Does the Balance Sheet Change Which Stocks Look Cheap? P/E Against EV/EBIT in Python
September 12, 2026
What’s the question?
A price-to-earnings ratio prices one claim on a company. It divides the market value of the shares by the profit left after interest and tax, so it describes the equity alone. Enterprise value takes the other view: it adds debt to the share value and subtracts cash, pricing the whole operating business whoever funded it. Set that against operating profit, measured before any interest is paid, and the result is EV/EBIT.
Take two companies earning identical operating profit from identical assets. One funds itself with shareholders’ money and holds cash; the other borrows. Interest charges cut the borrower’s net income and raise its P/E, while the borrowed money sits inside enterprise value and raises its EV/EBIT. Which force wins decides whether the borrower looks cheaper or dearer after the swap.
What matters is how often the disagreement is large enough to change a decision, and whether it lands where capital actually gets committed.
The approach
One cross-section of the S&P 500, valued at the 2025 fiscal year end.
- Take the index membership as it stood on 31 December 2025 and carry every member by entity id, so that a recycled ticker cannot swap one company for another.
- Drop Financials and Real Estate, where debt is raw material rather than funding and an enterprise multiple has no clean meaning.
- Keep companies whose latest annual report covers a fiscal year ending between 1 December 2025 and 31 January 2026, putting every valuation within five weeks of one date.
- Require a reported cash balance, a total debt figure, a diluted share count, positive net income and positive operating profit.
- Require the profit line to reconcile: net income within 10% of pre-tax income minus tax. Minority interests, discontinued operations and preferred dividends break that link, so those names sit outside the comparison.
- Value each company at its own fiscal year end, using that day’s closing price times diluted shares, so price and balance sheet share a date.
That leaves 212 companies, each ranked twice; the gap between the two rank positions is the quantity of interest.
Code
import pandas as pd
from scipy import stats
import xfinlink as xfl
xfl.set_api_key("YOUR_API_KEY") # free at https://xfinlink.com/signup
AS_OF = "2025-12-31"
FY_LO, FY_HI = "2025-12-01", "2026-01-31"
FIELDS = ["net_income", "ebit", "pretax_income", "income_tax_expense", "total_debt",
"cash_and_short_term_investments", "weighted_avg_shares_diluted"]
ids = sorted(xfl.index("sp500", as_of=AS_OF)["entity_id"].dropna().astype(int))
a = pd.concat(
[xfl.fundamentals(entity_id=ids[i:i + 60], period_type="annual",
start="2024-06-01", fields=FIELDS, max_rows=200000)
for i in range(0, len(ids), 60)], ignore_index=True)
a = a.dropna(subset=["period_end"]).drop_duplicates(["entity_id", "period_end"])
f = a.sort_values("period_end").groupby("entity_id").tail(1).set_index("entity_id")
f = f[~f["gics_sector"].isin({"Financials", "Real Estate"})]
f = f[(f["period_end"] >= FY_LO) & (f["period_end"] <= FY_HI)]
f = f.dropna(subset=FIELDS)
f = f[(f["net_income"] > 0) & (f["ebit"] > 0) & (f["weighted_avg_shares_diluted"] > 0)]
gap = (f["net_income"] - (f["pretax_income"] - f["income_tax_expense"])).abs()
f = f[gap <= 0.10 * f["net_income"]] # profit line must reconcile
px = pd.concat(
[xfl.prices(entity_id=list(f.index)[i:i + 60], start="2025-12-15", end="2026-02-02",
fields=["close"], max_rows=200000)
for i in range(0, len(f), 60)], ignore_index=True)
px = px.dropna(subset=["close"]).sort_values("date")
d = pd.merge_asof(f.reset_index().sort_values("period_end"),
px[["entity_id", "date", "close"]], left_on="period_end",
right_on="date", by="entity_id",
tolerance=pd.Timedelta("7D")).set_index("entity_id")
d = d.dropna(subset=["close"])
d["market_cap"] = d["close"] * d["weighted_avg_shares_diluted"]
d["net_debt"] = d["total_debt"] - d["cash_and_short_term_investments"]
d["ev"] = d["market_cap"] + d["net_debt"]
d = d[d["ev"] > 0]
d["pe"] = d["market_cap"] / d["net_income"]
d["ev_ebit"] = d["ev"] / d["ebit"]
d["nd"] = d["net_debt"] / d["market_cap"]
d["r_pe"] = d["pe"].rank()
d["r_ev"] = d["ev_ebit"].rank()
d["move"] = d["r_pe"] - d["r_ev"] # positive: cheaper on the enterprise multiple
n, k = len(d), len(d) // 5
q = pd.qcut(d["nd"], 5, labels=False)
cheap = len(set(d.nsmallest(k, "pe").index) & set(d.nsmallest(k, "ev_ebit").index))
dear = len(set(d.nlargest(k, "pe").index) & set(d.nlargest(k, "ev_ebit").index))
print(n, k, d["pe"].median(), d["ev_ebit"].median(), d["nd"].median())
print(stats.spearmanr(d["pe"], d["ev_ebit"]).statistic, cheap, dear, (d["move"].abs() > 50).sum())
print(d.groupby(q).agg(n=("pe", "size"), nd=("nd", "median"), pe=("pe", "median"),
ev=("ev_ebit", "median"), move=("move", "median")))
print(d.nlargest(8, "move")[["ticker", "pe", "ev_ebit", "nd", "r_pe", "r_ev"]])
print(d.nsmallest(8, "move")[["ticker", "pe", "ev_ebit", "nd", "r_pe", "r_ev"]])
