📈 economics

Government Reputation and Fiscal Capacity

This paper analyzes how a state's reputation for an executive's willingness to implement public spending governs the allocation of fiscal resources, demonstrating that dynamic testing through mandates can justify positive taxation even when statically undesirable and revealing how endogenous auditing and spending shocks shape equilibrium fiscal capacity and reputation risk.

Original authors: Emin Ablyatifov, Georgy Lukyanov

Published 2026-07-31

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: Emin Ablyatifov, Georgy Lukyanov

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are the captain of a massive ship, but you can't steer it yourself. You have to hire a first mate to navigate the waters. You know the ship has a treasure chest full of gold (tax revenue) that you can legally collect from the passengers. However, you don't know if your first mate is a honest sailor who will use that gold to fix the sails and feed the crew, or a scoundrel who will steal the gold to buy a private island. This is the heart of a problem economists call "delegation": how much power do you give to someone when you aren't sure they will use it for your benefit?

In the world of government, this is the difference between having the ability to collect taxes and having the willingness to hand that money over to a leader. If the leader is a thief, giving them more money just means more stolen gold. If they are honest, it means better roads and schools. The tricky part is that the leader knows their own intentions, but the government (the captain) does not. This paper explores a fascinating question: Can a government use a small amount of money as a "test drive" to figure out if a leader is honest, even if that test drive seems like a bad idea at first glance?

The Big Idea: The "Test Drive" for Leaders

This paper, written by Emin Ablyatifov and Georgy Lukyanov, dives into the mechanics of government reputation and fiscal capacity. In simple terms, "fiscal capacity" isn't just about how good a country is at collecting taxes; it's about how much of that collected money the government feels safe enough to actually let the leader spend.

The authors build a model where a "Fiscal Authority" (the government) tries to decide how much money to give an "Executive" (the leader). The leader can be two types:

The Committed Type: Always uses the money for public goods (like parks and schools).
The Opportunistic Type: Wants to steal the money for themselves, but might pretend to be honest if it helps them keep their job later.

The government doesn't know which type they have. They only have a "reputation score"—a guess, or a probability, that the leader is honest.

The Static Trap: Why Small Tests Usually Fail

In a simple, one-shot scenario (the "static benchmark"), the math is pretty strict. If the government's guess that the leader is honest is too low, they won't give them any money at all. Why? Because if the leader is likely a thief, giving them even a tiny bit of money is a waste. The cost of raising that money (through taxes that annoy people) is higher than the tiny chance of getting a good result. It's like refusing to buy a lottery ticket because the odds of winning are too low to justify the cost of the ticket.

The Dynamic Twist: The "Square" Rule

Here is where the paper gets exciting. The authors ask: What if the game isn't just one round, but goes on forever? What if the leader knows that if they steal today, they lose their job tomorrow, but if they behave today, they get to keep their job and the power to spend money tomorrow?

The paper proves a surprising result: A government might choose to give a small "trial mandate" (a test amount of money) even when the odds of the leader being honest are very low.

In fact, the authors prove a specific mathematical rule: The point at which it becomes worth running this "test" is no higher than the square of the point where you would normally start giving money.

If the "safe" threshold for giving money is 50% (0.5), the "test" threshold might be as low as 25% (0.5 squared).
This means that even when the government is quite skeptical (say, only 25% sure the leader is honest), it might still be smart to run a small experiment. The leader, wanting to keep their future power, might behave well just to prove themselves, even if the government is still doubtful.

The authors show this is exactly true in a specific, simplified world where the benefits of spending are linear and the costs of taxes are quadratic (a standard economic shape). In this case, the "square" rule is a hard fact, not just a guess.

The Real-World Simulation: When Things Get Noisy

The real world isn't as clean as the math model. Sometimes you can't perfectly see if a leader is stealing; you only get noisy signals (rumors, partial audits). The authors ran massive computer simulations to see what happens in this messy reality.

