Should Uber modify its driver compensation algorithm to address the detected gender pay gap?
Despite having an alleged gender-neutral algorithm, male Uber drivers earn 7% more than female drivers. Should Uber act to correct this gender pay gap?
The study
In a 2018 study conducted by University of Chicago, Stanford and Uber economists, where 740 million trips and 1.8 million drivers in the US were analyzed, the researches concluded that, on average, men earn 7% more per hour than women [1].
The explanation for the pay gap
Three were the reasons behind the gap:
- Speed: Women generally drive slower than men (or tend to use longer routes), and hence are able to complete less trips per hour.
- Experience: Women tend to churn more often (65% in the first 6 months for men vs. 76% in the case of women), and hence have lower average experience. More time in the platform helps the driver act strategically regarding which trips to accept/decline, routes and other aspects that affect pay.
- Location: Women tend to avoid areas with higher crime rates, into which the algorithm derives some drivers through higher compensation.
No illegal discrimination?
As part of the conclusions, the researches highlight that Uber uses a gender-blind algorithm and that compensation is tied to output, an objective criterion. In other words, there is no intentional/direct discrimination embedded into the algorithm.
However, some legal experts – especially in Europe – claim that, based on the proven gender pay gap, the algorithmic-based compensation could constitute what the European Court of Human Rights and the Court of Justice of the European Union defines as indirect discrimination [2].
Indirect discrimination occurs when there is a policy or rule that apparently applies in the same neutral manner for everyone, but in practice disadvantages or has a worse effect on a group of individuals who share a protected characteristic – gender in this case. Unless there is an objective and reasonable justification strong enough to supersede the interest of gender pay equality (in this case), such policy or rule should be deemed discriminatory and hence illegal [3].
In the Uber case, is the algorithm-determined differentiated pay based on neighborhood safety a powerful enough reason that should prevail over pay equality? Similarly, is it reasonable to use a compensation tool that rewards speed if it economically penalizes women for slower (and safer!) driving?
In my view, the compensation algorithm is used to incentivize productivity, output (number of trips) and hence profits (both for Uber and the driver), all of which are valid and legitimate reasons. However, given that the implementation of the tool leads to a gender pay gap, it is questionable that the individual goals of company productivity and profit maximization should prevail over the societal benefit and fundamental right to equal pay. In any case, the weighting of these interest is tough and the answer is not evident.
Moreover, some people could argue that, as opposed to what the study claims, the criteria of the algorithm itself is not gender neutral. For example, by rewarding driving in area with higher crime rates, the system fails to account the fact that, unfortunately, women have traditionally had higher personal safety concerns than men [2].
Change the algorithm?
In light of this discussion, should Uber change the algorithm? In my view it could, by (i) removing any parameters in the algorithm that negatively affect women (eg. surge pricing of unsafe zones); and/or (ii) including parameters that could potentially positively affect women (eg. safety, speed regularity). Plus, Uber could raise awareness of the biases of its algorithm when onboarding new drivers and provide training and best-practice advice on how to mitigate them. I believe this is both legally and ethically sensible, but also acknowledge the complexity and cost of implementing such measures.
LPA class, what is you view?
[1] Cook C., et al., The Gender Earnings Gap in the Gig Economy: Evidence from over a Million Rideshare Drivers, https://web.stanford.edu/~diamondr/UberPayGap.pdf , accessed 04/13/2020
[2] Todolí, A., Uber saca un estudio que culpa a las mujeres de la brecha salarial, https://adriantodoli.com/2018/02/27/uber-saca-un-estudio-que-culpa-a-las-mujeres-de-la-brecha-salarial/ , accessed 04/13/2020
[3] Eurofund, https://www.eurofound.europa.eu/observatories/eurwork/industrial-relations-dictionary/indirect-discrimination , accessed 04/13/2020
Hi Counselor! Very interesting thoughts on the existing incentive metrics for drivers and how they could influence the pay gap. I agree that the compensation algorithm should be thoughtfully crafted to discourage risky driving behaviors or gender-based barriers, and I think to truly understand the data there needs to be more user research on the female driver population to understand their personal choices and preferences to provide effective support. Instead of being “gender-neutral”, I wonder if having the option of tailored support and education programs could benefit some drivers. I also recently learned that it’s very sensitive for apps to have any sort of geography-based “safety” metrics, as safety could be a subjective/relative concept that could lead to accessibility issues for certain neighborhoods / socioeconomic groups, so definitely curious how the surge rate policies could be adjusted to serve both economic and societal needs.
To Serene’s point, I also wondered how a “safe” location could be conflated with a location “filled with people that look like me” since that is so often seen as a driver of whether someone “feels” safe or not. I’m not sure I have an answer on this, but I’m personally leaning towards leaving the algorithm the way it currently stands.
Interesting article, and great points Serene! I actually find it pretty concerning that speed could be such a significant driver of income, since route guidance is provided by Uber in the first place – so faster trip times than expected would mean that the driver is likely speeding (which is breaking the law, and should not be encouraged). I also found it interesting in the post that women have a significantly higher churn rate than men, and I would want to know more about why this is the case. If fears regarding physical safety are a component of this, that is certainly something that Uber should address.
I actually found the results of this paper very encouraging, since it detailed precisely the mechanism through which Uber could act to further reduce the gender wage gap. However, to answer the question raised in the title of your blog post, I do not think that Uber should adjust the driver compensation algorithm to address the wage discrepancy, but rather implement other initiatives to help close the gap.
First, approximately 33% of the wage gap is attributed to experience on Uber’s platform. The presents a lot of opportunities for intervention. While there is no substitute for experience, Uber could partner with drivers to develop educational material to help ramp up inexperienced drivers with the platform. As part of this, Uber could provide additional information on the mechanisms of the Surge map, since 16% of the gap is attributed to male driver’s decisions to to drive more in areas with higher surge and lower wait times.
If anything, I think this article reveals the value in collecting, measuring and analyzing data. It enables to identify when gender gaps emerge and identify interventions to help address these issues!