Discrimination on online platforms: a call for regulation

March 15, 2019 417754 Leave a comment

Context

In the housing and rental market, anti-discrimination laws in the US gradually reduced discrimination through the legal system for the past two decades. However, academic scholars (Edelman, Luca & Svirsky, 2017) argue that the emergence of online platforms facilitate discrimination, as these laws do not reach smaller property owners using online platforms. Airbnb, the world largest online platform for short-term rental and housing, adopts a design choice that enables discrimination on its platform. Hosts decide whether or not to accept a guest after seeing the name and profile picture of the guest.

Methodology and experiment

In order to test whether Airbnb facilitated discrimination through its design choice, the authors (Edelman, Luca & Svirsky, 2017) conducted a field experiment across five different cities, including: Baltimore, Dallas, Los Angeles, St. Louis and Washing DC between 7 July 2015 and 30 July 2015 (see Figure 1). Originally, the experiment aimed to gather data from the top 20 cities in the US, but the experiment was halted due to Airbnb’s systems detected and blocked the automated tools used to gather the data.

The experiment gathered a wide range of information about hosts and their listings (see Figure 2). Information of hosts include but are not limited to profile image, gender, age, number of properties listed and previous guests that visited the host. Information on listings include price, number of rooms, cancellation policy, cleaning fee, rating and whether the room was shared or not to control for interaction between the guest and the host.

After gathering data, the experiment sent 6,400 messages with 20 Airbnb accounts. Hosts who offered multiple listings on the platform were contacted for one of their listings to prevent the host from receiving identical e-mails and to reduce the imposed burden. The accounts used to send messages are identical except for two variables: i) race and ii) gender. Race and gender were indirectly embedded in the profiles through the use of names based on Bertrand and Mullainathan (2004). Additionally, to alleviate confounds that would arise from using profile pictures, accounts did not include a profile picture. Finally, the experiment tracked the response over 30 days after the message was sent.

Results

The authors found that guests with distinctive White American sounding names were accepted ±50 of the time, while guests with African American sounding names were accepted at ±42 of the time. The ±8% gap persists across characteristics of the hosts and listings as control variables. More important, the results infer that the discrimination effect occurs in differences of a simple “Yes” or “No” response and not because of intermediate response and non-response. The authors further found that the discrimination effect disappears when hosts previously accepted African American guests. Control variables including homophily concerning race, age categories, price of the listing and demographics of the vicinity are however of no significant influence on the discrimination effect. Discrimination further cause financial consequences, as host incur costs when rejecting guests causes a unit to remain unrented.

Discussion: strengths and weaknesses

This paper provides clear evidence of the presence of discrimination in online platforms. The relevance of this paper is also strengthened by the way it emphasizes discrimination in the online channels, while in the past the focus was primarily on discrimination in offline channels. The results are consistent with other studies on discrimination in the online rental and housing market. Ge, Knittel, MacKenzie and Zoepf (2016) found a similar pattern of discrimination in peer transportation companies such as Uber and Lyft; African American passengers face longer waiting times and more frequent cancellations compared to their White-American counterparts.

The research also has a few flaws. First, the research is not able to detect the type of discrimination that occurs (e.g. statistical discrimination and taste-based discrimination) and whether discrimination is based on socioeconomic status or race that is associated with the name. Second, the paper suggests that the discrimination effect occurs when users of these platforms gain the choice to accept or to reject guests and passengers, which suggests that the problem lies in the platform’s design choice. The suggestions to alleviate discrimination by limiting design choice such as removing information of guests and passengers such as concealing names and profile photos or to eliminate the screening procedures by introducing instant book options as the only option, may harm the user experience for both (hosts and guests) sides. For hosts it is desirable if they can maintain control on who they allow to stay at their place, while for guests the platform is attractive if they can choose the place and host of their liking. When choosing to reduce discrimination by lowering the user experience for either party, online platforms run the risk of becoming less attractive than their competitors and jeopardizing their own competitiveness. Ultimately, discrimination will continue to occur on competing platforms that do not change their design in benefit of combatting discrimination and the non-discriminating company will lose its competitive edge and fail. Third, the inferences made by the paper are to a certain extent limited to the US. A recent study found that racial discrimination is more prominent in the US than in Europe (Pitner, 2018). The focus on metropolitan areas also questions whether the same effect will occur in rural areas. On the assumption that metropolitan areas are more globally connected and face higher exposure to other races, one can logically assume that metropolitan areas are more tolerant and discriminate less against other races.

