Or, How Do I Fill Out CalculateBeta's Create New Case Form?
There are a number of choices someone must make when calculating betas. When you use pre-calculated betas those choices are made for you. CalculateBeta™ allows you to make those choices yourself based on your task and your judgment. The options are briefly discussed below.
If instead you are looking for information about:
please follow the respective links above.
Citations are to the Bibliography/Suggested Readings.
Choosing the As-of Date
Betas are calculated as of a particular date. Sometimes, that date will be the current date. Other times, past dates are relevant. For example, when valuing a company as of a past date, that date will typically be the best date as of which to calculate betas. Or again, an analyst might want to compute a series of rolling betas, in order to determine if a particular company’s beta has been stable over time and, if not, when shifts occurred and whether there are trends. You can specify calculation dates as early as January 1, 2000 using CalculateBeta.
Some beta sources provide only current, recent or infrequently calculated betas. If a company changes significantly, even a beta which is only a few months old may no longer reflect the company's risk very well. A major acquisition or a large debt pay-off could change the operating or financial leverage of a company significantly in a short time. That, in turn, may change the company’s future expected beta.
Choosing the Look-back Period
The time period over which returns are measured is another important choice in calculating betas. Free and commercial beta sources often use a three-year to five-year look-back period and monthly returns to calculate betas. Bloomberg defaults to two years of weekly returns.
A longer measurement period will increase the number of return observations, regardless of whether daily, weekly or monthly returns are used, and that larger number of observations will in some respects improve the beta estimate. However, a longer look-back period may decrease the relevance of the beta estimate.
Historical betas are usually calculated to estimate forward-looking betas. Because companies change, return data from periods preceding important company events (e.g., changes in lines of business, or leverage) can make historical betas less indicative of future betas.
For example, General Electric was formed in the late nineteenth century and was part of the original Dow Jones Industrial Average. Therefore, even if computed on an annual basis, long series of prices and returns are available. However, GE's history includes several technological eras, a multi-decade period of heavy involvement in financial services, a couple of extended periods of involvement in media (e.g., through ownership of NBC, from 1986 through 2013, and also prior to 1930), and other changes. GE's beta and its comparability to any other given company as of a particular date and measurement period may be influenced by these changes across the company's history.
Industry-wide factors may also affect risk and return, and therefore beta. Such factors may include new laws (consider healthcare) or regulatory changes (consider the deregulation in past decades of industries such as airlines, communications, utilities, etc., or the many regulatory changes in finance). Other external events, such raw materials shortages or gluts, or new disruptive technologies, may also influence entire industries' risk characteristics and betas.
It is often more reasonable to use long look-back periods for companies which have had operationally and financially consistent businesses.
Also, when computing industry or peer group betas it may be desirable to use the same look-back period for all companies. In that case, the least common denominator may rule - that is, the company with the shortest meaningful look-back period may determine the look-back period for all companies. However, CalculateBeta makes it easy to sensitivity test the impact of these types of decisions by considering different cases.
Choosing the Periodicity
The time interval, or periodicity, of the return measurements must also be chosen. Returns can be measured daily, weekly, monthly, quarterly, annually, or over some other period (hours, minutes, every six-months, etc.).
Shorter return intervals will provide more observations. However, with shorter return intervals some securities may pose measurement problems due to a lack of trading. This problem tends to bias calculated beta results downward for less traded securities. Sum betas, which are provided by CalculateBeta, have been proposed to help with this problem.1 Simply using longer return intervals also helps.
In practice, most commercial beta services use monthly returns, which are less likely to cause bias from lack of trading, or weekly returns. Daily return data will tend to give rise to such bias for smaller companies, and may be undesirable if the look-back period allows use of weekly or monthly returns. Practically, many firms change too much over time for annual or quarterly returns to provide a sufficient number of data points. CalculateBeta does not, therefore, support calculating such betas.
Weekly returns may be desirable when companies have changed too much to allow sufficient monthly observations. For example, for a firm which completed a major restructuring of operations and financial posture 7 months ago, a beta computed using 60 monthly returns (common among commercial beta providers) may not be relevant. On the other hand, six or seven monthly returns is a small number of data points. In such a case, weekly returns may well be a better choice - 30 weekly returns would both capture a relevant period and likely provide sufficient observations.2
Choosing the Market Proxy
Prof. Roger Ibbotson and others have found that the choice of market proxy is of limited importance provided that the index chosen is fairly broad and value-weighted. Such indexes are highly correlated with each other. When desired, you can test the impact by considering different indexes supported by CalculateBeta.
It is important to note that much widely available stock and index data does not include dividend reinvestment and does not, therefore, reflect total returns. Sometimes, some data is total return and other data is not, potentially resulting in a mishmash of return data. Beta calculations are best performed using total returns, although the impact of excluding income returns from the calculations will vary in part with the company and the industry (i.e., where dividend payments are common, the impact can be substantial). CalculateBeta uses total returns for stocks and total return indexes when calculating betas.
1 One assumption in using sum betas concerns the duration of the lag which a thinly traded stock's price requires before reacting to changes in the market index. CalculateBeta supports calculating sum betas with monthly, weekly or daily periodicity, and therefore, lags, but what may make sense in a given case will be a matter of facts and circumstances. For more information about sum betas, see Ibbotson (1997), Pratt and Grabowski (2014) and other sources.
2 Actually, CalculateBeta puts certain limits on the inputs it will accept, and one of those is the minimum number of observations. Regardless of whether you specify daily, weekly or monthly return measurement, you must request at least 30 observations (although if a company has a short trading history, it is possible that fewer than 30 observations will be returned).