Should you buy a home with a box spread?
The first post put longs and shorts in one margin account—a brokerage account in which the broker lends against the securities it holds—and read the ledger line by line. This one adds options to the account, and with them a second equity. A short box spread is a loan written in four index options: sell it and you receive the present value of a fixed sum today, and owe the sum itself, in cash, on a fixed date. The option market sets the rate: since 2004, a median 35 basis points over the one-year Treasury, which, on the schedules read in September 2026, made it about half the list margin rate at a full-service retail broker. That is why the box now appears in advisors' securities-based-lending comparisons, and why the title's question is being asked. The question turns on three things the box does inside a margin account: what the margin rules hold as collateral under Reg T (the full face of the loan, not the proceeds), what a typical balances page counts as equity (the cash you received, but not the debt you took on), and what those two facts do to anything else sharing the account. None of this needs a long-short book. A box needs a margin account approved for spreads; a long-only account of index funds is the plain case, and the one the home example uses. The long-short case is here because it is where the accounting misleads most. No option experience is assumed: the first section builds the instrument from two payoff diagrams, and every term of art is defined where it first appears. The worked example is priced from four option quotes at one close, the history behind it from twenty-two years of the same quotes; the rest follows from the rules. It is not advice about whether to borrow, or how.
(General disclaimers apply.) I am not your accountant, broker, or financial advisor, and this is not investment, legal, or tax advice. Nothing here recommends any instrument, account structure, or borrowing; the examples exist to show how the margin rules and the balances page interact, and the rates and schedules quoted were read on the dates given and will have changed since. Options involve risk and are not suitable for everyone; read the Characteristics and Risks of Standardized Options before trading them. The tax treatment sketched below is a simplified summary of intricate, fact-specific rules whose application to box spreads is debated among practitioners; it is not something to rely on. Margin rules vary by broker and change without notice; your custodian's schedule and margin agreement govern.
What you should know
Parity
(A call, a put.) An option is a contract on the level of an index at a fixed future date, the expiry. Write \(S_T\) for the S&P 500 on the expiry date \(T\), and \(K\) for the strike, the level the contract is written at. A call pays \(\max(S_T - K, 0)\) on that date: whatever the index finishes above \(K\), or nothing. A put pays \(\max(K - S_T, 0)\): whatever it finishes below \(K\), or nothing. Each payoff is a hinge at \(K\)—flat on one side, a line of slope one on the other. Buying either costs a premium today, and the buyer is said to be long the option. Selling either collects the premium and takes on the payoff as a debt; the seller is short, and holds the buyer's line with the sign reversed. The options here are European, meaning the payoff is settled once, at expiry; an American option can be exercised at any time before, which will matter later.
Payoff at expiry, one strike K
S_T < K S_T >= K
long call 0 S_T - K -- the upside above K
long put K - S_T 0 -- the downside below K
short call 0 -(S_T - K) -- the same lines, owed
short put -(K - S_T) 0
(Two options, one forward.) Hold the call long and the put short, both at \(K\), and whatever the index does you end up buying it at \(K\) on date \(T\). Above \(K\) you exercise the call and pay \(K\) for something worth \(S_T\); below \(K\) the put is exercised against you and you pay \(K\) for the same thing. Either way you have bought at \(K\) what is worth \(S_T\), a payoff of \(S_T - K\): the two hinges join into one straight line through \(K\). That is a forward—a commitment made today to buy the index at \(K\) on date \(T\), nothing paid now, the difference settled then—built from two options with no forward market involved. From here on the argument is about forwards, because a box is two of them. Reverse both legs, sell the call and buy the put, and you are committed to sell at \(K\) instead: \(K - S_T\), the same forward, sold.
Long call, short put, same K
S_T < K S_T >= K
long C(K) 0 S_T - K
short P(K) -(K - S_T) 0
-----------------------------------------
total S_T - K S_T - K -- a forward bought at K: linear in S_T
What is the forward worth today? Its payoff at \(T\) is \(S_T - K\). Price the two pieces separately, each by building it from something that trades now.
The first piece, \(S_T\), is a claim to the index at \(T\). Write \(S\) for the index level today. Buy the index for \(S\) and hold it: at \(T\) you have \(S_T\), and along the way you also collect the dividends, which the claim does not include. So the claim is worth a little less than \(S\)—the index less the dividends forgone, \(S e^{-qT}\) for an index yielding \(q\). This is the prepaid forward. Notice what became of \(S_T\). Nobody knows where the index will finish, but a claim on wherever it finishes has a price today, and that price is what the payoff is worth.
The second piece, \(-K\), is a fixed payment at \(T\). It is worth what you would put in Treasuries today to have \(K\) then: \(K e^{-rT}\), the strike times the discount factor at the riskless rate \(r\).
The two options together cost what the two pieces cost: were they cheaper, buying the options and selling the pieces would keep the difference without risk; were they dearer, the reverse would. So
\[C(K) - P(K) = S\,e^{-qT} - K\,e^{-rT}\]This is put-call parity: the forward built from two options costs what the forward built from the index and a Treasury costs.
