Restated vs Recast Financials: Why Historical Numbers Change—and Which Version Should You Use?

The ways historical financials can change

Several different events can make an old period look different in a later filing.

They should not be treated as interchangeable.

There is also a filing term that often causes confusion:

Amendment

A 10-K/A or 10-Q/A tells you that a filing was amended.

It does not by itself tell you why.

An amendment can contain a financial restatement, correct disclosures, add omitted information, revise exhibits, or address something else entirely.

Restatement: when the accounting itself was wrong

A restatement corrects an error in previously issued financial statements.

That can involve:

• revenue recognized in the wrong period;
• an incorrect reserve;
• consolidation errors;
• depreciation or capitalization mistakes;
• improper expense recognition;
• tax accounting;
• classification errors that require correction.

The distinguishing feature is that the previously issued financial statements contained an accounting error.

The later filing is not merely reorganizing the old information.

It is correcting it.

Big R and little r restatements

Not every accounting error produces the same type of restatement.

The SEC has described two commonly used categories.

"Big R" — reissuance restatement

When an error is material to previously issued financial statements, those statements must be corrected. The SEC notes that this is commonly called a reissuance restatement or Big R restatement.

Previously issued financial statements can no longer be relied upon.

"little r" — revision restatement

An error may be immaterial to the previously issued financial statements but still need to be corrected because either correcting it entirely in the current period or leaving it uncorrected would materially misstate the current period.

In that case, the company can revise the comparative prior-period information in a current filing and disclose the error.

This is commonly called a revision restatement or little r restatement.

The SEC emphasizes that both are corrections of errors in previously issued financial statements.

That distinction matters because an investor should not assume:

Historical figures can be revised even when the correction is not a Big R restatement.

Primary source: SEC Office of the Chief Accountant, Assessing Materiality: Focusing on the Reasonable Investor When Evaluating Errors
SEC primary source ↗

Valeant: two versions of the same 2015 results

Valeant's Philidor episode is a useful case because it shows how quickly the meaning of "historical earnings" can become version-dependent.

Its July 28, 2015 Q2 filing contained the figures investors had at the time.

A later Q2 filing in August 2016 presented those same six months on a restated basis.

SourceState preserves both filing vintages:

SIX MONTHS ENDED JUNE 30, 2015

                               Original 10-Q        Later restated view
Revenue                           $4.9233bn              $4.9025bn
Operating income                   884.9m                 912.5m
Net income                          22.9m                  46.9m
Basic EPS                            0.06                   0.13

The later filing explicitly described a "Restatement of Previously Issued Financial Statements."

Among the adjustments was Valeant's treatment of sales to Philidor.

Before Philidor was consolidated, Valeant had recognized certain revenue when products were delivered to Philidor—what the company called a sell-in basis.

The review later concluded that certain transactions should instead have been recognized when Philidor dispensed the products to patients—a sell-through basis. Valeant also concluded that collectability had not been reasonably assured when some of the revenue was originally recorded.

That is more than a taxonomy or presentation issue.

It changes the answer to a basic investment question:

The restatement also illustrates why one revised line cannot summarize an entire correction. Revenue fell in the six-month comparison above, while operating income, net income, and EPS increased after the full set of adjustments.

The point-in-time consequence

Suppose a quantitative screen had been run on July 29, 2015.

The correct inputs are the figures Valeant had actually filed by then.

Using the later restated EPS of $0.13 would give the model information that did not exist in the public record on that date.

The later number may be the better corrected history.

It is the wrong point-in-time history.

Primary sources:

• Valeant Q2 2016 Form 10-Q, Note 2 — Restatement of Previously Issued Financial Statements
  SEC primary source ↗

• Valeant 2015 Form 10-K — Restatement Background
  SEC primary source ↗

Recast: when the business history is presented differently

A recast often solves a different problem.

The company may reorganize its segments, sell a business, or change how management views the organization.

To make the current structure comparable with prior years, the company presents historical periods using the new structure.

The old accounting may have been perfectly valid.

The question changed.

This creates an unusual kind of historical truth:

The original structure answers:

The recast structure answers:

Both can be useful.

Fiserv: same company, same revenue, different segments

Fiserv provides a particularly clean example.

In its original FY2023 filing, the segment presentation was:

Acceptance                      $8.132bn
Fintech                          3.171bn
Payments                         6.696bn
Corporate and Other              1.094bn
                               ---------
Consolidated revenue           $19.093bn

A year later, after a segment realignment, the FY2023 comparative was recast as:

Merchant                        $8.722bn
Financial                        9.101bn
Corporate and Other              1.270bn
                               ---------
Consolidated revenue           $19.093bn

The consolidated total is identical:

$19.093 billion.

