it direct access to underlying value.
Beta also assumes that the upside potential and downside risk of any investment are essentially equal, being simply a function of that investment's volatility compared with that of the market as a whole. This too is inconsistent with the world as we know it. The reality is that past security price volatility does not reliably predict future investment performance (or even future volatility) and therefore is a poor measure of risk."
I've had the same view for a long time. People look at volatility because it's easy to measure, but what they are really interested in is the maximum drawdown and chance of a permanent loss. For many assets, that tail risk also tends to be negatively correlated with the day-to-day volatility.
Yeah, you're right, it makes intuitive sense to me that the more you engineer your portfolio to eliminate volatility, the higher your risk of a blow-up. My reasoning is twofold. Firstly, portfolio engineering requires the use of leverage, and the more you leverage yourself, the more sensitive and vulnerable you become to market swings. Secondly, the more complicated the composition of your portfolio (just imagine adding derivatives, shorts, etc.), the more ways for things to go wrong in case the market conditions shift and break your assumptions. And markets are a well known chaotic, nonstationary and highly unpredictable system...
This whole financial engineering business makes no sense to me (though it might be because I'm new). Why do hedge funds bother engaging in it? To impress potential investors with how fancy their maths are and thus grow AUM?
They're making a ton of money, irregardless of whether you or I can understand it.
As for the engineering aspects, the math is obviously wildly complex but that's why they hire Masters and PHDs from Ivy League schools to do it. As for the assumptions, usually if the assumptions shift, the positions shift (look at high frequency trading as an example).
Beyond my smooth brain but interested in the discussion.
I would think a good manager would view risk holistically, i.e., volatility being one of a number of data points to gauge risk.
That makes sense. Do most of the major investors/traders think in this way?
I would hope so. Again, outside my area but I would think as an investor, you'd want to have a holistic view beyond market risk. Some of the fundamental factors come to mind around risks to the business and growth strategy, competitive moat that sort of thing.
Yes. See my response to OP below.
Bump and interested
We do basically what you say, sharpe/var are just nice high level things to track. In reality when you cut/add we do exactly what you said.
This is for ficc world not equtiies. But imagine everyone similar.
Are there any funds that do *not* do this? I'd like to know, so I can avoid applying to them after my graduation.
.
They would rather be slow and steady wins the race vs take on undue risk that leads to losses they will never make back. Why they would use a P/E a screening metric.
Basically you overthought the question so bad cultural fit.
.
That's the trick isn't it? How do you calculate the odds you will make/lose money on an investment, and to what degree?
If you can accurately measure risk in the financial markets, you can get rich beyond your wildest dreams.
cc: thebrofessor
it can be measured if your realized return is -100%
/snark
in all seriousness the likelihood you experience a total loss is probably incalculable, but there are a few ways to think about it
1. don't use leverage or unhedged shorting equities, minimizing the probability that temporary movements can give you a total loss
2. ensuring you have collateral/recourse if you're on the fixed income side
3. buying into ridiculously overpriced names (e.g. CSCO in March 2000, while it still exists it's showing a negative return 21+ years later, not a permanent loss, but close enough)
4. don't make your returns rely on assumptions spit out by models a la LTCM
5. avoiding things that sound too good to be true (like 15% investor returns on senior secured loans/whatever's being pitched to me today that I scoff at)
6. diversification minimizes the probability that one bad idea tanks the whole portfolio
7. tail risk hedging - taleb/spitznagel's strategy which is basically S&P 500 index + deeeeeeeep OTM puts for super cheap. managed properly this seems to be the easiest (although option trading isn't easy) solution, however since neither one of them offer any fucking hints as to how an individual investor can do this, I can't recommend it.
sidebar: I've been wanting to have a discussion on tail risk hedging and the applicability for the little guy, but wonder if anyone would have any interest since it's so niche. part of my beef with taleb (FD: he's my favorite author, but like everyone he's imperfect) and spitznagel is they call foul on all of these parts of the financial markets (risk measurement being one of them) and then NEVER EVER offer a solution. when pressed, spitz says he won't give it away, so it's basically like "haha individual investors, everything you know is wrong and there's nothing you can do about it!" which isn't all that helpful unless you have $100mm (their minimum, I've called)
I've seen a couple of blowups from where I sit and thankfully have avoided them for my clients by and large (or if they've suffered, it's been a forgettable % of their folio)
1. concentration in a single named stock that never recovered (ORCL, C, CSCO, GE, worldcom/MCI, and more)
2. leverage, whether it's a private credit deal that lends $6 for every $1, 200% long hedge funds trying to play the momentum game, LBOs when a recession hits, or people doing cash out refis and buying the bitcoins and NFTs of yesteryear, and so on
3. speculation, be it land, real estate, commodities, options (particularly naked puts, short calls, etc.), home flipping when you have no clue what you're doing
Expected return per unit of standard deviation (or volatility) is fine for most people’s understanding of the underlying math and financial risk. You’re correct that it’s fundamentally not real, but no model is.
