Much noise has been made recently about “fixing” the incentive structures that predominate in our financial markets. Never fear, there has been a lone voice of sanity crying in the wilderness, defending the way things have evolved in the markets, and to my mind, the way things probably have to be. I wish to stand with him and the doers, against the white-coated theorizers and commentators, and their vicious personal attacks on his friend in blogging, The Epicurean Dealmaker, stock symbol TED (who actually made the personal attacks first, but never mind, he’s a blogger). Having marshalled my thoughts, I do so here.
What are the carpers saying? Firstly that pay incentives are skewed to the short term, incentivising “traders” to take risky bets which look much less smart further out in time, like, say, CDOs. This is Martin Wolf’s major point, and I have some sympathy with it. Secondly, that traders should be paid for what they produce, not what the market makes you; they should be paid only for their “alpha-generation”. Instead a lot of investment returns are based on taking higher beta bets in rising markets. This is “fake alpha”. At some point those beta strategies — again, like leveraged CDO investing — will fail, and fail big, when the market drops. This is Professor Rajan’s substantial point, and is the one I have most trouble with. Both Wolf and Rajan suggest changes to compensation structures in financial markets: a realignment of incentives in favour of sustainable non-fake alpha-generation, compensation based on longer time periods (Wolf suggests a 12 years’ horizon) and “clawback” of escrowed compensation when things go wrong.
I have 3 problems with this:
Traders find it very hard to distinguish between beta and alpha when they are trying to make money
the concept of beta may even be unhelpful, a historical standard deviation-based measure with little predictive ability. This undermines the legitimacy of alpha
we don’t really know what is beta and alpha in investment returns, and how much is plain luck. Luck should be compensated if it persists
One quick point needs to be made before we hit the jump. Martin Wolf said in one response to TED
“Academic” is often used as a term of abuse. But pretty well all the fundamental ideas in modern monetary and financial economics were invented by academics.
Not in my field, they haven’t. Pure academics in portfolio theory brought us the flawed Black Scholes Merton option pricing model, LTCM, and all the useless bits of the Capital Asset Pricing Model. Everything useful in portfolio theory has come from academic practitioners, people like Benjamin Graham and Philip Fisher, who took on risk and got their hands dirty in the market. Can anyone think of exceptions? And no, enforcing economic orthodoxy at the IMF for a bit doesn’t count as proper work.
The first problem then: “alpha”, for most traders, simply doesn’t exist on a day to day basis; there is only “making money”, and this is hard enough. A good trader will use any edge to increase returns for his punters, and not care whether he is making money because of his view on the market, or because he has picked super-duper stocks. I think most traders know what alpha is:
where R¹ is actual return and Rm is return on market/benchmark.
Look, there’s a formula and everything — I don’t mean there is no such thing. But it is not possible to think of “alpha” when you are running a portfolio; it is not a useful concept. It is after the fact that “alpha” and “beta” exist, when your results are viewed by others and compared. And here is an interesting thing: most of the time we don’t actually know even after the fact whether our returns consisted of alpha or beta and in what proportions.
For instance, I can make alpha out of beta. So I think the stockmarket is going to go up. Imagine a 2 stock benchmarked portfolio, the benchmark of which consists of say, Ciena and Microsoft. I am not allowed to have any other stocks in it. I can change the beta of that portfolio, and make money without changing the stock selection, by buying more of the riskier one, Ciena, and less of the “safe” one, Microsoft. In the rally, Microsoft and Ciena move exactly (in real life they almost certainly would not, see below) according to their historical betas to the NASDAQ Composite, say CIEN has a beta of 1.5, and MSFT 1.0. I just fiddled with the beta of my portfolio and created value, which should will be counted as performance when it comes to bonus time, but presumably counts as “fake” or bad alpha, as defined by Professor Rajan. I didn’t need to understand the stocks, he would say, I just needed a view on the market. But you know what? Getting the market right is really hard too. The trader who made that bet deserves to be paid for it as much as the other guy who took the same position on fundamental valuation. Both could have gone just as badly wrong. It takes as much (of an admittedly different) skill to get this right consistently.
This is a point well made by TED, and others. Furthermore, beta and alpha are inextricably linked. I am a stock guy. I look for trends, find stocks, analyse them intensely, take big bets (which define my portfolio) and make money. But my tactical allocation is driven by my view of markets. I like the prospects of certain stocks and sectors, but understanding we are likely in a bear market means for some stocks it is not the time; they don’t make it into my portfolio. In that sense my alpha decisions are driven by my beta view. There is an “allocation effect” as well as a “selection effect”, in the argot, in creating alpha.
