Written and maintained by CASRAI Editorial Board
Published
On this page
The World Bank Enterprise Surveys (WBES) are firm-level surveys of business owners and top managers, run by the World Bank Group’s Enterprise Analysis Unit, that collect comparable data on the business environment, firm performance, access to finance, corruption and infrastructure across more than 160 economies. Researchers use the microdata to compare firms and countries; the sampling design and survey type decide how far the numbers can be generalised.
What are the World Bank Enterprise Surveys?
The Enterprise Surveys team describes its formal-sector survey as its main product, and says the surveys cover more than 160 economies with plans to reach 180. The microdata library describes them as firm-level surveys carried out since 2002, with data collection centralised in the Enterprise Analysis Unit since 2005-06. Respondents are business owners and top managers, and the surveys keep respondent anonymity.
Fieldwork is not done by the World Bank itself. NORC, for example, describes its role as a contractor running WBES interviews, and its project page lists 2024 fieldwork in Ecuador (360 interviews), Uruguay (360) and the United States (3,030). The same page says that, under the World Bank’s B-Ready initiative, the surveys are to run every three years in each country, and that follow-up interviews with firms from earlier rounds are part of the 2025 work.
For a researcher the practical meaning is simple: WBES is a design-based survey programme. The numbers are only as good as your handling of strata, weights and survey type, which is why this guide concentrates on the design rather than on any single indicator. If you mainly need country-level indicators rather than firm microdata, the World Bank Data guide and the Worldwide Governance Indicators guide cover those sources.
Formal, informal and micro surveys: which one is which?
Formal-sector Enterprise Surveys
The formal survey is the core WBES product. Its eligible population is registered firms in the ISIC Rev.4 sectors the methodology page lists, with at least five workers and at least 1% private ownership; since August 2022 Section M is also eligible. Interviews are face to face.
Informal-sector Enterprise Surveys
Informal surveys target entrepreneurs of unregistered businesses and use a different design. Because no register exists, the sample is drawn in two stages (described below), and the interview is shorter, at 20-25 minutes.
Micro surveys and topic surveys
Micro surveys cover firms with one to four workers and follow the formal methodology; the methodology page lists 35 economies for them. The programme has also run COVID-19 follow-up surveys in 2020-2022 and topic surveys, such as innovation and tourism. The data table below sets the three main tracks side by side.
How is the WBES sample designed?
Formal surveys: stratified random sampling
The formal surveys use stratified random sampling. Firms are stratified by size (small, 5-19 workers; medium, 20-99; large, 100 or more), by business sector and by region within the economy. The ideal sampling frame is the national statistics office’s firm universe; where that is unavailable, the frame is built from tax or licensing agencies and business associations. Frame quality therefore varies by country, which matters when you compare countries.
Informal surveys: adaptive cluster sampling
Informal surveys use a two-stage design. First, Block Areas in a grid laid over urban centres are selected at random, often within strata, using Adaptive Cluster Sampling. Second, businesses inside the selected blocks are drawn at random. Weights follow the same two stages: first-stage weights come from the cluster design, and second-stage weights are adjusted for assumed refusals and unavailable businesses.
Panel firms
The programme aims to re-interview at least half of the firms from the previous survey in a new round. Because a panel firm’s chance of being in the sample differs from that of a fresh draw, the sampling weights are adjusted for the altered inclusion probabilities.
Stratifying on size, sector and region serves a clear analytical purpose: it guarantees that each group of interest appears in the sample in enough numbers to be analysed on its own, which a simple random draw from a skewed firm population would not. The cost is that the sample no longer mirrors the population, and that is exactly what the weights repair. When you pick a research question, check first whether the strata match the groups you want to compare; a comparison across, say, large firms in one sector may rest on very few interviews. The implementation report that accompanies each dataset is where sampling details and variable names for that country are documented, so read it before the questionnaire.
What does the questionnaire cover?
The formal questionnaire spans more than 15 topics, including firm characteristics, gender participation, access to finance and annual sales. About 90% of the questions are fact-based; the remaining 10% or so collect opinions on obstacles to growth. That split is useful for choosing variables: fact-based items such as sales or employment are better candidates for hard comparisons than opinion items, which reflect respondents’ perceptions. The informal questionnaire asks about business characteristics, owner background, electricity and water use, supplier and customer linkages, business practices, access to finance, workforce, and sales and assets.
Why weights matter: an illustrative calculation
Stratified samples deliberately over- or under-represent groups, so a raw average of the sample is not an estimate for the whole population. The sampling weight corrects that. Illustrative composite, not WBES data: suppose a country’s frame holds 900 small firms and 100 large firms, and the sample takes 30 of each.