The historical business activity did not suddenly change.

The analytical structure did.

The later filing said 2023 and 2022 results "have been recast to reflect the Segment Realignment."

This is the opposite of the Valeant example.

Valeant changed historical accounting values because the original accounting contained errors.

Fiserv changed the historical lens through which the same consolidated business was presented.

For a current segment model, the recast structure is usually more useful.

For research into what investors saw in February 2024, the original Acceptance / Fintech / Payments structure is the relevant one.

Same total. Different history.

Primary filings:

• Original FY2023 Form 10-K, filed February 22, 2024
  SEC primary source ↗

• FY2024 Form 10-K with recast FY2023 segments, filed February 20, 2025
  SEC primary source ↗

Discontinued operations can rewrite the historical income statement

A company does not need to discover an accounting error for historical revenue to change.

A later transaction can alter how an old period is presented.

Sabre provides a clean example.

Its original FY2024 10-K reported revenue of:

$3.029565 billion

After Sabre sold its Hospitality Solutions business in 2025, its FY2025 filing presented Hospitality Solutions as discontinued operations for all periods shown.

The FY2024 continuing-operations revenue shown in that later filing became:

$2.744845 billion

The business activity that occurred in 2024 did not disappear.

The later filing separated part of it from continuing operations because of a transaction that happened afterward. The company also noted that general corporate overhead that did not qualify as discontinued operations remained outside the discontinued-operations presentation.

This creates another point-in-time distinction:

WHAT AN INVESTOR SAW AFTER FY2024
Revenue including the business then presented within continuing operations

vs.

HOW FY2024 IS PRESENTED AFTER THE 2025 SALE
Continuing operations separated from discontinued operations

Neither representation should silently overwrite the other if the goal is to reconstruct historical information states.

Primary filings:

• Sabre FY2024 Form 10-K, filed February 20, 2025
  SEC primary source ↗

• Sabre FY2025 Form 10-K, filed February 18, 2026
  SEC primary source ↗

Retrospective adoption can move history between lines

Sometimes the economics remain unchanged while classification shifts.

DXC provides an unusually clean example.

For the quarter ended June 30, 2017, the original filing reported:

Net cash from operating activities    $534m
Net cash from investing activities     859m

In a later comparative filing, after retrospective adoption of ASU 2016-15:

Net cash from operating activities    $519m
Net cash from investing activities     874m

Exactly $15 million moved from operating to investing.

The later filing explained that retrospective adoption caused:

• a $15 million decrease in operating cash flow;
• a corresponding $15 million increase in investing cash flow.

The reclassification did not change the combined operating-plus-investing cash flow represented by those two lines.

For a historical cash-flow-factor backtest, however, the classification can matter materially.

A strategy screening for operating cash flow in August 2017 should not quietly receive the later $519 million figure if the $534 million classification was what investors actually had.

Primary filings:

• DXC Q1 FY2018 Form 10-Q, filed August 9, 2017
  SEC primary source ↗

• DXC Q1 FY2019 Form 10-Q, filed August 8, 2018
  SEC primary source ↗

An amended filing is not automatically a restatement

Assured Guaranty's Q2 2011 filing shows an amendment that did change the financial statements.

The original 10-Q was filed August 9.

A 10-Q/A followed on November 14.

Among the changes SourceState captures:

                                  Original        Amended/restated
Net investment income             $100.827m          $101.153m
Total expenses                     215.645m           206.650m
Total assets                       $19.239bn          $19.300bn

The amended filing presented "as previously filed" and restated values and attributed the correction to intercompany eliminations involving consolidated financial-guaranty variable-interest entities and insurance subsidiaries.

That is a substantive amendment.

But the lesson is broader:

Some amendments change financial statements.

Others do not.

A historical database needs to inspect what changed, not just the form type.

Primary filings:

• Original Q2 2011 Form 10-Q
  SEC primary source ↗

• Q2 2011 Form 10-Q/A
  SEC primary source ↗

Hertz: the restatement that kept expanding

Some restatements begin with one accounting issue and end as a much broader reconstruction of the reporting system.

Hertz is a useful example.

In 2014, while preparing its first-quarter filing, Hertz identified errors involving the capitalization and timing of depreciation for certain non-fleet assets and allowances for doubtful accounts in Brazil.

The review did not stop there.

The company subsequently identified additional errors involving damaged-vehicle receivables and lease-restoration obligations.

