The ledger does not lie, only the noise obscures.
This week the noise arrived as a single number: $80.7 billion. According to a statistical briefing circulating through English-language news wires, Americans lost $80.7 billion to cryptocurrency scams in 2025. The same briefing concedes, in subordinate clauses that most headlines will never reach, that the reported, documented, verifiable loss figure stands at $11.4 billion. The gap between these two numbers is not evidence of hidden crime. It is the product of a multiplier.
A sevenfold multiplier, to be precise. Borrowed from a 2017 survey of general consumer fraud. Applied without re-validation to a technological landscape that did not exist when the survey was conducted. The source of the report is unnamed. The methodological appendix, if one exists, has not been published. No on-chain data anchors the estimate. No address clustering supports it. No de-duplication across reporting channels was acknowledged. No sensitivity analysis accompanies the headline figure.
Liquidity is a phantom; solvency is the skeleton. The skeleton of this story has nothing to do with actual fraud patterns. It concerns how an unverified number becomes congressional testimony, how congressional testimony becomes rulemaking, and how rulemaking redraws the competitive map of an entire industry. I have lived through this cycle before. It begins always with a number that flatters the reporter. It ends always with a regulatory burden that the statisticians never pay.
The context that matters is the fraud-reporting infrastructure against which these figures must be measured. In 2025, that infrastructure is unrecognizably more mature than in 2017. The FBI's Internet Crime Complaint Center operates a dedicated cryptocurrency unit. The SEC's Crypto Assets and Cyber Unit has processed hundreds of enforcement referrals. Major exchanges maintain suspicious-activity reporting systems mandated by the Bank Secrecy Act. Chainalysis and its competitors have mapped attribution for a majority of significant on-chain exploit proceeds. The Treasury's Financial Crimes Enforcement Network has extended customer-identification obligations to a widening set of virtual asset service providers.
Each of these institutions generates data. Each feeds the recorded complaint ecosystem. When the 2017 survey measured consumer fraud reporting rates, it was measuring a world in which victims had few channels, little knowledge, and minimal confidence in the complaint process. The 2025 victim can file a complaint in minutes, attach a transaction hash as evidence, and track the case through a public portal. The reporting environment has changed. A static multiplier cannot capture change; it can only deny it.
The category confusion embedded in the estimate deserves its own flag. "Cryptocurrency fraud" is a bucket that contains investment scams, pig-butchering operations, phishing, fake wallets, compromised keys, Ponzi schemes that merely denominate in tokens, and outright theft. Reporting rates vary sharply across these vectors. A romance-scam victim may never report out of shame. A phishing victim whose exchange account was drained is reported automatically by the exchange's compliance team. A Ponzi participant may wait a year before acknowledging the loss. The 2017 multiplier cannot account for these differences because it was not built to.
I want to be precise about what this document actually is before I assess what it does. The source material is a sector news analysis — a fast-turnaround statistical briefing rather than a peer-reviewed study or a law-enforcement report. Its economic content is limited to three data points: the estimated total loss, the reported loss, and the extrapolation ratio. Zero of its content touches on protocols, code, network architecture, token design, custody mechanics, or any technical dimension of the crypto ecosystem. It is not a technology story. It is a political artifact wearing the costume of data journalism.
THE AUTOPSY OF THE 7X MULTIPLIER
Let me perform the arithmetic that the headline writers skipped. $11.4 billion multiplied by 7.077 yields $80.7 billion. The multiplier is not a rounded seven; it carries three decimal places of false precision. That precision is a tell. It suggests the author divided 80.7 by 11.4 and reported the quotient as if it were an independently estimated parameter rather than a back-derived artifact. In quantitative analysis, this is called circular reasoning. The estimate and the multiplier are not independent. One is the shadow of the other.
The 2017 survey that supposedly justifies the multiplier examined general consumer fraud — telemarketing solicitations, deceptive lending, identity theft, traditional financial exploitation. It found that roughly 86% of victims did not file a complaint. The inverse of 14% reporting yields the approximate sevenfold factor. Whether the survey was methodologically sound for its own purpose is beside the point. The point is that it was never designed to measure cryptocurrency fraud reporting in a decade in which cryptocurrency itself had become an institutional asset class.
The adaptation is invalid on three separate grounds.