They found some cool patterns:

The Reputation Trap: If a leader starts with a very bad reputation, the government might stop giving them money entirely. If no money is given, no new information is learned, and the reputation stays stuck at zero. It's a trap where the government gives up, and the leader never gets a chance to prove themselves.
The "High Need" Spark: Sometimes, a sudden crisis (like a natural disaster or a war) creates a huge need for spending. The simulations show that this high need can "wake up" the system. Even if the leader has a low reputation, the government might say, "We need this money spent now, so let's try a test!" This can jump-start the relationship and get the leader behaving better.
Patience Matters: If the government is very patient (they care a lot about the future), they are more willing to run these tests. The simulations showed that increasing the "patience" factor from 0.88 to 0.92 lowered the threshold for starting a test, meaning more leaders got a chance to prove themselves.

What the Paper Says and Doesn't Say

The authors are very careful. They prove the "square rule" for the simple, perfect-information world. For the complex, noisy world, they simulate the results. They don't claim to have solved the problem of corruption forever, but they show a clear mechanism: Reputation isn't just a feeling; it's a tool. By carefully choosing how much money to delegate, a government can force a leader to reveal their true nature.

They also rule out a few things. They show that you can't just "wait and see" if a leader is honest without giving them some power to act. And they show that simply having the legal ability to tax doesn't mean you will actually use that power; the fear of theft can freeze the system completely.

The Takeaway

This paper tells a story about trust and testing. It suggests that governments shouldn't just sit back and wait for a leader to become perfect. Instead, they can use small, calculated risks—giving a little bit of power to see if it's used well. Even if the odds look bad, the promise of future power can be enough to make a "bad" leader act like a "good" one. It's a bit like giving a kid a small allowance to see if they can manage it before handing over the family car keys. The math says that sometimes, you should give that small allowance even when you're pretty sure the kid might spend it on candy, because the chance they might learn to be responsible is worth the risk.

Technical Summary: Government Reputation and Fiscal Capacity

1. Problem Statement

The paper addresses a fundamental disconnect in fiscal governance: the administrative ability to collect revenue does not equate to the willingness to delegate that revenue to an executive. The central question is how an uninformed fiscal authority determines the scale of a tax-financed mandate when the executive's willingness to implement public spending is privately known.

The model distinguishes between administrative fiscal capacity (the feasible set of revenue determined by tax bases and enforcement) and effective fiscal capacity (the portion of that revenue the authority is willing to delegate). The core tension arises because an executive may be either "committed" (converting revenue to public goods) or "opportunistic" (diverting revenue). Since the authority cannot observe the executive's type or, in the dynamic setting, the implementation action directly, it must rely on reputation to govern delegation.

2. Methodology

The paper employs a combination of analytical game theory and computational equilibrium analysis.

Analytical Framework

Static Benchmark: A one-period model where the authority chooses a revenue mandate $r$ . The executive's type is fixed ( $H$ or $O$ ), and the authority maximizes expected welfare $pB(r) - K(r)$, where $p$ is the belief that the executive is committed.
Two-Period Reputation Game: A dynamic extension where a mandate serves a dual purpose: funding public goods and testing the executive. The opportunistic executive faces a trade-off between current diversion rents and the future value of maintaining a reputation that allows for future mandates.
Infinite-Horizon Model: A stationary Markov equilibrium model featuring:
- Noisy Monitoring: Implementation is not perfectly observed; signals depend on mandate size and audit intensity.
- Endogenous Auditing: The authority jointly chooses the mandate size and audit intensity.
- Spending-Need Shocks: Exogenous shocks to the marginal benefit of public goods.
- Recursive Incentives: The executive's decision to deliver or divert depends on the spread between future posteriors, which is endogenous to the aggregate delivery probability.

Computational Approach

The authors compute stationary equilibria for the infinite-horizon model using a finite state-action approximation.

