How Quantitative Is Each Finance Exam? % Computational Across 40 Credentials (2026 Data)

The most quantitative finance exam is SOA Exam P at 90.2% pure-numeric multiple choice, with SOA Exam FM right behind at 88.5%. The least quantitative is CPA ISC at 0.2%, followed by CPA AUD at 0.4% and Series 63 at 0.8%. Across 40 finance credentials, the FreeFellow team measured the % of multiple-choice questions where all answer choices are pure numeric (currency, percent, decimal). This is the only cross-credential dataset of question-format computational density we are aware of.

The most common question I get from candidates picking a finance credential is some version of "how math-heavy is this exam." The honest answer varies a lot by credential, and even within the same family (CFA L1 vs L2, FRM Part 1 vs Part 2) it varies more than people expect.

So I measured it directly. Across the FreeFellow question bank (41,799 multiple-choice questions covering 40 finance credentials), I classified each question by whether every answer choice is a pure numeric value (currency, percent, decimal, scientific notation, with optional LaTeX wrapping). That fraction is a clean proxy for "this question requires computational reasoning to answer," because you cannot pick a numeric answer by reading prose alone. You have to derive a number.

It is not a perfect proxy. A few caveats sit in the methodology section below. But it gives you a more honest picture than the exam outlines, which tend to describe content categories without telling you how the questions are actually shaped.

For a job-mapping and family-comparison view, see the Finance Credentials Map cornerstone post. This page is the data view.

The Full Table (40 credentials, sorted by % quantitative)

Investment & Risk Family

This is where the heaviest variance lives. The actuarial exams dominate the top of the list. SOA Exam P is 90.2% quantitative, meaning roughly nine out of every ten multiple-choice questions ask you to compute a numeric answer, with Exam FM right behind at 88.5%. That tracks with what every actuarial candidate will tell you: the preliminary path is calculus, probability, and financial mathematics applied at speed under exam conditions.

CFA shows a clean L1 to L2 step-up. Level I is 30.9% quantitative; Level II jumps to 38.7%. The reason is structural. L2 introduces the item-set vignette format, which loads up derivation-heavy equity valuation, fixed income, and quantitative methods questions. L3 falls back to 23.6% in the core because the pathway-specific sections (PM, PW, PMKT) emphasize judgment, frameworks, and constructed-response essays over computation.

FRM Part 1 (52.1%) vs Part 2 (41.6%) tracks the exam's deliberate structure: Part 1 covers quantitative methods, foundations of risk, financial markets, and valuation models. Part 2 shifts to market risk, credit risk, operational risk, liquidity risk, and Basel regulation, where the questions are more about identifying the right framework or regulatory interpretation than running a calculation.

CAIA L1 vs L2 has a similar fall-off (21.1% to 12.2%). L1 covers asset classes with valuation mechanics for hedge funds, private equity, real assets, structured products. L2 leans heavily on portfolio construction frameworks, ethics, due diligence, which read more like CFA L3 in shape.

SOA ALTAM and ASTAM look low (21.3% and 7.8%), but the number is misleading: these are constructed-response exams. Most of their scored content is multi-part written-answer questions where you derive a numeric answer in narrative form. The multiple-choice portion is small. The actual content is heavily computational. The same caveat applies to the three constructed-response exams new to this snapshot: CAS Exam 5 (62.3%), CAS Exam 6U (33.2%), and SOA Exam PA (21.8%) are measured here by their multiple-choice concept-check banks, while the graded exam work is written-answer.

CAS MAS-II (75.2%) is the biggest mover since the first snapshot of this dataset. CAS publishes MAS-II answer choices in interval-bucket format (e.g., "At least 0.40, but less than 0.50"). Earlier versions of the classifier read those buckets as prose and reported MAS-II near zero; the classifier now parses bucket choices as numeric, and MAS-II lands third overall, which matches its computational reputation.

Accounting & Tax Family

This family has the biggest surprises.

CMA Part 2 (43.8%) is the most computational accounting exam, and CMA Part 1 measures 28.1%. The IMA's CMA exam looks similar to CPA on paper, but the question shape is closer to FRM. The bigger surprise is that CPA FAR (37.0%) now sits between the two CMA parts: the "CPA is all rules, no math" shorthand does not survive contact with FAR's computation density.

EA Part 1 (40.3%) and Part 2 (39.8%) vs Part 3 (13.4%) tracks the IRS exam structure. Part 1 (individual taxation) and Part 2 (business taxation) are filled with "calculate the basis," "compute the deduction," "determine the gain or loss" prompts with pure-numeric choices. Part 3 (representation, practices, procedures) is procedural and ethical content where the choices are usually prose.

