In this post, I will walk through step by step on a survey budget design and will show you the case of the upcoming National Survey on Microfinance Demand in Madagascar.
When designing a survey budget , one must take care of 02 main aspects:
Technical budget encompasses underlying costs related to sample size . In the national microfinance survey, we used a systematic random sampling methodology with unequal probabilities sampling to select in total 336 area zones to be surveyed.
For each randomly selected area zone, pollsters have to go on the field then count the number of households. At the end of the count, a predetermined number of 15 households per area zone are randomly selected using systematic random sampling with equal probability which makes a sample of 5040 households across Madagascar.
When thinking about the cost, most of the time, we start with the following simple equation:
Effective data collection duration x Households per day x Number of pollsters = Total number of household to be surveyed
From this equation, one can approximately derive the effective duration of data collection which differs from the data collection duration but also can control the number of pollsters that will be involved in the survey.
Duration data collection = Effective duration + Trips duration + random events
Trips duration should also be included in the budget. In fact, area zone are geographically dispersed across the country and costs related to transportation should be carefully considered. However, transportations differ depending on the distance from cities where pollsters come from but also the state of the roads in the area zone which may delay local arrival time. Transportation costs are often multiple of the number of human resources involved in data collection: pollsters, controllers, supervisors.
Budget designers should also take care of the fact that cars location may be involved during data collection then they should include fuel costs per kilometers in their budget. This often introduces unequal cost per human resource . For example, pollsters may not benefit from car location as opposed to supervisors.
There exist 02 main types of random events: expected events like the waiting time before reaching interviews which are most likely to appear in any (household/non household) surveys but also unexpected ones such as illness, material losts or damages which all contribute to delay a data collection.
Questionnaires printing is directly function of the number of households to be surveyed then are included in the technical budget. One should keep in mind the following formula when calculating budget related to questionnaires printing:
Unitary cost per page x number of page x (number of households + reserves)
In a nutshell, one must carefully determine duration of the survey which will served to estimate the cost related to human resources as often times, they are paid on a human day basis (Number of days times number of human resources).
Costs related to non technical budget include:
Additional expenditures are also involved during human resources training such as coffee break, room rental and furnitures such as pens and paper sheets.
Pilot survey is also an important phase of a national survey. It is conducted in order to test questionnaire. It implies additional spending mainly car location, fuels, earnings of pollsters but also questionnaire printing. This step serves as well to forecast effective duration of the upcoming data collection then adjust the budget.
Questionnaire forms and database design and implementation require specific skills that statisticians may not have especially if survey is powered by new technology such as questionnaire from tablets or Smartphone.
Statisticians and some experts such as Economists are required to ensure quality of surveys. They are present in all steps: from questionnnaire design, data cleaning, analysis and interpretation of results. They are the core of the survey.
Last but no least, security in the terrain should not be underestimated then budget associated with it must be carefully reviewed . In the south of Madagascar, insecurity is an issue and pollsters should be associated with police and local authorities during their data collection in order to ensure their own security but also to avoid non-responses from households.