I’d argue the bigger issue is relying on probabilities based on an assumed Gaussian distribution. This is empirically wrong and meaningfully undercounts tail risk, as others like Taleb have pointed out.
Apart from Gaussian returns, there is also the dependency in tail events that cannot be measured. Assets and factors that look uncorrelated in normal times don't stay that way in big crashes, like March 2020. I guess people are aware of this but think it's too high a cost to hedge against everything, and just hope something really bad doesn't happen (or hope the Fed bails everyone out).
Yes, imo right now the Fed is effectively bailing out any and all institutional investor regardless of how crappily they managed their risk. Scary stuff, as the can is just kicked down the road.
Great point, perfect example of this is the peanut industry. You’d think it would be a fairly consistent consumer staples play to hedge risk, but turns out it’s highly correlated to airline performance and TANKS in a pandemic.
It is easy to construct strategies whose returns look i.i.d. normal. For instance, suppose that every month a manager boosts the fund’s returns by some factor y where y is a lognormally distributed error with mean 1 and small variance. The boost comes at the cost of going bankrupt with probability 1/y each month. Assuming that the variance of y is small, the scheme can run for hundreds of months (25+ years) before the fund goes bankrupt, and the residuals will look very convincing. Thus, in this case the t‐test would seem to be appropriate, and the estimate of alpha will be 5%+ per year at a very high level of statistical significance. This is completely misleading, however, as in reality, the distribution of returns is not approximately normal ‐‐ there is a large potential loss hidden in the tail. Models need to correct for this “hidden volatility”... all the quant shops have models that do this though, or so I would imagine...
You send it to a quant to work on fitting an appropriate (sometimes custom) distribution when you see via graphs that normal distribution is ridiculous.
How they do it... no idea lol
The standard way of estimating alpha is to correct for correlation with the market by regressing the asset’s returns against the market returns over an extended period of time and then apply the t‐test to the intercept. The difficulty is that the residuals often fail to satisfy independence and normality; in fact, portfolio managers may have an incentive to employ strategies whose residuals depart by design from independence and normality.
http://www-stat.wharton.upenn.edu/~stine/research/markov_test.pdf
I have never understood variance as a measure of risk. Sure volatility can be bad but it also means higher return. I would measure risk by knowing expected returns given different probabilities.
Technically it's the same thing, because an assumption that basic valuation uses is that returns follow a normal distribution so having a mean return and standard deviation of returns would be "enough" to reverse engineer what you are saying.
Obviously the normal distribution assumption doesn't hold out all the time, and when it doesn't you ask the quants to figure it out lol.
yea no respectable systemic strategy runs on models that assume normal distribution lol
CAPM is extremely outdated. It "works" for instances where ballpark valuations are enough like pricing shares for an IPO.
For investing, firms use a variety of metrics to price returns. These are proprietary and will vary from firm to firm.
For an example, you can look at a publicly released example in academia - Carhart four-factor model which uses factors that are dependent on the company itself, rather than just the variability of returns to predict future performance
https://en.wikipedia.org/wiki/Carhart_four-factor_model
Hi there, to directly answer your question and not go on a long tangent about the conundrums of finance... Risk is the quantity of possible outcomes. Something high risk has many significantly different outcomes, while something with no risk, has only 1 guaranteed outcome. Think A to B as a final end point without time as a variable. The token example for this is treasuries... but they do have duration and rate risk. Volatility on the other hand, while similar, is the what happens in between point A and B, and is an attempt to measure it. Usually models do a decent job of predicting this during "normal times".. but it's the outliers where things get interesting.
Side note, most intelligent and insightful person I work with has a physics phd background, I applaud you for making the switch and wish you luck on the journey.
