This story draft by @escholar has not been reviewed by an editor, YET.
Authors:
(1) Muhammad Zia Hydari, Katz Graduate School of Business, University of Pittsburgh and Corresponding author;
(2) Idris Adjerid, Pamplin College of Business;
(3) AAaron D. Striegel, Department of Computer Science and Engineering, University of Notre Dame.
2. Background and 2.1. Leaderboards
3. Effect of Leaderboards on Healthful Physical Activity and 3.1. Competition
3.3. Moderating Effects of Prior Activity Levels and Leaderboard Size
4. Data and Model
5. Estimation and Robustness of the Main Effects of Leaderboards
5.2. Robustness Check for Leaderboard Initiation
5.4. Fitbit Attrition, Leaderboard De-Adoption, and Additional Robustness Checks
6. Heterogeneous Effect of Leaderboards
6.1. Heterogeneity by Prior Activity Levels
6.2. Interaction of Leaderboard Size, Rank, and Prior Activity Levels
6.3. Summary of Findings from Heterogeneous Effect Analysis
7. Conclusions and Discussion, Endnotes, and References
As an additional robustness check, we also considered leaderboard initiation as it may be a proxy for confounded leaderboard adoption. Specifically, if the focal user is the primary inviter to the leaderboard, this leaderboard may be more likely to be driven by unobserved motivation changes. For the purpose of this analysis, we consider focal users to be of the “inviter-type” if they initiate most, not necessarily all, of the invitations to other users on their leaderboard. Although we do not have access to direct measures of who initiated a leaderboard, we construct a proxy variable that we argue identifies users who are more likely to be inviter types. We leverage two aspects of leaderboard creation to construct this proxy variable. First, per the discussion in Section 2.1, Fitbit does not use a leaderboard that is defined centrally as a group of individuals that others can join or leave. Rather, each leaderboard is owned by the user and each user pair must agree to share their step information for them to be joined on their individual leaderboards. In addition, Fitbit does not advertise to the user’s friends that they have joined the platform.
Based on these aspects of Fitbit leaderboards, we designated InviterLB using two criteria: (i) whether the leaderboard had three or more individuals when it was first adopted and (ii) whether the leaderboard was such that the other users (excluding the focal users) had been on the Fitbit platform for longer than 90 days. The first criterion is useful because the size of the leaderboard at leaderboard initiation can be indicative of the likelihood of initiation by the focal user. If there are two people when the leaderboard is started, it is unclear who initiated. However, if three people (or more) are on the leaderboard at initiation, a leaderboard fully initiated by others would require that two other users actively searched and invited the focal user in the same week and that the user accepted both invitations. However, this criterion may still include mixed leaderboards that were only partially
Note. (a) Please see Equation (2) for charts’ specification, (b) 90% confidence intervals, (c) vertical axes use different scales.
initiated by the focal users (e.g., another user initiates but the focal user invites the third person). Thus, we add the second criterion that the other users on the leaderboards have been on the platform for more than 90 days. The rationale behind this criterion is that users on the platform for longer periods of time are more settled on the platform and less likely to be actively scouring the platform for new connections. The 90-day threshold was chosen based on data suggesting that Fitbit abandonment happens in the first few months of adoption.[18] Among the leaderboard adopters, 15.3% met these criteria. In Table 2, column 6, we add an interaction term between leaderboard and an indicator for InviterLB and identify a negative coefficient that is close to zero and insignificant (p=0.8). This result suggests that users who are more likely to have initiated the leaderboard do not see different treatment effects and is further evidence that time-varying changes in motivation are unlikely to be confounding our results.
This paper is available on arxiv under CC BY 4.0 DEED license.