Reputation and Delegated Public Spending

arXiv:2509.03087 · econ.TH · Submitted 2025-09-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: We're looking at the paper: "Reputation and Delegated Public Spending".

Warren: Here's the gist of the paper. How does uncertain implementation shape fiscal policy over time?

Ruth: The dynamic activation threshold is no greater than the square of the static reputation cutoff; this bound is exact and independent of discounting in a linear-benefit, quadratic-cost economy <ref:2509.03087#pg2> Core Mechanism The paper studies an uninformed fiscal authority choosing a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds <ref:2509.03087#pg2>.

Warren: Reputation is treated as a Bayesian state variable linking current fiscal control to future capacity <ref:2509.03087#pg2>.

Ruth: The dynamic activation threshold is no greater than the square of the static reputation cutoff <ref:2509.03087#pg2>.

Paper summary: Ruth: So we're looking at this paper called "Reputation and Delegated Public Spending." It deals with how uncertain implementation shapes fiscal policy over time. Basically, they're studying an uninformed fiscal authority that sets a distortionary tax-financed mandate and an audit intensity while a privately informed executive decides whether to deliver or divert the money.

Warren: And what really matters here is this concept of reputation being a Bayesian state variable linking current fiscal control to future capacity. It means how you look right now affects what you can actually do later, which is interesting because it moves beyond just static parameters in tax problems.

Iris: I'm curious about the core mechanism they describe. They claim that the dynamic activation threshold is no greater than the square of the static reputation cutoff, and this bound is exact even in a linear-benefit, quadratic-cost economy, and crucially, it’s independent of discounting when patience scales both the trial mandate and the future rent that sustains delivery.

Ruth: Exactly. So in their two-period benchmark with perfect observation, they find that a positive trial mandate can be optimal even at beliefs for which every single positive tax is statically undesirable. That suggests you can start something if you believe it might pay off later, even when the immediate cost seems too high, and that's true in a linear-benefit, quadratic-cost economy where the threshold doesn't change because patience scales both the trial mandate and the future rent that sustains delivery.

Warren: That’s a strong finding for dynamic activation; it suggests you can initiate fiscal activity based on future possibilities rather than just immediate static desirability. But then they move to infinite-horizon dynamics, where things get more complex because posterior reputation becomes a bounded martingale, and its one-step conditional variance has an exact policy-dependent formula.

Iris: And that variance formula is policy dependent because the expected amount of authorized revenue converted into public goods is what they call effective fiscal capacity, which depends on whether you're inducing pooling at a certain level or not <ref:2509.03087#pg2>.

Ruth: That brings up a big question about how that executive makes their choice; the paper points out that probability affects the signal distribution and the posterior that supports that same implementation choice, so you can't just look at one number <ref:2509.03087#pg3>.

Paper summary: Warren: And they show this by computing equilibria on a finite state-action approximation, reporting only cases that pass a hard-policy certificate on the declared grids, which leads to some specific results about where audits concentrate and how spending need shifts the activation frontier.

Iris: They also point out that increasing the discount factor from zero point eight eight to zero point nine two lowers the highest belief at which both need states remain inactive from zero point two one seven to zero point one eight three, which shows how patience interacts with reputation and need states <ref:2509.03087#pg3>.

Ruth: What that means for the real world is that a high-need realization raises effective capacity without changing the mean reputation, but it changes belief dispersion and exposure to the inactive region, which is a subtle distinction when you're trying to manage public spending <ref:2509.03087#pg3>.

Warren: And if you look at the computational verification part, they confirm that this mechanism separates administrative fiscal capacity from the revenue that can credibly be placed under executive control <ref:2509.03087#pg2>. That separation is key for understanding how much spending is actually feasible versus just what's available in the budget.

Iris: They also highlight that zero mandates can create closed inactive classes where fiscal activity and learning stop together, which suggests a situation where no one acts and no one learns at all <ref:2509.03087#pg2>.

Ruth: So when we look at the implications, these certified equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need shocks <ref:2509.03087#pg2>. That means the system doesn't just react linearly when you change how much money you want to spend; it changes based on what happened before.

