Most UK undergraduate accounting dissertations do not recruit participants at all — they build a sample of companies, filings or firm-years from archival data. That single fact changes what “sample size” even means for the subject: the question is not how many people you can recruit, but how many companies, years or filings your population and access actually support. Getting this framing right from the proposal stage saves a rewrite of the whole methodology chapter later.
Why Is Accounting Sampling Different From Most Other Subjects?
A psychology or business dissertation samples people; an accounting dissertation more often samples companies, and the unit of analysis is frequently a company-year rather than a single entity — a study of ten companies across five years of filings is a sample of fifty firm-years, not ten. Getting this distinction right in your methodology chapter avoids a common examiner correction: confusing the number of companies with the actual number of observations your statistical tests are run against. It also changes how you should read the word “sample” itself when planning your dissertation: it refers to the observations your analysis actually runs over, not simply the count of named organisations you mention.
How Do I Define My Population?
Start by naming four things explicitly: the sector (or sectors), the size band (e.g. FTSE 350, AIM-listed, a defined SME turnover band), the listing status (public vs private, since private company filings are far less detailed), and the time window. A vague population (“UK companies”) is not workable; a defined one (“FTSE 350 non-financial companies, 2021–2025”) is. Writing the population definition as a single sentence you could hand to a stranger and have them independently reproduce your company list is a useful, quick test of whether it is actually specific enough.
| Population type | Where the filings come from | Typical constraint |
|---|---|---|
| Listed companies (FTSE 350, AIM) | Companies House + company investor-relations annual reports | Detailed, standardised disclosure; smaller total population |
| Private/SME companies | Companies House abbreviated/micro-entity accounts | Far less disclosure detail; larger population but thinner data per firm |
| Sector-specific (e.g. retail, banking) | Companies House filtered by SIC code + sector-specific regulator data | May need a second source for sector-specific metrics regulators publish separately |
How Do I Access the Data?
Companies House gives free public access to filed accounts and company details for every UK-registered company, but it does not check the accuracy of what companies file — treat filed figures as the company’s own reported position, not independently verified fact, and say so explicitly in your methodology. For panel-style research needing many companies’ data pre-structured for analysis rather than read filing-by-filing, FAME (via Bureau van Dijk) is the standard route UK university libraries commonly provide access to — check your own library’s database list, since access is institutional rather than a personal subscription. Our guide to data analysis software for an accounting dissertation covers what to do with the data once you have it, including where FAME exports fit against Excel, SPSS, Stata, R and Python.
How Do I Build a Defensible Sample From a Population Too Large to Use in Full?
Purposive (judgemental) sampling is the norm in accounting archival research, not random sampling — you deliberately select companies matching your population criteria rather than drawing randomly, because the research question usually concerns a specific, defined group (a sector, a size band, a regulatory regime) rather than a representative cross-section of the whole economy. State your inclusion and exclusion criteria explicitly: which SIC codes, which size band, which listing status, and — critically — how you handled companies that delisted, merged or stopped filing partway through your window, since silently dropping them introduces survivorship bias into your results.
What Does “Sample Size” Actually Mean for Archival Accounting Research?
There is no participant-recruitment equivalent of a power calculation here — the relevant question is whether your number of firm-year observations supports the statistical test you plan to run, not whether you have “enough” companies in an abstract sense. A panel regression across 40 companies over 5 years (200 firm-years) supports considerably more statistical power than a cross-sectional comparison of the same 40 companies in a single year. State the unit of analysis explicitly in your methodology — firms, firm-years, or individual filings — since this is exactly the distinction examiners check first.
A Worked Example: Building a Sample From Scratch
Illustrative, not a real study: a dissertation asking whether ESG disclosure quality relates to audit fees among UK-listed retailers might define its population as “all companies listed on the London Stock Exchange under the retail SIC code range, 2020–2025.” From that population, the sample is built by excluding companies that delisted before the window closed (noting this exclusion explicitly, and discussing the survivorship-bias implication in the limitations chapter), giving a final purposive sample of, say, 25 companies across 5 years — 125 firm-year observations. Data comes from each company’s own annual report ESG disclosures (hand-collected or via FAME where available) and Companies House–filed audit fee disclosures. The methodology chapter states this exact chain: population definition, exclusion criteria, final sample, and unit of analysis — in that order, so an examiner can follow the reasoning without asking you to explain it in the viva.
How Should This Appear in Your Methodology Chapter?
