From the raw data, variables are calculated to represent the research model. This model consist of the variables team membership, lending behavior and the modelling effect. Lending behavior is split into lending amount and lending frequency. The following describes operationalization of these variables.
All variables are based on lending actions saved in the dataset. Lending actions describe lending behavior because they administer actual lending. All lending actions are linked to their respective lender and loans. Loans are linked to burrowers. Lenders are, in case of memberships, linked to teams.
Lending behavior (frequency)
Lending frequency describes lending behavior for lenders as the lender’s lending activity in a set period, expressed as the amount of lending actions per day. E.g. 10 lending actions per day. Calculated as follows:
Sum Lending Actions for Lender ID / tijd
Proxy lending amount (per Loan ID)
The proxy for lending amount is required to create the variable lending amount. This proxy describes the average loan amount of the lending actions for a loan (Loan ID). When calculating the lending amount for a lender, the median of the average loan amounts per loan is calculated. The median is less sensitive to outliers.
Lending amount (for Lender ID)
Lending amount describes the amount lent by a lender as the lending activity for this lender in a set period, expressed in US dollars per day. E.g. $ 250 per day. For this variable, the calculated proxies, for the loans that the lender contributed to, are used. Calculated as follows:
Sum Proxies lending amounts for Lending Actions by Lender ID / tijd
Team members/Lending team memberships
Team members are lenders who have joined one or more teams and are not team captain. The status change, regarding amount of teams joined during the period of data collection, is recorded as the amount of team memberships at the time of change, based on retroactively gathered data that are gathered because the lender was linked to one or more lending actions within the period of data collection.
Team captains/Leadership status
Team captains are team members that are marked as captain of one or more teams they have joined. This mark is recorded as a dummy variable for each lender’s team memberships.
Lending actions by team members that have taken place after a team captain made a lending action, are marked as conditioned. This operationalizes the modelling effect. It is unsure for how long the team captain’s lending activity potentially influences the team member’s lending activity. Literature study on this topic is not entirely conclusive. The duration of conditioning is set to 24 hours, for the practical reason that lending behavior is also measured in days.