CCC’s: What do we know about algorithmic collusion? (Guest Post by Ai Deng PhD)

 by  Leave a Comment

Dr. Ai Deng of Bates White Economic Consulting has been a long time and frequent contributor to Cartel Capers.  He is a leading voice in the area of artificial intelligence and algorithmic collusion.  You can follow him on LinkedIn (here).  HIs most recent post is below:

************************************

I had the pleasure of speaking about artificial intelligence and algorithmic collusion at the American Bar Association Section of Antitrust Law Spring Meeting 2018 last month. The star war-themed session seemed to have gone very well. I want to thank again Paul Saint-Antoine, Lesli Esposito, Professors Maurice Stucke and Joshua Gans for putting together the panel with me.

I have just posted another article on algorithmic collusion on SSRN. The paper is partially based on my remarks at the Spring Meeting but expands on several fronts. Below is the abstract. You can download the full working paper at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3171315

Abstract

The past few years have seen many legal scholars and antitrust agencies expressing interest in and concerns with algorithmic collusion. In this paper, I survey and draw lessons from the literature on Artificial Intelligence and on the economics of algorithmic tacit collusion. I show that a good understanding of this literature is a crucial first step to better understand the antitrust risks of algorithmic pricing and devise antitrust policies to combat such risks.

Keywords: algorithmic pricingalgorithmic collusionartificial intelligenceantitrust

This is one of a series of papers I have written in the past year about the general topic of machine learning and artificial intelligence, and their implications on antitrust issues.

As always, I appreciate your thoughts and comments. You can reach me at [email protected] or connect with me on LinkedIn [here]

Ai Deng, PhD

Principal at Bates White Economic Consulting

Lecturer at Advanced Academic Program, Johns Hopkins University

direct: 2022161802 | fax: 2024087838

1300 Eye Street NW, Suite 600, Washington, DC 20005

CCC’s: An Antitrust Lawyer’s Guide to Machine Learning (Guest Post by Ai Deng PhD.)

 by  Leave a Comment

Below is a post by valued guest contributor, Ai Deng, PhD. of Bates White Economic Consulting.

*******************************************************************************

There has been growing interest in the legal community in artificial intelligence (AI), and more specifically in machine learning (ML). This recent interest in AI is at least in part driven by concerns about algorithmic collusion, i.e., the possibility that computer algorithms could ultimately collude on their own, without human facilitation.

There is no question that the antitrust community is largely playing catch-up when it comes to the technical subject matters of AI and ML. As the Acting Chair of the Federal Trade Commission Maureen K. Ohlhausen noted, “The inner workings of these tools are poorly understood by virtually everyone outside the narrow circle of technical experts that directly work in the field.”

While there is no point to antitrust attorneys understanding the nuts and bolts of AI and ML technology, a basic understanding is necessary to better understand and assess the implications of the AI/ML research on antitrust and related legal and economic issues. That is the motivation behind my latest article. Through a series of simple examples, I introduce some fundamental concepts in ML. Along the way, I also discuss a wide variety of ML applications in the law and economics field. I conclude with a brief discussion of the hot topic of algorithmic collusion.

You can download the paper here https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3082514

As always, I appreciate your thoughts and comments. You can reach me at [email protected] or connect with me on LinkedIn [here].

Ai Deng, PhD

Principal

direct: 2022161802 | fax: 2024087838

1300 Eye Street NW, Suite 600, Washington, DC 20005

[email protected]

BATESWHITE.COM

CCC’s: A New Article on Algorithmic Collusion (Guest Post by Ai Deng PhD.)

 by  Leave a Comment

Below is a post by valued guest contributor, Ai Deng, PhD. of Bates White Economic Consulting.

*****************************

In a new article I published in Law360 last week, I discussed the following four reasons why the scope of colluding algorithms, even if they are technologically possible, could be limited:

  • Algorithmic asymmetry
  • Robust compliance
  • Observable collusive outcomes
  • Risk of class actions

The paper is titled “Four Reasons We May Not See Colluding Robots Anytime Soon” and is available here. If you do not have a subscription to Law360 but would like to have a copy, please feel free to email me at [email protected]

As always, I appreciate your thoughts and comments. You can reach me at the email above or connect with me on LinkedIn [here].