Warren: And they also mention that with noisy monitoring, the mandate scale and audit intensity jointly determine current implementation and the law of motion for reputation, which is a bit of a mouthful but important for anyone trying to design an effective oversight system <ref:2509.03087#pg2>.

Iris: The paper also touches on robustness by establishing why dynamic activation can happen and why a zero-mandate region, when it exists, is absorbing, which is important because it explains the boundaries of policy options <ref:2509.03087#pg2>. However, they also note that the analytical bound on p20 rests on perfect verification of arbitrarily small projects, whereas noisy monitoring makes informative experiments lumpy <ref:2509.03087#pg2>.

Paper summary: Ruth: And the numerical inactive frontier isn't just some mechanical discretization of a two-period cutoff, but it emerges from the joint choice of scale, audit intensity and strategic implementation <ref:2509.03087#pg2>. That suggests that in practice, figuring out where you can push spending is about making choices across all those variables simultaneously.

Warren: It’s interesting to see how they handle the tax foundation too; the fiscal authority levies a proportional labor tax tau in the range of zero to one and the private welfare cost of raising r is defined by K(r) = one minus

one minus tau (r): (one/epsilon + one/epsilon) <ref:2509.03087#pg9>. The objective function in their model is B(g)-K(r) <ref:2509.03087#pg9>.

Iris: And they find that zero is optimal exactly when p less than or equal to p0, and the envelope theorem gives S'(p0) equals zero, which simplifies things a bit regarding the boundary conditions of the tax structure <ref:2509.03087#pg9>.

Ruth: So we've covered how this paper, "Reputation and Delegated Public Spending," explains how uncertainty in implementation creates dynamic fiscal capacity through reputation feedback, showing that policy choices aren't just about the immediate budget but about the long-term credibility of delivery <ref:2509.03087#pg1>.

Warren: And if you think about what this means for policymakers, it suggests that designing a system needs to account for how reputation evolves based on implementation choices, not just static rules or simple discounting <ref:two thousand five hundred nine point zero three zero eight seven#pg2.

Iris: It implies that auditing policies might need to be non-monotone because they affect the signal distribution, and spending need shifts the activation frontier, which are things you can't ignore when trying to model real-world fiscal capacity <ref:two thousand five hundred nine point zero three zero eight seven#pg3.

Ruth: That’s what we were discussing today with "Reputation and Delegated Public Spending." It shows that fiscal capacity is partly a dynamic problem of credible implementation, and with noisy monitoring, mandate scale and audit intensity jointly determine current implementation and the law of motion for reputation <ref:two thousand five hundred nine point zero three zero eight seven#pg2.

Warren: So we've seen how these dynamics lead to inactive classes and history-dependent capacity, which is a bit more nuanced than just looking at a single budget constraint <ref:two thousand five hundred nine point zero three zero eight seven#pg2.

Conclusion: Ruth: So we're wrapping up this look at "Reputation and Delegated Public Spending." This paper is about how uncertainty in how you actually implement a policy changes what you can actually spend over time.

Warren: It boils down to this idea that reputation isn't just a number; it’s a state variable, linking your current fiscal control to your future capacity for action.

Iris: I think the core mechanism they build is about that dynamic activation threshold, which they show is exactly bounded by the square of the static reputation cutoff.

Ruth: That's what makes me think this moves beyond just static budget constraints; it's about how a system learns and adapts over time.

Warren: And if you look at the infinite horizon, they find that posterior reputation acts like a bounded martingale, which gives them an exact formula for the variance of future reputation changes based on their policy choices.

Iris: That variance formula depends on whether they're inducing pooling at a certain level or not, which shows how the implementation strategy directly affects uncertainty.

Ruth: What this means for us is that fiscal capacity isn't just about having money in the bank; it’s deeply tied to the credibility of delivery over a long period.

Warren: And if you want to run this kind of system, you have to worry about how that reputation evolves based on what people actually choose to do with the funds.

Iris: They also show that this can lead to history-dependent limiting fiscal capacity, meaning what you can spend tomorrow depends on the whole timeline leading up to it.

Ruth: That's a big picture point; it suggests we need models that think about how past choices shape future possibilities in a complex way.

Warren: It also means auditing policies might not be steady; they can be non-monotone because the signal distribution itself changes depending on whether the executive chooses to implement or divert the money.