First, the population base has changed. In 2017, the average crypto participant was betting a modest discretionary allocation on an experimental asset. By 2025, the American crypto market includes institutional allocators, pension funds, registered investment advisers, and an entire class of ETF investors — all of whom hold crypto through regulated intermediaries with mandatory reporting obligations. When BlackRock's IBIT experiences a security event, the SEC learns about it through a different channel than a consumer complaint.
Second, the reporting inducements have changed. Exchange-level KYC means that compromised accounts are detected by automated systems and reported without victim initiative. The 2021 FinCEN guidance on convertible virtual currency suspicious activity reporting pulled a large share of apparent fraud into the official pipeline. Some portion of the $11.4 billion reported total was captured not because victims filed complaints but because institutions filed reports.
Third, the countervailing shame dynamic has been partially offset by community infrastructure. The crypto community, for all its faults, is relentless in publicizing scam alerts. Discord, Telegram, and X channels circulate address blacklists within hours. Registration databases like CryptoScamDB and Chainabuse provide structured complaint intake. Web3 wallet plugins warn users before approvals that interact with flagged contracts. A 2025 victim has encountered three independent warnings before losing funds, converting many potential recovery claims into averted losses — which, naturally, never enter the statistics.
The multiplier fails on all three counts. Yet the estimate built on it is being quoted with the same confidence as an audited balance sheet. The algorithm reveals what the story hides. And the story hides the absence of an algorithm.
WHAT 2017 AND 2020 TAUGHT ME ABOUT NUMBERS
I have been a professional skeptic of crypto statistics since before most of the current analyst corps entered the industry. In late 2017, during the ICO mania, I conducted forensic due diligence on five Ethereum-based projects that had collectively solicited hundreds of millions of dollars in retail capital. The project I have always referred to as Project Alpha had secured $50 million of commitments on the strength of a website, a whitepaper, and the founder's charisma. It took me nine days to find the reentrancy vulnerabilities in its token contract — the same bug class that had drained The DAO of $60 million the previous year. The code spoke unambiguously. The marketing was the lie.
I published a technical breakdown on GitHub. The social media reaction was instructive. The project's defenders attacked my motives, my identity, and my timing. They did not attack the bytecode, because the bytecode was unassailable. I saved a meaningful group of early investors from a catastrophic loss. It would be pleasant to say that this vindicated the power of technical rigor. What it actually demonstrated is that the ecosystem's ability to verify claims was — and remains — catastrophically underdeveloped.
The $80.7 billion estimate is the exact inverse of the Project Alpha situation. In Project Alpha, the code was the verifiable truth and the narrative was the fraud. Here, the narrative is the fraud, and the code — or rather, the documented data infrastructure — is nowhere to be found. I do not dispute that fraud is real; I have tracked its evolution across every market cycle since 2017. What I dispute is the confidence with which an extrapolation without a verification layer has been promoted as fact.
The 2020 DeFi Summer sharpened my approach. When I modeled the initial token emission schedules of the Curve ecosystem, the mathematics was unambiguous: the early APYs were not yields; they were liquidity rents — subsidized participation fees paid from an inflating supply. I stress-tested my portfolio accordingly, shorting volatile governance tokens and rotating into stablecoin yield aggregators. When Harvest Finance collapsed in July 2020, my firm absorbed no material exposure. The lesson that carried forward: incentive structures decay on predictable schedules, and any model that treats a static parameter as immortal will get you killed.
The 7x multiplier is exactly such a static parameter. It assumes that the reporting ratio has been constant from 2017 through 2025, across an asset-class maturation that has no parallel in comparable eight-year periods. That assumption costs nothing to make — until someone acts on it.
My 2022 experience reinforced the same discipline. After the Terra-LUNA collapse, I shifted my research framework from crypto-specific metrics to global macro liquidity indicators, specifically Federal Reserve balance-sheet contractions. I authored a report correlating stablecoin supply shrinkage with S&P 500 correlation, demonstrating that crypto had become a leveraged bet on global M2 expansion. That systemic view preserved 80% of our capital during the winter. It worked because it was built on observable data — central bank balance sheets, money supply, stablecoin flows — rather than narrative constructs. The $80.7 billion estimate belongs to the narrative construct family. It cannot survive contact with observable data.
THE REGULATORY TRANSMISSION MECHANISM
Here is where the phantom becomes policy. The mechanism has six stages, and I can trace each one from prior experience.