- Small-firm design weight = 900 / 30 = 30 firms represented per sampled firm.
- Large-firm design weight = 100 / 30 = 3.33 firms represented per sampled firm.
- If sampled small firms report a mean of 40% capacity use and large firms 70%, the unweighted mean is (40 + 70) / 2 = 55%.
- The weighted mean is (900 × 40 + 100 × 70) / 1,000 = (36,000 + 7,000) / 1,000 = 43%.
The 12-point gap shows why you should always apply the weight variable that ships with the dataset. In R the same logic is weighted.mean(x, w), where x is your indicator and w the dataset’s weight column; in Python, numpy.average(x, weights=w). For standard errors you also need the design information (strata and weights), not just the point estimate; the guide to sampling errors explains why. One further rule from the methodology page: since May 2025 all informal-survey indicators use the “wweak” weights, which include refusals and unavailable businesses.
How to access the data
Datasets are distributed through the World Bank’s data portal and the Microdata Library, together with implementation reports on fieldwork, sampling and variable names. The microdata library lists country, aggregated and panel datasets plus documentation, and registration is required but free. Informal-survey datasets require a portal login. Downloadable resources include indicator descriptions, the WBES manual, the questionnaire and a replication archive containing the firm-level standardised and indicator databases.
Limitations and reading guidance
The table below turns the design rules into reading guidance. Each row is a restatement of a statement on the methodology page, followed by what it means for your analysis.
| Design feature | What it means for your analysis |
|---|---|
| Formal sample covers registered firms with 5+ workers | Results do not describe unregistered businesses or firms with fewer than five workers (micro and informal surveys cover those groups). |
| Sample stratified by size, sector and region | Use the weights; unweighted sample shares do not match the population. |
| Sampling frame depends on national sources | Frame coverage differs between countries, so check the implementation report before ranking economies. |
| About 10% of questions are opinions | Treat obstacle rankings as perceptions, not measured conditions. |
| Informal surveys before 2015 used a convenience sample | Do not treat pre-2015 informal data as representative or pool it with later rounds without care. |
| Not every survey followed the Global Methodology | The methodology page advises caution when comparing raw data across surveys that did and did not follow it. |
Two further habits help. First, check which survey year and round a country file belongs to before building a time series; panel firms exist only where a follow-up round re-interviewed them. Second, report the sample sizes and weights you used. For wider survey-design context, compare the European Social Survey guide (a household survey with a different design) and the advantages and disadvantages of surveys, and see sampling methods for the vocabulary used here.
How to cite Enterprise Surveys data
The methodology page does not prescribe a citation format, and the microdata collection links separate terms-of-use and citation pages that should be followed for the dataset you download. Whatever style your journal uses, record these elements for each file: the economy, the survey year, the WBES round, the survey type (formal, informal or micro), the publisher (World Bank Group, Enterprise Analysis Unit), the dataset’s persistent link, and the date you downloaded it. Naming the survey type and weight variable in your methods section lets readers reproduce your numbers.
Sources
- World Bank Group Enterprise Surveys: Methodology, checked 2026-11-17
- World Bank Group Enterprise Surveys: About the surveys, checked 2026-11-17
- World Bank Microdata Library: Enterprise Surveys collection, checked 2026-11-17
- NORC: World Bank Enterprise Surveys, checked 2026-11-17
Data: World Bank Enterprise Surveys: formal, informal and micro survey designs compared
Compares three World Bank Enterprise Surveys tracks (formal, informal, micro) by target population, sampling design, respondents and weighting notes, compiled from the programme's methodology page and microdata library collection page.
Coverage: Over 160 economies.
| Survey track | Target population | Sampling design | Respondents | Weighting and comparability notes |
|---|---|---|---|---|
| Formal | Registered firms with 5+ workers in eligible ISIC Rev.4 sectors, at least 1% private ownership | Stratified random sampling by firm size (5-19, 20-99, 100+), sector and region; face-to-face interviews | Business owners and top managers, sometimes accountants or HR managers | Panel firms: weights adjusted for altered inclusion probabilities; at least half of prior firms re-interviewed |
| Informal | Unregistered businesses | Two-stage: Adaptive Cluster Sampling of grid Block Areas in urban centres, then random businesses; 20-25 minute interview | Entrepreneurs of unregistered businesses | Two-stage weights; from May 2025 indicators use wweak weights; pre-2015 samples were convenience samples, not representative |
| Micro | Firms with 1-4 workers | Follows the formal methodology | Business owners and top managers (formal methodology) | Covers 35 economies |
Sources
- World Bank Group Enterprise Surveys: Methodology (accessed 2026-11-17)
- World Bank Microdata Library: Enterprise Surveys (accessed 2026-11-17)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