On June 6, 2014, Hertz disclosed that its Audit Committee had concluded that the 2011 financial statements should no longer be relied upon and needed to be restated.

The review ultimately expanded across multiple years.

When Hertz filed its 2014 Form 10-K, it said the investigation and accounting review had identified material misstatements in the 2011, 2012, and 2013 consolidated financial statements.

The restatement reduced pre-tax earnings by approximately:

• $81 million in 2012
• $72 million in 2013

The episode illustrates a different reason versioned financial history matters.

A restatement is not always a neat event in which one old number is replaced by one new number on one date.

It can unfold over months:

Error discovered
      ↓
Filing delayed
      ↓
Prior statements deemed unreliable
      ↓
Scope of review expands
      ↓
Multiple years reconstructed
      ↓
Restated financial statements filed

During that interval, "the historical financials" are not a single static dataset.

There are original filings, non-reliance disclosures, unresolved periods, and ultimately restated numbers.

A robust data system should distinguish those states rather than rewrite the archive after the fact.

Primary sources:

• Hertz June 2014 Form 8-K — non-reliance determination
  SEC primary source ↗

• Hertz 2014 Form 10-K — description of restatement matters
  SEC primary source ↗

Latest-known history vs point-in-time history

This is the distinction that ties all of these examples together.

Latest-known history

Ask:

For Valeant's 2015 results, use the corrected restated values.

For Fiserv's 2023 segments, use the later recast structure if you want a segment history consistent with the current organization.

For Sabre, use the continuing-operations presentation when analyzing today's historical continuing business.

Latest-known data is extremely useful.

It gives you a clean current historical series.

But it is not necessarily what investors knew at the time.

Point-in-time history

Ask:

For a July 2015 Valeant screen, use the July 2015 filing.

For a February 2024 Fiserv segment analysis, use the segment structure available then.

For a 2017 DXC cash-flow signal, use the pre-ASU-2016-15 classification that had actually been filed.

This requires the database to preserve filing vintages.

Conceptually:

Fiscal period: FY2023
        │
        ├── Filing vintage A
        │      what investors originally saw
        │
        ├── Filing vintage B
        │      later recast / restated comparative
        │
        └── Filing vintage C
               possibly another later revision

The fiscal period alone is not sufficient to identify the historical observation.

The look-ahead trap

Imagine a backtest run as of August 1, 2015.

The model asks:

A modern database that stores only the latest-known history might answer:

$0.13

But the filing available at the time showed:

$0.06

The model has just learned the future.

The same problem can occur even when the consolidated number never changes.

Suppose a historical factor model screens companies by segment growth.

If it uses Fiserv's later recast 2023 segment structure in a simulation dated before that structure was disclosed, the model is applying today's organizational knowledge to yesterday's investment decision.

That may be the correct dataset for a current historical comparison.

It is not a true point-in-time backtest.

Restatement, recast, and reclassification are not synonyms

A useful shorthand:

Restatement

Example: Valeant's revenue-recognition correction.

Recast

Example: Fiserv's segment realignment.

Discontinued-operations presentation

Example: Sabre.

Reclassification / retrospective presentation

Example: DXC moving $15 million from operating to investing cash flow after retrospective adoption.

Amendment

These distinctions matter because they answer different questions about historical truth.

Which version should you use?

There is no single answer.

It depends on the task.

The important thing is to make the choice explicit.

A dataset labeled simply:

can hide too much.

A better representation knows:

Period:
FY2023

Value:
$X

Filing:
accession ...

Filed:
date ...

Version:
original / later comparative

Status:
as reported / restated / recast / reclassified

That makes the history inspectable.

How SourceState treats revised history

SourceState stores filings as separate accession-level fact sets rather than silently overwriting an earlier filing with a later comparative.

That makes it possible to preserve:

• the original value;
• the later restated or recast value;
• the filing date for each version;
• the source filing;
• the reported label and concept;
• dimensional structure;
• the reason for the change where it can be established from the filing.

For Valeant, that means both the July 2015 filing and the later restated comparative remain available.

For Fiserv, the original FY2023 segment structure and the later recast FY2023 structure can coexist.

For DXC, the original cash-flow classification and the later retrospective presentation can both remain part of the historical record.

This supports two different queries:

LATEST-KNOWN
Give me the best current historical series.

POINT-IN-TIME
Give me the latest information that was actually public
as of a specific historical date.

Those queries should not necessarily return the same data.

That is not a defect.

It is the reality of financial reporting.

The simplest way to remember the difference

A restatement says:

A recast says:

A reclassification says:

A point-in-time database asks:

A latest-known database asks:

Serious historical financial data should be able to answer both.