(Two forwards, one box.) Now do it twice, at \(K_1 < K_2\), in opposite directions: buy the forward at \(K_1\), sell the forward at \(K_2\). At expiry the position pays \((S_T - K_1) - (S_T - K_2) = K_2 - K_1\) whatever the index does, and in parity the index terms cancel:
\[\mathrm{Box}(K_1, K_2) = \big[C(K_1) - P(K_1)\big] - \big[C(K_2) - P(K_2)\big] = (K_2 - K_1)\,e^{-rT}\]A long box pays \(K_2 - K_1\) at \(T\) with certainty and costs its present value today: a zero-coupon bond. Its obligor is the clearinghouse—the Options Clearing Corporation stands between every buyer and seller of a listed option, so the promise runs through the clearinghouse and one's own broker, not through whoever took the other side. A short box is the mirror—sell the forward at \(K_1\), buy it at \(K_2\), receive the present value now, pay \(K_2 - K_1\) at \(T\)—a zero-coupon loan. The index never enters. The only thing being priced is \(r\). The name is the picture. A broker's option chain is the table of every option on one underlying: calls in one column and puts in the other, strikes down the rows. The four legs of a box are the four corners of a rectangle two strikes tall and two types wide, and inside that box the payoff is flat.
Short box 7000/8000, per contract, at expiry (index points)
S_T < 7000 7000 <= S_T < 8000 S_T >= 8000
short C7000 0 -(S_T - 7000) -(S_T - 7000)
long P7000 +(7000 - S_T) 0 0
long C8000 0 0 +(S_T - 8000)
short P8000 -(8000 - S_T) -(8000 - S_T) 0
----------------------------------------------------------------------
total -1000 -1000 -1000
forward sold at 7000 + forward bought at 8000 = a constant
x $100 x contracts = F, the face: owed on the expiry date, cash-settled, automatic
(Why SPX.) SPX is Cboe's option on the S&P 500 index itself, and four of its properties are load-bearing. European exercise: the seller of an American option can be assigned—made to perform—whenever the buyer chooses, and a European one only at expiry, so no counterparty can pull one leg out early and leave you holding the other three. Cash settlement: the payoff is paid in dollars against an index level fixed at settlement, so there is nothing to deliver and no pin risk, the last-day uncertainty over whether a contract finishing near its strike will be exercised against you.The standard SPX contracts, the December expiries here among them, settle in the morning of the third Friday to the special opening quotation, an index level built from each constituent's opening print; the last day to trade them is the Thursday before. The weeklys, SPXW, settle to the close. A $100 multiplier: each index point is $100 a contract, so a 1,000-point box is $100,000 of face and a mortgage-sized loan is a handful of contracts. Deep markets in wide boxes, because both sides are pricing \(r\) and neither cares where the index is. And the tax treatment bears on the rate. Broad-based index options are Section 1256 contracts—marked to market at year-end, with gains and losses 60% long-term and 40% short-term capital—so the lender may book a capital gain rather than interest, one reason the rate sits near Treasuries. The conversion-transaction rule, which treats a pure time-value return as ordinary income, may reach the trade instead; whether it does is disputed. On the same sketch, the borrower's cost arrives as a net mark-to-market capital loss across the four legs, not as margin interest. (More on the tax below.) Cboe presents the instrument to advisors as a securities-based-lending alternative, and not, in its words, as one "for long-term financing or lifestyle leverage."Cboe Insights, March 2026, puts short box spreads "in the comparison matrix for any securities-based lending discussion, whether a margin loan, pledged asset line or similar collateralized borrowing solutions." The sentence does not say where the requirement lands in dollars, though the page notes that Reg T requirements "can consume a large amount of collateral"; it also names the account advisors "frequently" prefer—portfolio margin—and the margin section below shows what each rulebook asks of the box.
