Thin capitalization rules ask a single question in two different ways. The older way looks at the balance sheet and asks whether the borrower carries more debt than its equity supports. The newer way looks at the income statement and asks whether the interest deduction is large relative to the earnings the borrower reports. The United States has used both, in that order, and the OECD recommendation that reshaped international practice is of the second (income statement ratio) kind.
Test 1: the balance-sheet ratio
Section 163(j), enacted in 1989 as the earnings-stripping provision, disallowed a deduction for disqualified interest only if two threshold tests were both satisfied. The first was a debt-to-equity test: the provision did not apply unless the payor’s ratio of debt to equity exceeded 1.5 to 1 at the close of the taxable year. A corporation inside the ratio was outside the statute, which is why the ratio came to be called the safe harbor.
\begin{aligned}
&(1)\quad \frac{D}{E}\leq 1.5
\qquad \text{(old section 163(j) safe harbor)}
\end{aligned}The Compustat analogues are the long-term debt item DLTT, the current portion DLC, and stockholders’ equity SEQ. Two cautions attach. The statutory ratio was computed on the adjusted tax basis of assets and on a consolidated basis, so book equity is a proxy for the tax measure rather than the measure itself. And a numerator confined to DLTT omits DLC, where intercompany borrowing frequently sits. A screen built on DLTT alone will find a taxpayer inside the safe harbor whose related-party leverage is short-dated.
Test 2: the income-statement ratio
The second threshold test of old section 163(j) was an earnings test: the provision applied only where net interest expense exceeded fifty percent of adjusted taxable income, computed without regard to net interest, net operating losses, depreciation, amortization, and depletion. Adjusted taxable income was, in substance, a tax-basis EBITDA.
The Tax Cuts and Jobs Act of 2017 rewrote the provision. The debt-to-equity safe harbor was repealed. The rule now applies to all taxpayers above the gross-receipts threshold, related and unrelated lenders alike, and caps the deduction of net business interest at thirty percent of adjusted taxable income. What had been one of two conjunctive tests became the whole rule:
\begin{aligned}
&(2)\quad \frac{XINT-IDIT}{ATI}\leq 0.30
\qquad \text{(current section 163(j))}
\end{aligned}The numerator is net interest, not gross. Business interest income is subtracted from business interest expense before the ratio is formed, so XINT alone overstates the tested quantity by the amount of IDIT. In an interest-rate environment where cash balances earn a return, the difference is not small.
The denominator changed twice
This is where the choice between OIBDP and OIADP is settled, and the answer is that both are correct, for different tax years. Adjusted taxable income has passed through three regimes:
For tax years 2018 through 2021, depreciation, amortization, and depletion were added back, so adjusted taxable income was a tax-basis EBITDA. The book analogue is OIBDP.
For tax years beginning after December 31, 2021, the add-back expired and adjusted taxable income became a tax-basis EBIT. The book analogue is OIADP.
For tax years beginning after December 31, 2024, the One Big Beautiful Bill Act restored the add-back permanently. The book analogue reverts to OIBDP.
Since OIBDP = OIADP + DP, the two denominators differ by the depreciation and amortization charge, and the ratio tested against the thirty percent threshold differs accordingly. The gap is widest in capital-intensive industries, which is why the 2022 change bound in real estate, construction, and manufacturing and was largely invisible elsewhere. Any panel that spans 2018 to the present and applies a single denominator is measuring three different statutes with one instrument.
The Compustat mnemonic DLTT alone omits DLC, which is where intercompany borrowing often sits. The numerator of Test 2 is net interest: XINT − IDIT, since business interest income is subtracted before the ratio is formed.
