Henry Cust



The path of development through industrialisation is a well-trodden one. Countries in Asia, such as China, have made huge strides in development over the last 40 years and are now reaping the rewards with average GDP per capita growth of 9.4% from 1978 to 2012[1] and are now the world leaders in manufacturing exports. In Tanzania, and Africa more widely, economic development has taken a different, slower path. McMillan et al. (2017) described how structural change through the reallocation of labour from low productivity to high is the driving force behind much of the recent growth in Africa.[2] Understanding firms, how they begin and how they grow (or fail to) is vital to informing policy to ensure this recent rapid growth is not a flash in the pan but becomes the new normal for Africa.

In 2015 my colleague Steven Broadbent participated in a conference in Geneva focused on the cutting edge of establishment and firm level surveys at the time presenting on the potential of good quality data as well as learning a wealth about the challenges and new methods of collecting data. More can be read at here. Three years on, EDI has taken these insights and put them into practice.

We, at EDI, understand how to collect top quality firm data. We understand the subtleties of working with firms of different sizes and from different sectors of the economy. For example, working with informal firms poses a particular challenge since they are not officially registered with government and tax authorities, particularly when it makes up a significant section of the overall economy. Medina et al. (2017) estimate the informal sector of the Tanzanian economy could be as high as 50%-65%[3] so knowing how to engage these firms in data collection activities is key to good data. The challenges and solutions in this blog post can be applied across sectors including, agriculture, industrial, service and tertiary sectors in a variety of country contexts.

How firm surveys differ from household surveys

Building a Sample

Unlike a household survey where each interview takes place with a fairly homogenous unit of study, the selected household respondent(s), a firm has much more variety, from small single person enterprises which are akin to households and actually are often based in the household of the owner, to worldwide globalised conglomerates. A key part in the preparation of a project is building a sample frame and stratification strategy taking account for the wide variety of firms in the economy. Government and non-government organisations working in certain sectors of the economy are the best place to find lists of firms currently operating to start the sample frame. However, in developing countries, these lists are often incomplete, out of date, have multiple entries for the same firm, with varying details making it difficult to differentiate duplicates, or contain firms that do not exist at all. Informal firms are unlikely to be listed officially so different methods are needed to ensure inclusion of these firms.

After the best possible sample frame has been constructed remotely using lists, a sample needs to be drawn based on the overarching research questions. Key variables that are included in the stratification could include: the firm’s sector, size (employees and/or revenue/market share), ownership structure and/or location, dependent upon the information available. Unlike many households and agricultural establishments, e.g. farms or mills, many firms and establishments in the secondary and tertiary sectors are located in and around centres of population where a marketplace for their goods and workers to supply labour are located. Because of this it is often more appropriate in these sectors, to sample the firms and establishments at a regional or city level rather than the community level typically used for households.


Compared to a household survey, there are more possible respondents within a firm, with a variety of responsibilities and skills. Selection methods such as randomisation[4] or ‘most knowledgeable’ are not appropriate for firm surveys. A more targeted approach is required where the scope of the research is important for deciding who within the firm is the most suitable respondent. How much exposure the respondent has to the operations of a firm or establishment along with the role of the respondents are among some important questions that should be considered. For example, a survey about ownership would benefit from the owners of firms as respondents, or, a survey about factory conditions would benefit from having factory floor workers and supervisors as respondents.

Value of Time

For obvious reasons the more responsibility and power a respondent has within a company, the more valuable their time and the less likely they have spare time to complete surveys. However, this applies to all workers as their time is not their own. This creates an obstacle for workers to take time out of their day to answer questions to researchers. Less senior members of staff often require permission and may not be paid for their time answering questions.