Stage one: an alarming estimate is published, sourced loosely, and cleansed of its uncertainties. Stage two: news wires pick it up; headlines compress the range into a single point. Stage three: mainstream financial media cites the point figure without caveats; the multiplier's provenance vanishes. Stage four: a legislator in the Senate Banking Committee or the House Financial Services Committee reads the figure into the congressional record. Stage five: the SEC, CFTC, FBI, or FinCEN incorporates the figure into enforcement narratives, budget justifications, and proposed rule text. Stage six: compliance officers, exchange counsel, and institutional allocators encode the number into risk models and procurement decisions.
The figure's provenance no longer matters beyond stage one. Once established in official channels, the number acquires the solidity of precedent. I have watched this dynamic mis-shape policy before. The "90% of ICOs are scams" statistic was fabricated from a convenience sample of token listings, statistically indefensible, and repeated so often that it was eventually cited in academic papers and congressional testimony as established fact. The "Terra collapsed with $40 billion in investor losses" narrative survived for years despite conflating market-cap destruction with realized cash flows. Each false number served an agenda. Each had the same shape: big, scary, and directionally aligned with restraint.
There is also a deeper historical analogy. In the aftermath of the 2008 financial crisis, regulators cited aggregated estimates of consumer fraud and predatory lending that were later shown to be inflated by sloppy data aggregation. But the regulatory architecture built on those inflated numbers — the Consumer Financial Protection Bureau, the expanded SEC enforcement mandate, the Dodd-Frank derivatives regime — did not collapse when the statistics were corrected. Institutions, once empowered, do not surrender authority because the numbers that justified them were soft. The crypto industry should expect the same asymmetry.
The specific regulatory consequences that could flow from an $80.7 billion figure are concrete and legible.
One, expanded KYC requirements for self-hosted wallets. Congress has repeatedly entertained bills that would impose counterparty identification on non-custodial protocols. A fraud-loss narrative of this magnitude provides the political cover to overcome technical feasibility objections and civil-liberties concerns.
Two, privacy-tool suppression. The Treasury's sanctions designations of Tornado Cash were legally contested, and the courts delivered at least partial relief to the developers. An $80.7 billion fraud narrative changes the litigation calculus. The courts are not trained to distinguish credible sanctions designations from indefensible ones. They are trained to be alarmed by loss magnitudes.
Three, expanded securities classification. The Howey test, as I have argued in institutional memos, is a brittle instrument for smart-contract-native assets. It was designed in 1946 for citrus groves. To apply it rigorously requires analyzing whether a token holder's expectation of profit derives from the efforts of others. An $80.7 billion loss narrative encourages regulators to skip the nuanced application and declare entire asset classes to be investment contracts by default.
Four, enforcement budgets. The SEC and CFTC have repeatedly requested expanded appropriations for digital asset enforcement. Large loss statistics are the primary budget justification. More funding produces more investigations; more investigations produce more publicized enforcement actions; more enforcement actions validate the original statistical narrative. A self-reinforcing loop, insulated from the underlying data-quality question.
I conducted a comparative custody audit of BlackRock's IBIT and Fidelity's FBTC in early 2024, prior to the spot ETF approval, and I remain convinced that the ETF structures were approved on the strength of genuinely rational data about surveillance sharing and custody infrastructure. That is the system working correctly. But the same system, fed fabricated statistics, produces fabricated urgency. When urgency is fabricated, the burdens fall disproportionately on decentralized protocols that cannot speak in the accent of Washington compliance. Arbitrary enforcement punishes the compliant and exempts the connected.
MARKET MICROSTRUCTURE AND THE TRANSMISSION OF FEAR
Will the $80.7 billion figure crash the market? No. Crypto markets have absorbed shocks of far greater structural magnitude: the 2022 credit contagion, the May 2022 stablecoin depeg cascade, the 2024 long-liquidation waves. A statistical briefing, however scary, does not carry the operational force to trigger systemic leverage liquidation. But the transmission of fear operates on a slower and more dangerous channel.
The first link in that channel is attention allocation. When mainstream media leads with a triple-digit billion loss figure, retail investors' cognitive bandwidth shifts from accumulation to threat assessment. They hold. They hesitate. They stop adding to positions. The marginal buyer disappears at precisely the moment that liquidity is thinnest.
