1. The question hidden inside a loss number
A bank-failure headline often contains a number large enough to crowd out every other detail. An asset discount, an estimated insurance-fund cost, a remaining loan balance and a payment to creditors can all be described as money lost or recovered. Yet they answer different questions. Adding them together can count the same economic exposure twice. Comparing them without dates can turn an ordinary revision into an apparent contradiction. Treating an estimate as a settled payment can make a recovery look finished before the money arrives.
The useful starting question is not whether the number is big. It is whose record contains it, what the number measures, and where it sits between a transaction and a final outcome. That approach makes bank resolution less mysterious without pretending that every document supplies a complete explanation. It also gives readers something better than an argument over a single headline: a way to identify the missing reconciliation.
The FDIC describes its individual receivership financial statements as unaudited, prepared on a modified cash basis, and including statements of assets and liabilities in liquidation, operations, and accompanying notes. Its publication page distinguishes annual from quarterly releases. Those descriptions matter before any number is copied into a comparison. FDIC receivership financial statements
That is a different reporting object from the insurance funds themselves. GAO's March 26, 2026 audit reported that the Deposit Insurance Fund and FSLIC Resolution Fund financial statements for 2025 and 2024 were fairly presented in all material respects under generally accepted accounting principles. An opinion on those funds is not an audit opinion on every individual estate's separate statement. GAO's 2025 and 2024 fund audit
Imagine a reader assembling three folders: the announcement describing a resolution transaction, the records following the assets that remain, and the records showing payments or changes in an insurance fund. The folders can be connected, but their headings should not be erased. A cash receipt in one may discharge a claim held by another. A reduction in an expected loss may involve no contemporaneous receipt at all. The work is to identify those connections, not to assume that identical words imply identical measures.
This article uses dated official examples and an invented estate called Harbor Ledger. Harbor Ledger is not a bank, a forecast, a real receivership or a model of anyone's claim. Its deliberately simple records let us test arithmetic that would otherwise be buried inside a much larger transaction. No example supplies instructions for filing a claim, establishes an individual's priority or recommends a financial decision.
2. Read the transaction before constructing the subtraction
On March 26, 2023, the FDIC announced the First Citizens transaction involving Silicon Valley Bridge Bank. The announcement described about $72 billion of assets purchased at a $16.5 billion discount, approximately $90 billion in securities and other assets remaining in receivership, and equity appreciation rights potentially worth up to $500 million. These were different components of a dated transaction announcement, not three interchangeable measures of a final loss. FDIC's March 26, 2023 announcement
The words around those amounts perform real accounting work. Purchased assets identify a transaction perimeter. A discount compares a purchase measure with a specified reference amount. Assets remaining describe something outside that sale. Potential proceeds identify a contingent amount, not necessarily a receipt already recorded. A reader who removes those labels can produce a precise-looking calculation that does not correspond to any of the underlying documents.
There is also a buyer's reporting perspective. First Citizens' SEC-filed statement addresses assets acquired and liabilities assumed on March 27, 2023, using estimated fair values based on preliminary valuations. Its auditors explicitly distinguish that statement from a complete presentation of the company's assets and liabilities. The filing also reports that the appreciation instrument was exercised on March 28 and a $500 million payment made on April 4. That last sequence separates the right, its exercise and the later payment. First Citizens' acquisition statement and notes
This is not permission to equate every buyer-side valuation with a seller-side recovery. It is a reason to label each perspective before attempting a bridge between them. A number can be accurate in an acquisition statement without being the amount of cash a receivership collected on the announcement date. A document can be audited without answering a question outside its stated scope.
For an original illustration, suppose a package contains assets with a reference amount of 40 units and liabilities of 35 units. Calling the package a 40-unit purchase tells us nothing sufficient about the net settlement. Was the asset amount a carrying balance, a valuation or a gross contractual balance? Did the buyer assume all 35 units? Was consideration cash, a note, or several components? Without those fields, subtracting a quoted discount from the asset figure cannot produce an independently verified payment.
A useful transaction worksheet therefore retains the exact source label for each amount rather than translating everything into purchase price. It has separate spaces for assets included, liabilities assumed, retained items, cash consideration, noncash consideration and unresolved settlement adjustments. Blank spaces are informative. They mark questions that the announcement cannot answer, rather than inviting a reporter to invent the missing entries.
