We have talked about all the steps for sampling a coverage survey and also about some alternative sampling methods and some of their strengths and limitations. Now, I want to talk a little bit about potential barriers to good sampling. What are the most common obstacles that people run into when they are trying to use or implement good quality sampling for their survey? One, probably the most commonly cited barrier to good sampling is around budget or resources. It is doing this kind of sampling, particularly where you are mapping out and enumerating households in all of your survey clusters is not cheap. It costs money to send, cost money to train the folks who are going to do that. It costs money to send them to the clusters and obviously for them to do the work in the clusters. You're more or less going to the field twice. This is more of an issue for smaller surveys. For cases where the survey money, the evaluation funds may be coming out of the program funds and so may be quite limited. It is not a huge amount of money. Ideally, as we've talked about in previous lessons, you would prioritize spending that money on sampling so that you make sure that you have a good sample. But if you have a very limited survey budget, it may be the case that you decide that you just don't have the funds to do that, to send people to the field twice. I think there has been some work on alternative approaches using, for example, GIS that we've talked about and that can help in the sense that the development of the sampling frame can be done from the central office. You don't need to go to the field to do your enumeration of households. It still requires a fair amount of person-time and technical expertise in your central office, so it's not free. It may be somewhat less expensive though than traditional sampling. Obviously, if you're doing a random walk that's more or less free in the sense of you don't have to do any additional sampling besides selecting your clusters so there's no work to develop the sampling frame. But again, that is going to give you potentially a sample that is not representative. Another issue that I actually think may be a bigger issue than budget or resources, but sometimes this gets kind of folded into or conflated with the budget issue is access to technical expertise. If you've never done this kind of sampling before and you don't have anybody on your team who's done it before, and you just don't know how to do it. You don't know where to get mappers or enumerators, people with experience, you don't know how to train them. That can feel like it can be a big obstacle, right? This is, it's not a simple thing to do to go into clusters and to map them and to numerate all households and then to use that to generate the sampling frame. It does require technical expertise. If you're in that situation where you don't have access to the experience or the technical expertise on your team, the best solution is probably to reach out to somebody who does. Every country that does demographic and health surveys, that does Multiple Indicator cluster surveys, does this kind of sampling. By this kind of sampling, I mean, they sample clusters, and then they send folks to the clusters to do the mapping and the enumeration of households. In every country where they have done those types of large household surveys, they do have experts, usually at the National Institute of Statistics, which is called different things in different countries, in some countries might be National Bureau of Statistics, National Statistics Office. They have people there who have done this kind of sampling. They often had a pool of experienced cartographers on whom you might be able to draw. Using country expertise to do this is often a good solution if the issue is access to technical expertise, reaching out to the National Institute of Statistics, and seeing if they're available to consult on your survey and the sampling piece. In particular, seeing if you would be able to recruit your mappers from their pool of mappers. This again is not free. These are experts who will need to be paid for their time. But it can help kind of plug that hole in your team's experience if you don't have that particular experience on your team. Also if you do that a few times you can use it to train folks on your team so that you do eventually have some people who have some expertise in doing this kind of sampling and can provide that expertise when you're doing these surveys. A third barrier is time. It takes time to do this. We'll talk about maybe some potential approaches to that. There aren't a ton of ways to get around this. It takes time to construct a sampling frame. It takes more time than just showing up in the cluster and doing a random walk. Often, these surveys, particularly when they're done sort of at the request of a donor, a funder, for an evaluation, you're often under significant time pressure. What are some things you can do if time is an issue? One approach that I think we talked about during the sampling lessons is that it is possible, although quite logistically challenging, to do the sampling, the mapping and the enumeration, and the sampling in parallel with data collection. What happens there is essentially you start your team's of mappers and enumerators a few days, maybe a week before your data collectors go to the field. You make sure that you map out the trajectory, the itinerary for the data collection teams, and you make sure that the mapping teams are in those clusters a few days ideally before your data collection teams so that they can do the mapping, the enumeration, and then the sampling frame can be constructed, then the sample can be drawn and provided to the data collection team. That is logistically complex. It's not for the faint of heart. But if you're under time pressure and you have a strong team, it's a possibility. Another potential possibility that I mentioned earlier is using an alternative like GIS sampling. Again, these are methods that are under development. They still do require some time, but because you don't have to go to the field, you have people who can stay at the central office and do this. In some situations, you may have some time savings, not always, so that's something to be aware of that. Those are two potential ways to save a bit of time if you want to do the sampling and you are under pressure to get your survey done quickly. A final issue for doing this kind of sampling is insecurity. Some countries, certain countries that I work in, like Burkina Faso, Mali, right now in 2020, there are some areas that are seeing lots of attacks from different groups. It, in some cases, maybe you may want to limit the time that data collectors spend in certain areas because you don't want to put people at risk. You don't want to put your data collectors at risk needlessly. Sending people to the field twice wants to do the enumeration and then once again to do the survey is not ideal. This is again, a situation in which potentially, I think a sampling approach that's based on satellite images and GIS coordinates may have some value in that you don't have to go to the field to construct your sampling frame and so you can limit data collectors time in the field. This is not an exhaustive list of the issues that you may run into, and certainly there are others, but these are the ones I think that have come up most frequently and some potential ways to address them. Although there are always situations in which the solutions won't necessarily work.