CPA REG (18.0%) and CPA TCP (14.0%) still measure below their computational feel. CPA tax questions often present a fact pattern with one or two numeric inputs but the answer choices are mixed-format: things like "$15,000 LTCG taxed at 15%" or "Basis of $80,000 after the section 754 adjustment." The reasoning is computational but the answer format is prose-with-numbers, not pure-numeric, so the classifier reads it as conceptual.

CPA AUD (0.4%) and CPA ISC (0.2%) are genuinely conceptual. Audit is about evidence, judgment, control assessment, and reporting standards. ISC is about IT general controls, SOC reports, and access controls. Both test pattern recognition against frameworks, not arithmetic.

CPA FAR (37.0%) is the calculation section: compute goodwill, the lease liability, the deferred tax asset, the consolidation elimination. BAR (12.4%) is far more conceptual (managerial accounting frameworks, business analysis), which makes the FAR-to-BAR gap the widest within any single credential's sections.

Wealth & Securities Family

The lightest family in terms of arithmetic.

CFP (18.0%) is lighter than any CFA level but not the near-zero exam it is sometimes assumed to be. Most CFP questions are scenario-based fact patterns where you select the most appropriate planning recommendation, and those choices are prose ("convert all to Roth," "leave as is"). But a real slice of the exam computes: education funding amounts, insurance needs, retirement withdrawals, and tax figures with bare-numeric choices.

Series 79 (38.5%) is the most computational FINRA exam: the Investment Banking Representative exam runs on valuation multiples, merger math, and offering economics. Series 7 (22.5%) is next, with options P/L, margin calculations, accrued interest, and sales-charge math. Real arithmetic, just at the scale of single-question application rather than CFA-level derivation.

Series 65 (16.2%) and 66 (14.5%) are middle-of-the-road. Investment Adviser Representative exams have some quant content (Sharpe, alpha, beta interpretation) but most of the exam is regulatory rules under the Uniform Securities Act and the Investment Advisers Act.

Series 63 (0.8%), Series 6 (7.0%), and SIE (8.9%) are the lightest. Series 63 is state Blue Sky law, almost entirely rules. Series 6 and the SIE are broadly conceptual with a handful of sales-charge and margin questions.

Five Most Surprising Numbers

1. SOA Exam P is 90.2% quantitative. Nine of every ten questions end in a number. No other credential family starts its pipeline this computationally.
2. CPA FAR (37.0%) out-computes CMA Part 1 (28.1%) and closes on FRM Part 2 (41.6%). FAR is the section that breaks the "CPA has no math" assumption.
3. Series 79 (38.5%) is the most computational FINRA exam, well above Series 7 (22.5%). Investment-banking valuation math outweighs brokerage arithmetic.
4. CPA AUD is 0.4% quant despite being a CPA section. Candidates who choose AUD as their easiest section often do so because there is no math.
5. EA Part 3 (13.4%) is dramatically less quant than Parts 1 and 2 (both about 40%). The IRS representation exam is procedural, not computational.

What This Means for Picking a Credential

If you do not enjoy multi-step arithmetic and probability manipulation, the SOA/CAS preliminary path will be a painful slog. If you prefer reading rules and applying them to fact patterns, CPA AUD, CPA ISC, Series 63, and Series 6 are where you live. Most of finance lives in the middle: enough arithmetic to require comfort with formulas, but not so much that the exam is a calculation marathon.

If you are comparing two specific credentials, the Finance Credentials Map cornerstone post has the family-to-job mapping plus cost data. For exam-pair deep-dives, the CFA vs CFP vs CPA comparison post is a good starting point.

Methodology

The classifier checks each multiple-choice question's answer-choice array. A question is counted as "all-numeric" if every choice in the array parses as a pure numeric value after stripping optional LaTeX delimiters, currency symbols, percent signs, thousands separators, and whitespace. The classifier accepts integers, decimals, scientific notation, signed and parenthesized-negative values, magnitude suffixes (%, bps, million, billion), and the CAS interval-bucket format ("At least 0.40, but less than 0.50"); it rejects anything with embedded prose.

What the classifier does NOT count as quantitative:

• Mixed-format choices like $15,000 LTCG taxed at 15%. Pure-numeric requires the answer to be ONLY a number after stripping symbols. The 15% is a number, the prose label disqualifies it.
• Constructed-response questions (TBS on CPA, ALTAM/ASTAM on SOA, essays on CFA L3, etc.). These have no answer-choice array.
• Choice formats with embedded units like 4.5 years or $X per share. Same rule: any non-numeric character disqualifies.

What this means: the % quantitative number is a floor on actual computational density. The true rate is higher in cases where the reasoning is computational but the published choice format includes prose (the CPA tax sections are the clearest case). The number is most accurate for exams that publish bare-numeric answer choices (SOA preliminary, FRM, CFA quantitative methods sections).

Question-bank totals are FreeFellow originals only. We exclude verbatim reproductions of CFA Institute end-of-reading items and SOA sample questions from the count because the question-format profile of third-party content can differ from our original generation.

