Surveys are a good way of gathering a lot of quantitative and qualitative data from a large or small sample of poject participants, beneficiaries and/ or stakeholders. Click here to see the between surveys and questionnaires. See also the website entitled key differences.
Here the focus is on face-to-face surveys which lend themselves particularly well to theory-based approaches that seek to understand both why and how change takes place.
How to Conduct Face-to-Face Surveys
There are lots of sources of advice on how to- conduct face-to-face surveys, see for example the document by Information systems services. The following steps are usually important:
- Define the aims of your survey. How can a survey shed more light on your project or some aspect of it? You may need to do some basic research about similar projects in similar regions. Have similar surveys or questionnaires already been tried? What were they aiming to show?
- Identify the population and sample The population refers to all members of the target group. A sample is a sub-set of the population. Samples are usually preferred as it would be too costly to survey the whole of your target group. See our earlier discussion on sampling.
What kind of sample should you go for? A representative sample may for example include a selection of respondents drawn randomly from the overall population. Ideally, it should include the same number of males and females or under and over 40s as in the overall target population.
Needless to say, this degree of precision is hard in some African villages and you may end up having to accept that some people (perhaps women) are more literate and hence more qilling to fill in a questionnaire/ surve than others (e.g. men, chldren). If so, you should acknoweldge this bias when referring to your findings.
When surveying any human respondent and particularly vulnerable adults or children, particular care is needed and you should be particularly alert to ethical guidelines
- Decide how to collect replies. Should the survey be completed by the respondent directly or through an interviewer? In either case, explain why it is important. If you are sending your survey/ questionnaire by post or e-mail, it is even more important to include a message or letter explaining what usual practice to provide a letter that explaining what the survey/ questionnaire is about and why it is essential to complete it.
- You may need to add incentives: good for the community n that it will do X or Y, prize draw. Keep stressing that it will soon be over as some people will give up. If you expect respondents to send you a reply by post or email you must make this as easy as possible: stamped address envelope. ‘Submit’ Icon on your computer if you are sufficiently computer savvy.
- Design your survey or questionnaire, making sure you ask concise and clear questions that help you answer the aims of your survey/ questionnaire. This will ensure you collect relevant data. Do NOT ask too many questions or you will reduce your response rate.
Think about the aim of your survey and what you are trying to show and then write yur questions accordingly.
|Trying to Show||Questions to ask|
|Whether men or women are making more use of an alcohol abuse treatment centre||Are you male or female?|
|Whether learning outcomes have been achieved?||On a scale of 1 to ten how much more familiar are you now with family planning techniques|
|Whether assumptions are correct||Should this project concentrate primarily on the victims or the perpetrators of petty theft?or On a scale of 1 to 10 with one being ‘not at all’ and ten being ‘very strongly’) how far do you agree with the following statement: It is because of the location of our village that we do not have adequate health services|
- Make sure your survey or questionnaire has a clear title and date/ revision date, a brief statement of its aims and contact information/ address/ email.
- Number your questions and potentially your sections but don’t have too many sections.
- Do not cram your questions together and do NOT make the survey/ questionnaire too long.
- Do not use tiny font or jargon.
- Do start with easier questions but perhaps leave personal questions to the end.
- Do use bold font, italics, images and colour if you can afford to do so. Keep your respondent interested.
- Do group questions together by theme/ topic.
The Lickert scale (from Strongly Agree (SA) to Agree (A) to No view (NV) to Disagree (D) through to Strongly Disagree) is very useful in surveys and questions. You could for example say: Indicate the answer that reflects your view of the latrine installation project. Circle the answer which is most appropriate:
|The new latrines will help to stamp out diseases |
in the village
|The new unisex latrines do not offer enough |
privacy to users
Or you can ask respondents to choose from a range of answers. You could also invite respondents to rank answers in terms of importance
Which of the following is most important to you:
- Small class sizes
- Classes of the same age
- Classes streamed in terms of ability
Which of the following are important to you in terms of your daily commute to work? You may tick more than one box.
- A regular bus service into town
- WiFi access on the bus
- Opportunities to talk with fellow passengers
- Storage space for luggage
- Opportunities to see the local countryside
- Other. Please specify below
Yes/ No Answers
These may ask respondents if they agree with a statement.
- We need to accept more refugees from neighbouring countries. YES/ NO (delete as appropriate)
Avoid Ambiguous or Leading Questions
- Are you against de-criminalising abortion? YES/ NO
- Do you like meat, fish and seafood?
- Would you accept the general consensus on the part of experts, namely that …?
Ask for precise information
- Either: Please state your age
- Or: Which age category are you in:
Note that open ended questions can elicit rich answers, useful quotations and (sometimes) helpful testimonials but do not ask more than one or two. They are not suitable for coding and can discourage respondents from finishing the survey.
Key Steps to Take
First of all, run a pilot survey. This step is only touched upon here as this toolkit is designed for smaller to medium sized NGOs. It involves testing the survey on a small sample of the population or at least a few colleagues. This can spote ovious design flaws and rephrase some questions.)
Second, conduct the main survey. Make sure you explain when sending, handing out or conducting the survey how important it is to fill it in. Why should respndents respond? How might that help them or their community?
If conducting a survey decide how, where and on whom to do this. Perhaps you (and possibly your co- surveyors) could stand at the same place (outside a University or in a quiet part of a busy market place) at different times of the day. Or you may go to different places: one wealthier and one poorer district of town. Are you approaching everyone or just men/ women/ market-stall holders?
You could for example be doing a purposive sample of teachers, conducting the survey at a teachers’ conference. Put respondents at ease, reassuring them that the survey will be anonymous and that it will take no more than 10 to 15 minutes. Stress the benefits of filling it in.
Finally, analyse the data. If you are not computer savvy then you may wish to enter your data on a sheet of A3 or possibly in a Word document. Respondents will be given a number or letter and represented along the top row or left column.
Answers will be abbreviated and written out along rows or columns (depending on where you have placed the respondents)
Alternatively, enter your data on to an Excel spreadsheet. You will have one row for each survey/ questionnaire respondent and one (or more column) column for each answer. ou may wish to summarise briefly in each column the sense of the individual question.
You may find it easier to use coding. The following advice is quite basic. You may well wish to consult more detailed guidance.
- Attribute numbers to each answer.
- So True would be 1 and False 2.
- Male might be 1 and Female 2.
- In rated answers, strongly agree would be 1 and strongly disagree would be 5.
- In multiple response questions:
- Where an answer has to be selected (say out of 4), you should enter 1, 2, 3 or 4
- Where multiple responses (MR) are possible you may need a column for each possible answer.
|Respondent||M/F||T/F||Rated Answer||Selected answer||MR|
|MR (Answer 2)||MR (Answer 3)|
On the above basis, it should be easily to count how many males/ females. A quick way of doing sort is to use the Sort function in Excel and look across at the row number. Remember to decuct one row as that is used for your heading
It should also be posssible to work out how many men or women reponded true or false or gave particular ratings to particular answers. You can compare the percentages responding in a particular way to a partcular question.
You can also show that 18 out of 20 females (90%) strongly agreed that gender abuse was a serious issue in the local community whereas only four out of ten male (40%) recipents strongly agreed.
If you want to go further and draw up correlations between your answers (most people who answered 2 on the first question also answered three on the tenth question, then see advice on CORREL.
You can also turn your answers into bar charts or graphs. See It Still Works for instructions