Measuring Social Media Network Effects: Evidence from Online Choice Experiments

How to cite

Benzell, Seth; Collis, Avi; Nicolaides, Christos; Aral, Sinan. 2023. "Measuring Social Media Network Effects: Evidence from Online Choice Experiments." AEA RCT Registry. AEARCTR-0012300. https://doi.org/10.1257/rct.12300-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Seth Benzell* , Avi Collis , Christos Nicolaides , Sinan Aral

* Primary investigator

Registry ID
AEARCTR-0012300
DOI
https://doi.org/10.1257/rct.12300-1.0
First registered
Last updated
Status
In development
Trial dates
2023-10-18 – 2023-12-15
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/12300

Abstract

We have already conducted a series of surveys through Lucid on individuals’ demographics, connections on a digital platform, and their valuations of these connections and digital platforms.

In this document we are pre-registering a follow-up survey and analysis plan. The survey will collect information on 2,500 users of each of four popular social media platforms (Instagram, Facebook, Twitter/X, and LinkedIn), with a focus on their number of contacts, valuation of the platform, social media usage, and other socio-economic and demographic information.

The first question we are interested in is the relationship between the number of contacts and platform valuation. Our main specification will be a linear regression with an intercept term. In secondary specifications we will replace the number of contacts with the log of the number of contacts, and include additional linear controls.

We are also interested in how platform valuations vary with socio-economic, demographic and platform usage characteristics, as well as question framing. Our main analysis will first select covariates (and their interactions) with explanatory power using LASSO. We will then run a regression of platform valuation on the LASSO selected coefficients and interactions. As secondary specifications, we will estimate how the average platform valuation varies by demographically defined sub-groups.

Experimental design

See attached file describing planned data collection and analysis.

Design details

Randomization method
There are two levels of randomization in the study: 1) Survey takers will be randomly selected into one of three groups, each with a different top-coded value for platform valuation. This randomization will be performed through qualtrics. 2) We will then use an office computer to randomly select a subset of users to participate in a follow-up offer (See attachment for details) 3) We will finally use an office computer to randomly select an offer to make the survey taker to give up the platform for a month.
Randomization unit
Individual survey taker
Clusters
10,000 individual survey takers
Observations
10,000 survey takers
Arms
See attached document for details
Primary outcomes
Research Questions for the Data Collection and Analysis: >How does platform valuation vary with the number of contacts on the platform >How does platform valuation vary with other user characteristics Our primary outcome for both of these research questions is the monthly platform valuation of a survey taker for the platform we are asking about.

Cited by

No citation data available for this registration yet.

Keywords