Peer-to-peer Discrimination in the Sharing Economy
Received the Franco Nicosia Award
Background
Despite the rapid expansion of the sharing economy, it has become evident that not all consumers have equal access to it. They are subject to screening by peer providers based on race, gender, sexual orientation, age, religion, occupation, and even marital status. These discriminatory practices curtail opportunity for some consumer groups to use the sharing economy.
Research Question
In addition to these demographic characteristics, whether the location/residency information displayed on consumers’ profiles (e.g., San Francisco, CA on Airbnb & Turo, CA on Lending Club & Prosper), affects their vulnerability in discriminatory screening.
More particularly, how does economic inequality in a consumer’s region, raging from local community to country, influence their access to sharing economy?
Hypotheses
-
Consumers from regions with higher economic inequality are less likely to be served by peer providers, experiencing reduced to access to the sharing economy.
-
These consumers experience reduced access to the sharing economy because they are perceived to be less trustworthy and more financially risky by peer providers.
-
The negative effect of economic inequality on consumers’ equal access to the sharing economy can be mitigated by an increase in perceived similarity between the peer provider and the consumer.
Studies
1. Observational data
Data: cross-sectional transaction data from a leading P2P lending platform (2010 - 2014, 48 US states, N = 54k+)
Analysis: fixed-effect robust linear model & fixed-effect quantile regression
Key Insights:
With each 0.05 increase in the Gini index within a state, the average lending amount from peer lenders (providers) for a loan requested by a borrower (consumer) from that state decreases by approximately 14%.
2. Survey
Sample: individuals with real-world experience as peer providers recruited from Cloud Research
Weighting Method: Raking based on gender, age, and race
Analysis: Pearson correlation (weighted & unweighted
Key Insights:
A significant negative correlation between peer hosts’ (providers’) perception of inequality in a guest's (consumer) region and their willingness to accept the booking request from the guest was observed (r = -.28, p = .007).
3a. Thought Protocol Analysis
Sample: individuals recruited from Amazon Mechanic Turk
Design: one-way between-subject experiment with manipulated economic inequality, and open-ended question for thought protocol
Analysis: content coding, logistic regression, casual mediation analysis
Key Insights:
On average, peer tool owners (providers) are 7.7% less willing to rent a large toolbox to the tool renter (consumer) when the renter is from a region with high (vs. low) economic inequality.
On average, peer tool owners (providers) are 51% less likely to express trust in tool renter (consumer) when the renter is from a region with high (vs. low) economic inequality.
A casual mediation analysis confirms the mediating role of perceived trustworthiness .
3b. Experiment
Sample: individuals recruited from Cloud Research
Design: one-way between-subject experiment with manipulated economic inequality
Analysis: t-test, Welch’s ANOVA, Moody’s median test, & mediation analysis
Key Insights:
On average, peer lenders (providers) intend to fund 35.8% less to a borrower (consumer) when the borrower is from a region with high (vs. low) economic inequality.
A significantly lower proportion (-21.2%) of peer lenders intend to fund an amount above the sample median to a borrower when the borrower is from a region with high (vs. low) economic inequality.
A serial mediation analysis supports Hypothesis 2.
4. Experiment
Sample: individuals recruited from Cloud Research
Design: mixed-design experiment with manipulated economic inequality and measured perceived similarity
Analysis: t-test, OLS regression
Key Insights:
As peer hosts (providers) perceive a guest (consumer) as more similar, it MITIGATES the observed regional economic inequality based discrimination.
Conclusion
Despite the explicit policy of the platform, which prohibits the refusal of service to consumers based on criteria such as race, color, ethnicity, national origin, religion, sexual orientation, gender identity, or marital status, our findings add to the mounting evidence of peer-to-peer discrimination within the sharing economy.
They underscore the pressing needs for more robust measures to combat this issue.
Managerial Implications
Introduce an enhanced provider-consumer MATCHING algorithm
Integrate self-disclosed information from both providers and consumers as an additional feature to calculate the similarity score and facilitate matching (in addition to the preference and booking history features)
Possibly include a set of demographic characteristics, but caution should be exercised to ensure this inclusion does not introduce any biases
Introduce the display of the provider-consumer SIMILARITY score
To increase the salience of the similarity between a provider and a potential consumer, thereby diverting providers’ attention from other bias-susceptible characteristics of the consumer
To partially address the cold-start issue by assisting discrimination-prone consumers in obtaining their initial service and reviews.