Research shows that the raw beta of firm tends to revert to the systemic average asset. But why does this happen? I can’t think of a reason for this phenomenon, even though statistically, it was shown to exist in a few published papers.
bro, did you not cover this topic in your 1-hour long monologue on beta in a recent interview? shame
Who cares about betas…we are all alphas here…
Can’t be bothered to look up the papers. Unlikely that I’d find them either.
There is Blume’s paper from 1975, the one Bloomberg uses to adjust betas. Then there is Jorion’s regression weighing technique from 1986 where the raw regression beta receives the most weight when the standard error of beta from the regression is smallest.
Anyone know why systemic risk of firms revert to the broad market average?
You could think of the beta as being from two sources. The first is general risk of the business, such as the industry or sector. An industrials stock will have a higher beta than a consumer staples stock, about every time. You wouldn’t expect that aspect of the beta to mean-revert, but it may be that the sector/industry beta declines if it was high previously. The other is more related to security-specific issues. For instance, the stock might have been a young company, but now it is more established. In that case, you’d expect the beta to decline to be more consistent with other stocks in the sector/industry.
Estimating a time-varying beta requires rather sophisticated econometrics. However, there are simple things you could do, like assume the beta follows a mean-reverting process, supply the average beta of the industry or sector as the level it mean-reverts to and then make some assumption about how fast it will mean-revert.
Another reason to use the approach is that beta is only estimated with some error. The further your forecast horizon, the greater the error. The less confidence you have in your high beta estimate, the more you should bring it closer to 1.
I always found this assumption problematic, and I suspect the research looked at large conglomerates, which tend to be more like microcosms of the economy. At least thats how I justified it to myself. If the research ended in the 1970s and 1980s and people just ran with the conclusions, that would make sense, since it was really only in the mid 80s that firms tried to go the “core competencies” way rather than turn themselves into giant conglomerates.
The other possibility is that corporate finance incentives push firms to manage their leverage to bring the stock beta more in line with index beta. I forget if this conclusion was about stock beta or unlevered beta.
Completely unfounded theory based on reading none of the appropriate materials: the mean reversion being detected is the result of survivor bias in one way or another.
In all seriousness, I think refering to this adjustment as “mean-reversion” is a misnomer because it implies that the true beta was reliably estimated from the historical data but we expect it to change in some fashion. I don’t believe you would find a theoretical justification of why betas should converge towards 1.0 over time.
A better way to think of the adjustment, in my opinion, is from a Bayesian standpoint. Absent any other information and prior to making any measurements, your best educated guess for the beta is that it is 1.0. You then update this prior with new information after you measured beta, recognizing that your measurement is not 100% accurate, and produce an “adjusted beta” which combines the prior info and the observed data. Arguably, this is your best estimate for the true beta rather than simply taking the direct result of your measurement and ignoring your prior “beliefs.”
Bayesian inference is in the eye of the beholder - the probability distribution you pick for your prior will determine the amount of adjustment. Blume’s adjustment seems static whereas the Jorion technique you describe appears to shift the “degree of belief” you place in your data vs. your prior based on the regression error - that seems reasonable to me, even though I am not familiar with the details of the approach.
I tind Twice the man’s thoughts the most realistic. I do not know the methods and measurments used to arrive at that conclusion, but it could very well be survivour bias. The assets with too high betas either drop out from data, or mature. The ones with too low betas attempt to increase leverage and acquire companies in order to create value. In both instances, we assume that beta here is a real measure of systemic risk, and unsystemic risk can be ignored. Not realistic, but for argument’s sake, nessecary.
But on the other hand, a mature firm with optimal capital structure and low sensitivity to the market has no reason to increase it’s systemic risk towards the market average, so that’s a blow to the theory.
Since M&As make up a significant part of most firms’ operations (moreso today than back then?), it seems sensible that a typical firm would have it’s total beta revert to one with that in mind, that’s how I’d explain it so far. The answer could also lie in a multi-factoral modified CAPM model like Fama, where all the betas converge towards the market average, including the market beta?
_ Most _ firms?
Seems farfetched.
Well that depends on the magnitude of most, and the universe of firms. I don’t have data to support it.
But I’d say that around 70% of firms invlove in regular M&A activity, I’d be counting both parties as part of the statistic. I can’t remember the source the at the moment, but inorganic growth has been playing a bigger part of operations with every passing cycle. There was a couple of M&A waves during the time the papers were published, and they have gotten bigger since.