betaFeedback

WorksEconomics, Econometrics and Finance

Optimal hybrid tax enforcement

Abstract

Modern tax design and enforcement rely on classification data reported by taxpayers. Data improve the targeting of tax policy and audits but also open avenues for evasion. How can tax authorities use such data while accounting for their potential strategic distortion? We demonstrate that restoring deterrence generates an endogenous global verification frontier: every report below the frontier is verified through either a profit audit or data verification. Optimal hybrid tax enforcement requires selecting which data classes are subject to profit audits up to the global frontier and which are profit-audited only up to lower class-specific cutoffs, with data verification taking over beyond these cutoffs. This costly data verification requirement limits the benefit from differentiated tax enforcement. Optimal implementation is characterized by profit bunching but involves no data bunching.

Introduction

Modern tax systems make extensive use of self-declared data to differentiate tax obligations and enforcement. Corporate tax returns, for example, contain detailed expense information and balance-sheet items that firms must report under the appropriate categories to claim deductions and exemptions. Such data can improve the targeting of tax policy and audits, thus mitigating the regressivity bias of tax enforcement [21, 24, 31]. Yet, data-driven tax policies also create natural incentives to distort reported information.2

A growing empirical literature documents widespread manipulation of classifications. Firms respond to size-dependent enforcement thresholds by adjusting actual or reported revenue or organizational form, generating sharp distortions and bunching patterns [2, 5, 13] (Almunia et al. 2024). Similarly, a differentiated tax regime across production technologies—such as deductions or exemptions for specific activities—induces misclassification, relabeling of activities, and business splitting [1, 10, 15] (Kuchumova et al. 2025). While such findings are consistent with evidence that reporting behavior responds endogenously to the design of tax policy and enforcement [14, 20, 32, 33], tax theory largely abstracts from strategic data reporting.

In this paper, we study how tax enforcement should optimally account for self-reported data that are both informative and manipulable. The tax authority has two verification instruments: costly profit audits that verify income and cheaper data audits that detect misclassification and may trigger subsequent profit verification. The central tradeoff is that, while data allow the authority to differentiate enforcement across firms, self-classification allows firms to manipulate both enforcement and the tax base. Preserving truthful data reporting thus places significant limits on differentiation.

Our framework provides a systematic approach to studying the structure of optimal enforcement policies under strategic data reporting. We find that restoring deterrence introduces a global verification frontier. When enforcement differs across classes, firms assigned to more strongly enforced classes have an incentive to jointly misreport profits and classification data to access more favorable tax treatment. Eliminating these joint deviations requires that every report below the frontier is verified, either directly through a profit audit or through the verification of the classification data. Firms reporting the same profit may face different forms of enforcement depending on their data class: a profit report may trigger a profit audit in one class and data verification in another.

A practical implication of this frontier structure is that optimal enforcement is inherently nonseparable across data classes and cannot be optimized class by class: differentiated tax treatment generates an endogenous verification requirement that links profit audits and data verification and calls for integration between enforcement functions. For example, the IRS organizes enforcement through specialized units, including Tax Exempt and Government Entities, Large Business and International, and Small Business/Self-Employed, while coordinating across divisions through data-driven workload selection and cross-divisional case classification. Our results highlight the importance of such collaboration and efficient allocation of the enforcement budget across departments.3

More specifically, our characterization reveals the costs and benefits of differentiated enforcement across data classes. Building on Sanchez and Sobel [30], we derive a local enforcement index that captures the marginal gains from extending profit audits within each class. We demonstrate how this relatively overlooked index can be leveraged to assess the efficiency of class-by-class profit-audit allocation under strategic data reporting.4 The index can be readily estimated from administrative data.5

The form of optimal enforcement depends on the shape of the enforcement index. When the index is decreasing in profits (e.g., under a linear tax schedule with a decreasing inverse hazard rate), enforcement operates on the intensive margin. As the audit cutoff rises, each additional audit deters underreporting among fewer high-profit firms, thus generating diminishing returns and favoring class-specific cutoffs. We demonstrate that optimal hybrid tax enforcement combines profit audits and data verification, with the mix of instruments varying across classes. In some classes, profit audits stop at class-specific cutoffs, where data verification takes over and continues up to the global verification frontier. In other classes, the authority relies exclusively on profit audits, thus avoiding the additional costs of data verification on profit reports below the frontier. These alternative outcomes show how the costs of preserving truthful data reporting can outweigh the benefits of differentiated profit-audit cutoffs, despite their promise for improved audit targeting.6

When the enforcement index increases in profits (e.g., under a sufficiently progressive tax schedule), by contrast, enforcement operates on the extensive margin. As the audit cutoff rises, the gains from deterring underreporting increase with profits. Whenever a class is audited, profit audits thus extend all the way up to the global frontier. Conditional on the global verification frontier, the enforcement decision therefore shifts from choosing class-specific audit cutoffs to deciding which classes are audited for profits up to the frontier and which are wholly exempt from profit audits. Importantly, exemption from profit audits does not imply the absence of enforcement altogether: these classes still undergo systematic data verification to prevent firms in profit-audited classes from obtaining weaker enforcement through misclassification. The resulting division between profit-audited and data-verified classes echoes the IRS’s division-based approach to enforcement and highlights how optimal enforcement may involve fundamentally different verification regimes across taxpayer classes.

Our characterization also provides a diagnostic for whether data verification is sufficient. Optimal implementation generates bunching in reported profits at profit-audit cutoffs, but no bunching in classification data: firms respond to the limits of profit enforcement, whereas verification removes the gain from strategically changing data classes. Empirical findings of bunching around size-dependent thresholds and strategic relabeling of activities and expenses [2, 10] may therefore indicate that data-contingent policy is not successfully implemented.

Finally, we provide a measure to quantify the value of self-reported data for enforcement. More informative data improve targeting, but strategic reporting requires costly verification that diverts resources from profit audits. We decompose the gross targeting value of data into the net reduction in the tax gap and the implementation cost. This yields a simple adoption criterion: the targeting value must exceed the implementation cost together with the cost of establishing the information system. Naively conditioning on manipulable data can perform worse than ignoring the data altogether.

Illustrative example

A simple example illustrates the fragility of data-contingent enforcement when classification data are manipulable. Suppose firms self-classify an activity as either eligible for special tax treatment (e.g., R&D expenses) or ordinary (e.g., other expenses). Based on this classification, the tax authority sets a lower profit-audit cutoff, ye′y'_e, for eligible activities and a higher cutoff, yo′y'_o, for ordinary activities. As shown in Figure 1, this generates two effective tax schedules: the blue solid line for eligible activities and the blue dashed line for ordinary activities.

Figure 1. Data-contingent enforcement

If classification were taken at face value, a firm with an ordinary activity could obtain the more favorable treatment by reporting its activity as eligible and simultaneously reducing its declared tax base to ye′y'_e. The firm, therefore, manipulates both dimensions of its report: classification data determine which enforcement schedule it faces, while the profit declaration determines its position on that schedule. Such joint deviations are the central incentive problem in our model.

Credible enforcement of different tax schedules for eligible and ordinary activities requires data verification. With two classes, the higher cutoff, yo′y'_o, defines the global verification frontier: every report below it must be verified either through a profit audit or through verification of the reported classification. Ordinary activities are profit-audited up to yo′y'_o; eligible activities are profit-audited up to ye′y'_e, while their classification data are verified for y∈(ye′,yo′]y \in(y'_e,y'_o] to induce truthful reporting.7

The paper is organized as follows. The model is set up in § 2. § 3 studies non-strategic data reporting, and § 4 investigates strategic data reporting. § 5 studies the value of data and compares the tax gap in alternative information environments. § 6 discusses related literature, and § 7 concludes. Proofs and omitted results are in Appendices § 8–§ 10; extensive supplementary material—on MVPF, comparative statics, analytical examples, and a quantitative illustration—is delegated to Appendices § 11–§ 15.

Model

Setup. The economy consists of a unit mass of firms maximizing after-tax profits net of expected penalties for misreporting. A tax authority (hereafter, the IRS) designs an audit policy to minimize the expected tax gap, subject to a budget constraint.

Firms. Each firm has a private type (y,d)(y,d), where yy denotes taxable profit and dd denotes correlated self-reported data (e.g., turnover, payroll, asset holdings, expenses within a given category) that are used for improved targeting of tax enforcement.8

Types are drawn from a joint cumulative distribution function HH with a differentiable density function hh, supported on S:=[y‾,y‾]×[d‾,d‾]S := [\underline{y}, \overline{y}] \times[\underline{d}, \overline{d}]. Marginal distributions are denoted by (HY,hY)(H_Y, h_Y) and (HD,hD)(H_D, h_D), and conditional distributions by (HY∣D,hY∣D)(H_{Y|D}, h_{Y|D}).9

Reports. After observing its realized type (y,d)(y,d), each firm simultaneously chooses a profit declaration xx and a data declaration cc that it reports to the IRS. The declared profit determines the firm’s statutory tax liability T(x)T(x), where TT defines an exogenous, increasing tax schedule. A reporting strategy maps each type (y,d)(y,d) into declarations (x,c)(x,c), i.e., x:=x(y,d)x := x(y,d) and c:=c(y,d)c := c(y,d).

Audits. The IRS commits to an audit policy AA that firms can observe before submitting reports (x,c)(x,c):

A:S→[0,1]2,A(x,c):=(αY(x,c),αD(x,c)),A : S \to[0,1]^2,\qquad A(x,c) := (\alpha_Y(x,c),\alpha_D(x,c)),

where αY(x,c)\alpha_Y(x,c) is the probability of a profit audit and αD(x,c)\alpha_D(x,c) is the probability of a data audit for (x,c)(x,c).

Audits impose costs. A profit audit entails a cost ρY\rho_Y, while a data audit entails a lower cost ρD<ρY\rho_D < \rho_Y. Data audits are cheaper because they can rely on partially automated cross-checks with third-party information, whereas profit audits require more detailed verification of accounting information. Costless data verification, ρD=0\rho_D = 0, provides a natural benchmark in which classification data can be verified without using enforcement resources (§ 3). We depart from this benchmark by allowing data verification to be arbitrarily cheap but costly, ρD>0\rho_D > 0 (§ 4).

Given reporting strategies (x(y,d),c(y,d))(x(y,d),c(y,d)), expected audit expenditure is

ρ=ρY∫αY(x(y,d),c(y,d)) dH(y,d)+ρD∫αD(x(y,d),c(y,d)) dH(y,d).\rho= \rho_Y \int\alpha_Y(x(y,d),c(y,d))\,\mathrm{d}H(y,d) + \rho_D \int\alpha_D(x(y,d),c(y,d))\,\mathrm{d}H(y,d).

We assume that audits perfectly detect misreporting in the dimension being verified.

When a data mismatch is detected, the IRS escalates the case to a profit audit, enabling penalties proportional to evaded tax.101112

Penalties. If a firm of type (y,d)(y,d) reports (x,c)(x,c), its expected penalty is

Φ(x,c∣y,d,A)=π(αY(x,c)+αD(x,c)1{c≠d})(T(y)−T(x)),\Phi(x,c \mid y,d,A) = \pi\left(\alpha_Y(x,c) + \alpha_D(x,c)\mathbf{1}_{\{c\ne d\}}\right)(T(y)-T(x)),

where π>1\pi> 1 is the marginal penalty per unit of evaded tax.13 As in Yitzhaki [34], the penalty is proportional to the evaded tax, so incentives scale with the extent of profit underreporting. Moreover, following Sanchez and Sobel [30], we assume that firms make truthful reports unless strategic reporting strictly increases expected payoffs. Under this standard assumption, firms either report both dimensions truthfully, misreport only profits, or misreport both profits and data; misreporting data alone does not occur.

A firm’s problem. Each firm chooses a reporting strategy (x,c)=(x(y,d),c(y,d))(x,c)=(x(y,d),c(y,d)) to maximize expected net profits after taxes:

Π(y,d)=max⁡x,c{y−T(x)−Φ(x,c∣y,d,A)}.(1)\Pi(y,d)=\max_{x,c}\{y-T(x)-\Phi(x,c\mid y,d,A)\}. \tag*{(1)}

The IRS’ problem. The IRS chooses an audit policy A(x,c)A(x,c) to minimize the expected tax gap subject to a budget constraint:

G(B)=min⁡αY,αD∫(T(y)−T(x(y,d))) dH(y,d)s.t.ρ≤B.(2)G(B)=\min_{\alpha_Y,\alpha_D}\int(T(y)-T(x(y,d)))\,\mathrm{d}H(y,d)\quad\text{s.t.}\quad\rho\le B. \tag*{(2)}

We assume B<ρY/π(<ρY)B < \rho_Y/\pi(<\rho_Y), so full deterrence via profit audits is infeasible.

Timing. Anticipating firms’ reporting responses, the IRS commits to AA. Firms observe their private types (y,d)(y,d) and submit reports (x,c)(x,c) optimally given AA. The IRS conducts audits according to AA and enforces taxes and penalties.14

Benchmark–Truthful data reporting

Consider as a benchmark the case where data are truthfully reported. The IRS can then condition enforcement directly on dd and optimally choose a budget B(d)B(d) allocated to audit class dd, with B=∫B(d)dHD(d)B=\int B(d)dH_D(d).

In a given class dd, the optimal audit policy can be represented by a cutoff y′(d)y'(d), with profit reports below y′(d)y'(d) audited at the minimal deterrence probability αY=1/π\alpha_Y=1/\pi.15 The within-class budget constraint is

B(d)=ρYπHY∣D(y′(d)∣d),(3)B(d)=\frac{\rho^Y}{\pi}H_{Y\mid D}(y'(d)\mid d), \tag*{(3)}

and the resulting expected tax gap is

G(d)=∫y′(d)yˉ(T(y)−T(y′(d))) dHY∣D(y∣d).(4)G(d)=\int_{y'(d)}^{\bar{y}}(T(y)-T(y'(d)))\,dH_{Y\mid D}(y\mid d). \tag*{(4)}

Differentiating the tax gap with respect to the audit budget yields

dG(d)dB(d)=−πρYT′(y′(d))γ(y′(d)∣d).\frac{dG(d)}{dB(d)}=-\frac{\pi}{\rho_Y}T'(y'(d))\gamma(y'(d)\mid d).

where

γ(y∣d):=1−HY∣D(y∣d)hY∣D(y∣d)\gamma(y\mid d):=\frac{1-H_{Y\mid D}(y\mid d)}{h_{Y\mid D}(y\mid d)}

denotes the inverse hazard rate of profits conditional on data dd. This expression gives an enforcement index, familiar from optimal-audit models with heterogeneous taxpayer classes ([23, 30]),

I(y∣d):=T′(y)γ(y∣d).(5)I(y \mid d) := T'(y)\gamma(y \mid d). \tag*{(5)}

Up to the common factor π/ρY\pi/\rho_Y, the index measures the reduction in the tax gap from extending audit coverage within class dd. It combines the marginal tax liability T′(y)T'(y) with the information rents generated by private information, summarized by γ(y∣d)\gamma(y \mid d); Appendix § 8 provides the derivation and incentive-theoretic interpretation.

