Identification in Auctions with Truncated Transaction Prices

arXiv:2510.04464 · econ.EM · Submitted 2025-10-06 · 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: "Identification in Auctions with Truncated Transaction Prices".

Warren: Here's the gist of the paper. Identification in auctions with truncated transaction prices establishes nonparametric identification results in first- and second-price auctions when transaction prices are truncated by a binding reserve price under a range of information structures When the number of potential bidders is fixed and known across all auctions, if only the transaction price is observed, the bidders’ private-value distribution is identified in second-price auctions but not in first-price auctions Identification in first-price auctions can be achieved if either the number of active bidders or the number of auctions with no sales is observed When the number of potential bidders varies across auctions and is unknown, the bidders’ private-value distribution is identified in first-price auctions but not in second-price auctions, provided that both the transaction price and the number of active bidders are observed.

Ruth: Pruning redundant reasoning tokens The paper addresses the identification challenge arising when observed bids or transaction prices are truncated by a binding reserve price or entry costs This truncation occurs because bidders with values below the reserve price may not submit bids, leading to only valid auctions being observed.

Warren: The authors focus on first- and second-price auctions, noting that descending and ascending auctions are strategically equivalent to these formats.

Ruth: How the training signal is built The key to identification with truncated transaction prices lies in the truncation threshold, which is related to the seller’s optimal screening level Without knowledge of this threshold, the unconditional distribution of bidders’ values cannot be recovered from the distribution of transaction prices conditional on a valid auction.

Title and authors: Ruth: The paper "Identification in Auctions with Truncated Transaction Prices" is essentially about getting a handle on bidder value distributions when the transaction prices are truncated by things like a reserve price. It’s not just about standard identification; it’s about making sure the identification still works even when we only see those valid transactions.

Warren: And they point out that this truncation is a real problem because bidders with values below that cutoff simply don't participate, which means our observed data isn't representative of everyone who *could* have bid.

Iris: I think the authors are laying out a framework where they examine how different information structures—like whether you know the total number of potential bidders or not—affect whether we can successfully identify that underlying distribution.

Ruth: They specifically focus on first-price and second-price auctions because those are the ones we usually use, but they also note that these results apply to descending and ascending auctions, which is a nice bit of consistency for the real world.

Warren: So what this means for us as traders or market observers is that if you only see prices above some threshold, you can still make inferences about what people value behind those observed prices, provided you have enough context.

The paper's summary: Iris: To summarize the main thrust of "Identification in Auctions with Truncated Transaction Prices," the authors are showing that nonparametric identification results *can* be achieved in both first- and second-price auctions when the transaction prices are truncated by a binding reserve price, depending on what we observe.

Ruth: That’s the big claim: that you can recover the bidders' private value distribution even with this truncation, but only if you have specific auxiliary information about the auction setup.

Warren: The paper uses a quantile-based approach where they define a bidder’s type by an alpha, which is basically where their value falls on a unit interval from zero to one. This lets them simplify the notation and focus on how that cutoff threshold relates to the observed transaction prices.

Iris: The key mechanism they highlight is that without knowing this truncation threshold—which is tied to the seller’s optimal screening level—you just can't recover the distribution from the transaction price data alone, which is a pretty strong point of their analysis.

Ruth: And they build on previous work by showing how observing specific things, like the number of invalid auctions or active bidders, unlocks those identification results in different auction settings.

The paper's improvements: Warren: Now let's talk about what the paper suggests improving for future work. One big point they make is that if you have a fixed number of potential bidders known across all auctions, the optimal screening level, which is that truncation threshold, doesn't actually depend on how many bidders there are; it stays the same.

Iris: That’s a neat simplification because it makes things much more tractable when N is fixed and known. They also show that in second-price auctions specifically, observing the number of invalid auctions helps facilitate identification of alpha star and V of alpha for values greater than or equal to alpha star.

Ruth: For first-price auctions, they suggest that if you observe either the number of active bidders or the number of invalid auctions, you can achieve point identification for alpha star and V of alpha for values above it. That’s a practical way to get those results in a real auction setting.

Warren: They also extend these ideas to entry costs, showing similar identification results hold there too, though the entry cost model introduces some complexity because the threshold depends on how many potential bidders there are.

Conclusion: Ruth: So wrapping up on "Identification in Auctions with Truncated Transaction Prices," the main implication is that we don't have to give up on identifying bidder values just because the transaction prices look weird due to a reserve price; we just need to be smart about what other data we collect.

Warren: The paper shows that for second-price auctions, if you know the number of potential bidders and observe the transaction prices, you can get a point identification of alpha star and V of alpha for values above it. But they make a key distinction: if the truncation is due to entry costs instead of a reserve price, things get harder because the threshold changes with N.

