valuation-multiples-overview-pe-evebitda-ps — evaluating investments

The practical point: a multiple is a price you pay today for 1 unit of a business metric over the last 12 months (TTM)—and your return over the next 36–60 months is usually dominated by (1) metric growth (%) and (2) multiple change (x), not by your "story."

Why Valuation Multiples Matter

Multiples turn a messy valuation problem into a 3-number comparison: price, fundamentals, and time (typically TTM and a 3–5 year horizon). Empirically, mean reversion is not a slogan: when stocks trade >20% below their sector median P/E, they historically beat the market by +3.2% per year over 1980–2010, while >20% above lagged by -2.7% per year, with convergence taking about 3–5 years (Damodaran, 2012). The point is: your baseline assumption should be partial reversion, not permanent exception.

P/E Ratio (Price-to-Earnings): "How many dollars for $1 of earnings?"

Definition (use a 12-month denominator)

P/E = Market Capitalization ÷ Net Income (or Price per share ÷ EPS) for the last 12 months. If a company earns $5.00 per share and trades at $100, your P/E is 20x—you are paying $20 for $1 of annual earnings.

Quantified thresholds you actually use

Broad-market cheap: Trailing P/E < 12x (this occurred 18% of S&P 500 trading months since 1950)
Broad-market expensive: Trailing P/E > 22x (this occurred 23% of months since 1950; subsequent 10-year returns averaged 4.2% per year)
Best-fit sectors (typical ranges): utilities 14–20x, consumer staples 18–25x, financials 8–14x

When P/E works (and when it breaks)

P/E is most defensible when earnings volatility is low (single-digit % swings) and capital structure is reasonably stable over 3–5 years. It breaks fast when net income is near 0% margin (a 1% revenue change can swing earnings by >50%), or when leverage changes the equity claim materially.

EV/EBITDA: "What you pay for the operations, regardless of how they're financed"

Definition (enterprise value forces you to count debt)

EV/EBITDA = (Market Cap + Net Debt) ÷ EBITDA (TTM). You're paying for the whole firm (equity + debt holders), then dividing by an operating cash-flow proxy.

Two empirical reasons it's used:

Cross-industry pricing error is lower: 18.3% for EV/EBITDA versus 23.7% for P/E, and for capital-intensive sectors EV/EBITDA error drops to 14.1% (Liu, Nissim & Thomas, 2002).
Return spread is large in the tails: lowest EV/EBITDA decile averaged 13.2% annual returns vs 6.8% for the highest decile (6.4 percentage points spread) (Loughran & Wellman, 2011).

Quantified thresholds and sector ranges

Value/distress flag: EV/EBITDA < 6x can indicate value or distress—so you screen covenants and refinancing risk, not just the multiple.
Growth premium rule-of-thumb: every +5 percentage points of EBITDA growth has historically commanded about +1.2x EV/EBITDA.
Sector ranges (medians):
- Technology 12–25x (median 17x)
- Industrials 8–14x (median 10.5x)
- Healthcare 10–18x (median 13x)
- Energy 4–8x (median 5.5x)
- Real estate 14–22x (median 18x)

The hidden trap: "cheap" EV/EBITDA can be an earnings collapse

From June 2014 to February 2016, S&P 500 Energy EV/EBITDA expanded from 5.8x to 14.2x because EBITDA collapsed 61% while enterprise values fell only 34%; investors buying "cheap" at 6.5x later saw 25x as fundamentals deteriorated, with sector return -42% vs S&P 500 +11% (S&P Capital IQ; Bloomberg aggregation). The point is: a low trailing multiple is not cheap if the denominator is peaking.

P/S Ratio (Price-to-Sales): "When earnings are unusable, anchor on revenue"

Definition (simple, but you must correct it)

P/S = Market Cap ÷ Revenue (TTM). You use it when net income is negative or net margin is <2%, because in that zone P/E is either undefined or violently unstable.