The enforcement index determines how profit audits are allocated. For interior cutoffs, optimal allocation equalizes marginal enforcement gains across data classes:

I(y′(d)∣d)=κ,(6)I(y'(d) \mid d) = \kappa, \tag*{(6)}

where κ>0\kappa> 0 is the shadow value of the audit budget BB. Thus, the enforcement index summarizes in which classes additional profit-audit coverage is most valuable.

The slope of the index determines the margin of adjustment. As depicted in Figure 2, when I(y∣d)I(y \mid d) is decreasing in yy, marginal returns decline as audit coverage expands and enforcement adjusts on the intensive margin through class-specific cutoffs satisfying I(y′(d)∣d)=κI(y'(d) \mid d) = \kappa [Panel (a)]. When I(y∣d)I(y \mid d) is non-decreasing, the interior solution satisfying I(y′(d)∣d)=κI(y'(d) \mid d) = \kappa may not be optimal; instead, the IRS chooses between the boundary cutoffs.16 Optimal enforcement then operates on the extensive margin, with some classes fully enforced, y′(d)=yˉy'(d) = \bar{y}, and others exempt from profit audits, y′(d)=yy'(d) = y [Panel (b)].17

Figure 2. The enforcement index and optimal audit coverage. (a) Intensive margin (b) Extensive margin

Together, the preceding analysis explains how auxiliary data facilitate audit target-

Iˉ(d):=∫y‾yˉI(y∣d) dHY∣D(y∣d).\bar{I}(d) := \int_{\underline{y}}^{\bar{y}} I(y \mid d)\,\mathrm{d}H_{Y \mid D}(y \mid d).

A class is fully enforced if Iˉ(d)>κ\bar{I}(d) > \kappa and exempt from profit audits if Iˉ(d)<κ\bar{I}(d) < \kappa; at equality, the IRS is indifferent. With strategic data, the analogous condition depends on the endogenous verification frontier; see § 4. ing by identifying classes in which resources have the highest marginal value. When data become strategic, however, differentiated enforcement creates incentives for firms to misclassify their information into more leniently enforced classes. The upcoming section specifies how data verification resolves this incentive problem while preserving differentiated audit targeting.

Equilibrium–Strategic data reporting

We now allow firms to manipulate the data used for targeting optimal enforcement. As in the truthful-data benchmark, profit deterrence is characterized by class-specific cutoffs y′(d)y'(d): firms report profits truthfully below the cutoff and bunch at y′(d)y'(d) above it.

Strategic data reporting introduces a cross-class incentive constraint. When cutoffs differ across classes, a firm assigned to a more strongly enforced class can jointly under-report profits and claim membership in a more leniently enforced class (see the example offered in the introduction). Preventing such deviations requires verification of the reported data wherever a lower-cutoff class would otherwise provide an escape from enforcement.

Let

y~:=inf⁡dy′(d),y^:=sup⁡dy′(d),\tilde{y} := \inf_{d} y'(d), \qquad\hat{y} := \sup_{d} y'(d),

and call y^\hat{y} the (global) verification frontier.

Lemma 1 (Global verification frontier) Let y~<y^\tilde{y}<\hat{y}. Every report with x<y^x<\hat{y} must be verified in equilibrium through either a profit audit or data verification. Without loss of generality, the optimal hybrid audit policy can be represented as

αY(x,d)=1π1{x<y′(d)},αD(x,d)=1π1{y′(d)≤x<y^}.\alpha_Y(x,d)=\frac{1}{\pi}\mathbf{1}_{\{x<y'(d)\}}, \qquad\alpha_D(x,d)=\frac{1}{\pi}\mathbf{1}_{\{y'(d)\le x<\hat{y}\}}.

Proof. By definition, reports with profits below y~\tilde{y} are profit-audited in every data class. Now suppose that some report x∈[y~,y^)x\in[\tilde{y},\hat{y}) receives neither form of verification in a given data class cc. A firm with profit y>xy>x belonging to a class dd with cutoff y′(d)>xy'(d)>x could then jointly report data cc and profit level xx and escape enforcement. Hence, every report below the global frontier must be covered by either a profit audit or data verification. Data audits need not extend beyond y^\hat{y}: any firm (y,d)(y,d) can report its true data dd and profit min⁡{y,y′(d)}≤y^\min\{y,y'(d)\}\le\hat{y} without suffering penalties, and thus cannot gain from reporting x>y^x>\hat{y}, which would carry a higher tax. A profit audit directly deters profit underreporting, while data verification deters joint misreporting by triggering profit verification when a data discrepancy is detected. Under the maintained penalty structure, both instruments provide deterrence at probability 1/π1/\pi, yielding the stated representation.

For a given cutoff schedule, the resulting hybrid policy assigns profit audits to reports

x<y′(d)x < y'(d)

and data verification to

y′(d)≤x<y^.y'(d) \le x < \hat{y}.

Consequently, the audit technologies jointly cover all reports below the global frontier.

Note also that firms with true profits above their class-specific profit-audit cutoff bunch at that cutoff. Since the bunching report is subject to data verification when y′(d)<y^y'(d) < \hat{y}, all such firms undergo data verification, including those with true profits above the global frontier. Therefore, the cost of data verification depends on the mass of firms above the class-specific cutoff, rather than merely those with profits between the cutoff and the frontier. However, whenever y′(d)=y^y'(d) = \hat{y}, the bunching report lies outside the verification region, and no data verification is thus required in that class.

Intensive-margin enforcement

Consider first the case in which the enforcement index I(y∣d)=T′(y)γ(y∣d)I(y \mid d) = T'(y)\gamma(y \mid d) is strictly decreasing in profits (e.g., a linear tax schedule T(y)=τyT(y) = \tau y and Weibull profits with k>1k > 1). Given a fixed frontier y^\hat{y}, expected enforcement expenditure is

ρ=1π∫[ρYHY∣D(y′(d)∣d)+ρD(1−HY∣D(y′(d)∣d))1{y′(d)<y^}] dHD(d).(7)\rho= \frac{1}{\pi}\int\left[\rho_Y H_{Y\mid D}(y'(d)\mid d) + \rho_D\left(1-H_{Y\mid D}(y'(d)\mid d)\right)\mathbf{1}_{\{y'(d)<\hat{y}\}}\right]\,dH_D(d). \tag*{(7)}

The first component captures expenditure on profit audits. The second captures data-verification expenditure in classes whose profit-audit cutoff lies below the global frontier.

As firms with profits above their class-specific cutoff bunch at that cutoff, in these classes, data verification applies to the entire upper tail of the profit distribution. When the cutoff coincides with the global frontier, no data verification is required.

For classes with profit-audit cutoffs below the global frontier, the optimal allocation follows the same marginal condition as in the truthful-data benchmark:

I(y′(d)∣d)=σ,(8)I(y'(d)\mid d)=\sigma, \tag*{(8)}

whenever the cutoff is interior. The difference is that expanding profit audits now replaces data verification, reducing their marginal cost from ρY\rho_Y to ρY−ρD\rho_Y-\rho_D, as reflected in the shadow value of profit-audit coverage σ=λ(ρY−ρD)/π\sigma=\lambda(\rho_Y-\rho_D)/\pi.18

However, satisfying this marginal condition need not be globally optimal: extending profit audits to the global frontier eliminates data-verification expenditure altogether. The following proposition characterizes this additional choice.

Proposition 1 (Intensive-margin enforcement) Suppose that I(y∣d)I(y\mid d) is continuous and strictly decreasing in yy for every data class dd. For a given global frontier y^\hat{y}, let q<y^q<\hat{y} be the class-specific cutoff implied by the local enforcement index I(y∣d)I(y\mid d).

Extending profit audits to the global frontier is preferred to stopping at qq if and only if

∫qy^[σ−I(y∣d)]hY∣D(y∣d) dy≤λρDπ[1−HY∣D(y^∣d)].\int_{q}^{\hat{y}}[\sigma-I(y\mid d)]h_{Y\mid D}(y\mid d)\,dy\le\frac{\lambda\rho_D}{\pi}[1-H_{Y\mid D}(\hat{y}\mid d)].

Proposition 1 characterizes the tradeoff between differentiated audit targeting and the cost of preserving truthful data reporting. The left-hand side measures the loss from extending profit audits beyond the cutoff favored by marginal enforcement returns. The right-hand side captures the resulting savings in data-verification expenditure. As in the truthful-data benchmark, diminishing marginal returns favor differentiated, class-specific profit-audit cutoffs. Under strategic data reporting, however, the savings from eliminating data verification may outweigh these gains. Consequently, the optimal policy may combine classes with lower profit-audit cutoffs and data verification with classes in which profit audits extend all the way to the global frontier.

Proposition 2 (Global verification frontier) Suppose that I(y∣d)I(y\mid d) is continuous

and strictly decreasing in yy for every data class dd. Let D={d:y′(d)<y^}\mathcal{D} = \{d : y'(d) < \hat{y}\}. If the optimal global verification frontier y^\hat{y} is interior and the set D\mathcal{D} remains unchanged under sufficiently small perturbations of y^\hat{y}, then

∫Dc[I(y^∣d)−λρYπ]hY∣D(y^∣d) dHD(d)=0,\int_{\mathcal{D}^{c}} \left[I(\hat{y} \mid d) - \frac{\lambda\rho_Y}{\pi}\right] h_{Y \mid D}(\hat{y} \mid d)\,dH_D(d) = 0,

where λ>0\lambda> 0 is the multiplier on the enforcement budget constraint.

The condition balances the marginal gains from extending profit audits against their marginal costs, aggregated over classes audited up to the global frontier. For classes with lower profit-audit cutoffs, the frontier does not directly affect equilibrium reporting or verification expenditure. Thus, Proposition 2 demonstrates that the optimal frontier is determined by the classes that rely exclusively on profit audits.

Figure 3 describes the allocation of audits across data classes and true profits. Profit audits can be concentrated in classes with high enforcement returns, while data verification sustains this differentiation by preventing firms from selecting into more leniently enforced classes. The two instruments also reach different parts of the true-profit distri bution: profit audits tend to target lower-profit firms, whereas data verification reaches higher-profit firms whose reports bunch at class-specific audit cutoffs. As demonstrated by Macho-Stadler and Pérez-Castrillo [24], access to data can reverse the regressive bias in tax enforcement. The figure illustrates how hybrid enforcement can mitigate this bias by extending enforcement to higher-profit firms.

Cross-sectional audit allocation by true profit

Figure 3. Cross-sectional audit allocation by true profit. No audit; Profit audit; Data audit; Frontier y^\hat{y}.

Table 1 further exemplifies the resulting policy for three classes with

Reported profit xxClass d1d_1Class d2d_2Class d3d_3
x<y′(d1)x < y'(d_1)Profit auditProfit auditProfit audit
x∈[y′(d1),y′(d2))x \in[y'(d_1), y'(d_2))Data auditProfit auditProfit audit
x∈[y′(d2),y^)x \in[y'(d_2), \hat{y})Data auditData auditProfit audit
x≥y^x \ge\hat{y}No auditNo auditNo audit

Table 1. Cross-sectional audit allocation by reported profit.

y′(d1)<y′(d2)<y′(d3)=y^.y'(d_1) < y'(d_2) < y'(d_3) = \hat{y}.

Within each class, profit and data audits partition the verified region of reported profits: profit audits apply below the class-specific cutoff and data verification between that cutoff and the global frontier. Across classes, firms reporting the same profit may therefore face different audit instruments. The class with the highest cutoff relies exclusively on profit audits, while the remaining classes combine both instruments.

Remark 1 (Implications for bunching patterns) The policy also has a distinctive implication for reporting behavior: profit reports bunch at class-specific audit cutoffs, whereas correctly implemented data verification eliminates incentives to manipulate classification data.

Empirical studies document bunching of declared amounts around tax and enforcement thresholds, with excess mass concentrated on the side favored by the incentives created by the threshold [11, 19, 29]. Our model similarly predicts strategic bunching in profit declarations at the endogenous audit cutoff, on the side that avoids profit auditing. Bunching on the audited side would not be explained by these strategic incentives alone but may reflect optimization. frictions. By contrast, bunching or manipulation in classification data, as documented by Almunia and Lopez-Rodriguez [2] and Chen et al. [10], may be consistent with insufficient data verification.

Extensive-margin enforcement

The structure of the hybrid policy changes when the enforcement index is increasing in profits (e.g., truncated Pareto profits under a linear tax schedule, or a sufficiently progressive tax schedule), favoring boundary cutoffs. With strategic data reporting, however, the choice is between profit auditing up to the global verification frontier and relying entirely on data verification within a class.

For this comparison, define the frontier-dependent average enforcement index

Iˉ(d;y^):=∫yy^I(y∣d) dHY∣D(y∣d)HY∣D(y^∣d).(9)\bar{I}(d;\hat{y}) := \frac{\int_{y}^{\hat{y}} I(y \mid d)\,\mathrm{d}H_{Y\mid D}(y \mid d)}{H_{Y\mid D}(\hat{y} \mid d)}. \tag*{(9)}

This index measures the average return to profit auditing over the region below the global verification frontier. Classes with sufficiently high average enforcement returns are profit-audited up to the frontier, whereas classes with lower returns rely on data verification. Unlike in the truthful-data benchmark, the choice also depends on the relative costs of the two enforcement technologies.

Proposition 3 (Extensive-margin enforcement) Suppose that I(y∣d)I(y \mid d) is continuous and strictly increasing in yy for every data class dd. For a given global frontier y^\hat{y}, a class is either profit audited up to the frontier or instead subject to data verification.

Profit auditing up to the global frontier is preferred to data verification if and only if

Iˉ(d;y^)≥λπ[ρY−ρDHY∣D(y^∣d)].\bar{I}(d;\hat{y}) \ge\frac{\lambda}{\pi}\left[\rho_Y-\frac{\rho_D}{H_{Y\mid D}(\hat{y} \mid d)}\right].

The global verification frontier and the allocation of enforcement across classes are determined jointly, subject to the aggregate budget constraint. The proposition reveals

a distinction between the two enforcement technologies. Without profit audits, all firms in a class bunch at the lowest profit report. Their reports are hence subject to data verification, making the resulting verification expenditure independent of the global frontier. By contrast, profit-audit expenditure also depends on the fraction of firms below the frontier whose profits are audited. Consequently, other things equal, profit auditing becomes more attractive in classes with a smaller fraction of firms below the global frontier. Profit audits thus concentrate in higher-profit classes while targeting lower-profit firms within them.

Example 1 (Extending enforcement across classes or profit levels) Consider two equally likely classes dH,dLd_H,d_L, with profits y∈{0,1,2}y \in\{0,1,2\} and conditional distributions

y=0y=0y=1y=1y=2y=2
dHd_H0.60.20.2
dLd_L0.80.20

Table 6.

