07.3 · Cyclicals & commodities¶

Why this matters: steel, cement, metals, chemicals, sugar, shipping, refining — a large part of the Indian market earns its profit in bursts. Valuing a cyclical on last year's earnings is how investors buy at the top (when the P/E looks lowest) and sell at the bottom (when it looks highest or infinite). The tools here — normalised earnings, mid-cycle margins, EV per tonne, replacement cost, the cost curve — exist to see through the cycle, and the trader's instinct for regimes is an advantage.

Learning objectives — after this lesson you can:

Explain the P/E paradox for cyclicals and why trailing multiples invert their usual meaning.
Build normalised (mid-cycle) earnings from a through-cycle margin or EBITDA-per-unit history.
Use P/B, EV/replacement cost and EV/tonne (or EV/MW, EV/room) at cycle troughs.
Read a cost curve and identify the marginal producer that sets the price.
Build a commodity price deck and translate it into earnings sensitivity.
Apply the toolkit to a fictional steel company and know where to check the real Indian names.

Prerequisites: 04.2 Margins & cost structure, 05.4 The capital cycle, 06.7 Other valuation methods · Time: ~90 min

1. What makes a business cyclical¶

Earnings are cyclical when demand swings (autos, capital goods, housing) and/or price swings (commodities: steel, aluminium, crude, sugar, chemicals) meet high fixed costs and long-lived capacity (operating leverage, 04.2) and lagged supply (the capital cycle, 05.4). Price-takers in commodities are the purest case: the product is identical everywhere, so profit is set by the gap between a global price and the company's cost, and both move.

The consequences for analysis:

Any single year's earnings is a draw from a wide distribution, not an estimate of "earning power".
Margins mean-revert violently; the reversion is the thesis, not the noise.
The balance sheet must survive the trough; leverage that looks modest at the peak is lethal at the bottom.
The right multiples are the ones anchored to things that do not cycle: book value, capacity, replacement cost, normalised earnings.

2. The P/E paradox¶

At the top of the cycle, earnings are at their highest and the market — knowing they will fall — pays a low multiple: the stock looks cheapest exactly when it is most dangerous. At the bottom, earnings are minimal or negative and the P/E is enormous or meaningless: the stock looks most expensive when it is cheapest. The old rule: buy cyclicals at high P/Es, sell them at low P/Es.

Narmada Steel (fictional): 7 MT capacity, market cap ₹30,000 Cr, net debt ₹12,000 Cr, equity ₹20,000 Cr.

FY	EBITDA/tonne (₹)	Volume (MT)	EBITDA (₹ Cr)	PAT (₹ Cr)	P/E at ₹30,000 Cr mcap
FY17	3,200	4.0	1,280	(720)	n/m
FY18	4,500	4.2	1,890	(110)	n/m
FY19	6,800	4.5	3,060	832	36x
FY20	9,500	4.8	4,560	1,920	16x
FY21	12,500	5.0	6,250	3,188	9x
FY22 (peak)	15,000	5.5	8,250	4,688	6.4x
FY23	7,000	5.8	4,060	1,395	22x
FY24 (trough)	4,200	6.0	2,520	15	2,000x
FY25	5,500	6.2	3,410	570	53x
FY26	8,000	6.5	5,200	1,875	16x

(EBITDA = ₹/t × MT ÷ 10 gives ₹ Cr; PAT = (EBITDA − D&A − interest) × 0.75 with D&A rising from ₹900 to ₹1,800 Cr as capacity was added and interest ₹650–1,100 Cr.) At the FY22 peak the stock was "cheap" at 6.4x; two years later earnings had fallen 99.7%. An investor who bought the 6.4x P/E bought the top.

3. Normalised earnings¶

Normalised (mid-cycle) earnings replace the current year's margin with a through-cycle average, holding current volume and cost structure:

\[\text{Normalised EBITDA} = \text{Current volume} \times \overline{\text{EBITDA/tonne}}_{\text{cycle}}\]

Narmada: ten-year average EBITDA/tonne = ₹7,620 (median ₹6,900); at FY26 volume of 6.5 MT → normalised EBITDA ₹4,953 Cr; less D&A 1,800 and interest 900 → PBT 2,253 → normalised PAT ≈ ₹1,690 Cr; normalised P/E = 17.8x. FY26's actual (₹1,875 Cr PAT, 16x) is close to mid-cycle — neither peak nor trough — which is the useful conclusion: the stock is roughly fairly priced on through-cycle earnings, and a bet on it is a bet on the next phase of the cycle, not on cheapness.

