CFA Level III essay and item set format

Two formats, both tied to a vignette

Level III has two kinds of question set. An item set starts with a vignette and asks four multiple-choice questions connected to the case. A constructed-response set also begins with a vignette, then asks you to type written responses. Both formats test whether you can use evidence in the case, not simply recall a definition.

There are 11 sets of each type, 22 total. One item set and one essay set are trial content and do not count toward the score. Trial sets are not identified, so treat every set as live. Both sessions last 132 minutes. Each contains either six item sets and five essays or five item sets and six essays. Each set is worth 12 points, making 12 minutes per set a useful average pacing reference.

Read the vignette with a purpose

A vignette may describe a client, portfolio, institution, manager, or market conditions. First identify the decision maker and the task. Then mark facts that define the decision: objective, constraints, horizon, benchmark, tax or liquidity needs, and relevant return or risk data. Some details will be irrelevant to one question and essential to another.

For an item set, reading each question stem before studying every paragraph can focus attention. Return to the case to locate the needed facts, then solve only what is asked. One item might use bond values; another might ask which risk measure matches the client's concern. Do not carry a calculation assumption into a later item unless the case says it applies.

Watch measurement basis. Questions can distinguish arithmetic from geometric return, gross from net, beginning from ending values, active from total risk, or local from base currency. A number may be calculated accurately but still answer the wrong question if it belongs to another time period or return definition. The client's objective and policy constraints often matter more than a generic claim about which asset class has higher expected return.

Multiple-choice item sets

Each item set contains four multiple-choice questions worth three points each. The questions share a vignette, but answer each independently. Use the facts supplied, not outside assumptions. If transaction costs are excluded, do not introduce them. If the case names a liability benchmark, do not substitute a familiar market index just because it is easier to calculate.

A disciplined sequence is: identify the requested output; find the case input; select the method; check sign, units, and timing; then compare with the choices. Distractors commonly reflect a recognizable mistake, such as using the wrong denominator, ignoring hedge direction, or treating relative return as absolute return. In review, diagnose the mistake rather than memorizing a correct letter.

Original item-set example: liability duration

A pension plan has assets worth $480 million and liabilities worth $500 million. Liability duration is 12.0. It considers adding bonds with duration 8.0, replacing cash, with no change in total asset value. Assume parallel small yield changes and ignore convexity. If the full asset portfolio had duration 8.0, its dollar duration would be $480m times 8.0, or $3.84bn-years. Liability dollar duration is $500m times 12.0, or $6.00bn-years. The negative $2.16bn-years gap means liabilities are more rate-sensitive. A yield decline would tend to increase liability value more than asset value.

The distractor “duration 8 is longer than the liability duration” reverses the comparison: 8 is less than 12. “The funded ratio is 96%, so rates have no effect” confuses funding level with interest-rate sensitivity. “A fall in yields lowers liabilities” reverses the bond-price relationship. In a full portfolio, cash flows, convexity, and curve changes matter, but the prompt excludes them.

Constructed-response sets

Command words determine the response. Determine asks for a choice or result. Justify asks why the choice fits. Explain asks for a cause or implication. Discuss calls for a supported account. If a prompt says determine and justify, provide both parts. If it requests two reasons, make two distinct points, preferably in separate bullets or paragraphs.

The answer box accepts typed words and numbers. Bullets are acceptable. When multiple responses are requested, they are evaluated up to the number asked for, in the order you give them. Extra choices are not evaluated. Lead with your best answer instead of listing alternatives and hoping one receives credit.

For a calculation, a correct number alone can receive full credit. Showing a correct formula and substitution may give a chance for partial credit if the final arithmetic is wrong. Include units and signs when they clarify the result. A long explanation is not automatically stronger than the requested value.

Original essay example: net active return

A balanced portfolio returned 7.2% before fees and 6.8% after fees. Its policy benchmark returned 6.0%. The prompt asks whether the portfolio added value after fees and requests net active return in percentage points. A direct response is: “Yes. Net active return was 0.8 percentage points: 6.8% minus 6.0%.” Use after-fee return because that is what the question asks. Gross active return is 1.2 points, but is not the requested measure. Reversing the subtraction gives the wrong sign; using 7.2% ignores fees.

If a separate question asks for fee drag, calculate 7.2% minus 6.8%, or 0.4 percentage points. Fee drag is not active return. Naming the measure before calculating prevents applying correct arithmetic to the wrong inputs.

