WB Trading Labs / Research / Analyst ratings
I tested 295,000 rating and price-target changes on 1,307 US companies since 2012. An upgrade moves a stock +1.7% against the market on the day it's published, a downgrade −1.9%. Over the next 45 sessions there's nothing more, within the margin of error. And over a year, the median company with the most promised upside from price targets did about as well as the one with the least (+13% against +11%, within the margin of error).
How this study happened
In early October @PawRze posted analyst ratings he had collected for 240 companies. I wanted to see what the same question looks like on much more data, so I tested 295,000 rating and price-target changes. The thread went out on X on 4 October, in Polish.
Section 5, on less liquid stocks, exists because Paweł asked about them in the discussion under the thread.
Wojciech, October 2026
upgrade, return vs the market on the day
downgrade, on the day
further move over 45 sessions, within the margin of error
of companies are at or above the price target a year later: least vs most promised upside
| Claim tested | Finding | Verdict |
|---|---|---|
| An upgrade keeps paying after the day | +1.7% on the day, then zero over 45 sessions | No lasting edge |
| More promised upside means more return | Median promise from −4% to +39%; median 12-month return +11% to +13% in every fifth | Does not hold |
| Some firms' calls are worth more | A firm's past accuracy does not predict its next call | Does not hold |
| Less liquid stocks react more | About 30% stronger on the day; gone afterwards | Holds, on the day |
1 · The day of the note
A firm's rating change clearly moves the stock on the day it comes out. But anyone who bought at that day's close got nothing more: over the next 45 sessions the average return against the market stayed flat, and the individual cases spread evenly around zero.
The "after the day" test starts at the close of the first session after the note and compares the stock with other stocks in the same price state (same year, similar return over the last month and year, similar distance from the 52-week high). Upgrades, downgrades, target raises and cuts, initiations and the number of firms changing their mind: all within the margin of error. The study does not test entering before the open or during the session.
2 · Twelve months later
I rebuilt the consensus price target every month from 2015 to 2025, with at least 3 analyst firms per company, and split the companies into fifths by promised upside - the gap between the price and the target. The median promise goes from −4% to +39%. The median company in every fifth returned +11% to +13% over the next year. A year later, 68% of the companies with the least promised upside were at or above their target, against 29% of those with the most.
The most "promising" fifth does have a higher average (+36.3% against +16% to +20% in the others), but a handful of rockets carry it, mostly from 2021-2025. Take out the top 1% of outcomes and the gap shrinks to about 4 percentage points and stops being statistically significant. And the data misses companies that went bust along the way, which flatters exactly this fifth.
3 · Why
Before a big price-target cut, the stock had on average been at the 29th percentile of the market; before a big raise, at the 73rd. The bigger the change in target, the harder the price had moved first. So a high "upside" is most often a price that fell and targets that haven't caught up yet: promised upside correlates negatively with the previous 12 months' return (−0.40).
4 · Does it depend on who writes it?
On the day, firms differ a little: notes from Morgan Stanley, UBS, Jefferies and BofA move the price more, Citi and Wells Fargo less. After that day no group of firms - big banks, mid-tier or boutiques - has an edge. I also checked whether a firm's past accuracy, using only earlier calls that had already played out, predicts its next one. The best third of firms did no better than the weakest third, over weeks or over a year.
5 · Less liquid stocks
Paweł asked under the thread whether less liquid stocks react more. On the day, yes, and not just because they're more volatile: measured in each stock's own daily volatility, the reaction is about 30% stronger. It holds in both 2012-2020 and 2021-2026.
| least liquid third | middle third | most liquid third | |
|---|---|---|---|
| Upgrade | +2.1% | +1.5% | +1.5% |
| Downgrade | −2.4% | −1.8% | −1.7% |
After the day the edge is gone: over the next 45 sessions, net of the usual drift of similarly liquid stocks, +0.8 percentage points remain for upgrades and about 0 for downgrades, which is noise. Important: the sample is the 1,500 most liquid US companies, so "less liquid" still means about $50 million traded a day. Micro caps may behave differently, and this data can't tell.
Method and caveats
Correction, 11 October 2026: the first version compared price targets with share prices adjusted afterwards for later splits and dividends, not the prices analysts saw, and judged the target on the return with dividends. That overstated the promised upside of the most promising fifth (+105% on average; correctly +50%, median +39%) and understated how many companies were at or above the target a year later (least and most promising fifth: 62% and 15%; correctly 68% and 29%). Fig. 3 now compares median with median (before, it set the average promise against the median return). The finding stands: a bigger promise didn't bring a bigger median return. The results on the day a rating is published were not affected.
I post new studies on X first.
Follow @WojciechBernPL