Alternatives in the Portfolio: Allocation, Due Diligence and Real-Asset Valuation

Finance-Plus Ring · Branch AL (Alternative Investments), Node AL1.04 · ~18 focused hours · Mastery gate >= 85%

A private property fund and a listed property trust can own the same kind of building, collect rent from the same kind of tenant, and report volatilities that differ by a factor of two. Neither number is a lie. The private fund is valued four times a year by an appraiser who anchors partly on last quarter's appraisal and partly on comparable sales that closed months ago. The trust is priced every few seconds by people with opinions.

Feed both numbers into the same two-asset optimiser and the difference is not academic. On the inputs worked through below, a private real-estate sleeve reporting a 9.00% annual standard deviation and a 0.22 correlation with equities takes 65.30% of the risky book. Correct the same two numbers for the smoothing that produced them, to 16.70% and 0.263, and the sleeve takes 24.91%. Forty points of portfolio weight, and not one economic assumption changed between the runs.

This is the shape of almost every problem in the alternatives sleeve. The asset is usually fine. The number describing it arrives through a measurement process built for accounting rather than for risk, and an allocator who accepts it at face value is not being optimistic, just uncorrected. So the work runs in four passes, and they are the four halves of what follows.

First, fix the numbers. Unsmooth what was smoothed, replace the standard deviation with statistics that survive a fat left tail, and subtract the survivorship, backfill and self-reporting biases that a hedge-fund index carries by construction. Second, pace the money. A commitment is not an investment; it is a contingent liability that turns into an investment on somebody else's schedule, and holding a stable 20% private weight is a ten-year exercise in arithmetic that has almost nothing to do with picking managers.

Third, check the operations. The auditor, the administrator and the valuation policy decide whether the reported number means anything at all, and a manager can be right about every position and still lose your money through the plumbing. Fourth, value the real assets, because a building, a stand of pine and a token each have a valuation method, and only one of the three has a cash flow you can discount.

The honest limit, stated once. Nothing here tells you to hold alternatives. The corpus's own position is that concentration in things you understand beats diversification into things you do not, and a 20% private sleeve is a choice with a liquidity price attached that this node will make you compute rather than assume. What follows is the toolkit for judging the claim when somebody else makes it, and for running the programme properly if you decide the claim is good.

One warning, because the region deserves it. An assistant will produce a pacing table, a de-smoothed volatility and a property valuation in seconds, and each of them can be confidently wrong in ways that leave no visible trace. A pacing table that does not reconcile still looks like a pacing table. The defence is the tie-out, run by hand, every time, and the gate rewards exactly that.

Learning objectives

By the end you can:

1. State the three honest roles an alternatives sleeve can play (diversification, return enhancement, inflation linkage), say which sleeve delivers which, and attach the caveat that each role carries.
2. Read the same fund through the asset-class lens and the risk-factor lens, compute the share of its variance that is equity beta, and price the fee against the alpha that survives the beta.
3. Compute the five statistics an illiquid, non-normal sleeve actually needs: skewness, excess kurtosis, maximum drawdown, the Sortino ratio and the first-order autocorrelation, naming the convention used for each.
4. Unsmooth a reported return series, both by the exact reverse recursion and by the two closed forms, estimate the smoothing parameter from the data, and show what the correction does to a mean-variance answer.
5. Quantify survivorship, backfill and self-reporting bias separately, reconcile the three to the total overstatement, and compound it into a terminal-wealth error.
6. Benchmark a private fund with a public market equivalent and a direct alpha, and read a vintage-year quartile claim knowing exactly which convention and which peer set produced it.
7. Build a ten-year commitment-pacing table that reaches and holds a target private weight, derive the steady-state commitment from first principles, and prove the table reconciles.
8. Size a liquidity buffer from unfunded commitments and the spending call, run the stress case in which listed markets fall, and price the secondary-market discount that a forced seller pays.
9. Run an operational due-diligence walk on the service-provider stack and the valuation policy, and convert a marking failure into rupees of overstated net asset value and overpaid performance fee.
10. Value a property three ways (direct capitalisation, the income multiplier and a five-year cash flow with a terminal cap rate), reconcile them, and decompose a timberland or farmland return into its sources so the components tie to the total.
11. Grade the four digital-asset valuation approaches on falsifiability, and say what each one would have to observe to be shown wrong.
12. (Productivity objective.) Use code and an assistant to build the pacing model, the unsmoothing routine and the property cash flow, while running the reconciliations that make the output admissible.

The duality, stated once. The gated skill is the arithmetic: a pacing table that ties, an unsmoothing that agrees by two routes, a property valuation whose three approaches you can explain the gaps between. The payoff is doing it in thirty lines of code across forty managers. The trap is sharper here than in accounting, because a wrong pacing table and a right one look identical on a page. There is no red squiggle under an unfunded balance that does not reconcile.

Prerequisites & connections

Builds on. AL1.01 supplies hedge-fund strategy classification and the fee mechanics used without re-derivation here: 2/20, hurdles, high-water marks, gates and lock-ups. When Worked Example 5 crystallises a performance fee on a mis-marked gain, the fee arithmetic is AL1.01's and the marking failure is this node's. AL1.02 supplies commodities, roll yield, infrastructure and the appraisal-smoothing observation that Section 4 turns into a method; its eight-quarter Sahyadri Realty series is reused deliberately so the two treatments can be compared. AL1.03 supplies crypto mechanics: what a token is, how staking and fee burn work, what custody costs. PW1.02 supplies commitments, calls, distributions, the J-curve, DPI/RVPI/TVPI and the waterfall. PW1.03 supplies the concession-life discounted cash flow for infrastructure and the discount-for-lack-of-marketability machinery. AA1.01 supplies the frontier, the tangency portfolio and the illiquidity cap as a policy constraint. AA1.04 supplies capital market expectations and consumes this node's de-smoothing closed forms when it builds a covariance matrix. M5.09 supplies the REIT operating toolkit: net operating income, funds from operations, adjusted funds from operations and cap-rate net asset value.

Feeds forward. AA1.05's institutional-investor allocations need the pacing and liquidity arithmetic built here before a private target can be written into a policy statement. AA1.07's manager-selection work assumes the operational due-diligence split taught in Section 10. PW1.04 owns the public market equivalent computation in full; Section 6 uses it as a benchmarking tool and hands the venture-valuation and cap-table machinery straight over. E11.05's index-default argument is the honest counterweight to everything here.

Deliberate non-overlaps. AL1.01 owns hedge-fund strategies and fee mechanics. AL1.02 owns commodities, roll yield, infrastructure as an asset class and the InvIT and MLP wrappers. AL1.03 owns crypto mechanics, custody and regulation. PW1.02 owns fund economics and the distribution waterfall, including the J-curve sense of the phrase "denominator effect" that Section 9 explicitly distinguishes from its allocation sense. PW1.03 owns the concession-life valuation of an infrastructure asset. PW1.04 owns the public market equivalent's derivation and the venture-valuation build. AA1.01 owns mean-variance optimisation; Section 4 applies a two-asset closed form once and does not rebuild the optimiser. M5.09 owns listed REIT and developer analysis; this node owns the private-property valuation approaches. Where those nodes own a concept, what follows cites and moves on.

Core teaching content

2. The asset-class lens and the risk-factor lens

Hold a long-short equity fund and you can describe it two ways. Under the asset-class lens it is an alternative: a separate line in the policy portfolio, sized against a target weight, benchmarked against a peer universe. Under the risk-factor lens it is a bundle of exposures, and the question is how much of its return came from exposures you already own.

Run the regression. A fund with an 11.4% annual standard deviation and a 0.62 beta to an equity index whose own standard deviation is 15.0% carries systematic variance of (0.62 x 15.0)^2 = 86.49 against a total variance of 11.4^2 = 129.96. The R-squared is 0.6655, the implied correlation with equities is 0.8158, and the idiosyncratic standard deviation is sqrt(129.96 - 86.49) = 6.5932 points. Two thirds of what looks like an alternative is the thing the rest of the portfolio already holds.

Now price it. If the index returned 9.0% over the period, the beta alone delivered 0.62 x 9.0 = 5.58%. The fund returned 8.6%, so 3.02 points sat above the beta line.

Deduct the 1.5% management fee and 1.52 points remain, and 1.44 points once you charge 8 basis points for buying the same beta through an index future. The fee takes 49.67% of the pre-fee alpha, and the investor would pay nothing at all for the two thirds of the risk that a futures position replicates.

Neither lens is wrong. The asset-class lens is how policy portfolios are written and how trustees think. The factor lens is how the risk actually behaves, and it is the one that tells you whether a 12% allocation to hedge funds has quietly become a 7% addition to your equity weight. Run both. When they disagree, the factor lens wins for risk and the asset-class lens wins for governance.