Another aspect to consider in the questionnaire design is the type of questions you are asking and who you are asking them to. In the private sector, there is a real consequence of releasing sensitive, commercial information about your firm, and unfortunately, this is often the information researchers want to extract. Therefore, strict confidentiality needs to be ensured with pledges that sensitive information will not be published for competitors, or anyone, to see. Even with confidentiality procedures in place, respondents may refuse many questions for fear of their manager disciplining them, worry about the consequences of truthful answers or they simply cannot respond because of privacy agreements. Questions directly asking about the operational finances, stock or operating numbers of firms are sure to raise suspicions and should be avoided. Research analysis plans should account for questions that are expected to receive a high level of resistance and contingencies made.

However, even before you begin asking questions, firms will often treat interviewers with suspicion. Firms will often assume interviewers are from competitors trying to seek out key information, or from the government and tax authorities trying to find reasons to extract additional money from the firm. Stating the purpose of your visit and reassuring there are no links to government or competitors is vital in creating a positive atmosphere where open and honest discussions can take place. Obtaining government clearances is vital for all studies and knowledge that the government is aware of the study and the confidentiality of answers may help contributing to a positive interview environment.

External Influences

Because a respondent’s position is not guaranteed, external factors to the interview may influence the respondent’s answers differently to household surveys. An example would be the belief that giving more positive answers would result in being rewarded by the company’s hierarchy or a respondent may have grievances that affect their answers. Whilst this is true of household interviews where confidentiality can be ensured, this may not be enforceable in a professional environment and the immediate consequences of revealing certain pieces of information or giving certain opinions will weigh more heavily on the minds of respondents.

Multi-round Surveys

Building panel datasets with households is relatively straightforward. Measures to track households can be put in place and EDI often achieves a tracking rate of greater than 95%. Read more here. Tracking firms or particularly respondents within firms poses more challenges. Firstly, people change jobs more often than changing households meaning trying to find a particular person in a firm after several years have passed is difficult. Secondly, the survey may require tracking a position rather than a person. However, positions may be dissolved, or responsibilities change whether the same person remains in the role or not. Thirdly, if nonessential for the study, it is important to prevent information from previous rounds entering later rounds. In some cases, even knowledge of previous round can influence the answers provided.

How EDI successfully implements firm surveys


To build a comprehensive sample frame EDI conducts the following in Tanzania:

  • Gather lists of all firms from government sources including the Tanzania Revenue Authority (TRA)[5] and the Ministry of Trade, Industry and Investment.[6]
  • Contact non-governmental bodies for lists such as Small Industries Development Organisation (SIDO), the Association of Tanzania Oil and Gas Providers (ATOGS)[7] or the Confederation of Tanzania Industry (CTI).[8]
  • All lists are cross-referenced to create a working sample frame, but this list is not complete.
  • Include measures to add to the sample as we conduct fieldwork similar to snowball sampling.

Once teams have entered the field and data collection has begun, we include a module at the end of our questionnaires asking about the target firm’s network of suppliers and customers to build into our lists. This is a good activity to access small and informal firms which are most likely to be missing from our lists. Whilst many of the firms described in this module of the questionnaire are often already known to us, the benefit of those few that are not makes the additional time spent worth it for the best possible sample. Another method is to look at physical products or services offered, finding their manufacturer or retailer, and contacting them. For example, visiting a market and asking where certain goods are produced – this is especially useful for handmade goods that sellers buy directly from those who produce them. These methods build lists by approaching the sample from multiple angles and result in a better more complete sample overall.

Questionnaire Adaptation

A vital part of the project preparation is ensuring questionnaires are suitable and will address the research questions. Therefore, we use our wealth of experience to adjust and adapt questions that may raise suspicions in the respondent. EDI always conducts a pilot of the questionnaire with respondents from firms we know well, and those we do not, to obtain honest feedback on areas they would be unwilling to reveal details about and why. Being creative with questions allows the desired information to be extracted whilst avoiding sensitive topics which undoubtedly leads to better quality data.