Pricing one
Take five 7000/8000 boxes with fifteen months to run—the December 2027 expiry—sold at the close of 17 September 2026 with the index at 7,637.76. A mark is the price the broker carries a position at after the close—taken here as the midpoint of the closing bid and ask; brokers set their own—and a leg's value is its mark times the $100 multiplier times the contracts, negative for the legs you sold, because closing them would cost that much. These are the four:
Five SPX 7000/8000 boxes, Dec 2027 expiry, sold at the close of 17 Sep 2026
the mark is the midpoint of bid and ask; value = mark x $100 x contracts
leg contracts bid / ask mark value
Dec-27 7000 C -5 1,235.90 / 1,248.50 1,242.20 -621,100 -- sold: owed
Dec-27 7000 P +5 267.60 / 274.30 270.95 +135,475 -- bought: owned
Dec-27 8000 C +5 570.50 / 578.80 574.65 +287,325 -- bought: owned
Dec-27 8000 P -5 538.60 / 546.70 542.65 -271,325 -- sold: owed
------------------------------------------------------------------
net -469,625 -- P: to close all four
F = (8000 - 7000) x 100 x 5 = 500,000 T = 456 days = 1.249 y
DF = P / F = 0.9393
r = DF^(-1/T) - 1 = 5.15% a year -ln(DF) / T = 5.02% continuous
(Reading the marks.) Four marks make one price only if they agree with each other, and parity says how to check. At each strike \(C - P\) is the same prepaid forward less the strike times the same discount factor: 971.25 at 7000 and 32.00 at 8000, a difference of 939.25 for 1,000 points of strike, so \(\mathrm{DF} = 0.9393\). Put the discount factor back and both strikes return the same prepaid forward, 971.25 + 7000 × 0.93925 = 7,546.0, next to a spot of 7,637.76: the index less 1.2% of dividends over fifteen months, 0.97% a year, where the trailing yield is 1.08%. The other strike pairs quoted that day, 4000 to 9000, return 5.09% to 5.15%. Two tests, then: whether the rate sits near the Treasury curve, and whether the dividend yield the same marks imply is the index's. Each point of error in any one mark moves the rate by about 9 bp and the implied yield by about 8 bp, so the tests are equally sensitive to the marks; they differ in how well the reference is known.With \(r = \mathrm{DF}^{-1/T} - 1\) and \(\mathrm{DF}\) the box price over 1,000, a point on any leg moves \(\mathrm{DF}\) by 0.001 and \(r\) by \(\mathrm{DF}^{-1/T - 1} / (1{,}000\,T)\), 9.0 bp here. The same point moves the prepaid forward \(C - P + K_1\,\mathrm{DF}\) by \(1 + K_1/(K_2 - K_1) = 8\) points for a 7000 leg (\(K_1/(K_2 - K_1) = 7\) for an 8000 leg), and \(q = -\ln(\mathrm{PF}/S)/T\) by \(8/(\mathrm{PF}\,T)\), 8.5 bp. A fair box rate is known only to within a percentage point or so—somewhere from Treasuries to a point and a half over them—while the index's yield is published to the tenth. Five points of error in one mark, out of 1,242, keep the rate inside the plausible band and move the dividend yield four-tenths of a point—to the edge of its own band or past it. When the marks fail either test the rate is noise; when they pass both, it is a price.
The same day's Treasury curve read 4.40% at one year and 4.67% at two: 4.47% at the box's fifteen months, at par, or 4.52% as an effective annual rate, the basis the box's 5.15% is quoted on—so the box is about 63 bp over Treasuries.Constant-maturity Treasury yields are bond-equivalent, compounding twice a year: 4.47% at par is \((1 + 0.0447/2)^2 - 1 = 4.52\%\) a year. A SOFR swap pays its fixed leg once a year on an actual/360 count, so a quoted 4.476% is \(4.476 \times 365/360 = 4.54\%\). The box is quoted with annual compounding, so the three sit on one basis. Treasuries are FRED's DGS1 and DGS2; the swaps Bloomberg's USOSFR1 and USOSFR2. The wholesale term rate is the SOFR swap, the fixed rate exchanged for a year or two of overnight SOFR; it read 4.45% at one year and 4.55% at two, 4.54% at fifteen months as an effective rate, so the box is 61 bp over that. Overnight SOFR itself, the rate at which cash is lent overnight against Treasury collateral, was 3.85%.
(Margin rates.) A margin rate is the annual interest a broker or custodian charges on a debit balance, the cash a client borrows against securities held in the account. It is quoted as a base or reference rate plus a spread, accrues daily, usually on a 360-day year, and posts monthly, so it floats with policy rates. The quote is therefore not yet on the box's basis: interest charged day by day and added to the debit compounds, and a schedule's 10.325% is \((1 + 0.10325/360)^{365} - 1 = 11.03\%\) over a year, the annual-compounding figure the box's 5.15% already is.The 360-day count alone adds \(365/360 - 1 = 1.4\%\) of the rate, 14 bp on 10.325%; daily compounding adds the other 56. The same arithmetic takes prime-broker margin from 4.08% to 4.22% and the custodian program's 4.32% to 4.48%; the ledger below carries both columns. Interest actually posts monthly, so the exact figure is \((1 + r \times 30.4/360)^{12} - 1\), 10.98% here, a few basis points under the daily statement; the difference does not matter for anything that follows. The references are the overnight rates the Fed publishes each morning. The Fed funds target is the range the FOMC sets for banks lending reserves to one another overnight, 3.75% to 4.00% from 17 September, the FOMC having raised it a quarter point at the meeting that ended the day before; several schedules key off its upper bound. EFFR, the effective Fed funds rate, is the volume-weighted median of what banks actually paid, 3.88% that day, a little under the top of the range. SOFR, above, is the secured cousin, and the one bank lines use. The large full-service retail brokerages publish their own base rate, set at their discretion and currently about six percentage points above the Fed funds target, and add balance-tiered spreads on top, which puts their published margin rates at roughly 10% to 13% for balances under half a million