Pith. sign in

REVIEW 3 major objections 7 minor 45 references

Enhancing the Merger Simulation Toolkit with ML/AI

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A data-driven supply model beats Bertrand-Nash at predicting post-merger airline prices.

desk verdict VMM-based flexible supply estimation is a solid methodological contribution with convincing Monte Carlo evidence, but the empirical merger prediction claim rests on an implausible exclusion restriction. read the letter →

arxiv 2506.05225 v1 pith:GSX3Z4QD submitted 2025-06-05 econ.EM

Harold D. Chiang , Jack Collison , Lorenzo Magnolfi , Christopher Sullivan This is my paper
classification econ.EM MSC 62P2091B24
keywords mergersimulationnonparametricsupplyvariationalmethodofmomentsmachinelearninginstrumentalvariablesairlineindustrypost-mergerpricepredictionBertrand-Nashconduct
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Standard merger simulation predicts post-merger prices by first assuming a model of firm conduct, usually Bertrand-Nash price competition with constant marginal costs; when that assumption is wrong, predictions can be badly off. This paper tries to replace the assumed conduct with a flexible supply function, estimated from market data, that nests Bertrand-Nash and many other conduct models. The paper claims that when rich data are available, this flexible model can be identified with instrumental variables, estimated with machine learning, and used to simulate post-merger prices without giving up the equilibrium structure. In the American-US Airways merger, the flexible model's passenger-weighted post-merger price mean squared error is 66.93 versus 365.71 for Bertrand-Nash, and its median predicted price increase of 2.05% is close to the difference-in-differences estimate of 2.92%.

What carries the argument

The central object is the flexible supply function $h(s_t, D_t, w_t; H_t)$, which bundles markups and marginal costs into one unrestricted function of market shares, the matrix of demand derivatives, observed cost shifters, and the ownership matrix; it nests Bertrand-Nash and many other conduct models. The argument is carried by the moment condition $E[\omega_{jt} | z_{jt}, w_{jt}] = 0$, which identifies $h$ under an exclusion restriction and a completeness condition, and by estimating $h$ with the Variational Method of Moments as a neural-network min-max problem. Post-merger prices are then solved from the fixed point $\tilde{p}_t = h(s(\tilde{p}_t), D(\tilde{p}_t), w_t; \tilde{H}_t) + \hat{\omega}_t$.

What would settle it

Estimate a cost equation with marginal costs (or Bertrand-implied costs) regressed on the excluded instruments, the share of nonstop flights and squared average distance; any nonzero coefficient would violate the exclusion restriction and overturn identification of the supply function.

Watch

Extended reading notes

Core claim

The paper's central claim is that the price equation $p_{jt} = h(s_t, D_t, w_t; H_t) + \omega_{jt}$, with $h$ an unrestricted function of market shares, demand derivatives, and cost shifters under a given ownership structure, can be identified and estimated, and that counterfactual post-merger prices obtained from it are more accurate than those from the standard toolkit. The identification argument adapts nonparametric instrumental-variable logic: demand shifters excluded from the cost function act as instruments for endogenous shares and demand derivatives, and a completeness condition rules out alternative functions. Estimation uses the Variational Method of Moments, a min-max procedure that replaces classical nonparametric instrumental variables with neural networks and thereby avoids the curse of dimensionality. The paper's headline evidence is the American-US Airways retrospective: passenger-weighted MSE 66.93 for the flexible model versus 365.71 for Bertrand-Nash, and a median predicted price increase of 2.05% versus a difference-in-differences benchmark of 2.92%. Monte Carlo simulations also show the flexible model outperforming misspecified standard models and nearly matching the true model's post-merger price predictions.

Load-bearing premise

The load-bearing premise is the exclusion restriction in the empirical application: the instruments 'share of nonstop flights' and 'squared average distance' are assumed to affect demand but not marginal costs; if they move costs, the moment condition $E[\omega|z,w]=0$ fails and the supply function is not identified.