3. Measuring a sleeve that is neither liquid nor normal

The mean and the standard deviation describe a normal distribution completely. Alternatives sleeves are frequently not normal, and many of them fail in the same direction: many small gains and occasional large losses, the payoff of a written option. Five statistics catch what the standard deviation misses, and each needs its convention named because the conventions differ and the differences are large.

• Skewness. The third standardised moment. The moment convention divides m3 by m2^1.5 using n in both. The sample-adjusted convention that a spreadsheet returns multiplies that by sqrt(n(n-1))/(n-2). On the 24-month series in Worked Example 3 the two give -3.6606 and -3.9093. Report which one you used.
• Excess kurtosis. The fourth standardised moment less 3. The same two conventions give 12.4421 and 15.7840 on that series. A normal distribution scores zero; anything above about 3 means the tails carry far more probability than a volatility number implies.
• Maximum drawdown. The largest peak-to-trough fall of the compounded wealth path, computed month by month rather than from calendar-year returns. It is the only one of the five that a client actually experiences.
• The Sortino ratio. Excess return over downside deviation rather than over total volatility. Downside deviation squares only the shortfalls below a minimum acceptable return, and then divides by something. Divide by the full sample count and you get one answer; divide by the count of periods below target and you get a very different one. On the same series the two denominators give downside deviations of 1.4920% and 5.1684%, a factor of 3.4641 apart, which turns a respectable Sortino ratio into a terrible one. State the denominator or the number means nothing.
• First-order autocorrelation. The correlation of the series with itself lagged one period. In a liquid, correctly marked series it sits near zero. A reading of 0.3 or more is the fingerprint of smoothing, and it is the input the whole of Section 4 runs on.

The reason to compute all five is that the Sharpe ratio is blind to the shape of the distribution. Two funds can post the same mean, the same standard deviation and therefore the same Sharpe ratio while one of them loses 6.5% in a month and the other never loses more than 3.2%. Worked Example 3 builds exactly that pair and shows the Sortino ratio and the drawdown separating them by 63% and 3.35 points while the annualised Sharpe ratios differ by 0.0006.

6. Benchmarking the unbenchmarkable

A private fund cannot be benchmarked the way a long-only equity account can, because the manager controls the timing of the cash flows and there is no investable index of the same thing. Three devices are used, and each answers a different question.

The public market equivalent asks whether the same money, moved on the same dates, would have done better in a public index. Take each contribution and each distribution, future-value it at the index to a common date, and divide the future value of the distributions plus the residual net asset value by the future value of the contributions. Above 1.00, the fund beat the index on the money it actually used; below 1.00, it did not.

The Kaplan-Schoar form of this is the one to learn first, and Worked Example 7 computes it at 1.3148 for a fund whose net internal rate of return was 17.50% against an index annualising 10.09%.

Direct alpha converts the same comparison into an annual rate. Future-value every flow at the index, then take the internal rate of return of the adjusted series. On the same fund it comes out at 6.02% a year, which is materially below the 7.41 points you get by subtracting the index's annualised return from the fund's internal rate of return. The naive subtraction overstates the outperformance by 1.39 points, because an internal rate of return is not a quantity you can subtract from another internal rate of return computed on a different pattern of money.

The vintage-year quartile asks how the fund did against its peers of the same age. It is the most-quoted and the least reliable of the three, for reasons that are arithmetic rather than philosophical. On a fifteen-fund peer set, the same fund at a 17.50% net internal rate of return sits above the top-quartile break of 17.40% computed by linear interpolation on position (n-1)q, and below the break of 18.00% computed on position (n+1)q. One convention makes it top quartile and the other does not, and neither convention is wrong.

Remove the three worst funds because they stopped reporting and the inclusive break moves to 18.60%, and the fund is no longer top quartile on either method.

Ask four questions of any quartile claim, in this order. Which peer set, and does it include funds that failed? Which quartile convention? How much of the fund's stated multiple is still unrealised net asset value rather than cash returned, since 16.58% of this fund's total value was still a mark? And is the vintage defined by first close, by first capital call or by the year the fund was legally formed, because managers have been known to shop for the definition that flatters.

11. Monitoring: what changes a rating

Due diligence at the point of investment is the easy half. Monitoring is what catches the manager who was excellent in 2021 and is now running a different strategy with the same name.

Style drift is the commonest and the most measurable. Run a rolling twelve-month regression of the fund against the exposures it claims not to have. A long-short fund whose offering document describes a 0.28 equity beta and whose rolling beta now reads 0.71 has drifted by 0.43, which on a ₹120 crore sleeve is ₹51.60 crore of equity exposure the investor did not authorise and, in a 25% equity fall, ₹12.90 crore of loss, or 10.75% of the sleeve. That is a rating change, not a conversation.

Terms deserve a standing review of their own. A gate limiting redemptions to 20% of a position per quarter takes 14 quarters, three and a half years, to release 95% of a holding; 40.96% is still trapped after a year and 16.78% after two. Read the gate before you need it, because a gate is priced at zero on the way in and is the only thing that matters on the way out.

The triggers that should force a written review, each of which is a fact rather than a feeling: a change in the named portfolio manager or the head of operations; a change of auditor or administrator; a fee or liquidity term amended by side letter; assets under management doubling or halving in a year; a Level 3 share that rises for two consecutive quarters; a first-order autocorrelation that rises above 0.3; a strategy note that describes an instrument absent from the last one. Write the trigger list before you invest and the decision is a check rather than an argument.

14. Farmland and timberland

Both are land with a crop on it, and their return decompositions are genuinely different because one crop is harvested annually and the other stands for thirty years.

Farmland decomposes into two pieces. Income return is net operating income, usually cash rent less property tax and management, over the opening land value. Appreciation is the change in land value. On a 2,400-acre Iowa row-crop tract at $11,800 an acre earning $295 of cash rent against $46 of costs, the income return is 249 / 11,800 = 2.1102% and a 2.00% land appreciation gives a total of 4.1102%. Just over half the return, 51.34%, arrives as cash.

Timberland decomposes into four, and the fourth is the one that makes it a different asset. Value changes because the trees grow, because the price of standing timber moves, because land appreciates, and it falls because harvesting removes inventory. Income arrives when timber is sold. The identity that must hold:

change in value = growth valued at the opening price
                + price change on the closing inventory
                - harvest valued at the opening price
                + land appreciation

Worked Example 4 ties that to the cent on a 12,000-acre Arkansas pine tract, and the total return comes to 5.2985%, of which only 27.18% arrives as cash. The three appreciation components, growth, price and land, came to 5.8793 points between them, which is 110.96% of the total return, because harvest depletion took 2.02 points back and the harvest net of its own depletion contributed -0.58 points.

Harvest timing is an option, and it can be priced. A tree not cut this year is a tree that grows and can be sold next year. Defer the 22,000-ton harvest and the volume becomes 22,990 tons at 4.5% growth; at next year's $28.60 price that is $657,514 against $617,100 today, a one-year return of 6.5490%. So the break-even discount rate is 6.5490%: below it, defer; above it, cut. At a 7% required return the deferral loses money, and the price at which it would break even is $28.72 a ton.

The option there is a real one rather than a figure of speech. A timberland owner can store inventory inside an asset that keeps growing, at no storage cost, and can wait out a weak stumpage market in a way no farmer with a harvested crop can. The exercise rule is a comparison of two numbers that anybody can run.

Both sleeves are marketed as inflation hedges, and both linkages are observed rather than contractual. At an illustrative 3.10% consumer price inflation over the same year, verify against the current release, the timberland tract's real return is 2.1324% and the farmland tract's is 0.9798%. Positive, but neither is the mechanical pass-through that a regulated tariff formula provides, and neither should be sold as one. The defensible claim is that both have historically held real value across long horizons, with land supply as the mechanism, and that a single year proves nothing.

Two access notes worth carrying. In the United States both sleeves are reachable through institutional separate accounts, timberland investment management organisations and listed vehicles. In India, agricultural land ownership is restricted to agriculturists in several states, and the restrictions differ state by state, so a portfolio route usually means a listed agri-business or an overseas fund rather than direct title. Treat the specific state restrictions as perishable and verify against the state's own land-revenue statute before advising anybody.

Common mistakes & how experts think differently

1. Reading a reported volatility as a risk estimate. The novice compares a private fund's 9% standard deviation with an equity index's 16% and concludes the fund is safer. The expert computes the first-order autocorrelation first, and only then decides whether the 9% is a measurement or an artefact. The reflex is to look at the autocorrelation before the volatility, every time, because the second number is meaningless until the first is known.

