Valuation models for Indian stocks, and when each one works
Which valuation model suits which kind of company?
Start from the business, then choose the model. A model built for a steady manufacturer gives nonsense for a bank or a loss-making start-up.
| Kind of company | Models that fit | Why |
|---|---|---|
| Steady compounder | DCF, earnings power value, owner earnings | Cash flows are predictable enough to forecast and discount |
| Profitable fast grower | Reverse DCF, PEG, Lynch fair value | The value is in growth; ask how much growth the price already assumes |
| Loss-making or young | EV / sales, reverse DCF on sales | There are no earnings to put a multiple on |
| Cyclical: metals, cement, autos | Earnings averaged over a cycle, P/B, EV / EBITDA against history | One year's profit is the top or the bottom of a cycle, not the norm |
| Bank, lender, insurer | P/B against return on equity | Debt is the raw material, so free cash flow has no clear meaning |
| Regulated utility | P/B, dividend yield | Returns are set by a regulator on the asset base |
| Holding company | Sum of the parts, with a holding discount | Its value is the stakes it owns |
Valarth sorts every company into a business type before valuing it, for this reason. Business type, explained.
What is a discounted cash flow, and when does it work?
The value of a business is the cash it will hand its owners in future, each year's cash marked down for the wait and the risk.
Free cash flow is operating profit after tax, plus depreciation, minus the capital spending and working capital the business needs. The discount rate for an Indian company starts from the yield on the 10-year Government of India bond and adds a premium for the risk of owning shares.
Works for: businesses with steady, positive cash flow. Fails for: banks, loss-makers, and cyclicals valued at the peak. The trap: most of the value sits in the terminal value, which rests on two guesses.
What is a reverse DCF?
A DCF run backwards: take today's price as given and solve for the growth the market must be assuming.
A forward DCF asks you to forecast growth, and a small change in the forecast gives any answer you like. The reverse form asks an easier question. If the price implies 22% growth a year for ten years, has this company, or any company in its industry, done that? If it implies 4% for a business that has grown 15%, the market is gloomy.
What is earnings power value?
What the company would be worth if its profit never grew again.
Bruce Greenwald's method removes the growth forecast altogether. Normalised profit is the average margin over a cycle applied to current sales. If the share trades below its EPV, you are paying nothing for growth. If it trades far above, all of the gap is a bet on growth.
What are the Graham number and the Graham formula?
Two quick tests from Benjamin Graham: one for a ceiling on price, one for a value that allows for growth.
The 22.5 is a P/E of 15 times a P/B of 1.5, the most Graham would pay for a defensive share. With earnings of ₹40 a share and book value of ₹250, the number is √(22.5 × 40 × 250) = ₹474.
8.5 is the P/E Graham gave a company with no growth. At ₹40 of earnings and 8% growth: 40 × (8.5 + 16) = ₹980.
Works for: profitable companies with real assets. Fails for: asset-light businesses such as software and consumer brands, where book value says little, so the Graham number looks far too low. Both formulas were written for a different market and interest-rate era; read them as a floor, not a target.
What are Lynch fair value and PEG?
Peter Lynch's rule of thumb: a fairly priced company has a P/E about equal to its growth rate.
A company growing profit 18% a year on a P/E of 27 has a PEG of 1.5. With earnings of ₹50 a share its Lynch fair value is 50 × 18 = ₹900, against a price of ₹1,350.
Works for: companies growing roughly 10% to 25% a year. Fails for: slow growers, where it gives absurdly low values, very fast growers, where the growth will not last, and cyclicals. It also says nothing about how much capital the growth consumes. PEG, explained.
What are owner earnings?
Warren Buffett's measure of the cash a business really produces for its owners.
Reported profit charges depreciation, an accounting estimate. Owner earnings replace it with what the company must actually spend to keep its factories and market position. A value is then the owner earnings divided by the return you require.
The hard part: companies do not report maintenance capex separately from growth capex, so it has to be estimated.
How do you value a share against its own history and its peers?
Compare today's multiple with what the market has usually paid for this company, and with what it pays for similar ones.
If a company has traded at a median P/E of 30 for five years and is at 22 today, either the business has got worse or the price has got better. Its peers tell you which: if they have de-rated too, it is the sector; if not, look for what changed in this company.
Works for: any profitable company with a few years of history and true peers. Fails when: the business has changed, the whole sector is in a bubble or a slump, or the peers are not really alike. A relative model can only say cheaper or dearer than something else, never cheap in absolute terms.
P/E ratio · P/E vs own 5-year median · P/E percentile in sector · EV/EBITDA
How are banks and lenders valued?
On price-to-book, judged against the return the bank earns on that book.
A bank earning 16% on equity, growing 8%, with a 12% cost of equity deserves (16 − 8) ÷ (12 − 8) = 2 times book. One earning 10% deserves half of book: it earns less than its capital costs, so growth destroys value.
Works for: banks, non-bank lenders, insurers and other businesses whose assets are carried near market value. The trap: book value is only as good as the loan book. Bad loans not yet recognised make book value, and so the P/B, look better than they are. RoE-justified P/B upside · P/B ratio · Return on equity for lenders
How do you use several models together?
Run every model that fits the business and look at where they agree. A range is more honest than a single number.
Each model is wrong in its own way. A DCF is sensitive to the discount rate, a Graham number to book value, a peer multiple to the mood of the sector. When five of seven applicable models say a share is below its value, that is stronger evidence than any one of them. When they split, the disagreement tells you what the valuation depends on.
Valarth does this for each company. It runs nine models: a fair value from its own engine, earnings power value, the Graham number, the Graham formula, Lynch fair value, owner earnings, P/E against the company's own history, P/E against peers, and the P/B that its return on equity justifies. It leaves out any that do not apply, such as owner earnings for a lender or any earnings model for a loss-maker, and reports how many of the rest say cheap. Beside that it gives a low, middle and high value and one sentence of evidence for its cheap, fair or dear view.
Valuation models saying cheap · Our valuation view · All 38 valuation methods, with calculators
A valuation is an estimate built on assumptions. None of this is a recommendation to buy or sell.