Choices that matter:

Which average? A full cycle (peak to peak or trough to trough), not "the last five years". Ten years usually spans one; check that it includes at least one bad year.
Mean or median? The mean is pulled up by super-cycles (FY21–22 above); the median is more conservative. Report both.
Has the structure changed? New capacity at lower cost, a captive mine, a tariff regime, China's export behaviour — any of these shifts the mid-cycle level. Normalise the unit economics, then adjust for known structural change, and say which.
Volume is not the current year's either if capacity utilisation is cyclical (cement at 65% vs 85%).

Multiples on normalised earnings: EV/normalised EBITDA (Narmada: 42,000 / 4,953 = 8.5x) and P/normalised E are the cyclical analyst's working multiples; compare with the company's own through-cycle history and with peers at the same cycle stage.

4. Asset-anchored multiples at troughs¶

When earnings are near zero, value the assets:

Multiple	Narmada	Reading
P/B	30,000 / 20,000 = 1.5x	Through a cycle, commodity producers oscillate roughly between 0.5–0.8x book at troughs and 2–3x at peaks (India-specific bands differ by sector — verify on Screener for steel/cement). 1.5x is mid-range, consistent with the normalised-earnings read
EV / tonne of capacity	(30,000 + 12,000) / 7 MT = ₹6,000 Cr/MT	Compare with replacement cost: a greenfield integrated steel plant in India has cost roughly ₹7,000–9,000 Cr per MT in recent projects (verify against announced capex — e.g., recent JSW/Tata/AM-NS expansions); at ₹6,000 Cr/MT the market values Narmada's capacity below what it would cost to build, a mild trough signal
EV / replacement cost	≈ 0.7–0.85x	Below 1x: no rational entrant builds new capacity at these prices → supply tightens → prices recover (the capital-cycle mechanism). Above 1.5x: everyone builds
EV / EBITDA (peak)	42,000 / 8,250 = 5.1x	The "cheap" peak multiple — a warning, not an invitation

For cement the same logic runs on EV/tonne (₹ Cr per MT of capacity) against a replacement cost that has risen with land and limestone costs; for power EV/MW; hotels EV/room; shipping EV vs fleet scrap/newbuild value; oil & gas EV/boe of reserves. In every case the question is the same: is the market pricing this capacity below what it would cost to replace, and if so, why?

5. The cost curve and the marginal producer¶

For a commodity, price is set where demand meets the cost curve — producers ranked from lowest to highest cash cost. In a downturn, price falls to the cash cost of the marginal producer needed to meet demand; the high-cost tail loses money and closes; the low-cost producers keep earning. In an upturn, price rises until the highest-cost producer needed is profitable.

flowchart LR
  A[Low-cost producers<br/>captive ore, scale<br/>cash cost ₹30k/t] --> B[Mid-cost<br/>₹38k/t]
  B --> C[High-cost / marginal<br/>bought-in ore, old plants<br/>₹46k/t]
  D[Demand] -.->|sets price near<br/>marginal cash cost| C

What this means for a stock:

Position on the curve is the moat. A first-quartile producer (captive raw materials, scale, logistics) earns through the cycle; a fourth-quartile producer earns only at peaks and may not survive troughs. Coal India and NMDC sit low on their curves because of reserve quality; some Indian steelmakers moved down the curve by acquiring captive iron ore mines in the 2020 auctions (verify).
Price floors are analysable. If the marginal producer's cash cost is ₹46k/t and the price is ₹44k/t, capacity is closing and the floor is near; if the price is ₹60k/t, the whole curve is profitable and capacity is coming.
China sets many curves. For steel, aluminium and chemicals, Chinese capacity and export policy define the global marginal producer; Indian tariffs (safeguard duties, anti-dumping) move the domestic price relative to it. A view on Chinese exports is a view on Indian steel margins.

