This post draws on FERDI Working Paper P376: Arezki, Rota-Graziosi, and Le (2026), “On the Unintended Consequence of Fiscal Rules: Evidence from Fiscal Expenditures.”
Over the past three decades, fiscal rules have emerged as an institutional instrument to discipline public finances. The underlying diagnosis is now widely accepted: when left unconstrained, governments tend to run persistent deficits, increasing spending today while shifting the resulting costs onto future taxpayers. Setting ex ante ceilings on debt, deficits, or expenditures is meant to neutralize that bias.
The empirical evidence confirms that these rules can indeed rein in fiscal profligacy—provided they are well designed and effectively enforced. Our recent work (Arezki, Rota-Graziosi and Le, 2026), which relies on the Global Tax Expenditures Database (GTED), documents an unintended consequence, however: when fiscal rules become binding, they induce governments to reallocate policy effort toward a less visible, harder-to-discipline instrument—tax expenditures.
Tax Expenditures as Creative Accounting
Fiscal rules have long been associated with circumvention. Milesi-Ferretti (2003) shows that stringent rules and weak transparency encourage creative accounting, while von Hagen and Wolff (2006) find that European deficit ceilings increased reliance on stock–flow adjustments without improving underlying fiscal positions.
We argue that tax expenditures are a powerful and overlooked form of such circumvention. Exemptions, tax credits, preferential rates, and special regimes deliver targeted political benefits while remaining less visible and less scrutinized than direct spending. As Surrey (1973) emphasized, they amount to “hidden spending” through the tax system. Our contribution is to identify tax expenditures as a quantitatively important channel of fiscal circumvention and to test this mechanism using a global panel.
A Simple Model
We formalize this intuition in a three-period model that adapts the framework of Halac and Yared (2014). The government is benevolent ex ante but exhibits a present bias ex post: it overvalues the immediate political benefits of tax expenditures and underweights the future erosion of the tax base they generate.
The mechanism turns on an essential distinction: fiscal rules can only be written on verifiable aggregates—debt, deficits, on-budget spending—but not on tax expenditures, whose fragmented legal status, heterogeneous reporting, and often implicit character make them difficult to contract on.
The model yields three predictions:
- Present bias implies an overuse of tax expenditures, even in the absence of any rule.
- Introducing a binding rule on debt, deficits, or spending pushes tax expenditures above their already-inflated no-rule level. As the observable fiscal space contracts, governments substitute toward the non-verifiable margin: tighter rules mean strictly more tax expenditures at equilibrium. The circumvention effect is thus additive to the pre-existing present-bias distortion, not corrective of it.
- Spending rules exhibit a distinctive—and troubling—feature: because past tax expenditures erode the tax base and mechanically shrink current revenues, they tighten the fiscal space defined by the rule, thereby amplifying incentives to lean once again on the hidden instrument. A self-reinforcing dynamic ensues.
Only revenue rules escape this logic: by imposing a floor on realized revenues, they discipline the hidden instrument directly.
The GTED: The Foundation for our Empirical Test
Taking these predictions to the data required a comparable, cross-country, long-horizon measure of tax expenditures—a challenge that was long insurmountable. The GTED fills that gap, gathering official revenue-forgone estimates for 218 jurisdictions since 1990 at the level of the individual legal provision, with information on the beneficiary type and the tax base concerned. This granularity is what makes it possible to distinguish direct tax expenditures (personal and corporate income taxes, property taxes) from indirect ones (VAT and other taxes on goods and services), and to track their evolution before and after the introduction of a fiscal rule.
We pair the GTED with the IMF Fiscal Rules Dataset (1985–2024), which documents the presence, type, and characteristics of fiscal rules for 123 economies, and with the control-of-corruption indicator from the WB Worldwide Governance Indicators—our proxy for how binding a rule actually is in practice. As we put it in the paper,
“Fiscal rules can restrict only contractible (observable and verifiable) objects—here, on-budget spending, deficits, or public debt—but cannot condition on the non-verifiable tax-expenditure instrument.”
What the Data Reveal
Our estimates rely on the synthetic difference-in-differences estimator (Arkhangelsky et al. 2021), complemented by local-projection difference-in-differences for robustness.
First result. The adoption of a fiscal rule is followed by a statistically significant and persistent rise in tax expenditures, peaking roughly sixteen years after adoption. On average, rule adoption is associated with an increase of about 1.1 percentage points of GDP in revenue forgone.
Second result. The effects are markedly heterogeneous across rule types. Expenditure rules, budget-balance rules, and debt rules all produce a significant rise in tax expenditures—expenditure rules exhibiting the largest effect (+1.6 percentage points of GDP). Revenue rules alone generate no statistically significant effect, exactly as the theory predicts: by directly disciplining revenues, they close the door on the hidden instrument.
Third result, arguably the most telling: the effect operates both on direct tax expenditures (corporate income tax, personal income tax, property taxes) and on indirect tax expenditures (VAT exemptions and reduced rates). Circumvention, in other words, is not confined to one channel—it plays out across the entire tax code.
Fourth result, counterintuitive on its face: the effect emerges only in countries with strong control of corruption. The apparent paradox dissolves once read through the model: fiscal rules can only be circumvented if they are actually binding. In countries where public-finance oversight is less effective or corruption control is weaker, enforcement remains lax—as confirmed by our finding that conventional expenditures are not effectively constrained—and the innovation channel fails to emerge. By contrast, where fiscal rules are genuinely binding, governments primarily circumvent them by substituting tax expenditures for conventional spending.
Policy Implications
These findings in no way undercut the case for fiscal rules. They do, however, restate a lesson long familiar from financial regulation: any constraint imposed on an observable instrument creates incentives to substitute toward the adjacent unobservable one.
Three practical implications follow.
First, a fiscal rule that is not paired with a robust tax-expenditure transparency framework is a rule that invites circumvention. The Global Tax Expenditures Transparency Index (GTETI) as well as initiatives championed by the Coalition on Tax Expenditure Reform and the systematic-reporting recommendations of the IMF (2019) acquire their full meaning in this perspective.
Second, the choice of target matters as much as the stringency of the rule. Revenue rules—rare in our sample, with only twenty-four countries—deserve renewed consideration, not as a substitute for spending or deficit rules, but as a complement that disciplines the hidden margin.
Third, in developing countries, where administrative capacity is often more limited, the effective and comprehensive implementation of fiscal rules should remain the primary objective. At the same time, systematic reporting on tax expenditures is essential to prevent a gradual and opaque shift in public spending toward instruments that are less transparent, potentially more distortionary, and more vulnerable to capture by vested interests—a pattern documented in developed economies (Burman and Phaup 2012; European Commission 2014; OECD 2025).
Fiscal rules, like capital requirements, are effective only insofar as they simultaneously close the exits that institutional innovation inevitably opens. Without more and better tax expenditure data, that closure remains out of reach.