3. Identify the estate, the date and the accounting boundary
A current search result can lead to a statement covering an earlier period. The period written on the statement is the relevant date for its balances. The retrieval date establishes when the researcher encountered it; the two dates should appear in different fields. Nor does a quarterly filename guarantee that the document represents the latest quarter available elsewhere. This article's examples are explicitly dated records, not a claim to have established the newest balance of every estate.
One official indexed statement illustrates why those distinctions matter. San Diego National Bank's fund 10135 statement identifies an October 30, 2009 failure and a March 31, 2026 reporting period. It lists $7,460,768 in cash and investments, zero assets in liquidation, and $58,689,164 in total liabilities. The statement's age since failure and its remaining balances belong to the same record; an old failure does not by itself tell us what remains on a later statement. FDIC's March 31, 2026 San Diego National Bank statement
Those figures alone do not establish when another distribution will occur, why a particular balance remains, or how much any individual can receive. The arithmetic difference between assets and liabilities is not a substitute for those missing facts. The important lesson is narrower: a historical failure date and a later liquidation statement can coexist without contradiction. Reading one while ignoring the other produces an incomplete timeline.
The FDIC's separate Silicon Valley Bank financial-statement reading-room page offers another dated caution. It says that some aspects of the June 30, 2023 statements might not yet reflect sale transactions in progress even though cash received from those pending sales appears in the statements. The page also notes presentational changes. That warning belongs to those identified statements, not to an allegation about every later release. FDIC's Silicon Valley Bank financial-statement notice
In Harbor Ledger's imaginary archive, a cover sheet says Estate H, quarter ending June 30, dollars in thousands. An attached sale report says Portfolio H-2, settled July 3, dollars. A worksheet that imports both amounts as if they were June cash in the same units will fail even if every digit was transcribed correctly. Multiplying the wrong entry by a thousand fixes one error but not the other. The July settlement still lies outside June's cash period.
This is why a record identity should include the institution or estate identifier, document type, statement date, unit, accounting basis and version. The identity is part of the evidence, not an administrative ornament. It prevents a bank, its bridge institution, its purchaser and a related corporate entity from silently becoming one row simply because their names share familiar words.
4. A balance sheet can shrink while collections succeed
Start Harbor Ledger with 100 units of recorded assets: cash of five, loans of 75, securities of 15 and other assets of five. Its toy obligations total 110. The resulting negative residual is ten. These are invented starting values, not a representation of FDIC accounting conventions or an actual institution's condition. We use them to distinguish an inventory from what happens to it.
Harbor sells a subset of loans with a recorded amount of 40 and receives 30 in cash. The loan line falls from 75 to 35; cash rises from five to 35. Total assets fall from 100 to 90 because the sale realizes ten less than the recorded amount removed. Yet the estate has collected 30. Saying that it recovered nothing because assets declined would confuse a balance change with a cash flow.
Next it pays two for services and distributes 20 against obligations. Cash is now 13: five plus 30, minus two, minus 20. The remaining assets are cash of 13, loans of 35, securities of 15 and other assets of five, totaling 68. Obligations fall from 110 to 90 after the assumed distribution. The residual is negative 22. It changed by the ten-unit sale shortfall and the two-unit expense, not by the 20-unit distribution that reduced both cash and obligations.
The example creates several true but non-equivalent statements. Thirty was collected from a sale. Assets declined by 32 from their starting total. Twenty reached the stipulated obligation holders. Two was spent on services. The residual worsened by 12. None is a synonym for the others. If a story calls all five measures recovery or loss without retaining their definitions, the reader cannot reconstruct what happened.
There is a second trap in the remaining 68. It includes 55 of noncash assets. Calling 68 available cash would treat the remaining loans, securities and other assets as though they had already become receipts. Calling only the 13 of cash the ultimate recovery would make the opposite mistake by assigning no possible future proceeds to what remains. Our toy has not specified those future proceeds, so neither conclusion is available.
A proper bridge preserves the beginning balances, the items removed, the proceeds received, the expenses paid, the distributions made and the ending balances. It should be possible to follow the same sale through the asset record and the cash record without counting it twice. That is a stronger test than whether a total seems reassuring or alarming in isolation.