Airbnb adjusted its non-discrimination policy in 2018. Hosts are no longer allowed to request a guest’s photo before accepting a booking agreement (Thinkprogress, 2018). Based on the research (Edelman, Luca and Svirsky, 2017), the adjustment will not help as hosts can still view names prior to the selection procedure. A potential solution is to increase the prevalence of reviews in the selection procedure. Cui, Li and Zhang (2016) found that encouraging credible peer-generated reviews mitigates the discrimination effect of guests with African American-sounding names on Airbnb. However, we argued that the action of one platform may not suffice as a solution to stop discrimination and call for more regulation on online platforms from authorities.

Sources

Bertrand, M. & Mullainathan, S. (2004). “Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” American Economic review 94 (4): 991–1013.

Cui, R., Li, J., & Zhang, D. (2016). Discrimination with incomplete information in the sharing economy: Evidence from field experiments on Airbnb.

Edelman, B., Luca, M., & Svirsky, D. (2017). Racial discrimination in the sharing economy: Evidence from a field experiment. American Economic Journal: Applied Economics, 9(2), 1-22.

Ge, Y., Knittel, C. R., MacKenzie, D., & Zoepf, S. (2016). Racial and gender discrimination in transportation network companies (No. w22776). National Bureau of Economic Research.

Pitner, B. H. (2018, May 17). Viewpoint: Why racism in US is worse than in Europe. Retrieved March 5, 2019, from https://www.bbc.com/news/world-us-canada-4415809

Thinkprogress. (2018). Airbnb announces booking policy change to head off outcry over persistent racial discrimination. Retrieved fromhttps://thinkprogress.org/airbnb-changes-photo-policy-combat-racial-discrimination-4f71c375553a/

Articles

The influence of the gig economy on entrepreneurial activity

March 15, 2019 koenwalsteijn Leave a comment

TAt the moment, more than one third of the workforce in the United States works (part-time) as freelancer (Muhammed, 2019). Almost half of the freelancers over the age of 55, use supplemental gig work as a means to finance their retirement. It is expected that by 2020, almost half of the entire population will (part-time) perform gig work.

These facts are caused by the rise of the gig economy. A gig economy is an economy in which temporary, flexible jobs are common. In the gig economy, employees usually work part-time as contractors and freelancers, whilst in a classical economy employees work full-time, rarely switch jobs and often work for the same company their whole career (Kenton, 2018). During the last decade the gig economy has experienced a significant rise, caused by three major developments: First, Millennials have a different attitude towards working than the traditional workforce, as they prefer to have variety in work and a good balance between working and private life. Second, since the financial crisis of 2008, companies increasingly use gig workers to lower labor costs and to improve flexibility. Third, due to technological advances and the digitalization of the world, gig economy platforms were introduced, improving the access to gig work and workers for both firms as the labour force (Muhammed, 2018).

However, whereas the rise of the gig economy has several welfare benefits (Jennings, 2018), there are also possible downsides to it. In the paper “Can you gig it? An empirical examination of the gig economy and entrepreneurial activity.”, authors Gordon Burtch and Seth Carnahan researched a possible downside, whether the introduction of a gig economy platform in a geographical area would lead to a decrease in entrepreneurial activity.

Context & methodology

The authors used the introduction of Uber X into a specific geographical area as measure of an introduction of a gig economy platform. Uber X was chosen for two reasons: First, Uber enters the economy city-wise, meaning that for a specific area, the service might be available in one city, but might not be in another. This makes it possible to directly analyse the results of the entry of Uber X in a certain area. Second, Uber X is used instead of the (premium) Uber Black service because of the lower entry barrier and a larger network of drivers. Additionally, to measure the influence of entry barriers of the gig economy on the observed effect, the results gained by researching Uber X are compared to those of Uber Black.

The research was conducted as a natural experiment. The authors used real life examples of entries of the gig economy and measures of entrepreneurial activity. The researchers measured entrepreneurial activity by combining 2 data from 2 sources:

First, the US Census Current Population Survey (CPS). In this survey, self-employment is measured.
Second, Kickstarter campaign data was used, where the volume and size of Kickstarter campaigns were measured.

Results

The research found a significant negative correlation between the introduction of the gig economy and the entrepreneurial activity within the same geographical area. This negative correlation can be explained by the fact that gig economy platforms offers a source of income for the unemployed with a very low entry barrier, so there is no direct need for them to engage in entrepreneurial activity in order to ensure an income.