1. Treating an unfunded commitment as if it were nothing. It appears on no portfolio report and earns no return, so it drops out of the liquidity conversation until the capital call arrives. The expert carries a single figure at all times: net asset value plus unfunded, as a share of the portfolio. In the year-ten programme built below that figure is 30.43% against a stated private weight of 20.14%, and the ten-point difference is the part that can force a sale.

1. Committing the steady-state amount during the ramp. A programme that commits 16% of its target net asset value every year while the portfolio compounds at 4.3% will never reach target, because the requirement is 20.3% rather than 16.0%. The novice concludes the managers are underperforming. The expert derives the steady-state formula before writing the first cheque and knows within one line of arithmetic whether a proposed programme is under-committed.

1. Comparing a fund's internal rate of return with an index's annualised return. They are computed on different patterns of money and the difference is not an alpha. On the fund in Worked Example 7 the naive subtraction gives 7.41 points where the direct alpha is 6.02, an overstatement of 1.39 points. The expert future-values the flows at the index and takes the internal rate of return of the adjusted series, or quotes a public market equivalent and says which one.

1. Accepting a top-quartile claim without asking which convention. The two standard quartile methods put the same fund on opposite sides of the line. Removing three failed funds from the peer set moves the bar by 1.20 points. The expert asks for the peer set, the convention and the share of the multiple that is still an unrealised mark before the claim means anything, and treats a manager who cannot supply all three as having answered the question.

1. Running operational due diligence inside the investment team. The team that likes the strategy is not the team to ask whether the administrator strikes the net asset value. The expert separates the two functions, gives the operational side a veto, and treats the list of documents a manager will not provide as more informative than the ones it will.

1. Valuing a property by whichever approach gives the preferred answer. Direct capitalisation and a discounted cash flow disagree for a reason, and the reason is almost always the implied perpetual growth inside the two cap rates. The expert reconciles them explicitly, solves for the terminal cap rate that would close the gap, and reports that number alongside the valuation.

1. Confusing a total return with a cash return in a real-asset sleeve. Timberland returned 5.30% in the year worked below and paid out 1.44 points; the rest was a change in the appraised value of standing trees and land. An investor who needs cash and reads the total return has misread the asset. The expert reports the cash component separately in any real-asset programme intended to fund spending.

1. Believing an index of things that self-report. Live-fund averages, backfilled track records and the quiet disappearance of funds that closed are three separate distortions, and the first two point the same way. The expert treats a hedge-fund index as an upper bound on what was available and asks what the attrition rate was, because a 41% attrition rate is itself the most useful number in the database.

1. Grading a valuation model by its elegance instead of by what could refute it. The novice is impressed by a tight fit. The expert asks what observation would prove the model wrong, and notices that a model refitted after every miss has no such observation. Applied to digital assets that single question sorts the four approaches faster than any amount of debate about the technology.

Worked examples

The figures in these examples are computed at full precision and rounded only for display. Synthetic entities are labelled synthetic. Figures that move with time carry a verify flag.

Worked example 1: A ten-year commitment-pacing programme (India, ₹)

Vindhya Family Office (synthetic) holds ₹500 crore, all of it in listed markets, and spends 4.5% of the prior year's total portfolio annually. It wants a 20% private-markets weight measured on net asset value.

Planning assumptions, all stated and all the office's own: the listed sleeve compounds at 8.0% before spending, private net asset value compounds at 12.0%, capital is called at 25% of the opening unfunded balance each year, and distributions run at 0% of grown net asset value in years 1 and 2, then 5%, 10%, 15%, 20% and 25% thereafter. The whole portfolio is planned to compound at 4.3% net of spending. No single vintage may exceed 10% of the portfolio.

Solution. Step 1, the steady state, before any table. At g = 12% and a mature distribution rate d = 25%, (1+g)(1-d) = 1.12 x 0.75 = 0.84. A flat net asset value therefore needs C = NAV x (1 - 0.84) = 0.160 x NAV, and unfunded settles at C / 0.25 = 4.00 x C. A net asset value growing at 4.3% needs C = NAV x [1.043 - 0.84] = 0.203 x NAV. Commit 16% of target and the programme drifts down forever; commit 20.3% and it holds.

Step 2, the ramp. Commit the growing steady-state amount plus 35% of the gap between the target net asset value and the current one, capped at 10% of the portfolio. Rounded to the nearest ₹1 crore the controller produces: 50, 52, 47, 38, 29, 23, 21, 22, 24, 27.

Step 3, run the table. All figures in ₹ crore.

Step 4, the two tests. Cumulative commitments of ₹333.00 crore less cumulative calls of ₹256.207060 crore gives ₹76.792940 crore, which is the closing unfunded balance to nine decimal places. Rebuilding net asset value from calls and distributions alone, with no reference to the table, returns ₹150.214030 crore, matching the table's closing figure to nine decimals. Both residuals are zero. Cumulative distributions were ₹216.206690 crore and cumulative spending ₹268.933218 crore.

Step 5, read the path. The programme crosses the 20% target in year five, peaks at 22.27% in year six and settles back to 20.14% by year ten, at which point commitments have turned upward again to track the growing portfolio. Year-10 unfunded of ₹76.79 crore is 10.30% of the portfolio, so total private exposure, marks plus promises, is 30.43%.

Read it. The overshoot in year six is the model's own J-curve, not an error: commitments signed during the ramp keep being called after the target is reached. Two numbers deserve to leave this table and go on a monitoring sheet.

The first is the year-ten commitment of ₹27 crore. The controller sized it on the ₹142.99 crore target implied by the opening ₹714.95 crore portfolio: the growing steady state asks for 20.3% of that, or ₹29.03 crore, and 35% of the ₹5.36 crore by which net asset value already sat above target trims ₹1.88 crore off it, for ₹27.15 crore before rounding. Against the ₹149.18 crore target that a 20% weight implies on the year-ten portfolio of ₹745.9 crore, that cheque is 18.10%, below the steady-state 20.3% precisely because the programme is running a little rich. The second is the 30.43%, because that is the exposure that behaves like a commitment when markets fall, and no report the office receives will show it.

Worked example 2: A US office building, valued three ways, and a core fund unsmoothed (US, $)

Kestrel Plaza (synthetic) is a 180,000 square foot suburban office building in Charlotte, North Carolina. In-place rent is $28.00 a square foot on the full rentable area, vacancy and collection loss runs at 8%, other income is $180,000, operating expenses are $1,780,000 and the replacement reserve is $54,000. Three comparable sales are available:

Land is worth $6,400,000, replacement cost new is $285 a square foot, and accrued depreciation is judged at 34%. The property discount rate is 8.75%, net operating income is expected to grow 3.0% a year, the exit cap rate is 8.00% and selling costs are 2.0%.

Solution. Step 1, build net operating income. Potential gross income = 180,000 x 28.00 = $5,040,000. Vacancy and collection loss = 8% x 5,040,000 = $403,200. Effective gross income = 5,040,000 - 403,200 + 180,000 = $4,816,800. Net operating income = 4,816,800 - 1,780,000 - 54,000 = $2,982,800, which is $16.5711 a square foot, at an operating expense ratio of 38.0751%.

Step 2, extract the cap rates. Comp A 2,904,000 / 41,200,000 = 7.048544%; B = 7.400000%; C = 6.900000%. Mean 7.116181%. Direct-capitalisation value = 2,982,800 / 0.07116181 = $41,915,740, or $232.87 a square foot.

Step 3, the income multiplier, and the check that makes it usable. A 41,200,000 / 4,900,000 = 8.408163; B = 8.005618; C = 8.595041. Mean 8.336274. Applied to the subject's effective gross income: 4,816,800 x 8.336274 = $40,154,166, which is 95.80% of the direct-cap value. The identity multiplier x cap rate = 1 - expense ratio explains the entire gap: the comparables average a 40.7291% expense ratio against the subject's 38.0751%, and (1 - 0.407291) / (1 - 0.380751) = 0.9571, which predicts the 0.9580 observed almost exactly.

Step 4, the cost approach. 6,400,000 + 285 x 180,000 x (1 - 0.34) = 6,400,000 + 33,858,000 = $40,258,000, which is 96.05% of the direct-cap value.

Step 5, the five-year cash flow. Net operating income grows at 3.0% from $2,982,800:

The five present values sum to $12,338,477. Year-6 net operating income is 2,982,800 x 1.03^5 = $3,457,883, so the reversion is 3,457,883 / 0.08 = $43,223,534, and after 2.0% selling costs $42,359,063, discounted five years at 8.75% to $27,848,385. Total discounted-cash-flow value $40,186,862, or $223.26 a square foot, of which the reversion is 69.30%.