The OECD recommendation
BEPS Action 4 is the reason the income-statement test displaced the balance-sheet test internationally. The 2015 final report, updated in December 2016, recommends a fixed ratio rule limiting an entity’s net interest deductions to a benchmark percentage of tax EBITDA, with the corridor set between ten and thirty percent; a group ratio rule permitting a higher deduction based on the net third-party interest to EBITDA ratio of the consolidated group; and targeted rules for specific risks. The choice of EBITDA over a debt-to-equity ratio is deliberate. A fixed ratio of debt to equity constrains the stock; a fixed ratio of interest to earnings constrains the flow that the deduction actually reduces.
The European implementation, Article 4 of the Anti-Tax Avoidance Directive, sets the limit at thirty percent of EBITDA with a de minimis of EUR 3 million of exceeding borrowing costs. It also preserves the balance sheet, in a comparative form: under the equity escape rule a member of a consolidated group may deduct in full where its own ratio of equity to total assets is equal to the group ratio or lower by no more than two percentage points. The test is no longer whether the entity is leveraged, but whether it is leveraged more than the group of which it is part.
What the two ratios measure
The pair is worth holding together because they answer to different economics. The debt-to-equity ratio is a stock relation, indifferent to profitability; a firm inside 1.5 to 1 passes whether it earns anything or not. The interest-to-earnings ratio is a flow relation, indifferent to capital structure; a profitable firm can carry leverage that would fail the older test. Neither is a measure of arm’s-length pricing, and neither was offered as one. They are administrative screens, and their thresholds are conventions rather than findings.
For empirical work the consequence is that the two ratios do not rank firms the same way, and the ranking is not stable across the 2018, 2022, and 2025 regimes. Before drawing any inference from a distribution of either ratio, one should ask what the denominator was in the year observed.
References
OECD, Limiting Base Erosion Involving Interest Deductions and Other Financial Payments, Action 4 – 2015 Final Report (Paris: OECD Publishing, 2015), doi 10.1787/9789264241176-en. https://www.oecd.org/en/publications/limiting-base-erosion-involving-interest-deductions-and-other-financial-payments-action-4-2015-final-report_9789264241176-en.html
OECD, Action 4 – 2016 Update: Inclusive Framework on BEPS (Paris: OECD Publishing, 2016), doi 10.1787/9789264268333-en. Full text, free: https://www.oecd.org/content/dam/oecd/en/publications/reports/2016/12/limiting-base-erosion-involving-interest-deductions-and-other-financial-payments-action-4-2016-update_g1g745d0/9789264268333-en.pdf
Council Directive (EU) 2016/1164 (ATAD), Article 4, interest limitation rule. See also the Commission’s implementation report, COM(2020) 383, CELEX 52020DC0383: https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52020DC0383
IRS, Notice 2018-28, Initial Guidance Under Section 163(j) as Applicable to Taxable Years Beginning After December 31, 2017. States the two threshold tests of old section 163(j), including the 1.5 to 1 safe harbor ratio. https://www.irs.gov/pub/irs-drop/n-18-28.pdf
IRS, LB&I Training, Tax Cuts and Jobs Act: IRC Section 163(j) Participant Guide. Contrasts old and new section 163(j) directly. https://www.irs.gov/pub/newsroom/lbi-tcja-participant-guide-163j-13301.pdf IRS, REG-106089-18, Notice of Proposed Rulemaking under section 163(j). https://www.irs.gov/pub/irs-drop/REG-106089-18-NPRM.pdf
Testing for Thin Capitalization Under Section 163(j): A Flawed Safe Harbor. A critique of the debt-to-equity safe harbor, including its failure to accommodate industries with different asset mixes. Free full text: https://activityinsight.pace.edu/pcohen/intellcont/Article-Testing%20for%20Thin%20Capitalization%20under%20Section%20163(j)%20%20A%20Flawed%20Safe%20Harbor%20-Final%20Article%20-Jan%202014-1.pdf
Not So Fast: 163(j), 245A, and Leverage in the Post-TCJA World, Yale Law Journal Forum. Worked numerical example of the pre-TCJA safe harbor. https://yalelawjournal.org/forum/not-so-fast-163j-245a-and-leverage-in-the-post-tcja-world