Let T(0)=0T(0)=0, T(1)=1T(1)=1, T(2)=7T(2)=7, π=2\pi=2, ρY=2\rho_Y=2, and ρD=1.2\rho_D=1.2. Profit reports strictly below the audit cutoff are audited, while firms with higher profits bunch at the cutoff.

Suppose initially that y^=1\hat{y}=1, with dHd_H profit-audited up to the frontier and dLd_L subject to data verification. The IRS can either extend profit auditing to dLd_L or raise the frontier to y^=2\hat{y}=2, retaining the original class allocation. Both expansions cost 0.10 in aggregate, but their reductions in the aggregate tax gap are, respectively, 0.22T(1)=0.1\frac{0.2}{2}T(1)=0.1 and 0.22[T(2)−T(1)]=0.6\frac{0.2}{2}[T(2)-T(1)]=0.6. Thus, extending the frontier yields the greater return. If instead T(2)=1.5T(2)=1.5, this benefit falls to 0.05, reversing the ranking in favor of expanding audits across classes.

While the shape of the enforcement index determines how profit audits are allocated, Propositions 1 and 3 reveal a common principle. The verification frontier is driven by the cross-class incentive constraint: whenever taxpayers can manipulate the tax base by misreporting their class, targeted enforcement requires data verification in more leniently enforced classes.

Remark 2 (Profit-audit exemptions) *With strategic data, exemption from profit audits

dits does not imply the absence of enforcement. An unverified class otherwise creates an enforcement loophole, because firms can avoid profit-audits through data misclassification. Exempt classes must, therefore, be data-verified wherever they provide an escape from profit audit.19

Our characterization identifies both the benefit and the cost of data-contingent enforcement. Data improve targeting by allowing profit audits to be directed toward classes with higher enforcement returns. When data are strategic, however, sustaining this differentiation requires additional verification. The following section investigates this tradeoff more closely.

Value of data

We characterize the value of introducing self-reported data into enforcement by comparing the expected tax gap across three enforcement regimes: GAG^{A} when the IRS ignores data, GBG^{B} when data are truthful, and GCG^{C} when data are strategic and optimally verified. The corresponding tax gaps are

GA=∫y^Ayˉ(T(y)−T(y^A)) dHY(y),G^{A} = \int_{\hat{y}_{A}}^{\bar{y}} \left(T(y) - T(\hat{y}_{A})\right) \, dH_{Y}(y),

and, for j∈{B,C}j \in\{B,C\},

Gj=∫∫yj′(d)yˉ(T(y)−T(yj′(d)))dHY∣D(y∣d)dHD(d),G^{j} = \int\int_{y'_{j}(d)}^{\bar{y}} \left(T(y) - T(y'_{j}(d))\right) dH_{Y\mid D}(y \mid d)dH_{D}(d),

where the cutoffs yB′(d)y'_{B}(d) and yC′(d)y'_{C}(d) are characterized in § 3 and § 4, respectively. Formal derivations are provided in Appendix § 10.

Because the IRS can always ignore classification data, and truthful data relax the incentive constraints imposed by strategic reporting,

GB≤GC≤GA.G^{B} \le G^{C} \le G^{A}.

The difference between GBG^{B} and GAG^{A} identifies the targeting value of data:

VT:=GA−GB.V_{T} := G^{A} - G^{B}.

It measures the reduction in the tax gap attainable when classification data can be used without cost to direct profit audits toward classes with the highest enforcement returns.

When data are strategic, part of this targeting value must be sacrificed to make differentiated enforcement incentive compatible. Under the optimal hybrid policy, the enforcement budget is divided between profit audits,

BY=ρYπ∫HY∣D(yC′(d)∣d)dHD(d),B_{Y} = \frac{\rho_{Y}}{\pi} \int H_{Y\mid D}(y'_{C}(d)\mid d)dH_{D}(d),

and data verification,

BD=ρDπ∫[1−HY∣D(yC′(d)∣d)]1{yC′(d)<y^C}dHD(d),B_{D} = \frac{\rho_{D}}{\pi} \int[1-H_{Y\mid D}(y'_{C}(d)\mid d)]\mathbf{1}_{\{y'_{C}(d)<\hat{y}_{C}\}}dH_{D}(d),

with BY+BD=BB_{Y}+B_{D}=B. The opportunity cost of this verification for the IRS is

CS:=GC−GB.C_{S} := G^{C} - G^{B}.

Hence, the net enforcement value of strategic data admits the decomposition

GA−GC=GA−GB⏟targeting value−(GC−GB)⏟implementation cost.(10)G^{A} - G^{C} = \underbrace{G^{A} - G^{B}}_{\text{targeting value}} - \underbrace{(G^{C} - G^{B})}_{\text{implementation cost}}. \tag*{(10)}

Proposition 4 (Value of data) The tax gaps under truthful data, optimally verified strategic data, and no data satisfy

GB≤GC≤GA.G^{B} \le G^{C} \le G^{A}.

The gross targeting value of classification data is GA−GBG^A-G^B, while their net enforcement value under strategic reporting is

GA−GC=(GA−GB)−(GC−GB).G^A-G^C=(G^A-G^B)-(G^C-G^B).

Example 2 (Informative data without enforcement value) Consider two equally likely data classes, d1d_1 and d2d_2, with profits y∈{0,1,2}y \in\{0,1,2\} distributed as follows:

y=0y=0y=1y=1y=2y=2
d1d_10.400.050.55
d2d_20.900.010.09

Let T(y)=yT(y)=y, π=2\pi=2, ρY=2\rho_Y=2, ρD=1.9\rho_D=1.9, and B=0.65B=0.65. Reports are restricted to the three profit levels, and profit reports strictly below the cutoff are audited.

Without data, the optimal uniform cutoff is 1, yielding GA=0.32G^A=0.32. With truthful data, the IRS instead chooses yB(d1)=2y_B(d_1)=2 and yB(d2)=0y_B(d_2)=0, yielding GB=0.095G^B=0.095.

When data are strategic, however, every differentiated cutoff allocation requires data verification and exceeds the budget. The IRS therefore retains the uniform cutoff 1, implying GB<GC=GAG^B<G^C=G^A. Thus, informative classification data can have no enforcement value when reporting is strategic.

Suppose, in addition, that introducing classification data into the enforcement system entails a fixed cost K≥0K\geq0. This may capture any costs of data administration or IT infrastructure not included in the audit budget. Data should then be introduced if and only if

GA−GB⏟targeting value≥GC−GB⏟implementation cost+K⏟data-introduction cost.(11)\underbrace{G^A-G^B}_{\text{targeting value}} \geq \underbrace{G^C-G^B}_{\text{implementation cost}} + \underbrace{K}_{\text{data-introduction cost}}. \tag*{(11)}

Thus, introducing informative data into tax enforcement is not necessarily valuable to the authority: their targeting gains must be sufficiently large to compensate both for the resources required to sustain truthful classification and for the cost of introducing the data.

Verification is essential for realizing targeting gains. If the IRS conditions enforcement on manipulable data without verifying them, firms select into the most leniently enforced class. Writing the resulting pooling cutoff as y~=inf⁡dy′(d)\tilde{y} = \inf_d y'(d), the tax gap becomes

GD=∫y~yˉ(T(y)−T(y~))dHY(y).G^D = \int_{\tilde{y}}^{\bar{y}} (T(y) - T(\tilde{y}))dH_Y(y).

For a budget-feasible pooling policy, GD≥GA≥GCG^D \ge G^A \ge G^C. The first inequality follows because budget feasibility requires y~≤y^A\tilde{y} \le\hat{y}_A, with strict inequality whenever y~<y^A\tilde{y} < \hat{y}_A. Thus, naive reliance on manipulable data can perform worse than ignoring them altogether.

Corollary 1 Introducing manipulable data at cost KK is optimal if and only if

GA−GB≥(GC−GB)+K.G^A - G^B \ge(G^C - G^B) + K.

If data are used naively without verification, then GD≥GA(≥GC≥GB)G^D \ge G^A(\ge G^C \ge G^B).

Figure 4. Tax gaps across enforcement regimes.

Figure 4 describes the tax-gap comparisons in Proposition 4 and Corollary 1, based on the calibrated quantitative illustration in Appendix § 15. Moving from the no-data benchmark (A) to the truthful-data benchmark (B) reduces the tax gap from GA=0.151G^A = 0.151 to GB=0.140G^B = 0.140, yielding a gross targeting value of 0.011. Under strategic reporting, optimal verification (C) yields GC=0.148G^C = 0.148, whereas naive reliance on unverified data (D) yields GD=0.186G^D = 0.186. At the baseline relative verification cost ρD/ρY=0.03\rho_D/\rho_Y = 0.03, verification hence absorbs about 9.5 percent of the enforcement budget, and strategic reporting dissipates approximately 67 percent of the gross targeting value of classification data. Interestingly, as illustrated in Figure 5, the optimal policy tailored for strategic data reporting features partial pooling of enforcement cutoffs: two of the three classes share a profit-audit cutoff, while the third is treated differently. Thus, verification costs affect not only the value of data but also the optimal degree of class-by-class differentiation.

Profit-audit cutoffs across enforcement regimes

Figure 5. Profit-audit cutoffs across enforcement regimes.

Literature

Our analysis builds on the literature studying how auxiliary information can improve the design of tax enforcement. Scotchmer (1987) shows that audit probabilities should condition on taxpayer characteristics and bounds the maximum audit probability by the inverse of penalties. Sanchez and Sobel (1993) establishes the familiar cutoff structure in which evasion is deterred up to an income threshold, above which taxpayers are not audited. Macho-Stadler and Pérez-Castrillo (1997) extends the analysis to multiple income sources: audit policies differ across sources, while within each source only incomes below a source-specific cutoff are audited. Macho-Stadler and Pérez-Castrillo (2002) further demonstrate how signals correlated with income can be used to target audits and may reverse the standard regressive bias of optimal enforcement policy. Cremer and Gahvari (1996), Aronsson and Xu (2026), and Canta et al. (2024) study optimal taxation with tax evasion in the Mirrleesian tradition with multiple taxpayer types.

Our model bridges the cutoff structure of Sanchez and Sobel [30] and the differentiated enforcement of Macho-Stadler and Pérez-Castrillo [24]. As in Macho-Stadler and Pérez-Castrillo [24], auxiliary information allows profit enforcement to differ across data classes. Once the information used for targeting can itself be manipulated, however, these class-specific policies can no longer be chosen independently. Cross-class incentive compatibility generates a single global verification frontier, much like in Sanchez and Sobel [30]: every report below this frontier must be verified. Data verification connects differentiated profit enforcement to this common frontier. When marginal returns to profit auditing are decreasing, profit audits extend to class-specific cutoffs, with data verification filling the remaining gap to the frontier. When marginal returns are increasing, the hybrid structure instead operates class by class: some classes are profit-audited up to the frontier, while others are enforced through cheaper data verification.20

Kuchumova [21] also endogenizes the information available for enforcement by allowing the tax authority to invest in the accuracy of external information reporting.21 The authority must then decide how much of a limited enforcement budget to devote to improving information and how much to audits. More accurate information improves targeting, but its value falls with audit coverage. Our approach instead takes the informativeness of classification data as given but allows taxpayers to manipulate the information used for targeting. Data verification then has a direct implementation role: it prevents taxpayers from selecting into more leniently enforced classes and thereby sustains differentiated profit enforcement. This creates a tradeoff between the targeting gains from classification data and the verification resources required to use them. We characterize the resulting net value of data by comparing optimal enforcement with strategic data to the implementable benchmark in which the authority ignores classification data.

Our treatment of classification data is motivated by a broad empirical literature on bebehavioral responses to the design of tax policy and enforcement. This literature indicates that taxpayers respond to different policies not only along the conventional margin of income misreporting, but also by changing reported characteristics on which tax rates, deductions, exemptions, and audit treatment are later conditioned [1, 2, 5, 10, 13, 15] (Almunia et al. 2024; Kuchumova et al. 2025).

Conclusion

How can tax authorities use detailed taxpayer data without introducing new opportunities for evasion? When classification data are potentially manipulable, we demonstrate that optimal enforcement is organized around a global verification frontier, with data verification preventing firms from dodging enforcement through misclassification. The structure of optimal hybrid tax enforcement depends on whether profit auditing operates on the intensive or extensive margin, with the costs of data verification restricting differentiation across data classes. Data are therefore valuable only when the associated targeting gains exceed the costs of verification and those of introducing data into enforcement; without credible verification, reliance on self-classified data can perform worse than disregarding data altogether. When successfully implemented, the optimal policy generates bunching in profits but involves no data bunching, providing an empirical diagnostic for insufficient data verification. Taken together, these results provide the first unified account of how the possibility of manipulating classification data affects optimal tax enforcement and the resulting patterns of reporting behavior.

Omitted results

Truthful data reporting

For completeness, the following analysis derives the enforcement index and information rents directly from the IRS problem (Appendix § 8.1.1) and then provides the mechanism-design interpretation of optimal audit policy (Appendix § 8.1.2).

Enforcement index and information rents

Fix a data class dd and write y′:=y′(d)y^{\prime} := y^{\prime}(d). From

G(d)=∫y′yˉ(T(y)−T(y′)) dHY∣D(y∣d),G(d) = \int_{y^{\prime}}^{\bar{y}} \left(T(y) - T(y^{\prime})\right) \, \mathrm{d}H_{Y \mid D}(y \mid d),

Differentiating this using Leibniz’s rule gives

dG(d)dy′=−(T(y′)−T(y′))hY∣D(y′∣d)−T′(y′)∫y′yˉdHY∣D(y∣d),\frac{\mathrm{d}G(d)}{\mathrm{d}y^{\prime}} = -\left(T(y^{\prime}) - T(y^{\prime})\right)h_{Y \mid D}(y^{\prime} \mid d) - T^{\prime}(y^{\prime}) \int_{y^{\prime}}^{\bar{y}} \mathrm{d}H_{Y \mid D}(y \mid d),

and hence

dG(d)dy′=−T′(y′)(1−HY∣D(y′∣d)).\frac{\mathrm{d}G(d)}{\mathrm{d}y^{\prime}} = -T^{\prime}(y^{\prime})\left(1 - H_{Y \mid D}(y^{\prime} \mid d)\right).

Increasing the cutoff by dy′dy^{\prime} raises the reported tax liability of every firm above the cutoff by approximately T′(y′)dy′T^{\prime}(y^{\prime})dy^{\prime}. The mass of these firms is 1−HY∣D(y′∣d)1 - H_{Y \mid D}(y^{\prime} \mid d). The boundary term vanishes because a firm at the cutoff has no tax gap.