5. Reconcile the cash account without calling every inflow a recovery
A second Harbor Ledger exercise starts a quarter with five units of cash. Its receipts register shows 12 from asset sales, eight from collections on retained assets, one of interest and two from new borrowing. Its payments register shows two in expenses, 18 in distributions and one repaying earlier borrowing. All amounts are invented, and the categories are stipulated to be non-overlapping. The ending cash balance is seven: opening five plus receipts of 23 minus payments of 21.
A tempting headline would call the 23 a recovery. But two came from borrowing, which creates a repayment obligation rather than demonstrating successful collection of the old assets. Another headline might call the cash increase of two the recovery, discarding the money that arrived and then left in distributions. Both fail to describe this register. The informative account separates 20 from sales and collections, one of interest, two of financing and the different uses of the money.
Even netting needs a label. Receipts from sales, collections and interest total 21; subtracting the two in expenses gives 19. That is a defined operating cash subtotal in this fictional exercise, not a universal FDIC performance measure. It is not the same as ending cash, total distributions, or profit on the assets sold. We have not specified the assets' acquisition values or the periods over which the collections were earned, so this worksheet cannot supply those additional measures.
Now suppose one sales receipt appears twice in the evidence packet: a broker's settlement statement records 12, and a bank statement records the same 12 arriving. Two documents corroborate one event; they do not establish two receipts. The reconciliation should match an event identifier, amount, counterpart and settlement date across the documents, then preserve their distinct evidentiary roles. The settlement statement explains the transaction; the cash record supports the movement of funds.
Dates can create a smaller but equally consequential mismatch. Harbor's fictional sale agreement is signed September 28, the settlement is scheduled September 30, and the cash record shows October 2. A document saying the transaction closed in September does not itself prove that the receipt belongs in September's cash subtotal. The accounting note and actual entry must resolve the treatment. We should not repair the mismatch by silently moving either date.
A useful publication can report a bounded result before every ambiguity is resolved: identified receipts total 23, of which two are borrowing; documented payments total 21; the cash balance reconciles to seven. It can then identify a specific missing field, such as whether a listed fee is already netted from a settlement. That is more useful than either declaring fraud from an unexplained difference or announcing a clean reconciliation by plugging the difference into miscellaneous income.
6. Keep cash, carrying amounts and estimates in separate columns
Harbor's next fictional portfolio begins with a contractual reference balance of 50. For this exercise, an expected-loss adjustment of 20 produces an estimated net amount of 30. Nothing about that subtraction places 30 in a cash account. It expresses an expectation attached to the assets. We stipulate the figures rather than applying them to any official statement or presenting them as the required accounting method for all receiverships.
During the next period, Harbor receives 12 from part of the portfolio. Its analysts estimate that the remaining assets will produce another 22. Cash received plus estimated future proceeds now totals 34, four above the earlier expectation. A summary saying expected proceeds improved by four is supported by these assumptions. A summary saying Harbor has already received 34 is not. Twenty-two still depends on future events in the constructed example.
This distinction becomes especially important when explaining a revision. The total can improve because collection results exceeded an earlier assumption, because the remaining assets' outlook improved, or because both happened. Our numbers alone do not identify the contributions. To do that, the analyst would need the earlier expectation for the particular assets that produced the 12, not just the portfolio's initial total of 30.
Suppose that earlier allocation expected ten from the assets now collected and 20 from those remaining. Then two of the improvement comes from realized receipts above their assigned expectation, and two from raising the expected remaining amount to 22. The four-unit change can now be decomposed. Without the original allocation, this explanation would have been a plausible story, not a demonstrated reconciliation. The extra document matters because it selects between competing explanations that produce the same total.
A downside scenario is also easy to state precisely. If the remaining assets eventually produce only 16, lifetime proceeds become 28, below both the revised expectation of 34 and the original expectation of 30. That outcome would not mean the previously recorded 12 was imaginary. It would mean the future component disappointed. The timeline separates an actual receipt, two forecasts and an eventual result rather than allowing the final outcome to rewrite what each earlier record said.
The practical reading habit is to preserve three verbs: recorded, received and expected. They should not be interchangeable even when the quantities happen to be numerically equal. A statement can combine cash and estimated values in a total while still identifying them in separate lines. A reader's simplified chart should retain enough of that distinction to show how much of its conclusion depends on events that have not yet occurred.