The second link is risk premium. As perceived danger rises, the required return for holding a risky asset rises with it. Higher required returns compress valuation multiples. In an asset class as sentiment-driven as digital assets, the compression can be immediate.
The third link is capital flow. Retail investors who are unsure whether the industry is inherently fraudulent do not sell at once; they simply stop buying. Institutions, meanwhile, update their regulatory risk models and delay allocation decisions. The consequence is a liquidity decay curve, not a price crash. And liquidity decay, as I wrote in my 2020 stress-testing framework, is the mechanism by which the market actually redistributes capital between protocols. Macro tides drown micro-waves without warning.
I tracked the post-Terra stablecoin outflows through the second half of 2022. Initial aggregate declines were attributable to the direct depeg; the sustained outflows were not. They were attributable to a learned narrative — the house-of-cards thesis — which persisted long after the UST-specific mechanics had been settled. Every subsequent data point was filtered through that interpretive lens. The same filtering will happen for the $80.7 billion report. Every future scam headline will be cited as confirmation of an already massive problem.
Institutionally, I continue to view crypto as a leveraged bet on global M2 expansion. When central bank balance sheets contract, crypto assets historically underperform to the downside. But the regulatory narrative compounds the macro effect. Allocators weight regulatory risk alongside liquidity risk. In an environment where the regulatory narrative is driven by inflated loss statistics, capital rotates toward the most compliant vehicles: ETF wrappers, regulated futures, prime brokerage. Decentralized protocols bearing residual compliance risk face a relative capital outflow.
The asymmetry between headlines and corrections ensures that this rotation persists. A debunking requires reading a methodology section, understanding confidence intervals, and caring about statistical validity. A headline requires only a headline. By the time the academic rebuttal arrives — if it arrives — the capital has already moved.
THE BOUNDARY OF THE DATA
Let me set out what the data actually supports, so we do not lose the thread of what is knowable.
The reported loss figure of $11.4 billion is best understood as a floor. It aggregates complaints submitted through official channels: FBI IC3 submissions, FTC consumer complaints, SEC and CFTC whistleblower intakes, state financial regulators, and institutional suspicious-activity reports. A floor is still valuable. It tells us that recognized fraud is nontrivial and growing. It tells us that enforcement and education have not kept pace. But a floor is not a total. The actual number is certainly higher than $11.4 billion, because some victims never file. The honest question is: how much higher?
A responsible estimate would present a range. It would incorporate stratified sampling of the crypto user population, a dedicated survey of crypto-specific reporting behavior, cross-referencing of exchange-banned addresses with regulatory complaints, on-chain address clustering to identify multi-asset victims, and a time-series analysis of how reporting rates have shifted with regulatory consolidation. That effort would produce a central estimate with a confidence interval. My own instinct, based on the institutional data I have reviewed across multiple regulatory regimes, is that the true figure likely lies between $25 billion and $45 billion, with a central estimate around $33 billion. That range is urgent enough to justify aggressive enforcement without being inflated enough to justify draconian restrictions.
The false precision of a single point estimate with three decimals is a signal of an unwarranted identity between the model and the world. Precision is not accuracy. A number that appears exact while being systemically miscalibrated licenses the most aggressive policy response. A range that acknowledges uncertainty forces policymakers to weigh tradeoffs. The report has chosen the former.
I also flag the recovery accounting. The $11.4 billion reported total does not disclose how much was recovered through law-enforcement freezes, exchange clawbacks, insurance payouts, or legal judgments. Crypto's on-chain transparency creates a recovery infrastructure that traditional fraud victims do not have: stolen assets can be frozen at centralized exchange entry points, identifiable through chain analysis, and seized through court orders. The gross-to-net loss ratio in crypto is meaningfully different from the general consumer fraud context that the 2017 multiplier assumes. The report does not address this distinction, and the distinction changes everything.
THE COMPLIANCE TECHNOLOGY OPPORTUNITY
A false number can still move real markets. The $80.7 billion figure, regardless of its accuracy, will accelerate procurement of fraud-detection and compliance infrastructure. I see four beneficiary classes, and I am evaluating each with the same rigor I apply to any investment thesis.