print(d.groupby("gics_sector").agg(n=("pe", "size"), nd=("nd", "median"),
move=("move", "median")).sort_values("move"))Full script with formatting and visualisation: pe-versus-ev-ebit-sp500-python.py
Output
Equity multiple against enterprise multiple, S&P 500 at 31 December 2025
Universe: 497 index members on 2025-12-31, carried by entity id;
212 non-financial companies with a fiscal year ending
1 December 2025 to 31 January 2026 and a reconciling profit line
P/E = diluted market capitalisation / net income
EV/EBIT = (diluted market capitalisation + total debt - cash) / operating profit
Median P/E 25.58 median EV/EBIT 21.26 median net debt / market cap +0.15
Spearman rank correlation between the two multiples +0.8726
Cheapest quintile (42 names) 32 shared, 10 replaced
Dearest quintile (42 names) 35 shared, 7 replaced
Companies moving more than 50 rank places: 25 of 212
By net debt quintile (net debt / market capitalisation)
quintile net debt median P/E median EV/EBIT median rank move
1 ( 43 names) -0.01 32.35 25.48 +7.0
2 ( 42 names) +0.06 31.70 25.83 +5.0
3 ( 42 names) +0.15 26.92 21.13 +9.0
4 ( 42 names) +0.26 21.09 18.51 +2.5
5 ( 43 names) +0.70 18.60 20.09 -18.0
Largest moves toward cheap on the enterprise multiple
ticker company P/E EV/EBIT net debt P/E rank EV rank
HUM HUMANA INC 26.05 8.66 -0.24 108 5
SWK STANLEY BLACK & DECKER INC 28.90 17.82 +0.43 129 59
BKNG BOOKING HOLDINGS INC 32.35 19.78 -0.00 154 85
NCLH NORWEGIAN CRUISE LINE HLDGS 25.19 15.49 +1.27 104 41
MAR MARRIOTT INTERNATIONAL INC 32.63 20.42 -0.00 157 95
DPZ DOMINOS PIZZA INC 24.21 15.16 -0.01 99 40
EPAM EPAM SYSTEMS INC 30.50 19.71 -0.11 140 83
SW SMURFIT WESTROCK PLC 29.10 19.21 +0.62 131 75
Largest moves toward expensive on the enterprise multiple
ticker company P/E EV/EBIT net debt P/E rank EV rank
UBER UBER TECHNOLOGIES INC 17.23 31.74 +0.02 40 178
FANG DIAMONDBACK ENERGY INC 26.12 45.69 +0.33 109 198
XEL X C E L ENERGY INC 21.56 29.66 +0.76 79 166
D DOMINION ENERGY INC 16.72 22.35 +0.97 36 121
LNT ALLIANT ENERGY CORP 20.69 26.58 +0.63 67 150
GM General Motors Company 29.34 65.23 +1.40 133 205
PCG P G & E CORP 13.09 19.95 +1.68 16 88
REGN REGENERON PHARMACEUTICALS IN 18.61 23.11 -0.01 55 126
By sector
sector names median net debt median rank move
Utilities 26 +0.75 -35.5
Communication Services 8 +0.50 -4.0
Energy 13 +0.21 -2.0
Health Care 30 +0.14 -0.5
Information Technology 30 +0.03 +0.0
Materials 16 +0.19 +6.5
Consumer Staples 14 +0.20 +12.5
Industrials 52 +0.10 +13.0
Consumer Discretionary 23 -0.01 +16.0What this tells us
The two rankings agree on the broad shape and disagree at the edges. A Spearman rank correlation of +0.8726 says a company near the cheap end on one measure is usually near the cheap end on the other, yet 25 of the 212 move more than 50 places, and ten of the 42 names in the cheapest quintile on P/E leave it once the balance sheet enters.
The quintile table shows where the disagreement lives. Four of the five net debt buckets barely move, with median rank changes of +7.0, +5.0, +9.0 and +2.5 places against a 212-name list. The most indebted fifth, carrying net debt equal to 0.70 times market value at the median, drops 18 places, and that same bucket holds the lowest median P/E in the sample at 18.60. Borrowing lifts earnings per share and pulls the equity multiple down; the enterprise multiple takes that flattery back out. Utilities carry most of the effect, losing 35.5 rank places on median net debt of 0.75, while Consumer Discretionary gains 16.0 places on net debt near zero.
Two individual cases warn against applying the enterprise multiple mechanically. General Motors moves from 133rd to 205th, but most of its $110bn of net debt funds the loan book at its finance arm rather than its factories, and Humana moves from 108th to 5th on a cash pile that backs insurance claims rather than sitting idle. Uber is cleaner, at 17.23 times earnings against 31.74 times operating profit, because a large tax credit lifted net income above what the business itself earned.
So what?
Screen on P/E for the four-fifths of the market carrying little net debt. An enterprise multiple buys a median of nine rank places or fewer there, which will not change a shortlist.
Inside the most indebted fifth the equity multiple is measuring the funding as much as the business, and a low P/E in that group deserves the enterprise check before anything else. That fifth is mostly regulated utilities and a scattering of heavily borrowed operators, so a value screen filling up with utilities has probably found borrowing rather than value.
Two checks belong with the enterprise multiple. Confirm the cash is genuinely spare rather than regulatory float, and confirm the debt funds operations rather than a captive lending book. Either failure moves EV/EBIT for a reason unrelated to the business being valued.
Built with xfinlink — free financial data API for Python. pip install -U xfinlink
pip install -U xfinlink