Iris: I agree that knowing whether it's a reserve price or an entry cost makes a big difference in what identification we can achieve, and it shows that observing invalid auctions gives us more leverage in first-price settings.

Ruth: Overall, the paper gives us constructive results for estimation methodologies by showing exactly when and how to get point identification under these truncated conditions. It’s a solid piece of work on identifying auction models with binding reserve prices and entry costs.

National University of Singapore

econ.EM

Submitted: 2025-10-06

Updated: 2026-11-10

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 85/100

The gist: Identification in auctions with truncated transaction prices establishes nonparametric identification results in first- and second-price auctions when transaction prices are truncated by a binding…

Key concepts

Truncated Transaction Prices
This occurs when a binding reserve price prevents bidders with values below it from submitting bids. Only valid auctions are observed, meaning the observed transaction prices are not representative of all potential bids. The paper focuses on how this truncation affects identification in different auction types.
Optimal Screening Level (Truncation Threshold)
This threshold is related to the seller's optimal strategy for screening bidders based on their values. Recovering this threshold from observed transaction prices is key to identifying the underlying bidder distribution. Without knowing it, the true distribution cannot be recovered.
Second-Price vs. First-Price Auctions
The paper analyzes identification results differently for these two auction formats. In second-price auctions, the transaction price exhibits a mass point at the reserve price, aiding identification. In contrast, first-price auctions have continuous transaction prices that offer less direct information about the truncation threshold.
Information Structures (N and Active Bidders)
Identification success varies based on what information is available. If the total number of bidders (N) is fixed and known, or if both N and the number of active bidders are observed, identification becomes easier in certain scenarios. When N varies or is unknown, the problem becomes more complex.

Terminology

Summary

Identification in auctions with truncated transaction prices establishes nonparametric identification results in first- and second-price auctions when transaction prices are truncated by a binding reserve price under a range of information structures When the number of potential bidders is fixed and known across all auctions, if only the transaction price is observed, the bidders’ private-value distribution is identified in second-price auctions but not in first-price auctions Identification in first-price auctions can be achieved if either the number of active bidders or the number of auctions with no sales is observed When the number of potential bidders varies across auctions and is unknown, the bidders’ private-value distribution is identified in first-price auctions but not in second-price auctions, provided that both the transaction price and the number of active bidders are observed. p. 0

Pruning redundant reasoning tokens

The paper addresses the identification challenge arising when observed bids or transaction prices are truncated by a binding reserve price or entry costs This truncation occurs because bidders with values below the reserve price may not submit bids, leading to only valid auctions being observed. The authors focus on first- and second-price auctions, noting that descending and ascending auctions are strategically equivalent to these formats. p. 2

How the training signal is built

The key to identification with truncated transaction prices lies in the truncation threshold, which is related to the seller’s optimal screening level Without knowledge of this threshold, the unconditional distribution of bidders’ values cannot be recovered from the distribution of transaction prices conditional on a valid auction. The authors use a quantile-based approach to simplify notation, where the bidder’s type is defined by alpha i = F(v i) and is distributed as Unif[0, 1]. p. 6

Where the gains come from

The paper derives several identification results across different information structures and auction formats. For instance, in auctions with a common and known N, the optimal screening level alpha star remains invariant to N. p. 2 Identification in second-price auctions is facilitated by observing the number of invalid auctions Identification in first-price auctions can be achieved if either the number of active bidders or the number of invalid auctions are observed. p. 2

Identification in Auctions with Entry Costs

The analysis extends to auctions with entry costs, showing that similar—though not identical—identification results hold. In the entry model, the bidder’s equilibrium bidding threshold depends on the number of potential bidders N, which makes identification more challenging when N varies. p. 3

Identification in Auctions with Varying Bidders

When N varies across auctions and is known by the econometrician, alpha star and V(alpha) for alpha >= alpha star are point identified in both first- and second-price auctions if only transaction prices strictly above the reserve price are observed. This result relies on comparing quantile functions derived from different known N values to identify the threshold.

Identification in Auctions with Unknown Bidders

When N is random, finite, and unknown, observing both transaction prices and the number of active bidders allows for point identification of alpha star and V(alpha) for alpha >= alpha star in first-price auctions. However, this identification is not achieved in second-price auctions under these conditions. p. 3

Identification with Invalid Auctions

If invalid auctions are also observed, they provide crucial information for identification in first-price auctions. Proposition 2 shows that if Tl and L invalid are observed, alpha star and V(alpha) for alpha >= alpha star can be identified in first-price auctions. p. 2

Identification with Entry Costs

In the entry model, observing both transaction prices and the number of active bidders allows for point identification of N, alpha star, F, and V(alpha) for alpha >= alpha star in both first- and second-price auctions. This is demonstrated through Proposition 8.