Quantified benchmarks (especially for SaaS)

For software-like unit economics, revenue growth maps into P/S bands:

>30% revenue growth: 8–15x P/S
15–30% revenue growth: 4–8x P/S
<15% revenue growth: 2–4x P/S

Accuracy (why P/S is not "worse," just conditional)

P/S had 31.4% average valuation error for profitable firms, but 22.8% for loss-making firms where P/E cannot be computed (Schreiner & Spremann, 2007). The point is: P/S is a "best available" tool when earnings are structurally uninformative.

Mandatory adjustment: margins turn P/S into an implied P/E

You convert P/S into an earnings equivalent with one division:

Implied P/E = (P/S) ÷ (Net margin)

Example: at 1.2x P/S and 8% net margin, implied P/E is 1.2 ÷ 0.08 = 15x. At 0.5x P/S and 2% margin, implied P/E is 0.5 ÷ 0.02 = 25x—the "cheaper" P/S is 67% more expensive on earnings power.

When You Use Each Multiple (a decision rule with numbers)

Use this 3-step filter on every screen of 20–200 comps:

If net margin is >2% and stable (±2 percentage points) over 3–5 years—start with P/E.
If leverage differs by >2.0x Net Debt/EBITDA across comps—switch to EV/EBITDA for comparability.
If net income is negative or <2% margin—use P/S, then margin-adjust to implied P/E.

If you want fewer errors, move from "industry median" to "warranted multiples": adjusting for profitability, growth, and risk reduced valuation errors by 35–45% versus simple medians (Bhojraj & Lee, 2002).

Sector Comparisons: the only comparisons that don't lie

You compare multiples within sectors because capital intensity and margins differ by >10 percentage points across industries. Your anchors are the medians above (e.g., Industrials 10.5x EV/EBITDA vs Energy 5.5x), and your discipline is to treat ±20% from the sector median as "needs a reason," not "a bargain" (Damodaran, 2012).

You also model compression explicitly:

Every +100 bps increase in the 10-year Treasury has historically compressed growth-stock P/E by 2.1x.
The 1st earnings miss compresses the multiple by 8–12%; the 2nd consecutive miss by another 15–20%.
At recession onset, defensive sector multiples expand 10–15%, while cyclical multiples compress 20–30%.

Historical example: in the Tech bubble compression from March 2000 to October 2002, the NASDAQ Composite P/E fell from 175x to 21x while the index fell 78% (from 5,048 to 1,114); buying at 40x P/E in March 2001 still produced a -52% loss over the next 18 months (Shiller, 2005). The point is: multiple risk can overwhelm "good companies" for 6–24 months.

Worked Example: Comparing two industrial companies using EV/EBITDA

You are a value-oriented institutional investor with $2,000,000 allocated to Industrials, a 3–5 year horizon, and a maximum acceptable drawdown of 25%.

Step 1: Calculate normalized EV/EBITDA

You compute Company A: EV $4.4B ÷ EBITDA $400M = 11x.
You check history: A's trailing 5-year average is 9.2x, so today is +20% premium (11.0 ÷ 9.2 - 1).
You compute Company B: EV $3.6B ÷ EBITDA $200M = 18x.
You check history: B's trailing 5-year average is 15.4x, so today is +17% premium (18.0 ÷ 15.4 - 1).

Step 2: Benchmark to sector medians

Industrial machinery median EV/EBITDA is 10.5x; automation subsector median is 16.2x.
You find A at +5% to sector (11.0 ÷ 10.5 - 1).
You find B at +11% to subsector (18.0 ÷ 16.2 - 1).

Step 3: Pay attention to growth per turn of multiple

A has 4% revenue CAGR and 6% EBITDA CAGR.
B has 18% revenue CAGR and 22% EBITDA CAGR.
You compute a PEG-like ratio: EV/EBITDA ÷ EBITDA growth—A = 11 ÷ 6 = 1.83, B = 18 ÷ 22 = 0.82.

