Data Collection Dates start 2018-04-8 end 2018-05-05
Data Collection Mode Face-to-face [f2f]
Data Collection Notes
Pre fieldwork: Six teams – each consisting of 1 field supervisor, 3 interviewers, and a driver – were recruited to carry out the survey over a period of 4 weeks from April 8th to May 5th. Six persons were recruited to conduct and code the respective fields for occupations and industries in each questionnaire. Experienced field staff who had successfully completed at least two rounds of the LFS were given a two-day refresher training and new or relatively new field staff were given a week of training. A one-day field practice was scheduled in a selected urban area in each district for each team to familiarize themselves with conducting an actual interview using all materials including tablets, MiFi devices, and electronic maps. Teams were asked to upload all complete assignments so that headquarters could review the following day and point out errors not captured by the software. All temporary staff, including editors/coders met for a one-day discussion of the practical field work. An additional day was scheduled to train editors/coders on how to properly code occupations and industries and look for errors in each questionnaire.
Fieldwork: After a thorough training, the six teams started data collection on April 8th using 7-inch tablets installed with Survey Solutions, a Computer-Assisted Personal Interview (CAPI) software developed by the World Bank. Corozal, Orange Walk, Belize, and Cayo districts were each assigned 450 households, while Stann Creek and Toledo districts were allotted 500 each. The teams worked six days a week, taking advantage of the weekends when most persons could be found at home. Through Survey Solutions, field supervisors were able to assign households to interviewers using internet connection. Field supervisors were given access to electronic maps through an ArcGIS application installed on their tablet that would assist them in identifying the Enumeration Area (EA) boundary. Along with a list of sampled households, a print out of a visitation record for each EA assisted supervisors in locating households. Once the household was identified, interviewers conducted face to face interviews and uploaded complete assignments at the end of the day. Data was collected from a responsible adult from the household; a proxy was allowed in cases where the other members of the household were not present. Editors/coders and headquarters had access to the data once this process was complete. Constant field supervision was performed during the four weeks conducting live interviews and re-interviews to improve data quality.
Field Checking: In April 2018, 6 temporary Field Supervisors were hired to each supervise the work of a team of 4 persons. The role of the Field Supervisor (FS) consisted of primarily identifying the enumeration district with the use of a GIS application and identifying each household in the sample listing with the use of printed visitation record. They also conducted at least 2 re-interviews in every cluster and witnessed a live interview for each interviewer during the first two weeks of data collection. With support from HQ, re-interviews were completed at random or specific selected households that needed further clarification or where inconsistencies were spotted by editors/coders. If there were major changes captured from a re-interview conducted, the data was changed and the field supervisor and/or the interviewer was made aware of the error. If further re-training was needed in the field, then this was also done at the time of HQ visit. HQ also supported in locating households that were difficult to find, persuading households that refused to participate in the survey, and verifying selected households that had changed to businesses, vacant lots, under construction and so on. The FSs and HQ staff had similar roles but HQ had scheduled visits only one time per the week to conduct spot checks in both urban and rural areas.
Sensitization: An advertisement was announced over six radio stations: Love FM (Belize), Estereo Amor (Belize), Hamanali (Dangriga), Wamanali (Punta Gorda), Radio Bahia (Corozal) and Fiesta FM (Orange Walk). The announcement was about a minute in length informing the public that interviewers would need their cooperation, the purpose of the survey and the periodicity of the data collection. The announcement was aired at prime time for 9 days (Mondays, Wednesdays, and Fridays).
Data Processing and Editing and Coding: Six temporary editors/coders and one head editor/coder were hired to commence work on April 10th, 2018. They were trained in survey solutions and how to identify and enter occupation and industry codes accurately. The role of the editor/coder was to check each questionnaire for consistency and enter codes where necessary. At the supervisory level of checks, an account was created so both FSs and editors/coders could have access to all questionnaires that were completed and synchronized. Since survey solutions covered most consistency checks, the editing part was mostly to verify that the data that was collected was logical. For instance, if the head of the household and spouse were both Mennonites, and the child was Creole, then survey solutions would not capture this. In editing, they also rejected responses in the ‘Other category’ that was already listed in one of the previous categories. Once error-free, the questionnaires were approved and ready for payment. The head editor/coder gave support to all six coders, especially new ones. An HQ account was created for her to also have access to questionnaires where she was initially responsible to spot check six out of 25 questionnaires and approve these cases.
Each editor/coder was assigned a district, but if they had approved all synched questionnaires for their district, then they would move on to a different district. Although Field Supervisors were asked to check all descriptions to ensure that the response could be codes, there were still vague or broad descriptions that were rejected by editors/coders. There were many difficulties with coding questionnaires from both Cayo and Stann Creek districts which caused delays since FSs were reminded on a daily basis to advise the teams to make corrections. The LFS team at HQ was comprised of 3 persons who also supported editors/coders and pointed out their errors using an HQ level survey solutions account. Errors were pointed out using the comments sections and were rejected back to editors/coders where they either needed to make a coding correction or it had to be further rejected to the interviewer. The data processing stage took about 6 weeks.
After approving all questionnaires at the HQ level, the data was exported to SPSS where 4 useful datasets were merged and a final document was prepared to be cleaned. In the cleaning stage, there were some missing information (especially codes) that needed to be entered. Once the final dataset was cleaned, it was ready for data analysis.
Data Collectors: Data collection was completed by six teams consisting of 3 interviewers, one FS and a driver in each team. Each team was assigned a rented vehicle and were required to work 6 days of the week, including weekends. A total of 31 field staff were recruited for training. There were three training dates to cater for experienced and new interviewers recruited. In each district, a relief person was trained in the event that a person from the team was not able to work. For districts where all or most interviewers were new, the best interviewer was selected after seeing their performance during training. Some relief interviewers were discouraged with this idea when it was mentioned the first day of training, so some persons dropped out. The main role of the interviewer was to travel to the assigned area and conduct face-to-face interviews using a 7-inch tablet and edit and synchronize completed assignments at the end of each day. They were also to make corrections rejected by editors/coders and make call backs to the households wherever necessary.
Estimates of Sampling Error None