Can you Go to Jail if your AI Engages in Collusion?

The stereotype of price fixing is fat, cigar-chomping executives meeting secretly in smoke-filled rooms to agree to raise prices. But the Justice Department's recent consent decree with RealPage anticipates that AI collusion may look different. RealPage sold revenue-management software that used confidential pricing information from competing landlords to recommend rents. Although the software generated the recommendations, the DOJ alleged that the system facilitated unlawful coordination among competitors. I had my doubts. Nevertheless, the settlement requires RealPage to stop using certain competitively sensitive data and to change features of its pricing software.

AI can still recommend prices. Firms have used sophisticated pricing software for decades. A broader implication is that firms cannot avoid antitrust liability by outsourcing pricing decisions to an algorithm. If competing firms provide confidential information to a common AI system that helps coordinate pricing decisions, regulators may view the arrangement much like traditional collusion.

This distinction will become increasingly important as companies deploy AI agents to make autonomous business decisions. An AI pricing system that independently analyzes a firm's own costs, demand, and inventory is generally very different from one that relies on competitors' confidential information or otherwise facilitates coordination among rivals. The RealPage consent decree is consistent with antitrust law focusing on economic outcomes rather than how those outcomes were produced.

Use It or Lose It: The Perverse Incentives Draining the American West

If you wanted to design a property rights regime that guaranteed the waste of a scarce resource, you could hardly do better than Western water law's doctrine of prior appropriation. It requires rights holders to put water to "beneficial use" or—after five consecutive years of non-use—forfeit the right. [1] The result is a textbook perverse incentive: a farmer who irrigates more efficiently risks losing the conserved portion of her right, so the private return to conservation is negative even when the social return is big.

Oregon tried to eliminate the perverse incentive with its 1987 Instream Water Right Act that made instream flow a beneficial use. [2] On the Deschutes, conservancies now lease and buy water for the river. [3], but legalizing the trade didn't create a liquid market. Rights are mostly held by irrigation districts rather than individual farmers, so every deal needs board approval, and transaction costs run high: quantifying how much water actually reaches crops through leaky, century-old canals is expensive, and farmers still fear that proving they can conserve invites a future challenge to their right.

In the Colorado River, states face the same problem. If an Upper Basin state conserves, the saved water flows downstream to be consumed by someone else which creates the risk of becoming the baseline for future cuts. [4] Instead, the feds pay farmers billions to fallow fields—a costly government subsidy trying to address the perverse consequences of another government policy. [5]

BOTTOM LINE: any rule tying an asset's ownership to its continuous consumption—budget lines that vanish if unspent, headcount that shrinks if unfilled—will be consumed regardless of value. 

NOTES

[1] Schwabe, Williamson & Wyatt, "Oregon Water Law Questions and Answers" — beneficial use requirement and five-year forfeiture rule. https://www.schwabe.com/publication/oregon-water-law-questions-and-answers/

[2] Oregon Legislature, "Background Brief on Water Rights" — the 1987 legislation adding instream water rights as a beneficial use. https://www.oregonlegislature.gov/lpro/Publications/2004HM_Water_Rights.pdf

[3] University of Oregon School of Law, "Evaluating Instream Flow Programs" — Oregon's 1,100+ instream leases and transfers, including the Deschutes River Conservancy's role. https://law.uoregon.edu/sites/default/files/ai61_ch._22_with_legend1.pdf

[4] High Country News, "Why Colorado River negotiations are so difficult" — prior appropriation dynamics and the interstate stalemate. https://www.hcn.org/articles/why-colorado-river-negotiations-are-so-difficult/

[5] Congressional Research Service, "Management of the Colorado River" (R45546) — federally compensated conservation, including 2.3 million acre-feet paid for with congressionally approved drought funds. https://www.congress.gov/crs-product/R45546

HT:  Claude Fable

Extracting Information from Supply Disruptions

On Sept. 16, 2025, a fire at Novelis's aluminum rolling mill in Oswego, New York disrupted roughly 40 percent of the automotive aluminum sheet used in North America. The importance of this supply disruption can be determined using standard stock market event study analyses. Ford was one of the firms most exposed to the disruption because it relied heavily on automotive aluminum sheet. As analysts began estimating the impact on vehicle production and profitability, Ford's stock price fell sharply. Similarly, investors also penalized Hindalco, Novelis's parent company. The interesting question is what happened to competitors. If one supplier exits temporarily, shouldn't rivals benefit?