dollars; the discount brokers' lists run near 5%. Advisor and institutional accounts are priced differently: custodians, the brokers that hold client accounts for registered investment advisers—RIAs, the fee-charging firms that manage money without holding it—negotiate margin terms with each advisory firm, strategy-level programs price directly off the Fed funds target upper bound, and prime brokers, the bank desks that finance and hold hedge funds' positions, price off EFFR, with spreads measured in tens of basis points rather than percentage points.Published schedules as read on 24 September 2026: Schwab, Fidelity, Interactive Brokers, Robinhood, Altruist, Goldman Sachs Custody Solutions, BNY Pershing, Vanguard, E-TRADE and Schwab Bank's pledged-asset line. Pershing clears for introducing firms, brokers that keep the client relationship and rent Pershing's back office, each adding its own spread to Pershing's base. A pledged-asset line is a bank loan secured by a brokerage account rather than a margin loan inside one, which is why it prices off SOFR like other bank credit. The prime-broker, program and advisor-negotiated rates come from confidential schedules and correspondence and are left unnamed. That day the Fed funds target upper bound was 4.00% (FRED's DFEDTARU), the effective rate 3.88% and SOFR 3.87%, the print for the day before. In a long/short portfolio at a prime broker the debit is largely offset by short-sale proceeds, which earn a short interest rebate—the broker pays interest on the cash a short sale brings in—quoted as Fed funds minus a spread; the economic cost of financing is therefore the gap between the margin rate and the rebate, applied to the overlay, not the headline margin rate. Against that range the box's 5.15% midpoint sits five to seven points below the full-service retail lists and two-thirds of a point to a point above the prime-broker and program rates, before the cost of crossing the quote.
Fifteen months, 469,625 borrowed. Rates restated: annual compounding, the box's basis
quoted effective cost over the term
short box, implied 5.15% 5.15% 30,400
SOFR swap at 1.25 y 4.48% 4.54% 26,800 -- the box is +61 bp
Treasury at 1.25 y 4.47% 4.52% 26,700 -- the box is +63 bp
prime-broker margin, EFFR + 20 bp 4.08% 4.22% 24,900 -- the box costs 5,500 more
custodian program, target + 32 bp 4.32% 4.48% 26,400 -- the box costs 4,000 more
retail list, best published tier 10.325% 11.03% 65,600 -- the box costs 35,200 less over the term
(What the market has charged.) One close is one close; the record is longer. Minute-level SPX quotes, assembled by van Binsbergen, Diamond and Grotteria, give a daily box rate from 2004 to March 2018, and closing quotes on the December expiries carry it to today; where the two overlap they agree to within seven basis points.Jules van Binsbergen, William Diamond and Marco Grotteria, "Risk-free interest rates", Journal of Financial Economics 143 (2022); their daily series, deposited at LBS, runs from January 2004 to March 2018 and is continuously compounded, converted here to annual. From January 2018 the series is Bloomberg's closing bid and ask on SPX December options at the 1,000-point strikes: for each expiry the median rate over every strike pair with all four legs quoted, interpolated to one year, taken as weekly medians. The touch is what selling all four legs at their quoted bid or ask would have implied, smoothed over five weeks. Treasury yields, SOFR, prime and the mortgage rate are FRED's DGS1, DGS2, DGS5, SOFR, DPRIME and MORTGAGE30US. Over the 1,146 weeks since January 2004 in which a box was quoted, the one-year box has priced a median 35 bp over the one-year Treasury—19 to 55 bp four weeks in five; 37 bp before 2018 and 32 since. Against the one-year SOFR swap, which exists from 2018, the median is 36 bp, 17 to 48 four weeks in five: the box has borrowed at the wholesale term rate plus a third of a point. The spread over the Treasury is a convenience yield—the premium paid for a Treasury's safety, liquidity and usefulness as collateral—and it widens when safe assets are scarce: 148 bp in the week of 17 October 2008, when Treasury yields collapsed and box rates did not. In March 2020 the midpoint held 33 bp over a Treasury at 0.2%, but selling the four legs at their quoted bid and ask would have cost over 3%—the market was there at mid, and only there. The only weeks the box priced below the Treasury were the last of May and the first of June 2023, by one and five basis points, when bills cheapened into the debt-ceiling deadline. And the market has grown: open interest in the December SPX options at the nine round strikes went from 176,000 contracts in March 2018 to 3.3 million in September 2026, calls and puts in near-equal numbers at each strike from 4,000 to 8,000—the footprint of boxes, which need one of each.
(Three cautions.) The first is about the marks. Deep in the money an option's price is almost all intrinsic value—the 7000 call is 638 points in—and it trades a few times a day, so the day's last trade says little: over the fortnight around this close the last trades on the four legs implied anything from 4.4% to 6.5%, while the midpoints, on the days all four legs were quoted tight, stayed between 4.9% and 5.2%. Nor are the quotes narrow. The four legs' touch—selling each at its bid, buying each at its ask—implied 6.8% to sell the box that day and 3.6% to buy it. A package order is quoted well inside that width,Cboe's complex order book takes the four legs as one order at one net price. The market makers who quote it are pricing \(r\), not four options, so the package is quoted far inside the legs' combined width, which adds up to three hundred basis points. The touch in the figures is therefore a ceiling on the cost of crossing, not an estimate of it. but the rate paid is the one on the confirm, not the one on the screen.
