Point-in-Time Data for Academic Finance Research: A Reproducibility Guide

In empirical finance, a backtest should use only information available on the simulated decision date. If it uses financial statements based on fiscal period-end dates, current database values, or later restatements, the model may incorporate facts that had not yet been published.

That is lookahead bias. It can affect estimated predictability, reported coefficients, and reproducibility.

Point-in-time, or PIT, fundamentals record when each value became public and allow researchers to reconstruct the information set available as of a historical date. For accounting-based asset-pricing signals, PIT treatment can be an important part of research design.

Why ordinary fundamentals may not be enough

Suppose a company has a fiscal year ending December 31. Its annual revenue does not necessarily become public on December 31. The filing may arrive weeks later, and an amendment or restatement may subsequently change the value.

A conventional panel might contain:

• the fiscal period end;
• the latest known value;
• a standardized concept;
• the company identifier.

That panel may be useful for descriptive tasks, but it does not necessarily answer the historical question: What value could a researcher have observed on the portfolio formation date?

A PIT row needs an availability date or timestamp. Tradevo Data exposes first_filed, defined as the date the value became public. Its API applies first_filed <= as_of, including same-day filings. Each row also contains:

• original_value: the first-reported, PIT-safe value;
• latest_value: the current revision;
• restated: a flag for changes above 0.5% on the same tag, including amendments;
• qa_status: a data-quality field.

The distinction between original and latest values matters. Replacing the original observation with a later revision can introduce future information even if the fiscal period itself is correctly lagged.

For a deeper treatment, see https://tradevodata.com/blog/lookahead-bias-fundamental-backtests?utm_content=blog-point-in-time-data-for-academic-finance-research.

PIT design belongs in the methods section

A reproducible paper should state more than “fundamentals were lagged.” A fixed lag is a convention, not proof that the information was public.

A stronger methodology documents:

1. Availability rule: Which filing date or timestamp determines eligibility?
2. Boundary rule: Are filings published on the formation date included?
3. Revision policy: Does the study use first-reported or latest values?
4. Fiscal-period policy: How are annual observations aligned across firms?
5. Concept mapping: Which source tags map to each standardized variable?
6. Missingness policy: Are missing values dropped, imputed, or interpreted economically?
7. Universe policy: Does the source include delisted companies and non-US issuers?
8. Snapshot procedure: Can another researcher reconstruct the same extract later?

Tradevo Data's server-side endpoint is:

/v1/fundamentals?ticker&as_of[&concept][&period=annual|quarterly]

Annual is the default. Bulk access is available through /v1/download and /v1/snapshot?as_of, each accepting period=annual|quarterly. The snapshot route makes the historical cutoff explicit. Parquet format is not included.

What Tradevo Data covers—and what it does not

Tradevo Data provides point-in-time US equity fundamentals derived from public-domain SEC EDGAR filings. Its annual dataset contains 632,466 point-in-time rows across 5,168 US companies, with 16 annual concepts and up to 12 fiscal years. Its quarterly dataset contains 1,089,635 rows across 5,382 companies and seven quarterly concepts. Across the data, 37,804 restatements are labeled.

The annual concepts are Revenue, NetIncome, Assets, StockholdersEquity, OperatingCashFlow, EPSDiluted, DilutedShares, GrossProfit, OperatingIncome, PretaxIncome, IncomeTaxExpense, CapitalExpenditures, CashAndCashEquivalents, CurrentAssets, CurrentLiabilities, and NetPPE.

Quarterly data is also available with seven concepts. Q4 is reported where tagged or derived and labeled where supported. Derived Q4 EPS and shares are excluded.

The limitations are material:

• US companies only;
• no non-US coverage;
• no delisted-company coverage;
• no TTM fields;
• a focused concept set rather than the breadth of a large institutional platform;
• no included Parquet format.

These constraints should appear in a paper's data section. In particular, the absence of delisted companies may create survivorship concerns depending on sample construction and the research question.

Tradevo Data versus WRDS and Compustat

WRDS is an institutional research platform through which universities and other organizations may access datasets such as Compustat, subject to their subscriptions and licenses. It may be the stronger choice when a study requires broader historical coverage, standardized identifiers across databases, delisted securities, international accounting data, or established institutional workflows.