Firm	Relationship to Event	Approximate Stock Market Reaction
Ford	Customer of automotive aluminum sheet	-7% abnormal return
Hindalco (owner of Novelis)	Directly affected supplier	-6% abnormal return
Kaiser Aluminum	Potential substitute supplier	+3% to +5% short-term gain
Constellium	Potential substitute supplier	Little measurable effect

Investors appeared to believe that Ford would suffer large costs from the disruption, but they did not assign equally large gains to Novelis's competitors. While aluminum itself is a commodity, automotive aluminum sheet is not. Suppliers must invest in specialized equipment, satisfy demanding quality standards, and undergo lengthy qualification processes with automakers. Even if competitors wanted to absorb Novelis's lost volume, they may not have had sufficient spare capacity or approved production lines to do so quickly. Ford's stock market loss was several times larger than analysts' estimates of immediate production costs because investors recognized that replacing a critical supplier is expensive and time-consuming. There are no doubt benefits to Ford and Novelis from contract exclusivity. This episode highlights the costs.

Rethinking Strategy at Cracker Barrel

Julie Felss Masino was recruited to become the new CEO of Cracker Barrel in 2023 to reinvigorate the brand. Should strategy focus on growing demand by increasing attractiveness to new customers or should it focus on keeping existing customers comfortable? As the WSJ reports, it opted for the former:

The goal was to modernize its stores, menu and design to help bring new customers to the vintage chain. The company had seen some promising results from food updates and marketing tie-ins when, as part of a fall marketing campaign in 2025, it unveiled a new, simplified logo.

And then Cracker Barrel became an unexpected flashpoint in the culture wars. The new logo was described as "going woke" and the changes alienated many of its more traditional patrons. Visits plummeted and the company lost $100 million in market capitalization in a week.

Well that didn't work. To her credit, Masino quickly reversed course.

She cut ties with the marketing firm behind the chain’s rebranding campaign and revamped the company’s leadership structure, bringing back a former vice president for menu strategy and elevating a veteran field operator to oversee store operations. 

They doubled-down on the traditional experience. Even the old logo has returned. The pivot seems to be working. Customers are beginning to return and the share price is rebounding, though not to the $60-$70 range prior to these events.

A strategic decision is a sunk cost. Abandon it when new information indicates a mistake.

Hidden Cost of Environmental Protection

 REASON:

Greenpeace and its activists allies have blocked for more than two decades the adoption of Golden Rice, which is genetically enhanced to produce the vitamin A precursor beta-carotene. The result, according to new calculations by DC Abundance founder and research director at the Golden Gate Institute for AI Abi Olvera, is that "delay has killed about 106,000 children and left another 210,000 to 425,000 blind."

What Set-Asides Cost: A Lesson from Timber Auctions.

California's Public Utilities Commission is pressuring utilities to steer about 1.5% of procurement toward state-certified "LGBT-owned" firms.  Setting aside the certification debate, what will this cost the state? 

In every auction, the winner has to outbid the second-best bidder, so the second-best bidder sets the price. Weaken the field and you weaken that price-setter. But which way the price moves depends on whether the government is buying or selling.

When the government buys, it runs a procurement auction: bidders compete to sell to the government, and the lowest-cost bidder wins, while second-lowest-cost bidder sets the price. Restrict who can bid and the price the government pays goes up.

