This is for the practitioners who actually rely on beta for any kind of valuation.
What do you use and why? Bloomberg beta or … ? Also, over what time period?
I am too lazy (actually pressed for time) to google the research on the stability of individual company beta; but I suspect it cannot be very stable. Yet CAPM assumes it to be pretty much a constant.
Maybe someone knows some research on the “volatility in volatility” i.e. std dev of beta or std dev and so on.
At our valuation practice, the betas we use for our subject public company approach, for cost of equity, we use 5 years prior to the valuation date, monthly betas (so 60 data points), off the S&P 500 index. We also have an option to use 2 year weekly beta for younger companies. Both come from CapIQ.
At the prior small firm I was at, we just pulled the pricing data of yahoo finance and ran the regression ourselves, 5 year monthly.
Over a 5 year period, it should be relatively stable, because of the way beta is calculated, with 60 data points. A few outliers may screw it but with 60 data points, it shouldn’t be thaaat much, correctly me if I’m wrong.
We use these approaches, not to make money, but to not get in trouble with regulators. It’s acceptable practice.
Also, other valuations I’ve reviewed from other firms use something close, usually 5 yr, either monthly or weekly data, S&P 500, MSCI world index, and recently I saw one that used local market index for each company in the comp set… I’m still thinking about that…
hpracing I still don’t work in an instiutional firm, but irregardless of common practice. You are right in the sense that 60 data points on a five year period is probably a good balance between reducing error, and explaining underlying fundamentals. But most of the time, the R-squared is just too low to explain the beta properly. This is why the best way to derive beta is by a bottom up approach by carefully choosing comparables. This attempts to minimize regression errors, and idiosyncratic influences on the measure of systemic risk.
^ damodaran preaches that approach, which sounds good to me, anyone ever see it used in practice?
I’ve not done it, because I tend to work with indexes rather than individual stocks. I assume what’s happening is that you delever the observed betas to neutralize the effect of financial leverage, then come up with a beta for the business, using comparable companies, then relever using current financial leverage to figure out how sensitive the stock is, yes?
In my experience with indexes, beta does seem to change a disturbingly large amount depending on what time period you are using. Remember that a good estimate of beta should ideally be at least one and preferably an integer number of business cycles, which is typically 6-8 years, and not the more conventially used 5.
I may be wrong but I think what you describe is a market compriable or industry beta. I don’t know that term I just do them haha
My understanding of a bottom up beta is you split the company, example damodaran gives is for Disney, into lines of business, say theme parks, movie business, theme park, broadcasting, ect. You try to find pure play comps or ones with similar economics for each line, delever and relever, and weight based on revenue of each business line, that way you are not just performing a single beta calc of Disney as a while.
if in free today, I’m going to see if capiq lets me graph a beta over time and see just how much it fluctuates, I’m interested now.
Yes, if the company has multiple business lines or is a conglomerate, then you would redo that approach for each line of business, using separate comparables and then merge them by relative share (by dedicated assets I would think, though possibly by revenue). I don’t see what I’ve said and what you’ve said as being incompatible. I was just describing what I understood as the bottoms-up approach to finding beta using comparables that you were asking about. If your company has more than one business line that is substantially different from the others, then you would break the lines apart and merge the betas together at the end.
Reading your post and mine again, and see what you’re saying.
Yes, Damodaran preaches that approach, and the general finance literature agrees as well.
bchad, as for using a whole business cycle to measure the beta, this comes down to the analyst’s subjective approach. But generally, five years attempts to grasp one business cycle on average when you consider all industries. If the underlying business has changed considerbly in the past 6-8 years (assuming it’s one business cycle for the company in question), let’s say, has matured. Then measuring a whole business cycle might overestimate beta in this case. Another case in point, is that if the business in question exhibits the same behaviour of cyclicality compared to the previous one, then it would be prudent to derive a bottom up beta of other companies with smillar size and operating structure, for a better reflection of the leading measure of beta. In honesty, it would be prudent to do the same for almost all cases where comparables are confidently defined.
Even then, you will unlikely get a precise measure, but it still remains the best estimation of systemic risk in a CAPM model.
For a particular company, beta over time will likely fluctuate significantly, but the beta of a portfolio / the average beta for a group of comps will be a lot more stable. If you look at the R2 for the OLS beta for each comp, it may be very low - but the R2 for the beta of the comps portfolio may well be higher than each of the individual R2’s. So if you’re using a bunch of comps and taking the average beta for your target, it is not particularly meaningful to report individual regression stats or look at the individual variability of each comp beta.