Step 6, reconcile. The cash flow value is $1,728,877 below the direct-cap value, a gap of 4.1246%. The going-in cap rate of 7.1162% embeds perpetual growth of 8.75 - 7.1162 = 1.634%; the 8.00% exit cap embeds 0.750%. That 0.884-point difference is the whole explanation. Solve for the terminal cap rate that closes the gap and it is 7.5324%, which sits 46.76 basis points below the assumed exit cap and 41.62 basis points above the going-in cap.

A reconciled value weighting the cash flow 60% and direct capitalisation 40% is $40,878,413, or $227.10 a square foot, implying a 7.2968% cap rate on year-one income.

Step 7, unsmooth a core fund. Cardinal Core Property Fund (synthetic) publishes twenty quarters of returns: 1.02, 0.33, 1.12, -0.97, 0.02, -1.03, 0.10, 1.22, 1.49, 1.86, 2.47, 2.52, 2.75, 2.97, 0.58, 1.23, 1.82, 1.33, 0.24 and 0.98 percent. The reported mean is 1.1025% a quarter and the reported quarterly standard deviation is 1.1252%, so the reported annualised volatility is 2.2505%. The first-order autocorrelation is 0.599949, which is the estimate of alpha.

Run the reverse recursion at that alpha and the unsmoothed series is 1.0200, -0.7048, 2.3047, -4.1043, 1.5047, -2.6047, 1.7946, 2.8996, 1.8949, 2.4149, 3.3848, 2.5950, 3.0949, 3.2999, -3.0042, 2.2048, 2.7048, 0.5952, -1.3946 and 2.0898. Its quarterly standard deviation is 2.2495%, an annualised 4.4990%. The closed form gives 2.250482 x sqrt(1.599949 / 0.400051) = 2.250482 x 1.999839 = 4.5006%, so the two routes agree to 0.0016 points.

Step 8, the consequence. At a 0.55% quarterly cash rate the reported annualised Sharpe ratio is 0.9820 and the corrected one is 0.4886, a factor of 2.0100: a doubled volatility halves a Sharpe ratio. Maximum drawdown moves from 1.97% on the reported path to 5.20% on the unsmoothed one. A reported 0.24 correlation with equities becomes 0.3000.

Read it. Three valuation approaches on the same building landed within 4.2% of each other, which is what a competent appraisal looks like, and the reconciliation named the single assumption responsible. The fund half of this example is the more dangerous one, because nothing about the reported series looks wrong. It has a plausible mean, a modest drawdown and an attractive Sharpe ratio, and every one of those is a property of the marking process rather than of the buildings. The 0.5999 autocorrelation is the only visible symptom.

It takes one line of code to compute, and almost nobody computes it.

Worked example 3: Two funds, one Sharpe ratio (US, $)

Sandhill Premium Partners (synthetic) sells volatility. Cedar Bluff Diversified (synthetic) does not. Twenty-four monthly returns each, in percent. The cash rate is 0.35% a month.

Sandhill: 1.25, 1.40, 1.10, 1.35, 1.20, 1.45, 1.15, 1.30, -6.50, 1.20, 1.35, 1.25, 1.40, 1.10, 1.30, 1.20, 1.45, 1.15, 1.35, -2.20, 1.10, 1.40, 1.30, 1.20.

Cedar Bluff: 1.94, -0.34, 0.12, -1.25, 1.49, 2.40, 1.26, -1.48, -1.70, 1.71, 1.03, -0.56, -0.11, 3.54, 0.35, -1.02, 2.85, 3.08, -0.79, -1.93, 2.17, 2.62, 3.31, 0.57.

Solution. Step 1, the first two moments. Sandhill: mean 0.802083%, sample standard deviation 1.712263%. Cedar Bluff: mean 0.802500%, sample standard deviation 1.712682%. By construction the two are indistinguishable.

Step 2, the Sharpe ratios. Sandhill (0.802083 - 0.35) / 1.712263 = 0.264027 a month, annualised x sqrt(12) = **0.914616**. Cedar Bluff 0.264205 and 0.915235. The annualised gap is 0.0006. On the single most-quoted statistic in the industry these two funds are the same fund.

Step 3, the shape. Sandhill's skewness is -3.6606 on the moment convention and -3.9093 on the sample-adjusted one; its excess kurtosis is 12.4421 and 15.7840. Cedar Bluff's skewness is 0.0005 and its excess kurtosis is -1.2750. One of these distributions has a tail and the other does not.

Step 4, the drawdown. Sandhill's compounded path peaks after month 8 and troughs in month 9, for a maximum drawdown of 6.5000%. Cedar Bluff's worst peak-to-trough fall is 3.1548%. The gap is 3.35 points, and it is the number a client actually lives through.

Step 5, the Sortino ratio. Against a 0.35% minimum acceptable return, Sandhill's downside deviation on the full-sample denominator is 1.491993%, giving a monthly Sortino of 0.303006 and an annualised 1.049645. Cedar Bluff's is 0.914125%, a monthly 0.495009 and an annualised 1.714761. Cedar Bluff's Sortino ratio is 1.6337 times Sandhill's on identical Sharpe ratios.

Step 6, the denominator warning. Only two of Sandhill's 24 months fall below the target, so dividing by that count instead of by 24 gives a downside deviation of 5.168414%, larger by a factor of 3.4641, and a Sortino ratio that collapses. Two funds cannot be compared on Sortino ratios computed on different denominators.

Step 7, the autocorrelation check. Sandhill's first-order autocorrelation is -0.0740 and Cedar Bluff's is 0.0934. Both are near zero, so neither low volatility is a marking artefact. Sandhill's smoothness is a real property of a short-volatility payoff rather than an appraisal effect, which matters because the two look identical in a report and have completely different remedies.

Read it. Over 24 months Sandhill compounded to 20.72% and Cedar Bluff to 20.75%, so the difference is not in the outcome that happened. It is in the outcomes that did not. Sandhill's excess kurtosis of 12.4 says its tail carries far more probability than its volatility admits, and a third loss month of the same size would have separated the two decisively. A Sharpe ratio is a statement about the middle of a distribution.

A written option lives in the tail.

Worked example 4: A timberland year, decomposed (US, $)

Ouachita Pine Partners (synthetic) owns 12,000 acres of Arkansas pine carrying 38 tons an acre of merchantable inventory. Net biological growth runs at 4.5% of standing volume. During the year 22,000 tons are harvested at an average $28.05 a ton, and management, taxes and roads cost $186,000. Stumpage opens at $27.50 and closes at $28.60; bare land opens at $1,450 an acre and closes at $1,508.

Solution. Step 1, the inventory roll. Opening volume = 12,000 x 38 = 456,000 tons. Growth = 4.5% x 456,000 = 20,520 tons. Closing volume = 456,000 + 20,520 - 22,000 = 454,520 tons.

Step 2, the value roll. Opening timber = 456,000 x 27.50 = $12,540,000; opening land = 12,000 x 1,450 = $17,400,000; opening total $29,940,000. Closing timber = 454,520 x 28.60 = $12,999,272; closing land = 12,000 x 1,508 = $18,096,000; closing total $31,095,272.

Step 3, income. Harvest revenue = 22,000 x 28.05 = $617,100, less $186,000 of costs, gives net income $431,100.

Step 4, the total return. (31,095,272 - 29,940,000 + 431,100) / 29,940,000 = 1,586,372 / 29,940,000 = **5.298504%**.

Step 5, the decomposition, and the tie-out that proves it.

The four value terms sum to $1,155,272, which is the change in total value to the cent, with a residual below one nanodollar. Add income and the five components equal the total return exactly.

Step 6, the harvest as an option. Harvesting now grosses $617,100. Deferring one year grows the volume to 22,000 x 1.045 = 22,990 tons, worth 22,990 x 28.60 = $657,514 at next year's price. The one-year return to waiting is 6.5490%, so that is the break-even discount rate. At 5%, 6% or 6.5% the deferral wins; at 7% and 8% it does not. At a 7% required return the price at which deferral would break even is $28.72 a ton, which is above the $28.60 assumed, so the tract should be cut.

Read it. Only 27.18% of this year's return arrived as cash, and the harvest, once its own depletion is charged against it, contributed -0.58 points. An investor funding spending from this asset has 1.44 points of income and 3.86 points of appraisal.

Against an illustrative 3.10% inflation rate, verify against the current release, the real total return is 2.1324%, and a US farmland tract earning $249 an acre net on $11,800 land appreciating 2.0% returns 4.1102% nominal and 0.9798% real, with 51.34% of it in cash. Two real assets, two very different cash profiles.

The difference matters enormously to a portfolio that has to pay somebody every quarter.