Data snapshot: 2026-08-01. Earlier snapshots (through 2026-05-21) predate the classifier's interval-bucket and parenthesized-negative parsing, which is why previously published figures for CAS MAS-II and the calculation-heavy CPA sections were materially lower; bank counts also moved with the CFA CY2027 curriculum cutover and routine bank growth. The classifier is deterministic, rule-based numeric parsing with no ML judgment calls, and the parsing rules listed above are the complete specification.

How to Pick a Credential Based on Your Tolerance for Arithmetic

Step 1: Honestly assess your relationship with multi-step calculation

Spend an hour working through 5-10 questions from an exam you are considering. Track how you feel after each one. Did you enjoy the derivation? Were you frustrated by the arithmetic? Were you bored by the lack of conceptual reasoning? Your gut reaction at the end of an hour is a much better signal than any aspirational self-image.

Step 2: Match your tolerance to the right band

If you genuinely enjoy multi-step quantitative work, target the 40-90% quantitative band: the actuarial exams (P, FM, FAM, MAS-I, MAS-II, CAS Exam 5), FRM (both parts), CMA Part 2, and EA Parts 1 and 2.

If you prefer a moderate mix of computation and concept, target the 15-40% band: CFA (any level), CPA FAR, Series 79, Series 7, CMA Part 1, CAIA Level I, SOA SRM, CFP, and CPA REG.

If you prefer reading rules, applying frameworks, and reasoning through fact patterns, target the under-15% band: CPA AUD, CPA ISC, CPA TCP, CPA BAR, CAIA Level II, EA Part 3, Series 63, Series 6, Series 65, Series 66, SIE.

Step 3: Cross-check your tolerance against your target job

A common mistake is picking a credential because it sounds easier and then ending up in a job that requires the underlying quantitative skill anyway. If you want quantitative risk at a bank, you cannot avoid the math by picking CFP instead of FRM; you will fail the actual job interview. Match the credential to the job FIRST, then evaluate whether you can tolerate the underlying quant load.

Step 4: Plan study time proportional to the quant density

Computational exams (the 40-85% band) need more practice hours per topic than conceptual exams, because you are building procedural fluency, not just recognition. Budget at least 1.5x the typical study-time estimate for these exams if you are not already comfortable with the underlying math. Conceptual exams reward reading and re-reading; computational exams reward repeated problem-solving under exam conditions.

Frequently Asked Questions

Q: Is the CFA quantitative?

Moderately. CFA Level I is 30.9% pure-numeric MCQs, Level II is 38.7% (vignette derivation), Level III drops back to 23.6% in the core (with constructed-response essays in pathway sections). Comfort with derivative pricing, fixed-income duration, equity valuation, and statistics is required. Calculus is not.

Q: Is the CPA quantitative?

Less than its reputation suggests, with one big exception. FAR measures 37.0% pure-numeric, real calculation density. The other sections test rules application against fact patterns: AUD is 0.4%, ISC is 0.2%, and the tax sections read as prose-with-numbers.

Q: How quantitative is the FRM?

Part 1 is 52.1% pure-numeric (quantitative methods, valuation models, financial markets). Part 2 drops to 41.6% (regulatory and qualitative risk). Part 1 is the heavier math lift; Part 2 is the heavier reading lift.

Q: Which finance exam is most quantitative?

SOA Exam P at 90.2% pure-numeric multiple-choice questions. SOA Exam FM (88.5%) and CAS MAS-II (75.2%) are the next two. All three are actuarial exams.

Q: Which finance exam is least quantitative?

CPA ISC at 0.2% pure-numeric, CPA AUD at 0.4%, Series 63 at 0.8%. Among the wealth-management credentials, Series 6 (7.0%) and the SIE (8.9%) are the lightest; CFP measures 18.0%.

Q: Why did CAS MAS-II jump from 0.2% to 75.2% between snapshots?

CAS publishes MAS-II answer choices in interval-bucket format ("At least X, but less than Y"). The classifier used for the 2026-05-21 snapshot read those buckets as prose, so MAS-II reported near zero. Bucket parsing has since been added, and MAS-II now measures 75.2%, third overall, which matches its computational reputation.

Q: How does this compare to published exam blueprints?

The SOA and CFA Institute publish topic weights but not question-format breakdowns. AICPA's CPA Blueprints describe skill levels (Remembering & Understanding, Application, Analysis, Evaluation) which loosely correlate with computation but don't quantify it. This data is the first cross-credential measurement of question-format computational density we are aware of.

Practice Free on FreeFellow

Every exam in this table has a free question bank with practice questions, condensed outlines, and formula sheets. The CFA materials are validated as a CFA Institute Prep Provider.

For the family-to-job-mapping view, the Finance Credentials Map cornerstone has the job-by-credential breakdown, cost data, and study-time estimates.