The first is blockchain analytics. Chainalysis, Elliptic, TRM Labs, and a cohort of smaller forensic specialists will face sustained demand growth. Regulators cite loss statistics; then they demand tools to trace the losses. The procurement cycle typically opens within six to twelve months of a major fraud narrative, as enforcement agencies and financial institutions allocate their budgets.
The second is identity and reputation infrastructure. Wallet risk-scoring protocols, on-chain identity verification, and compliance-focused KYC products will be integrated by exchanges under pressure to demonstrate fraud mitigation. The 2024 ETF custody standards, which I analyzed in detail for institutional clients, set an expectation of continuous monitoring that will now migrate to a wider set of platforms.
The third is insurance. Crypto insurance products — exchange custody coverage, smart-contract cover, protocol indemnification — become more valuable when the perceived threat environment worsens. The report's failure to disclose recovery rates is relevant here: gross-loss narratives drive insurance purchases even when net exposure is materially lower.
The fourth is user education and protective tools. Web3 wallet risk-plugins, scam-alert services, approval-simulation tools, and investor-protection platforms have historically struggled to monetize. A sustained fear cycle changes the economics. Users will pay for protection when they believe the threat is real.
I must be clear that none of these theses depends on the accuracy of the $80.7 billion figure. They depend on its propagation. In the attention economy, a compelling narrative moves budgets more effectively than a precise one. The compliance sector's growth does not require the fraud narrative to be correct; it requires the narrative to be sticky. Due diligence is the only hedge against asymmetry. The asymmetry here is information. I intend to hedge accordingly.
WHAT A RESPONSIBLE ESTIMATION WOULD REQUIRE
Let me reconstruct what a defensible fraud-loss estimate would actually involve, because it provides the benchmark against which all future claims should be judged.
Step one is reconciliation. Every reported complaint must be de-duplicated across datasets. The same victim who files an FTC complaint, an IC3 report, and an exchange support ticket appears three times in the raw counts. Without identity resolution, even the reported floor is inflated.
Step two is vector classification. Each complaint must be coded by fraud mechanism: phishing, romance scam, investment fraud, credential compromise, malicious dApp, fake wallet, private-key theft, social engineering. Reporting rates vary by an order of magnitude across these vectors. Blending them into a single rate is statistically incoherent.
Step three is latency modeling. The 2017 survey measured a contemporaneous snapshot. Fraud complaints in crypto arrive on a delayed schedule: a victim may not discover the loss for months, and the discovery may not be reported for further months. Any point-in-time estimate must adjust for the accrued-but-unreported inventory.
Step four is on-chain verification. A modern estimate should deploy address clustering, transaction-graph analysis, and — where practicable — symbolic execution of known scam contracts to confirm their roles. It should distinguish losses to external hacks from losses to internal collusion, and both from losses that were never crypto-native but merely crypto-denominated.
Step five is recovery accounting. The estimate should report gross losses, frozen losses, recovered losses, and insurance reimbursements separately. The substantive question — how much net wealth did Americans permanently lose — is answerable only after this decomposition.
Step six is sensitivity analysis. The final output should be a range. A single point estimate with three decimal places of precision is a political artifact, not a statistical finding. If the report cannot publish its confidence interval, it is not ready for congressional testimony.
I have applied this framework to the $80.7 billion claim. It fails every step.
THE AI FRAUD FRONTIER
One final dimension demands attention. In 2026, I developed a valuation model for Machine-to-Machine economy tokens — infrastructure that enables autonomous AI agents to transact without direct human supervision. What I discovered during this work is that the fraud landscape is shifting beneath the feet of every existing estimation methodology.
The next generation of crypto fraud is agent-mediated. Deepfake video conferencing clones executives and instructs treasury staff to authorize transfers. Generative AI deploys phishing sites with pixel-perfect fidelity to legitimate interfaces. AI agents analyze on-chain behavior and dynamically deploy honeypot contracts tailored to individual users' transaction histories. Autonomous botnets conduct social engineering at scales and speeds that human-run operations cannot match.
Each of these vectors has a different forensic signature. Each has a different reporting profile. Each will be mis-measured by a 2017 multiplier optimized for telephone-based consumer fraud.
The 2026 framework I built valued tokens according to algorithmic utility and data-verification cost rather than human social metrics. The same logic applies to fraud defense: machine-mediated fraud requires machine-speed defense. Static rule-based systems will be obsolete. The demand for real-time, AI-augmented on-chain risk assessment will grow as fast as the processing power of the attackers.
