Field surveys are one of the most commonly used methods used by researchers for the process of primary data collection. In cases where secondary sources of data do not provide sufficient information, field surveys allow researchers to better monitor and evaluate the impact of field experiments. For example, consider a study that aims to evaluate the impact of micro-loans on farm output in a small village. It is possible that data on farm output for the last 10 years is not available, or is insufficient. In this case, researchers can conduct a field survey among local farmers to collect data on farmer incomes and farm outputs.
- Primary data is vital for conducting empirical inquiry in the field of development economics.
- The research team can either conduct the survey directly, or indirectly through a survey firm.
- Researchers conduct surveys by asking respondents to answer a survey instrument (questionnaire).
- With increasing availability of specialized survey firms, ODK-based tools like CAPI, and standardized field management practices, it is important for researchers to follow certain best practices to gather data.
Field surveys allow research teams to collect data through in-person interviews. These can either be done in the form of computer-assisted personal interviews (CAPI) or computer-assisted field entry (CAFE). One of the biggest advantage of field surveys over other forms is that it allows for human interaction. This can result in higher response rates compared to other forms of data collection (like phone surveys), and improve the quality of collected data.
However, in some situations in-person data collection is not feasible, and the research team must look at other options like remote surveys. Examples of such cases are studies involving areas that are hard to physically reach or conflict areas that might be unsafe for fieldwork.
If the research team decides that conducting a field survey is in fact feasible, they can move to the process of preparing for field surveys. This process involves multiple stages such as drafting, piloting, programming, and translating, with clearly defined timelines for each step.
Pre-pilot and draft
The pre-pilot and draft stage of implementing a survey starts by defining rules and guidelines in the form of survey protocols. Clear protocols ensure that fieldwork is carried out consistently across teams and regions, and are important for reproducible research.
The pre-pilot, which is the first component of piloting a survey, involves answering qualitative questions about the following:
- Selection of respondents.
- Tracking mechanism.
- Number of revisits.
- Dropping and replacement criteria.
Based on the pre-pilot, the research team designs a questionnaire to generate a first draft. For this purpose, avoid starting from scratch, and try using existing studies and questionnaires as a point of reference.
The next step is to conduct a content-focused pilot. This stage involves answering questions about the structure and content of the questionnaire. Global best practices recommend conducting the content-focused pilot on paper, which makes it easier to revise and refine the survey instrument. It is equally important to simultaneously pilot survey protocols like interview scheduling, infrastructure, and sampling.
Once the questionnaire content and design have been finalized, the next step is to program the survey instrument. Researchers should not program the instrument before finalizing the design of the questionnaire, otherwise they will waste crucial time and resources in going back and forth. Researchers must set aside 2-3 weeks for coding the instrument, and another 2-3 weeks for testing and debugging.
Before conducting a data-focused pilot, the research team must procure a survey firm and sign a contract with the selected firm. The data-focused pilot test for the following:
- Survey design. Re-check the design of the survey. Check if comments from earlier reviews have been resolved.
- Interview flow. Re-check the time taken for each module of the interview. Ensure that the interview has a flow to it, and both the interviewer and respondent are clear about the question.
- Survey programming. Check if questions display correctly. Check if modules need re-ordering. Ensure that built-in data checks are working.
- Data. Check whether all variables appear correctly. Check for missing data. Check for variance in data.
- High frequency checks. Monitor data quality through high frequency checks.
DIME Analytics has also created the following checklists to assist with piloting:
- Checklist: Preparing for a survey pilot
- Checklist: Refining questionnaire content
- Checklist: Refining the questionnaire data
The next step is to translate the questionnaire. Translation will be considered good or complete only when enumerators and respondents have the same understanding of each question. Often, with repeated translations, the research team must list out adequate version control norms.
The final step before launching the survey is enumerator training. The research team must complete the previous steps before the training of enumerators starts. This minimizes enumerator effects, which arise due to differences in the way a question is asked to each respondent because different enumerators may share different translations of the same question.
After preparing the survey, the research team launches the final instrument for fieldwork. This step concludes the process of data collection, and sets the stage for the next step in conducting a field evaluation, analysis.
The process of implementing field surveys comes with its own set of challenges.
- Measurement challenges. Sometimes some questions are sensitive, or aim to measure things that seem hard to quantify. For instance, while employee-satisfaction is very important for various studies to assess labor-market conditions, it is also very hard to measure objectively.
- Data-quality assurance. It is very important to have a data quality assurance plan. Researchers must ensure that this plan is followed by field coordinators and enumerators to avoid problems during the subsequent steps of data cleaning and data analysis.
- Translation errors. If translation is poor or incomplete, the meaning of the questionnaire can change, resulting in incorrect data.
- Gaps in enumerator training. This can affect the quality of responses during a survey, and therefore negatively affect the results of a field evaluation.