Worked example 5: An operational due-diligence walk, priced (India, ₹)

Nilgiri Long-Short Fund (synthetic) is an Indian Category III alternative investment fund reporting ₹840 crore of net asset value against ₹700 crore at the start of the year. It charges 20% over an 8% hard hurdle, crystallised annually in cash.

The due-diligence pack discloses: an audit by a two-partner firm with no other fund clients, which also audits the fund's administrator; a net asset value calculated by the manager and "reviewed" by that administrator; a valuation policy permitting "manager discretion where no observable market price exists", with no escalation and no independent valuer; 18% of net asset value in unlisted positions marked by the chief investment officer; custody with an entity affiliated with the sponsor; no key-person clause; and two of three named investment staff departed in eighteen months.

Solution. Step 1, test the auditor. Independence fails on two counts. A firm auditing both the fund and its administrator cannot independently corroborate the administrator's work, and a firm with no other fund clients has no comparable base against which to judge a valuation policy. Ask for the engagement letter and the last three opinions; a qualification or an emphasis-of-matter paragraph is what you are looking for.

Step 2, test the administrator. "Reviewed" is the word doing all the work. An administrator that recalculates the net asset value from primary records is a control. One that checks the manager's arithmetic is a spreadsheet. Where the manager both marks and computes, the fee is a self-assessment and the audit is the only external check, which Step 1 has already disqualified.

Step 3, test the valuation policy, and price it. Level 3 holdings are 18% x 840 = **₹151.20 crore**. Apply a 20% haircut, which is modest for positions with no independent mark, and they are worth ₹120.96 crore. Corrected net asset value is ₹809.76 crore, so the reported figure overstates by ₹30.24 crore, which is 3.60% of the reported and 3.7344% of the corrected net asset value.

Step 4, price the fee. The hurdle amount is 8% x 700 = ₹56.00 crore. On the reported gain of ₹140.00 crore the fee is 20% x (140.00 - 56.00) = **₹16.80 crore**. On the corrected gain of ₹109.76 crore it is 20% x (109.76 - 56.00) = **₹10.752 crore**. The difference, ₹6.048 crore, is fee paid on gains that do not exist: 74.69 basis points of corrected net asset value, and 5.51% of the real gain.

Step 5, the reversal. If the mark reverses in full next year, the high-water mark stops the manager earning a second fee on the same ground. It returns nothing. Cash paid is gone, so the leakage is permanent at ₹6.048 crore, and the combined damage, overstated net asset value plus the fee, is ₹36.288 crore, or 4.48% of the corrected figure.

A high-water mark protects the manager's next fee. It returns none of the last one.

Read it. Not one of these failures is about the strategy, and an investment analyst reading the pitch book would have seen none of them. Three red flags from the standing list fire at once: the administrator reviews rather than calculates, the auditor is small and conflicted, and Level 3 exceeds a tenth of net asset value while being marked by the investment team. The fund is not necessarily a fraud.

It is a fund where the only person checking the number is the person paid on it, and the arithmetic above shows what that costs in a single ordinary year with a haircut most valuers would call conservative.

Worked example 6: Three biases in a hedge-fund index, separated (US, $)

A commercial database holds 1,240 live funds averaging 9.80% a year and 860 funds that stopped reporting, averaging 3.10%. Of the live funds' monthly observations, 22% are backfilled, and those months averaged 13.40% against 8.10% for the post-listing months. A further 180 funds closed to new money and stopped reporting; independent work puts their average at 12.10%.

Solution. Step 1, survivorship. All-fund mean = (1,240 x 9.80 + 860 x 3.10) / 2,100 = (12,152 + 2,666) / 2,100 = **7.056190%**. Survivorship bias = 9.80 - 7.05619 = **2.7438 points**. Attrition is 40.95%, which is itself the headline: two funds in five did not make it.

Step 2, backfill. Mixed mean = 0.22 x 13.40 + 0.78 x 8.10 = 2.948 + 6.318 = **9.266%**. Backfill bias = 9.266 - 8.10 = **1.166 points**.

Step 3, self-reporting. Full-universe mean = (1,240 x 9.80 + 860 x 3.10 + 180 x 12.10) / 2,280 = **7.454386%**. As a component of the overstatement this contributes 7.05619 - 7.454386 = **-0.3982 points**, because the funds that vanished for good reasons were above average.

Step 4, reconcile. Headline = 9.80 + 1.166 = **10.966%**. Universe = 7.454386%. Gap 3.511614 points. Components 2.743810 + 1.166000 - 0.398195 = 3.511614, residual zero to twelve decimals.

Step 5, compound it. 1.10966^10 = 2.830736 against 1.074544^10 = 2.052303. Terminal wealth is overstated by 37.93%, or $77.84 for every $100 invested over ten years.

Read it. The three biases do not point the same way, which is exactly why they must be measured separately rather than lumped into a single fudge. Two of them inflate and one deflates, and the net is still 3.5 points a year. The practical consequence for an allocator is blunt: a fund that beats a hedge-fund index has beaten a number nobody could have bought, and the comparison that means something is against a public index on the same money, which is what the public market equivalent computes.

That is the comparison to insist on.

Worked example 7: A public market equivalent, a direct alpha and a quartile claim (US, $)

Harlow Ridge Partners VII (synthetic) is a 2016-vintage US buyout fund. Cash flows in $ millions, with the index level at each date:

Residual net asset value at year 7 is $32 million.

Solution. Step 1, the multiples. Contributions $90m, distributions $161m, residual $32m. DPI = 161/90 = **1.7889**; RVPI = 32/90 = **0.3556**; TVPI = 193/90 = **2.1444**. The net internal rate of return, with the residual treated as a year-7 inflow, is 17.5049%.

Step 2, future-value every flow at the index. Multiply each by 196 / index(t):

Contributions future-value to $164.520407m; distributions plus net asset value to $216.313957m.

Step 3, the public market equivalent. 216.313957 / 164.520407 = **1.314815**. The same money in the index would have grown to $164.52m; the fund produced $216.31m, so $51.79m of value was added.

Step 4, direct alpha. Take the internal rate of return of the index-future-valued series and it is 6.0204% a year. The index itself annualised at (196/100)^(1/7) - 1 = **10.0908%**, and subtracting that from the fund's 17.5049% gives 7.4142 points, which overstates the direct alpha by 1.3938 points. Internal rates of return computed on different money patterns are not subtractable.

Step 5, the quartile. Fifteen 2016-vintage buyout peers report 4.1, 6.8, 7.9, 9.2, 10.4, 11.1, 12.6, 13.0, 14.3, 15.8, 16.8, 18.0, 20.4, 23.6 and 28.1 percent. Interpolating at position (n-1) x 0.75 = 10.5 gives a top-quartile break of 17.40%. Interpolating at position (n+1) x 0.75 = 12 gives 18.00%. The fund's 17.5049% is top quartile on the first convention and not on the second. Remove the three worst-performing funds because they stopped reporting and the first convention's break moves to 18.60%, so the fund is top quartile on neither.

Read it. Three numbers describe the same fund and only one of them survives a change of peer set. The public market equivalent of 1.3148 says the fund beat the public alternative on the money it used, and that statement survives any peer-set problem because it needs no peers. The direct alpha of 6.02% converts it to an annual rate.

The quartile claim depends on a convention nobody quotes and a peer set that quietly excludes failures, and 16.58% of the fund's stated multiple is still an unrealised mark, so even the internal rate of return is partly a valuation opinion. When a manager leads with the quartile, ask for the other two.

Worked example 8: Four digital-asset valuation approaches, graded (global, $)

All inputs are illustrative at September 2026 and every one is perishable; verify against the network's own data and a named market source before quoting. Network hashrate 720 EH/s, fleet efficiency 22 J/TH, electricity 5.5 cents a kilowatt hour, block subsidy 3.125 coins across 144 blocks a day, circulating supply 19.9 million, network value $1,180 billion, daily on-chain transfer value $9.4 billion.

Solution. Step 1, cost of production. 720 x 10^6 TH/s x 22 J/TH = 15.84 GW, and 15.84 GW x 24 h = 380,160,000 kWh a day, costing x $0.055 = **$20,908,800**. Daily issuance = 144 x 3.125 = 450 coins, so the electricity cost per coin is $46,464. Widen fleet efficiency to 16 to 30 J/TH and power to 3.5 to 7.5 cents and the range runs $21,504 to $86,400, a factor of 4.02.

Step 2, network scaling. At 42 million active addresses the network was worth $380 billion, implying a Metcalfe constant of 380e9 / (42e6)^2. Applied to 61 million addresses that predicts $801.58 billion against $1,180 billion observed, an error of -32.07%. The exponent that fits both points is 3.0361, not 2. The network-value-to-transactions ratio is more honest: 1,180 / 9.4 = **125.53 days**.