7. Recovery percentages require a cohort, not just a denominator
Consider a third invented Harbor portfolio with an original reference balance of 100. By the reporting date, assets representing 70 of that original balance have been disposed of and have produced cash of 50. Assets representing the remaining 30 are still held, with estimated future proceeds of ten. The cash recovered relative to the entire starting reference is 50 percent. Cash relative to the disposed portion's reference is about 71.43 percent. Cash plus expected remaining proceeds relative to the starting reference is 60 percent.
All three percentages are arithmetically valid. They describe different things. The first is a realized-to-date measure for the whole cohort. The second describes the completed portion, whose characteristics might differ from what remains. The third combines actual and expected amounts. A headline that selects the largest percentage without its definition is not merely rounding aggressively; it is changing the question the statistic answers.
The word cohort is important because portfolios change. Suppose Harbor adds a newly identified asset with a reference amount of 20 halfway through the period. The denominator becomes 120 if that asset is included in the reporting population. Keeping the numerator at 50 produces 41.67 percent, even though no previously recorded cash disappeared. An unchanged-definition comparison would either keep the original cohort separate or restate the earlier period transparently. It would not interpret the percentage decline as a deterioration in collections without examining the added asset.
Weighted comparisons introduce another trap. Fictional portfolio A produces nine from a reference of ten, while portfolio B produces 30 from a reference of 90. Averaging their individual rates gives approximately 61.67 percent: the mean of 90 percent and 33.33 percent. Combining their actual amounts gives 39 from 100, or 39 percent. The difference is not a disagreement about the payments. The unweighted average assigns the small portfolio the same influence as the portfolio nine times its size.
Neither calculation needs to be hidden. If the question is how a typical portfolio performed when each portfolio counts equally, the unweighted measure has a defined purpose. If the question concerns dollars recovered across both portfolios, the combined amounts answer it. What is unacceptable is labeling one as the other. An article should give the numerator, denominator and weighting choice before asking readers to interpret the result.
There is a final boundary around every percentage here: expenses. The invented 50 is gross cash from the assets. If obtaining it required five in separately recorded expenses, a net measure would use 45, subject to a clear allocation rule. Shared costs cannot be assigned to each portfolio in full without duplication. A comparative chart that appears exact to two decimal places can still be fundamentally ambiguous about which costs belong in its numerator.
8. A distribution waterfall is not a promise to a particular claimant
For ordinary statutory priority, 12 U.S.C. 1821(d)(11) lists receiver administrative expenses, deposit liabilities, other general or senior liabilities, subordinated obligations, then shareholder-status obligations. The text excludes secured claims to the extent of their security and expressly references another statutory provision. This is a bounded description, not a complete determination of anyone's rights. Current official text of 12 U.S.C. 1821
To examine the arithmetic without deciding any legal status, create three purely fictional payment buckets: A has 80 units of already established obligations, B has 20, and C has ten. Stipulate that the example's rules require A before B and B before C. Also stipulate that the amount available is already net of every cost and reserve relevant to the toy. These are model inputs, not findings about a real creditor or the treatment of a particular instrument.
With 60 available, the example pays 60 into A and nothing into B or C. With 90 available, it pays 80 into A and ten into B. With 110 available, all three buckets are filled. Total resources increased by 30 between the first two cases, but the improvement is not divided equally across the buckets. A receives 20 more and B receives ten more. A report that applies a single recovery rate to every obligation would erase the stipulated order.
Now hold the available amount at 90 but change the established amount in A from 80 to 85. B's allocation falls from ten to five even though available resources have not changed. The cause is a different obligation denominator, not a decline in asset receipts. This is why an explanation of a distribution change needs both sides of the record: what became available and which obligations were established in the relevant class.
An internal reserve assumption can change the timing too. Suppose 90 of cash is on hand, but the toy sets aside eight for a specifically documented future expense. Only 82 enters the current calculation; A receives 80 and B receives two. If the expense later totals five rather than eight and nothing else changes, three more becomes available. Those are conditional arithmetic results. They do not establish that an actual reserve is excessive, that it should be released, or that a payment is due on a particular date.
A responsible public explanation can show these sensitivities while leaving legal eligibility to the actual record and applicable process. Its contribution is to distinguish a change in resources, a change in established obligations and a change in a stated reservation. It should not turn a model into a calculator for someone's check. The missing inputs are not minor details when they determine which bucket, if any, a real obligation occupies.