6. Price decks and sensitivity¶

A price deck is the set of commodity price assumptions a model uses by year. Build it from forward curves where they exist (LME for base metals, Brent), broker/consultant consensus where they don't (steel, cement realisation), and always show the sensitivity:

\[\Delta\text{EBITDA} \approx \Delta\text{Price} \times \text{Volume} \times (1 − \text{pass-through to costs})\]

Narmada: a ₹1,000/t change in realisation on 6.5 MT with ~15% of it absorbed by cost linkages (input prices move with output prices) changes EBITDA by ≈ 1,000 × 6.5 / 10 × 0.85 = ₹553 Cr, or ~11% of FY26 EBITDA and ~22% of PAT (operating and financial leverage stacked). Present the sensitivity per ₹1,000/t, per $10/t or per 1% of price, and show what price the market is implying (a reverse DCF on the price deck — 06.6).

7. Indian sectors and where to look¶

Sector	Cycle driver	Unit metric	Anchor multiple	Representative listed names (for study; verify facts)
Steel	Global price (China), domestic demand, iron ore/coking coal	EBITDA/tonne; cash cost/tonne	EV/tonne; P/B	Tata Steel, JSW Steel, SAIL, Jindal Steel
Aluminium, zinc, copper	LME price, power cost, captive mines	EBITDA/tonne; cost curve position	EV/tonne; EV/EBITDA on LME deck	Hindalco, Vedanta, Hindustan Zinc, NALCO
Iron ore / coal mining	Reserve quality, government pricing, volumes	Realisation/tonne; cost/tonne	P/E on normalised; dividend yield (cash-rich PSUs)	NMDC, Coal India
Cement	Regional demand/supply, utilisation, fuel (petcoke/coal)	EBITDA/tonne; utilisation	EV/tonne	UltraTech, Ambuja, Shree, Dalmia
Refining & petrochemicals	GRMs, crack spreads, polymer margins	GRM $/bbl	EV/EBITDA on mid-cycle GRM	Reliance, IOC, BPCL, HPCL
Commodity chemicals	Chinese capacity, crude derivatives, capex waves (2021–24 specialty chemicals)	Spread over feedstock	EV/EBITDA normalised; P/B at troughs	Verify current names and cycle stage
Sugar	Cane pricing (FRP/SAP), ethanol policy, monsoon	Recovery %, realisation	P/B; EV/tonne of crushing	Regulated pricing dominates
Shipping, ports	Freight rates (Baltic indices), trade volumes	TCE $/day	EV vs fleet value; NAV	Verify
Autos, capital goods (demand cyclicals)	Interest rates, incomes, capex cycle	Volume, utilisation	Normalised P/E; mid-cycle margin	08.5, 08.6

Trader's lens

Cyclicals are regime trades, and the multiple is the regime indicator read backwards: low trailing P/E = the market pricing a regime shift down; high P/E = pricing a shift up. The useful analogy is vol of vol — you are not trading the level of earnings but the change in the level, and the entry signals are the ones vol traders know: extremes relative to a long history, capacity (supply) behaviour, and positioning. Normalised earnings are the long-run mean; the capital cycle is the mean-reversion mechanism; the cost curve tells you where the floor is. Position sizing must respect that the floor is only approximate and the path to it can bankrupt the levered.

India notes

Government pricing intrudes: coal (Coal India's notified prices and e-auction premia), iron ore (NMDC's monthly prices), sugar (FRP/SAP), ethanol, fertiliser (subsidy), fuel (OMC marketing margins). Model the policy as a driver, with a scenario for change.
Duties move the domestic price relative to the global cost curve: safeguard duties on steel (2025), export duties (2022), anti-dumping on chemicals — check the current schedule for any sector you model.
PSU miners and metals companies pay large dividends at cycle peaks; the dividend yield at the peak is not sustainable and should not anchor valuation.