Write the population, sampling method and final sample as three distinct, explicitly labelled sub-sections rather than folding them into one paragraph — markers specifically look for each element to be separately identifiable. State the population definition first, then the sampling approach and any exclusions, then the final sample with its unit of analysis (firms vs firm-years) stated in the same sentence as the number itself, so there is no ambiguity about what the number actually counts. A short summary table — population, sampling method, exclusions, final N, unit of analysis — at the end of this section gives an examiner everything they need in one place without re-reading three paragraphs of prose to check it.
What If My Dissertation Needs Primary Data Instead?
Some accounting dissertations do involve primary data — a survey or interviews with practising accountants, auditors, or SME finance staff about a practice, attitude or judgement question archival data cannot answer (for example, how accountants interpret a new standard’s ambiguous wording). Where this is your design, the sample-size logic reverts to the more familiar participant-based reasoning covered in our general guide to what sample size an undergraduate dissertation needs. Access is usually the binding constraint rather than sample-size theory: professional bodies and individual firms are commonly approached through personal or university contacts rather than open recruitment, and response rates from busy practitioners are typically lower than from a general population sample — build this into your timeline from the proposal stage, not as an afterthought once recruitment has started.
How Does Enforcement and Regulatory Data Fit In?
Where your dissertation examines audit quality, compliance or reporting failures, the Financial Reporting Council (FRC) publishes its enforcement cases and their sanction outcomes, which is a citable, free, publicly available source for that specific angle — distinct from the company-level filing data covered above, and worth checking directly rather than relying on a secondary summary of what the FRC has published.
What Mistakes Cost the Most Marks Here?
- Confusing the number of companies with the number of observations. A panel of 40 companies over 5 years is 200 firm-year observations — state which number your statistical test actually uses.
- Silently dropping delisted or merged companies. This introduces survivorship bias; state how you handled attrition in your sample explicitly.
- Treating Companies House filings as independently verified fact. State clearly that filed figures are the company’s own reported position.
- Applying participant-based sample-size logic to archival research. A “power calculation” framed around number of participants does not translate directly to a firm-year panel design — the relevant question is whether your observation count supports your specific statistical test.
- Assuming FAME access without checking your library first. It is institutional, not a personal subscription — confirm access before your proposal assumes it.
- Folding population, sampling and final sample into one vague paragraph. Keep the three elements explicitly separate and labelled so an examiner can follow the chain without asking for clarification.
Once your population and sample are defined, Tesify drafts your methodology chapter around the design you specify. Everything stays 100% written by you, and it is free to start.
Frequently Asked Questions
Do I need ethics approval for an archival accounting dissertation?
Usually a lighter form than primary research involving people, since you are not collecting data from participants — but check your department’s specific policy, since secondary-data research can still require a declaration.
How many companies do I need for an accounting dissertation?
There is no fixed number — what matters is whether your total firm-year observations support the statistical test you plan to run, and whether your population definition (sector, size, listing status) is defensible and clearly stated.
Is Companies House data reliable enough to build a dissertation around?
Yes, as a source of what companies have officially filed, but state clearly that Companies House does not independently verify the accuracy of what is submitted.
What is FAME and do I need it?
FAME (via Bureau van Dijk) is a commercial database that structures UK and Irish company financial data for analysis; it is useful where you need many companies’ data pre-organised rather than reading individual filings, and is usually accessed through your university library.
What is survivorship bias in accounting research?
The distortion that results from only including companies that survived your whole study window, silently excluding those that delisted, merged or ceased filing — state explicitly how you handled this in your sample.
Can I survey practising accountants for my dissertation?
Yes, where your research question needs a judgement or attitude archival data cannot answer, but expect access and response rates to be the binding constraint rather than sample-size theory, and build recruitment time into your plan early.
What does “firm-year” mean?
One company observed in one year — a sample of 40 companies over 5 years produces 200 firm-year observations, which is the number that determines your statistical power, not the count of companies alone.
Where can I find data on accounting or audit enforcement failures?
The Financial Reporting Council publishes its enforcement cases and sanction outcomes directly and freely — a more targeted source than general company filings for this specific angle.
Should population, sampling method and final sample be written as separate sections?
Yes — keep the three elements explicitly labelled and separate rather than combined into one paragraph, so an examiner can follow your reasoning chain without needing to ask for clarification.
Does the same firm-year logic apply to a finance dissertation?
Broadly yes — finance dissertations using company or market data face the same population, sampling and unit-of-analysis questions, though the specific data sources and metrics differ by topic.