💰 quantitative finance

Reputation and institutional certification as complementary trust mechanisms in a single online market

This paper analyzes nearly one million eBay Pokemon card listings to demonstrate that reputation and institutional certification function as complementary trust mechanisms that partition the market into distinct equilibrium regimes, where self-grading becomes more effective as seller reputation increases while third-party grading remains a fixed-cost alternative.

Original authors: Yuta Kido, Yohsuke Ohtsubo

Published 2026-08-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yuta Kido, Yohsuke Ohtsubo

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

In the vast, invisible marketplace of the internet, where strangers trade goods without ever shaking hands, trust is the only currency that matters. When you cannot touch a product or see the seller, you face a fundamental problem: how do you know you are not being defrauded? For decades, economists have identified two main ways people solve this puzzle. The first is reputation, a digital record of past behavior that builds up over time, much like a neighborhood watch where everyone knows who has been honest and who has not. The second is institutional certification, a badge of approval from a third-party expert that guarantees quality, similar to a health inspector's stamp on a restaurant. While we know both methods work, it has remained unclear how they work together when they are both available at the same time. Do they compete, or do they help each other?

A team of researchers from the University of Tokyo decided to find out by looking at the world of Pokémon trading cards on eBay. This market is a perfect laboratory for studying trust because the value of a card depends entirely on its condition, which is hard to verify without an expert, and the risk of fraud is high. The researchers analyzed nearly one million listings for these cards, observing how sellers chose to signal the quality of their items. Sellers had three options: they could do nothing, they could grade the card themselves with a written description, or they could send the card to a professional service to be graded and certified by an outside institution. By tracking which sellers chose which option, what prices they set, and whether the items actually sold, the team mapped out a complex landscape of trust.

What they discovered was that reputation and certification do not simply compete; they occupy different territories based on the value of the item and the seller's history. The researchers found that the market naturally splits into distinct zones. When a card is worth a lot of money but the seller has a low reputation, the seller almost always relies on the third-party certification. The expensive, official grade acts as a shield, proving the card is real and in good condition, effectively replacing the need for a personal reputation. However, as a seller builds a strong history of honest transactions, the rules change. For high-reputation sellers, the need for expensive certification fades, even for valuable cards. Instead, these trusted sellers rely on their own word. Their reputation acts as a form of collateral; because they have so much to lose by lying, their self-graded claims become credible enough to command high prices without the need for a third party.

This dynamic creates a fascinating pattern where the two mechanisms function as complements across the entire market, even if they seem like substitutes in a single transaction. The study showed that a seller's reputation amplifies the price they can get for a self-graded card, but it does not help them get more money for a professionally graded one. In fact, for items that are already certified by an institution, having a high reputation does not add extra value to the price. This suggests that once a third party has vouched for the quality, the seller's personal history becomes redundant. The researchers built a theoretical model to explain this, showing that the cost of lying is different for each method. Lying on a self-graded card hurts a seller's future earnings by damaging their reputation, a penalty that grows larger as their reputation gets better. Lying on a certified card is impossible because the institution catches the fraud before the sale. Therefore, the two systems cover different ground: the institution handles the high-value, low-trust scenarios, while reputation handles the high-trust scenarios, allowing the market to function smoothly across a wide range of prices and seller histories.

The implications of this finding extend far beyond collectible cards. The researchers suggest that this structure helps explain how online platforms can sustain massive amounts of trade. Reputation systems are powerful, but they have a limit; a new seller with no history cannot easily build trust on their own, especially for expensive items. Institutional certification provides a way for these new sellers to enter the market and prove their worth. Once they have made enough honest sales to build a reputation, they can eventually stop paying for certification and rely on their own standing. This creates a developmental path for trust, where institutions help build the very reputation that eventually makes them unnecessary. The study confirms that these two systems are not rivals fighting for dominance, but rather partners that allow the market to expand, ensuring that both new and established sellers can find a way to be trusted.

Technical Summary: Reputation and Institutional Certification as Complementary Trust Mechanisms

Problem Statement
In markets characterized by information asymmetry, buyers face the risk of adverse selection, where high-quality goods are driven out by low-quality "lemons" because quality cannot be verified. Two primary mechanisms have traditionally been proposed to solve this: reputation (endogenous trust built through accumulated transaction history) and institutional certification (exogenous trust provided by third-party verification). While both are widely available on contemporary online platforms, their interaction remains poorly understood. Existing literature often treats them as substitutes or studies them in isolation. This paper investigates how these two mechanisms interact when simultaneously available to sellers, specifically examining whether they function as substitutes within a transaction or as complements across the market.

Methodology
The study employs a mixed-methods approach combining large-scale empirical analysis with a theoretical signaling game model.

