📈 economics

Delegated Monitoring in Public-Private Sector Credit Programs: Underinvestment, Overinvestment, and the Design of Subsidized Lending

This paper develops a mechanism-design model demonstrating that delegated monitoring in public-private credit programs can generate either underinvestment or overinvestment distortions depending on monitoring curvature and subsidy intensity, a theoretical framework validated through descriptive analysis of SBA SBIC data and calibrated simulations.

Original authors: G. Charles-Cadogan

Published 2026-08-05

📖 7 min read🧠 Deep dive

CC BY 4.0

Original authors: G. Charles-Cadogan

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

The Great Credit Game: When Good Intentions Go Wrong

Imagine a world where getting a loan is like trying to buy a ticket to a secret, exclusive club. The club is run by a strict bouncer (the bank) who is terrified of letting in troublemakers who will never pay back their entry fee. Because the bouncer can't read minds, he often turns away good, hardworking people just to be safe. This is a classic problem in economics called credit rationing: when good borrowers can't get money because the lender is too scared of the bad ones.

To fix this, governments often step in. They say, "Don't worry, I'll help you find the good people!" They create special programs where they give money to private investors (like venture capitalists) to lend to small businesses. The government hopes these private investors will act as delegated monitors—super-sleuths who can tell the difference between a diamond and a piece of glass. The government provides a safety net (subsidies) to make the investors brave enough to lend. But here's the twist: what if the private investors, trying to please the government and make a profit, accidentally start playing the game wrong? They might either ignore the people they were supposed to help, or they might give too much money to the wrong people. This paper explores exactly how that happens.

The Paper's Big Discovery: The Double-Edged Sword of Subsidies

This paper, written by G. Charles-Cadogan, dives into the messy reality of these public-private partnerships. It asks a simple but tricky question: When the government pays private investors to lend money to small businesses, does it actually help, or does it create new problems?

The author builds a mathematical model to simulate this scenario, treating the government as a "Principal" (the boss) and the private investors as "Agents" (the workers). The goal is to see how the agents behave when they have to screen applicants, decide who gets a loan, and watch over them to make sure they pay it back.

The paper finds that the system is a bit of a paradox. It can fail in two opposite ways, depending on how the rules are set up:

The "Underinvestment" Trap (Too Little Help):
Imagine the private investors are trying to be super careful. Because they can't perfectly tell who is a "good" borrower and who is a "bad" one, they get scared. To avoid losing money, they might raise the "interest rate" (the cost of borrowing) or just say "no" to everyone who looks even slightly risky.
- The Result: The very people the government wanted to help—the small, struggling, but honest businesses—get locked out. The investors decide the risk isn't worth it, and the money stays on the shelf. This is called underinvestment. It's like a bouncer who is so afraid of letting in a troublemaker that he refuses to let in the entire neighborhood, including the good kids.
The "Overinvestment" Trap (Too Much Help):
Now, imagine the government tries to fix the first problem by giving the investors a huge subsidy (free money) to lend to risky businesses. The investors, seeing this free money, might get too excited. They start lending to high-risk businesses that they normally wouldn't touch because the government is covering their losses.
- The Result: The risky businesses get way more money than they should, and the investors stop checking if the businesses are actually doing well. This is called overinvestment. It's like giving a teenager an unlimited credit card because "it's for their education," only for them to buy a jet ski and never pay it back. The paper suggests that if the subsidies are too strong, they reduce the investors' motivation to be careful, leading to a flood of bad loans.

The Secret Sauce: Curvature and Classification

The paper argues that which of these two disasters happens depends on two main things:

The "Curvature" of Monitoring Costs: How much harder does it get to watch a borrower as they get riskier? If it gets really expensive to monitor a risky firm, the investors will stop lending to them (Underinvestment). If the cost doesn't rise fast enough, they might lend too much (Overinvestment).
The Intensity of the Subsidy: How much free money is the government giving? A little helps, but too much makes the investors less diligent.