The within-class budget satisfies

B(d)=ρYπHY∣D(y′∣d),dB(d)dy′=ρYπhY∣D(y′∣d),B(d) = \frac{\rho_Y}{\pi} H_{Y \mid D}(y^{\prime} \mid d), \qquad\frac{\mathrm{d}B(d)}{\mathrm{d}y^{\prime}} = \frac{\rho_Y}{\pi} h_{Y \mid D}(y^{\prime} \mid d),

which gives

dG(d)dB(d)=−πρYT′(y′)1−HY∣D(y′∣d)hY∣D(y′∣d)=−πρYI(y′∣d).\frac{\mathrm{d}G(d)}{\mathrm{d}B(d)} = -\frac{\pi}{\rho_Y} T^{\prime}(y^{\prime}) \frac{1 - H_{Y \mid D}(y^{\prime} \mid d)}{h_{Y \mid D}(y^{\prime} \mid d)} = -\frac{\pi}{\rho_Y} I(y^{\prime} \mid d).

The same expression gives the information-rent interpretation of the index. A marginal increase in the cutoff by dy′dy^{\prime} reduces the reporting rent of every firm above y′y^{\prime} by approximately T′(y′)dy′T^{\prime}(y^{\prime})dy^{\prime}. The mass of these inframarginal firms is 1−HY∣D(y′∣d)1-H_{Y\mid D}(y^{\prime}\mid d), while the marginal mass that must be brought under audit coverage is hY∣D(y′∣d)dy′h_{Y\mid D}(y^{\prime}\mid d)dy^{\prime}. Thus the rent reduction per marginal firm brought under enforcement is

T′(y′)1−HY∣D(y′∣d)hY∣D(y′∣d)=T′(y′)γ(y′∣d)=I(y′∣d).T^{\prime}(y^{\prime})\frac{1-H_{Y\mid D}(y^{\prime}\mid d)}{h_{Y\mid D}(y^{\prime}\mid d)}=T^{\prime}(y^{\prime})\gamma(y^{\prime}\mid d)=I(y^{\prime}\mid d).

The inverse hazard rate therefore summarizes the information rents generated by incomplete enforcement.

Mechanism design and optimal enforcement

The optimal profit-audit policy can also be derived using standard mechanism-design arguments.

Fix a data class dd. Consider a firm with profit yy contemplating a lower profit report x<yx<y. Relative to truthful reporting, underreporting of profit yields

(1−παY(x,d))[T(y)−T(x)].(1-\pi\alpha_Y(x,d))[T(y)-T(x)].

As TT is increasing, tax evasion is deterred if and only if

αY(x,d)≥1π.\alpha_Y(x,d)\geq\frac{1}{\pi}.

Because audits are costly, the IRS applies the minimum audit probability 1/π1/\pi required for deterrence on enforced reports and otherwise sets a zero probability. Therefore, in an optimal audit policy,

αY(x,d)∈{0,1π}.\alpha_Y(x,d)\in\left\{0,\frac{1}{\pi}\right\}.

Incentive compatibility further requires the profit-audit probability assigned to equilibrium reports to be non-increasing in profits. To see this, consider y2>y1y_2>y_1. The two incentive constraints imply

y1−T(x(y1,d))−Φ(x(y1,d),d∣y1,d,A)≥y1−T(x(y2,d))−Φ(x(y2,d),d∣y1,d,A),y_1 - T(x(y_1,d)) - \Phi(x(y_1,d),d \mid y_1,d,A) \ge y_1 - T(x(y_2,d)) - \Phi(x(y_2,d),d \mid y_1,d,A),
y2−T(x(y2,d))−Φ(x(y2,d),d∣y2,d,A)≥y2−T(x(y1,d))−Φ(x(y1,d),d∣y2,d,A).y_2 - T(x(y_2,d)) - \Phi(x(y_2,d),d \mid y_2,d,A) \ge y_2 - T(x(y_1,d)) - \Phi(x(y_1,d),d \mid y_2,d,A).

Adding these constraints and using the definition of Φ\Phi gives

π[αY(x(y1,d),d)−αY(x(y2,d),d)][T(y2)−T(y1)]≥0.\pi[\alpha_Y(x(y_1,d),d) - \alpha_Y(x(y_2,d),d)] [T(y_2) - T(y_1)] \ge0.

Since taxes are increasing,

αY(x(y1,d),d)≥αY(x(y2,d),d).\alpha_Y(x(y_1,d),d) \ge\alpha_Y(x(y_2,d),d).

Together with αY∈{0,1/π}\alpha_Y \in\{0,1/\pi\}, this implies a cutoff rule: αY(x,d)=1/π\alpha_Y(x,d) = 1/\pi for x<y′(d)x < y'(d) and αY(x,d)=0\alpha_Y(x,d) = 0 for x≥y′(d)x \ge y'(d).

The envelope theorem applied to the firm’s problem yields

∂Π(y,d)∂y=1−παY(x(y,d),d)T′(y).\frac{\partial\Pi(y,d)}{\partial y} = 1 - \pi\alpha_Y(x(y,d),d)T'(y).

Integrating from yy to y‾\underline{y},

Π(y,d)=Π(y‾,d)+(y‾−y)−π∫yy‾αY(x(s,d),d)T′(s) ds.\Pi(y,d) = \Pi(\underline{y},d) + (\underline{y} - y) - \pi\int_y^{\underline{y}} \alpha_Y(x(s,d),d)T'(s)\,ds.

Thus, when profit auditing is positive at αY=1/π\alpha_Y = 1/\pi over [y,y‾][y,\underline{y}], it reduces the information rent of type yy by ∫yy‾T′(s) ds\int_y^{\underline{y}} T'(s)\,ds.

For a fixed data class, minimizing the tax gap is equivalent to maximizing expected reported tax payments, since the statutory tax liability ∫T(y)dHY∣D(y∣d)\int T(y)dH_{Y\mid D}(y \mid d) is fixed.

From the firm’s equilibrium payoff,

T(x(y,d))=y−Π(y,d)−Φ(x(y,d),d∣y,d,A).T(x(y,d)) = y - \Pi(y,d) - \Phi(x(y,d),d \mid y,d,A).

Substituting the envelope representation and taking expectations gives

∫y‾yˉT(x(y,d)) dHY∣D(y∣d)=y‾−Π(y‾,d)+π∫y‾yˉ∫y‾yαY(x(s,d),d)T′(s) ds dHY∣D(y∣d)−∫y‾yˉΦ(x(y,d),d∣y,d,A) dHY∣D(y∣d).\int_{\underline{y}}^{\bar{y}} T(x(y,d))\,dH_{Y\mid D}(y\mid d)=\underline{y}-\Pi(\underline{y},d) +\pi\int_{\underline{y}}^{\bar{y}}\int_{\underline{y}}^{y}\alpha_Y(x(s,d),d)T'(s)\,ds\,dH_{Y\mid D}(y\mid d) -\int_{\underline{y}}^{\bar{y}}\Phi(x(y,d),d\mid y,d,A)\,dH_{Y\mid D}(y\mid d).

Applying Fubini’s theorem,

∫y‾yˉT(x(y,d)) dHY∣D(y∣d)=y‾−Π(y‾,d)+π∫y‾yˉαY(x(y,d),d)T′(y)[1−HY∣D(y∣d)] dy−∫y‾yˉΦ(x(y,d),d∣y,d,A) dHY∣D(y∣d).\int_{\underline{y}}^{\bar{y}} T(x(y,d))\,dH_{Y\mid D}(y\mid d)=\underline{y}-\Pi(\underline{y},d) +\pi\int_{\underline{y}}^{\bar{y}}\alpha_Y(x(y,d),d)T'(y)[1-H_{Y\mid D}(y\mid d)]\,dy -\int_{\underline{y}}^{\bar{y}}\Phi(x(y,d),d\mid y,d,A)\,dH_{Y\mid D}(y\mid d).

Under the optimal implementation, expected penalties vanish on path: profit-audited reports are truthful, while reports involving profit understatement are not audited. Hence,

∫y‾yˉT(x(y,d)) dHY∣D(y∣d)=y‾−Π(y‾,d)+π∫y‾yˉαY(x(y,d),d)T′(y)γ(y∣d) dHY∣D(y∣d).\int_{\underline{y}}^{\bar{y}} T(x(y,d))\,dH_{Y\mid D}(y\mid d)=\underline{y}-\Pi(\underline{y},d)+\pi\int_{\underline{y}}^{\bar{y}}\alpha_Y(x(y,d),d)T'(y)\gamma(y\mid d)\,dH_{Y\mid D}(y\mid d).

Consequently, apart from the lowest type’s payoff, the IRS’s objective assigns profit-audit coverage the mechanism-design weight

I(y∣d)=T′(y)γ(y∣d).I(y\mid d)=T'(y)\gamma(y\mid d).

The term T′(y)T'(y) captures the marginal tax liability, while γ(y∣d)\gamma(y\mid d) reflects the information rents reduced by extending enforcement at profit yy.

In other words, the enforcement index measures the return to extending profit-audit coverage. Let κ>0\kappa>0 denote the shadow value of the aggregate enforcement budget. With a decreasing index, the IRS acts on these returns pointwise, extending profitaudit coverage until

I(y∣d)=κ.I(y \mid d) = \kappa.

With an increasing index, an interior class-specific cutoff is not optimal. The IRS instead chooses between the boundary policies,

y′(d)∈{y‾,yˉ},y'(d) \in\{\underline{y}, \bar{y}\},

so that a class is either fully profit-audited or exempt from profit audits. The choice is governed by the average enforcement index

Iˉ(d)=∫y‾yˉI(y∣d) dHY∣D(y∣d).\bar{I}(d) = \int_{\underline{y}}^{\bar{y}} I(y \mid d) \, dH_{Y\mid D}(y \mid d).

A class is fully enforced if Iˉ(d)>κ\bar{I}(d) > \kappa and exempt from audits if Iˉ(d)<κ\bar{I}(d) < \kappa, with indifference at equality.22 Thus, the same enforcement index governs both margins: pointwise returns determine the intensive-margin allocation, whereas their average determines the extensive-margin allocation across classes.

Strategic data reporting

Proof of Proposition 1. Fix the global frontier y^\hat{y} and a data class dd. For a profit-audit cutoff q<y^q < \hat{y}, expected enforcement expenditure is

Cd(q)=1π[ρYHY∣D(q∣d)+ρD(1−HY∣D(q∣d))].C_d(q) = \frac{1}{\pi}\left[\rho_Y H_{Y\mid D}(q \mid d) + \rho_D\left(1 - H_{Y\mid D}(q \mid d)\right)\right].

Increasing qq replaces data verification by profit auditing. The marginal reduction in the tax gap is

T′(q)[1−HY∣D(q∣d)],T'(q)\left[1 - H_{Y\mid D}(q \mid d)\right],

while the marginal resource cost is

ρY−ρDπhY∣D(q∣d).\frac{\rho_Y - \rho_D}{\pi}h_{Y\mid D}(q \mid d).

Let Ld(t)=Gd(t)+λCd(t)\mathcal{L}_d(t)=G_d(t)+\lambda C_d(t) denote the class-specific Lagrangian, where Gd(t)G_d(t) is the expected tax gap and λ\lambda is the multiplier on the aggregate enforcement budget. For t<y^t<\hat{y}, its derivative is

dLd(t)dt=hY∣D(t∣d)[σ−I(t∣d)],σ=λ(ρY−ρD)π.\frac{d\mathcal{L}_d(t)}{dt}=h_{Y\mid D}(t\mid d)[\sigma-I(t\mid d)],\qquad\sigma=\frac{\lambda(\rho_Y-\rho_D)}{\pi}.

Since I(⋅∣d)I(\cdot\mid d) is strictly decreasing, the Lagrangian decreases where I(t∣d)>σI(t\mid d)>\sigma and increases where I(t∣d)<σI(t\mid d)<\sigma. Thus, the preferred cutoff below the frontier is uniquely determined by

I(q∣d)=σ,I(q\mid d)=\sigma,

if interior, or by the corresponding lower boundary. If the Lagrangian is decreasing throughout the interval below the frontier, extending profit audits to the frontier is optimal.

At the global frontier, data-verification expenditure is eliminated altogether, so that

Cd(y^)=ρYπHY∣D(y^∣d).C_d(\hat{y})=\frac{\rho_Y}{\pi}H_{Y\mid D}(\hat{y}\mid d).

Since the tax gap is continuous in the cutoff, the Lagrangian has a downward jump at the frontier of size λρD[1−HY∣D(y^∣d)]/π\lambda\rho_D[1-H_{Y\mid D}(\hat{y}\mid d)]/\pi. This may make extending profit audits to y^\hat{y} preferable to the cutoff qq implied by the enforcement index. Comparing these two cutoffs gives

Ld(y^)−Ld(q)=−∫qy^I(y∣d)hY∣D(y∣d) dy+λπ[ρYHY∣D(y^∣d)−ρD−(ρY−ρD)HY∣D(q∣d)]=∫qy^[σ−I(y∣d)]hY∣D(y∣d) dy−λρDπ[1−HY∣D(y^∣d)].\begin{aligned} \mathcal{L}_d(\hat{y})-\mathcal{L}_d(q) &=-\int_q^{\hat{y}} I(y\mid d)h_{Y\mid D}(y\mid d)\,dy \\ &\quad+\frac{\lambda}{\pi}\left[\rho_Y H_{Y\mid D}(\hat{y}\mid d)-\rho_D-(\rho_Y-\rho_D)H_{Y\mid D}(q\mid d)\right] \\ &=\int_q^{\hat{y}}[\sigma-I(y\mid d)]h_{Y\mid D}(y\mid d)\,dy \\ &\quad-\frac{\lambda\rho_D}{\pi}[1-H_{Y\mid D}(\hat{y}\mid d)]. \end{aligned}

Thus, extending profit audits to the global frontier is preferred if and only if this difference is nonpositive, yielding the stated condition.

Proof of Proposition 2. By Proposition 1, conditional on the global frontier y^\hat{y}, each class either chooses a lower cutoff y′(d)<y^y'(d) < \hat{y} or extends profit audits to the frontier.

Consider a local change in y^\hat{y}. For classes in D\mathcal{D}, the optimal cutoff is determined by I(y′(d)∣d)=σI(y'(d) \mid d) = \sigma, or by the corresponding lower boundary, and is independent of y^\hat{y}. Their contribution to the Lagrangian therefore remains unchanged.

For classes in Dc\mathcal{D}^{c}, the profit-audit cutoff equals y^\hat{y} and no data verification is required. Using the class-specific Lagrangian defined in the proof of Proposition 1, we obtain

dLd(y^)dy^=hY∣D(y^∣d)[λρYπ−I(y^∣d)].\frac{d\mathcal{L}_{d}(\hat{y})}{d\hat{y}} = h_{Y\mid D}(\hat{y}\mid d)\left[\frac{\lambda\rho_{Y}}{\pi} - I(\hat{y}\mid d)\right].

