*460 Justice Breyer

delivered the opinion of the Court.

A State may tax a proportionate share of the income of a nondomiciliary corporation that carries out a particular business both inside and outside that State. Allied-Signal, Inc. v. Director, Div. of Taxation, 504 U. S. 768, 772 (1992). The State, however, may not tax income received by a corporation from an “ ‘ “unrelated business activity” ’ which constitutes a ‘“discrete business enterprise.”’” Id., at 773 (quoting Exxon Corp. v. Department of Revenue of Wis., 447 U. S. 207, 224 (1980), in turn quoting Mobil Oil Corp. v. Commissioner of Taxes of Vt., 445 U. S. 425, 442, 439 (1980)). California’s rules for taxing its share of a multistate corporation’s income authorize a deduction for interest expense. But they permit (with one adjustment) use of that deduction only to the extent that the amount exceeds certain out-of-state income arising from the unrelated business activity of a discrete business enterprise, i. e., income that the State could not otherwise tax. We must decide whether those rules violate the Constitution’s Due Process and Commerce Clauses. We conclude that they do.

The legal issue is less complicated than may first appear, as examples will help to show. California, like many other States, uses what is called a “unitary business” income-calculation system for determining its taxable share of a multistate corporation’s business income. In effect, that system first determines the corporation’s total income from its nationwide business. During the years at issue, it then averaged three ratios — those of the firm’s California property, payroll, and sales to total property, payroll, and sales— to make a combined ratio. Cal. Rev. & Tax Code Ann. *461 §§ 25128, 25129, 25132, 25134 (West 1979). Finally, it multiplies total income by the combined ratio. The result is “California’s share,” to which California then applies its corporate income tax. If, for example, an Illinois tin can manufacturer, doing business in California and elsewhere, earns $10 million from its total nationwide tin can sales, and if California’s formula determines that the manufacturer does 10% of its business in California, then California will impose its income tax upon 10% of the corporation’s tin can income, $1 million.

The income of which California taxes a percentage is constitutionally limited to a corporation’s “unitary” income. Unitary income normally includes all income from a corporation’s business activities, but excludes income that “derive[s] from unrelated business activity which constitutes a discrete business enterprise,” Allied-Signal, 504 U. S., at 773 (internal quotation marks omitted). As we have said, this latter “nonunitary” income normally is not taxable by any State except the corporation’s State of domicile (and the States in which the “discrete enterprise” carries out its business). Ibid.

Any income tax system must have rules for determining the amount of net income to be taxed. California’s system, like others, basically does so by asking the corporation to add up its gross income and then deduct costs. One of the costs that California permits the corporation to deduct is interest expense. The statutory language that authorizes that deduction — the language here at issue — contains an important limitation. It says that the amount of “interest deductible” shall be the amount by which “interest expense exceeds interest and dividend income ... not subject to allocation by formula,” i. e., the amount by which the interest expense exceeds the interest and dividends that the non-domiciliary corporation has received from nonunitary busi-nessor investment. Cal. Rev. & Tax Code Ann. § 24344 (West 1979) (emphasis added); Appendix, infra. Suppose the Illi *462 nois tin can manufacturer has interest expense of $150,000; and suppose it receives $100,000 in dividend income from a nonunitary New Zealand sheep-farming subsidiary. California’s rule authorizes an interest deduction, not of $150,000, but of $50,000, for the deduction is allowed only insofar as the interest expense “exceeds” this other unrelated income.

Other language in the statute makes the matter a little more complex. One part makes clear that, irrespective of nonunitary income, the corporation may use the deduction against unitary interest income that it earns. §24844. This means that if the Illinois tin can manufacturer has earned $100,000 from tin can related interest, say, interest paid on its tin can receipt bank accounts, the manufacturer can use $100,000 of its interest expense deduction to offset that interest income (though it would still lose the remaining $50,000 of deduction because of income from the New Zealand sheep farm). Another part provides an exception to the extent that the subsidiary paying the dividend has paid taxes to California. §§ 24344,24402. If the sheep farm were in California, not New Zealand (or at least to the extent it were taxable in California), the tin can manufacturer would not lose the deduction. We need not consider either of these complications here.

One final complication involves a dispute between the parties over the amount of interest expense that the California statute at issue covers. Hunt-Wesson, Inc., claims that California (at least during the years at issue here) required interstate corporations first to determine what part of their interest expense was for interest related to the unitary business and what part was for interest related to other, nonunitary matters. It says that the statute then required it to put the latter to the side, so that only interest related to the unitary business was at issue. California agrees that the form it provided to corporations during the years at issue did work this way, but states that the form did not interpret the statute correctly. In its view, the statute takes all interest ex *463 pense into account. Apparently California now believes that, if the tin can manufacturer had $100,000 interest expense related to its tin can business, and another $50,000 interest expense related to the New Zealand sheep farm (say, money borrowed to buy shares in the farm), then California’s statute would count a total interest expense of $150,000, all of which California would permit it to deduct from its unitary tin can business income if, for example, it had no non-unitary New Zealand sheep farm income in that particular year. This matter, arguably irrelevant to the tax years here in question (Hunt-Wesson reported no nonunitary interest expense), is also irrelevant to our legal result. Therefore, we need not consider this particular dispute further.