Tradevo Data is a narrower, budget-tier PIT source for researchers who need US filing fundamentals without requiring university-gated access.

This is not an argument that one source universally replaces the other. The appropriate source depends on the identification strategy, required universe, licensing constraints, and replication plan. Sharadar, Tiingo, and QuantConnect are also credible alternatives; compare their current documentation, coverage, PIT methodology, and pricing pages directly.

Public proof before signup

Tradevo Data publishes a no-signup proof pack at:

https://github.com/christianpichichero-max/pit-fundamentals

It contains 225 rows for five companies over their latest three fiscal years, plus the full methodology. Researchers can inspect the schema, test availability logic, and evaluate whether the concept mapping is suitable before entering payment details.

On reliable-filing rows in that five-company proof pack only, measured lookahead averaged 35.2 days and reached a maximum of 48 days. These figures are not evidence about the full dataset. They demonstrate within the sample how filing availability can differ from fiscal period-end dating.

More background on the data model is available at https://tradevodata.com/blog/point-in-time-fundamentals-data?utm_content=blog-point-in-time-data-for-academic-finance-research.

A practical research workflow

A defensible workflow can be simple:

1. Pre-register or document the formation-date rule.
2. Query or snapshot the data with the required as_of date.
3. Use original_value for the historical information set unless the research question explicitly concerns revised accounts.
4. Retain first_filed, restated, and qa_status in intermediate files.
5. Archive query parameters, code, and extraction dates.
6. Report coverage exclusions, especially delisted and non-US companies.
7. Run sensitivity checks using alternative availability rules where justified.
8. Make a small, license-compliant replication extract available when possible.

The card-backed seven-day free trial covers 10 companies, a rolling three-year history, and 100 requests per day, without bulk access. It renews at $29/month unless canceled. Pro provides the complete available history, 5,000 requests per day, and bulk access through /v1/download and /v1/snapshot?as_of. Documentation is at https://tradevodata.com/docs?utm_content=blog-point-in-time-data-for-academic-finance-research, and the trial is available at https://tradevodata.com/?utm_content=blog-point-in-time-data-for-academic-finance-research.

When WRDS or another institutional source wins

Use WRDS, Compustat, or another appropriate institutional dataset when the paper requires:

• delisted-company coverage to address survivorship bias;
• non-US or cross-country samples;
• history beyond Tradevo Data's available horizon;
• accounting fields outside Tradevo Data's focused concept set;
• linked returns, analyst estimates, ownership, executive, or security-master data;
• identifiers and conventions already standard in a target literature;
• a licensed source mandated by coauthors, reviewers, or the institution.

Researchers should inspect the exact subscribed product, documentation, PIT fields, and licensing terms rather than assume every dataset exposed through a platform has identical historical properties. For current access and cost information, see each provider's pricing page.

When to build it yourself

Building directly from SEC EDGAR can be the better choice when the extraction pipeline is itself a research contribution, when unusual XBRL tags are central to the hypothesis, or when complete control over transformation rules matters more than implementation time.

A credible build may need to handle filing chronology, amendments, duplicate facts, taxonomy changes, units, dimensions, fiscal calendars, concept mapping, accession-level provenance, and historical snapshots. It may also require tests for revisions and malformed filings.

The advantage is transparency and customization. The disadvantage is the required data-engineering work. One approach is to validate a managed dataset against raw filings for a stratified sample and publish the audit code.

The standard is reconstructability

Using a familiar database does not by itself make academic finance reproducible. The core question is whether another researcher can reconstruct the information set available on each simulated date and understand the transformations applied.

Tradevo Data offers a focused option for annual and quarterly US filing fundamentals, with explicit availability dates, original and revised values, restatement labels, snapshots, and a public methodology. It is not a substitute for broad institutional databases in every research design. It is an inspectable budget-tier option when its coverage matches the paper.

Start by auditing the free sample at https://github.com/christianpichichero-max/pit-fundamentals. If the schema and methodology fit the study, the card-backed seven-day free trial is available at https://tradevodata.com/?utm_content=blog-point-in-time-data-for-academic-finance-research and renews at $29/month unless canceled.

Not investment advice.