https://www.youtube.com/watch?v=etlv7qTQUSY
For those who don't know, the man in that video is Tim Bennett - a truly wonderful teacher who's videos should be mandatory for anyone seeking financial knowledge - and for those who can't be bothered to watch the video, beta is basically a numerical illustration of the risk of an asset in relation to the wider market.
I'm now going to write with the assumption that you've watched that video:
One of the problems I have with regression betas in portfolio management is that we take our regression calculation and get a slope. We then take the gradient of this slope and it becomes a beta. The issue is that when we take the gradient of this slope we get a level of standard error. Now, in the US for example, the typical standard level of error for equity betas is between 0.2-0.25.
This is a bit of a problem, because it means that if Apple has a beta currently of 0.90, it may actually have a beta anywhere between 0.65-1.15 and it gets worse because depending on how you calculate your regression beta, you can end up with a cornucopia of betas and then pick and chose which ever one you want to believe.
Regardless of this single issue, the much bigger fall back with betas for me is that only about 20-25% of the risk in a company (according to Aswath Damodaran - an expert on betas) is market risk and a beta will only ever capture that portion of risk. This then can get even more fiddly, because you then have to analyse the extent to which company specific risk becomes market risk. The example that Aswath Damodaran uses was the whole Lehman Brothers debacle, which began with internally odd CDS bets (a company specific risk), but ended with a global financial collapse.
Leading on from this, betas are often calculated for whole investment portfolios, which we will assume have been diversified. This leads us to another issue, which is that diversification (ironically) only works as a hedging method when it's required least. So, in periods of crisis this iconic hedging method blows up because all of the global markets begin to move in tandem with each other. For those who don't believe me, this can be proved:
If you run a regression of stock returns against market returns you get R² and as I've said this averages at about 20-25% in the US, but between September 12th and December 31st of 2008 the average R² for US stocks rose to 70%, meaning that stocks were moving primarily with the market rather than with company specific news and events.
All is not lost though, because with a simple trick we can reduce not only this inherent statistical standard error significantly, but also get a better idea of the real levels of risk to our portfolios when we see rapidly declining markets and R² levels of 70% or more.
This certainly isn't an original idea, but if we get multiple regression betas (the more the better) under industry sectors and average them out, we get a much better picture of what levels of risk we're really seeing, in part because that level of standard error is reduced with the more sample regressions we take, but also because of the potential to use of many years of data in a beta calculation.
What's really good about this, is that sector average betas generally stay the same over long periods of time, which helps us personal portfolio managers to have a much better idea of our "actual beta" in periods of market turmoil. Although, this being said they do change, with the financial industry seeing large changes in average sector betas in the 2008 financial crisis, so it's not fool proof, but it is significantly better.
To conclude, I would say that it's worthwhile having a rethink regarding how you calculate and think about the betas you see for portfolios and stocks that you're viewing. The sector based beta calculation is certainly not a magic wand to success in risk calculating, but I feel that it does a much better job of giving realistic figures than standard regression betas do.
Enjoy,
The Masked AIM Trader
No comments:
Post a Comment