CFA Level III　考古題與解析

(A)Correct. Engagement is the use of ownership rights to influence how the issuer is run, through dialogue with the board and management, proxy voting and shareholder proposals; pressing a board committee on pay design is exactly that.

(B)Incorrect. Selling down on a changed model score is exit rather than voice: it is a portfolio construction decision that adjusts exposure and leaves the company's governance untouched.

(C)Incorrect. Overlaying a short index future manages the portfolio's factor and sector risk. It changes the manager's economic exposure, not the issuer's behaviour, so it is risk management rather than engagement.

(A)Incorrect. Deliberate overweights are an active decision. A fund that took them would no longer be passive, and the resulting tracking error would be far larger than a dozen basis points a year.

(B)Incorrect. A fund that holds a materially different constituent list is sampling or optimising, not fully replicating. That approach carries sampling error as an additional source of tracking difference.

(C)Correct. An index is a costless paper portfolio. A real fund pays a management fee, pays spread and commission on reconstitution and flow trades, and holds a small cash balance for redemptions, and those frictions are what a replicating fund's tracking error measures.

(A)Incorrect. 0.60 is the unconstrained information ratio, the ratio the manager would earn if every insight could be expressed at its optimal size. Real long-only and position-size limits mean only part of each insight reaches the portfolio.

(B)Incorrect. 1.50 divides the unconstrained ratio by the transfer coefficient instead of multiplying. Constraints can only destroy information ratio, never create it, so the constrained figure must be below the unconstrained one.

(C)Correct. The unconstrained information ratio is IC x the square root of breadth = 0.05 x 12 = 0.60, and the transfer coefficient scales it: 0.60 x 0.40 = 0.24.

(A)Incorrect. Matching cash flows date by date is cash flow matching, a different and stricter technique. Equal basis point values only equate the first-order rate sensitivities; the individual cash flows can still be timed quite differently.

(B)Correct. For multiple liabilities the money durations must match, which is what equal basis point values express, and asset convexity must slightly exceed liability convexity so the assets outperform on a parallel shift, while still being minimised to contain structural risk.

(C)Incorrect. Setting asset basis point value above the liability's leaves the portfolio deliberately long duration, which is an active rate bet rather than an immunisation, and asset convexity below the liability's would cause a loss on any large parallel move.

(A)Incorrect. Extending duration is a directional bet that rates will fall, and it also breaks the requirement to keep duration unchanged. A large rise in rates would produce a substantial loss.

(B)Correct. For a given duration, a barbell has greater convexity than a bullet because its cash flows are more widely dispersed. Higher convexity gains more when rates fall and loses less when they rise, which is the payoff to a large move of unknown sign.

(C)Incorrect. This is the reverse trade. A bullet concentrates cash flows near a single maturity and therefore has the lower convexity of the two at equal duration; buying it is the way to express a view that rates will barely move.

(A)Incorrect. This reverses both effects. More observations lower sampling error, and older data are by construction less representative of current conditions, not more.

(B)Correct. The standard error of a mean falls with the square root of the number of observations, so length helps precision. But a long window is more likely to straddle changes in policy regime, market structure or index composition, and precisely estimating an average across two different regimes is not the same as estimating the current one.

(C)Incorrect. Length reduces sampling error but increases, rather than reduces, exposure to regime change. Treating a longer sample as unambiguously more reliable is the assumption the regime-change problem exists to challenge.

(A)Incorrect. 5.5% omits the 0.6% liquidity premium. A corporate bond that cannot be sold quickly without a price concession must be paid for that, separately from its credit risk.

(B)Incorrect. 4.6% omits the 1.5% default risk premium, which is the compensation for the possibility that the issuer fails to pay coupon or principal in full and on time.

(C)Correct. The build-up sums the compensation demanded for each distinct source of risk: 1.0 + 2.2 + 0.8 + 1.5 + 0.6 = 6.1%. The first two terms together give the nominal short rate, and the three premia carry it out along maturity, credit and liquidity.

(A)Incorrect. That describes an unconstrained active risk problem, not mean-variance optimisation. Standard MVO works on total return and total variance and need not reference a benchmark at all.

(B)Incorrect. Rounding is not the issue; the sensitivity is to genuine estimation error in the inputs, above all in the expected returns. MVO is also a single-period model, so it does not model a sequence of periods.

(C)Correct. The objective function rewards a high expected return and a low covariance with everything else, and an asset whose mean has been overestimated scores well on exactly that test. Estimation error is therefore concentrated rather than diversified away, which is why unconstrained output is so often implausibly concentrated.

(A)Incorrect. This inverts the logic: reaching for return to close a funding gap raises the chance of missing the goal outright. Where a goal is genuinely unaffordable, the response is to revise the goal or the contributions, not the risk.

(B)Incorrect. Expected returns are indeed not a function of the investor's horizon, but the allocation is. The horizon governs the investor's ability to bear the risk required to earn those returns.

(C)Correct. A shorter horizon leaves fewer periods over which an adverse outcome can be made good, and the shrinking of human capital removes the bond-like asset that supported risk taking earlier. Illiquidity also becomes harder to justify once the money is needed on a known date.

(A)Incorrect. Representativeness is judging by resemblance to a stereotype, typically treating a short run of recent results as typical of the long run. Lengthening the sample addresses that, but not a client who overrides any sample.

(B)Correct. Illusion of control is the belief that one's own inputs and effort can steer an outcome largely beyond influence, and it shows up as excessive trading and concentrated positions. Anchoring the process on the global market portfolio and the returns it implies forces every deviation to be justified.

(C)Incorrect. Mental accounting is treating separately labelled pots of money as unrelated portfolios. This client runs one portfolio; the problem is the confidence he places in his own forecasts, not how he partitions the money.