Iris: And they show that spending need shocks don't just change how much you spend right now; they shift where you are on that activation frontier entirely.

Ruth: So, "Reputation and Delegated Public Spending" tells us that if you model policy this way, the system itself starts to determine its own limits through these feedback loops.

Warren: It’s a bit of a heavy lift computationally because they had to use finite state-action approximations to check their results.

Iris: And they flag that the analytical bound on the activation threshold relies on perfect verification, which is different from what we see when monitoring is just noisy and lumpy in reality.

Ruth: So if you want to understand the real world application of this paper, you have to consider how imperfect monitoring changes those theoretical bounds.

Warren: That's exactly where I want to go next; we need to look at how that noise actually affects the strategy of a quantitative trader trying to deploy capital under these kinds of uncertain incentives.

Emin Ablyatifov, Georgy Lukyanov

Geneva Business School · Toulouse School of Economics

econ.TH

Submitted: 2025-09-03

Updated: 2026-10-27

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 78/100

The gist: How does uncertain implementation shape fiscal policy over time? The dynamic activation threshold is no greater than the square of the static reputation cutoff…

Key concepts

Reputation
Reputation is treated as a Bayesian state variable linking current fiscal control to future capacity. It represents how the public perceives the authority's honesty and competence, which influences future decisions about whether to implement or divert funds.
Dynamic Activation Threshold
This threshold determines when fiscal activity begins. The paper proves that this dynamic threshold is no greater than the square of the static reputation cutoff, meaning uncertainty in implementation limits how high the authority can set its activation bar.
Fiscal Capacity
Fiscal capacity is shown to be a dynamic problem dependent on credible implementation. It means that an authority's ability to spend money over time is not fixed but depends on how reliably they implement their policies and maintain public trust.

Terminology

Summary

How does uncertain implementation shape fiscal policy over time? p. 0

Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

The dynamic activation threshold is no greater than the square of the static reputation cutoff; this bound is exact and independent of discounting in a linear-benefit, quadratic-cost economy <ref:2509. p. 0
Where the paper says it

In a two-period benchmark, a mandate is also an experiment: delivery sacrifices current rents but preserves future access. We prove that the dynamic activation threshold is no greater than the square of the static reputation cutoff; the bound is exact and independent of discounting in a linear-benefit, quadratic-cost economy.

03087#pg2>

Core Mechanism

The paper studies an uninformed fiscal authority choosing a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds <ref:2509.03087#pg2>. Reputation is treated as a Bayesian state variable linking current fiscal control to future capacity <ref:2509. p. 0

Where the paper says it

We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds. Reputation is the Bayesian state variable linking current fiscal control to future capacity.

03087#pg2>. The dynamic activation threshold is no greater than the square of the static reputation cutoff <ref:2509. p. 0
Where the paper says it

In a two-period benchmark, a mandate is also an experiment: delivery sacrifices current rents but preserves future access. We prove that the dynamic activation threshold is no greater than the square of the static reputation cutoff; the bound is exact and independent of discounting in a linear-benefit, quadratic-cost economy.

03087#pg2>.

Static Benchmark Results

In a two-period model with perfect observation, the dynamic activation threshold satisfies pD ≤ p20. A positive trial mandate can therefore be optimal at beliefs for which every positive tax is statically undesirable <ref:2509. p. 1

Where the paper says it

We solve this mimicry decision and prove that the dynamic activation threshold satisfies pD ≤ p 2 0 < p0. A positive trial mandate can therefore be optimal at beliefs for which every positive tax is statically undesirable.

03087#pg7>. The bound is exact in a linear-benefit, quadratic-cost economy where the threshold is independent of discounting because patience scales both the trial mandate and the future rent that sustains delivery <ref:2509. p. 1
Where the paper says it

The bound is exact in a linear-benefit, quadratic-cost economy. In that benchmark the threshold is independent of discounting because patience scales both the trial mandate and the future rent that sustains delivery.

03087#pg9>.