Editorial extensions

If this is right

  • If the flexible model is right, merger review need not commit to a single conduct assumption such as Bertrand-Nash when the data are rich enough to estimate conduct.
  • Standard simulations that assume Bertrand-Nash can materially overstate price increases in markets that approach monopoly after a merger, because the misspecified markup is the main source of price pressure.
  • The method yields confidence intervals on counterfactual post-merger prices, so agencies can report uncertainty around predictions rather than a single point.
  • The same estimated supply function can be used to simulate cost pass-through and to quantify welfare and profit effects of a merger, not just prices.
  • The approach is portable to other industries with standard price and quantity data and with variation in market structure.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An editorial extension: the estimated $h$ could be compared with Bertrand-Nash, Cournot, or profit-weight markups to test which conduct model fits the data, turning the predictor into a conduct diagnostic.
  • Editorial caution: the application's exclusion restriction, that the share of nonstop flights and squared average distance do not shift marginal costs, is not directly tested in the paper; adding those variables to the cost side in a sensitivity analysis would show how much of the 66.93 versus 365.71 gap depends on that assumption.
  • An out-of-sample test the paper does not run: apply the estimator to a later merger with known outcomes, such as Alaska-Virgin America, and compare its post-merger price predictions with realized prices.
  • If conduct itself changes because of the merger, the assumption that the same supply function applies before and after the merger would fail; detecting such a shift would require post-merger supply data or a model that lets the conduct parameters vary.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. This paper develops a flexible nonparametric model of supply for merger simulation, specified as p_jt = h(s_t, D_t, w_t; H_t) + ω_jt, and estimates it with the Variational Method of Moments (VMM) of Bennett and Kallus (2023). The supply function nests Bertrand, Cournot, and other conduct models, and h is identified under a completeness condition and an exclusion restriction that demand shifters are excluded from costs, following Berry and Haile (2014) (Theorem 3.1). The paper also proposes an inference procedure for post-merger price predictions based on the numerical delta method, Holm's step-down procedure, and a permutation construction (Algorithm 1). Monte Carlo experiments with logit demand under Bertrand and profit-weight conduct show VMM outperforming misspecified Bertrand, monopoly, and perfect-competition models in out-of-sample fit, merger-simulation MSE, and pass-through recovery. In an application to the American Airlines–US Airways merger (Section 6), the flexible model yields a post-merger price prediction MSE of 66.93 versus 365.71 for the standard Bertrand toolkit, with a median predicted price increase of 2.05% versus a difference-in-differences estimate of 2.92%.

Significance. The paper is a genuine methodological contribution at the intersection of IO and machine learning. If the statistical claims hold, it substantially broadens the supply-side assumptions that merger simulation can accommodate while retaining an explicit equilibrium structure. The Monte Carlo evidence is the strongest part of the paper: across sample sizes T=100 to 10,000 and two conduct DGPs, VMM's prediction MSE is within a small factor of the true model's and an order of magnitude below the misspecified benchmarks, with the evaluation done on hold-out markets and on strictly out-of-sample triopoly mergers (Tables A3–A4). The pass-through exercise (Table 5) is a valuable model-validity check that speaks to the 'black box' criticism. The authors are also appropriately explicit about the scope of the method in Section 7.