9. Follow a payment across both sides without doubling it
The FDIC's dividend explanation distinguishes a claim asserting indebtedness from a proven claim and describes receivership dividends as cash distributions to proven claimants generated through liquidation. Its policy explanation links distributions to sufficient cash accumulating in an estate. A remaining asset estimate is therefore not the same record as a completed distribution. FDIC's receivership-dividend explanation
There is another connection between a depositor's payment and the receiver's records. Section 1821(g) provides for FDIC subrogation to the extent of the specified payment or assumption, including corresponding dividend rights, while preserving the depositor's claim for an uninsured or unassumed portion. The statutory scope matters; it is not an additional duplicate entitlement to the same amount. Official subrogation provisions
An original, non-legal bookkeeping example makes the duplication problem visible. A fictional protection fund pays 80 to customers, and the toy assigns it a corresponding claim of 80 against an estate. Later the estate sends 30 to that fund. From the estate's perspective, 30 has left in a distribution. From the fund's perspective, 30 has arrived as a recovery on its claim. Combining those views should not create 60 of new value. There was one transfer, observed from two sides.
Nor does the 30 demonstrate that the customers have received another 30. In this stipulated example they already received the 80, and the later transfer runs to the fund. Keeping an explicit sender and recipient on each entry prevents a statement about replenishing the fund from becoming a false statement about additional customer payments. This is a record-reconciliation example, not an analysis of any actual depositor's coverage.
Timing adds a further distinction. If the estate records a distribution instruction on June 30 and the fund records cash on July 2, a consolidated timeline needs both dates and a description of the unsettled item. It should not declare one side wrong just because their period-end presentations differ. Equally, the existence of an instruction does not prove the recipient received the money. The evidence must establish the status claimed in the sentence.
The same discipline applies when the public record lists a percentage distribution. The researcher needs to know whether it is an additional payment for that round or a cumulative percentage across rounds, and which proven amount serves as the base. In a toy with an unchanged base of 100, successive additional distributions of 20 percent and ten percent produce 30 in total. If the second listing instead means a cumulative 30 percent, adding 20 and 30 would wrongly report 50. The label determines the calculation.
10. An estimated insurance-fund cost is not an estate's cash balance
The FDIC's special-assessment page, marked updated April 17, 2026, described approximately $16.7 billion as the estimated cost associated with the Silicon Valley Bank and Signature Bank failures that had to be recovered through that assessment. It explained that estimates change as assets are sold, liabilities satisfied and expenses incurred, and described final adjustment mechanisms. That dated estimate concerns the specified assessment scope; it is not a claim that $16.7 billion is the latest October figure, a single estate's cash balance or its total original assets. FDIC's dated special-assessment explanation
To see how a forecast can change without a new cash receipt, return to the imaginary protection fund. Suppose its simplified exposure consists of 100 already paid out, an expectation of 70 coming back and five of future costs. Its expected net cost is 35: 100 minus 70 plus five. If projected recovery rises to 78 while the other assumptions remain unchanged, expected net cost falls to 27. No money need have arrived that day for the forecast to improve by eight.
Now make eight of the previously expected 78 arrive. If the remaining forecast falls to 70 because that eight has been collected, the total expected recovery remains 78. The cash position has improved, but the expected net cost is still 27. Counting the receipt as another eight-unit improvement would double count a recovery already anticipated. This is precisely why an analyst needs a bridge from prior expectations to actual receipts and revised remaining expectations, rather than a list of favorable developments.
The example can move in the other direction. If future costs rise from five to nine, expected net cost becomes 31 even with the improved recovery forecast. Receipts and expected recoveries may have improved while the overall cost estimate worsens. A narrative that attributes the change to asset performance alone would miss the expense component. The correct explanation follows the individual inputs rather than choosing whichever headline sounds most dramatic.
Assessment collections are a separate line in this constructed ledger. If the fund collects 27 through an assessment, that receipt changes how the stipulated cost is financed; it does not transform the portfolio's expected 78 of recoveries into 105. Combining assessment cash with asset proceeds might answer a fund-liquidity question, but it would not answer how much the assets returned. The source and purpose of each receipt remain part of its meaning.
None of these invented calculations reproduces the FDIC's complete estimation methods. Their purpose is narrower: to demonstrate that cash movement, estimate revision and financing can occur on different schedules. A useful report states which changed, identifies the dated comparison and avoids calling a provisional estimate a final bill.