So, we can see that it may not be possible to pay traders or managers for “non-fake” alpha fairly, as we don’t really know where it begins and beta ends. In making and managing a portfolio under Rajan’s rules, the practitioners would lack a clear definition of what it is they should be aiming at. This is not a good start if we are to reconstruct compensation along new, fairer lines. Spinoza insists in the Ethics that clear definition of concepts is a sine qua non of being able to make informed decisions, and so should we. It might be that Prof. Rajan has a longer treatment of his proposal elsewhere which deals with this issue, if so he must forgive me.
My second major point is that I find stock beta as classically defined to be worse than useless as a concept. Please note the Rm in the equation above refer to “expected returns” for the market. In my experience, reliance on “expected”anything in equity portfolios tends to lead to disaster, unless the widest ranges of outcomes are used, and the wider the range of outcomes, the less useful the formulation becomes.
Beta, moreover, is a standard deviation-based measure. Regular readers of this site will know how I feel about standard deviation . Pace Taleb, in excluding the outcomes that lie beyond the standard deviation, the measure of beta completely misses the point, for it is those outlying events (compounded) that are often the most meaningful in describing the security’s actual return, relative to something else or not.
I hate multiple analysis and think it lazy. I am a DCF jockey. As such I am probably more acquainted with stock beta in the real world than the average bear, as it were, because beta figures so heavily in my “application” of the Capital Asset Pricing Model. I promise you, were I to use actual beta as defined by CAPM I would get nowhere, my DCF results would be meaningless and I would be groping around in the dark no better than the poor souls in Plato’s Cave, and the majority of my peers. Far from using beta as a measure of a stock’s correlation with the stock market to find its cost of equity, the beta in my models is an entirely subjective measure of the underlying business risk, or in other words, the riskiness of the fundamentals relative to the average listed company (it tends to center around 1.2 for most of my tech stocks, in case you are interested). But this is entirely subjective, and as such useless for Prof Rajan’s purposes. Classical stock beta is backwards looking,
and in looking backwards misses out the most meaningful movements of the stock, it describes what has been, and says nothing about what the actual covariance of a stock with an index will be in the future.
If α=R¹-βRm where β=bollocks, this undermines the very meaning of “fake” alpha. We are left with R¹-Rm, where the trader’s alpha return is simply whatever he made, over and above what we thought he would. In other words, we set him a target, and if he does better than that, we pay him more and ask no questions. This is what we do now. It is the current system.
So for my third and final major point, we have to recognise that we don’t really know, in the absence of any measurable standard of beta, what is fake and what is true alpha. We don’t really know what it is that drives certain individual traders’ persistent returns over long and ephermeral returns over short periods. We come back to the older distinction of “luck” and “skill”, and defer to Napoleon’s Law, that, in the absence of empirical data on skill, persistent luck is the best quality to look for in a general, and in a trader. Which makes it all the more reasonable to pay up for it, and, on the part of the bank, to pay up if a single trader has a good year, for now that trader is more likely to be one of the few who may do well the next, who have that magic persistence. That fat bonus is the bank paying up for the option on that guy, rolling him over to next year by paying his opportunity cost, and in that sense is perfectly rational behaviour.
Rajan and Wolf may not fundamentally misunderstand the bind the bank finds itself in. It is why they call for strong, invasive regulation to break that market imperative, the one where you have to pay for short term performance however it was achieved. But when you consider our fundamental ignorance of what drives traders’ returns, any system based on the lights they have suggested is likely to be hugely more flawed than the current one, which, remember, is one that has evolved over time. It is one which recognises our lack of knowledge.
As TED notes, there are parts of the financial industry where clawback already exists, and even in hedge funds some traders’ compensation is at least partially deferred, gets invested in the fund itself and vests on a 3 or more year horizon. But this is more a response to the perceived short termism of normal compensation structure, and is more than likely not a response to the fake alpha problem.
Sarbanes Oxely and Reg FD, the legislative responses to the last crisis, have not proven to be the wholly Good Things they seemed at the time. Sarbox has led to huge burdens in smaller US companies, contributed to the boom-bust in private equity, and helped foreign exchanges gain the upper hand vs the US capital markets. Reg FD, far from levelling the playing field in terms of disclosure, shut down informal information flows to the capital markets, reducing visibility and increasing volatility. Any regulation on traders’ pay along the lines proposed by Wolf and Rajan would probably be equally bad, but we don’t know in exactly what way yet. A very obvious case of “legislate in haste, repent at leisure”.