Original essay example: recommendation

A taxable investor expects to fund a home purchase in three years. Strategy A has a higher long-term expected return but substantial interim volatility and limited liquidity. Strategy B has a lower expected return, shorter-duration bonds, and liquid holdings sized for the purchase. Asked to select the more appropriate strategy and justify it, a concise response is: “Strategy B is more appropriate because its liquidity and shorter duration better support the known three-year cash need.” A generic claim about diversification does not explain the fit. Choosing A solely for its long-term return ignores the spending date.

Pacing a 132-minute session

Each session has 11 sets, so the average is 12 minutes per set. Use that number to notice drift, not as a rigid time limit. A calculation-heavy case may take longer; a short constructed response may take less. Avoid spending 25 minutes polishing one answer while leaving later items blank. If stuck, give the best supported response, move forward, and return if time remains.

Timed practice shows where time goes: locating facts, choosing a method, calculating, or drafting. If essays are slow, practice one-sentence conclusions with explicit case evidence. If item sets take too long, read stems first and extract only relevant vignette inputs. Practice a full mixed session to build stamina because format proportions vary between the two sessions.

Common avoidable errors

• Giving a general definition instead of the recommendation the prompt requests.
• Providing a selection but omitting its justification, or giving fewer reasons than requested.
• Writing more alternatives than requested; excess responses are not evaluated.
• Using the wrong return basis, time period, benchmark, currency direction, or denominator.
• Treating every vignette fact as relevant to every linked question.
• Leaving an item blank despite no penalty for an incorrect answer.
• Assuming a topic or response format appears only in one session.

How to learn from practice

For item sets, record which vignette facts supported each answer and classify every miss as knowledge, fact selection, calculation, or reading. For essays, compare your response with the command and the number of requested parts. A correct insight can still miss the task if it does not state the requested determination or explain the case implication.

Vary practice cases rather than memorizing keyed language. Change the time horizon, reverse the foreign-currency exposure, or use a different benchmark, then ask whether the recommendation should change. Use a familiar calculator and check units before interpreting the output. Practice percentages versus percentage points and local versus base currency explicitly.

The 12 points assigned to a set do not create a raw percentage that predicts the official result. CFA Institute sets a Minimum Passing Score and reports a scaled score relative to it. Practice percentages help locate weak areas and pacing problems; they are not an official pass conversion.

Answer every part in a clear order

A constructed-response prompt can contain a decision and a supporting task. “Determine the most appropriate strategy and justify your selection” calls for both the selected strategy and a reason tied to case evidence. “Explain two reasons” calls for two different points, not one point expressed twice. Count the requested outputs before typing. A complete short answer can be stronger than a long answer that obscures the decision.

Separate responses where it helps the reader follow them. If asked for one weakness for each of two firms, label Firm A and Firm B and make each answer distinct. Do not add a third weakness when only two are requested; excess answers are not evaluated. Put your strongest points first because the response order matters when the maximum number of evaluated answers is limited.

When calculation work helps

The answer box can accept a numerical result without a written derivation, and a correct value alone can receive full credit. Showing a valid setup can help when there is a chance to receive partial credit after an arithmetic slip. Keep each intermediate figure labeled. For example, distinguish market value, duration, and dollar duration rather than writing a sequence of unexplained numbers.

Use the measure requested. A difference of 6.8% minus 6.0% is 0.8 percentage points. A relative percentage change would divide by a base value and answer a different question. In a currency hedge, state whether the investor buys or sells the foreign currency before computing notional. A correct magnitude with the wrong direction can reverse the portfolio effect.

Pair the two formats during review

One effective drill uses the same case twice. First, answer a multiple-choice question that asks you to calculate or select. Then write a brief constructed response that justifies the choice from the vignette. Change one fact, such as the investor horizon or liability amount, and ask whether the conclusion changes. This tests transfer rather than recall of one keyed answer.

A compact review log can record the prompt command, case fact used, answer method, and error to avoid. If you knew the concept but chose the wrong benchmark, your next task is case interpretation rather than rereading the entire topic. If you wrote too much, practice reducing a response to the requested conclusion plus the strongest evidence. This targeted work improves both accuracy and timing.

Treat the session mix as flexible

The five-versus-six set mix means a session can lean slightly toward item sets or essays. The total remains 11 sets and 132 minutes. Do not assign all of your essay practice to one half or assume the second half is mainly multiple choice. Practicing both formats in the same sitting builds the attention and pacing needed to switch response modes.