The second is duration. The box is a bond, so it has a DV01, the change in its value for a one-basis-point move in the rate: $56 here. If the fifteen-month rate has fallen since the sale, closing early costs $56 a basis point; if it has risen, closing early shows a gain of the same size.
The third is bookkeeping. The interest never appears as interest. The marked liability walks from 469,625 to 500,000 over fifteen months, about $2,025 a month, and it arrives as a widening net loss across the four legs: interest in the costume of a mark-to-market loss and, on the tax sketch above, generally a capital loss rather than an interest deduction—usable against capital gains and otherwise only $3,000 a year against ordinary income, the rest carried forward, or back by election against earlier Section 1256 gains.
Margin: three regimes
(Three rulebooks.) A margin account is scored three ways at once; a reader of the first post can skip this paragraph. Reg T is the Federal Reserve's initial-margin rule—50% of a stock's value at purchase—extended to options by exchange rules that assign each position, or each recognised combination of positions, a requirement from a schedule. This is strategy-based margin, and a spread's requirement—for a box, its maximum loss—is the same on the day it is opened and the day before it expires. Maintenance margin is the floor the account must keep meeting afterward—FINRA's 25% of long stock and 30% of short, with brokers' own house rates on top. Portfolio margin replaces the schedule with a stress test: the account's positions are repriced across a range of moves in their underlyings and the requirement is the worst loss. A cash account borrows nothing at all. The box lands differently in each.Reg T is 12 CFR Part 220, the 50% in §220.12; the maintenance percentages are FINRA Rule 4210(c); portfolio margin is 4210(g). Section 1256 is 26 U.S.C. §1256, under which options on broad-based indexes are nonequity options, marked to market at year-end with 60% of the gain or loss long-term whatever the holding period.
A short box of face F: what each rulebook asks for
regime requirement why, and what follows
Reg T (strategy-based) F, flat the spread's maximum potential loss;
marks irrelevant, maintenance = initial
portfolio margin ~ 0 no market risk for the stress to find;
$37.50 a contract, the rule's floor
cash account F, in cash European, cash-settled only; the loss
sits on deposit: borrowing P to freeze F
long European box 50% of F the lender's side, with loan value;
the bond can itself be borrowed against
(Reg T.) FINRA 4210(f)(2)(H) margins a spread—a combination of long and short options on one underlying with limited risk—at the lesser of the requirement the short options would carry on their own and the spread's maximum potential loss; the long options are paid for in full, and the proceeds of the shorts may be applied. For a box the maximum loss is the strike difference times the multiplier times contracts—the face. Cboe's margin manual gives the short box its own row with no lesser-of: in a margin account the requirement is the maximum potential loss—for a box, the aggregate difference between the exercise prices—with the long legs paid in full, the short proceeds applied, and the initial requirement maintained. So the requirement is $500,000 the day you sell and $500,000 the day before expiry. It does not accrete, it does not mark, and it is not a call. It is the loan's collateral, posted in full, on day one.
You borrowed \(P = 469{,}625\) and posted \(F = 500{,}000\) of Reg T capacity. A margin loan of \(P\) would have consumed \(P\). The difference, \(F - P = 30{,}375\), is the term's interest, collateralized in advance.
(Five legs, two equities.) The first post's ledger had four legs—long stock, short stock, cash held, cash owed—and one equity. Options add a fifth. Call it \(OMV\): the options at their mark, which for a short box is what it would cost to close them, \(P\) today and \(F\) at expiry—a liability. At the large retail brokers the balances page prints a margin equity built from the four stock-and-cash legs alone, \(ME = LMV + CR - SMV - DR\), and options are not in it; conventions differ by firm, and the rules let a listed option with more than nine months to run be given market value, so the account's own page governs. What you own subtracts the options: \(E = ME - OMV\). The ledger below fixes the notation and the colours; every figure in this post uses them.
Read the ledger as a balance sheet. The left column is what the account holds: long stock and cash. The right column is every claim on it: the stock owed back on the shorts, the debit owed to the broker, the options at the cost of closing them, and the residual, \(E\), which is whatever makes the two columns equal. Equity is never a sum of parts; it is what remains after every claim is subtracted, so adding a liability lowers it. Margin equity stops one line short: it subtracts \(SMV\) and \(DR\) but not \(OMV\), so \(ME = E + OMV\). Sell a box and the proceeds land in cash, raising the left column, while the debt lands in \(OMV\), a line the page leaves out of its sum. The page reports the left column's gain and none of the right column's claim. With the box marked at \(P\), \(ME = E + P\): the equity the page runs its Reg T tests against is your true equity plus the money you borrowed.