I do not know whether the $80.7 billion figure includes meaningful AI-mediated losses. I suspect it does not, because the report appears methodologically locked to a pre-AI statistical infrastructure. But the next iteration of the fraud-estimation literature must be built for a world where the attackers are autonomous. That estimation does not exist yet. Whoever builds it first will shape the next decade of crypto security policy.
THE CONTRARIAN VIEW: DECOUPLING, NOT COLLAPSE
Let me invert the surface reading of this story. The dominant interpretation in market commentary will be bearish: giant scam losses, more regulation, weaker sentiment, lower prices. That interpretation is too simple. And simple interpretations, in this industry, are usually how people lose money.
The $80.7 billion figure is not primarily a market event. It is a structural event. It will accelerate the separation of the crypto industry into two tiers: the compliant, institutionally legible operators, and the gray-market survivors. I have seen this separation occur in every major fraud-panic cycle since 2017. The 2017 ICO fraud wave produced the custody, audit, and institutional service layer that now underpins Wall Street's engagement. The 2022 FTX collapse destroyed one centralized exchange but catalyzed the regulatory framework that made the 2024 spot ETF approvals possible. The 2025 fraud narrative, however distorted, will catalyze the consolidation of compliance infrastructure into a durable economic sector. Inversion is the only constant in chaos: what appears as a threat to the industry's survival is often the mechanism of its institutionalization.
The institutionalization thesis has a macro corollary. Fraud panics tend to cluster at the inflection points of liquidity cycles. They emerge when retail enthusiasm has overstayed its welcome and when the macro backdrop is shifting from expansion to contraction. In this reading, the $80.7 billion narrative is not a cause of the next downturn. It is a symptom of a market already transitioning from risk-on to risk-off. The most sophisticated allocators will treat it as a confirming signal, not a new one, and will position their portfolios accordingly — overweight the compliant infrastructure names, underweight the gray-market exposure.
The second inversion concerns the data itself. The inflated $80.7 billion number contains a hidden gift: it sets the expectation bar so high that the official annual reports from the FBI and the FTC — when they publish their 2025 fraud data — will of necessity appear as a refutation. When the official number lands at $15 billion or $18 billion, the public will be forced to confront the gap between media narrative and documented reality. That gap is the seed of credibility repair.
The risk is that the correction never arrives. The "90% of ICOs are scams" figure was never officially withdrawn. The "$40 billion Terra loss" has been recalibrated only in specialized outlets. A false statistic, once weaponized, has a long half-life. And for every actual victim of fraud, the inflated estimate is a second victimization: it erodes the credibility of the very apparatus designed to report, track, and recover their losses. The boy who cried wolf is not a myth. It is an operational risk.
So my contrarian assessment is not that the number is bullish or bearish. It is that the number is epiphenomenal. The real variable is institutional reaction. And institutionally, the reaction to a fraud panic is always the same: consolidate, standardize, regulate, absorb.
TAKEAWAY: WATCH THE SIGNALS, NOT THE HEADLINES
Clarity emerges from the subtraction of noise. Subtract the $80.7 billion entirely from your analytical model, and what remains is a set of measurable signals that will determine whether this narrative becomes policy or expires as trivia.
Signal one: the original report's provenance. If the data is traced to an authoritative institution — the FBI, the FTC, the Treasury — its policy weight increases by an order of magnitude. If it is traced to a marketing firm or an obscure publication, it should be discounted to near zero.
Signal two: regulatory citation. Watch for the $80.7 billion figure in SEC enforcement press releases, CFTC public statements, congressional hearing transcripts, and budget justifications. Each citation ratchets the compliance burden.
Signal three: the official 2025 annual reports. The FBI and FTC will publish their documented totals in early 2026. The gap between those totals and the $80.7 billion narrative will define the credibility of the report that produced it.
Signal four: exchange behavior. If major platforms announce new fraud-detection features, mandatory address screening, or reimbursement insurance products within the next six months, the number has already moved the operating environment.
The fraud is real. The losses are real. The response must be proportionate. But in an industry where the difference between $11.4 billion and $80.7 billion is the difference between a problem and a panic, precision is not an academic luxury. It is a survival requirement.
The ledger does not lie. The question is whether the next number that crosses the wire will be taken from the ledger — or from the phantom.