Since the allocation of classes between the two regimes is locally unchanged, the derivative of the aggregate Lagrangian is obtained by integrating this expression over Dc\mathcal{D}^{c}. At an interior optimum, this derivative must vanish, yielding the stated condition. □\square

Proof of Proposition 3. Fix the global verification frontier y^\hat{y} and a data class dd. For any cutoff q<y^q < \hat{y}, expected enforcement expenditure is

Cd(q)=1π[ρYHY∣D(q∣d)+ρD(1−HY∣D(q∣d))].C_{d}(q) = \frac{1}{\pi}\left[\rho_{Y}H_{Y\mid D}(q\mid d) + \rho_{D}\left(1-H_{Y\mid D}(q\mid d)\right)\right].

Let

Ld(q)=Gd(q)+λCd(q)\mathcal{L}_{d}(q) = G_{d}(q) + \lambda C_{d}(q)

denote the class-specific Lagrangian, where Gd(q)G_{d}(q) is the expected tax gap and λ\lambda is the multiplier on the aggregate enforcement budget.

For cutoffs below the frontier, differentiation gives

dLd(q)dq=hY∣D(q∣d)[λ(ρY−ρD)π−I(q∣d)].\frac{d\mathcal{L}_{d}(q)}{dq} = h_{Y\mid D}(q\mid d)\left[\frac{\lambda(\rho_{Y}-\rho_{D})}{\pi} - I(q\mid d)\right].

Since I(⋅∣d)I(\cdot\mid d) is strictly increasing, any interior stationary point is a local maximum rather than a minimum. Hence, the preferred cutoff among those below the frontier lies at a boundary, either yy or the limit as qq approaches y^\hat{y} from below.

Tax-gap comparisons and the value of data

At the global frontier, data verification is no longer required, so that

Cd(y^)=ρYπHY∣D(y^∣d).C_d(\hat{y}) = \frac{\rho^Y}{\pi} H_{Y\mid D}(\hat{y}\mid d).

The tax gap is continuous in the cutoff, while the enforcement cost falls discontinuously at y^\hat{y} by

ρDπ[1−HY∣D(y^∣d)].\frac{\rho^D}{\pi}\left[1-H_{Y\mid D}(\hat{y}\mid d)\right].

Consequently, the frontier cutoff dominates the limit from below, leaving only two candidates:

y′(d)∈{y,y^}.y'(d) \in\{y,\hat{y}\}.

At the lower cutoff, no profit audits are conducted. All firms bunch at yy and undergo data verification, giving

Cd(y)=ρDπ.C_d(y) = \frac{\rho^D}{\pi}.

The reduction in the tax gap from choosing profit auditing up to the frontier instead is

Gd(y)−Gd(y^)=∫yy^I(y∣d)hY∣D(y∣d) dy.G_d(y)-G_d(\hat{y}) = \int_y^{\hat{y}} I(y\mid d)h_{Y\mid D}(y\mid d)\,\mathrm{d}y.

The corresponding difference in enforcement expenditure is

Cd(y^)−Cd(y)=1π[ρYHY∣D(y^∣d)−ρD].C_d(\hat{y})-C_d(y) = \frac{1}{\pi}\left[\rho^Y H_{Y\mid D}(\hat{y}\mid d)-\rho^D\right].

Thus,

Ld(y^)−Ld(y)=−∫yy^I(y∣d)hY∣D(y∣d) dy+λπ[ρYHY∣D(y^∣d)−ρD].\mathcal{L}_d(\hat{y})-\mathcal{L}_d(y) = -\int_y^{\hat{y}} I(y\mid d)h_{Y\mid D}(y\mid d)\,\mathrm{d}y +\frac{\lambda}{\pi}\left[\rho^Y H_{Y\mid D}(\hat{y}\mid d)-\rho^D\right].

Profit auditing up to the global frontier is preferred to data verification if and only if this difference is nonpositive, yielding the stated condition. □\square

Tax-gap comparisons and the value of data

This appendix derives the tax-gap expressions used in § 5 and proves Proposition 4. We consider the four enforcement regimes introduced in the main text.

Tax gaps across enforcement regimes

Regime A: no data. When the IRS ignores classification data, all firms face the same profit-audit cutoff y^A\hat{y}_A. The budget constraint is

ρYπHY(y^A)=B.(12)\frac{\rho_Y}{\pi} H_Y(\hat{y}_A) = B. \tag*{(12)}

Equilibrium profit reporting is

xA(y)=min⁡{y,y^A}.x_A(y) = \min\{y,\hat{y}_A\}.

Firms below the cutoff report truthfully and generate no tax gap, whereas firms above the cutoff report y^A\hat{y}_A. Hence

GA=∫y^Ayˉ(T(y)−T(y^A)) dHY(y).(13)G^A = \int_{\hat{y}_A}^{\bar{y}} \left(T(y)-T(\hat{y}_A)\right) \, \mathrm{d}H_Y(y). \tag*{(13)}

Regime B: truthful classification data. When classification data are truthful, the IRS can condition profit audits directly on dd. Let yB′(d)y'_B(d) denote the optimal class-specific cutoff. Equilibrium reporting is

xB(y,d)=min⁡{y,yB′(d)}.x_B(y,d) = \min\{y,y'_B(d)\}.

Therefore

GB=∫∫yB′(d)yˉ(T(y)−T(yB′(d))) dHY∣D(y∣d) dHD(d).(14)G^B = \int\int_{y'_B(d)}^{\bar{y}} \left(T(y)-T(y'_B(d))\right) \, \mathrm{d}H_{Y\mid D}(y\mid d) \, \mathrm{d}H_D(d). \tag*{(14)}

The cutoff schedule is chosen subject to

ρYπ∫HY∣D(yB′(d)∣d) dHD(d)=B.(15)\frac{\rho_Y}{\pi}\int H_{Y\mid D}(y'_B(d)\mid d)\,dH_D(d)=B. \tag*{(15)}

Regime C: strategic classification data with optimal verification. Under strategic reporting, let yC′(d)y'_C(d) denote the optimal profit-audit cutoff and let

y^C:=sup⁡dyC′(d)\widehat{y}_C:=\sup_d y'_C(d)

denote the associated global verification frontier. Proposition 1 and Proposition 3 imply the equilibrium reporting rule

xC(y,d)=min⁡{y,yC′(d)}.x_C(y,d)=\min\{y,y'_C(d)\}.

The resulting tax gap is therefore

GC=∫yC′(d)y‾(T(y)−T(yC′(d))) dHY∣D(y∣d) dHD(d).(16)G^C=\int_{y'_C(d)}^{\overline{y}}\left(T(y)-T(y'_C(d))\right)\,dH_{Y\mid D}(y\mid d)\,dH_D(d). \tag*{(16)}

The hybrid policy uses profit audits below yC′(d)y'_C(d) and data verification between yC′(d)y'_C(d) and y^C\widehat{y}_C. Its budget constraint is

1π∫[ρDHY∣D(y^C∣d)+(ρY−ρD)HY∣D(yC′(d)∣d)] dHD(d)=B.(17)\frac{1}{\pi}\int\left[\rho_DH_{Y\mid D}(\widehat{y}_C\mid d)+(\rho_Y-\rho_D)H_{Y\mid D}(y'_C(d)\mid d)\right]\,dH_D(d)=B. \tag*{(17)}

Equivalently, expenditure can be decomposed as

B=ρYπ∫HY∣D(yC′(d)∣d) dHD(d)⏟BY: profit audits+ρDπ∫[HY∣D(y^C∣d)−HY∣D(yC′(d)∣d)] dHD(d)⏟BD: data verification.(18)B=\underbrace{\frac{\rho_Y}{\pi}\int H_{Y\mid D}(y'_C(d)\mid d)\,dH_D(d)}_{B_Y:\ \text{profit audits}} +\underbrace{\frac{\rho_D}{\pi}\int\left[H_{Y\mid D}(\widehat{y}_C\mid d)-H_{Y\mid D}(y'_C(d)\mid d)\right]\,dH_D(d)}_{B_D:\ \text{data verification}}. \tag*{(18)}

Thus BDB_D is the direct resource expenditure on data verification. Its opportunity cost in terms of the IRS’s objective is instead measured by the difference between the tax gaps under strategic and truthful data,

CS:=GC−GB.C_S := G^C - G^B.

Regime D: strategic data without verification. Finally, suppose that the IRS conditions profit auditing on classification data but does not verify those data. Firms can then select into the most leniently enforced class. Let

y~:=inf⁡dy′(d)\tilde{y} := \inf_d y'(d)

denote its profit-audit cutoff. The resulting reporting rule is

xD(y)=min⁡{y,y~},x_D(y) = \min\{y,\tilde{y}\},

and the tax gap is

GD=∫y~yˉ(T(y)−T(y~)) dHY(y).(19)G^D = \int_{\tilde{y}}^{\bar{y}} (T(y)-T(\tilde{y}))\,dH_Y(y). \tag*{(19)}

Tax-gap rankings

We first establish the ordering among the optimally designed regimes.

Because Regime A is feasible when data are truthful—the IRS can simply ignore dd and apply the uniform cutoff y^A\hat{y}_A—Regime B optimizes over a weakly larger set of policies. Hence

GB≤GA.(20)G^B \le G^A. \tag*{(20)}

Similarly, Regime A is feasible under strategic data: the IRS can ignore classification reports altogether, in which case no data verification is required. Therefore

GC≤GA.(21)G^C \le G^A. \tag*{(21)}

Conversely, the truthful-data problem is a relaxation of the strategic-data problem. Any policy feasible under strategic reporting must satisfy the additional incentive constraints required to prevent firms from manipulating their classification, whereas these constraints are absent under truthful data. Removing them cannot increase the minimized tax gap. Therefore

GB≤GC.(22)G^{B} \le G^{C}. \tag*{(22)}

Combining the three comparisons gives

GB≤GC≤GA.(23)G^{B} \le G^{C} \le G^{A}. \tag*{(23)}

Thus

GA−GB≥0,GC−GB≥0,GA−GC≥0.G^{A} - G^{B} \ge0,\qquad G^{C} - G^{B} \ge0,\qquad G^{A} - G^{C} \ge0.

Targeting value and strategic implementation cost

The gross targeting value of classification data is the improvement that would be attainable if the classification were truthful:

VT:=GA−GB.(24)V_{T} := G^{A} - G^{B}. \tag*{(24)}

Strategic reporting imposes an additional implementation constraint. Its opportunity cost in terms of the tax gap is

CS:=GC−GB≥0.(25)C_{S} := G^{C} - G^{B} \ge0. \tag*{(25)}

It follows immediately that

GA−GC=(GA−GB)−(GC−GB)=VT−CS.(26)\begin{aligned} G^{A} - G^{C} &= \left(G^{A} - G^{B}\right) - \left(G^{C} - G^{B}\right) \\ &= V_{T} - C_{S}. \tag*{(26)} \end{aligned}

Notice that BDB_{D} in (18) and CSC_{S} in (25) are different objects. The former is the direct expenditure on data verification, while the latter is its equilibrium opportunity cost: the increase in the minimum attainable tax gap caused by having to make data-contingent enforcement incentive compatible.

The cost of introducing data

Suppose that introducing classification data into enforcement entails an additional fiscal cost K≥0K \ge0, measured in the same monetary units as the tax gap. This can capture data administration, information systems, or other fixed costs that are not included in the enforcement budget BB.

Relative to ignoring data, adoption generates the net gain

(GA−GC)−K.(G^{A} - G^{C}) - K.

Hence introducing strategic classification data is optimal if and only if

GA−GC≥K.G^{A} - G^{C} \ge K.

Using (26), this condition can equivalently be written as

GA−GB≥(GC−GB)+K.(27)G^{A} - G^{B} \ge(G^{C} - G^{B}) + K. \tag*{(27)}

Thus the gross targeting value must cover both the strategic implementation cost and the additional cost of introducing the data system.

Naive use of manipulable data

It remains to compare Regime D with the no-data benchmark. Define

F(z):=∫zyˉ(T(y)−T(z)) dHY(y).F(z) := \int_{z}^{\bar{y}} (T(y) - T(z))\,\mathrm{d}H_Y(y).

Then

F′(z)=−T′(z)(1−HY(z))<0,F'(z) = -T'(z)(1 - H_Y(z)) < 0,

so the tax gap is strictly decreasing in the effective profit-audit cutoff.

Under Regime A, the budget is exhausted at y^A\hat{y}_A:

ρYπHY(y^A)=B.\frac{\rho Y}{\pi}H_Y(\hat{y}_A)=B.

Under Regime D, all firms select into the lowest-cutoff class. Feasibility of the induced policy therefore requires

ρYπHY(y~)≤B.\frac{\rho Y}{\pi}H_Y(\tilde{y})\le B.

Since HYH_Y is increasing,

y~≤y^A.\tilde{y}\le\hat{y}_A.

Consequently,

GD=F(y~)≥F(y^A)=GA,(28)G^D=F(\tilde{y})\ge F(\hat{y}_A)=G^A, \tag*{(28)}

with strict inequality whenever y~<y^A\tilde{y}<\hat{y}_A. Together with GC≤GAG^C\le G^A, this gives

GD≥GA≥GC.G^D\ge G^A\ge G^C.

Proof of Proposition 4. Equations (20)–(22) establish

GB≤GC≤GA.G^B\le G^C\le G^A.

Hence the gross targeting value is GA−GBG^A-G^B, the strategic implementation cost is GC−GBG^C-G^B, and

GA−GC=(GA−GB)−(GC−GB).G^A-G^C=(G^A-G^B)-(G^C-G^B).

If introducing data entails an additional cost KK, adoption is optimal if and only if

GA−GC≥K,G^A-G^C\ge K,

which is equivalent to

GA−GB≥(GC−GB)+K.G^A-G^B\ge(G^C-G^B)+K.

Finally, (28) gives GD≥GAG^D \ge G^A whenever y~≤y^A\tilde{y} \le\hat{y}_A, while optimality of Regime C gives GA≥GCG^A \ge G^C. This establishes all claims.

Audit returns, shadow values, and the MVPF

This appendix relates the enforcement index in § 3 to empirical measures of audit returns and the marginal value of public funds (MVPF). The key distinction is between the fiscal return from auditing a marginal taxpayer and the marginal return from expanding enforcement. In our model, the latter also includes the information rents eliminated among taxpayers above the enforcement cutoff.

Consider an interior cutoff y′(d)y'(d). An increase in the class-specific audit budget expands the set of deterred firms at the cutoff, but also raises the reported profits of all firms above it. From (4),

−dG(d)dB(d)=πρYT′(y′(d))1−HY∣D(y′(d)∣d)hY∣D(y′(d)∣d)=πρYI(y′(d)∣d).-\frac{\mathrm{d}G(d)}{\mathrm{d}B(d)} = \frac{\pi}{\rho_Y}T'(y'(d))\frac{1-H_{Y\mid D}(y'(d)\mid d)}{h_{Y\mid D}(y'(d)\mid d)} = \frac{\pi}{\rho_Y}I(y'(d)\mid d).