The author also looks at what happens over time. Imagine a loan isn't just one big check, but a series of payments. If a business does well in the first year, the investor might give them more money later. If they do poorly, the investor stops the flow. The paper uses a sequential model (a step-by-step game) to show that the best way to avoid mistakes is to update your beliefs constantly. If a business owner has a specific trait (like a great credit score or a solid business plan) that helps the investor guess correctly whether the business will succeed, the investor should spend more money checking that specific trait. If a trait is expensive to check but doesn't help much, they should ignore it.

What the Data Says (And What It Doesn't)

The paper doesn't just sit in a math classroom; it tries to see if this matches the real world. The author looks at data from the Small Business Administration (SBIC) in the United States from 2018 to 2025.

The Real-World Clues: The data shows that SBIC money isn't spread out evenly across the country. In 2025, the top five states received 42.8% of all the financing. The distribution is very concentrated, with a Gini coefficient of 0.61 (a number that measures inequality; 0 is perfect equality, 1 is total inequality). This suggests that private investors are indeed picking and choosing where to put the money, likely avoiding places where it's hard to find good deals or hard to monitor the borrowers.
The Simulation: Since the public data doesn't show the details of every single loan application (like who applied, what their risk score was, or exactly how much the investor watched them), the author couldn't prove the theory with a "smoking gun." Instead, they built a computer simulation with 60,000 fake loan applications across 36 states and 12 years.
The Simulation Results: When the computer ran the model, it successfully recreated the patterns the theory predicted. It showed that when monitoring is hard, loans get rationed. When subsidies are high, risky firms get too much money. This suggests the theory is "empirically recoverable"—meaning, if we had perfect data, we would likely see these exact patterns in real life.

The Bottom Line

The paper concludes that having good intentions isn't enough. Just because the government wants to help small businesses doesn't mean the system will work. If the rules aren't designed perfectly, the system can either starve the good businesses of cash or feed the bad businesses until they burst.

The key takeaway is that design matters. Governments need to think carefully about how much they subsidize and how they pay for monitoring. They need to make sure the private investors still have a reason to be careful. The paper suggests that the best way to do this is to focus monitoring resources on the specific details that actually help predict success, rather than wasting time on things that don't matter.

In short, the paper warns us that in the complex game of public-private lending, you can't just throw money at the problem. You have to understand the game's rules, or you might end up with a club that's either empty or full of people who never pay their dues.

Technical Summary: Delegated Monitoring in Public–Private Credit Programs

Problem Statement
This paper addresses the contracting problem inherent in public–private partnerships (PPPs) designed to expand access to credit for small firms. While the public sector aims to relax credit rationing and support socially valuable firms, it delegates the screening, allocation, and monitoring of subsidized loans to private equity or venture-capital intermediaries. The core tension lies in the misalignment of incentives: the public sector seeks to maximize social surplus, while delegated monitors retain private risk, effort, and screening incentives. The paper investigates whether benevolent public intent can be efficiently implemented when information is asymmetric, monitoring is costly, and subsidies alter the risk-return profile for both borrowers and intermediaries.

Methodology
The paper employs a mechanism-design framework to model the interaction between a risk-neutral public principal and risk-averse delegated monitors (venture capitalists/portfolio managers).

Static Model (Sections 2–4):
- Setup: Firms possess private information regarding their type ( $\theta$ ), which determines their productivity and risk. The delegated monitor observes noisy signals and allocates loan sizes ( $x$ ) and transfers ( $y$ ).
- Constraints: The model incorporates Individual Rationality (IR) and Incentive Compatibility (IC) constraints. It distinguishes between symmetric information (First Best) and asymmetric information (Second Best).
- Cost Structure: Monitoring costs are modeled as a convex function of the loan portfolio. The curvature of this cost function is critical; it determines the feasible spread between loans offered to high-risk and low-risk firms.
- Equilibrium Concepts: The analysis derives separating and semi-separating equilibria. It utilizes a "participation-composition condition" to determine when a program is socially beneficial (i.e., when the mass of low-risk firms offsets the costs induced by high-risk participation).
Sequential Extension (Section 5):
- The model is extended to a two-period dynamic setting involving credit scoring and Bayesian updating.
- At $t=0$ , firms are classified based on observable characteristics ( $c$ ). At $t=1$ , performance signals (e.g., sales, profits) are revealed, allowing the lender to update posterior probabilities of firm type.
- The welfare loss from misclassification is analyzed as a function of error probabilities and the marginal cost of monitoring specific borrower characteristics.
Empirical Component (Section 6):
- Due to the unavailability of granular borrower-level data, the paper provides descriptive state-year evidence using SBA SBIC reports (2018–2025) to validate allocation patterns.
- A calibrated simulation is constructed to demonstrate that the model's comparative statics are empirically recoverable if granular data (application, monitoring, and performance levels) were available.