Common mistakes

Buying a cyclical because its trailing P/E is low.
Normalising over a period that is all peak or all trough.
Ignoring balance-sheet survival: the trough is where the equity option can expire worthless (06.7, Merton).
Treating EV/tonne as a valuation rather than a cross-check (capacity ≠ cash).
Forgetting that structural changes (captive mines, new low-cost capacity, tariffs) move the mid-cycle.
Extrapolating a super-cycle (FY21–22) into "the new normal".

Key terms¶

Term	Meaning
Cyclical	A business whose earnings swing with demand and/or commodity prices
Price-taker	A producer of an undifferentiated product sold at a market price it cannot influence
Normalised / mid-cycle earnings	Earnings at through-cycle average margins on current volumes
P/E paradox	Cyclicals show low P/Es at earnings peaks and high P/Es at troughs
EV/tonne (EV/MW, EV/room)	Enterprise value per unit of physical capacity
Replacement cost	Cost to build equivalent capacity today; EV below it signals a trough
Cost curve	Industry producers ranked by unit cash cost; sets the marginal price
Marginal producer	The highest-cost producer needed to meet demand; its cash cost anchors the price floor
Cash cost	Cost per unit excluding depreciation and financing
Price deck	The commodity price assumptions by year in a model
Realisation	Average selling price per unit achieved
GRM	Gross refining margin ($/bbl): product prices minus crude cost
Super-cycle	An extended period of above-mid-cycle prices driven by structural demand or supply shocks

Check your understanding¶

Narmada Steel's FY22 P/E was 6.4x and FY24's was ~2,000x. Which year was the better time to buy, and what metric would have said so?

Answer

FY24. EV/tonne (well below replacement cost), P/B (≈1x or below at the FY24 price, if the price had fallen with earnings), and normalised P/E (which uses ₹1,690 Cr of mid-cycle PAT, not ₹15 Cr) all pointed to a trough; the 2,000x trailing P/E was the paradox at work.

Compute Narmada's normalised PAT using the median EBITDA/tonne instead of the mean, and the normalised P/E.

Answer

Median ₹6,900/t × 6.5 MT ÷ 10 = ₹4,485 Cr EBITDA; PBT = 4,485 − 1,800 − 900 = 1,785; PAT ≈ 1,339; P/E ≈ 22.4x. The choice of mean vs median moves the normalised P/E from 17.8x to 22.4x — report the range.

Steel is at ₹52,000/t; the marginal producer's cash cost is ₹46,000/t; new capacity announcements have surged. What does the cost curve predict?

Answer

The whole curve is profitable, so capacity is being added; when it arrives, price falls toward the marginal cash cost (~₹46,000) or below it temporarily until high-cost capacity closes. Margins for mid-curve producers compress by ~₹6,000/t; only first-quartile producers keep earning well.

Why does a cyclical need a stronger balance sheet than a stable business with the same average profit?

Answer

Its cash flow is intermittent: debt service is fixed but EBITDA can fall 70% in a trough (Narmada FY22→FY24). Leverage that is 1.5x EBITDA at the peak becomes 5x at the trough, covenants trip, and the equity — a call option on the assets — can expire worthless before the cycle turns.

A cement company trades at EV/tonne of ₹4,500 Cr/MT while recent acquisitions were at ₹8,000–10,000 Cr/MT. List two reasons it might not be cheap.

Answer

Its capacity may be poorly located (high freight to markets), old and inefficient (high cost per tonne), or under-utilised in an oversupplied region; and acquisition prices include control premia and synergies a minority shareholder does not get. EV/tonne must be read with utilisation, cost and regional pricing.

Go deeper¶

Edward Chancellor (ed.), Capital Returns: Investing Through the Capital Cycle (Marathon Asset Management) — the supply-side discipline applied to commodities and cyclicals.
Peter Lynch, One Up on Wall Street — the chapter on cyclicals and the P/E paradox.
Company "cost curve" slides in Hindalco/Vedanta/Tata Steel investor presentations, and Wood Mackenzie/CRU curves cited in DRHPs — learn to read where a producer sits.
Case I15 Tata Motors & JLR — a demand cyclical with acquisition leverage through two full cycles.

← Previous: 07.2 Insurers, AMCs, exchanges & brokers · Module index · Next: 07.4 High-growth & loss-making companies →