(What the page prints.) The Reg T excess is margin equity less the Reg T requirement: the account's spare capacity, and twice it is the buying power the page shows for stock, give or take the special memorandum account, Reg T's running ledger of past excess. A negative excess created or increased by the day's transactions is a Fed call—money due within the Reg T payment period (the settlement cycle plus two business days), or positions sold. The 469,625 you owe reaches this arithmetic only through the requirement, as \(F\); the proceeds reach it as cash, inside margin equity. Write the excess in those terms, with \(G = (L + S)/E\) the gross stock over true equity:
\[\mathrm{Ex\,EQ} = ME - \tfrac{1}{2}(L + S) - F = \Big(1 - \tfrac{G}{2}\Big)E - (F - P)\]Three consequences. First, relative to a margin loan of \(P\), the box costs \(F - P\) of Reg T capacity—30,375 here—shrinking to zero as \(P\) accretes to \(F\). Second, a box never creates withdrawable cash. On the day you sell, the proceeds add \(P\) to margin equity and the requirement adds \(F\), so excess falls by \(F - P\): your buying power goes down when you borrow, which surprises people the first time. Third, when you then withdraw \(P\)—the point of the loan—excess falls by \(P\) more, \(F\) in total, against \(P\) for a margin loan. \(G\) is the gross over \(E\); the page prints \(ME\); a book sized off "cash plus stock" is sized off \(E + P\).
(Portfolio margin.) The OCC's TIMS model reprices the whole account at ten index levels between −8% and +6% and charges the worst loss. Stressing the four legs together it finds nothing: each level pays \(K_2 - K_1\). What remains is the rule's per-contract minimum—$0.375 times the multiplier, $37.50 per SPX contract—which is economically zero.The floor is 4210(g)(7)(B); the OCC's TIMS does the stressing. A broker may still apply a house add-on to boxes; its schedule says whether it does. Portfolio-margin equity carries the options at their mark, so the tested equity is \(E\) and the loan is simply \(P\) of equity gone—a margin loan of \(P\) without the debit line. Here the test charges \(P\) of equity and no \(F\); it is also the account Cboe's advisor piece reports advisors "frequently" prefer.
(Cash account.) A cash account may carry a spread only in European-style, cash-settled index options—SPX qualifies—and only with the maximum loss on deposit. The short box's requirement is the same \(F\), held in cash or cash equivalents: selling it there borrows \(P\) in order to have \(F\) frozen, so the loan funds its own collateral and nothing leaves.
(The lender.) The long box is a zero-coupon bond with the clearinghouse as obligor, and it is margined like one. 4210(f)(2)(H)(v)e sets a long box in European-style options at 50% of the aggregate strike difference and lets the position count at up to 100% of that difference for margin purposes: loan value on a position built from options, which an option with nine months or less to run never has. The lender's side of the trade is a bond in a margin account that can itself be borrowed against—a description that also fits a Treasury.
What you might want to know
The box inside a long-short account
The box itself involves no short stock: a long-only account with \(LMV \ge 2F\) has the Reg T capacity to sell it, withdraw the proceeds, and carry it, and the home section works that case. This section puts the box beside a long-short overlay—a set of long and short stock positions run on top of the funding stock—because that is where the balances page misleads.
The first post's mixed-netting section said the margin account aggregates: one equity, three rulers. It does—over what it can see. The page's margin equity is defined over stock and cash; the box's debt lives in a line the page does not add. Put a box and a long-short overlay in the same account and every party in the room sizes the book on a different number.
A stylized account, per 100 of true equity: hold 75 of stock and 25 of cash. Sell a box for 25, at 5% for a year, so \(F = 26.25\).
The target is a 75%-net book with a 45/45 market-neutral overlay: 120 long, 45 short, 25 cash—of the basis. Net is long less short over equity, so 120 against 45 is 75% net, and an overlay of 45 long against 45 short adds none. Which basis?
One target, two bases: 120 / 45 / 25 of ...
... ME = 125 ... E = 100
LMV 150.00 120.00
SMV 56.25 45.00
gross stock 206.25 165.00 -- 206% of true equity, not 165%
Reg T requirement .5 gross + F 129.38 108.75
margin equity ME 125.00 125.00
Reg T excess (4.38) 16.25 -- Fed call on the fill | clears
cure by liquidation 8.75 of longs sold -- -- 2 x the deficiency, at 50%
maintenance requirement 80.63 69.75 -- .25 L + .30 S + F
maintenance excess 44.38 55.25 -- no maintenance call either way
(Read the two columns.) Sized on margin equity, the book is 25% larger than intended in every line, and the requirement carries an \(F\) the basis knew nothing about. The account is in a Fed call the moment the last order fills—with 44 of maintenance excess. That is a Fed-call regime, not a maintenance-call regime: a different clock (the Reg T payment period), a different cure (twice the deficiency by liquidation, since selling \(X\) of longs turns stock into cash, leaves margin equity unchanged, and lowers the requirement by only \(X/2\)—a cure 4210(f)(7) bars as a practice), and a different label on the balances page (a Fed call, not a house call). Sized on true equity, the same target clears with 16 to spare.
