How fast does supply-chain read-across get priced? Five years of evidence
We measured how much of a company's earnings move reaches its suppliers, customers and competitors across 608,141 event-company pairs. Same-day transmission is real and stable; the slow catch-up the literature is famous for is not reliable in today's large caps.
rejo research ·
One of the best-known results in empirical finance is that news travels slowly along economic links. Cohen and Frazzini (2008) showed that when a company's major customers have a good month, the supplier's stock tends to follow the next month, and a long-short strategy built on that earned more than 150 basis points a month. Barrot and Sauvagnat (2016) showed that shocks to suppliers of specialised inputs spread to their customers. Madsen (2017) found that the catch-up concentrates just before the linked company reports.
Those results are why read-across matters. But they were measured mostly on samples from the 1980s to the 2000s, heavy in small stocks that few analysts followed. Pinchuk (2023) re-ran customer momentum and found that after 2005 the value-weighted spread fell to about 0.6% a month and lost statistical significance. Ali and Hirshleifer (2020) argue the effects that survive run through links analysts don't share. So we asked a narrower, practical question for investors in large caps today: when a company reports, how much of its move reaches the companies linked to it, and when?
What we measured
We took 18,905 earnings prints from the 607 large caps rejo covers, over 5 years of daily prices. For each print we measured the reporting company's two-day move against its own sector. We then followed its links (suppliers, customers, competitors, and companies two or three steps away through rejo's supply graph, including private suppliers as relays) and measured each linked company's move against its sector over three windows: the same two days, days 2 to 10, and days 11 to 40. Linked companies reporting within three days were excluded so their own news couldn't contaminate the result. The most recent year was held out.
Result 1: competitors move with the reporting company, the same day
Competitors move about 19.3% of the reporting company's move on the same two days (t = 15.9, 15,597 pairs), and the held-out year agrees: 20.2% (t = 8.7). Customers (1.3%, t = 2.4), suppliers (2.3%, t = 3.6) and companies that sell to the same customers (1.4%, t = 3.1) also move the same days, more weakly.
Result 2: the slow catch-up does not reliably survive
In the days and weeks after, there is no robust drift. Competitors over days 2 to 10: 0.8% (t = 0.3). Customers over days 11 to 40: 3.0% (t = 1.5), and the held-out year has the opposite sign. Suppliers show a drift in the held-out year (6.8%, t = 2.9) but none in the four years before (0.3%, t = 0.2), which looks like a regime (the recent AI capex cycle) rather than a law.
We also tested the sharper version an investor would trade: if a linked company under-reacted on the event days relative to what its link implied, did it catch up later? Of 1,663 large historical gaps, 50.9% closed, with an average capture of 0.45%. That is a coin flip.
What it means for investors
- Speed and completeness are the edge, not drift. In large caps the market prices most read-across within two days. The value is knowing, the moment something happens, which of your names it reaches and through which link, especially the second- and third-step ones a sector screen won't show.
- Risk, not just return. If competitors reliably move a fifth of the reporting company's move the same day, a book's exposure to a print is measurable before the print. That is what rejo's Book X-ray and its pre-print view use.
- Be suspicious of drift claims. Many read-across signals are sold on results from samples where they no longer hold. We will only make a drift call for a link type once its live, out-of-sample track record earns it, and we publish the record.
Caveats
Links come from today's graph. Applying them to five years of history assumes the relationships existed then; supply relationships are sticky, but this is a known look-ahead. International listings without earnings-date history appear only as linked companies, not as reporters. Moves are measured against a market-cap-weighted index of each company's own sector built from our universe.