Monday’s Analyst Calls: Nine Tickers, Zero Alpha—Screening the Signal From the Noise

Sell-side research calls are, statistically speaking, close to noise. The empirical literature — Womack, Barber et al., and the more recent machine-learning replications — puts post-upgrade drift in the 50–200 bps range over months, which gets largely eaten by transaction costs and slippage unless you’re routing intelligently. So when Monday’s headlines tout analyst moves on Block, Boise Cascade, Equifax, FirstEnergy, Lockheed Martin, Lumentum, NetApp, Roper, and Spotify, the right response isn’t to read the calls. It’s to backtest them.

The workflow is straightforward in Python. Pull historical rating and price-target changes from a data provider (IBES-style datasets via Refinitiv, or Nasdaq Data Link for retail-grade feeds), join against adjusted OHLCV, and compute event-study abnormal returns using a simple market-model benchmark. A hundred lines with `pandas`, `statsmodels`, and `yfinance` will tell you whether a given analyst community’s upgrades carry any alpha at all — or whether they’re momentum-chasing dressed up in a DCF.

A few things worth measuring before treating any of these calls as signal:

**Revisions vs. levels.** Price-target changes cluster with recent price action, meaning most “upgrades” are just analysts extrapolating trailing returns. Regress target revisions on the prior 60-day stock return; if the R² is high, the call contains little incremental information beyond what your momentum factor already captures.

**Analyst persistence.** Group calls by analyst or firm and check rank information coefficients over time. Some houses are consistently early; most are consistently late. A rolling IC series sorted by firm separates the two — and the persistence, where it exists, is weak enough that you want a decay-weighted blend, not a binary follow.

**Sector mechanics.** Defense names like Lockheed move on contract flow and budget cycles, not rating changes — the actionable signal there is order announcements and appropriation timing, both scrapable and backtestable. Utilities like FirstEnergy trade on rate-case calendars and rate spreads versus the 10-year; an analyst upgrade is a rounding error next to a regime shift in real yields. Boise Cascade is a housing-starts and lumber-futures derivative wearing an equity ticker — check the correlation against `LBLQ` futures and building permits before caring what anyone at a bank thinks.

**AI-adjacent calls deserve extra skepticism.** Anything touching Lumentum or networking optics gets priced on narrative enthusiasm with fat left tails. If you’re trading those, size positions off implied volatility and realized skew, not target prices.

If you insist on trading the calls anyway: build a factor-neutral event portfolio — long upgraded names, short downgraded, beta- and sector-hedged — hold 20 to 60 days, and let the out-of-sample data decide. My prior, informed by every replication I’ve run: the drift exists, it’s small, and it dies at retail commission-free scale. The analysts’ contribution to your P&L is best treated as another input feature, not a thesis.

The genuinely systematic takeaway from a day like Monday isn’t any individual call — it’s that consensus revision breadth across a ticker is a measurably better predictor than any single house’s conviction. Track the aggregate, ignore the headline.