The numerator 1−HY∣D(y′(d)∣d)1-H_{Y\mid D}(y'(d)\mid d) measures the mass of firms whose reporting rents are reduced when the cutoff rises, while hY∣D(y′(d)∣d)h_{Y\mid D}(y'(d)\mid d) measures the marginal mass that must be brought under audit coverage. The inverse hazard rate

γ(y∣d)=1−HY∣D(y∣d)hY∣D(y∣d)\gamma(y\mid d)=\frac{1-H_{Y\mid D}(y\mid d)}{h_{Y\mid D}(y\mid d)}

therefore converts the fiscal stakes at the cutoff into the return from expanding enforcement.

This distinction complements the empirical evidence in Boning et al. [6]. They estimate both average and marginal fiscal returns to IRS audits and document substantial heterogeneity across the income distribution. They also show that the fiscal consequences of an audit extend beyond the revenue collected in the audited return: individual deterrence generates additional tax payments for many years after the audit. Our model captures a different spillover from enforcement. Tightening enforcement at a cutoff changes the reporting incentives of firms that are not themselves marginally added to the audit pool. Consequently, observed returns to the marginal audited taxpayer need not by themselves identify the marginal value of expanding enforcement.

This distinction has an immediate allocation implication. Let κ\kappa denote the shadow value of the aggregate enforcement budget. For classes with interior cutoffs,

πρYI(y′(d)∣d)=κ,\frac{\pi}{\rho Y} I(y'(d) \mid d) = \kappa,

and hence, for any two such classes dd and d′d',

T′(y′(d))γ(y′(d)∣d)=T′(y′(d′))γ(y′(d′)∣d′).T'(y'(d))\gamma(y'(d) \mid d) = T'(y'(d'))\gamma(y'(d') \mid d').

Thus heterogeneity in fiscal returns across taxpayer groups need not by itself indicate misallocation. What must be equalized is the marginal value of enforcement resources, which also depends on the information rents generated by the conditional profit distribution.

Relation to the MVPF

The distinction from the MVPF is equally important. The framework of Hendren and Sprung-Keyser [17] evaluates a marginal policy change by comparing individuals’ willingness to pay with its net fiscal cost. Applying this framework to tax enforcement, Boning et al. [6] incorporate audit costs, revenues, deterrence, and taxpayer compliance burdens to evaluate the welfare consequences of marginal audit expansions.

Our enforcement index answers a different question. Taking the enforcement budget and the tax schedule as given, it identifies where marginal enforcement resources have the greatest effect on the tax gap. It is therefore an allocation statistic rather than a welfare statistic. A complete welfare analysis could in principle place the enforcement effects characterized here inside an MVPF calculation, but would additionally require taxpayer welfare, compliance costs, and the incidence of enforcement.

Empirical implementation

The allocation condition is empirically operational. The tax schedule identifies T′(y)T'(y), while administrative data on profits conditional on the classification variables identify

γ(y∣d)=1−HY∣D(y∣d)hY∣D(y∣d).\gamma(y \mid d) = \frac{1 - H_{Y\mid D}(y \mid d)}{h_{Y\mid D}(y \mid d)}.

Together they identify

I(y∣d)=T′(y)γ(y∣d).I(y \mid d) = T'(y)\gamma(y \mid d).

At observed interior audit cutoffs, optimal allocation requires the index to be equalized across data classes. Systematic differences in I(y′(d)∣d)I(y'(d) \mid d) therefore provide a diagnostic for the allocation of profit-audit resources.

This also clarifies what additional information is needed beyond estimates of audit returns. Evidence such as Boning et al. [6] disciplines the fiscal consequences of marginal audit expansions; evaluating targeting across data classes additionally requires the conditional profit distributions that determine the information rents affected by enforcement.

Comparative statics

This section records several comparative statics implied by the optimal audit policy. The benchmark results follow directly from the cutoff representation. Under strategic data reporting, audit costs additionally affect the allocation between profit audits and data verification and, through this channel, the global verification frontier.

Audit resources and the profit-audit cutoff

Fix a data class dd. In the truthful-data benchmark, the within-class budget constraint is

B(d)=ρYπHY∣D(y′(d)∣d).B(d) = \frac{\rho_Y}{\pi} H_{Y\mid D}(y'(d) \mid d).

Differentiating with respect to B(d)B(d) gives

dy′(d)dB(d)=πρYhY∣D(y′(d)∣d)>0.(29)\frac{dy'(d)}{dB(d)}=\frac{\pi}{\rho_Y h_{Y|D}(y'(d)\mid d)}>0. \tag*{(29)}

A larger class-specific budget therefore extends profit-audit coverage further into the profit distribution. The response is larger where the conditional density at the cutoff is low.

Combining (29) with

dG(d)dy′=−T′(y′)(1−HY∣D(y′∣d))\frac{dG(d)}{dy'}=-T'(y')(1-H_{Y|D}(y'\mid d))

yields

dG(d)dB(d)=−πρYT′(y′(d))γ(y′(d)∣d)=−πρYI(y′(d)∣d)<0.(30)\frac{dG(d)}{dB(d)}=-\frac{\pi}{\rho_Y}T'(y'(d))\gamma(y'(d)\mid d)=-\frac{\pi}{\rho_Y}I(y'(d)\mid d)<0. \tag*{(30)}

Thus, the enforcement index determines the marginal reduction in the tax gap generated by additional audit resources. Classes with a higher index have a larger marginal return to profit-audit spending.

Profit-audit costs

Holding the class-specific budget B(d)B(d) fixed, differentiation of the budget constraint with respect to ρY\rho_Y gives

dy′(d)dρY=−HY∣D(y′(d)∣d)ρYhY∣D(y′(d)∣d)<0.(31)\frac{dy'(d)}{d\rho_Y}=-\frac{H_{Y|D}(y'(d)\mid d)}{\rho_Y h_{Y|D}(y'(d)\mid d)}<0. \tag*{(31)}

Higher profit-audit costs therefore reduce audit coverage. The magnitude of the response depends on the mass already covered relative to the density at the cutoff.

Under strategic data reporting, the marginal cost of increasing a class-specific profit-audit cutoff depends on whether the class is subject to data verification. In a verified class, raising the cutoff replaces data verification with profit auditing, so its incremental resource cost is ρY−ρD\rho_Y-\rho_D. In an unverified class, the marginal cost is ρY\rho_Y. For a verified class with an interior cutoff, the optimality condition

I(y′(d)∣d)=λρY−ρDπI(y'(d)\mid d)=\lambda\frac{\rho_Y-\rho_D}{\pi}

shows that, holding the multiplier λ\lambda and verification status fixed, a larger ρY−ρD\rho_Y-\rho_D lowers the cutoff when I(y∣d)I(y\mid d) is decreasing. This is a conditional response: the optimal multiplier and the set of verified classes may also change.

Data-verification costs and the global frontier

Strategic reporting introduces an additional cost margin because all reports above the profit-audit cutoff in a verified class are subject to data verification. Write

q(d)=1{y′(d)<y^}q(d)=\mathbf{1}_{\{y'(d)<\hat{y}\}}

for the indicator that class dd is verified. Total enforcement expenditure is

B=1π∫[ρYHY∣D(y′(d)∣d)+ρD(1−HY∣D(y′(d)∣d))q(d)] dHD(d).(32)B=\frac{1}{\pi}\int\left[\rho_Y H_{Y\mid D}(y'(d)\mid d)+\rho_D\left(1-H_{Y\mid D}(y'(d)\mid d)\right)q(d)\right]\,dH_D(d). \tag*{(32)}

Thus, unlike a policy that verifies only reports between y′(d)y'(d) and y^\hat{y}, the verification expenditure depends on the entire upper tail of reports in each verified class.

Holding the cutoff schedule and verification status fixed, the direct resource effect of a higher verification cost is

∂B∂ρD=1π∫[1−HY∣D(y′(d)∣d)]q(d) dHD(d)≥0,\frac{\partial B}{\partial\rho_D}=\frac{1}{\pi}\int\left[1-H_{Y\mid D}(y'(d)\mid d)\right]q(d)\,dH_D(d)\geq0,

with strict inequality whenever verified reports have positive mass. Likewise, the marginal expenditure from raising a cutoff within a class, without changing its verification status, is

∂B∂y′(d)=hY∣D(y′(d)∣d)π(ρY−ρDq(d)),\frac{\partial B}{\partial y'(d)}=\frac{h_{Y\mid D}(y'(d)\mid d)}{\pi}\left(\rho_Y-\rho_Dq(d)\right),

where the expression is understood as the class-specific integrand of the aggregate budget (and is weighted by the class probability for a discrete distribution of classes).

Analytical examples

The global frontier y^\hat{y} determines which classes require verification, but it is not an upper limit on the reports verified in those classes. Consequently, holding the cutoff schedule and the set of verified classes fixed, expenditure has no direct dependence on y^\hat{y}. When the frontier crosses a class-specific cutoff, verification status changes and the budget can change discontinuously. It would therefore be incorrect to obtain the response of the optimal frontier to ρD\rho_D by differentiating the budget constraint with respect to y^\hat{y} while holding the cutoffs fixed.

In the full optimization problem, a change in verification costs may alter the cutoffs, the verification status of classes, and the optimal global frontier. The resulting comparative statics are therefore conditional on the prevailing enforcement regime; in particular, the optimal policy may switch between differentiated enforcement, partial pooling, and ignoring classification data.

Remark 3 (Comparative statics and the enforcement index) The comparative statics emphasize two distinct roles of the profit distribution. First, the density at the cutoff determines how far a given change in audit resources moves the enforcement boundary. Second, the enforcement index

I(y∣d)=T′(y)γ(y∣d)I(y \mid d) = T'(y)\gamma(y \mid d)

determines the resulting marginal reduction in the tax gap.

Under strategic reporting, profit auditing in a verified class replaces data verification, giving it incremental resource cost ρY−ρD\rho_Y-\rho_D. Verification itself applies to the entire upper tail above the class-specific cutoff. Its cost therefore affects not only the allocation between enforcement instruments but also which classes are verified and how far their profit-audit cutoffs can be differentiated.

Analytical examples

This appendix provides analytical examples of the enforcement index and the intensive- and extensive-margin enforcement regimes characterized in the paper. The exercise is not intended as a structural calibration. Rather, it indicates how empirically meaningful differences in profit distributions and tax schedules map into optimal enforcement.

We consider three data classes, denoted by dS,dM,dLd_S,d_M,d_L, corresponding loosely to small, medium, and large firms. Representative profit levels are normalized to

yS=1,yM=10,yL=50.y_S = 1,\qquad y_M = 10,\qquad y_L = 50.

Throughout, the enforcement index is

I(y∣d)=T′(y)γ(y∣d),γ(y∣d)=1−HY∣D(y∣d)hY∣D(y∣d).I(y\mid d) = T'(y)\gamma(y\mid d),\qquad\gamma(y\mid d) = \frac{1-H_{Y\mid D}(y\mid d)}{h_{Y\mid D}(y\mid d)}.

The exercise emphasizes that the shape of the conditional profit distribution determines whether enforcement operates primarily on the intensive or extensive margin.

Intensive-margin illustration: Weibull profits

Suppose profits conditional on data class dd follow a Weibull distribution with shape parameter k>1k > 1 and scale parameter λ(d)>0\lambda(d) > 0:

HY∣D(y∣d)=1−exp⁡[−(yλ(d))k],y≥0.H_{Y\mid D}(y\mid d) = 1-\exp\left[-\left(\frac{y}{\lambda(d)}\right)^k\right],\qquad y\ge0.

The inverse hazard rate is

γ(y∣d)=λ(d)kkyk−1,\gamma(y\mid d) = \frac{\lambda(d)^k}{k y^{k-1}},

which is strictly decreasing in yy. Under a linear tax schedule, T′(y)=τT'(y)=\tau, the enforcement index is therefore

I(y∣d)=τλ(d)kkyk−1,(33)I(y\mid d) = \tau\frac{\lambda(d)^k}{k y^{k-1}}, \tag*{(33)}

and is decreasing within every data class. This is the intensive-margin case characterized in Proposition 1.

For example, set

k=2,τ=0.25,k = 2,\qquad\tau= 0.25,

and choose the scale parameter in each class so that its mean equals the representative profit ydy_d. Since the mean of a Weibull random variable is λ(d)Γ(1+1/k)\lambda(d)\Gamma(1 + 1/k), this gives

λ(d)=ydΓ(1+1/k).\lambda(d) = \frac{y_d}{\Gamma(1 + 1/k)}.

Evaluating the enforcement index at a common relative position in each class illustrates how differences in the conditional profit distribution translate into differences in marginal enforcement returns.

More generally, Proposition 1 implies that interior profit-audit cutoffs satisfy

I(y′(d)∣d)=σ.I(y'(d) \mid d) = \sigma.

For the Weibull specification with k=2k = 2, this gives the explicit cutoff

y′(d)=τλ(d)22σ.(34)y'(d) = \frac{\tau\lambda(d)^2}{2\sigma}. \tag*{(34)}

Thus, classes with larger scale parameters receive higher profit-audit cutoffs at a given shadow value σ\sigma. The numerical exercise therefore shows the intensive-margin targeting mechanism directly: information about the conditional profit distribution generates differentiated audit coverage across data classes.

Extensive-margin illustration: Pareto profits

As a contrasting case, suppose profits conditional on class dd follow a Pareto distribution with shape parameter α>1\alpha> 1 and scale parameter θ(d)>0\theta(d) > 0:

HY∣D(y∣d)=1−(θ(d)y)α,y≥θ(d).H_{Y\mid D}(y \mid d) = 1 - \left(\frac{\theta(d)}{y}\right)^\alpha,\qquad y \ge\theta(d).

Its inverse hazard rate is

γ(y∣d)=yα\gamma(y \mid d) = \frac{y}{\alpha}

Under a linear tax schedule,

I(y∣d)=τyα,I(y \mid d) = \tau\frac{y}{\alpha},

which is increasing in profits. The Pareto specification therefore provides a simple illustration of the extensive-margin case in Proposition 3.

Conditional on a global verification frontier y^\hat{y}, the authority does not optimally choose an interior profit-audit cutoff in the extensive-margin case. Instead,

y′(d)∈{y,y^},y'(d) \in\{y,\hat{y}\},

subject to the support of the conditional profit distribution. A class with y′(d)=y^y'(d) = \hat{y} receives profit audits below the frontier and does not require data verification. A class assigned the lower cutoff is instead sustained by verifying all reports above that cutoff, including reports exceeding y^\hat{y}.