Conclusion

The key to identification with truncated transaction prices hinges on knowing the truncation threshold When N varies and is known, the bidder’s private value distribution can be point-identified from transaction prices in both first- and second-price auctions if truncation arises from a reserve price. The paper concludes that for second-price auctions, the distribution is not identified if truncation is due to reserve price, but it is at least set-identified if truncation is due to entry costs. p. 0

The key to identification with truncated transaction prices hinges on knowing the truncation threshold When the number of potential bidders N is fixed, and both N and the transaction prices are observed, the distribution of bidders’ private values is identified in second-price auctions but not in first-price auctions regardless of whether truncation arises from binding reserve prices or entry costs. The main distinction between the identification results for first- and second-price auctions lies in the behavior of the transaction price. p. 2 In first-price auctions, the transaction price is continuous and provides no direct information about the bidder’s truncation threshold In contrast, in second-price auctions, the transaction price exhibit a mass point at the reserve price, which helps reveal the truncation threshold and thus facilitates identification. p. 2

The authors establish that for second-price auctions, a two-step approach is required: one must first recover the seller’s screening level before identifying the distribution of bidders’ private values. The identification strategy differs between auction formats

The paper contributes to the existing literature by extending results to auctions with a binding reserve price and analyzing auctions with entry costs. The positive identification results are constructive and should pave the way to the development of corresponding estimation methodologies. p. 3 p. 3 The key insight is that for second-price auctions, the support of the transaction price varies with N under entry costs, but remains the same in the binding reserve case. p. 1

While the presence of a binding reserve price is common in real-world auctions, research on the resulting problem of data truncation is surprisingly limited This paper addresses this gap by establishing identification results for both first- and second-price auctions when the observed transaction price data are truncated by a binding reserve price. Moreover, it extends these results to auctions with entry costs, where only auctions with at least one entrant are observed. p. 1 The identification results are summarized in the following table which shows that when N is fixed and both N and transaction prices are observed, the distribution of bidders’ private values is identified in second-price auctions but not in first-price auctions regardless of whether truncation arises from binding reserve prices or entry costs. p. 0 The key to identification with truncated transaction prices hinges on knowing the truncation threshold. p. 2

The key to identification with truncated transaction prices hinges on knowing the truncation threshold When N varies and is known, the bidder’s private value distribution can be point-identified from transaction prices in both first- and second-price auctions if truncation arises from a reserve price. However, this paper only establishes set identification for the bidder’s private value distribution if truncation arises from entry costs. p. 4 p. 4 The difficulty in identifying the entry cost model is due to the dependence of the truncation threshold on N. p. 3 p. 3 When N varies and is unknown, the identification problem becomes significantly more complex. p. 0

In first-price auctions, the bidder’s private value distribution can be identified from Tl and Nobs l regardless of whether truncation arises from binding reserve prices or entry costs In second-price auctions, if Tl and Nobs l are observed, the bidder’s private value distribution is not identified if truncation is due to reserve price, but it is at least set-identified if truncation is due to entry costs. The difference in identification results for second-price auctions stems from the fact that the support of the transaction price varies with N under entry costs, but remains the same in the binding reserve case. p. 2

The paper contributes to identifying auction models with a binding reserve price and entry costs. The positive identification results are constructive and should pave the way to the development of corresponding estimation methodologies. The key insight is that for second-price auctions, observing invalid auctions can facilitate identification of alpha star and V(alpha). p. 2

The key to identification with truncated transaction prices hinges on knowing the truncation threshold When N varies and is known, the bidder’s private value distribution can be point-identified from transaction prices in both first- and second-price auctions if truncation arises from a reserve price. However, this paper only establishes set identification for the bidder’s private value distribution if truncation arises from entry costs. The difficulty in identifying the entry cost model is due to the dependence of the truncation threshold on N. When N varies and is unknown, the identification problem becomes significantly more complex

The paper contributes to identifying auction models with a binding reserve price and entry costs. The positive identification results are constructive and should pave the way to the development of corresponding estimation methodologies The key insight is that for second-price auctions, observing invalid auctions can facilitate identification of alpha star and V(alpha). p. 2

The paper contributes to identifying auction models with a binding reserve price and entry costs The positive identification results are constructive and should pave the way to the development of corresponding estimation methodologies The key insight is that for second-price auctions, observing invalid auctions can facilitate identification of alpha star and V(alpha).

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. Improved systems can perform nonparametric identification of bidder private value distributions in first-price auctions when only transaction prices are observed, provided that either the number of active bidders or the number of auctions with no sales is known.

  2. Systems can recover a seller's optimal screening level, which is the truncation threshold, from observed transaction prices and the structure of second-price auctions through a mass point at the reserve price.

  3. For first-price auctions with entry costs, systems can be set identified for bidder private values when both transaction prices and the number of active bidders are observed, even if the total number of potential bidders is unknown.

Related papers