Analysts

Good-Bye I/B/E/S (or Not?)

Journal of Financial Reporting 2023, 8(1): 41 to 61

Five questions. Flip each card to see what the paper finds.

If you use analyst forecasts, this changes what you can safely infer from non-U.S. samples.

Problem Non-U.S. forecasts carry the risk U.S. samples are mostly intact. International samples need care.
Reshuffling Only 2.87% reshuffled The feared 30.7% reassignment problem does not show up in the matched forecast records.
Fix About 55.7% can be recovered A separate file makes many anonymized IDs recoverable.
Published version Request access Stata tutorial Preprint 2023 update 2024 update 2025 update Start the five-question quiz

Test yourself

Predict first. Then flip.

Each card asks for your guess before showing the answer. Click to flip. Or reveal all at once.

Q01

What share of analyst forecasts actually had reshuffled IDs?

click to flip
A01
2.87% 99% of those were previously anonymized forecasts being restored in later database versions, not real reshuffling.
click to flip back
Q02

Anonymization was announced October 2018. When did it actually start?

click to flip
A02
Jan 2017 At least a full year earlier than disclosed. The structural break is unmistakable in the data.
click to flip back
Q03

At its worst, what share of non-U.S. forecasts are anonymized?

click to flip
A03
21.06% One in five annual NON-US EPS forecasts by August 2021. That is up from roughly 3% before it all began.
click to flip back
Q04

Is U.S. analyst data still usable for research?

click to flip
A04
Yes Fewer than 0.2% of annual U.S. EPS forecasts since 1990 have anonymized IDs. The problem is almost entirely non-U.S.
click to flip back
Q05

Can the lost analyst identities be recovered?

click to flip
A05
55.7% By cross-referencing a separate file from the same data source that escaped the anonymization.
click to flip back

All figures from Law (2023, Journal of Financial Reporting), Sections II to IV.

↓

That 55.7% from Q05 shows what recovery looks like.

10 of 18 recovered (55.6%) 8 still anonymous

Synthetic demo. Tiles are illustrative. 18 cards stand in for the population, with 10 flipping (55.6%) to approximate the empirical 55.7% recovery rate from Law (2023, JFR), Section IV. Labels and identifiers shown below are not real records.

See it happen

Some flip. Some stay anonymous.

Each orange card is one anonymized analyst forecast. Press Recover and they flip one at a time, like an airport departure board, revealing the identities the method could rescue.

18anonymized 0recovered

On average, the method recovers 55.7% of forecasts with anonymized analyst IDs.

Request access to Stata code

Peak annual EPS anonymization

The split is U.S. versus non-U.S.

0.96%
U.S.firms
Annual EPS forecasts at the August 2021 peak
22×
higher for non-U.S.
21.06%
Non-U.S.firms
Annual EPS forecasts at the August 2021 peak

At the peak, annual non-U.S. EPS forecasts were about 22 times more likely to have anonymized analyst IDs than annual U.S. EPS forecasts.

Annual EPS forecasts with anonymized analyst IDs

Source: Table 2, columns (1) U.S. and (2) Non-U.S. Vertical guides mark January 2017 (real start) and October 2018 (claimed start).

0% 5% 10% 15% 20% 2015 2016 2017 2018 2019 2020 Aug ’21 Jan 2017 · real start Oct 2018 · claimed 21.06% 0.96%
U.S. firms Non-U.S. firms

Key takeaways

The concern

In October 2018, a product change notice announced that the identifications (IDs) of 88 brokers and their analysts had been anonymized in the Detail History File. A subsequent claim that 13.8% of broker IDs and 30.7% of analyst IDs had been reassigned raised the prospect that decades of analyst research, which depends on tracking individual analysts over time, could no longer be reliably conducted.

What I found

The ID anonymization predominantly affects forecasts of non-U.S. firms. U.S. firms are largely untouched (less than 0.2% of annual U.S. forecasts since 1990, per Table 2). The change began in January 2017, well before the announced October 2018 effective date, and extends beyond the 88 brokers in the Product Change Notification.

There is no evidence of the large-scale ID reshuffling implied by the 30.7% figure. Across the 2015 and 2021 Detail History Files, only 2.87% of forecasts have reshuffled IDs, and roughly 99% of those were forecasts that had been anonymized to begin with. For U.S. firms specifically, the reshuffling rate is 1.20% (annual EPS) and 0.57% (quarterly EPS).

How to recover anonymized IDs

Because the anonymization had very little effect on the Recommendations Detail File, anonymized analyst IDs in the Unadjusted Detail History File can be reverse-engineered by cross-referencing the non-anonymized analyst IDs in the Recommendations file. On average, this approach recovers about 55.7% of forecasts with anonymized analyst IDs.

For Stata code and a broker linking table, request access here. For a related guide to matching I/B/E/S vintages, see the Stata matching tutorial.

Practical resources

This analysis is updated annually. The most recent update (2025) is available via the links at the top of the page.

← Back to all research