The point is: B is "cheaper per unit of growth" by ~55% (0.82 ÷ 1.83 - 1).

Step 4: Quantify leverage risk under rates

A: Net Debt/EBITDA = 2.8x and debt service is 23% of operating cash flow.
B: Net Debt/EBITDA = 0.4x and debt service is 4% of operating cash flow.

If the 10-year yield rises +100 bps, you treat A as the more fragile equity claim because refinancing risk hits the equity first.

Step 5: Implement sizing and exits

You allocate $1.4M (70%) to B and $0.6M (30%) to A.
You dollar-cost average over 6 weeks.
You set a stop-loss for B at 22x EV/EBITDA, which is +22% above 18x (22 ÷ 18 - 1).

What the next 3 years look like (quantified scenarios)

Baseline: portfolio +56% to $3.12M; A +18%, B +73%.
Good: portfolio +112% to $4.24M; A +41% (to 12.5x), B +142% (to 22x).
Poor: portfolio -30% to $1.40M; A -36% (to 8.3x), B -28% (to 13.5x).

Common Implementation Mistakes (and what they cost you)

1) You compare P/E across different leverage and think you found value

If you ignore leverage, you can understate the true valuation cost by 72%: across 847 LBO transactions, average pre-acquisition P/E was 18.2x, but post-leverage P/E adjusted for debt service was 31.4x; investors relying on P/E overpaid $2.3M per $10M invested. Fix: you use EV/EBITDA when net debt differs materially, or you compute an "unleveraged" earnings measure.

2) You use trailing multiples at cyclical peaks and buy the denominator

If you buy cyclicals at peak earnings, you mistake 6x trailing P/E for value when normalized P/E is 14x. In autos at the 2018 peak, this error preceded declines of 38% and 29% (Ford and GM) over 18 months. Fix: you normalize with 5–7 year average margins applied to current revenue.

3) You use P/S without correcting for margin, and you invert "cheap" and "expensive"

If you buy 0.5x P/S at 2% net margin, you are effectively paying 25x earnings-equivalent; 1.2x P/S at 8% margin is 15x. That inversion corresponded to 8.3% per year underperformance for "cheap P/S" selections in retail. Fix: you convert P/S ÷ net margin (or margin-adjust relative to industry median).

Implementation Checklist (tiered by ROI)

Highest ROI (do this in 30–60 minutes)

Compute TTM P/E, EV/EBITDA, and P/S for 10–20 comps, then flag ±20% versus sector median.
Apply hard thresholds: P/E < 12x, P/E > 22x, EV/EBITDA < 6x, and SaaS P/S bands (2–4x, 4–8x, 8–15x by growth).
Run a 2-trigger compression check: +100 bps rates → -2.1x growth P/E; 1–2 earnings misses → -8–12% then -15–20% multiple.

Medium ROI (adds 2–4 hours, reduces error by ~35–45%)

Build "warranted multiple" comps using growth (%), profitability (margin %/ROE), and risk (beta) rather than SIC labels; this historically reduced errors 35–45% (Bhojraj & Lee, 2002).
Replace trailing with normalized where cycles exist: use 5–7 year averages for margins and test ±15% EBITDA shocks.

Lower ROI (do this only when sizing is large, e.g., >$1M)

Stress-test the equity claim with leverage: model debt service as % of operating cash flow, and cap position size so a 25% drawdown is survivable by design.
Pre-commit exits: set a multiple-based stop (e.g., 22x from 18x is +22%) and a fundamental stop (e.g., 2 consecutive misses).

The durable lesson

Your edge is not picking the "right" multiple; it's forcing every decision into a quantified triangle: multiple (x), growth (%), and mean reversion time (3–5 years). When you do that, "cheap" becomes a testable claim—not a feeling—and the penalty for being wrong becomes a number you can size.

Valuation Multiples Overview: P/E, EV/EBITDA, P/S