Beta: Leverage You Did Not Ask For
Beta is the slope of a regression of an asset's returns against a market index over a chosen window. A beta of one means it has moved roughly with the index, above one means more, below one means less. It is fitted, historical, and measured against a single index.
How it works
Beta measures how much a holding moves when the market moves. Take its returns and an index’s returns over the same window and fit a straight line through them.
The slope of that line is the beta. It is a regression coefficient: one number describing how steeply an asset has responded to the index.
A beta of one means it has moved roughly with the index. Above one it amplified the index in both directions, below one it damped it, and a negative figure means it moved the other way.
And a fitted number depends on how you fit it. Change the window length, or measure weekly returns instead of daily, and the same asset gives a different beta.
The leverage nobody named
High beta is leverage you did not ask for. A holding that moves half again as much as the index in both directions has the risk profile of a levered position, bought without anyone calling it that.
The decision is made by the sizing, not by the ticker. Hold several high-beta names at ordinary sizes and the account is a levered bet on the market, chosen by nobody. A leveraged exchange-traded fund at least says so.
It is measured against one index, and the choice matters. A share’s beta against a broad market index and against its own sector answer different questions. Choosing a benchmark is a decision, which index funds and ETF investing make for you.
And it describes the past, not the next fall. The line is fitted on returns that have already happened, so a beta summarises its estimation window and promises nothing about the next.
It says nothing about costs or about business risk. Spreads, financing and the state of the underlying company are absent from the regression, and a low figure on a deteriorating business is still deteriorating. That is valuation work.
In practice
Participation is not in the calculation at all. Volume never enters the regression, so a beta on a thinly traded share looks as authoritative as one on the most liquid share on the exchange.
A longer window gives a calmer and staler number. Lengthening the estimation period smooths single events away, at the price of describing a business that may have changed.
In a crash every beta moves towards one. When everything sells at once the spread between defensive and aggressive holdings compresses, and one opening gap takes the whole book down together.
It never tells you where the risk actually sits. Beta says how far a holding tends to travel, not the price at which the reason for owning it stops being true — a stop loss and risk management question.
And holding beta steady has a running cost. On this site’s shared history a round trip costs 0.0098 price units, which is 2% of a median bar’s range and 45% of the smallest bar.
Working out a portfolio’s beta
A portfolio’s beta is the weighted average of the betas it holds. Multiply each holding’s beta by its share of the portfolio’s value and add the results; the total is the sensitivity of the whole account to the index.
That single number is more useful than any of the individual ones. It says nothing about which companies you rate, and everything about how much of your money is one bet on the market — the exposure your position sizes created rather than the one you chose.
It makes concentration visible without naming a sector. Several holdings that each look moderate can add to a figure well above one, and a portfolio near 1.5 behaves like a levered index position however carefully each name was picked.
Which turns beta from a selection number into a sizing number. That is where it earns its keep, alongside the Sharpe ratio for judging what the exposure returned.
What beta is not
It is not volatility. How much a holding moves on its own is standard deviation.
It is not correlation. Correlation measures the strength of the relationship; beta measures its slope.
It is not a measure of business risk. Debt, competition and customer concentration never enter the fit.
And it is not a valuation. A cheap high-beta share and an expensive one give the same slope.
When it fails
In a flat market the fit has almost nothing to fit. When the index barely moves the slope comes from a cloud of tiny returns, and a trading range produces a confident number made mostly of noise.
The failure that matters is convergence. In a sharp market-wide fall individual betas drift towards one as everything sells together, so the diversification the numbers implied disappears exactly when it was needed.
A second is treating it as a trading signal. Beta is an input to a required-return estimate inside a discounted cash flow, and a portfolio-level exposure measure. Both are legitimate. Neither says when to buy.
A third is comparing figures from different providers. Two services using different windows, intervals and benchmarks publish different betas for the same company, and none is wrong.
A fourth is assuming it holds still. A firm that issues debt or changes its mix of business changes its sensitivity, and the fit lags by the whole estimation window.
And a fifth is reading a low beta as a drawdown estimate. On this site’s shared history the deepest fall was 3.76% and the median 1.36%, yet 95% of bars sat below a prior peak, the longest stretch running 73 bars.
The original data
The honest analogue of high beta is a leverage measurement, and this site has one. In
research/series-measurements.json, built by site/measure_series.py over a shared 576-bar history,
three times exposure returned 8.93% rather than the naive 10.82%, while the deepest drawdown went from
3.76% to 11.08%. A high-beta holding behaves the same way, unlabelled.
So the downside scaled in full and the upside did not. This is daily-reset exposure on a synthetic series, an illustration of the mechanism rather than a claim about any real stock. Compute the weighted beta of everything you hold before adding another position, and treat that number as the size of your bet on the market.
Related
Correlation measures how tightly two series move together, which a slope does not answer. Risk management turns a portfolio-level beta into a position size. And standard deviation is what people usually mean when they call a share volatile.
For a stretch I held several names that all felt like separate ideas, and they were not. They were the same trade with different logos, and on the days the market fell I lost on every line at once. Nobody sold me leverage and I never asked for any, but that is what the sizing added up to. I add the betas up now before I add anything else.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.