Indeed, it makes good sense that risk can be understood as the set of possible negative outcomes multiplied by the $-loss associated with each of these outcomes. And indeed the historic volatility can be used to provide a measure on the historic risk, and furthermore, it seems reasonable that sometimes the volatility of an asset stays roughly the same during "normal" times" and that this can be inserted into models to earn money. However, the moment these "normal times" end, I suspect things can quickly turn ugly, maybe even so ugly that the previous profits get wiped out... You see what I mean? And I think referring to these events as outliers is a bit suspect, as that implies outcomes in finance are concentrated around some mean, while in reality they are typically fat-tailed. Do the longest-living funds really rely on using these models?
Thanks a lot for your kind words!
Couple of points...
i) risk also applies to positive outcomes. Risk as generally measured today applies to both left and right tails of returns.
ii) I might be wrong, but I don't think historic vol can be used to measure historic risk, at least not accurately.. You can look at how an ice cube melted, but you cant infer what the original cube looked like only seeing the puddle. It's two different pieces of information. Someone correct me if I'm wrong, this could be a shit analogy.
iii) Yes, I get what you mean, left tails (or the lack thereof) are what make or break you. It's what previous posters have referred to as drawdown risk / value at risk / left tail / etc, what happens when shit hits the fan. Normal times do end spontaneously and it is incredibly challenging to predict when this might occur. This is the whole premise of Taleb's book, The Black Swan.
iv) As you state, returns are left skewed. That being said, returns do cluster around the mean. With this, it is fair to still call these events outliers. Looking at graphed data, they certainly are.. they just happen more frequently than models predict.. but that's why everyone is trying to build a better wheel.
v) regarding long-living funds. I haven't been in the game long enough to know. I assume though the managers in these seats are prudent to mitigate left tail risk. That being said, given there have been so many 1,000's of funds created and investors... some will get lucky, others will not... and maybe thats all we're left with today.
You're completely right. a big part of risk management doesn't really exist to manage risk, but to give the appearance that they're on top of risks.
Taleb is a bit of an asshole, but his POV on risk, volatility andrandomness is really interesting, I recommend you read his books, it will probably connect a lot more with you.
Good to hear this has been studied.. Am going to read, thanks.
I heard Taleb was a trander and then fund manager. Did he do well? Or is most of his wealth from his books?
it's late and my memory is foggy, that being said I think he went from powerlifter -> trader -> risk manager -> author
I'm pretty sure most of his wealth came from his prop trading career and only then he started writing books.
I never understood these misconceived illustrations of risk and how some investors think about volatility in general.
Mr. Burry explained it best in one of his 2001 investor letters:
"And the better managers are conceived to achieve average returns while exhibiting below-average volatility. By this logic, however, a dollar selling for 50 cents one day, 60 cents the next day, and 40 cents the next somehow becomes worth less than a dollar selling for 50 cents all three days. I would argue that the ability to buy at 40 cents presents opportunity, not risk, and that the dollar is still worth a dollar...A wildly fluctuating dollar selling for 40 or 50 or 60 cents will always remain more attractive - and far less risky than a dollar that consistently sells at 1.1X face value".
The "riskiness" of an investment should be looked at on a per asset basis comparing the "quality" and the "price" that you have to pay. Characteristics such as illiquidity or volatility for that matter, do not quantify risk per se.
It's not a fair example because in this case the dollar that changes value each day has a predetermined outcome. So you would know at 40 cents that it is a buying opportunity because it only can go up from there. This assumption cannot be translated into real life. In real life, at the moment of purchasing at 40 cents, it is possible to go to the downside given the previous drop is 10 cents. Therefore, it cannot be said that wildly fluctuating dollar remains more attractive than a dollar consistently selling at 1.1X face value. Given that, I am not convinced that it is more attractive.
Without the entire context presented here as to what Mr Burry was referring to to the shareholder letter, what you quote could be literally interpreted as bitcoin and GME/AMC is more attractive than assets with a constant face value because it is wildly fluctuating. However, it may be the case your trying to refer to the same asset with different movements.
Further, looking at the quality and price of the asset to determine the riskiness of the investment is at best arbitrary because the "quality" and "price" is subjective. The assumptions behind your conviction could be drastically different from reality or the direction of the asset that it wishes to head. This means that what seems to be a quality pick for one may not be a quality pick for another
By looking at the past prices to understand the information made known about the perceived future outlook of the company (micro and macro factors), IMO, would be the prudent choice to invest and manage money.
