Analyst Calls Update: Block, Equifax, Lockheed—Signal or Noise? Check the Data.
Analyst research calls are a fascinating data source precisely because they’re so poorly calibrated. Monday’s roundup from 24/7 Wall St. — covering Block, Boise Cascade, Equifax, FirstEnergy, Lockheed Martin, Lumentum, NetApp, Roper, Spotify — is the standard format: ticker, rating change, price target, one sentence of justification. As signal, it’s nearly worthless. As a dataset, it’s mildly interesting.
The measurable problem with sell-side calls is well-documented: price target revisions lag price moves, upgrades cluster after momentum has already run, and the distribution of target-implied returns skews optimistically in a way that survives no honest out-of-sample test. If you’re building a factor around analyst revisions, the edge — where one exists at all — sits in the *change* and the *dispersion*, not the level. A fresh Buy on Block after the stock has repriced 15% is the market telling you something happened; the analyst is confirming it, at retail prices, for free.
There’s a slightly better use: consensus dispersion as a risk overlay. Names like Lumentum and Spotify historically carry wide analyst target spreads, which correlates with realized volatility in ways a simple ATR filter doesn’t capture. Feeding target dispersion into a position-sizing model — `pandas` for the merge, nothing fancier — is a legitimate, if modest, risk-management tweak. It won’t generate alpha, but it will stop you from sizing a consensus-conflicted name like a high-conviction one.
For the industrial and utility names in the list — FirstEnergy, Boise Cascade, Lockheed — analyst calls matter even less. Those price off rate expectations, commodity curves, and defense budget cycles, none of which respond to a target bump from a desk that covers forty tickers. If your systematic model trades them, trade the macro factors directly; the analyst layer is noise with a byline.
The honest backtest nobody on the sell side wants you to run: take five years of rating changes, sort by whether the call confirms or contradicts trailing 3-month momentum, and measure forward returns. The confirm-late pattern shows up almost every time. Run it yourself before allocating anything to “analyst sentiment” as a signal class.
The real takeaway from Monday’s list isn’t any single call — it’s that a roundup of ten tickers with no stated methodology, no hit-rate disclosure, and no risk framing is marketing collateral, not research. Treat it accordingly.