Step 3, addressable market. Success state = 20% x $18 trillion / 20.6 million coins = **$174,757**. At a 25% probability the expected terminal value is $43,689, discounted at 12% for ten years by a factor of 0.32197324 to $14,067. Discounting instead at a 25% venture rate while keeping the probability gives $4,691, an understatement of 66.65% from double-counting the same risk.

Step 4, invert it. The observed price is 1,180e9 / 19.9e6 = **$59,296**. Holding the capture share at 20%, the price requires a probability of 1.0538, which cannot exist. At certainty it requires a capture share of 21.08% of the entire above-ground gold market.

Step 5, token cash flows. Net supply growth = 0.55% - 0.40% = 0.15%, so the staking yield net of dilution is 3.20% - 0.15% = **3.05%**. At a 12% required return and 4% growth the Gordon multiple is 12.50, so justified value is 3.05% x 12.50 = **0.3813 of the current price. Equivalently the price requires 8.95% growth, or a net yield of 8.00%, and the current yield falls short by a factor of 2.62**.

Step 6, the standard. Stock-to-flow = 19,900,000 / (450 x 365) = 19,900,000 / 164,250 = **121.16**, against gold's 216,265 / 3,644 = **59.35**, a ratio of 2.04. Four more years of issuance take the stock to 20,557,000 coins against a halved flow of 82,125 a year, so at the next halving the ratio jumps to 250.31, a 2.07 times step produced entirely by a schedule known years in advance.

Read it. Grade them by what could refute each one. Token cash flows can be refuted next quarter by a fall in fee revenue, and they produce a claim in the same grammar as an equity valuation. The addressable-market approach can be refuted by its own inversion, which is why Step 4 is the most useful line here: it converts a vague bullish thesis into "you are pricing a 21% capture of the gold market with certainty", which is a sentence somebody can disagree with. Cost of production refutes nothing about value and everything about supply.

Network scaling, with a free constant and an adjustable exponent, refutes nothing at all, and stock-to-flow regresses price on a calendar. Falsifiability is the ranking criterion.

It sorts these four in about ninety seconds, which is less time than any of them takes to explain.

Verify. The figures in these eight examples come from one computation at full precision, and six of them check themselves by an independent route. The eight-quarter reverse recursion returns the true series to twelve decimal places. The twenty-quarter unsmoothing gives 4.498973% of annualised volatility against the closed form's 4.500602%, agreeing to 0.0016 points. The three bias components sum to 3.511614 points, which is exactly the headline-minus-universe gap. The pacing table's commitments less calls, 333.000000 - 256.207060 = 76.792940, equals its closing unfunded to nine decimals, and the net asset value rebuilt from calls and distributions alone equals the table's to nine decimals. The timberland decomposition's four value terms sum to $1,155,272, matching the change in total value with a residual below 1e-9. The remaining headline values at full precision: Sandhill Sharpe 0.914616 and Sortino 1.049645, Cedar Bluff 0.915235 and 1.714761; tangency real-estate weights 0.653001 reported and 0.249106 corrected; Kaplan-Schoar 1.314815 and direct alpha 6.020414%; Kestrel direct capitalisation $41,915,739.69, multiplier $40,154,165.51, cost $40,258,000.00 and cash flow $40,186,862.26, reconciling at a 7.532377% terminal cap rate; timberland total return 5.298504% and break-even deferral rate 6.549020%; farmland 4.110169%; and the electricity cost of $46,464.00 a coin.

Practice set

Work each problem before reading the solution. A calculator is assumed. Numeric answers within 1% of the stated figure count as correct, except where a reconciliation is demanded, in which case it must tie exactly. Markets alternate between India and the United States deliberately.

P1 (guided). The three multipliers. For alpha of 0.20, 0.35, 0.50, 0.65 and 0.80, give the volatility multiplier, the correlation multiplier and the covariance multiplier, and state the closed form for the third.

Solution. Volatility sqrt((1+a)/(1-a)): 1.224745, 1.441153, 1.732051, 2.171241, 3.000000. Correlation 1/sqrt(1-a^2): 1.020621, 1.067521, 1.154701, 1.315903, 1.666667. Covariance is the product, and it collapses to 1/(1-a): 1.250000, 1.538462, 2.000000, 2.857143, 5.000000. At alpha 0.80 an appraised sleeve contributes five times the covariance its reported numbers admit to, which is why a heavily smoothed series is far more dangerous in an optimiser than in a return table.

P2 (guided). Correct a sleeve and reprice its Sharpe ratio. A private infrastructure fund reports a 6.40% annual standard deviation, a 0.31 correlation with a 17% equity book and a 6.9% return; cash is 3.5%. Its reported series has a first-order autocorrelation of 0.45. Give the corrected volatility, correlation, covariance and Sharpe ratio.

Solution. Volatility = 6.40 x sqrt(1.45/0.55) = 6.40 x 1.623688 = **10.391605%**. Correlation = 0.31 / sqrt(1 - 0.2025) = 0.31 / 0.893029 = **0.347133**. Reported covariance = 0.31 x 6.40 x 17.0 = **33.728** squared points; corrected = 0.347133 x 10.391605 x 17.0 = **61.323636**, a multiple of 1/(1-0.45) = **1.818182**, which checks. Sharpe falls from (6.9-3.5)/6.40 = **0.531250** to 3.4/10.391605 = **0.327187**.

P3 (guided). Two steady states. (a) A ₹2,000 crore endowment targets a 15% private weight, with 11% private growth, a 22% mature distribution rate, a 25% call rate and 4.0% planned portfolio growth. (b) A ₹900 crore family office targets 25%, with 13% growth, a 28% distribution rate, a 30% call rate and 5.0% planned growth. For each give the target net asset value, the annual commitment, the steady-state unfunded balance and the total private exposure as a share of the portfolio.

Solution. (a) Target NAV = 0.15 x 2,000 = **₹300 crore**. Commitment share = 1.040 - 1.11 x 0.78 = 1.040 - 0.8658 = **0.1742**, so C = **₹52.26 crore**. Unfunded = 52.26/0.25 = **₹209.04 crore**, which is 10.452% of the portfolio, so total private exposure is 25.452%. (b) Target NAV = **₹225 crore**. Share = 1.050 - 1.13 x 0.72 = 1.050 - 0.8136 = **0.2364**, so C = **₹53.19 crore**. Unfunded = 53.19/0.30 = **₹177.30 crore**, or 19.700%, and total exposure 44.700%. The second investor has committed nearly half its portfolio to reach a 25% marked weight, which is the sentence its investment committee needs to see.

P4. The denominator effect, with the cure priced. A portfolio holds ₹620 crore public and ₹155 crore private. Public markets fall 28% and private marks fall 6%. Give the new private weight, the overshoot, the private sale needed to restore the original weight, and what it raises at a 25% secondary discount.

Solution. Original weight = 155/775 = **20.000%**. After the shock: public = 620 x 0.72 = ₹446.40 crore, private = 155 x 0.94 = ₹145.70 crore, total ₹592.10 crore, weight = **24.6073%**, an overshoot of 4.6073 points. To restore 20%, sell 145.70 - 0.20 x 592.10 = **₹27.28 crore** of private. At a 25% discount that raises ₹20.46 crore and gives away ₹6.82 crore, so the investor pays 6.82 crore for a rebalance that changes no economic exposure it did not already have.

P5. Two indications on one property (US). A building produces $1,845,000 of net operating income on $2,930,000 of effective gross income. Comparable cap rates average 6.85% and comparable gross income multipliers average 9.6. Give both value indications, the cap rate implied by the multiplier indication, the subject's expense ratio and the comparables' implied expense ratio.

Solution. Direct capitalisation = 1,845,000/0.0685 = **$26,934,307**. Multiplier = 2,930,000 x 9.6 = **$28,128,000**, which is 4.43% higher. Cap rate implied by the multiplier value = 1,845,000/28,128,000 = **6.5593%**. Subject expense ratio = 1 - 1,845,000/2,930,000 = **37.0307%**. Comparables' implied ratio = 1 - 9.6 x 0.0685 = 1 - 0.6576 = **34.24%**, so the comparables run leaner buildings and the multiplier flatters the subject. Adjust or discard it.

P6. A property cash flow (US). Year-1 net operating income is $2,100,000 growing 2.5%, the discount rate is 9.00%, the terminal cap rate is 7.25% and selling costs are 1.5%. Give the present value of the five income years, the reversion, the total value, the implied going-in cap rate and the reversion's share of value.

