When a news story changes what a stock is worth

A stock is worth the cash it is expected to pay its owners, discounted for how risky those payments are and how far away they are. That is intrinsic value. A news story matters when it changes one of those inputs. The price can move for other reasons, including a change in the discount rate that has nothing to do with this firm, and a burst of trading that fades. Reading the story is how you tell those cases apart.

What you are updating

James Ohlson's 1995 residual-income model, published in Contemporary Accounting Research, is a useful way to see the inputs. Under clean surplus, value equals book value plus the present value of expected residual earnings. Residual earnings are earnings minus a charge for the capital already on the books. A story changes value if it changes expected earnings, the period over which those earnings stay above the cost of capital, the book they are earned on, or the cost of capital itself.

A discounted-cash-flow model has the same joints. You need a forecast of cash flows and a rate that turns those future dollars into a present value. News is not a third term in the model. News is the reason you revise a term that is already there.

This is why "the stock was in the news" is not an analysis. Coverage is not an input. Fang and Peress, in the 2009 Journal of Finance, found that stocks with no media coverage earned higher average returns than stocks with heavy coverage, after the usual risk controls. Being written about is not the same thing as a revision to cash flows. The premium they document sits with the stocks nobody wrote about.

Two kinds of news, with opposite signs

John Campbell's 1991 paper in The Economic Journal splits an unexpected return into two pieces. One is news about future cash flows. The other is news about future discount rates, which he also calls expected-return news. They enter with opposite signs. Higher expected cash flows raise the price. A higher discount rate lowers the price, even if every cash flow forecast is unchanged.

That split is the whole reason to read the story instead of the percentage change. A rate cut, a wider credit spread, or a rise in the equity risk premium can reprice a stock whose business did not change this morning. A lost customer, a cut in guidance, or a contract that adds three years of revenue can reprice a stock on a day when the market discount rate did not move.

At the level of one firm, cash-flow news is the larger piece. Tuomo Vuolteenaho, in the Journal of Finance in 2002, decomposed firm-level returns with a vector autoregression. For a typical stock, the variance of cash-flow news was more than twice the variance of expected-return news. Expected-return news was highly correlated across firms, so it survives in a portfolio. Cash-flow news mostly cancels when you diversify. If you own one name, the story about that name's cash flows is the news you cannot get from the index.

Long Chen, Zhi Da, and Xinlei Zhao revisited the question in the Review of Financial Studies in 2013, using direct cash-flow forecasts rather than a dividend regression. Cash-flow news was a large part of returns, and its share rose with the horizon. Past two years, cash-flow news was the more important component. Their result holds for firms and for the market as a whole. The practical reading is the same: the further out you say you are investing, the more the story has to be about cash flows rather than about today's discount rate.

The price is not the revision

Robert Shiller's 1981 American Economic Review paper asked whether prices move too much to be explained by later changes in dividends. They do. Volatility is larger than the volatility of the discounted dividend path. A headline and a price jump can sit next to each other without the jump equaling the change in intrinsic value.

There is a second, quieter reason. Vuolteenaho also found that shocks to cash flows and shocks to expected returns are positively correlated for a typical small stock. Part of the good cash-flow news is offset by a higher discount rate, so the price moves less than the cash-flow revision alone. You cannot recover the revision by dividing the return by anything simple.

Richard Roll's 1988 presidential address, published as "R2" in the Journal of Finance, made a related point from the other direction. Market and industry factors leave most of a firm's daily variance unexplained. He left open whether the remainder was private information or trading that was not information. Jacob Boudoukh, Ronen Feldman, Shimon Kogan, and Matthew Richardson, in the Review of Financial Studies in 2019, identified firm-specific news with a topic model instead of a raw headline count. Days with identified, relevant news account for much more of firm-level volatility than a loose headline screen suggests. The lesson for a reader is narrow: relevance is the screen. A mention is not a cash-flow revision, and a day with no relevant story is a weak day on which to invent one.

How to read one story against the model

Take the story apart in four questions. Write the answers down. If you cannot answer them from the text, the story has not given you a revision yet.

1. Which input moved? Near-term earnings, the growth rate after that, the capital the firm must invest to get those earnings, or the discount rate. A lawsuit accrual is earnings and sometimes a permanent capital hit. A refinancing that cuts the coupon is a discount-rate and cash-interest item. A product rumor with no customer and no date is none of these until the text supplies one.
2. Is the change in the firm's own words, or in a commentator's? Guidance, a contract value, a unit count, and a date are inputs. An adjective is not. Paul Tetlock, Maytal Saar-Tsechansky, and Sofus Macskassy, in the 2008 Journal of Finance, found that the share of negative words in firm-specific stories forecast lower earnings, and that this was strongest when the story was about fundamentals. The same author's 2007 paper, on a market column rather than a firm story, found pessimism that reversed. The difference is the object of the sentence. A sentence about this firm's orders is a candidate for the cash-flow forecast. A sentence about "the mood" is a candidate for a discount-rate or sentiment effect that may not last.
3. How long does the text say it lasts? Residual income is a stream, not a single quarter. A one-quarter tax item changes near-term earnings and should not be capitalized as if it repeats. A multi-year contract can be capitalized only to the extent the text supports renewal and margin. Ohlson's model makes persistence an explicit input. The story either gives it to you or it does not.
4. What did the price do, and what did it not prove? Record the move. Then leave it out of the value estimate until the three questions above have answers. Shiller's point, and Campbell's sign convention, are there to stop you from treating the return as the answer.

Getting the story, not a pile of mentions

None of this requires a particular vendor. It requires the text, the time it was published, and a way to know the story is about the firm rather than a passing mention of its ticker. That is the part a news feed is for.

MarketNewsAPI tags stories to tickers and splits stock items into news, press releases, and legal notices. The press release is the company's own account, which is the right first document for the four questions above. A legal notice is often a standing filing, not a revision to guidance. The short summary on each story is a neutral compression of the text the model saw, so you can discard items that never touch earnings, capital, or risk before you open them. The article is available over REST at /v1/news and on the live stream. The docs describe the filters, including tickers and type=press_release.

The feed does not estimate intrinsic value. It does not know your discount rate. What it does is put the story that might revise a cash-flow forecast next to the name, with a timestamp, so the revision is something you can actually do.

Frequently asked questions