Lesson 2 of 7
Four betas, one company
The same stored prices give one company betas of minus 0.05, 0.24, 0.37 and 0.41 by window and frequency. That is 2.51 points of cost of equity.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- Four ordinary measurement choices on one company's own price history produced betas from minus 0.049 to 0.408.
- That is 2.51 points of cost of equity, on a company whose business did not change between the four calculations.
- The choice is not resolvable from the data. It is a decision to be made, written down, and carried consistently.
A published beta arrives as a single number with no method attached. Here are four, computed from the same stored daily closes for one company against the same market proxy, under four choices that are all standard practice.
One company, four windows, four betas, four costs of equity
- 1 year of daily returns (beta -0.049)4.73%
- 2 years of weekly returns (beta 0.240)6.32%
- 5 years of monthly returns (beta 0.374)7.06%
- 5 years of weekly returns (beta 0.408)7.24%
The four choices, and why none is wrong
Window. One year, two years, five years. A short window is more current and noisier. A long one is steadier and describes a company that may no longer exist in that form.
Frequency. Daily, weekly, monthly. Daily gives the most observations and the most noise, including the effect of a stock that does not trade every minute. Monthly gives 61 points over five years, which is a small sample for a regression.
Index. The model asks for the market portfolio. Every implementation substitutes something, and this one uses an exchange traded fund tracking the S&P 500 rather than the index itself, which is a choice and is stated in the source trail.
Adjustment. Many published betas are pulled part way toward 1.00 on the argument that betas drift there over time. That adjustment would turn the minus 0.049 above into 0.30 and the 0.408 into 0.60.
What the negative one is telling you
The one-year daily regression returned minus 0.049. Taken literally it says this company's price moved slightly against the market over that year, and the model would then price it below the risk-free rate at 4.73 percent.
Nobody believes that. It is the measurement failing rather than the business changing: a large, defensive staples company whose price barely responds to index moves produces a slope near zero, and the sign of a slope near zero is noise. This is why the product holds beta inside a floor of 0.30 and a ceiling of 3.00 before it reaches a cost of capital, and records in the source trail that it did: for this company it says measured 0.24, held at 0.3.
What to do about it
Three habits. Pick one method and use it for every company, so comparisons are not contaminated by method drift. State it beside the rate: window, frequency, index, adjustment. And when the answer matters, run the valuation at the high and low beta rather than the middle one, because the range is the honest output.
Check yourself
Four betas for the same company came from what?
- Four different data sources
- Four combinations of window and frequency applied to the same stored prices
- Four different market indices
Show the answer
Answer: B.
One price history, four ordinary choices: one year daily, two years weekly, five years monthly, five years weekly.A one-year daily regression returns a beta of minus 0.049. What is the right reading?
- The company is a hedge against the market
- The slope is near zero and its sign is noise, which is why a floor is applied before the rate is built
- The regression was computed incorrectly
Show the answer
Answer: B.
A near-zero slope on a defensive company carries no information in its sign, and taking it literally would price the equity below the risk-free rate.What is the practical response to a 2.51 point spread in measured cost of equity?
- Average the four betas
- Fix one method, state it beside the rate, and run the valuation at both ends when the answer matters
- Use the lowest, since the company is defensive
Show the answer
Answer: B.
An average of four arbitrary measurements is a fifth arbitrary measurement. A stated method and a range are both defensible.Sources
- Computed for this lesson from the product's own stored daily closes for the company and for SPY, 2021-08-13 to 2026-09-14, under four window and frequency choices.
- Sharpe's capital asset pricing model: the cost of equity as the risk-free rate plus beta times the premium.
- Procter and Gamble, Form 10-K, fiscal year ended 30 June 2026, filed 2026-08-04: the cash flow, debt and share count behind the downstream figure.
What this leaves out
- Five years is the whole price history stored here, so a longer window cannot be measured at all.
- The market proxy is an exchange traded fund tracking the S&P 500, not the index and not the market portfolio the model asks for.
- No adjustment toward 1.00 was applied to the four figures. Applying one would move all four and is itself a choice.
- Nothing here is advice about any security.