Carry the same oversized book on a margin debit instead: margin equity 100, requirement 103.13, excess (3.13). The box's marginal contribution to the hole is 1.25—\(F - P\), the year's interest. Of the 4.38 deficiency, 1.25 is the box; the rest is the basis.
(Three equities for one account.) Where a box lives, three account values can circulate. The custodian's, options included—\(E\), and correct. A reporting layer that drops option rows because they do not fit a stock-and-cash schema—\(ME\), inflated by \(P\), and the one an optimizer sizes from. And a layer that carries a sold lot on a settlement-date basis while the cash side already holds the proceeds—\(E\) plus the sale, a double count to the cent. A client statement shows the first, a risk report the second, an order blotter the third. One sentence disposes of the second: an equity that excludes an option row is not an equity.
(Sizing, in numbers.) Every option row in the holdings enters the arithmetic. For each short box, \(F\) = width × multiplier × contracts and \(P\) = the net mark. The sizing basis in the example is \(E = ME - P\), not cash plus stock. A pre-trade check of the form \(\tfrac{1}{2}(L + S) + F \le (1 - c)\,ME\) carries a cushion \(c\); the example uses 5%. For a 75%-net book with an \(x/x\) overlay sized on \(E\), the check solves in closed form:
\[x_{\max} = 0.625 - c - \frac{F - (1 - c)\,P}{E}\]which is 55 at \(c = 5\%\) in the example. Sized on the inflated basis with \(F\) in the check, \(x_{\max} = 0.625 - c - F/ME\), which is about 41 at \(c = 0\): why 45/45 broke. Sized on the inflated basis with \(F\) not in the check, 62.5: why 45/45 looked safe.
(Where the box lands.) Under portfolio margin its location changes no test: it stresses to nothing and \(E\) is what is scored. Under Reg T the tests separate by account. A box sold against the stock in a long-only account opens, and lets the proceeds out, when \(LMV \ge 2F\), since the Reg T excess after the withdrawal is \(\tfrac{1}{2}LMV - F\); it holds against a house maintenance rate \(h\) while \((1 - h)\,LMV \ge F\). An overlay in a separate margin account is tested on its own equity. That is the first post's multiple-accounts architecture: the overlay's tests do not see the box, the box's account does not see the overlay, and a 4210(f)(5) consolidation consent, where one is signed, re-links the maintenance test only, not the Reg T test that bites here, which runs account by account. Reg T §220.4(a)(2) lists the cases in which a broker may open a second margin account for the same person, among them an account over which the broker or a third-party adviser has investment discretion.
A down payment against index funds
An account holds $2,000,000 of index funds and its owner needs $500,000 for a down payment. Four routes:
Raising 500,000 against 2,000,000 of index funds
(cost per year at each quoted rate, simple, on 500,000, before tax, points or fees;
mortgage: Freddie Mac's 30-year average, the week of the close;
tax lines: a simplified sketch, not advice)
cost / yr collateral clock
sell 500k of stock the gain none none
tax: capital gain, today
mortgage at 6.95% 34,750 the house 30 y, amortizing
tax: interest may deduct if itemizing, to the cap
retail margin at 10.3% 51,500 the stock, at h demand
advisor margin at 4.45% 22,250 the stock, at h demand; negotiated, not listed
tax, both: traced to a residence, generally personal, not deductible
short box at 5.15% 25,750 the stock, at h, for F T, then roll
tax: section 1256 capital loss, 60/40, marked each year; offsets gains, else $3,000 a year
The numbers run both ways: twenty-six thousand against thirty-five for the mortgage and fifty-two for a retail margin loan—but twenty-two for margin priced the way custodians price it for advisory firms, and less at a prime broker. Institutional margin sits below the box; the full-service retail lists sit five to seven points above it. At the midpoint the box has priced close to the wholesale rate whoever sells it—a median 36 bp over the one-year SOFR swap since 2018, and 61 bp at the September close used here—because the other side is a bondholder pricing a zero coupon cleared at the OCC, not a bank pricing a loan; the rate on a confirm sits above the midpoint by the cost of crossing. Whether an after-tax gap survives is a different question, and one for a tax adviser rather than this post: the mortgage's interest may deduct, within the acquisition-debt cap; the margin loan's, if its proceeds trace to the residence, generally will not; the box's cost is, on the sketch, a Section 1256 capital loss marked each year, usable against capital gains and otherwise only $3,000 a year against ordinary income, with the straddle rules in the neighbourhood when index funds sit in the same account. Nothing below depends on the tax answer.
