Actionable insights — Fairfax India: discount on discount?

1. Find the marking gap: compare unrealised private returns against realised returns from the same manager

1. Split the vehicle's portfolio into what is marked to market (listed) and what is marked to management's estimate (private). Get the percentage in each bucket.
2. Find three reported numbers: the unrealised return on the listed book, the unrealised return on the private book, and the return on monetised (sold) investments.
3. If realised returns match the listed-book returns and both far exceed the unrealised private returns, the private marks are conservative — the same people cannot be brilliant in public and mediocre in private.
4. The size of the hidden value scales with the share of the portfolio that is not marked to market.
5. Test the alternative explanation before accepting it: does the manager monetise selectively, selling only what worked? If so, part of the gap is selection, not conservatism.

• Adverse selection in what gets sold — realised returns are computed on the manager's own choice of exits.
• The mirror risk: a vehicle whose private marks are aggressive fails this test in the opposite direction, and the same three numbers detect it.

2. Compare the growth rate of the mark against the growth rate of the underlying business

1. Pick the largest private asset. Find its carrying value in an old annual report and in the current one.
2. Compute the CAGR of the carrying value.
3. Compute the CAGR of the asset's revenue, EBITDA or traffic over the same window.
4. A carrying value growing far slower than the business is the clearest possible evidence of a stale mark — and it is available from public documents without any modelling.
5. Multiply the gap by the asset's share of the portfolio to size what it is worth.

• Ownership-percentage changes between the two reports — the carrying values here are stakes (54% then, 74% now), so the per-unit comparison needs the grossing-up the post does inline.
• Revenue growth that has not reached profit: an airport in a heavy capex phase can compound revenue and still not be worth more.

3. Value an unlisted asset off recent private transactions, then sanity-check against the nearest listed peer

1. Assemble actual transaction multiples for comparable assets — deals, not analyst targets.
2. Take the average, and note the dispersion; a 13x-to-29x range is a warning that "comparable" is doing a lot of work.
3. Apply it to the asset's own metric and compare with the carrying value.
4. Cross-check against the closest listed peer, and be strict about comparing the same multiple on both sides.
5. Adjust for growth and asset quality explicitly, in words, rather than asserting the asset "deserves" a higher multiple.

• Step 4 done loosely: EV/Sales against EV/EBITDA is not a comparison, and the post presents it as one.
• Comparables supplied by the party arguing the asset is cheap — this table comes from Fairfax India's own deck.
• Transaction multiples struck in a different rate environment. Sydney and Copenhagen were bought when capital was cheaper.

4. Require a named event that converts an estimate into a price, and state what it does to the numbers

1. A discount to intrinsic value is not a thesis unless something can close it. Identify the specific event: an IPO, a sale, a spin-off, a wind-up.
2. Quantify the mechanical effect on reported book value at a conservative assumed price.
3. Restate the valuation multiple after that effect, so the discount is expressed in post-event terms.
4. Check who controls the event, and what still stands in its way.
5. Size the position for the possibility it never happens: a good business at a discount is survivable; a bad one waiting for a catalyst is not.

• Step 4 is where this one is weakest: "still in the process of obtaining regulatory approvals", with no date, and the same intention has been signalled before.
• An IPO that prices below the carrying value — the mechanism cuts both ways once the market, not management, sets the mark.

5. Read the share count as the honest record of capital allocation

1. Chart shares outstanding against the share price over a decade.
2. Look for issuance at highs and repurchase at lows. The pattern is rare enough to be a genuine screen.
3. Get the disclosed average repurchase price and compare it with book value at the time.
4. Express cumulative buybacks as a percentage of shares ever issued — that is the number that shows whether it was policy or gesture.
5. Apply the same test to the parent and to any related vehicle: the behaviour usually travels with the people.

• Buybacks at a discount to a book value that is itself an estimate — the discipline is only real if the estimate is conservative, which is the same claim being tested elsewhere.
• Succession: this whole test is a judgement about people. Here it doubles as the succession answer — Benjamin Watsa is both the allocator being assessed and Prem's designated successor.

6. Before buying a managed vehicle, work out who collects the fee — and whether you can own that side instead

1. Read the management agreement: base fee, performance fee, hurdle, and the period over which the hurdle is measured.
2. Convert it to a single annual drag on the portfolio, and check it against the fees actually paid over the vehicle's life.
3. Identify the recipient. If it is a listed entity, you now have two ways to express the same view.
4. Compare them: the fee is a cost to one and revenue to the other, and the payer needs to be enough cheaper to compensate.
5. Choose the side of the fee you want to be on, and say so plainly.

• A performance fee measured on book value per share, when the same analysis argues book value understates reality — the hurdle is struck on the conservative number.
• Your own incentives when you already own the collector: every argument that the payer is undervalued is also an argument for the fee stream you own.

7. Demand one completed transaction as proof before you underwrite a whole portfolio of estimates

1. Search recent disclosures for any private stake the company has actually sold.
2. Compare the realised price per share against the carrying value immediately before the sale.
3. Treat that ratio as the empirical estimate of the conservatism factor for the rest of the private book.
4. Apply it, discounted, to remaining private assets — one data point is evidence, not a multiplier.
5. Re-read the headline valuation of the parent in that light.

• Generalising from one sale. An 85% uplift on the asset the manager chose to sell says little about the assets they chose to keep.
• The direction of travel: once conservative marks are widely believed, the discount closes and the free option disappears — this argument works best while it is unpopular.