Infinite Horizon Dynamics

In the infinite-horizon model, posterior reputation is a bounded martingale and its one-step conditional variance admits an exact policy-dependent formula <ref:2509.03087#pg2>. Zero mandates can create closed inactive classes in which fiscal activity and learning stop together <ref:2509. p. 0

Where the paper says it

In the infinite-horizon model, posterior reputation is a bounded martingale and its one-step conditional variance admits an exact policy-dependent formula. Zero mandates can create closed inactive classes in which fiscal activity and learning stop together.

03087#pg2>. The stationary problem is nonlinear because the opportunistic executive’s implementation probability affects the signal distribution and the posterior that supports that same implementation choice <ref:2509. p. 2
Where the paper says it

probability affects the signal distribution and the posterior that supports that same implementation choice. We compute equilibria on a finite state-action approximation and report only cases that pass a hard-policy certificate on the declared grids.

03087#pg2>.

Policy Implications

The certified equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need shocks <ref:2509.03087#pg2>. Increasing the discount factor from 0. p. 2

Where the paper says it

In the certified equilibria, audits are concentrated at intermediate reputations, spending need shifts the activation frontier, and effective fiscal capacity depends on the history of public signals. Increasing the discount factor from 0.

88 to 0.92 lowers the highest belief at which both need states remain inactive from 0. p. 2
Where the paper says it

Increasing the discount factor from 0. 92 lowers the highest belief at which both need states remain inactive from 0.

217 to 0.183 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg7>. A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region <ref:2509. p. 2
Where the paper says it

92 lowers the highest belief at which both need states remain inactive from 0. A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region.

03087#pg7>.

Computational Verification

The paper computes stationary equilibria on a finite state-action approximation and reports only cases that pass a hard-policy certificate on the declared grids <ref:2509.03087#pg2>. The certified equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need <ref:2509. p. 0

Where the paper says it

On a finite approximation, we compute stationary equilibria with noisy signals, endogenous auditing, and spending-need shocks and certify them against every action in a common adaptive cloud. The equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need: shocks leave mean reputation unchanged but alter its dispersion and exposure to the inactive region.

03087#pg2>. The mechanism separates administrative fiscal capacity from the revenue that can credibly be placed under executive control <ref:2509. p. 1
Where the paper says it

We call the expected amount of authorized revenue converted into public goods effective fiscal capacity. Administrative institutions determine the feasible revenue set; reputation and dynamic incentives determine how much of that set is used successfully.

03087#pg2>.

Robustness and Scope

The analytical results establish why dynamic activation can occur and why a zero-mandate region, when it exists, is absorbing <ref:2509.03087#pg2>. The analytical p20 bound rests on perfect verification of arbitrarily small projects, whereas noisy monitoring makes informative experiments lumpy <ref:2509. p. 7

Where the paper says it

Perfect verification matters for the exact local bound. In the two-period benchmark even a vanishing project is fully informative, so the value of the experiment is first order in its size.

03087#pg2>. The numerical inactive frontier is therefore not a mechanical discretization of the two-period cutoff: it emerges from the joint choice of scale, audit intensity and strategic implementation <ref:2509.03087#pg2>.

Conclusion

Fiscal capacity is partly a dynamic problem of credible implementation <ref:2509.03087#pg2>. With noisy monitoring, mandate scale and audit intensity jointly determine current implementation and the law of motion for reputation <ref:2509. p. 1

Where the paper says it

Reputation is therefore not merely a parameter in a one-shot tax problem. It is a Bayesian state variable whose law of motion is itself affected by the scale and monitoring of fiscal activity.

03087#pg2>. If a zero mandate is uninformative, low reputation can shut down fiscal activity and learning simultaneously <ref:2509. p. 1
Where the paper says it

Third, when a zero mandate generates no information, beliefs at which both spending states prescribe zero activity form closed inactive classes. Fiscal activity and learning then stop together, and initial reputation can have permanent consequences for effective capacity.

03087#pg2>. The certified stationary equilibria show how this mechanism produces inactive classes, non-monotone audit policies, history-dependent limiting capacity, and nonlinear responses to spending need shocks <ref:2509. p. 0
Where the paper says it

On a finite approximation, we compute stationary equilibria with noisy signals, endogenous auditing, and spending-need shocks and certify them against every action in a common adaptive cloud. The equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need: shocks leave mean reputation unchanged but alter its dispersion and exposure to the inactive region.

