Sector context and the ratios that lie

Every ratio in this module has a sector where it is the right tool and a sector where it is nonsense. A P/E of 12 is expensive for a shipping company and cheap for a software company. A P/B of 0.8 is alarming for a retailer and normal for a bank in a bad year. Before any ratio means anything, it needs to be placed in its sector, and then checked for the handful of ways it can be made to lie.

Typical ranges by sector

The figures below are rough and shift with the rate cycle (module 6); the point is the spread between sectors, not the exact numbers.

ACME, an industrial, at 16.7 times forward earnings and 11.6 times EV/EBITDA, sits in the middle of its band. Compared with a software company at 30 times, ACME is not cheap; it is priced like an industrial. Compared with an energy producer at 10 times, ACME is not expensive; it is priced like a business whose earnings do not depend on a commodity price.

The lesson: the only fair comparison is with the same sector, and a stock that "screens cheap" across all sectors is usually just in a cheap sector.

Why sectors trade differently

Three reasons explain most of the spread.

Growth and duration. Sectors expected to grow for longer command higher multiples because more of their value sits in the future. This also makes them more sensitive to interest rates (module 6).

Cyclicality. Sectors whose earnings swing with the economy trade at low multiples at peak earnings and high multiples at trough earnings. A 6 P/E on a steel company at record prices is the market saying "this will not last".

Capital intensity. Sectors that must reinvest heavily produce less free cash per dollar of EBITDA, so they get lower EV/EBITDA multiples. The EBITDA is real; it is just spoken for.

Key idea: A ratio is a comparison. Compare only within the sector, and know whether that sector's multiple is low because of cyclicality, capital intensity or genuine cheapness.

The ratios that lie

Now the traps. Each of these is a place where the number is technically correct and practically misleading.

1. The peak-cycle P/E. Covered in module 1: earnings at the top of a cycle make the trailing and even forward P/E look low just before earnings fall. Fix: look at EV/EBITDA on an average of several years' EBITDA, or at the P/E on trough earnings.

2. Adjusted EBITDA with everything added back. Some companies present EBITDA that excludes stock compensation, "restructuring" that recurs every year, litigation, and anything else inconvenient. Fix: recompute from operating income plus D&A yourself, and compare the company's adjusted figure with yours. A large gap is information.

3. The P/E with a one-off gain. ACME sells a division for a $150M gain. Trailing net income jumps to $360M, EPS to $3.60, P/E to 11.1. Nothing about the ongoing business changed. Fix: use adjusted or "continuing operations" EPS, and check the cash flow statement for the proceeds under investing activities.

4. Book value stuffed with goodwill. ACME's $1,600M of equity includes $1,000M of goodwill. Tangible book is $600M, or $6 a share. The P/B of 2.5 becomes 6.7 on tangible book. For an industrial that may not matter; for a bank it would be everything. Fix: for anything where book value is the lead multiple, use tangible book.

5. Net cash that is not available. A company shows $2,000M of cash, but $1,800M of it is held overseas or committed to a pending acquisition. The EV calculation subtracts cash that cannot actually be used. Fix: read the liquidity section of MD&A.

6. Debt that is not on the balance sheet. Operating leases (now mostly on balance sheet under current rules, but check), pension deficits, guarantees, and receivable factoring. A retailer with "no debt" but $3,000M of lease commitments is levered. Fix: credit analysts add lease liabilities and pension deficits to debt; do the same.

7. Growth that is bought. Revenue up 40% because the company made three acquisitions. Organic growth was 3%. The PEG built on 40% growth is fiction. Fix: MD&A almost always separates organic from acquired growth. Use organic.

8. Diluted share count from last quarter. A company with a large convertible bond or a stock-heavy pay plan can add 5% to its share count in a year. Per-share numbers computed on last year's count overstate everything. Fix: use the most recent diluted share count from the cover of the 10-Q.

A worked check on ACME

Run ACME through the lie list:

Nothing serious. ACME's ratios can be taken roughly at face value, with a note that management's EBITDA is 4% more flattering than the computed figure.

Try it: Take a company whose ratios look unusually attractive versus its sector. Run it through the eight-item lie list. Most "screens too cheap" stocks fail at least two items; find which two.

Recap

• Ratios are only comparable within a sector; each sector has its own typical band and lead multiple.
• Sector spreads come from growth duration, cyclicality and capital intensity.
• The common lies: peak-cycle earnings, over-adjusted EBITDA, one-off gains, goodwill in book value, trapped cash, off-balance-sheet debt, acquired growth, stale share counts.
• Recompute EBITDA yourself and compare with the company's adjusted figure; the gap is information.
• A stock that screens cheap across all sectors is usually just in a cheap sector.