major comments (3)
  1. [Section 6.2, Assumption 5] The exclusion restriction in the application is asserted rather than tested, and as stated it cannot be satisfied by the instrument list provided. The paper reports that 'squared average distance in thousands of miles' is used as an excluded instrument while average distance appears in w; since h is nonparametric in w, a deterministic function of an included regressor cannot supply the excluded variation that Assumption 5 requires, so the only genuinely excluded own-product variation comes from the share of nonstop flights. That variable plausibly enters marginal costs through fuel, crew, aircraft utilization, and airport charges, in which case E[ω|z,w]=0 fails, Theorem 3.1 does not identify h as the supply relation, and the Section 6.3 comparison between VMM and Bertrand is not structural. The Monte Carlo cannot validate this assumption because there demand characteristics and cost shifters are drawn independently and ω is correlated only with ξ, so the exclusion holds by construction. The authors should report an overidentification test, re-estimate the supply function with the nonstop share removed from z, and/or explicitly re-frame the Section 6.3 claims as conditional on this untested restriction.
  2. [Section 3.1, Assumption 3; Section 6.2; Equation (7)] The identification and estimation theory treat the demand derivative matrix D_t as known (Assumption 3), but in the empirical section D_t is computed from an estimated nested logit demand system. Theorem 3.1 and the VMM asymptotics invoked in Section 4 do not account for the sampling error in D_t, and the counterfactual Equation (7) uses the estimated demand mapping s(·) together with the estimated h. The Monte Carlo either supplies D exactly or learns it from a correctly specified logit DGP, so it cannot speak to the size of the plug-in bias in the Section 6.3 predictions. The authors should either extend the theory to cover an estimated demand first step or add a Monte Carlo design that mirrors the empirical pipeline, for example by estimating nested logit demand on training markets and using the estimated D in supply estimation and counterfactuals.
  3. [Section 5.4, Table 10, Algorithm 1] The paper claims the procedure 'provides valid statistical inference,' but the evidence in Table 10 points in the opposite direction. In the profit-weight rows, the reported interval [11.991, 13.822] excludes the true value ψ=17.321 at N=253, and [15.261, 15.847] excludes ψ=17.375 at N=2,579; only the Bertrand rows cover the truth. The text presents these as 'quantifiable and tight confidence intervals' without noting the systematic under-coverage in the profit-weight DGP. Additionally, the simultaneous interval in Algorithm 1, constructed as the union of bounds over all d! permutations of the Holm critical values, has no coverage proof and no simulation validation; Figure C4 reports widths but not coverage. A coverage simulation across the two DGPs and sample sizes, with and without the permutation procedure, is needed before the inference claim can be maintained.
minor comments (7)
  1. [Section 3.3, Equation (9)] The sign in Equation (9) is reversed: since ω_jt = p_jt − h_j(·), we have E[ω_jt|z_jt,w_jt] = E[p_jt|z_jt,w_jt] − E[h_j(·)|z_jt,w_jt]. The subsequent conclusion in Equation (10) is unaffected, so the proof goes through, but the displayed equation is incorrect as written.
  2. [Section 5 (Data Generation)] The instrument vector z used in the VMM estimation of the Monte Carlo is never specified. Since identification (Assumption 5) and the estimator (Equation 12) both hinge on the instrument vector, the section should state exactly which variables enter z in the simulations so that a reader can verify that the exclusion is implemented there.
  3. [Section 5.4, Table 10] The table is labeled 'Table 10' even though the immediately preceding tables are numbered 1 through 5, and the in-text reference in Section 5.4 points to 'Table 10' as well; the table should be renumbered and the cross-reference corrected.
  4. [Section 4, Equation (12)] The sentence introducing the estimator, 'Given a preliminary consistent estimate θ̃_N estimator solves a min-max program,' is missing a verb and should be rewritten.
  5. [Section 5.1] The sentence 'Our method performs again greatly outperforms all misspecified models' is ungrammatical; it should read 'Our method again greatly outperforms all misspecified models.'
  6. [Section 6.2] The phrase 'own-product characteristics that do not directly impact marginal costs' should be presented as an identifying assumption to be defended or tested, not as a statement of fact; the current wording obscures that the empirical claim rests on it.
  7. [Section 1] The statement 'We provide detailed guidance and code for implementation' is not substantiated by a code archive, replication appendix, or link in the submitted version.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: central identification is external (Berry-Haile), estimation is external (Bennett-Kallus), and counterfactual predictions are evaluated out-of-sample; only a minor non-load-bearing self-citation appears.

full rationale

The central derivation chain is not circular. Theorem 3.1 identifies h from the moment condition E[omega|z,w]=0 and a completeness assumption, with the proof following Berry and Haile (2014), an external source; the identification does not define the instruments in terms of h or vice versa. The VMM estimator is taken from Bennett and Kallus (2023), also external. Counterfactual predictions in both Section 5 and Section 6.3 are evaluated on hold-out samples or observed post-merger prices, so the reported MSE improvements are not constructed from the fitted values themselves. The only self-citation bearing on the model is footnote 4, which points to Dearing et al. (2024) (overlapping authors) for a discussion of conduct models satisfying Assumption 4; this is illustrative, not load-bearing for Theorem 3.1 or the empirical predictions, so it does not make the result circular. The reviewer-flagged exclusion restriction in Section 6.2--using squared average distance as an excluded instrument even though average distance is an included cost shifter--is a substantive identification and robustness concern, but it is not a circularity: it does not make the post-merger prediction algebraically equal to the estimation inputs. Section 7 openly states the extrapolation limitation for novel market structures, further confirming that the operative risk is external validity rather than circularity.