03087#pg2>.

Tax Foundation

The fiscal authority levies a proportional labor tax τ ∈ [0, 1) and the private welfare cost of raising r is K(r) = 1 − [1 − τ (r)]1+/ε1 + 1/ε <ref:2509.03087#pg9>. The objective in (A. p. 4

Where the paper says it

(3) The value satisfies S(p) = 0 for p ≤ p0 and S ′ (p) = B(ρ(p)) for p > p0. The threshold is the sign of the derivative of a strictly concave objective at the origin.

4) is B(g)-K(r) <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>. Zero is therefore optimal exactly when p ≤ p0. p. 7
Where the paper says it

At p = 0, zero is uniquely optimal. The local comparison that delivers the bound is lim z↓p0 JM(p, z) δρ(z) = B ′ (0) p p0 − p0 , (15) which is positive exactly when p > p2 0 .

The envelope theorem gives S′+(p0) = 0. p. 1
Where the paper says it

Dynamically, however, a mandate is also an experiment. An opportunistic executive that delivers today gives up current diversion but preserves access to tomorrow’s mandate.

Two-Period Proofs

The authority’s period-1 value is V1(p) = max r∈[0,r¯] W(p, r). Conditional on inducing pooling at p > p0, the authority chooses rP(p) = δρ(p). p. 6

Where the paper says it

The three regions induce pooling, partial mimicry and separation respectively, and the expressions agree at their boundaries. Conditional on pooling, the authority chooses the largest sustainable mandate, r P (p) = δρ(p).

The local expansion of JM by δρ(z) yields (15), which is positive exactly when p > p20. p. 7
Where the paper says it

At p = 0, zero is uniquely optimal. The local comparison that delivers the bound is lim z↓p0 JM(p, z) δρ(z) = B ′ (0) p p0 − p0 , (15) which is positive exactly when p > p2 0 .

The ratio defining pM has a positive continuous extension at p0, and the ratio defining pS is positive and continuous on a nonempty compact interval. p. 8
Where the paper says it

So the squared cutoff is not merely an upper bound, and the closed form makes the absence of δ from the threshold visible: patience determines how large the trial mandate is and how much the authority gains from running it, but the belief at which running it becomes worthwhile is fixed by the benefit-cost ratio alone. The proofs are in B and C.

Infinite-Horizon Identities

The process of public reputation is a martingale under any equilibrium policy. The one-step conditional variance of reputation is Var(p'p, ξ, r, a, m) = p2(1 − q)2λ(r, a)2P + (1 − P +). The martingale convergence theorem gives almost-sure and L1 convergence to a limit p∞. p. 0

Where the paper says it

A positive trial mandate may therefore be optimal when every positive tax is statically undesirable. In the infinite-horizon model, posterior reputation is a bounded martingale and its one-step conditional variance admits an exact policy-dependent formula.

The variance response is positive and persistent in all six exercises, and at p init = 0. p. 0
Where the paper says it

A positive trial mandate may therefore be optimal when every positive tax is statically undesirable. In the infinite-horizon model, posterior reputation is a bounded martingale and its one-step conditional variance admits an exact policy-dependent formula.

50 its maximum is 0.0540 under β = 0.88 and 0.0483 under β = 0.92

Numerical Protocol

The authority then makes a pure, hard-max choice. The cloud was informed by dense screens and bounded continuous refinements, but we do not convert that search into a continuum optimality claim. p. 2

Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

The hard-policy certificate covers a finite reputation grid and a common adaptive action cloud. p. 0
Where the paper says it

On a finite approximation, we compute stationary equilibria with noisy signals, endogenous auditing, and spending-need shocks and certify them against every action in a common adaptive cloud. The equilibria display history-dependent limiting fiscal capacity, non-monotone auditing, and nonlinear responses to spending need: shocks leave mean reputation unchanged but alter its dispersion and exposure to the inactive region.

References

Aghion, P., Tirole, J. p. 2

Where the paper says it

There is also an organizational interpretation. Formal and effective control need not coincide (Aghion and Tirole, 1997); relational incentives make current conduct depend on future access (Baker et al.