Solution. The five discounted incomes sum to $8,550,628. Year-6 income = 2,100,000 x 1.025^5 = **$2,375,957**, so the reversion is 2,375,957/0.0725 = **$32,771,824**, net of selling costs $32,280,247, discounted five years at 9% to $20,979,946. Total $29,530,573. Implied going-in cap rate = 2,100,000/29,530,573 = **7.1113%**. The reversion is 71.04% of value, so this valuation is mostly an exit assumption wearing a five-year cash flow.

P7. A timberland year with a falling price (US). 9,000 acres carry 300,000 tons. Growth is 5.2%, 16,000 tons are harvested at $30.55, costs are $121,000, stumpage falls from $31.00 to $30.10 and land rises from $1,900 to $1,957 an acre. Decompose the return and check the tie-out.

Solution. Opening value = 300,000 x 31.00 + 9,000 x 1,900 = 9,300,000 + 17,100,000 = **$26,400,000**. Growth = 15,600 tons; closing inventory = 300,000 + 15,600 - 16,000 = **299,600 tons**. Closing value = 299,600 x 30.10 + 9,000 x 1,957 = 9,017,960 + 17,613,000 = **$26,630,960**. Components: biological growth 15,600 x 31.00 = **$483,600**; price 299,600 x (-0.90) = **-$269,640**; depletion -16,000 x 31.00 = **-$496,000**; land 9,000 x 57 = **$513,000**. Sum $230,960, which equals the change in value exactly. Net income = 16,000 x 30.55 - 121,000 = **$367,800**. Total return = (230,960 + 367,800)/26,400,000 = **2.2680%**, of which biological growth is 1.8318%, price -1.0214%, land 1.9432% and income-plus-depletion -0.4856%. A falling stumpage price took a whole point off a year in which the trees grew perfectly well.

P8. A second public market equivalent (US). Flows in $m by year: -30, -40, -20, +12, +26, +44, +38, with a $18m residual at year 6. Index levels 100, 108, 119, 106, 131, 150, 158. Give the public market equivalent, the fund's internal rate of return, the index's annualised return, the TVPI and the value added.

Solution. Contributions future-value to $132.473140m, distributions plus residual to $151.592238m, so the public market equivalent is 1.144324. The fund's internal rate of return, residual treated as a year-6 inflow, is 10.8963% against an index annualising (158/100)^(1/6)-1 = **7.9219%**. TVPI = (120+18)/90 = **1.5333**. Value added over the index $19.119097m. Note the fund beat the index and still returned only 1.53 times the money in six years, which is what a modest public market equivalent looks like in multiple terms.

P9. Two biases stacked. A database holds 640 live funds averaging 11.20% and 410 dead funds averaging 2.40%. Backfilled observations are 18% of the live sample and averaged 15.1% against 9.2% post-listing. Give the attrition rate, the survivorship bias, the backfill bias and the combined overstatement.

Solution. All-fund mean = (640 x 11.20 + 410 x 2.40)/1,050 = (7,168 + 984)/1,050 = **7.7638%**. Survivorship bias = 11.20 - 7.7638 = **3.4362 points**. Attrition = 410/1,050 = **39.05%**. Backfill = 0.18 x 15.1 + 0.82 x 9.2 - 9.2 = 10.262 - 9.2 = **1.062 points**. Combined 4.4982 points a year, which over a decade is the difference between a good manager and an imaginary one.

P10. Non-normality on twelve months (US). Monthly returns: 1.4, 1.1, 1.6, 0.9, 1.3, -7.2, 1.5, 1.2, 1.0, 1.4, 1.1, 1.3. Cash is 0.30% a month. Give the mean, standard deviation, skewness, excess kurtosis, first-order autocorrelation, maximum drawdown, downside deviation, Sharpe and Sortino ratios, all annualised where relevant, naming conventions.

Solution. Mean 0.550000%, sample standard deviation 2.449304%. Skewness on the moment convention -2.976803; excess kurtosis on the same convention 6.952405. First-order autocorrelation -0.141347, so nothing here is smoothed. Maximum drawdown 7.2000%. Downside deviation against a 0.30% minimum acceptable return, full-sample denominator, 2.165064%. Monthly Sharpe 0.102070, annualised 0.353580. Monthly Sortino 0.115470, annualised 0.400000. The twelve months compounded to 6.4383%, and the one bad month took the annualised Sharpe ratio to 0.3536 against the 15.2969 those eleven good months would have produced on their own.

P11. A mis-mark priced (India). A fund reports ₹1,250 crore of net asset value with 24% in Level 3 positions marked by the investment team. Apply a 15% haircut. Give the corrected net asset value, the overstatement as a share of it, the performance fee overpaid at 20% of the phantom gain, and that fee in basis points.

Solution. Level 3 = 0.24 x 1,250 = **₹300 crore**; write-down = 15% x 300 = **₹45 crore**; corrected net asset value ₹1,205 crore. Overstatement = 45/1,205 = **3.7344%**. Performance fee overpaid = 20% x 45 = **₹9 crore**, which is 74.69 basis points of the corrected figure. Both percentages match Worked Example 5 exactly, because the overstatement is share x haircut / (1 - share x haircut) and 0.24 x 0.15 = 0.18 x 0.20; two different funds, identical damage.

P12. Size a buffer (India). An office carries ₹260 crore of unfunded commitments against a ₹1,400 crore portfolio and calls run at 28% of opening unfunded. Spending is 4.5% of the portfolio. Give the three years of calls, the remaining unfunded, and the three-year liquidity requirement including spending.

Solution. Year 1 = 0.28 x 260 = **₹72.80 crore**; year 2 = 0.28 x 187.20 = **₹52.416 crore**; year 3 = 0.28 x 134.784 = **₹37.740 crore**. Three-year calls ₹162.956 crore; remaining unfunded ₹97.044 crore, still 37.32% of the original. Spending = 3 x 0.045 x 1,400 = ₹189.00 crore. Total three-year requirement ₹351.956 crore, which is a quarter of the portfolio and none of it optional.

P13. Cost of production (global). Network hashrate 640 EH/s, fleet efficiency 25 J/TH, electricity 4.8 cents. Give the power draw, daily energy, daily cost, cost per coin at a 3.125 subsidy, and cost per coin after the next halving.

Solution. Power = 640 x 10^6 x 25 = 1.6 x 10^10 W = **16.00 GW**. Daily energy = 16 x 10^6 kW x 24 = **384,000,000 kWh**. Daily cost = **$18,432,000**. Coins = 144 x 3.125 = 450, so cost per coin $40,960. After the halving the same electricity produces 225 coins, so the cost doubles to $81,920 with no change in anything physical, which is exactly why a cost floor cannot be a valuation.

P14. A token's justified value (global). Gross staking yield 4.10%, issuance 0.62%, burn 0.51%, required return 13%, growth 3.5%. Give net supply growth, yield net of dilution, the Gordon multiple, justified value as a fraction of price, and the growth the current price requires.

Solution. Net supply growth = 0.62 - 0.51 = **0.11%**. Yield net of dilution = 4.10 - 0.11 = **3.99%**. Gordon multiple = 1/(0.13-0.035) = **10.526316**. Justified value = 0.0399 x 10.526316 = **0.4200** of the current price. The price therefore requires 13% - 3.99% = **9.01%** growth in fee demand rather than 3.5%.

P15. A quartile that flips (US). Twelve peers report 2.9, 5.5, 7.1, 8.8, 9.9, 12.2, 13.4, 15.1, 16.6, 19.0, 22.3 and 26.7 percent. Give the median and both top-quartile breaks, and classify a fund returning 17.4%.

Solution. Median on the inclusive method = **12.800%**. Top-quartile break, inclusive, at position 11 x 0.75 = 8.25, gives 16.6 + 0.25 x 2.4 = **17.200%**. Exclusive, at position 13 x 0.75 = 9.75, gives 16.6 + 0.75 x 2.4 = **18.400%**. A fund at 17.4% is top quartile on the inclusive method and not on the exclusive one, and the two breaks sit 1.200 points apart. Ask which convention before you believe the claim.

P16. The factor lens (US). A fund with a 13.2% standard deviation has a 0.74 beta to an index whose standard deviation is 16.5%. The index returned 10.5% and the fund 9.4%, with a 1.5% management fee. Give the systematic standard deviation, R-squared, idiosyncratic standard deviation, the return the beta explains, the alpha before and after fees, and the fee as a multiple of the pre-fee alpha.

Solution. Systematic = 0.74 x 16.5 = **12.210** points. R-squared = 12.21^2/13.2^2 = 149.0841/174.24 = **0.855625**. Idiosyncratic = sqrt(174.24 - 149.0841) = **5.015566** points. The beta explains 0.74 x 10.5 = **7.770%**, so alpha before fees is 9.4 - 7.77 = **1.630 points** and after fees 0.130 points. The fee is 0.920245 of the pre-fee alpha. With 85.6% of the variance explained by an index, an investor is paying 92 paise of every rupee of alpha for a position that is mostly a cheaper thing they could have bought.