Empirical Analysis:
- Data: The authors analyzed a dataset of nearly one million ( $N = 960,029$ ) fixed-price listings of Pokémon trading cards on eBay (US marketplace) collected between June and December 2025.
- Variables:
  - Signals: Sellers chose among three options: Third-Party (TP) grading (institutional certification, e.g., PSA), Self-grading (costless condition description), or None.
  - Reputation: A composite score derived via Principal Component Analysis (PCA) from eBay's feedback score and positive feedback percentage.
  - Value: The market reference price (from PriceCharting) adjusted for card grade.
- Modeling: A Hierarchical Bayesian Model (HBM) was fitted to jointly model three sequential stages of a transaction: (1) Signal selection (Multinomial Logit), (2) Price setting (Log price ratio relative to market value), and (3) Purchase outcome (Bernoulli Logit). The model included seller-level random effects to account for unobserved heterogeneity and selection bias. Posterior predictive simulations were used to estimate realized premiums on sold items.
Theoretical Modeling:
- A signaling game was formulated where a seller privately observes product quality ( $H$ or $L$ ) and chooses a signal.
- Cost Structures:
  - TP Grading: Incurs a fixed ex-ante cost ( $c_{TP}$ ) but is only feasible for high-quality sellers (low-quality sellers cannot pass).
  - Self-Grading: Costless ex-ante but carries an ex-post penalty proportional to the seller's reputation ( $\phi\rho$ ) if a low-quality seller is caught misrepresenting the item.
- Equilibrium Analysis: The authors derived Perfect Bayesian Equilibria (PBE) using the D1 criterion to refine off-path beliefs, partitioning the reputation-value space into distinct regimes.

Key Results

Regime Partitioning: The empirical data reveals that the reputation-value space is not uniform but is divided into distinct regions based on the dominant trust mechanism:
- Self-Dominant Region: High-reputation sellers with low-to-moderate value items predominantly use self-grading.
- TP-Dominant Region: Low-reputation sellers with high-value items predominantly use third-party grading.
- Boundary: The boundary separating these regions is positively sloped; as seller reputation increases, the item value threshold required to justify the cost of third-party grading also increases.
Interaction Effects on Pricing:
- Self-Grading: Reputation significantly amplifies the price premium of self-graded listings. High-reputation sellers can command higher prices for self-claimed conditions because their reputation acts as collateral against dishonesty.
- TP Grading: Reputation does not amplify the price premium of third-party graded items; in fact, a negative interaction was observed. High-reputation sellers using TP grading tend to accept lower conditional margins, likely because the institutional signal renders the seller's reputation informationally redundant for that specific transaction.
Theoretical Validation:
- The signaling game successfully reproduces the empirical regime map. It demonstrates that the "Self-separating equilibrium" (where high-quality sellers self-grade and low-quality sellers do not) exists only when the item value is below a reputation-collateral bound ( $\bar{V}(\rho)$ ).
- Above this bound, a "Semi-separating" regime emerges where self-claims are partially credible, and eventually, a "TP-separating" regime takes over where institutional certification becomes the preferred signal for high-quality sellers.
- The model predicts that the two mechanisms are substitutes within a single transaction (a seller chooses one or the other based on the regime) but complements across the market (together they enable trade across a wider range of seller reputations and item values than either could alone).
Reputation Dynamics:
- Theoretical analysis of reputation evolution suggests a "low-trust trap" for sellers without certification: if a seller starts with low reputation and cannot access certification, misrepresentation may drive reputation to zero.
- Institutional certification provides an "escape path," allowing low-reputation high-quality sellers to build a track record, eventually transitioning to reputation-based credibility.

Significance and Claims
The paper claims to resolve the "reputation-versus-institution" dichotomy by proposing a regime structure determined jointly by seller reputation and item value.

Complementarity vs. Substitution: The authors argue that while the mechanisms are functionally substitutable in pricing terms (one replaces the other in a specific transaction), they are complementary at the market level. They extend the scope of credible exchange, allowing low-reputation sellers to enter via certification and high-reputation sellers to operate efficiently via self-claims.
Mechanism of Trust: The study highlights a temporal distinction in signaling costs: reputation imposes an ex-post penalty for misrepresentation, while certification imposes an ex-ante monetary cost. This dual structure allows online platforms to scale trust mechanisms that were previously limited to small-group interactions.
Platform Governance: The findings suggest that platform design (e.g., feedback protection, monitoring intensity) directly influences the "reputation-collateral bound," thereby determining which sellers can rely on reputation alone versus those requiring institutional intervention.

Limitations
The authors acknowledge that the study is cross-sectional, limiting causal identification of the reputation dynamics and the specific channels behind the negative reputation-TP interaction. The empirical setting is specific to a collectibles market with mature third-party grading infrastructure, and the results may not generalize to markets with different cost structures (e.g., used cars without equivalent certification). Additionally, the sample is restricted to the US market, leaving cross-cultural variations unexplored.

Technical Summary: Reputation and Institutional Certification as Complementary Trust Mechanisms

More like this