Talk:Beta (finance)/Archives/2015

calculation question

In calculation of Beta, what kind of return data do we use, daily, montly, or other? Do we use return data that is greater than 1, i.e, if the interest rate is 5%, we use return as 1.05 or 0.05 ? This data specification Thanks, Hua

depends on the purpose - all time frames are used. the returns are generally of the form .034. Its actually my understanding (not reflected in this article) that the technically correct returns to use are the market returns less the risk free rate. Is this usually ignored in practice, but I think the academic specifications call for it. 24.153.227.130 17:14, 21 February 2007 (UTC)

I am formally educated in finance. The answer is that you use the Beta which balances two tradeoffs. If you are investing long-term, then you look at the yearly betas. However, you use shorter betas whenever there has been a redirection in the risk of the company. For example, Boeing has a different beta in times of war, than in times of peace. Early in a long war, the beta for Boeing in theory would be small.

Going into more depth than you probably wish to know, when you are chosing assets to include for a portfolio, you look deeper into the underlying data to see which beta best matches what you would expect, for the portfolio. You basically have to understand the actual operations of the firms real assets which generate the cashflows, and you have to understand and categorize major risk groups and assign weighted averages to see which other assets have comparable expectations. Betas have so much underlying math, that you almost have to "un-Beta" the initial beta, and then recalculate a new beta from your tweaked data. It's messy, but by doing this process, you know every nuance of the Beta which you chose to use. Once you understand all the nuances of all your summary statistics of your assets, its like being keen to new insights and you'll believe in your data and you'll be more confident in your investment decisions. Sentriclecub (talk) 13:26, 10 September 2008 (UTC)

if i have the intercept, the standard deviation and the correlated, how can i calculate the beta??

To reiterate Hua's question from a user's perspective, ‘’What time interval between prices is used for calculating Beta as reported by common financial information providers?‘’ That is, if a site says Beta = 1.05 over a 3-year period, did they use the annual, monthly or daily close? 67.244.129.52 (talk) 11:42, 8 December 2011 (UTC)

Answer to Hua: The sampling rate for computing beta is monthly. And the return used would be .05 (the log of 1.05 -- because we're in a lognormal world). Now, that's just the convention used in the financial community, as most funds' returns are published (and audited) monthly. Also, the period of time used, i.e. the sample size, is five years. Again, that's just convention. Technically, beta is a function of all of those parameters (sample rate, sample size, etc..), so you can plug in whatever you want to do the calculation. But the beta values listed on finance sites like google, yahoo, msn, etc. are all computed using 5 years of monthly returns (in log form). — Preceding unsigned comment added by 172.90.196.198 (talk) 13:26, 3 November 2015 (UTC)

lack of correlation between beta and average return

Although I'm not a finance person, I just read

• Fama, Eugene F & French, Kenneth R. 1992. The cross-section of expected stock returns. The Journal of Finance, 47 (2), 427–465.

which reports that there is essentially no correlation between beta and average returns for 1941–1990. This is even true when beta is the only predictor in the regression model. This finding contradicts what this article claims, namely a positive correlation between beta and return (that is, more risk = greater return). (Which brings up the question: who says "Academic theory claims that higher-risk investments should have higher returns"?) There was a correlation for returns before 1963, but 1963–1990 has no relation between the two and 1941–1990 has only a weak correlation. So, it seems to me from this that beta is worthless. Perhaps there should be some explanation of this in the article, and perhaps it is worthwhile citing Fama and French. Maybe there is other more recent research that looks at this even more. peace – ishwar  (speak) 09:35, 3 December 2012 (UTC)

Answer to ishwar: There IS more recent research that looks at this more. The bottom line is this: Fama and French have published extensively on the CAPM pricing model, which predicts among other things that discount rates are linearly related to market beta. They have typically found that beta predicts discounts rates reasonably well, but not perfectly. They conclude that additional factors are needed to fully explain discount rates. They did NOT conclude that beta is "worthless."

Furthermore, recent work by Prof. Richard Roll articulates the problem of data sensitivity to measurement. He points out that by making only trivial corrections to observed data, discount rates can be shown to be FULLY explained by beta. In other words, it may be that beta DOES, in fact, perfectly predict discount rates, it's just that you have to get the measurements exactly right or it'll be off.

In any case, I recommend you do the experiment for yourself. Start with the top 10 market cap stocks on the S&P 500. Look up their beta values on msn finance. Then take the reciprocal of the P/E, and use this as your discount rate. So each stock gets a beta (your x-value) and a discount rate (your y-value). Plot the scatter in excel for all 10 stocks, then add a linear trend line. Tell excel to print the r^2 value, which tells you how well the x's explain the y's. Decide for yourself if beta is "worthless." — Preceding unsigned comment added by 172.90.196.198 (talk) 13:54, 3 November 2015 (UTC)