Key Contributions
The paper makes three primary contributions to the literature on credit rationing, delegated monitoring, and public finance:

Integration of Rationing and Delegation: It connects the classic credit-rationing theory (Stiglitz & Weiss, 1981) with delegated monitoring in PPPs, showing how public subsidies interact with private screening incentives.
Dual Distortion Mechanism: It demonstrates that the same delegated intermediation structure can generate both Stiglitz–Weiss underinvestment and De Meza–Webb overinvestment distortions.
- Underinvestment occurs when imperfect screening and interest-rate sorting exclude creditworthy target firms.
- Overinvestment occurs when subsidies weaken monitoring and repayment incentives, causing high-risk firms to obtain excessive credit.
- The operative distortion is determined by the monitoring cost curvature and subsidy intensity.
Sequential Welfare Optimization: It extends the static mechanism to a sequential contract, showing that welfare loss from misclassification is minimized when monitoring resources are allocated according to the marginal effect of borrower characteristics on posterior risk classification.

Key Results

Interest-Rate Sorting and Underinvestment: In the presence of asymmetric information, delegated monitors maximize expected surplus by offering contracts that sort firms by risk. If screening is imperfect, interest-rate sorting may exclude target firms, leading to underinvestment.
Overinvestment via Subsidies: If the government provides subsidies to "level the playing field" for high-risk firms, it can induce an overinvestment problem. High-risk firms become privately attractive to venture capitalists even when the allocation is socially inefficient, as subsidies mask the true risk.
Feasible Loan Spread: The model predicts that the feasible spread between loan contracts for high-risk and low-risk firms narrows as expected risk increases. This is a geometric consequence of operating on a steeper monitoring-cost schedule; higher monitoring intensity contracts the range of viable loans.
Optimal Monitoring Rule: In the sequential extension, the paper derives a condition for minimizing welfare loss: the marginal cost of verifying a borrower characteristic should be weighted by that characteristic's marginal effect on the posterior probability of misclassification. Characteristics that do not significantly improve classification should not be heavily monitored.
Empirical Patterns: The descriptive analysis of SBIC data reveals substantial geographic concentration in capital allocation (e.g., the top five states received 42.8% of financing in FY2025). This pattern is consistent with delegated allocation, where intermediaries deploy capital based on deal flow and monitoring technology rather than population proportion.

Significance and Claims
The paper claims that benevolent public intent does not by itself imply an efficient access-to-credit mechanism. The efficiency of PPPs depends critically on the design of the delegation contract and the management of agency costs.

Design Implications: Effective PPP design requires explicit attention to delegated-monitor agency costs, risk-based subsidy rules, and audit technology. Programs should not be evaluated solely by loan volume or the number of firms reached, but by whether they induce additional lending to target firms without weakening screening and repayment incentives.
Modesty of Empirical Claims: The author explicitly states that the empirical component serves as design validation rather than a causal test. The aggregate data supports the model's allocation patterns (geographic concentration, allocation gaps) but cannot identify the specific mechanisms of borrower sorting or welfare loss without granular, firm-level data. The simulation serves as a "backstop" to show that the model's sharper predictions are theoretically recoverable in data that is currently unavailable.
Scope: The paper focuses on the feasibility and comparative statics of the mechanism under specific regularity conditions (single-crossing, concavity, convex agency costs). It does not claim to characterize every perfect Bayesian equilibrium of an unrestricted dynamic game, nor does it model the political economy of subsidy choice.

In summary, the paper provides a theoretical framework explaining why public-private credit programs may fail to reach their intended targets due to the interplay of imperfect screening, interest-rate sorting, and subsidy-induced moral hazard, while offering a normative rule for optimizing monitoring resources in sequential lending environments.