ML

Dossier 008

Matt Levine

Columnist · Money Stuff author · Former investment banker and M&A lawyer

Ten-year sample score

673 hits · 2 partials · 1 miss
Review windowJun. → Nov. 2016
Eligible sample6 of 14 screened claims
Price-only calls0
Last reviewedJuly 22, 2026

The verdict

He explained the machine.
Mostly, the machine obliged.

Levine’s useful forecasts rarely arrived wearing forecast-shaped hats. They were mechanisms: impossible targets produce misconduct; balance-sheet costs preserve apparent arbitrage; thin bond liquidity can feed on itself; computers have the natural advantage in minute-by-minute trading. A decade later, the machinery had been considerate enough to leave receipts.

The weaker calls came when institutions mattered more than mechanics. Goldman built a real deposit franchise but retreated from its grander consumer experiment. And a Delaware trial judge’s finding that Michael Dell bought cheaply was reversed by a higher court. Even efficient markets occasionally insist on appellate review.

The receipts

Six claims, ten years on

Six mechanisms. No target prices. One appellate reversal.

How this sample was chosen

We screened 14 accessible Bloomberg Opinion columns and newsletter editions published from June through November 2016, selecting six claims before grading outcomes. We excluded contemporaneous diagnoses without a future-facing implication, including the DAO hack after it occurred, derivatives-resolution mechanics that have not faced a comparable US bank failure, and an IEX interview in which the forward-looking claims belonged to the interviewee. We also excluded jokes, moods and claims too dependent on an inaccessible continuation of a paywalled article.

01
Corporate law

Dell appraisal

Bought too cheaply

Michael Dell’s 2013 buyout acquired Dell for less than its fair value, as the Chancery Court’s appraisal analysis found

The headline chose the trial court’s discounted-cash-flow answer: Michael Dell bought his company too cheaply. Delaware’s Supreme Court chose the market check, reversing because the lower court gave the deal price no weight. The original judgment was real; it was also not the last judgment. Finance has footnotes. Law has appeals.

Bloomberg Opinion · Jun. 1, 2016
Observed realityMissThe Delaware Supreme Court reversed and held that the deal price had heavy, if not overriding, probative valueMiss
02
Bank governance

Wells Fargo

Impossible quotas

Unrealistic sales targets would generate fraud, abusive tactics or burnout even when management repeatedly told employees not to cheat

Levine identified the system rather than accepting the bad-apple brochure. Telling staff not to open fake accounts did not neutralise targets they could not honestly meet. The Justice Department later described the same mechanism, Wells paid $3 billion, and the Federal Reserve kept an asset cap in place until 2025. Eight did not rhyme with governance.

Money Stuff newsletter · Sep. 19, 2016
Observed realityHitWells admitted that onerous goals and management pressure led thousands of employees to create millions of unauthorized or fraudulent accountsHit
03
Bank strategy

Goldman Sachs

Boring-bank turn

Post-crisis constraints were pushing Goldman toward a more ordinary model: take retail deposits and use them to fund customer loans

The funding thesis worked. Marcus deposits became part of Goldman’s scaled franchise and its bank ranked sixth in US consolidated assets in early 2025. The consumer revolution attached to those deposits fared less elegantly: cards and mass-market lending were cut back after losses. Goldman became more bank-like, then remembered which parts of banking it preferred.

Bloomberg Opinion · Nov. 10, 2016
Observed realityPartialGoldman built a large Marcus deposit base, but abandoned much of its broader consumer-lending strategy after costly lossesPartial
04
FX market structure

Covered interest parity

Arbitrage stays broken

Currency-basis anomalies would persist because bank balance sheets were too costly and constrained to arbitrage them away

A textbook no-arbitrage law met a leverage ratio and lost. Levine’s account was that hedging demand could open the basis while expensive dealer balance sheets prevented it closing. The BIS was still modelling the resulting neutral band in 2024. The free money remained available chiefly to anyone for whom capital was free, a notably compact membership category.

Money Stuff newsletter · Sep. 19, 2016
Observed realityHitBIS research in 2024 still described covered-interest-parity deviations as no longer short-lived and linked them to post-crisis risk regulationHit
05
Credit markets

Bond funds

Liquidity feedback

Poor bond liquidity could push investors into passive, buy-and-hold funds, making individual bonds harder to trade and reinforcing the shift

The loop appeared, but it did not eat the market. In March 2020, fund outflows forced sales into worsening liquidity, dealers hit balance-sheet limits and the Fed intervened in corporate bonds and bond ETFs. Yet funds met redemptions and assets later exceeded pre-pandemic levels. The vulnerability was correctly wired; permanent paralysis was not.

Money Stuff newsletter · Sep. 19, 2016
Observed realityPartialPassive credit expanded and March 2020 redemptions produced forced sales, collapsing liquidity until Federal Reserve facilities restored market functionPartial
06
Trading technology

Human scalpers

Machine advantage

Physiological intuition would not give human short-horizon traders an advantage over algorithms that measure and react vastly faster

A small study linked heartbeat awareness to trader survival. Levine declined to infer victory over machines, noting that computers are exceptionally attentive to their own clock cycles. Subsequent market structure has been unromantic: algorithms became ordinary infrastructure while human judgment moved up the stack. The machines did not develop better instincts. They made milliseconds a hostile habitat for instincts.

Money Stuff newsletter · Sep. 19, 2016
Observed realityHitRegulators describe modern markets as fully electronic and algorithmically pervasive, with automation forming the dominant execution infrastructureHit

Outcome checks use the Delaware Supreme Court’s Dell record, the Justice Department’s Wells Fargo statement, the Federal Reserve’s 2025 asset-cap decision, Goldman Sachs’s 2025 annual report, Federal Reserve bank asset data, the BIS’s 2024 covered-interest-parity study, the Federal Reserve’s corporate-bond liquidity review and the SEC’s algorithmic-trading report. The sample contains no price-only investment calls. “Partial” receives half credit.

Open ledger

Discussion

Yes, everything is securities fraud. The grading still requires evidence.

SP
Sonia P.Jul 22, 2026

The Wells Fargo item captures why Levine is difficult to score. It was framed as an explanation of a scandal already underway, but the mechanism made a durable claim about what impossible quotas would keep producing.

TL
The Long ViewJul 22, 2026

That is the threshold used here: the event itself was not a forecast, but the claim that screening alone could not fix a target-driven management system was both forward-looking and testable.

ER
Edwin R.Jul 22, 2026

Partial for Goldman is right. Marcus proved retail deposits could diversify funding; it did not prove Goldman could become a mass-market consumer lender without setting fire to a respectable sum of money.

Join the discussionNew comment