The question before us then is reasonably straightforward: Does the Constitution permit California to carve out an exception to its interest expense deduction, which it measures by the amount of nonunitary dividend and interest income that the nondomiciliary corporation has received? Petitioner, Hunt-Wesson, Inc., is successor in interest to a non-domiciliary corporation. That corporation incurred interest expense during the years at issue. California disallowed the deduction for that expense insofar as the corporation had received relevant nonunitary dividend and interest income. Hunt-Wesson challenged the constitutional validity of the disallowance. The California Court of Appeal found it constitutional, No. A079969 (Dec. 11, 1998), App. 54; see also Pacific Tel. & Tel. Co. v. Franchise Tax Bd., 7 Cal. 3d 544, 498 P. 2d 1030 (1972) (upholding statute), and the California Supreme Court denied review, App. 67. We granted certio-rari to consider the question.

Relevant precedent makes clear that California’s rule violates the Due Process and Commerce Clauses of the Federal Constitution. In Container Corp. of America v. Franchise Tax Bd., 463 U. S. 159 (1983), this Court wrote that the “Due *464 Process and Commerce Clauses ... do not allow a State to tax income arising out of interstate activities — even on a proportional basis — unless there is a ‘“minimal connection” or “nexus” between the interstate activities and the taxing State, and “a rational relationship between the income attributed to the State and the intrastate values of the enterprise.”’” Id., at 165-166 (quoting Exxon Corp., 447 U. S., at 219-220, in turn quoting Mobil Oil Corp., 445 U. S., at 436, 437). Cf. International Harvester Co. v. Department of Treasury, 322 U. S. 340, 358 (1944) (Rutledge, J., concurring in part and dissenting in part) (“If there is a want of due process to sustain” a tax, “by that fact alone any burden the tax imposes on the commerce among the states becomes ‘undue’”). The parties concede that the relevant income here — that which falls within the scope of the statutory phrase “not allocable by formula” — is income that, like the New Zealand sheep farm in our example, by itself bears no “rational relationship” or “nexus” to California. Under our precedent, this “nonunitary” income may not constitutionally be taxed by a State other than the corporation’s domicile, unless there is some other connection between the taxing State and the income. Allied-Signal, 504 U. S., at 772-773.

California’s statute does not directly impose a tax on non-unitary income. Rather, it simply denies the taxpayer use of a portion of a deduction from unitary income (income like that from tin can manufacture in our example), income which does bear a “rational relationship” or “nexus” to California. But, as this Court once put the matter, a “ ‘tax on sleeping measured by the number of pairs of shoes you have in your closet is a tax on shoes.’ ” Trinova Corp. v. Michigan Dept. of Treasury, 498 U. S. 358, 374 (1991) (quoting Jenkins, State Taxation of Interstate Commerce, 27 Tenn. L. Rev. 239, 242 (1960)). California’s rule measures the amount of additional unitary income that becomes subject to its taxation (through reducing the deduction) by precisely the amount of nonuni-tary income that the taxpayer has received. And for that *465 reason, that which California calls a deduction limitation would seem, in fact, to amount to an impermissible tax. National Life Ins. Co. v. United States, 277 U. S. 508 (1928) (finding that a federal statute that reduced an insurance company’s tax deduction for reserves by the amount of tax-exempt interest the company received from a holding of “tax-free” municipal bonds constituted unlawful taxation of tax-exempt income).

However, this principle does not end the matter. California offers a justification for its rule that seeks to relate the deduction limit to collection of California’s tax on unitary income. If California could show that its deduction limit actually reflected the portion of the expense properly related to nonunitary income, the limit would not, in fact, be a tax on nonunitary income. Rather, it would merely be a proper allocation of the deduction. See Denman v. Slayton, 282 U. S. 514 (1931) (upholding Federal Tax Code’s denial of interest expense deduction where borrowing is incurred to “purchase or carry” tax-exempt obligations).

California points out that money is fungible, and that consequently it is often difficult to say whether a particular borrowing is “really” for the purpose of generating unitary income or for the purpose of generating nonunitary income. California’s rule prevents a firm from claiming that it paid interest on borrowing for the first purpose (say, to build a tin can plant) when the borrowing is “really” for the second (say, to buy shares in the New Zealand sheep farm). Without some such rule, firms might borrow up to the hilt to support their (more highly taxed) unitary business needs, and use the freed unitary business resources to purchase (less highly taxed) nonunitary business assets. This “tax arbitrage” problem, California argues, is why this Court upheld the precursor of 26 U. S. C. § 265(a)(2), which denies the taxpayer an interest deduction insofar as the interest expense was “incurred or continued to purchase or carry” tax-exempt obligations or securities. Denman v. Slayton, *466 supra, at 519. This Court has consistently upheld deduction denials that represent reasonable efforts properly to allocate a deduction between taxable and tax-exempt income, even though such denials mean that the taxpayer owes more than he would without the denial. E. g., First Nat. Bank of Atlanta v. Bartow County Bd. of Tax Assessors, 470 U. S. 583 (1985).