I think you're missing the point here. Even if the 40 cents drops to 15 cents, but the underlying value is a dollar, it would present an enormous buying opportunity. Sure, in real life you don't get the exact face value of assets but that's why you have to properly value them. If you can buy an asset for 400 that is worth a 1000 according to your own valuation, it presents a good investment opportunity (assuming that you made a proper rough ballpark valuation). Even if the price drops first to 200 before climbing up again its still a good investment. You never know the right point of entry of any investment but as long as you are doing your valuation properly, buy at a discount, even the most volatile assets can be good investments.
And yes, even though I buy no coins myself (shame on me looking at everyone around me making easy money), bitcoin could still be a good investment if the value is indeed a multiple of what it's currently trading at. All I tried to make clear is that volatility is no right measurement of risk per se. I would rather buy an enormous volatile asset for 40c on the dollar than buying a dollar for a dollar which trades constantly at face value.
Sure, quality and price is subjective. But so are market prices and past prices (which you seem a big fan of). Volatility of stocks, and metrics such as systemic risk are at best nonsense as well. Market prices can be completely out of whack too. Have a look at the alchemy of finance by Soros which was mentioned up top to have a better understanding of what reflexivity can do to prices. I agree that past prices can be helpful in valuing assets, but only as long as these past prices refer to real underlying drivers of the asset. You can buy a stock assuming the volatility is low, looking at historic vol, but you can lose a lot of money if you the position moves against or if you get squeezed and if you did not buy at a proper discount. As long as you bought the asset for a discount, eventually you will come out of top again.
But anyways, whatever floats your boat and makes you money at the end of the day right.
Since you're a physics guy, you might appreciate this paper on the subject: https://arxiv.org/abs/1003.2688. Written by two well known MIT profs (one finance, the other physics)
Thanks!
I did a bunch of shit (research, work experience, clubs, pitches) around asset management (fundy/macro) and this is a no-brainer topic that I like to discuss:
Volatility =/= Risk
I don't care if an asset class is literally a 45 degree straight line but can tank 70% at the drop of a hat, I'd rather hold an asset that looks like a fucking rollercoaster but doesn't stay down for years on end if it plummets.
Completely agree with you.
I'd say true risk is the variance of the distribution of the intrinsic value of a stock more than the volatility of the market price. The problem is that this distribution cannot be known with certainty in advance and we can only see the one realization of the outcome.
That's why risk management should be something that is done by PM when constructing their portfolio and not something that is outsourced to "risk managers" who are focused on meaningless metrics in their spreadsheet. You need to actually understand the stock and the thesis to manage risks.
Man I've worked on the quant side and the fundamental side, and I firmly believe that anyone who claims volatility =/= risk doesn't know what they're talking about.
Not that they're necessarily wrong, just that if this is something someone casually brings up in conversations they're about 2/10s of the way through Dunning Kruger.
Two levels of risk for a PM (not fund):
1. The riskiness of an individual security
2. The combined risk of a portfolio of assets
A good PM will evaluate these two separately.
1. How much can I lose on this stock/bond/etc.? Historical variance/std is helpful to see how new information and uncertainty impacts the stock. Backing this into a fundamental model is useful to evaluate the outcomes investors were pricing. Is there a risk of permanent capital loss in this asset? If I need to quickly dump this asset, will I be able to do so easily? Etc.
2. What does my portfolio risk look like? Am I over/under exposed to factors? Are the risk characteristics in-line with my thinking? What does my drawdown look like in various scenarios (using beta as a proxy for risk here)? How is my covariance? Am I properly diversified across my themes/views, or is my entire portfolio simply one big bet?
Replying because edit doesn't seem to work...
Pure alpha strategies will try to reduce ALL non-view items. E.g. If I stress-test my portfolio with SPX/any benchmark at various levels, the impact to my PNL is 0.
Relative value strategies are "factor loaders" in that they are over/underweight some type of characteristic (value, growth etc.). Therefore, SPX scenarios should also have a negligible impact on PNL. However, factor indices or benchmarks should have an impact to pnl. E.g. if you're a value investor, it should mean that your beta to value factor is >1.
Tl;dr
Thoughtful risk management incorporates quantitative measures of risk with qualitative measures -- but those don't mean anything without context (which for a professional investor is their mandate). Bashing the usefulness of this stuff (and things like CAPM, EMH, etc.) is a great signal that you don't know what you're talking about, while simply regurgitating axioms fails to demonstrate your thoughtfulness. Strive for balance!
Go through the CFA process (particularly the L3 curriculum) and you’ll get the idea
Depends on the duration of your liabilities man. For some people, the massive fluctuations are irrelevant.