Applied mini-project

Write the alternatives programme document for one investor. Budget about four hours. The deliverable is a document an investment committee could approve, plus a workbook behind it. Everything computed must reconcile, everything cited must trace to a named source with a date, and every synthetic entity must be labelled synthetic.

Choose your investor first, and state in two sentences their size, spending rate, tax position, governance and time horizon. Use a real institution's published figures where you can (an Indian family office's public disclosures, a US university endowment's annual report from its own site) or build a synthetic one and say so.

Part 1, the role statement and the target weights. For each sleeve you propose, name which of the three roles it plays and give the caveat. Compute the fee hurdle two ways at the proposed weight: the drag the incremental fee puts on the whole portfolio, and the outperformance the sleeve would need to move the portfolio's return by one point. Test 1: every sleeve has exactly one primary role and a stated caveat, and both halves of the fee-hurdle arithmetic are shown.

Part 2, the ten-year pacing table. Build it in a spreadsheet or in code: commitments, opening unfunded, calls, closing unfunded, distributions, net asset value, the liquid sleeve, the total and the private weight. State your call rate, distribution ramp, growth rates and spending rate, and say where each came from. Test 2: cumulative commitments less cumulative calls equals closing unfunded, to 1e-9. Test 3: net asset value rebuilt from calls and distributions alone equals the table's net asset value, to 1e-9. Test 4: derive the steady-state commitment from the roll-forward and show that your year-10 commitment is within 10% of it.

Part 3, the liquidity buffer and its stress justification. Shock the listed sleeve by a stated amount, mark the private book down by a smaller stated amount with the lag justified, and compute the post-shock private weight, the two-year call requirement, the spending requirement, the distributions on a stated haircut and the net cash need. Then price the alternative: the secondary sale required at a stated discount and the value given away. Test 5: the buffer is stated as a number and as a share of the liquid sleeve, and the secondary-discount cost is computed rather than described.

Part 4, a corrected risk table. For at least one appraised or self-reported sleeve, obtain or construct a return series, compute its first-order autocorrelation, unsmooth it by the reverse recursion, and report reported and corrected volatility, correlation and Sharpe ratio. Test 6: the reverse recursion and the closed form agree to within 0.05 points of annualised volatility, or you explain in one paragraph why they do not.

Part 5, an operational due-diligence checklist, filled. Take one manager, real or synthetic. Work the service-provider stack, the valuation policy, key-person and continuity, and the ten-item red-flag list. Where a red flag fires, price it: convert the marking exposure into overstated net asset value and overpaid performance fee. Test 7: every checklist line is answered "yes", "no" or "not disclosed", and the count of "not disclosed" lines is reported as a number, because that count is the finding.

Part 6, the monitoring triggers. A one-page list of facts that force a written review, each with a threshold and a data source, plus the redemption arithmetic for the fund's actual gate.

The deliverable is a six-page document plus the workbook, closing with one paragraph naming the single assumption the whole programme is most sensitive to, with the sensitivity quantified.

Rubric (20 points; pass at 16 with no zero on any row).

Any part whose test fails scores zero for that row regardless of how well the document reads. A pacing table that does not reconcile is a wrong table, whatever the prose around it says.

Reading & resources

The allocation spine.

• CFA Institute Level III, Alternative Investments for Portfolio Management: the asset-class and risk-factor lenses, the role of alternatives in an allocation, and the return and risk measures in exam form. [Paid] [Intermediate]
• CAIA Association, the Level I and Level II curricula: the most complete practitioner treatment of manager selection, operational due diligence and real-asset valuation available in one place. [Paid] [Intermediate to Advanced]
• Swensen, Pioneering Portfolio Management (revised edition): the endowment model argued by its author, and unusually honest about the illiquidity price and the governance it demands. [Paid] [Intermediate]
• Ilmanen, Expected Returns: the risk-factor lens applied across every asset class here, and the best single antidote to reading alternatives as separate boxes. [Paid] [Advanced]

Measurement and de-smoothing.

• Geltner's work on appraisal-based real-estate indices: the primary source for the unsmoothing model in Section 4. Read the original derivation before quoting the closed forms. [Paid] [Advanced]
• NCREIF methodology documents (ncreif.org): how the property and farmland and timberland indices are actually constructed, including the appraisal cycle that produces the smoothing. [Free] [Intermediate]
• Getmansky, Lo and Makarov, "An econometric model of serial correlation and illiquidity in hedge fund returns" (2004): the hedge-fund version of the same problem, and the paper that made autocorrelation a standard screen. [Free] [Advanced]
• Sortino and Price on downside risk: the primary source for the ratio, including the denominator question that trips up most published figures. [Paid] [Intermediate]

Benchmarking private funds.

• Kaplan and Schoar, "Private equity performance: returns, persistence and capital flows" (2005): the origin of the public market equivalent used in Section 6. [Free/Paid] [Advanced]
• Gredil, Griffiths and Stucke on direct alpha: the conversion of a public market equivalent into an annual rate, and why subtracting index returns from internal rates of return does not work. [Free/Paid] [Advanced]
• ILPA reporting templates and guidance (ilpa.org): what a limited partner is entitled to ask for, in the industry's own words, and the best checklist of fee and expense disclosure. [Free] [Intermediate]

Operational due diligence and valuation.

• AIMA's due diligence questionnaires (aima.org): the standard questionnaire, useful less as a form to send than as a list of what a complete answer looks like. [Free] [Intermediate]
• IPEV Valuation Guidelines (privateequityvaluation.com): how private positions are supposed to be marked, which is the document a valuation policy should be tested against. [Free] [Intermediate]
• The Appraisal Institute, The Appraisal of Real Estate: the three approaches, direct capitalisation, the income multiplier and the property cash flow, in full and with the adjustments this node compresses. [Paid] [Intermediate to Advanced]
• SEBI's Alternative Investment Fund Regulations and its valuation and reporting circulars (sebi.gov.in): the Indian regime for Category I, II and III funds, including independent-valuer requirements. Treat every threshold as perishable and verify the current text. [Free] [Intermediate]

One exercise beats reading any of these. Take any private fund factsheet you can obtain and compute its reported series' first-order autocorrelation. If it exceeds 0.3, unsmooth it and recompute the Sharpe ratio. Twenty minutes of that on a real document teaches more about alternatives than any chapter here.

The gap between the two Sharpe ratios is a number somebody is currently using to make an allocation.

Teach it back & journal

Feynman prompt. Explain to a trustee who has just been shown two funds why the one with the lower reported volatility may be the riskier holding. Use the actual mechanics rather than the conclusion: the appraiser who anchors on last quarter's number, the first-order autocorrelation that gives the anchoring away, and the arithmetic that turns a 9.00% reported standard deviation into 16.70%. Then take the second step, which is the one that decides the meeting. Show the trustee what the correction does to the allocation, not to the risk report: 65.30% of the risky book on the reported numbers and 24.91% on the corrected ones, with no economic assumption changed between the two runs. Then stop and let them ask the obvious question, which is whether the appraiser is wrong. The answer is no, and being able to say why not, that an appraiser values a building rather than forecasting a return series, is what separates somebody who has understood this from somebody who has memorised a formula. If your trustee can, unprompted, ask a manager "what is the autocorrelation of your reported series?", you have taught the substance.

Journal prompt. Find one real alternatives product you or somebody near you actually holds or has been pitched: a private fund, an alternative investment fund, a category III scheme, an interval fund, a real-asset sleeve inside a retirement account. Pull its own factsheet or offering document, not a summary of it. Then write down four things. First, its reported volatility and, if you can get a return series, its first-order autocorrelation; if the series is not published, write down that it is not, because that absence is the finding. Second, its unfunded commitment or its equivalent, and what share of your total portfolio the marks plus the promises represent. Third, its redemption terms, worked through arithmetically: at the stated gate, how many quarters to get 95% of your money back, and what fraction is still trapped after a year. Fourth, which of the four service providers the document names and which it does not.

Then write a paragraph on the thing this whole node is really about. Every number in an alternatives sleeve arrives through a process built for somebody else's purpose: an appraiser satisfying an accounting standard, a database serving marketing, a manager crystallising a fee. None of those processes is dishonest and none of them was designed to answer your question. The work is translation, and translation is slow, unglamorous and almost never done. So finish with a commitment that is small enough to keep. Which single check will you run, automatically, before any alternatives allocation you are ever asked to approve, and what will you do this week to make it a habit rather than an intention?