, 1997. Formal and real authority in organizations. p. 2
Where the paper says it

Formal and effective control need not coincide (Aghion and Tirole, 1997); relational incentives make current conduct depend on future access (Baker et al. , 2002); and reputation can support authority (Al-Najjar, 2001).

Journal of Political Economy 105 (1), 1–29 <ref:2509. p. 2
Where the paper says it

A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region. The paper contributes to dynamic public economics and political economy.

03087#pg9>.

Al-Najjar, N. p. 2

Where the paper says it

Formal and effective control need not coincide (Aghion and Tirole, 1997); relational incentives make current conduct depend on future access (Baker et al. , 2002); and reputation can support authority (Al-Najjar, 2001).

I., 2001. A reputational model of authority. p. 1
Where the paper says it

This paper studies the resulting dynamic control problem. An uninformed fiscal authority chooses a distortionary revenue mandate and, in the full model, an audit intensity.

Journal of Economic Behavior & Organization 46 (2), 165–191 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>.

Baker, G.P., Gibbons, R., Murphy, K.J., 2002. Relational contracts and the theory of the firm. p. 2

Where the paper says it

There is also an organizational interpretation. Formal and effective control need not coincide (Aghion and Tirole, 1997); relational incentives make current conduct depend on future access (Baker et al.

Quarterly Journal of Economics 117 (1), 39–84 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>.

Besley, T., Ilzetzki, E., Persson, T. p. 2

Where the paper says it

The mechanism complements leading accounts of state capacity. In Besley and Persson (2009) and Besley et al.

p. 2
Where the paper says it

The mechanism complements leading accounts of state capacity. In Besley and Persson (2009) and Besley et al.

, 2013. Weak states and steady states: The dynamics of fiscal capacity. p. 1
Where the paper says it

Third, when a zero mandate generates no information, beliefs at which both spending states prescribe zero activity form closed inactive classes. Fiscal activity and learning then stop together, and initial reputation can have permanent consequences for effective capacity.

American Economic Journal: Macroeconomics 5 (4), 205–235. p. 2
Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Besley, T., Persson, T., 2009. The origins of state capacity: Property rights, taxation, and politics. p. 0

Where the paper says it

We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds. Reputation is the Bayesian state variable linking current fiscal control to future capacity.

American Economic Review 99 (4), 1218–1244 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>.

Besley, T., Smart, M. p. 2

Where the paper says it

, 1997; Martin, 2010), reputation as a source of policy discipline (Phelan, 2006; Fudenberg et al. , 2022), and fiscal restraints as a way to shape political incentives (Besley and Smart, 2007; Dovis and Kirpalani, 2021).

, 2007. Fiscal restraints and voter welfare. p. 2
Where the paper says it

, 1997; Martin, 2010), reputation as a source of policy discipline (Phelan, 2006; Fudenberg et al. , 2022), and fiscal restraints as a way to shape political incentives (Besley and Smart, 2007; Dovis and Kirpalani, 2021).

Journal of Public Economics 91 (3–4), 755–773. p. 2
Where the paper says it

A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region. The paper contributes to dynamic public economics and political economy.

Bobonis, G. p. 2

Where the paper says it

Here administrative capacity is fixed and households comply mechanically; reputation operates on the implementation side by changing how much revenue is delegated and successfully converted into public goods. Monitoring can discipline or select political agents (Bobonis et al.

J., C´amara Fuertes, L.R., Schwabe, R., 2016. Monitoring corruptible politicians. p. 1
Where the paper says it

Reputation is therefore not merely a parameter in a one-shot tax problem. It is a Bayesian state variable whose law of motion is itself affected by the scale and monitoring of fiscal activity.

American Economic Review 106 (8), 2371–2405. p. 2
Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Brollo, F., Nannicini, T., Perotti, R., Tabellini, G., 2013. The political resource curse. p. 3

Where the paper says it

On the increasing side of the Laffer curve, revenue is an equivalent policy instrument, K is increasing and strictly convex, and K′ (0) = 1. The same appendix records the underlying resource accounting.

American Economic Review 103 (5), 1759–1796 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>.