(A)Correct. Zero cost requires the call premium to match the 3.10 paid for the put, which only the 103 call does. Equity index implied volatility is skewed, so an out-of-the-money put costs far more than a call the same distance away, forcing the cap nearer to spot.

(B)Incorrect. Put-call parity links a put and a call at the same strike, not at different strikes, and it does not imply equal premiums. The 112 call at 0.55 falls well short of funding the 3.10 put.

(C)Incorrect. Nothing requires the strikes to be equidistant; they are set by whatever premiums the manager chooses to exchange. The 106 call raises only 1.35, leaving a 1.75 net debit rather than zero cost.

(A)Incorrect. Negative gamma with positive theta describes a short straddle, which collects premium and loses from large moves. A bought straddle carries the opposite sign on both.

(B)Incorrect. A delta near one describes a long forward or a deep in-the-money call, and negative vega describes a written option. The two long legs offset in delta while reinforcing each other in vega.

(C)Correct. The long call and long put have offsetting deltas near the money, but both are long convexity and long volatility, so gamma and vega are positive. Both legs also decay, which makes theta negative.

(A)Incorrect. A short futures position would leave the fund holding cash and a negative equity exposure, doubling the shortfall against the policy portfolio rather than closing it.

(B)Incorrect. 7,500 is 12,000,000 divided by the index level alone, with the 250 multiplier left out. Contract notional is the index level times the multiplier, which is 400,000 here.

(C)Correct. Each contract represents 1,600 x 250 = 400,000 of index exposure, so 12,000,000 / 400,000 = 30 contracts are bought. The cash keeps earning its own return while the futures supply the equity beta.

(A)Incorrect. There is no premium to earn: the forward sits below spot, so the foreign currency is at a discount, and a seller of a currency at a discount gives up value rather than collecting it.

(B)Correct. The manager contracts to sell at 1.2250 and, with spot unchanged, gives up 0.0250 against a spot of 1.2500, or 2.0%. A forward discount on the foreign currency, which arises when its interest rate is the higher one, makes hedging costly.

(C)Incorrect. Covered interest parity is what creates the discount, by tying the forward to the interest rate differential; it does not neutralise it. Parity rules out arbitrage, not the realised cost of carrying a hedge.

(A)Incorrect. Physical separation supports a barrier but is not one. Information travels by e-mail, shared systems and committee papers, and staff of the two arms legitimately meet on firm-wide matters.

(B)Correct. The standard expects a firm to control material nonpublic information so that it reaches only those who need it, with a recorded process for bringing someone across and restricting them once crossed. That control, together with review of trading in affected securities, is what makes a barrier real.

(C)Incorrect. A permanent ban on every security the banking arm has touched is not required and would penalise the asset-management clients. The duty is to restrict dealing while material nonpublic information is held, not indefinitely.

(A)Incorrect. An approximate figure is still the client's information, and a range close enough to be useful is close enough to be a disclosure. The size of the leak does not change its character.

(B)Correct. Standard III(E) requires a member to keep information about current, former and prospective clients confidential. Being a client of the same firm gives the son no claim on his mother's information; her authorisation is what would permit the disclosure.

(C)Incorrect. Confidentiality runs account by account, not household by household. A shared adviser and a family purpose do not create consent, and the mother may have reasons of her own for not wanting the figure known.

(A)Incorrect. Saying that the process has changed, and in general terms how, does not require publishing the model's parameters. Proprietary detail may be withheld; the fact of a material change may not.

(B)Correct. Standard V(B) requires prompt disclosure of a material change to the investment process. Clients chose the manager partly for how it invests, and replacing the core signal changes the thing they bought even though the label is the same.

(C)Incorrect. An unchanged objective and benchmark say nothing about how the returns will now be generated. Holding the news until an annual review leaves clients invested in a strategy that is no longer the one they evaluated.

(A)Incorrect. A carve-out with no cash allocated overstates what a standalone portfolio in that strategy would have earned, so it cannot be treated as a portfolio. Non-discretionary portfolios are not given a composite of their own either — they are left out of composites altogether.

(B)Incorrect. Carve-outs are not prohibited; once cash is allocated they may be included. Attaching a note to a non-discretionary portfolio does not make its return attributable to the firm's investment decisions.

(C)Correct. A carve-out can be treated as a portfolio only once cash is allocated to it, because a return carrying no cash drag is not the return a client would actually have received. Non-discretionary portfolios are excluded because the firm did not control the decisions that produced the result.

(A)Incorrect. That describes the Brinson-Hood-Beebower specification, which measures the segment return against zero. In a strongly rising market it credits almost every overweight, even one to a segment that badly lagged the benchmark.

(B)Incorrect. Whether interaction is reported separately or folded into selection is the choice between a three-factor and a two-factor attribution. It is a separate decision from how the allocation term is benchmarked.

(C)Correct. Netting off the total benchmark return makes allocation measure the weighting decision itself. Overweighting a segment that returned 3% in a benchmark that returned 5% destroyed value, and Brinson-Fachler records that correctly as a negative allocation effect.

(A)Incorrect. 0.60 is the Sharpe ratio, (11.4% − 2.4%)/15.0%. It divides excess return over the risk-free rate by total standard deviation, so it appraises stand-alone rather than benchmark-relative performance.

(B)Correct. The information ratio is mean active return divided by active risk: (11.4% − 8.4%)/4.0% = 3.0/4.0 = 0.75. Numerator and denominator must both be measured against the same benchmark.

(C)Incorrect. 2.25 mixes the two measures, dividing excess return over the risk-free rate by tracking error. The information ratio's numerator has to be the return over the benchmark, not over cash.