The selection of classes for profit auditing consequently reflects both their enforcement returns and the full resource cost of verification. For a candidate class with lower cutoff a(d)a(d), replacing verification above a(d)a(d) by profit auditing up to y^\hat{y} changes enforcement expenditure by

1π[ρY(HY∣D(y^∣d)−HY∣D(a(d)∣d))−ρD(1−HY∣D(a(d)∣d))],\frac{1}{\pi}\left[\rho_Y\left(H_{Y\mid D}(\hat{y}\mid d)-H_{Y\mid D}(a(d)\mid d)\right)-\rho_D\left(1-H_{Y\mid D}(a(d)\mid d)\right)\right],

provided the alternative with profit auditing up to y^\hat{y} does not require verification. The second term includes the entire upper tail, not merely reports below the frontier. Thus, although the increasing enforcement index still favors concentrating profit audits on classes with high enforcement returns, class selection must account for the full verification expenditure.

The Pareto example therefore showcases a qualitatively different use of classification data. Rather than determining how far profit auditing extends within every class, data help determine which classes receive profit audits and which are sustained through data verification.

Tax progressivity

The tax schedule provides a second source of heterogeneity in enforcement returns. Since

I(y∣d)=T′(y)γ(y∣d),I(y \mid d) = T'(y)\gamma(y \mid d),

higher marginal tax rates increase the enforcement value of the corresponding profit levels.

For instance, consider the stylized three-bracket schedule

T′(y)={0.15,y<yM,0.25,yM≤y<yL,0.30,y≥yL.T'(y) = \begin{cases} 0.15, & y < y_M, \\ 0.25, & y_M \le y < y_L, \\ 0.30, & y \ge y_L. \end{cases}

Relative to a linear tax rate of 25 percent, this schedule lowers the enforcement index at low profits and raises it at high profits:

Iprog(y∣d)Ilinear(y∣d)=Tprog′(y)0.25.\frac{I_{\mathrm{prog}}(y \mid d)}{I_{\mathrm{linear}}(y \mid d)} = \frac{T'_{\mathrm{prog}}(y)}{0.25}.

Hence

Iprog(yS∣d)Ilinear(yS∣d)=0.6,Iprog(yL∣d)Ilinear(yL∣d)=1.2.\frac{I_{\mathrm{prog}}(y_S \mid d)}{I_{\mathrm{linear}}(y_S \mid d)} = 0.6, \qquad \frac{I_{\mathrm{prog}}(y_L \mid d)}{I_{\mathrm{linear}}(y_L \mid d)} = 1.2.

Progressivity therefore rotates enforcement incentives toward high-profit firms. In the intensive-margin case this changes the relative audit cutoffs across classes; in the extensive-margin case it changes the ranking of classes according to their average enforcement gains.

Quantitative illustration

This section provides a quantitative illustration of the model. The exercise corroborates the main theoretical results and assesses their quantitative importance. In particular, we quantify how much classification data can improve the targeting of profit audits, how much of this gain is lost when the data are strategically reported, and how the resulting value of data depends on their informativeness and the cost of verification.

It should be noted that the exercise is not intended as a structural estimation of a particular tax administration. We instead discipline the main parameters using estimates from the empirical tax literature and examine the robustness of the results over plausible parameter ranges. This allows us to put magnitudes on the mechanisms characterized in the paper while retaining the transparency of the theoretical model.

Quantitative environment

We take the intensive-margin environment as our baseline setting. Conditional on data class dd, profits follow a Weibull distribution,

HY∣D(y∣d)=1−exp⁡[−(yλ(d))k],y≥0,H_{Y\mid D}(y \mid d) = 1 - \exp\left[-\left(\frac{y}{\lambda(d)}\right)^k\right], \qquad y \ge0,

with k>1k > 1. Under a linear tax schedule T(y)=τyT(y) = \tau y, the enforcement index is

I(y∣d)=τλ(d)kkyk−1,I(y \mid d) = \tau\frac{\lambda(d)^k}{k y^{k-1}},

which is strictly decreasing in profits. Hence optimal enforcement follows the intensive-margin characterization of Proposition 1.

We consider three equally likely data classes,

d∈{0,1,2},Pr⁡(D=d)=13,d \in\{0,1,2\}, \qquad\Pr(D=d) = \frac{1}{3},

and parameterize the scale of the conditional profit distribution as

λ(d;β)=λ0exp⁡[β(d−1)]13(e−β+1+eβ)\lambda(d;\beta)=\lambda_0\frac{\exp[\beta(d-1)]}{\frac{1}{3}(e^{-\beta}+1+e^\beta)}

The parameter β≥0\beta\ge0 controls heterogeneity in the conditional profit distribution across data classes. The normalization keeps average profits fixed as β\beta varies. When β=0\beta= 0, profits have the same distribution in every class and classification data have no targeting value. As β\beta increases, the conditional profit distributions become more differentiated, increasing the potential value of conditioning enforcement on dd.

We set the linear corporate tax rate to τ=0.25\tau= 0.25, a representative statutory corporate tax rate in advanced economies (OECD, 2023). We normalize the cost of a profit audit to ρY=1\rho_Y = 1 and parameterize the relative cost of data verification by

rD:=ρDρY.r_D := \frac{\rho_D}{\rho_Y}.

Because the two enforcement technologies are difficult to map directly into existing administrative cost estimates, we treat rDr_D as a sensitivity parameter. Our baseline uses a relatively inexpensive data-verification technology and we report results over a broad range of values.

Empirical evidence suggests substantial heterogeneity in both enforcement returns and responses to classification thresholds. Boning et al. [6] for example, estimate total revenue returns to audit spending that differ by more than a factor of two across taxpayer groups, while Almunia and Lopez-Rodriguez [2] document economically significant reporting responses around the Spanish Large Taxpayer Unit threshold. We use this evidence to discipline the range of heterogeneity across data classes rather than to estimate β\beta directly.

Our baseline setting uses k=2k = 2, as in the analytical Weibull illustration above. We normalize the aggregate enforcement budget to B=0.05B = 0.05. Together with π=2\pi= 2 and ρY=1\rho_Y = 1, this generates a nondegenerate allocation of audit coverage across classes. The baseline is chosen to generate meaningful cross-class heterogeneity while remaining in the interior hybrid region in which both profit auditing and data verification are used. None of the main conclusions relies on this particular choice, as shown by the sensitivity exercises below.

In summary, our baseline sets

k=2,λ0=1,τ=0.25,π=2,ρY=1,rD=0.03,B=0.05,β=0.35.k = 2,\quad\lambda_0 = 1,\quad\tau= 0.25,\quad\pi= 2,\quad\rho_Y = 1,\quad r_D = 0.03,\quad B = 0.05,\quad\beta= 0.35.

The scale normalizations λ0=ρY=1\lambda_0 = \rho_Y = 1 are without loss for the comparative exercise. The value τ=0.25\tau= 0.25 is representative of statutory corporate tax rates, while π=2\pi= 2 follows the baseline used in related numerical tax-enforcement exercises. The baseline values of rDr_D and BB generate an interior hybrid policy with both profit auditing and data verification.

ParameterSymbolBaselineRangeDiscipline
Weibull shapekk21.5–3analytical benchmark
Mean-profit scaleλ0\lambda_01—normalization
Class heterogeneityβ\beta0.350–0.70empirical literature
Tax rateτ\tau0.25—OECD (2023) median
Penalty parameterπ\pi21.5–3numerical benchmark
Profit-audit costρY\rho_Y1—normalization
Relative verification costrDr_D0.030–0.20sensitivity parameter
Enforcement budgetBB0.050.03–0.08interior hybrid allocation

Table 2. Baseline parameters

Enforcement regimes and value of data

For each parameterization, we solve the four enforcement regimes introduced in § 5. Regime AA ignores classification data and chooses a common profit-audit cutoff. Regime BB conditions profit audits on truthful classification data. Regime CC allows classification data to be strategically reported and optimally combines profit audits with data verification. Regime DD conditions profit auditing on classification data without verifying those data.

Let the corresponding optimized tax gaps be

GA,GB,GC,GD.G^{A}, \quad G^{B}, \quad G^{C}, \quad G^{D}.

As established in § 5,

GB≤GC≤GA≤GD.G^{B} \le G^{C} \le G^{A} \le G^{D}.

The quantitative exercise therefore focuses not on this ranking itself, but on the magnitudes of the wedges between the regimes.

We decompose the value of classification data as

VT:=GA−GB,CS:=GC−GB,VN:=GA−GC=VT−CS.V_{T} := G^{A} - G^{B}, \qquad C_{S} := G^{C} - G^{B}, \qquad V_{N} := G^{A} - G^{C} = V_{T} - C_{S}.

Here VTV_{T} is the gross targeting value of classification data, CSC_{S} is the implementation cost generated by strategic reporting, and VNV_{N} is the net enforcement value of strategically reported data.

A useful summary statistic is

RV:=VNVT=GA−GCGA−GB,R_{V} := \frac{V_{N}}{V_{T}} = \frac{G^{A} - G^{C}}{G^{A} - G^{B}},

whenever VT>0V_{T} > 0. The statistic RVR_{V} measures the share of the potential targeting value of classification data that survives once strategic manipulation and the associated verification requirements are taken into account.

Baseline results

Table 3 reports the optimized tax gaps and enforcement allocations under the four regimes. The truthful-data benchmark isolates the targeting value created by conditioning profit audits on the classification variable. Moving from Regime BB to Regime CC then measures the amount of this targeting gain that must be sacrificed to make the classification incentive compatible.

Regime AARegime BBRegime CCRegime DD
Tax gap0.1510.1400.1480.186
Profit-audit expenditure0.0500.0500.0450.013
Data-verification expenditure000.0050
Profit-audit cutoff, d=0d = 00.2900.1440.1460.144
Profit-audit cutoff, d=1d = 10.2900.2900.3860.144
Profit-audit cutoff, d=2d = 20.2900.5850.3860.144
Global frontier0.290—0.386—

Table 3. Quantitative illustration: baseline results

The baseline results illustrate all three forces emphasized by the theory. First, truthful classification data improve the targeting of profit audits. Relative to ignoring data, the tax gap falls from GA=0.151G^{A} = 0.151 to GB=0.140G^{B} = 0.140. The corresponding gross targeting value of data is

VT=GA−GB=0.011,V_{T} = G^{A} - G^{B} = 0.011,

or 7.3 percent of the no-data tax gap.

Second, strategic data reporting limits but does not eliminate this targeting value. Under optimal verification, the tax gap is GC=0.148G^{C} = 0.148, implying an implementation cost of approximately CS=0.0075C_{S} = 0.0075. The remaining net value is approximately VN=0.0037V_{N} = 0.0037, or 32.9 percent of the gross targeting value. Thus strategic reporting dissipates about 67.1 percent of the potential benefit of classification data.

The enforcement allocation reveals how this loss arises. With truthful data, the entire enforcement budget is devoted to profit audits, with cutoffs

(yB′(0),yB′(1),yB′(2))=(0.144,0.290,0.585).(y'_{B}(0), y'_{B}(1), y'_{B}(2)) = (0.144, 0.290, 0.585).

With strategic reporting, approximately BY=0.0452B_{Y} = 0.0452 is devoted to profit audits and BD=0.0048B_{D} = 0.0048 to data verification. The optimal cutoffs become

(yC′(0),yC′(1),yC′(2))=(0.146,0.386,0.386).(y'_{C}(0), y'_{C}(1), y'_{C}(2)) = (0.146, 0.386, 0.386).

The two upper classes therefore share a profit-audit cutoff, while the lowest class is treated differently. Strategic reporting induces partial pooling rather than a uniform reduction in audit coverage across classes. Verification absorbs about 9.5 percent of the enforcement budget, while the loss in targeting value is substantially larger. This illustrates how verification expenditure and the incentive constraints associated with strategic reporting jointly restrict differentiated enforcement.

Finally, using manipulable classification data without verification performs substantially worse than ignoring the data. Under Regime DD, firms select into the most leniently enforced class, yielding GD=0.186G^{D}=0.186. The resulting tax gap is about 23 percent larger than under Regime AA. Hence, the baseline describes both sides of the value of data: classification can materially improve enforcement when it is used together with appropriate verification, but naive reliance on manipulable data can make enforcement less effective.

The informativeness of classification data

We next vary β\beta, which determines how strongly the conditional profit distribution varies across data classes. This exercise isolates the central tradeoff in the value of classification data.

When β=0\beta=0, the conditional profit distributions coincide. The IRS then has no reason to differentiate profit-audit coverage across data classes, so classification data have no targeting value:

VT=VN=CS=0.V_{T}=V_{N}=C_{S}=0.

As β\beta increases, classification becomes more useful for targeting profit audits. At the same time, the resulting differentiation in enforcement creates stronger incentives to manipulate classification reports and therefore increases the need for data verification.

Figure 6 plots

Line graph showing gross targeting value $V_T/G^A$, implementation cost $C_S/G^A$, and net value $V_N/G^A$ as functions of informativeness $\beta$

Figure 6. The value of data and their informativeness

VTGA,CSGA,VNGA\frac{V_{T}}{G^{A}},\qquad\frac{C_{S}}{G^{A}},\qquad\frac{V_{N}}{G^{A}}

as functions of β\beta. Normalizing by the no-data tax gap makes the magnitudes comparable across parameterizations.

Greater heterogeneity increases the potential value of targeting profit audits, but strategic reporting can eliminate that value at low levels of informativeness. At β=0.20\beta= 0.20, the gross targeting value is approximately 2.6 percent of the no-data tax gap, yet optimal strategic enforcement ignores data. At the baseline β=0.35\beta= 0.35, the gross targeting value is 7.3 percent, of which 4.9 percentage points are dissipated by strategic implementation, leaving a net gain of 2.4 percent. At β=0.70\beta= 0.70, the corresponding gross, implementation, and net values are 21.6, 5.2, and 16.5 percent.

Near β=0\beta= 0, however, the value of differentiated enforcement emerges more slowly than its verification requirement. To see this, write z=d−1∈{−1,0,1}z = d - 1 \in\{-1, 0, 1\}. The normalized scale parameter satisfies

λ(d;β)=λ0[1+zβ+O(β2)].\lambda(d;\beta) = \lambda_0 \left[1 + z\beta+ O(\beta^2)\right].

Under truthful data, an interior cutoff satisfies

I(yB′(d)∣d)=κ.I(y_B'(d) \mid d) = \kappa.

For Weibull profits this gives

yB′(d)=(τλ(d)kkκ)1/(k−1).y'_B(d) = \left(\frac{\tau\lambda(d)^k}{k\kappa}\right)^{1/(k-1)}.

Because the first-order changes in λ(d)\lambda(d) average to zero across the three symmetric classes, the budget condition implies κ′(0)=0\kappa'(0) = 0. Therefore,

yB′(d)=y0′[1+kk−1(d−1)β+O(β2)].y'_B(d) = y'_0\left[1 + \frac{k}{k-1}(d-1)\beta+ O(\beta^2)\right].

