Analyst Calls Stack Up Monday—Block, Equifax, Lockheed: Signals or Noise?
# Analyst Calls Are Just Noisy Labels — Treat Them Accordingly
Monday’s upgrade/downgrade roundup covered Block, Boise Cascade, Equifax, FirstEnergy, Lockheed Martin, Lumentum, NetApp, Roper, Spotify, and a handful of others. The sell-side issued its usual batch of price-target adjustments, and the financial media dutifully transcribed them. For anyone running a systematic book, the correct response is to treat these as data points with documented, unimpressive predictive value — not as signals.
The academic literature on analyst revisions is genuinely interesting; the headline calls themselves are not. Post-announcement drift following *earnings surprise* revisions is one of the more robust anomalies in the factor literature. Brokerage upgrades aggregated in bulk, by contrast, tend to be momentum-chasing with a lag — the analyst upgrades the stock after it has already run, which means the “signal” is largely a noisy, delayed proxy for price momentum you could have captured directly with a 12-1 month cross-sectional momentum screen in `pandas` and about forty lines of code.
A few practical observations on this batch. Defense-adjacent names like Lockheed Martin trade heavily on budget-cycle news flow, which is seasonal and legislative — build an event-study window around NDAA milestones and appropriations releases rather than reacting to rating changes. Lumentum and other optical names are high-beta plays on hyperscaler capex; if you have conviction there, express it through the capex guidance of the buyers, not a sell-side target on the supplier. Equifax and NetApp are cyclical, and their analyst calls mostly restate the ISM and enterprise-IT spending data you should be ingesting directly. Roper is a compounder that the sell-side perennially rates Buy at whatever the current multiple happens to be — a rating with roughly zero information content beyond “it’s still expensive and we still like it.”
If you insist on consuming analyst data systematically, do it properly: pull the full revision distribution from an estimates provider, compute breadth (net upgrades minus downgrades across covering analysts), and test whether revision breadth adds alpha *after* controlling for momentum and earnings drift. In most samples it doesn’t, or the effect decays to spread-covering within days. Backtest it yourself with point-in-time data — survivorship-biased analyst histories are a classic way to fool yourself.
The meta-point: a list of ten Monday morning calls is a taxonomy of consensus, and consensus is priced. Your edge lives in the inputs the analysts haven’t processed yet — alternative data, order-flow imbalance, cross-asset leads — not in their outputs.