Dynamics of Verification in Project Quality Selection

Core Problem and Model

• 💡 The model examines a principal selecting a project from an agent over time, where the agent has private information about project quality.
• 🎯 The agent is biased towards selection (wants any project chosen), while the principal values quality.
• 🔑 Costly verification is the primary tool for the principal to learn project quality, as transfers are unavailable.
• 🏢 A leading application is a company board (principal) evaluating acquisition proposals from a CEO (agent), with the board seeking shareholder value and the CEO seeking any equity event.

Optimal Mechanisms: Low Verification Cost

• ✅ For low verification costs and a long horizon, the optimal mechanism includes a deadline after which a project is selected regardless of quality.
• 📈 Before the deadline, the principal verifies agent claims of quality with an intermediate probability.
• 📉 A key finding is "decreasing skepticism": this verification probability declines over time as the deadline approaches.
• 📌 The selection rule is deterministic once high quality is verified, and the agent's indifference between truth-telling and lying drives the verification probability.

Optimal Mechanisms: High Verification Cost

• ⚠️ When verification costs are high, the optimal mechanism begins with an initial phase of screening by randomization.
• 🎲 During this initial phase, the principal randomizes the approval decision without incurring verification costs.
• 🔄 Verification is always backloaded, meaning it occurs in a later phase, even if its cost (C) exceeds the benefit of high quality (H).
• ⏳ The length of this initial randomization phase can be independent of discounting if the horizon is sufficiently long.

Dynamic Bias and Skepticism

• 🧠 The model identifies dynamic bias where the principal prefers to wait for high quality, while the agent wants immediate selection.
• 🔄 While decreasing skepticism is typical for long horizons, increasing skepticism can occur with significant static bias (low quality projects are very detrimental) and a short horizon.
• 💡 The intuition for backloaded verification is that agent benefits from verification accumulate over time, allowing for relaxation of earlier incentive constraints.