Assumptions & free parameters 2 free parameters · 8 assumptions · 0 invented entities

The central claim rests on the Berry-Haile identification assumptions (1-6), plus the correctness of the estimated demand model and the VMM regularity conditions. No new physical or economic entities are postulated. The only hand-chosen inputs are the neural network architecture and the Monte Carlo profit-weight parameter.

free parameters (2)
  • Neural network architecture (small vs large) = 3x3 or 100x100 hidden layers
    Chosen by the researchers; performance varies with architecture and sample size, but the central claim does not depend on a specific architecture.
  • Profit-weight parameter kappa in Monte Carlo DGP = 0.75
    Used to generate the profit-weight data; not a parameter of the proposed method.
assumptions (8)
  • domain assumption Assumption 1: Unique equilibrium or stable selection rule
    Ensures observed prices are stable equilibrium outcomes (Section 3.1).
  • domain assumption Assumption 2: Cost separability in unobservables: c(s,w,omega)=c~(s,w)+omega
    Defines the unobservable cost shock as residual (Section 3.1).
  • domain assumption Assumption 3: Known demand derivatives
    The supply function is identified conditional on demand derivatives; in practice demand is estimated, and this estimation error is ignored (Section 3.1).
  • domain assumption Assumption 4: Markup depends only on shares and demand derivatives
    Restricts the class of conduct models; satisfied by Bertrand, Cournot, and profit-weight models (Section 3.1).
  • domain assumption Assumption 5: Instrument exclusion: demand shifters excluded from cost
    Prevents the supply function from being confounded with inverse demand (Section 3.3).
  • domain assumption Assumption 6: Completeness of instruments
    Needed for identification of h; untestable (Section 3.3).
  • standard math Regularity conditions for VMM asymptotics (Bennett and Kallus 2023)
    Invoked for consistency and asymptotic normality of the estimator (Section 4).
  • domain assumption Demand model (nested logit) correctly specified in the empirical application
    The supply estimation and counterfactuals condition on the estimated demand system (Section 6.2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing the Merger Simulation Toolkit with ML/AI." pith.science (2026). https://pith.science/paper/GSX3Z4QD

@misc{pith2026250605225,
  author       = {Pith},
  title        = {Pith review of: Enhancing the Merger Simulation Toolkit with ML/AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GSX3Z4QD}},
  note         = {Machine review of arXiv:2506.05225}
}
read the original abstract

This paper develops a flexible approach to predict the price effects of horizontal mergers using ML/AI methods. While standard merger simulation techniques rely on restrictive assumptions about firm conduct, we propose a data-driven framework that relaxes these constraints when rich market data are available. We develop and identify a flexible nonparametric model of supply that nests a broad range of conduct models and cost functions. To overcome the curse of dimensionality, we adapt the Variational Method of Moments (VMM) (Bennett and Kallus, 2023) to estimate the model, allowing for various forms of strategic interaction. Monte Carlo simulations show that our method significantly outperforms an array of misspecified models and rivals the performance of the true model, both in predictive performance and counterfactual merger simulations. As a way to interpret the economics of the estimated function, we simulate pass-through and reveal that the model learns markup and cost functions that imply approximately correct pass-through behavior. Applied to the American Airlines-US Airways merger, our method produces more accurate post-merger price predictions than traditional approaches. The results demonstrate the potential for machine learning techniques to enhance merger analysis while maintaining economic structure.

Figures

Figures reproduced from arXiv: 2506.05225 by the authors.