However, the research also found a significant increase in the average quality of entrepreneurial activity. According to the findings, this can be explained by the fact that people with mediocre or bad ideas that would engage in entrepreneurial activity without the entry of a gig economy platform, are now participating in that gig economy. The people that are still convinced of their own entrepreneurial qualities are pursuing their idea and don’t give it up to start driving for Uber X. These people apparently have a higher quality entrepreneurial activity, which increases the average quality of the entrepreneurial activity.

Additionally, a significant positive correlation was found between lower entry barriers of a gig economy platform and its effect on entrepreneurial activity. The research found that the introduction of Uber X (which has a low entry barrier), leads to a bigger decrease in entrepreneurial activity and a higher increase in average quality of entrepreneurial activity, as opposed to the introduction of Uber Black (which has a higher entry barrier: you have to buy a black car).

Strengths

The paper written by Burtch and Carnahan (2018) included several strengths. First, the paper included several robustness checks regarding the validity of data. By making use of multiple measures for the same phenomena, it is more likely that the correct effect was captured. Second, another strength of this paper would be the novelty of the research, as the paper is the first to examine the supply side of the gig economy. By researching novel subjects, the added-value to literature of this particular research area is higher. Third, as the paper makes use of two different data sources to measure entrepreneurial activity, the reliability of the data used in this paper is higher.

Weaknesses

Besides several strengths, the research done by Burtch and Carnahan (2018) has several weaknesses as well. First, a weakness of this paper would be the low generalizability of the sample. The research is solely performed in the United States, which is a country with low unemployment benefits. According to Cowling and Bygrave (2002) has a significant correlation with entrepreneurial activity. Having only performed research within the United States, the generalizability of the results is lower, which reduces the findings’ value.

Second, the paper’s assumption that Uber drivers would otherwise engage in entrepreneurial activity, might be too big of an assumption. As Uber offers low-skilled jobs, its drivers might not be the people who would otherwise engage in entrepreneurial activity. People might even start driving Uber X to make some money, but use the flexibility in planning to actually start a business next to driving for Uber X. Moreover, the paper assumes that the entry and timing of Uber X is exogenous with respect to entrepreneurial activity. These assumptions both deteriorate the research quality, but proved necessary to perform the research.

Lastly, the paper contains data issues. For the first time, the volume of Kickstarter campaigns was used as a (partial) measure for entrepreneurial activity. Kickstarter campaigns may be run multiple times, which implies that the quality of the data might be lower than expected. Another reason why the data from kickstarter might not be generalizable, is that it was obtained during a period of economic recession, during which unemployment rates are higher than normal, which could mean that the impact might be different than it would be during a ‘normal’ period of time.

Implications

The paper proves worthy both in research and in practice. For academics, this paper proves valuable as it is one of the pioneers in researching the supply side of the gig economy. On top of that, the researchers also pioneered in using kickstarter as a measure for entrepreneurial activity, despite its limitations. For policymakers debating about the legality of gig economy platforms as Uber, this paper could provide relevant insights for crafting legislation around gig economy platforms.

In many countries Uber faces legal problems.

Conclusion

The paper provides new information towards research about the gig economy as it offers insights on the supply side, which was rarely researched previously. The insights of this paper are relevant but should be taken cautiously however, as there are multiple notable concerns towards the validity, despite the efforts of the authors to ensure the quality.

References

Burtch, G., Carnahan, S., & Greenwood, B. N. (2018). Can you gig it? An empirical examination of the gig economy and entrepreneurial activity. Management Science, 64(12), 5497-5520.

Cowling, M., & Bygrave, W. D. (2002). Entrepreneurship and unemployment: relationships between unemployment and entrepreneurship in 37 nations participating in the Global Entrepreneurship Monitor (GEM) 2002. In Babson College, Babson Kauffman Entrepreneurship Research Conference (BKERC) (Vol. 2006).

Gooch, K. (2018). More Americans turn to crowdfunding for medical bills: 6 things to know. Many Americans struggle to afford their medical bills and are increasingly turning to crowdfunding for support, reports Yahoo Finance. Retrieved from https://www.beckershospitalreview.com/finance/more-americans-turn-to-crowdfunding-for-medical-bills-6-things-to-know.html

Jennings, M. (2018). 7 Reasons Why the gig economy is a Net Positive. Retrieved from https://www.entrepreneur.com/article/310685

Kenton, W. (2018). gig economy. Retrieved from https://www.investopedia.com/terms/g/gig-economy.asp

Muhammed, A. (2018). 4 Reasons Why The gig economy Will Only Keep Growing In Numbers. Retrieved from https://www.forbes.com/sites/abdullahimuhammed/2018/06/28/4-reasons-why-the-gig-economy-will-only-keep-growing-in-numbers/#6344d33e11eb