11. Time changes a recovery comparison even when totals match
Two fictional estates each eventually collect 100 from equally defined starting portfolios. Estate Quick receives the full amount today. Estate Slow receives it five years later. Nominally, their totals match. That fact does not make their timing identical, nor does it establish which strategy was preferable without considering costs, uncertainty and the options actually available. The timing is part of the outcome rather than a decorative footnote.
For a purely illustrative time-value calculation, choose an assumed annual discount rate of five percent. One hundred received in five years has a present-value equivalent of about 78.35 under that chosen rule: 100 divided by 1.05 raised to the fifth power. The result is not a current market valuation, a recommended rate or an official estimate. It shows why comparing undiscounted totals can conceal a material difference in a question explicitly concerned with timing.
Sensitivity should be visible. At an assumed zero rate, the later 100 remains 100. At ten percent, it becomes about 62.09. Selecting a rate can therefore change the presentation substantially. A publication should identify the assumption and explain why that calculation is being shown instead of displaying one result with an air of inevitability. The actual receipt date remains a fact of the toy; the present-value amount is a result of a chosen analytical rule.
Expenses can also offset a larger gross collection. Suppose strategy Early produces 90 immediately with two of costs, while strategy Late produces 100 after five years with 12 of costs at that same later date. Both net to 88 before any time adjustment. Under the assumed five-percent rule, the later net amount is equivalent to about 68.95 today. The extra ten of gross proceeds does not establish an improvement in this constructed comparison because costs and time differ too.
But the exercise does not prove that a real receiver should always sell earlier. We stipulated certainty, one receipt date, a rate and a simplified expense schedule. Real records might show different risks, constraints or intermediate collections. The public-interest value is in exposing those required inputs, not using a classroom calculation to announce a strategy that the evidence has not evaluated.
A less ambitious comparison may be more defensible. Report nominal cumulative receipts at fixed elapsed intervals after the relevant starting event, alongside cumulative expenses and the remaining asset scope. A six-month-old estate should not be judged against a ten-year collection total as if both had the same opportunity to realize assets. Even age matching cannot eliminate differences in portfolios or transaction structures, but it removes one obvious source of confusion without pretending to solve all the others.
12. Turn the unresolved difference into a specific records question
Consider a final Harbor Ledger evidence packet containing a sale schedule, two receipt records, a distribution listing and an expense summary. The sale schedule shows gross consideration of 14. One receipt record shows nine and the other four, making cash of 13. The expense summary shows a one-unit fee. It is tempting to declare the difference explained: 14 minus one equals 13. The arithmetic works, but the documents have not yet established that this fee belongs to that sale or was withheld from its proceeds.
Suppose the summary's fee identifier instead belongs to a different portfolio and was paid separately. The apparent reconciliation collapses. Alternatively, a settlement statement could identify the same sale and explicitly show the one-unit deduction, completing the bridge. A third possibility is that the second receipt represents a partial payment with one still outstanding. The identical numerical gap can reflect several different states. The missing evidence is the link between the sale, the deduction and the receipts, not another general description of liquidation.
This gives an editor a concrete question to pursue: what record identifies the composition of the net settlement for sale H-14? It is more precise than asking why a unit disappeared. It neither presumes misconduct nor accepts a balancing number simply because it fits. A public explanation can state that gross consideration and identified cash differ by one, describe the possible classifications without selecting one, and mark the reconciliation incomplete.
The distribution listing can be tested separately. Imagine it authorizes ten, the payment register shows ten sent, and the recipient evidence establishes nine received with one unresolved at the cutoff. Those are three statuses, not three competing estimates of an identical event. A headline can responsibly report ten authorized or nine confirmed received, provided it says which. It cannot turn the authorization into proof of completed receipt merely because the final amount is expected to match.
The resulting article should preserve a small, reproducible chain: the identified estate and reporting period; the transaction perimeter; the original amounts and units; the movement from assets to cash; the established distribution base; and the distinction between realized results and remaining estimates. Each unresolved item should have a specific missing relationship or field. That makes the account correctable when another statement appears, instead of forcing the next reporter to rebuild it from an isolated headline.
Following a receivership is ultimately about retaining distinctions until the records justify combining them. A buyer's discount is not automatically an insurance-fund loss. An estimated asset value is not a receipt. A distribution from an estate and its receipt by a fund are not two recoveries. A change in a percentage may come from its denominator. Once those boundaries are visible, the public can ask a more demanding question than how large a number sounds: what, exactly, does this set of records establish?