Anmerkungen:
Nach Informationen von SSRN wurde die ursprüngliche Fassung des Dokuments May 31, 2022 erstellt
Beschreibung:
Online platforms, marketplaces and retailers typically use ranking algorithms to determine the order in which hundreds or thousands of choices are presented to consumers. While ranking algorithms may aid consumer choice, there are concerns they may also lead to socially undesirable outcomes. In this research, we ask two questions. First, we examine the impact of ranking algorithms on consumer choice and the degree to which researchers may obtain biased estimates of preferences if abstracting from the algorithmic code or the rank order of search results. Second, we ask whether ranking algorithms can further socially desirable outcomes. We use data and the ranking algorithm obtained from the US educational crowdfunding website DonorsChoose and develop a structural model of donors’ contributions using a multiple discrete continuous choice framework. We demonstrate that not accounting for the ranking algorithm leads to a systematic bias in estimated consumer preferences. In two sets of counterfactuals, we then test how well DonorsChoose’s algorithm serves its objectives to both benefit disadvantaged groups and achieve a high rate of project completion. First, we show that removing the parameters from the algorithm that prioritize projects from high and highest poverty schools reduces contributions to such schools by 12.98 percentage points. Second, we find that the inclusion of parameters designed to increase the number of projects that succeed on the platform do not substantially affect overall contributions to projects from schools with high and highest poverty. To the ongoing debate about algorithmic bias, we add empirical evidence that algorithms can positively affect disadvantaged groups without compromising a platform’s overall goals