Experimental Design Details
Simulation round of BDM procedure
Participants have the opportunity to practice the sequential Becker-DeGroot-Marschak (1964; hereafter BDM) selling mechanism described in stage 2. They are confronted with hypothetical sequential sales decisions on their place of residence, postal code and name and have the opportunity to repeat this three-step procedure. They obtain direct feedback on their decisions, which enhances their understanding of the BDM mechanism and ensures conscious decisions in stage 1.
After each simulation round, participants can decide to leave the simulation and continue with the experiment. We explicitly state that subjects should only proceed if they understand the mechanism and that we are available to answer questions via e-mail.
Stage 1a: Determining willingness to sell social media data
Questionnaire on platform affiliation
In this stage, participants reveal what social networks they use.
Stage 1b: Determining willingness to sell social media data
Now, participants have the opportunity to sell up to three data sets from their personal Facebook, Instagram, TikTok accounts (if existent). The order of networks is randomized in the experiment. Stage 1 is divided into three different treatments. Treatments are randomly assigned before the experiment starts and, thus, do not depend on participants’ answers.
If a participant has none of the aforementioned social network accounts, they skip stage 1 and proceed directly to stage 2.
By varying the (number of) possible buyers, the treatments test whether the willingness to sell data depends on the recipients’ type: a research institution (Research), an external private firm (One-Ext) or different external private firms for each sequential selling decision (Diff-Ext). In the external buyer treatments (One-Ext and Diff-Ext) we forward the data to the external company.
All three treatments use the BDM mechanism: Participants can choose a price between EUR 0 and EUR 30 for each data set they are willing to sell. We are aware that telling participants about the price range of the BDM-mechanism is not without loss of generality (Bohm et al, 1997). Nonetheless, previous work (e.g. Benndorf and Normann, 2018, Schubert et al. 2021) implies EUR 0 - 30 being an appropriate range for the elicitation mechanism in this case.
Then, the computer draws a random number between 0 and 30. If the random number is greater than or equal to the chosen price and if the lottery – described in the next paragraph – draws the person, the data will be sold and the seller receives the value of the random draw as a payout. Otherwise, no sale will occur.
After a participant has played all consecutive BDM-rounds his earnings enter a lottery, at the end of the experiment the lottery draw takes place and selects one out of ten participants to sell their data.
To illustrate, assume that a participant chooses a price of EUR 0. She will sell her data with a probability of 100 %. If, for example, the random draw is 10, the participant receives a payout of EUR 10 if she is selected by the lottery in the end.
Participants always have the option to explicitly opt out of the sale by clicking on the “Do not sell data” button. Correspondingly, they have to agree to the sale of their data of each network separately by ticking a checkbox and clicking on the button “Sell data at least for the selected price”. The combination of checkbox and button is a double opt-in structure that ensures that the participants fully understand the consequences of their decision.
In the Research treatment we inform participants that their data will be sold to us as Researchers. The One-Ext and Diff-Ext treatments focus on the external validity of the experiment. Here, participants not only sell their data to us, but explicitly agree that we may pass on their data sets to one external firm (One-Ext). The Diff-Ext treatment differs from One-Ext in that now each individual data set is passed on to a different external firm.
The three external private firms that cooperate with us and take participants data are all online retailers. Participants obtain access to additional information about the firms. The firms receive the data from us free of charge and comply with data protection regulations.
For the experiment, we focus on evaluating the behavior of participants in the experiment, i.e., what prices they choose for their data, how sequential selling decision influence their behavior and how the willingness to sell differs between treatments.
Stage 2: Questionnaires on personality traits, social network usage, privacy concerns, and demographic characteristics
To better explain the behavior of participants in stage 1, the second stage includes questionnaires on personality traits, social media usage, privacy concerns, risk preferences and demographic characteristics. For the questionnaire on personality traits, we employ the Big Five Inventory (German short version, BFI-K) scale (Rammstedt and John, 2005). The questions on social network usage behavior include, among others, the frequency of use, the number of friends and the age of the respective accounts. Further, we measure privacy concerns using the standardized German version (Pape, 2018) of the Concerns for Information Privacy (CFIP) scale by Smith et al. (1996), ask participants a general risk question (Dohmen et al., 2005) and collect demographic characteristics to control for age, income, or gender effects.
By intention, participants complete all questionnaires that might affect their valuation of personal data after the BDM mechanism. As evident from the literature, the elicited valuation of personal data can be easily influenced by providing additional information beforehand (e.g. Winegar and Sunstein (2019) or Collis et al. (2021)).
If we observe high dropout rates due to participants' exhaustion in stage 2 of our trial run, we might have to adjust the order of the experiment.