Figure 1
Figure 1. Merger Simulation Results Panel A. Simulated Price Changes 0 10 20 30 40 Differences from Pre-Period Prices (%) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Density DiD: 2.92% Model VMM Bertrand Panel B. Post-Merger Price Prediction Error 40 20 0 20 40 60 Differences from Observed Prices (%) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Density 1e 2 Model VMM Bertrand Notes: The figure reports merger simulation results for the flexible mo… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

45 extracted references · 43 canonical work pages

  1. [7]

    Steven Berry, James Levinsohn, and Ariel Pakes

    Tracing the woes: An empirical analysis of the airline industry.American Economic Journal: Microeconomics2, 3 (2010), 1–43. Steven Berry, James Levinsohn, and Ariel Pakes

    work page 2010

  2. [13]

    Marine Carrasco, Jean-Pierre Florens, and Eric Renault

    Using Policy Functions to Estimate Merger Impacts: an Application to Jet-Blue Spirit.Working Paper(2024). Marine Carrasco, Jean-Pierre Florens, and Eric Renault

    work page 2024

  3. [15]

    Xiaohong Chen

    Efficient estimation of average derivatives in NPIV models: Simulation comparisons of neural network estimators.Journal of Econometrics235, 2 (2023), 1848–1875. Xiaohong Chen

    work page 2023

  4. [16]

    Federico Ciliberto, Charles Murry, and Elie Tamer

    Large sample sieve estimation of semi-nonparametric models.Handbook of econometrics6 (2007), 5549–5632. Federico Ciliberto, Charles Murry, and Elie Tamer

    work page 2007

  5. [18]

    Giovanni Compiani

    Does multimarket contact facilitate tacit collusion? Inference on conduct parameters in the airline industry.The RAND Journal of Economics45, 4 (2014), 764–791. Giovanni Compiani

    work page 2014

  6. [21]

    Learning Firm Conduct: Pass-through as a Foundation for Instrument Relevance. (2024). working paper. Nishanth Dikkala, Greg Lewis, Lester Mackey, and Vasilis Syrgkanis

    work page 2024

  7. [22]

    Max H Farrell, Tengyuan Liang, and Sanjog Misra

    Minimax estimation of conditional moment models.Advances in Neural Information Processing Systems33 (2020), 12248–12262. Max H Farrell, Tengyuan Liang, and Sanjog Misra

    work page 2020

  8. [23]

    Amit Gandhi and Jean-François Houde

    Deep learning for individual heterogeneity: An automatic inference framework.arXiv preprint arXiv:2010.14694(2020). Amit Gandhi and Jean-François Houde. 2020.Measuring Firm Conduct in Differentiated Products Industries. Technical Report. Working Paper. Jerry Hausman, Gregory Leonard, and J Douglas Zona

    arXiv 2020

Show all 45 references
  1. [24]

    Sture Holm

    Competitive analysis with differenciated products.Annales d’Economie et de Statistique(1994), 159–180. Sture Holm

    1994

  2. [27]

    Tetsuya Kaji, Elena Manresa, and Guillaume Pouliot

    Organizational Structure and Pricing: Evidence from a Large US Airline.Quarterly Journal of Economics139, 2 (2024), 1149–1199. Tetsuya Kaji, Elena Manresa, and Guillaume Pouliot

    2024

  3. [28]

    Pauline Kennedy, Daniel P O’Brien, Minjae Song, and Keith Waehrer

    An adversarial approach to structural estimation.Econometrica 91, 6 (2023), 2041–2063. Pauline Kennedy, Daniel P O’Brien, Minjae Song, and Keith Waehrer

    2023

  4. [30]

    Sophia Li, Joe Mazur, Yongjoon Park, James Roberts, Andrew Sweeting, and Jun Zhang

    Adversarial generalized method of moments.arXiv preprint arXiv:1803.07164(2018). Sophia Li, Joe Mazur, Yongjoon Park, James Roberts, Andrew Sweeting, and Jun Zhang

    2018 arXiv

  5. [31]

    Luofeng Liao, You-Lin Chen, Zhuoran Yang, Bo Dai, Mladen Kolar, and Zhaoran Wang

    Repositioning and Market Power after Airline Mergers.RAND Journal of Economics53, 1 (2022), 166–199. Luofeng Liao, You-Lin Chen, Zhuoran Yang, Bo Dai, Mladen Kolar, and Zhaoran Wang

    2022

  6. [32]