When the government sells, bidders compete to buy, and the highest-value bidder wins, while the second-highest-value bidder sets the price. Restrict who can bid and the price the government receives goes down.

A set-aside moves price through two separate channels, and they push the same direction. 

• First, it shrinks the number of bidders, so the second-lowest cost is higher (or the second-highest value is lower). 
• Second, the set-aside bidders themselves may be higher-cost or lower-value than the bidders they replace. 

Both channels move price against the government.  An article by a pair of middling economists shows by how much.  Prices in Forest Service small-business set-aside auctions—where only small businesses may bid—run about 15% lower than in open auctions. 

The lesson applies to California. Fewer, weaker bidders mean a worse deal for the government. 

AI and the Shrinking Firm

The way that Artificial intelligence (AI) is changing worker productivity may also be changing the optimal scale of the firm itself. According to a recent Axios report, AI-powered tools are enabling entrepreneurs to launch and operate businesses with little or no staff. Tasks that once required specialists in coding, graphic design, marketing, customer support, and bookkeeping can increasingly be performed by a single entrepreneur assisted by AI. The result is a growing number of "one-person firms" capable of generating revenue levels that previously required a small team. This contrasts with the trend since the industrial revolution in which ever more mechanization generated ever greater economies of scale and ever larger enterprises.

Some aspects of AI fit nicely into theories of the firm.

•          Transactions costs can be reduced by engaging with AI rather than an employee.
• AIs eliminate employee principal/agent issues.
• However, a more personalized AI may represent a relationship-specific sunk cost that can lead to holdup.

Has the risk premium for owning stocks disappeared?

Axios reported that the equity risk premium (ERP) — the extra return investors expect for holding risky stocks instead of safe Treasuries — has shrunk to almost nothing. A safe 10-year Treasury bond pays about 4.5%. Stocks, measured against the companies' current earnings, return only about 3.7%. So right now the safe bond actually pays more than risky stocks, even though stocks have historically paid 3–5% extra. It looks as if you'd be taking on the risk without the usual reward.

HOWEVER: That 3.7% is based on what companies earn today. Stock prices are high because investors expect much bigger earnings in the future, driven by AI. If those bigger earnings actually arrive, then the price you pay today is reasonable, and stocks aren't really overpriced after all.

BOTTOM LINE: The risk premium you calculate depends on which earnings you use — today's or the future's. The historical rule assumes today's earnings are a good guide to the future. If AI changes how much companies earn, that assumption breaks, and the real risk premium is unknowable until we see whether the growth shows up.

Did incapacitation, deterrence, or rehabilitation reduce crime in Baltimore?

The Free Press:

Bates, ``a new tough-on-crime prosecutor, ... replaced a scandal-plagued `progressive.'” '' 

Incapacitation (selection): sometimes referred to as ``specific deterrence.''

Bates said that his office has identified about about 6,000 frequent, violent offenders and put between 3,000 and 3,500 of them in prison. The cooperation of federal law enforcement has helped take a number of these offenders off the streets.

Deterrence (incentives):

...more willingness to process felons in possession of a gun, and a more credible threat of punishment.

Rehabilitation:

“If you’re seeing your friends all going to prison, you’re going to go: ‘What? I don’t want to go to prison,’ ” Bates said. “Now all of a sudden, that job or that program someone’s offered you before that you didn’t want to talk about—now it looks pretty appealing.”

Wheelchair Fraud

WSJ: They Get Wheeled on Flights and Miraculously Walk Off. Praise ‘Jetway Jesus.’ 

One viral account had 30 travelers needing wheelchairs to board a Southwest flight — and all but two walking off on their own at arrival.  Another traveler, Carlos Gomez, described a flight delayed by 25 wheelchair passengers and said he sees more "wheelchair fraud" every trip. A flight attendant reportedly told him many able-bodied passengers request wheelchairs for "the VIP experience" — skipping lines, first crack at overhead bins — and that the "healing begins" once they learn they'd have to wait for assistance to get off.