Grid: 61 reputation nodes and a cloud of 2,123 fiscal actions (combinations of tax rates and audit intensities).
Equilibrium Selection: The model admits multiple stable implementation roots (due to the self-referential nature of reputation games). The authors select the lowest and highest stable roots and verify that the reported policies are robust to this selection.
Certification: The results are "hard-policy certified," meaning the authors verify that no deviation within the action cloud yields a profitable gain for the authority, and Bellman residuals are within machine precision.

3. Key Contributions and Results

Contribution 1: The Trial-Mandate Result (Static vs. Dynamic)

In the static benchmark, revenue is raised only if the reputation $p$ exceeds a cutoff $p_0$ . Below this threshold, the expected diversion cost outweighs the benefits, and the optimal mandate is zero.

In the dynamic two-period game, the authors prove that a positive trial mandate can be optimal even when $p < p_0$ .

Mechanism: A small mandate acts as an experiment. If the opportunistic executive delivers, they sacrifice current rents to preserve future fiscal access.
The Bound: The dynamic activation threshold $p_D$ is proven to be no greater than the square of the static cutoff: $p_D \leq p_0^2$ .
Exactness: In a linear-benefit, quadratic-cost economy, the bound is exact ( $p_D = p_0^2$ ) and is independent of the discount factor $\delta$ . The independence arises because patience scales both the size of the trial mandate and the future rents it secures, causing them to cancel out in the threshold calculation.

Contribution 2: Reputation Dynamics and Inactive Regions

In the infinite-horizon model with noisy monitoring, the authors establish two general properties:

Martingale Property: Public reputation is a martingale under any equilibrium policy. Policy can change the dispersion (risk) of future reputation but not its expected value.
Reputation Traps: If the mandate is zero, no information is generated (visibility $\lambda=0$ ). Consequently, reputation remains static. This creates a closed, inactive class of beliefs where delegation and learning cease simultaneously. Initial reputation can therefore have permanent consequences, determining whether the economy falls into a "trap" of zero fiscal capacity.

Contribution 3: Computational Equilibria and Policy Features

The computed stationary equilibria reveal specific structural features:

Audit Concentration: Audit intensity is concentrated at intermediate reputation levels. It is low at very low reputations (no mandate to monitor) and low at high reputations (the committed type is sufficiently likely).
Spending-Need Shocks: High spending needs lower the activation threshold for mandates. Crucially, a high-need shock increases effective fiscal capacity without changing the mean reputation. Instead, it widens the posterior distribution of reputation, increasing the probability mass in the active region while preserving the mean.
History Dependence: Because of the inactive regions, long-run outcomes are not ergodic averages but depend on the initial reputation. Starting with a low reputation increases the probability of ending in an inactive state.

4. Significance and Claims

The paper claims to provide a mechanism linking organizational authority to fiscal capacity, distinct from existing literature on institutional investment or tax morale.

Reversal of Control: Unlike literature where an agent's reputation constrains a principal, here an uninformed authority chooses the scale of resources delegated to an agent whose reliability is being learned. The scale of delegation determines the productive surplus, the rent from opportunism, and the stake in the relationship.
Distinction from Fiscal Capacity Literature: The paper argues that expanding legal/administrative capacity to tax is distinct from expanding the revenue that can be credibly placed under executive control. Reputation operates on the implementation side, determining how much collected revenue is delegated.
Policy Implications: The mechanism suggests that "monitoring institutions and limited trial mandates" can expand effective fiscal capacity even when administrative capacity remains fixed. The paper emphasizes that low reputation can shut down both fiscal activity and learning simultaneously, creating a self-reinforcing trap.

The authors are modest in their claims, noting that the analytical results (the $p_0^2$ bound) are continuum results, while the findings regarding history dependence and the shape of policy functions are established for the specific "certified computed cases" rather than proved as general theorems for all parameterizations. The paper explicitly avoids claiming that the numerical results represent a unique global equilibrium for the continuous state space, focusing instead on the robustness of the reported policies within the certified action cloud.