What Is the Pass Rate for the Enrolled Agent Exam?

For the most recently completed testing window — May 2024 through February 2025 — the IRS Special Enrollment Examination (SEE) part-level pass rates were:

That puts the simple three-part average around 66%. Those numbers are derived from the Prometric "Results Reported for Special Enrollment Examinations" disclosure for that testing window and are independently reported by both Gleim and Becker.

If you read nothing else, read the next section.

As of May 14, 2026, there is no such thing as a published 2026–2027 EA exam pass rate. Any page telling you otherwise is either:

The reason is simple. On March 1, 2026, the IRS transitioned the SEE from Prometric to PSI Services. The 2026 testing window for domestic candidates opens July 1, 2026, and for international candidates September 1, 2026. No PSI-era testing window has closed yet, so no PSI-era pass rate exists yet. The current passing score remains 500 on a 200–800 scaled scoring system.

When you see "2026 EA exam pass rate" in a search result this summer, check the date. If the page does not explicitly say where the number came from and which testing window it covers, treat it as marketing copy, not data.

This is the strongest year-by-year table publicly available. The figures are reported by Gleim and corroborated by Becker for 2024–2025; both cite Prometric as the underlying source.

The shape of this five-year record matters more than any single number:

Most published commentary on the EA exam still describes Part 2 as "the hard one" and Part 3 as "the easiest." Both characterizations are out of date.

A longer year-by-year table circulates online going back to 2015–2016, originally compiled by iPass EA Exam. It is not an official Prometric or IRS table, and it has small disagreements with the more recent Gleim figures in overlapping years (typically one or two percentage points). For directional historical context only:

The single most striking long-run pattern is the collapse of Part 3's pass-rate premium. In 2015–2016, the gap between Part 2 and Part 3 was 31 points. In 2024–2025, it was negative 1 point — Part 3 was actually slightly below Part 2. Whatever made Part 3 historically easy has been substantially calibrated away.

For anything earlier than 2015–2016, public year-by-year data is unreliable. And for anything before October 2006, comparisons are not meaningful at all: the SEE moved from an IRS-administered, paper-based format to Prometric's on-demand, three-part computerized format in October 2006. Pre-2006 numbers are a different exam.

A part-level pass rate is the percentage of exam sittings on a given part that resulted in a passing score during the testing window. It is not the percentage of people who eventually become Enrolled Agents. Three structural features make the headline number easy to misread:

Sittings, not people. A candidate who fails Part 1 twice and passes on the third attempt contributes two failures and one pass — three data points, one person. Each part can be attempted up to four times per testing window. Heavy retake activity pulls the pass rate down even when most candidates eventually succeed.

Different candidate pools by part. The parts are not taken by the same population. Part 1 has roughly double the candidate volume of Parts 2 and 3, because most candidates start there. That means Part 1 is disproportionately populated by first-time SEE candidates, including casual and underprepared ones. Parts 2 and 3 have more survivorship bias — by the time someone sits for Part 2 or 3, they have usually already passed something else and figured out how the exam works.

Sequencing effects. Candidates can take the parts in any order. The "easy first" candidate who passes Part 3, then Part 1, then Part 2 contributes to each part's pass rate at a different point in their preparation arc than the "linear" candidate who takes Parts 1, 2, and 3 in order. Both Gleim and Becker explicitly caution that part-to-part comparisons are confounded by candidate composition.

The practical implication: a 58% pass rate on Part 1 does not mean Part 1 is harder than Part 3 at 70%. It may partly mean that Part 1 is harder. It may also mean Part 1 is where the underprepared candidates show up first.

A reasonable intuition is that the SEE is one exam program, updated together on one annual cycle (the IRS pulls the exam down each March-April to update for the most recent tax-law cutoff), so when one part gets harder, the others should too. Candidate-pool variation would be noise around a common signal.

The data supports a weak version of that intuition and rejects the strong version.

Looking at the five most recent windows (2020–2021 through 2024–2025), the year-to-year changes in pass rates are highly correlated between Parts 1 and 2 (0.98) and moderately correlated between Parts 1 and 3 (0.53) and Parts 2 and 3 (0.67). When something moves the exam environment — annual rewrite difficulty, tax law turnover, candidate mix — it tends to move all three parts in the same direction. The 2021–2022 window is the clearest example: all three parts dropped sharply.

But the levels are a different story. The gap between Parts 1 and 3 was +16 points in 2020–2021 and +12 points in 2024–2025. The gap between Parts 2 and 3 was +10 points in 2020–2021 and negative 1 point in 2024–2025. These are not constant relationships. The historical pass-rate "personality" of each part has shifted materially in just five years.

The honest summary: the three parts share a common annual factor, but candidate-pool effects and part-specific calibration are large enough that you cannot reliably use one part's pass rate as a proxy for another part's difficulty.

This is also a small dataset — five annual observations — so the exact correlation numbers should not be taken too seriously.

You are about to read a lot of pages on this topic. Most of them are commercial pages from EA prep providers, and quality varies sharply.

If a page quotes you a pass rate without naming the testing window it covers, the page is not a usable source.

A few practical takeaways:

The current realistic expectation is roughly two-thirds. Across the three parts and the past five years, the simple average has run between 63% and 73%. Plan around the assumption that any given attempt is more likely to pass than fail, but not overwhelmingly so.

Do not let pass-rate folklore drive your study order. The advice you will read everywhere — "take Part 3 first, it's the easiest" — is built on data that is now five-plus years out of date. Part 3 has been the same difficulty as Part 2 for two years running.

Part 1 is not necessarily the hardest exam. It may be the hardest exam to walk into underprepared. Its low pass rate reflects, among other things, that it is the part most often attempted by candidates who have not yet learned how the SEE actually works.

Watch for 2026 disclosures in spring 2027. The first full PSI-era testing window closes February 28, 2027. The first credible PSI-era pass-rate data will not be available until after that. Anything called "2026 pass rates" before then is either 2024–2025 data with a refreshed headline or it is invented.

The EA Notebook is a free weekly newsletter on the Enrolled Agent exam — tax law updates, exam mechanics, and study notes. Subscribe here.