In March 2023, Laterite expanded to Sierra Leone, our first office in West Africa and I was hired as a Data Operation Associate. Collecting quality primary data is not an easy feat anywhere and Sierra Leone is no exception. In my first year, I have come to appreciate some of the differences between Laterite’s experience in East Africa and here. Here are some of the nuances and how we are approaching these as a team to ensure high quality data collection in Sierra Leone.
Language and Translation
Translating questionnaires to the respondents’ local languages is one of the first steps in any survey. It has been fascinating for me to learn about the streamlined translation process we have in East Africa. In contrast to Sierra Leone, countries like Rwanda, Tanzania, Ethiopia or Kenya have at least one dominant local language that everyone speaks, and our field teams can as well read. This makes it easier to find professional translators and for our field team to check the accuracy of translated text. In Sierra Leone, while we have dominant languages such as Krio, Mende and Temne there is no one language that every Sierra Leonean speaks. Additionally, very few people can read and/or write in these languages.
In the last year we addressed this challenge by sourcing translators from the Bible Society and The Institute of Sierra Leonean Languages. We are also dedicating time for oral translation exercises in our training. We also have a language-diverse enumerator team. All of these processes ensure consistency in survey administration and a field team who can adapt as needed on the field.
Working in Remote Areas
Sierra Leone has over half of the population living in rural and remote areas. With only a quarter of the country having access to grid connection, remote areas in Sierra Leone are notorious for their lack of electricity, poor cell reception and internet connectivity, hazardous terrain or all of the above. This leaves field teams completely disconnected from the monitoring team, which has implications for data collection, with the biggest by far being:
- Delayed uploads: It can be days before enumerators’ progress is available for monitoring.
- Updating instruments: It’s difficult to communicate changes made to the survey tool with enumerators
- Resolving real-time issues: It’s a challenge for data collectors to reach the monitoring team for time-sensitive solutions.
While we cannot remove these obstacles there are a few things we can do to reduce their impact on data collection. One important aspect is to identify a hub town for reliable data uploads, usually the district or chiefdom headquarters closest to the survey area. Field coordinators will then travel regularly to the hub for syncing and direct contact with the monitoring team in Freetown. This travel time needs to be accounted for during the planning phase.
A second aspect is the selection of the field team. In this type of scenario, we prioritize field coordinators and enumerators with experience in the same location or in a similar setting, and with an exemplary track record of working independently. The field team should be kept small to give field coordinators an opportunity for frequent (more than usual) spot checks. Field coordinators will have individual debrief sessions with enumerators to solve real-time issues, get an update on progress, and document any change to the work plan.
Phone surveys and connectivity
Phone surveys in rural areas can be tricky because of connectivity. Participant lists are also more likely to have inaccurate phone numbers in remote areas due to lack of access to sim replacement facilities. This means that if people switch or lose their phones between the time of listing and data collection they will likely have a new number by the time the survey starts. It is also common for participants to be secondary users of a phone owned by someone else. Lack of electricity to charge phones can also make it difficult to reach people. All of this combined, makes it very difficult and time consuming to reach respondents.
One way we work around these challenges is to have an adequate replacement list. It’s best practice to account for low response rate during the sampling for a phone survey, and to collect secondary contact numbers during listing to increase chances of reaching participants. In anticipation of difficulties reaching respondents we always include ‘mop-up’ days in the plan.
Secondary data sources
A lot of progress has been made in streamlining access to public data but it’s still tricky to find data to inform sampling, such as exact population, households, languages, or distribution of villages in a section. To access this data we usually have to visit the district councils and check their records backed up by a listing.
Given the above, listing is very important to build a sampling frame. But in person listing can be time-consuming, and expensive too, so we combine some in person data collection with geospatial data to bridge the gap. In a recent household listing design across two districts in Sierra Leone for instance, we used a gridded sampling approach with information from remote sensing to generate the household list from which the sample could then be determined. In practice, the output was not just of households or dwellings but of all buildings or structures in the districts. We therefore sampled 30% more households than needed to make up for buildings that were not households. This was complemented with an in person listing exercise to validate the sample and collect information on the households.
Laterite has ample experience employing gridded sampling techniques and combining these with remote sensing data in different country contexts. From East Africa, we have a robust database of geospatial features, which we are hoping to replicate in Sierra Leone, and West Africa more generally.
Personal identifiers
Participant personal identifiers in Sierra Leone can be unreliable when trying to track participants of a longitudinal study. Naming, in particular, can be very homogenous. I always shared my name with at least one other person throughout high school, but this is not surprising as there are supposed to be over half a million “Kamaras” living in Sierra Leone. Date of birth is usually helpful for tracking participants in a longitudinal study, but that is not the case in Sierra Leone’s rural areas. Documenting birth has only just started to gain traction and most adults over forty-five don’t know exactly when they are born. According to Multiple Indicator Cluster 2017, one third of children don’t have a birth certificate.
We collect unconventional identifiers (e.g., landmarks, nicknames) to mitigate this. For example: a participant can be identified with a comment that they live opposite the mosque or church. Accepted nicknames of participants are usually more established within communities than legal names, and this makes them valuable as an identifier. Household composition data (such as name of household head, spouse, sibling) can also be used to identify them in subsequent years. It’s also essential – where possible – to have a secondary contact to help re-establish contact with participants.
It has been fun thinking about how to overcome these nuances. Feel free to reach out to Laterite if you are working in Sierra Leone or want to learn about collecting data in the country.
—-
This blog post was contributed by Ishmail Kamara, Data Operations Associate at Laterite Sierra Leone
Cover photo: Lindsey Starck, via Wikimedia Commons