Chen, C., Li, C., 2019. Laboratories of democracy: Policy experimentation under decentralization. p. 0

Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

American Economic Journal: Microeconomics 11 (3), 125–154. p. 2
Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Cripps, M.W., Mailath, G.J., Samuelson, L., 2004. Imperfect monitoring and impermanent reputations. p. 2

Where the paper says it

, 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al. In our model auditing is chosen jointly with the fiscal exposure it monitors, and imperfect monitoring can make reputations impermanent (Cripps et al.

Econometrica 72 (2), 407–432.

Dovis, A., Kirpalani, R. p. 2

Where the paper says it

, 1997; Martin, 2010), reputation as a source of policy discipline (Phelan, 2006; Fudenberg et al. , 2022), and fiscal restraints as a way to shape political incentives (Besley and Smart, 2007; Dovis and Kirpalani, 2021).

, 2021. Rules without commitment: Reputation and incentives. p. 6
Where the paper says it

Conditional on pooling, the authority chooses the largest sustainable mandate, r P (p) = δρ(p). That fiscal-restraint result is worth recording, but closely related incentive-compatible rules already appear in reputation and political-agency models (Besley and Smart, 2007; Dovis and Kirpalani, 2021), and it is not our contribution.

Review of Economic Studies 88 (6), 2833–2856. p. 2
Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Fudenberg, D. p. 2

Where the paper says it

Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al. , 1997; Martin, 2010), reputation as a source of policy discipline (Phelan, 2006; Fudenberg et al.

, Gao, Y., Pei, H., 2022. A reputation for honesty. p. 0
Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

Journal of Economic Theory 204, 105508. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

Gadenne, L. p. 2

Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

, 2017. Tax me, but spend wisely? p. 0
Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

Sources of public finance and government accountability. p. 2
Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

American Economic Journal: Applied Economics 9 (1), 274–314. p. 2
Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Judd, K.L., 1997. Computational economics and economic theory: Substitutes or complements? p. 2

Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

Journal of Economic Dynamics and Control 21 (6), 907–942 <ref:2509.03087#pg9>.

Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in nonlinear multivariate models. p. 6

Where the paper says it

Conditional on pooling, the authority chooses the largest sustainable mandate, r P (p) = δρ(p). That fiscal-restraint result is worth recording, but closely related incentive-compatible rules already appear in reputation and political-agency models (Besley and Smart, 2007; Dovis and Kirpalani, 2021), and it is not our contribution.

Journal of Econometrics 74 (1), 119–147 <ref:2509. p. 0
Where the paper says it

lukyanov@tse-fr. eu 1 arXiv:2509.

03087#pg9>.

Kreps, D.M., Wilson, R. p. 4

Where the paper says it

It shows exactly when treating a mandate as both spending and a test changes the activation decision at low reputation. The strategic motive is the classical reputation-game logic of short-run sacrifice for future rents (Kreps and Wilson, 1982; Milgrom and Roberts, 1982); what is endogenous here is the fiscal stake attached to it.

, 1982. Reputation and imperfect information. p. 1
Where the paper says it

Policy can therefore change reputation risk without changing expected reputation. Third, when a zero mandate generates no information, beliefs at which both spending states prescribe zero activity form closed inactive classes.

Journal of Economic Theory 27 (2), 253–279 <ref:2509. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

03087#pg9>.

Krusell, P. p. 2

Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

, Quadrini, V., R´ıos-Rull, J.-V., 1997. Politico-economic equilibrium and economic growth. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

Journal of Economic Dynamics and Control 21 (1), 243–272 <ref:2509.03087#pg9>.

Larcinese, V. p. 2

Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

, Sircar, I. p. 2
Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

, 2017. Crime and punishment the British way: Accountability channels following the MPs’ expenses scandal. European Journal of Political Economy 47, 75–99. p. 2
Where the paper says it

A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region. The paper contributes to dynamic public economics and political economy.

Martin, F.M., 2010. Markov-perfect capital and labor taxes. p. 2

Where the paper says it

The paper contributes to dynamic public economics and political economy. Existing work studies sequentially determined fiscal policies as Markov functions of economic and political states (Krusell et al.

Journal of Economic Dynamics and Control 34 (3), 503–521 <ref:2509.03087#pg9>.