If your dick tunneled through your thigh and tunneled out within a millisecond, it wouldn't matter as much as if they remained merged for a month.
wow!
Other replies from people who are clearly more qualified than I am to answer this question have shared their views, so let me answer from a slightly different perspective.
I think using volatility as a risk measure has very different practical implications at an individual stock/asset level and at a portfolio level. I wholeheartedly agree that volatility does not measure the real risk associated with an asset. Yet at a portfolio level, controlling beta offers a few practical benefits.
Average active investor has a 'decent' selection ability but has a 'terrible' asset allocation ability (many academic papers show that and those who have interacted with sell-side ER analysts know that very often the analysts can pick winners within their respective sectors but are clueless / wrong on the sector/industry perf outlook relative to other sectors/industries). What I mean by this is that some LO investment management firms will keep tracking error relatively low and will not take major sector / country bets relative to the benchmark. If they have semi-decent stock pickers who can, say, pick outperformers within an industry or sector, they can typically generate outperformance by having a low beta strategy. This can work in theory since return distribution of the stock market is left skewed and black swan events are by definition rare, and when they do happen, they are all different in flavor, i.e., cannot be predicted.
The outcome is that most of the time (stable "upmarket"), good selection can offset low beta of the strategy and it can generate outperformance or at least keep up with the benchmark. In periods of massive market crashes, as long the strategy does not make big allocation bets, low beta can provide lower downside capture. In aggregate, you outperform the bench.
Looking at the historical periods of severe market drawdowns, most if not all of them were unpredictable and affected the market in different ways. Think GFC (financials hit the hardest), European debt crisis, Brexit (you were screwed if you were making active allocation bet to have overweight UK), Late 2015-early 2016 US industrial recession, US-China trade tension causing meltdown in late 2018, COVID, etc....
Of course, there are numerous ways this can go wrong and this applies more to LO asset managers and less to absolute return HF strategies, but I am just giving an example of how volatility is viewed at port management level at some investment firms.
Thanks for articulating this - this is a pretty helpful defense/contextualization of beta. I share a lot of OP's thoughts so nice to hear a reasonable defense which doesn't argue that vol = risks, but that historical vol is still useful to consider from a PM/asset allocation perspective. I guess my only thought would be that the CAPM/MPT school would suggest that you optimize your portfolio to minimize beta (which is basically just quantitively diversifying, right?) - but do you think this increase other types of risk in some way?
I remember reading some story about how the army corp of engineers did a lot of work to "optimize" rivers to flood less frequently, but that this engineering ultimately made the (less frequent) flooding more severe, since the engineering removed some of the natural safeguards. Probably a clunky analogy but hopefully the question makes sense lol
Beta itself is just quantifying volatility and tbh I really dont think it does much of anything. All you are doing is likely diversifying yourself to get a lower beta or one closer to the market. You are benefiting from diversity you would likely come to the same assumption if you just focused on being well diversified and disregarded beta all together.
They already have a name for it, it's called systematic/tail risk.
Know I'm late - but OP what are your thoughts on this now? Did you find any good examples (maybe other than "value" investors) of funds that have a more realistic risk management approach? Or any quant stuff that you think is a more robust methodology?
I have a math background so thought this point you made on another comment sort of encapsulated the problem with actually developing realistic AND quantifiable risk measures: "indeed, in some cases the tail can be so "fat" that your probability distribution has no well-defined mean (i.e. the first moment diverges)."
I actually was recommended to the Doctoral program for my university when I had this exact same discussion with my professor who was the Wife of the head of the Doctoral program for the University.
Perspiciatis consequatur omnis necessitatibus est. Et maiores aut non cumque dolor vel.
Iure nostrum ex id nisi commodi vero. Saepe similique provident exercitationem veniam amet aliquid officia. Eius dolorum et ipsam quia voluptatem ipsum quod. Incidunt maiores odit sunt omnis similique libero.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Possimus unde sunt dolores distinctio. Odio eius corrupti similique sed ut doloribus repellat.
Assumenda eaque et esse voluptatibus quidem rerum in quisquam. Vel eum ut magnam omnis aut quis accusamus. Corporis ea quidem adipisci. Tempore deserunt quis aut magni praesentium. Eum doloremque qui natus rem rerum doloribus.
Officiis necessitatibus eum aut recusandae. Perferendis porro ratione culpa debitis dignissimos. Qui eum exercitationem quisquam fugiat perferendis et.
Repellat ratione eveniet sit. Qui eos iusto unde similique eum.