Thanks for reading.

CCC’s: Antitrust and Artificial Intelligence, Empirical Analysis in Class Certification: A Research Update (Guest Post)

 by  Leave a Comment

By: Ai Deng, PhD,  Principal, Bates White Economic Consulting

Hope everyone had a wonderful Labor Day weekend. During my time off CartelCapers, I have been working on several research projects. In this post, I’d like to give the interested readers an update on two of them.

When Machines Learn to Collude: Lessons from a Recent Research Study on Artificial Intelligence

From Professors Maurice Stucke and Ariel Ezrachi’s Virtual Competition published a year ago, to speeches by the Federal Trade Commission Commissioner Terrell McSweeny and Acting Chair Maureen K. Ohlhausen, to an entire issue of a recent CPI Antitrust Chronicles, and a conference hosted by Organisation for Economic Co-operation and Development (OECD) in June this year, there has been an active and ongoing discussion in the antitrust community about computer algorithms. In a short commentary (downloadable here), I briefly summarize the current views and concerns in the antitrust and artificial intelligence (AAI) literature pertaining to algorithmic collusion and then discuss the insights and lessons we could learn from a recent AI research study. As I argue in this article, not all assumptions in the current antitrust scholarship on this topic have empirical support at this point.

Sub-regressions, F test, and Class Certification

Did the anticompetitive conduct impact all or nearly all class members? This question is central to a court’s class certification decision. And to answer the question, a methodology—known as sub-regressions (also labelled less informatively as simply the “F test” in the recent Drywall litigation)—is being increasingly employed, particularly by defendants’ expert witnesses. A key step of a sub-regression type analysis is to partition the data into various sub-groups and then to examine data poolability.[1]

Forthcoming in the Journal of Competition Law & Economics, my article titled “To Pool or Not to Pool: A Closer Look at the Use of Sub-Regressions in Antitrust Class Certification” focuses on three areas of interest pertaining to sub-regressions:

  • The related law and economics literature related to this methodology
  • Courts’ recent class certification decisions in cases where parties introduced sub-regression analysis
  • Several methodological challenges, many of which have not been previously acknowledged, as well as potential ways to address them. Specifically, what test should one use? How does one choose the subsets or partitions of data to test? Are individual estimates of damages always the most reliable approach when we believe the impact varies across customers or across some other dimensions?

This paper is currently being processed at the Journal. If you would like a copy, please feel free to reach out to me.

As always, I appreciate your thoughts and comments. You can reach me at [email protected] or connect with me on LinkedIn [here].

Thanks for reading.

Ai Deng, PhD
Principal, Bates White Economic Consulting
Lecturer, Advanced Academic Program, Johns Hopkins University
direct: 2022161802 | fax: 2024087838
1300 Eye Street NW, Suite 600, Washington, DC 20005
[email protected]
BATESWHITE.COM

[1] I first provided an update on this project on CartelCapers here.

Guest Post from Ai Deng, Bates White

Below is a guest post from economist Ai Deng, Phd. of Bates White Economic Consulting:

***************************************************************

Hope everyone had a wonderful Thanksgiving holiday.

I thought the readers may be interested in knowing about the recent Global Forum on Competition hosted by OECD.

The event, taking place just a month ago, had a session on cartels titled “Serial offenders: Why do some industries seem prone to endemic collusion?” The panelists included Professor Joseph Harrington (The Wharton School, University of Pennsylvania), Professor Robert Marshall (Department of Economics, Penn State University and Bates White), Professor Valerie Suslow (Carey Business School, Johns Hopkins University), and Mr. Robert Wilson (Webber Wentzel). I did not attend the program in person, but the program materials including the panelists’ presentations are available for download here.

The panelists Professors Harrington, Marshall, and Suslow have all done influential academic research on cartel-related topics. Their work was cited in my own recent research on cartel detection and monitoring. Mr. Wilson, a partner in the Competition Practice at Webber Wentzel, specializes in competition law and international trade.