The California statute, however, pushes this concept past reasonable bounds. In effect, it assumes that a corporation that borrows any money at all has really borrowed that money to “purchase or carry,” ef. 26 U. S. C. §265(a)(2), its nonunitary investments (as long as the corporation has such investments), even if the corporation has put no money at all into nonunitary business that year. Presumably California believes that, in such a case, the unitary borrowing supports the nonunitary business to the extent that the corporation has any nonunitary investment because the corporation might have, for example, sold the sheep farm and used the proceeds to help its tin can operation instead of borrowing.

At the very least, this last assumption is unrealistic. And that lack of practical realism helps explain why California’s rule goes too far. A state tax code that unrealistically assumes that every tin can borrowing first helps the sheep farm (or the contrary view that every sheep farm borrowing first helps the tin can business) simply because of the theoretical possibility of a hypothetical sale of either business is a code that fails to “actually reflect a reasonable sense of how income is generated,” Container Corp., 463 U. S., at 169, and in doing so assesses a tax upon constitutionally protected nonunitary income. That is so even if, as California claims, its rule attributes all interest expense both to unitary and to nonunitary income. And it is even more obviously so if, as Hunt-Wesson claims, California attributes all sheep-farm-related borrowing to the sheep farm while attributing all tin-can-related borrowing first to the sheep farm as well.

*467 No other taxing jurisdiction, whether federal or state, has taken so absolute an approach to the tax arbitrage problem that California presents. Federal law in comparable circumstances (allocating interest expense between domestic and foreign source income) uses a ratio of assets and gross income to allocate a corporation’s total interest expense. See 26 CFR §§ 1.861-9T(f), (g) (1999). In a similar, but much more limited, set of circumstances, the federal rules use a kind of modified tracing approach — requiring that a certain amount of interest expense be allocated to foreign income in situations where a United States business group’s loans to foreign subsidiaries and the group’s total borrowing have increased relative to recent years (subject to a number of adjustments), and both loans and borrowing exceed certain amounts relative to total assets. See §1.861-10. Some States other than California follow a tracing approach. See, e. g., D. C. Mun. Regs., Tit. 9, § 123.4 (1998); Ga. Rules and Regs. § 560-7-7.03(3) (1999). Some use a set of ratio-based formulas to allocate borrowing between the generation of unitary and nonunitary income. See, e. g., Ala. Code § 40-18-35(a)(2) (1998); La. Reg. § 1130(B)(1) (1988). And some use a combination of the two approaches. See, e. g., N. M. Admin. Code, Tit. 3, § 5.5.8 (1999); Utah Code Ann. § 59-7-101 (19) (1999). No other jurisdiction uses a rule like California’s.

Ratio-based rules like the one used by the Federal Government and those used by many States recognize that borrowing, even if supposedly undertaken for the unitary business, may also (as California argues) support the generation of nonunitary income. However, unlike the California rule, ratio-based rules do not assume that all borrowing first supports nonunitary investment. Rather, they allocate each borrowing between the two types of income. Although they may not reflect every firm’s specific actions in any given year, it is reasonable to expect that, over some period of *468 time, the ratios used will reflect approximately the amount of borrowing that firms have actually devoted to generating each type of income. Conversely, it is simply not reasonable to expect that a rule that attributes all borrowing first to nonunitary investment will accurately reflect the amount of borrowing that has actually been devoted to generating each type of income.

Because California's offset provision is not a reasonable allocation of expense deductions to the income that the expense generates, it constitutes impermissible taxation of income outside its jurisdictional reach. The provision therefore violates the Due Process and Commerce Clauses of the Constitution.

The judgment of the California Court of Appeal is reversed, and the case is remanded for proceedings not inconsistent with this opinion.

It is so ordered.

APPENDIX TO OPINION OF THE COURT

Cal. Rev. & Tax Code Ann. §24344 (West 1979).

“Interest; restrictions

“(a) Except as limited by subsection (b), there shall be allowed as a deduction all interest paid or accrued during the income year on indebtedness of the taxpayer.

“(b) [T]he interest deductible shall be an amount equal to interest income subject to allocation by formula, plus the amount, if any, by which the balance of interest expense exceeds interest and dividend income (except dividends deductible under the provisions of Section 24402) not subject to allocation by formula. Interest expense not included in the preceding sentence shall be directly offset against interest and dividend income (except dividends deductible under the provisions of Section 24402) not subject to allocation by formula.”

*469 “§24402. Dividends

“Dividends received during the income year declared from income which has been included in the measure of the taxes imposed under Chapter 2 or Chapter 3 of this part upon the taxpayer declaring the dividends.”