Tomorrow is the great day when we will be back in the field. Over the past few weeks, a new assistant has joined the team (welcome Ignatius!) and the questionnaire has received a nice haircut, from 18 to 14 pages. The main challenge was to give a “haircut” to our questionnaire,  not an amputation, and I believe we succeeded. If you happen to pass by Kariobangi around 10am, come say hello in Landmark Plaza, we’ll just have finished our initial training session.

We have done a lot of brainstorming during this break and I have received feedback from many colleagues around the world. A few interesting points have emerged on the “joys and sorrows” of doing surveys – in Kenya as compared to other countries, and using surveys as compared to other research methods.

First, our experience confirms that Kenya (and Kenyans) are particularly research-friendly. Many colleagues were impressed that we could handle an extensive questionnaire without paying the respondents. Let’s be clear: more than once we have been turned down by respondents and several times people received us with suspicion. But that was mostly our own fault.

Thanks to an interesting discussion with the market chairman in Kariobangi, we realized that showing up with a questionnaire in our hands was a disastrous way of approaching people and that “being introduced properly” was crucial for increasing both the response rate and the quality of answers. Once we learned the proper manners, we noticed a much higher willingness to participate to the survey and interest about the results.

Though, “being introduced” to entrepreneurs (aka a partial “snowballing technique“) reduces the representativeness of the sample, which is no longer “purely random“. Unfortunately, in our own fieldwork, we saw a tradeoff between the “quality of the answers” (arguably higher with snowballing) and the “quality of the sample” (arguably higher if random). We chose to maximize the quality of responses, though at a price.

There is also another finding: we noticed that respondents rarely answer “I don’t know” to any questions -although they actually do not know the answers. We tried to understand whether we are posing the questions in the wrong way or whether this is related to culture or something else. On the “plus side”, this means that we have a very low number of missing values -the greatest headache in econometric studies. On the minus side, it might imply inaccurate data -a complete disaster- especially because it adds up to the infamous recall bias typical of recall questionnaires.  Quality control of the data collection and research process will be key. But this topic deserves its own post.