Targeting Respondents and Approaching Firms

EDI employs specific techniques for approaching firms and desired respondents. If an interviewer is targeting a certain position within a firm, the decision about who the respondent is has to be made by the interviewer. We provide specific training to provide each interviewer with in-depth knowledge of the responsibilities and tasks of the desired position to find the right respondents. Firms will often lead interviewers to persons not suitable for the research (usually someone more junior than is required). Another problem with respondents, particularly more senior respondents, is finding time for interviews. Respondents will often rearrange appointments at the last-minute leading to a low rate of completed interviews per visit. EDI ensures teams have the resources and time to adopt a flexible working pattern to fit with a time that the respondents are available; often outside of typical working hours and in locations other than their place of work.

Another approach we use is to contact firms over the phone or by email to determine the true respondent, their role and availability, before visiting in-person. Alternative options are setting up microsites and project pages on the internet so that information about the project and its aims is available to firms to digest in their own time and allow them to consent in their own time. Up to two weeks should be allowed for this stage to ensure the correct respondent is selected but ultimately allows interviewers to use their time much more effectively once the fieldwork begins.

Bespoke Training and Professionalism

With firm surveys, first impressions are key to keeping rates of refusal as low as possible. At EDI we understand that receiving a refusal from a key player in a sector we are working (think Coca-Cola in a survey of soft drinks companies) can have serious consequences on the quality of data therefore we shape our training around the firms we will be working with. This includes ensuring our interviewers first impressions, and lasting impressions, are positive and professional. We ensure all interviewers are provided with personalised business cards, arrive in a timely manner for interviews and dress professionally. Creating a professional tone and feel to our interactions will encourage respondents to consent for interview and be open if they feel they are dealing with professionals which undoubtedly leads to a better atmosphere for the interview and better data as a result.

Field Organisation

After considering all these challenges, the most important factor to include in any field plan is time and flexibility. Many of these challenges and steps, e.g. gathering lists, contacting firms, arranging appointments, re-arranging appointments rarely occur when you want them to and require a flexible approach from field teams. At EDI we design in high degrees of flexibility, utilising smaller field teams for longer periods, having relaxed travel plans and ensure all staff have access to all possible forms on communication.

Concluding Remarks

EDI applies these lessons to all our firm surveys in all the countries where we work. In Tanzania, we have worked in firms and establishments for clients such as the World Bank – collecting data for the World Management Surveys[9], the President’s Office – Regional Administration and Local Government (PO-RALG), and the University of Oxford.[10] In Uganda, EDI is currently in the preparation stages of a national Small and Medium Agribusiness Development Fund (SMADF) survey which includes a mixture of households and firms. From preparation, through to data collection, to data cleaning each phase requires a different approach. Firms and their workers can be unpredictable, an external research survey is a low priority, therefore, the dynamic of our work shifts. As the survey providers, we understand this dynamic and how to harness it to collect quality data from firms.

There is vast potential for firm data to be used for numerous aims, sometimes in conjunction with government data by governments or independently of governments. Either way security and anonymity are key, and it will fall on the data collectors and analysts as the guardians of such data to protect the identities of the firms. Maintaining this trust between research organisations and firms will be key to ongoing engagement with industry to unlock the potential of this data.

[1] https://www.weforum.org/agenda/2015/07/brief-history-of-china-economic-growth/

[2] See McMillan et al. (2017), “The Recent Growth Boom in Developing Economies: A Structural Change Perspective” at the following link https://ideas.repec.org/p/nbr/nberwo/23132.html

[3] See Medina et al. (2017) IMF working paper: https://www.imf.org/en/Publications/WP/Issues/2017/07/10/The-Informal-Economy-in-Sub-Saharan-Africa-Size-and-Determinants-45017

[4] See www.edi-global.com/views-from-research-no-10-randomising-within-household-respondent-selection/

[5] https://www.tra.go.tz/

[6] http://mit.go.tz/

[7] https://atogs.org/

[8] https://www.cti.co.tz/cpages/home

[9] The main website is found here (http://worldmanagementsurvey.org/) where datasets and reports for all participating countries can be found.

[10] Follow the embedded link for EDI’s description of the project and the following link to see the research paper written using our data: https://www.edi-global.com/tracking-survey/