Milgrom, P. p. 4

Where the paper says it

It shows exactly when treating a mandate as both spending and a test changes the activation decision at low reputation. The strategic motive is the classical reputation-game logic of short-run sacrifice for future rents (Kreps and Wilson, 1982; Milgrom and Roberts, 1982); what is endogenous here is the fiscal stake attached to it.

, Roberts, J. p. 4
Where the paper says it

It shows exactly when treating a mandate as both spending and a test changes the activation decision at low reputation. The strategic motive is the classical reputation-game logic of short-run sacrifice for future rents (Kreps and Wilson, 1982; Milgrom and Roberts, 1982); what is endogenous here is the fiscal stake attached to it.

, 1982. Predation, reputation, and entry deterrence. p. 0
Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

Journal of Economic Theory 27 (2), 280–312 <ref:2509. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

03087#pg9>.

Phelan, C., 2006. Public trust and government betrayal. p. 0

Where the paper says it

Government Reputation and Fiscal Capacity∗ Emin Ablyatifov† Georgy Lukyanov‡ Abstract How does uncertain implementation shape fiscal policy over time? We study an uninformed fiscal authority that chooses a distortionary tax-financed mandate and audit intensity while a privately informed executive decides whether to deliver or divert the proceeds.

Journal of Economic Theory 130 (1), 27–43. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

Strˆımbu, O. p. 2

Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

, Gonz´alez, P. p. 2
Where the paper says it

Monitoring can discipline or select political agents (Bobonis et al. , 2016; Strˆımbu and Gonz´alez, 2018; Larcinese and Sircar, 2017), and tax-financed revenue can generate accountability differently from external transfers (Gadenne, 2017; Brollo et al.

, 2018. Does transparency reduce political corruption? p. 2
Where the paper says it

A high-need realization raises effective capacity without changing mean reputation; instead, it changes belief dispersion and exposure to the inactive region. The paper contributes to dynamic public economics and political economy.

Journal of Public Economic Theory 20 (2), 123–135. p. 2
Where the paper says it

Here the uninformed principal chooses how many public resources to place within reach of the agent whose reliability is being learned. Finally, the computational analysis follows the view that numerical methods can extend, but should not blur the warrant of, economic theory (Judd, 1997): continuum theorems, finite-cloud equilibrium certificates, and parameter-specific comparative exercises are kept separate.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Dynamic fiscal policy simulation based on reputation as a Bayesian state variable allows AI to model how an uninformed authority chooses a distortionary mandate and audit intensity over time. This system can predict fiscal capacity based on history-dependent limiting fiscal capacity, which is history dependent rather than just current state variables.

  2. Optimizing policy choices under uncertainty by incorporating the dynamic activation threshold derived from the square of the static reputation cutoff allows AI to determine when a positive trial mandate may be optimal even when every positive tax is statically undesirable. This enables AI to select mandates that are informative in a way that balances current fiscal cost with future access preservation.

  3. Developing an infinite-horizon reinforcement learning agent for fiscal policy where the state includes public probability of commitment and spending-need states allows the agent to learn optimal sequences of tax rates, audit intensities, and implementation policies. This system can determine if zero mandates create closed inactive classes where fiscal activity and learning stop together.

  4. Implementing a mechanism to adjust belief dispersion based on spending need shocks allows AI to model how a high-need realization changes belief dispersion and exposure to the inactive region without changing mean reputation. This enables AI to proactively manage risk exposure when external economic conditions change.

Abstract

This paper studies a repeated funding relationship between a grant administrator and an implementing agency. The agency is either committed to implementation or able to divert the advance. The administrator chooses whether to fund and whether to verify, but cannot commit to future decisions. We construct a stationary equilibrium in which agencies with intermediate reputations are audited. An inspection either reveals diversion and ends funding or leads to continued funding with intermittent auditing. In the latter case, audits sustain implementation without further revealing type, while a strategic agency diverts unaudited advances. Cheaper auditing makes funding worthwhile at lower reputations. For initial beliefs audited under both cost levels, it raises the administrator's value but reduces expected long-run implemented spending: a higher reputation is required to retain funding, and fewer strategic agencies meet that requirement. For beliefs audited under both programs, better detection at unchanged cost instead raises implementation without changing the probability of continued funding. These results show why project delivery alone need not measure the value of improved monitoring.

Related papers