    Nathan Miller and Gloria Sheu

    Provably efficient neural estimation of structural equation models: An adversarial approach.Advances in Neural Information Processing Systems33 (2020), 8947–8958. Nathan Miller and Gloria Sheu

    2020

  7. [33]

    Nathan Miller and Matthew Weinberg

    Quantitative Methods for Evaluating the Unilateral Effects of Mergers.Review of Industrial Organization58 (2021), 143–177. Nathan Miller and Matthew Weinberg

    2021

  8. [34]

    Ross Morrow and Steven Skerlos

    Understanding the Price Effects of the MillerCoors Joint Venture.Econometrica 85, 6 (2017), 1763–1791. Ross Morrow and Steven Skerlos

    2017

  9. [37]

    Robert Porter

    Evaluating the Performance of Merger Simulation: Evidence from the US Airline Industry.Journal of Law and Economics49, 2 (2006), 627–649. Robert Porter

    2006

  10. [38]

    Joseph Sill

    Mergers and Coordinated Effects.International Journal of Industrial Organization73 (2020), 102583. Joseph Sill

    2020

  11. [41]

    Gregory J Werden and Luke M Froeb

    Estimating parameters of structural models using neural networks.Marketing Science (2024). Gregory J Werden and Luke M Froeb

    2024

  12. [42]

    & Org.10 (1994),

    The effects of mergers in differentiated products industries: Logit demand and merger policy.JL Econ. & Org.10 (1994),

    1994

  13. [43]

    Seungil You, David Ding, Kevin Canini, Jan Pfeifer, and Maya Gupta

    The Welfare Effects of Dynamic Pricing: Evidence from Airline Markets.Econometrica90, 2 (2022), 831–858. Seungil You, David Ding, Kevin Canini, Jan Pfeifer, and Maya Gupta

    2022

  14. [44]

    Zhe Yuan and Panle Jia Barwick

    Deep lattice networks and partial monotonic functions.Advances in neural information processing systems30 (2017). Zhe Yuan and Panle Jia Barwick. 2024.Network Competition in the Airline Industry: An Empirical Framework. Technical Report. National Bureau of Economic Research. R...

    2017

  15. [45]

    Harold D

    Instrumental variable regression via kernel maximum moment loss.Journal of Causal Inference11, 1 (2023), 20220073. Harold D. Chiang, Jack Collison, Lorenzo Magnolfi, and Christopher Sullivan21 A Additional Simulation Results Table A1. MSE Across Models (Bertrand DGP, Large Net...

    2023

  16. [1979]

    Han Hong, Aprajit Mahajan, and Denis Nekipelov

    A simple sequentially rejective multiple test procedure.Scandinavian journal of statistics(1979), 65–70. Han Hong, Aprajit Mahajan, and Denis Nekipelov

    1979

  17. [1992]

    Steven T Berry

    Estimation of a Model of Entry in the Airline Industry.Econometrica(1992), 889–917. Steven T Berry

    1992

  18. [1993]

    Steven T Berry

    Some Applications and Limitations of Recent Advances in Empirical Industrial Organization: Merger Analysis.American Economic Review83, 2 (1993), 247–252. Steven T Berry

    1993

  19. [1994]

    Steven T

    Estimating discrete-choice models of product differentiation.The RAND Journal of Economics25, 2 (1994), 242–262. Steven T. Berry, Michael Carnall, and Pablo T. Spiller

    1994

  20. [1995]

    Steven Berry and Ariel Pakes

    Automobile Prices in Market Equilibrium.Econometrica63, 4 (1995), 841–890. Steven Berry and Ariel Pakes

    1995

  21. [1997]

    Antoine Wehenkel and Gilles Louppe

    Monotonic networks.Advances in neural information processing systems10 (1997). Antoine Wehenkel and Gilles Louppe

    1997

  22. [2000]

    Taisuke Otsu and Martin Pesendorfer

    Mergers with differentiated products: The case of the ready-to-eat cereal industry.The RAND Journal of Economics(2000), 395–421. Taisuke Otsu and Martin Pesendorfer. 2024.Conduct in the Soft Drink Market: A Mechanism Design Approach. Technical Report. Working Paper. Craig Peters

    2000

  23. [2006]

    Jose Betancourt, Ali Hortaçsu, Aniko Oery, and Kevin Williams

    Airline hubs: costs, markups and the implications of customer heterogeneity.Competition policy and antitrust(2006). Jose Betancourt, Ali Hortaçsu, Aniko Oery, and Kevin Williams. 2022.Dynamic price competition: Theory and evidence from airline markets. Technical Report. NBER w...