To give the readers a quick, high level overview,

Professor Harrington’s presentation provides his thoughts on when firms collude. He then describes a 3-step inductive approach to cartel screening and uses the cement market as an example to demonstrate how to apply such an approach in practice.
Using a dataset of cartel participants based on the European Commission (EC) decisions in cartel cases, Professor Marshall specifically notes the role of association management companies (AMC) in cartels. He argues that “it would be valuable to understand the role of AMCs…” and if AMCs compete “with one another to provide this [cartel] services to firms in a product/industry/market, then antitrust policy should be directed toward deterring the role of AMCs with regard to such anticompetitive activities.”
Mr. Wilson’s presentation overviews South Africa’s Competition Act and then specifically focuses on South African construction industry. He identifies possible reasons for the extensive collusion in that industry and makes policy recommendations.
Professor Suslow’s presentation is titled “Serial Collusion in Context: Repeated offenses by firm or by industry?” In addition to address the question raised in the title, she also discusses seven policy tools and emphasized the importance of understanding what leads to collusion in the first place to select appropriate policy tool.

There is a wealth of information in their presentations and supplemental materials. In addition to the panelists’ presentations, also available for download are a “background note by secretariat” and contributions from a number of jurisdictions.

Ai Deng, PhD

Principal

direct: 2022161802 | fax: 2024087838

1300 Eye Street NW, Suite 600, Washington, DC 20005

[email protected]

BATESWHITE.COM

CCC’s: Welcome Ai Deng, Phd. (Bates White)–Guest Post

 by 

I am pleased to welcome Ai Deng, Phd., to Cartel Capers as a guest poster.  Ai is an economist with Bates White Economic Consulting.  I met Ai at some of the antitrust conferences that Bates White sponsors and I’ve always enjoyed his economist’s insight on various cartel related issues.  Ai’s first post is below.

******************************************************

Competition authorities and regulatory bodies are resolute in encouraging companies to beef up their corporate compliance programs. As an example, in the LIBOR investigation, the DOJ required Barclay’s and other banks to “maintain or develop monitoring systems or electronic exception reporting systems that identify possible improper or unsubstantiated submissions.” [1]Similar agreements were reached between various banks and the CFTC. Good compliance effort by the corporation apparently also pays—in the recent FOREX investigation, the DOJ took notice of Barclay’s efforts and stated in its plea agreement: “The parties further agree that the Recommended Sentence is sufficient, . . . , in considering, among other factors, the substantial improvements to the defendant’s compliance and remediation program to prevent recurrence of the charged offense.”

To better detect various forms of market manipulation, corporate compliance officials can employ a data analytic technique called an “empirical screen.” This technique has already been used by antitrust authorities all over the world, and it is getting increased attention in recent academic literature. An empirical screen is a metric that is based on data and a pre-specified formulation. The value of the metric changes as the likelihood of market manipulation increases or decreases. When the value crosses a certain threshold, a “red flag” for suspicious activity goes up. When this occurs, additional investigation of the causes may be warranted.

“Detection” techniques similar to empirical screens are widely used in the credit card and telecommunications industries for fraud detection purposes. AT&T Labs’ researchers Becker, Volinsky, and Wilks (2010) noted that AT&T implemented its fraud detection system (the Global Fraud Management System) nearly 20 years ago, in 1998.[2] AT&T’s team of data experts continuously analyzes data and devises new techniques to detect fraud. Credit card companies also invest significantly in fraud detection efforts. Organizations that are contemplating establishing or strengthening their compliance programs can also benefit from adopting screening and detection analytics. In the recent Law360 article “What Compliance Officials Must Know About Market Screening” available here, I focus on two important practical issues that have not yet been adequately addressed but which are crucial for a successful deployment of empirical screen techniques.  If you don’t have access to Law 360 and would like a copy of my article, please contact me at [email protected].

_______________________________________

1.      Deferred Prosecution Agreement at § vi (“Monitoring and Auditing”), United States v. Royal Bank of Scotland (D. Conn. Feb. 5, 2013), available athttp://www.justice.gov/iso/opa/resources/28201326133127414481.pdf.

2.    Richard A. Becker, ChrisVolinsky, and Allan R. Wilks, “Fraud Detection in Telecommunications: History and Lessons Learned,” Technometrics 52, no. 1 (2010): 20–33.