The dispersion in class-specific cutoffs is consequently first order in β\beta. For a differentiated cutoff allocation with a first-order spread, verification between local cutoffs and the global frontier entails expenditure of order

BD=O(β).B_D = O(\beta).

By contrast, when β=0\beta= 0 the conditional profit distributions coincide and the uniform no-data allocation is already optimal. The envelope theorem therefore implies that a symmetric perturbation generates no first-order targeting gain:

VT=GA−GB=O(β2).V_T = G^A - G^B = O(\beta^2).

Hence, for sufficiently small β\beta, the second-order targeting gain cannot justify the first-order verification requirement of differentiated enforcement, and the IRS optimally ignores classification data. In the baseline parameterization, differentiated strategic-data enforcement becomes optimal at approximately

β∗≈0.24.\beta^* \approx0.24.

Once this threshold is crossed, the targeting value can exceed the cost of implementing differentiated enforcement. The retained share

RV=VNVTR_V = \frac{V_N}{V_T}

is zero at β=0.20\beta= 0.20, approximately 0.33 in the baseline, and about 0.76 at β=0.70\beta= 0.70. Thus greater informativeness can eventually make differentiated enforcement worthwhile despite the verification requirement.

The cost of data verification

We next vary the relative cost of data verification,

rD=ρDρY.r_D = \frac{\rho_D}{\rho_Y}.

The truthful-data benchmark GBG^B and the no-data benchmark GAG^A are independent of ρD\rho_D. Hence the gross targeting value

VT=GA−GBV_T = G^A - G^B

is constant in rDr_D, and changes in the value of strategic data operate entirely through their implementation cost.

Figure 7 plots the retained share

RV=GA−GCGA−GBR_V = \frac{G^A - G^C}{G^A - G^B}

as a function of rDr_D.

Robustness

The value of data and the cost of verification

Figure 7. The value of data and the cost of verification

When data verification is costless, strategic classification can be implemented without sacrificing profit-audit resources, so

GC=GBandRV=1.G^{C} = G^{B} \qquad\text{and} \qquad R_{V} = 1.

As verification becomes more costly, the implementation cost rises and the net value of classification data falls. At rD=0.02r_{D} = 0.02, approximately 53.5 percent of the gross targeting value is retained. At the baseline rD=0.03r_{D} = 0.03, this share is 32.9 percent, while at rD=0.05r_{D} = 0.05 it is about 9.2 percent.

The retained share reaches zero as the cost of verification rises beyond approximately rD=0.057r_{D} = 0.057 in the numerical grid. The IRS then optimally disregards classification data, yielding GC=GAG^{C} = G^{A} and RV=0R_{V} = 0. This illustrates the adoption criterion in Corollary 1: classification data are worth using only when the targeting gains exceed the cost of sustaining truthful classification.

The gross targeting value remains 7.3 percent of the no-data tax gap throughout this exercise. The implementation cost rises from 3.4 percent of that gap at rD=0.02r_{D} = 0.02 to 4.9 percent at the baseline and 6.7 percent at rD=0.05r_{D} = 0.05. Correspondingly, the net value falls from 3.9 to 2.4 and then to 0.7 percent of the no-data tax gap.

Robustness

Table 4 varies the Weibull shape parameter, the enforcement budget, and the penalty parameter around the baseline. The main quantitative conclusions are robust across all specifications.

SpecificationVT/GAV_T/G^ACS/GAC_S/G^AVN/GAV_N/G^ARVR_V
Baseline0.0730.0490.0240.329
k=1.5k = 1.50.0490.0370.0110.229
k=3k = 30.1410.0620.0790.562
B=0.03B = 0.030.0530.0480.0050.090
B=0.08B = 0.080.1010.0430.0580.572
π=1.5\pi= 1.50.0610.0480.0130.210
π=3\pi= 30.0960.0440.0520.540

Table 4. Robustness of the quantitative illustration

Across these robustness exercises, the retained share RVR_V ranges from approximately 0.09 to 0.57, compared with 0.33 in the baseline. Strategic reporting therefore absorbs a substantial fraction of the gross targeting value, although the extent varies considerably across specifications.

Changes in the Weibull shape parameter have a large quantitative effect. At k=3k = 3, the gross targeting value is 14.1 percent of the no-data tax gap, of which approximately 56 percent survives strategic reporting. At k=1.5k = 1.5, the gross targeting value is 4.9 percent and the retained share is about 23 percent.

The value of classification data also rises with the enforcement budget and with the penalty parameter in these exercises. Increasing the budget from B=0.03B = 0.03 to B=0.08B = 0.08 raises the net value from 0.5 to 5.8 percent of the no-data tax gap, and the retained share from 0.09 to 0.57. Increasing π\pi from 1.5 to 3 raises the net value from 1.3 to 5.2 percent and the retained share from 0.21 to 0.54. In all reported specifications, optimally verified strategic-data enforcement remains strictly preferable to ignoring classification data.

The results indicate that the baseline magnitudes are sensitive to the parameters governing enforcement returns and resources, while the qualitative distinction between gross targeting gains and the cost of strategic implementation persists.

As an additional distributional robustness check, we replace the Weibull distribution by a lognormal conditional profit distribution, preserving the same class-specific mean structure and baseline enforcement parameters. Across log standard deviations between 0.35 and 1, the retained share RVR_V ranges from approximately 0.04 to 0.62. The value of strategic data therefore remains positive in these specifications, although the retained share is highly sensitive to the distributional parameters.

The baseline uses three equally likely data classes for transparency. As a robustness check, we allow the class shares to be asymmetric. Holding the high-return class at p2=0.10p_2 = 0.10, we vary the relative masses of the remaining two classes. The value of strategically reported classification data is non-monotone in the class composition. Over an intermediate range of class shares, differentiated enforcement remains worthwhile, but when the classification becomes sufficiently unbalanced, the verification requirement exhausts the targeting gain and the authority optimally disregards the data.

Extensive-margin illustration

(p0,p1,p2)(p_0,p_1,p_2)GAG^AGBG^BGCG^CGDG^DRVR_VBD/BB_D/B
Equal shares(0.33,0.33,0.33)(0.33, 0.33, 0.33)0.1510.1400.1480.18632.9%9.5%
Weak asymmetry(0.40,0.50,0.10)(0.40, 0.50, 0.10)0.1320.1250.1310.1569.5%11.3%
Moderate asymmetry(0.60,0.30,0.10)(0.60, 0.30, 0.10)0.1230.1170.1230.1410.0%0.0%
Strong asymmetry(0.70,0.20,0.10)(0.70, 0.20, 0.10)0.1190.1130.1190.1330.0%0.0%

Table 5. Robustness to asymmetric class shares

For example, with

(p0,p1,p2)=(0.7,0.2,0.1),(p_0,p_1,p_2) = (0.7, 0.2, 0.1),

truthful classification reduces the tax gap from GA=0.119G^A = 0.119 to GB=0.113G^B = 0.113. Under strategic reporting, however, optimal enforcement reverts to the no-data policy, so that

GC=GA=0.119.G^C = G^A = 0.119.

Hence, in this quantitative illustration with a discrete number of classes, the targeting value of data can be entirely dissipated by the cost of sustaining truthful classification.

Extensive-margin illustration

The baseline quantitative exercise focuses on the decreasing-index case. To study the extensive-margin result, we retain the same three data classes, profit distributions, enforcement costs, and budget, but replace the linear tax schedule by the progressive schedule

T(y)=τy3.T(y) = \tau y^3.

With k=2k = 2, the enforcement index becomes

I(y∣d)=3τ2λ(d)2y,I(y \mid d) = \frac{3\tau}{2}\lambda(d)^2 y,

which is strictly increasing in profits. Proposition 3 therefore applies.

Under the baseline parameter values, the optimal policy profit-audits only class d=2d = 2, up to the global frontier

y^=0.714,\hat{y} = 0.714,

while classes d=0d = 0 and d=1d = 1 are sustained through data verification. At this frontier, the average enforcement indices are

Iˉ(0;y^)=0.073,Iˉ(1;y^)=0.155,Iˉ(2;y^)=0.322.\bar{I}(0; \hat{y}) = 0.073,\qquad\bar{I}(1; \hat{y}) = 0.155,\qquad\bar{I}(2; \hat{y}) = 0.322.

Thus the class with the highest average enforcement return is selected for profit auditing, as characterized in Proposition 3.

The resulting tax gap is 0.387. By comparison, profit-auditing classes d=1,2d = 1, 2 yields a tax gap of 0.401, while profit-auditing all three classes yields 0.407. The numerical exercise therefore illustrates the extensive-margin logic: when enforcement returns increase in profits, concentrating profit audits on the highest-return classes can dominate spreading audit coverage across all classes, with data verification sustaining the resulting cross-class differentiation. This mechanism is qualitatively consistent with the substantial cross-group heterogeneity in fiscal audit returns documented by [6], although their empirical return measure is not identical to the enforcement index used here.

References

  1. [1]Aghion, Philippe, Ufuk Akcigit, Maxime Gravoueille, Matthieu Lequien, and Stefanie Stantcheva, “Tax simplicity or simplicity of evasion? Evidence from self-employment taxes in France,” Technical Report, Centre for Economic Performance, LSE 2023.
  2. [2]Almunia, Miguel and David Lopez-Rodriguez, “Under the radar: The effects of monitoring firms on tax compliance,” American Economic Journal: Economic Policy, 2018, 10 (1), 1–38.
  3. [5]Best, Michael Carlos, Anne Brockmeyer, Henrik Jacobsen Kleven, Johannes Spinnewijn, and Mazhar Waseem, “Production versus revenue efficiency with limited tax capacity: theory and evidence from Pakistan,” Journal of Political Economy, 2015, 123 (6), 1311–1355.DOI
  4. [6]Boning, William C, Nathaniel Hendren, Ben Sprung-Keyser, and Ellen Stuart, “A welfare analysis of tax audits across the income distribution,” Quarterly Journal of Economics, 2025, 140 (1), 63–112.
  5. [7]Boserup, Simon Halphen and Jori Veng Pinje, “Tax Evasion, Information Reporting, and the Regressive Bias Prediction,” Working Paper 2013.
  6. [9]Castillo, Sebastián, “Tax policy design in a hierarchical model with occupational decisions: S. Castillo,” International Tax and Public Finance, 2024, 31 (5), 1295–1341.DOI
  7. [10]Chen, Zhao, Zhikuo Liu, Juan Carlos Suárez Serrato, and Daniel Yi Xu, “Notching R&D investment with corporate income tax cuts in China,” American Economic Review, 2021, 111 (7), 2065–2100.
  8. [11]Chetty, Raj, John N Friedman, Tore Olsen, and Luigi Pistaferri, “Adjustment costs, firm responses, and micro vs. macro labor supply elasticities: Evidence from Danish tax records,” Quarterly Journal of Economics, 2011, 126 (2), 749–804.
  9. [13]Elschner, Christina, “Special tax regimes and the choice of organizational form: Evidence from the European Tonnage Taxes,” Journal of Public Economics, 2013, 97, 206–216.DOI
  10. [14]Fack, Gabrielle and Camille Landais, “The effect of tax enforcement on tax elasticities: Evidence from charitable contributions in France,” Journal of Public Economics, 2016, 133, 23–40.DOI
  11. [15]Fisman, Raymond and Shang-Jin Wei, “Tax rates and tax evasion: Evidence from “missing imports” in China,” Journal of Political Economy, 2004, 112 (2), 471–496.
  12. [16]GAO, “Artificial Intelligence: IRS Actions Needed to Address Skills Gaps, Information Quality, and Strategic Management,” Technical Report GAO-26-107522, U.S. Government Accountability Office March 24 2026.
  13. [17]Hendren, Nathaniel and Ben Sprung-Keyser, “A Unified Welfare Analysis of Government Policies,” Quarterly Journal of Economics, 2020, 135 (3), 1209–1318.
  14. [18]Heyes, Anthony, “A theory of filtered enforcement,” Journal of Environmental Economics and Management, 2002, 43 (1), 34–46.DOI
  15. [19]Kleven, Henrik J and Mazhar Waseem, “Using notches to uncover optimization frictions and structural elasticities: Theory and evidence from Pakistan,” Quarterly Journal of Economics, 2013, 128 (2), 669–723.DOI
  16. [20]Kleven, Henrik Jacobsen, Martin B Knudsen, Claus Thustrup Kreiner, Søren Pedersen, and Emmanuel Saez, “Unwilling or unable to cheat? Evidence from a tax audit experiment in Denmark,” Econometrica, 2011, 79 (3), 651–692.
  17. [21]Kuchumova, Yulia, “The Optimal Deterrence of Tax Evasion: The Trade-off Between Information Reporting and Audits,” Journal of Public Economics, 2017, 145, 162–180.DOI
  18. [23]Macho-Stadler, Ines and David Perez-Castrillo, “Optimal enforcement policy and firms’ emissions and compliance with environmental taxes,” Journal of Environmental Economics and Management, 2006, 51 (1), 110–131.
  19. [24]Macho-Stadler, Inés and J David Pérez-Castrillo, “Auditing with signals,” Economica, 2002, 69 (273), 1–20.
  20. [26]Menichini, Anna Maria C and Peter J Simmons, “Sorting the good guys from bad: on the optimal audit structure with ex-ante information acquisition,” Economic Theory, 2014, 57, 339–376.DOI
  21. [28]Phillips, Mark D., “Deterrence vs. gamesmanship: Taxpayer response to targeted audits and endogenous penalties,” Journal of Economic Behavior & Organization, 2014, 100, 81–98.DOI
  22. [29]Saez, Emmanuel, “Do taxpayers bunch at kink points?,” American Economic Journal: Economic Policy, 2010, 2 (3), 180–212.
  23. [30]Sanchez, Isabel and Joel Sobel, “Hierarchical design and enforcement of income tax policies,” Journal of Public Economics, 1993, 50 (3), 345–369.
  24. [31]Scotchmer, Suzanne, “Audit classes and tax enforcement policy,” American Economic Review, 1987, 77 (2), 229–233.
  25. [32]Sharma, Rishi R, Joel Slemrod, Michael Stimmelmayr, John D Wilson, and Peter Choi, “Optimal Dual-Regime Business Tax Systems,” Technical Report, National Bureau of Economic Research 2025.
  26. [33]Slemrod, Joel, “Cheating ourselves: The economics of tax evasion,” Journal of Economic perspectives, 2007, 21 (1), 25–48.DOI
  27. [34]Yitzhaki, Shlomo, “Income tax evasion: A theoretical analysis,” Journal of Public Economics, 1974, 3 (2), 201–202.DOI
  28. [35]Zilberman, Eduardo, “Audits or Distortions: The Optimal Scheme to Enforce Self-Employment Income Taxes,” Journal of Public Economic Theory, 2016, 18 (4), 511–544.DOI

Paper details

vlnc:W_kzqxm9f8
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Economics, Econometrics and Finance

Not recorded for this paper

View statement
Looking for related papers…
Contents