    2006

  24. [2007]

    Jiafeng Chen, Xiaohong Chen, and Elie Tamer

    Linear inverse problems in structural econometrics estimation based on spectral decomposition and regularization.Handbook of econometrics6 (2007), 5633–5751. Jiafeng Chen, Xiaohong Chen, and Elie Tamer

    2007

  25. [2010]

    Andrew Bennett and Nathan Kallus

    Simulating the dynamic effects of horizontal mergers: Us airlines.Working Paper(2010). Andrew Bennett and Nathan Kallus

    2010

  26. [2011]

    Aviv Nevo

    Fixed-Point Approaches to Computing Bertrand-Nash Equilibrium Prices Under Mixed-Logit Demand.Operations Research59, 2 (2011), 328–345. Aviv Nevo

    2011

  27. [2013]

    José Azar, Martin C Schmalz, and Isabel Tecu

    The price effects of a large merger of manufacturers: A case study of Maytag-Whirlpool.American Economic Journal: Economic Policy5, 1 (2013), 239–61. José Azar, Martin C Schmalz, and Isabel Tecu

    2013

  28. [2014]

    Steven Berry and Panle Jia

    Identification in Differentiated Products Markets Using Market Level Data.Econometrica 82, 5 (2014), 1749–1797. Steven Berry and Panle Jia

    2014

  29. [2015]

    Harold D

    Extremum estimation and numerical derivatives.Journal of Econometrics(2015), 250–263. Harold D. Chiang, Jack Collison, Lorenzo Magnolfi, and Christopher Sullivan20 Ali Hortaçsu, Olivia Natan, Hayden Parsley, Timothy Schwieg, and Kevin Williams

    2015

  30. [2017]

    Greg Lewis and Vasilis Syrgkanis

    The competitive effects of common ownership: Economic foundations and empirical evidence.A vailable at SSRN 3008331(2017). Greg Lewis and Vasilis Syrgkanis

    2017

  31. [2018]

    C Lanier Benkard, Aaron Bodoh-Creed, and John Lazarev

    Anticompetitive effects of common ownership.The Journal of Finance 73, 4 (2018), 1513–1565. C Lanier Benkard, Aaron Bodoh-Creed, and John Lazarev

    2018

  32. [2019]

    Yanhao Wei and Zhenling Jiang

    Unconstrained monotonic neural networks.Advances in neural information processing systems32 (2019). Yanhao Wei and Zhenling Jiang

    2019

  33. [2020]

    RAND Journal of Economics51, 4 (2020), 1108–1161

    Best Practices for Differentiated Products Demand Estimation with pyblp. RAND Journal of Economics51, 4 (2020), 1108–1161. Adam Dearing, Lorenzo Magnolfi, Daniel Quint, Christopher Sullivan, and Sarah Waldfogel

    2020

  34. [2021]

    Federico Ciliberto and Jonathan W Williams

    Market Structure and Competition in Airline Markets.Journal of Political Economy129, 11 (2021), 2995–3038. Federico Ciliberto and Jonathan W Williams

    2021

  35. [2022]

    Christopher Conlon and Jeff Gortmaker

    Market Counterfactuals and the Specification of Multiproduct Demand: A Nonparametric Approach.Quantitative Economics13, 2 (2022), 545–591. Christopher Conlon and Jeff Gortmaker

    2022

  36. [2023]

    Steven Berry and Philip Haile

    The Variational Method of Moments.Journal of the Royal Statistical Society Series B: Statistical Methodology85, 3 (2023), 810–841. Steven Berry and Philip Haile

    2023

  37. [2024]

    Orley C Ashenfelter, Daniel S Hosken, and Matthew C Weinberg

    Price discrimination in international airline markets.Review of Economic Studies91, 2 (2024), 641–689. Orley C Ashenfelter, Daniel S Hosken, and Matthew C Weinberg

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.