(The collateral is not the house.) A mortgage lender does not mark your house nightly and cannot sell it because the market fell. Your broker does and can. Under strategy-based margin the box is a requirement of \(F\) against stock the broker will only count at \((1 - h)\) of its value, where \(h\) is the house maintenance rate—FINRA's floor is 25%, 25% to 30% is typical on index funds, and the example uses 30%. The maintenance test after a drawdown \(d\) is \((1 - h)\,LMV\,(1 - d) \ge F\), so the call arrives at
\[d^{*} = 1 - \frac{F}{(1 - h)\,LMV}\]With \(h = 30\%\): 29% if the stock is twice the face, 64% at four times, 62% in the example (\(F = 525{,}750\) against \(2{,}000{,}000\)).A guest piece on Cboe Insights by the cofounder of a box-spread platform, updated March 2025, arrives at the same line from the other end: for a client who borrows the maximum 50% under Reg T, "a margin call can be prevented if their portfolio declines by less than 28.5%". That is the twice-the-face case. The S&P 500 fell 56.8% on closing prices from 9 October 2007 to 9 March 2009, and took four more years to regain the high; since 1950 it has sat 20% or more below its high on one trading day in six, and 30% or more on one in twenty.Daily closes from Bloomberg's SPX Index history, which begins on 30 December 1927. Falls are measured close to close, so the intraday lows of October 1987 and March 2020 sit a little deeper than the figures here; a maintenance call is issued on the close, so the close is the series used here. Under portfolio margin the box's requirement is the floor, but the equity carries the debt, so the line is \((1 - s)\,LMV\,(1 - d) \ge P\) for a stress \(s\)—the rule's 8% for a high-capitalization broad-based index product; for a fund that tracks one, brokers set their own, and Schwab publishes 15%—about 71% here at 15%, before add-ons. The mortgage has no line at all.
(The loan is a bullet.) Nothing is repaid until the face falls due in full at expiry—a bullet—and then it is paid in cash or refinanced: a roll, a new box sold to raise the \(F\) owed, at whatever the one-year rate is then. The one-year Treasury went from 0.4% to 4.7% during 2022. Or the box is longer and carries duration: a five-year box on a 500,000 face has a DV01 near $185, so a 150 bp fall in rates after the sale makes an early close cost about $29,000—the asymmetry a mortgage's prepayment option removes, at a price inside its rate. Term has a price. The mortgage's thirty years—term, prepayment option and all—sit about 180 bp a year above the box; the box's one year carries the roll. The record, with neither side carrying points, fees or the tax difference, and the roll a constructed series—five one-year loans at the Treasury rate plus the box's 35 bp, not a box price before 2004—set against the first five years of a thirty-year mortgage fixed at the start, the lock: from 89% of the start months since 1971 the roll came out under the lock, and over all start months it ran a median 252 bp a year below it.For each start month from April 1971: the lock is Freddie Mac's Primary Mortgage Market Survey rate for the month, a 30-year conforming fixed-rate loan; the roll is the one-year constant-maturity Treasury, as an effective annual rate, averaged over the start month and the same month in each of the next four years, plus 35 bp. Neither side carries points, fees or the tax difference, and the roll ignores the touch. The mortgage survey begins in 1971, which sets the start. The exceptions are mostly the months that ran into rising rates—every start month from September 1976 to October 1979, and every one from July 2020 to February 2022, when a mortgage fixed at 2.7% to 3.8% beat rolling into the hikes by up to 100 bp a year—plus a handful of scattered months in 1971–72, 1976 and 1980 where the roll lost by less than a point. From March 2022, the last start months with five readable years, the roll came out under the lock again.
The proceeds land in the sweep—the bank deposit or money-market fund the broker parks idle cash in—looking like any other credit balance, and in margin equity as cash; the balances page prints both. The line that records the claim is the Reg T excess, which fell by \(F - P\) at the sale. In a long-only account with \(LMV \ge 2F\) the rule lets \(P\) be withdrawn—because the stock's loan value was there already. The box did not manufacture the withdrawal; it financed one the stock was always good for, changed the rate, and changed the size of the claim recorded against you: \(F\), not \(P\).
And proceeds left in the account: a box financing cash is a loan at 5.15% funding a deposit at the sweep rate—a fraction of a percent in a bank sweep, near the bill rate in a government money fund—or at about 4% in Treasury bills. Negative carry at every deposit rate below the box rate, and the box rate has sat above Treasuries in all but two weeks of the record; proceeds that stay in the account carry that gap from the day they arrive.
(The three facts, together.) The stock-to-face ratio and the house rate set the call line: at four times the face and a 30% house rate, a 64% fall, beyond the deepest fall since 1950, 2007–09's 57%; at twice the face, 29%, inside it. The term against the repayment date sets the roll count: none where a sale, a vesting or a bonus already dates the repayment, an open series where nothing does—over a mortgage's thirty years, thirty rolls, thirty rate resets, and a collateral test on every trading day in between. The regime and what else shares the account set the sizing basis: \(E\) under portfolio margin or in an account holding nothing else, \(E + P\) on a typical balances page otherwise. The balances page puts it its own way: a requirement of \(F\) against stock counted at a 30% haircut. That is what the broker has recorded.