Tax arbitrage, as the term is used in the literature, involves having offsetting long and short positions that are taxed asymmetrically.
In a "pure tax arbitrage," as Eugene Steuerle first (to my knowledge) dubbed it in 1985, the taxpayer has perfectly offsetting economic positions that may pair, say, deductibility of the payout with excludability (or at least deferral) for the receipt. A classic example is the Knetsch case, decided by the Supreme Court in 1960. Here the taxpayer purported to borrow and lend the same $4 million from the same insurance company counter-party. But he "borrowed" at 3.5% and "invested" at 2.5%, leading to $140,000 of accrued interest expense and only $100,000 of accrued annuity appreciation per year. So he'd write the company a $40,000 check for the difference and deduct $140,000 of interest. (Black letter law at the time unambiguously supported both the deduction and the exclusion, leaving aside economic substance doctrine).
In a regular tax arbitrage, the positions aren't perfect offsets. Hence, they may not be arbitrages in the way that finance people use the term, but they are still tax arbitrages. For example, if you borrow to hold non-dividend-paying stock, deducting the interest while deferring the offsetting appreciation, this would give you a tax arbitrage, even though you have a net position (e.g., bad news for you if the stock market plunges). However, the tax juice from this little game is addressed by the investment interest limitation, which generally limits deductions for investment interest to net investment income.
There would be no tax arbitrages under a pure Haig-Simons income tax that was consistently applied. But when there is a hodge-podge of different rules for different types of items, some of it becomes inevitable. A whole host of special rules in the Internal Revenue Code try to limit it in one setting or another. These rules add complexity if one defines it simply as the number of pages of tax law, but they may actually reduce complexity in practice if they sufficiently discourage taxpayers from going to great (or any) lengths to arrange tax arbitrages that would siphon money out of the Treasury to no socially or economically valuable end.
Asymmetric substantive rules are one way get to the creation of tax arbitrages that serve no good purpose. Having at least some of this is inevitable once you have a realization-based income tax. But the creation of pointless tax arbitrages becomes even easier for taxpayers, if one allows them to pick and choose between alternative marginal rates for the income and/or loss from the same activities.
It's amazing to me to what extent the House and Senate versions of tax "reform" pointlessly create huge tax arbitrages of this kind. The proposed special rate for pass-throughs, a key feature of both bills, has the potential to become the single greatest inducement to tax arbitrage ever enacted by a single Congress.
And of course I have been noting here how the Senate Finance Committee's idea of expanding expensing immediately while delaying the corporate rate increase by a year, creates a massive tax arbitrage loophole. As per my prior post, it is economically equivalent to providing 175% expensing - i.e., allowing deductions of $1.75 for every $1 the taxpayer spends - for 2018 only. Spending a dollar to earn a dollar, but getting to deduct $1.75 while only including $1, is a classic tax arbitrage example.
If I were willing to lengthen this post still more, I could explain in detail a further tax arbitrage that is possibly more important than the ones I have been emphasizing. The bills combine allowing expensing for new investment with partially retaining interest deductibility, and the literature has shown that this allows the creation of tax arbitrages that can drive the effective tax rate for new investment below 0 percent (i.e., to amount to a net payout from the government even if the investment offers a pretax return of zero).
Given these egregious problems in the bills, I am thinking that the Republicans should relabel what they are doing the Tax Arbitrage Act of 2018.
They may not have fully realized this, but it's an inevitable byproduct of trying to craft multiple rates that can apply to the same taxpayer and (with requisite planning) to the same offsetting outlays and receipts.