Nigeria - General Household Survey-Panel 2010-2011 (PostHarvest), First Round (Wave one)

This survey used a concurrent data entry approach. In this method, the fieldwork and data entry were handled by each team assigned to the state. Each team consisted of a field supervisor, 2-4 interviewers and a data entry operator. Immediately after the data were collected in the field by the interviewers, the questionnaires were handed over to the supervisor to be checked and documented. At the end of each day of fieldwork, the questionnaires were then passed to the data entry operator for entry. After the questionnaires were entered, the data entry operator generated an error report which reported issues including out of range values and inconsistencies in the data. The supervisor then checked the report, determined what should be corrected, and decided if the field team needed to revisit the household to obtain additional information. The benefits of this method are that it allows one to:

¨ Capture errors that might have been overlooked by a visual inspection only,
¨ Identify errors early during the field work so that if any correction required a revisit to the household, it could be done while the team was still in the EA

The CSPro software was used to design the specialized data entry program that was used for the data entry of the questionnaires.


6.12 Data Cleaning

The data cleaning process was done in a number of stages. The first step was to ensure proper quality control during the fieldwork. This was achieved in part by using the concurrent data entry system which was, as explained above, designed to highlight many of the errors that occurred during the fieldwork. Errors that are caught at the fieldwork stage are corrected based on re-visits to the household on the instruction of the supervisor. The data that had gone through this first stage of cleaning was then sent from the state to the head office of NBS where a second stage of data cleaning was undertaken.

During the second stage the data were examined for out of range values and outliers. The data were also examined for missing information for required variables, sections, questionnaires and EAs. Any problems found were then reported back to the state where the correction was then made. This was an ongoing process until all data were delivered to the head office.

After all the data were received by the head office, there was an overall review of the data to identify outliers and other errors on the complete set of data. Where problems were identified, this was reported to the state. There the questionnaires were checked and where necessary the relevant households were revisited and a report sent back to the head office with the corrections.

The final stage of the cleaning process was to ensure that the household- and individual-level data sets were correctly merged across all sections of the household questionnaire. Special care was taken to see that the households included in the data matched with the selected sample and where there were differences these were properly assessed and documented. The agriculture data were also checked to ensure that the plots identified in the main sections merged with the plot information identified in the other sections. This was also done for crop-by-plot information as well.

The cleaning process at the head office was impeded by the fact that the questionnaires were not immediately available for inspection when problems were identified in the data. The questionnaires were retained by the state in case there was the need for household revisits. So whenever problems were identified at the head office, the state office had to be contacted in order to determine if the suspect data were the same as the information on the questionnaire, and to ensure that changes were captured in both places. This was a very cumbersome and time consuming process since communication was difficult and in many instances the response was not timely. However, this is a necessary process to ensure that the households can be re-visited to provide the correct information to avoid having to make imputations. Also, this process allows the state officers to understand the key issues that arose during field work and will serve to enhance further rounds of data collection. It will be important, nonetheless, to find a mechanism to facilitate this process in the next round of data collection and cleaning.

A second challenge in data management and cleaning was the difficulty faced by state offices in sending the data from the state to the head office. There were difficulties in accessing internet facilities in many of the EAs and surrounding areas where the field teams were active. The consequence of this was that the data were not sent to the head office until the teams returned to state capitals where, due to the distance, it was difficult to return to the EAs for household revisits when requested by the head office. This issue will need to be addressed for future rounds of the survey.