1This article examines the ways in which companies developing digital platforms use opaque algorithms—which have agency over workers’ autonomy and their working hours—to organize work. It explores how we can obtain transparency about how drivers’ work is organized and how algorithms assign trips and transform the temporalities of work. We provide empirical evidence about the so-called “algorithmic management” (Möhlmann et al. 2021) phenomenon in Uber, a ride-hailing platform operating in the gig economy, from a case study exploring Uber’s functioning in Paris and the perspectives of its drivers, through an analysis of their personal data.
2Algorithmic management replaces human managerial functions of work with automated systems using algorithms, allowing the possibility of matching the supply and demand (Möhlmann et al. 2021), and the control of drivers’ working time in opaque ways (Van Doorn 2020). The literature on the gig economy in the fields of sociology of work, socio-economics and law has shown that platforms propagate a myth of entrepreneurship (Aloisi and De Stefano 2022), promising flexibility and autonomy in their working time via their independent contractor’s status (Rosenblat and Stark 2016). Although some workers learn to regulate their working time (Seo et al. 2022), algorithmic management actually constrains workers’ freedom (Cano Renau et al. 2021). Workers for such platforms are paid for each task they complete—which are often small actions completed in a short time—while suffering precarious working conditions and working excessively to achieve a certain revenue (Casilli 2019). Piasna (2023) explains that in digital capitalism, on the one hand:
[…] working hours are further shortened by the optimization and automation of some tasks. The outcome is atomized and punctuated working time; that is, a patchwork of ever-shorter units of paid working time scheduled in irregular and discontinuous patterns according to business demand and intertwined with unpaid or non-work periods. On the other hand, however, the time allocated to work is expanding and stretching in various forms, depending on the context of implementation (Piasna 2023: 6).
3Uber pays drivers per trip based on a combination of factors, including a base fare for each trip, charges per km driven, and per minute spent on the trip. Factors are not all clearly specified by Uber and vary per city (Rosenblat 2018; Pidoux et al. 2024). Consequently, drivers in Paris often work extended hours to achieve a certain revenue or a minimum subsistence requirement to live (Benvegnù and Bernard 2023).
4Among other types, Uber uses “surge pricing” algorithms – a dynamic pricing strategy that raises fares when there is high demand from clients and a shortage of available drivers, to equalize the supply and demand ratio.1 These algorithms use data to continuously learn about the market and fix the price of every trip in real-time, at the precise moment when it is ordered by a client (Eyert et al. 2022). To our knowledge, no studies have yet been conducted into how data feeds Uber’s algorithms, which would enable us to understand how these systems transform work organization and working time.
5Based on the personal data of Uber drivers in Paris – collected by exercising their data access rights under the EU’s General Data Protection Regulation – we provide transparency on how Uber’s algorithms allocate tasks (i.e. trips drivers accept), and use geolocation to calculate distance. This allows us to show that for drivers, work organization is segmented into three distinct units of working time, i.e. searching for; approaching; driving with passengers. Finally, we show that this work organization is directly linked to the behavior of the market, which has an impact on the fares generated by drivers, without the drivers having access to the information about the market that they would need in order to mitigate the financial risks they are forced to undertake as independent contractors.
6Our analysis is based on 12 data sets of drivers active between January 2019 and December 2022, with four to seven years of experience at Uber. This period was selected as the drivers indicated that the COVID-19 pandemic had affected their work organization, and therefore it allowed us to analyze the significant fluctuations in demand. Important to note, however, that our analysis does not focus on the impact of the pandemic. In partnership with NGO PersonalData. IO and company Hestia. AI, the data sets were extracted and analyzed with the drivers, following a participatory methodology in line with a “citizen social science” approach (Albert et al. 2021). This article draws on only a small part of the data collected and does not claim to cover all aspects of Uber’s algorithmic management. Its main aim is to understand how drivers can gain transparency and autonomy over the calculation of their working time, and the extent to which Uber’s opacity affects the fares they can generate.
7First, we review the relevant social science literature that analyses algorithmic management in relation to platforms’ opacity. Then, we summarize the sociopolitical context of regulating platforms and work, both globally and specifically in Paris. Second, we present our methods, including defining “digipower” – a participatory methodology that allows us to engage with drivers in data analysis. Third, the results provide a definition of working time according to how data is collected by the app and drivers’ practices. We also present descriptive statistics about the drivers’ overall activity and four types of analysis related to the organization of drivers’ work into three working-time units, and to France’s minimum wage. Finally, our discussion centers around the definition of working time for Uber drivers and a phenomenon we call “algorithmic predictability risk”. We conclude with a summary of the study’s main contributions and research perspectives.
8“Time efficiencies achieved by algorithmic management have tended to be predominantly realized in an asymmetric way, with employers indeed able to organize and purchase labor input in shorter time units (atomized and punctuated time) yet conditioned on workers shouldering the related costs and filling in unpaid gaps in punctuated time through incessant availability” (Piasna 2023). These time efficiencies have not been studied empirically for a set of drivers; platforms exercise “calculative power” (Van Doorn 2020) by hiding the formulas used by the algorithms, while workers lack the means to calculate their future income. Workers’ revenues depend on the algorithmic calculation of distance and time to complete a task, as this determines the fare price before the workers have accepted the task (ibid.). Gaining transparency would allow them to better assess which trips to accept, as well as how to organize their working hours.
9Rosenblat (2018) highlights the ways in which algorithmic management creates information and power asymmetries between drivers and the platform. Drivers claim to lose money through “up-front pricing” discrepancies. Up-front pricing is a strategy used by Uber “to calculate in advance the estimated trip time and distance from origin to destination”2, but the price may vary during the course of the trip. Drivers identified unexplainable discrepancies. They reported instances where passengers are charged a higher fare than what is displayed on the driver’s app, there are unpaid cancellation fees, missing tips, and manipulative surge pricing. Drivers describe the surge pricing system, i.e. increasing the fare prices when there is high demand and few drivers, as manipulative, as Uber uses gamification strategies to encourage them to drive to bonus zones displayed in the app’s map, where there is a possibility of obtaining more trips at better rates (Rosenblat 2018). However, drivers do not know how the surge pricing is calculated, and there is no mechanism that guarantees they get the trips. To our knowledge, no empirical study on Uber drivers’ work organization in Paris has yet been conducted to understand how their working hours are organized through algorithmic management, and the impact on fares.
10Uber considers drivers to be “independent contractors”, and Uber itself as a “neutral technological intermediary” (Rosenblat and Stark 2016), mainly in charge of a “matching” feature (Stark and Pais 2021; Piasna 2023), regulating supply and demand in the market through surge pricing algorithms. As such, drivers are in principle free to work hours that fit their personal schedule or during profitable periods (Casilli 2019). However, platforms incentivize workers to remain available during off-peak times, when there is low demand, through symbolic or economic rewards, and platforms might penalise those who do not comply. This effectively controls drivers’ working time, despite their supposed independence as independent contractors (Ibid.). Now courts around the world have ruled that drivers are formal employees (Rosenblat and Stark 2016). In a judgement of January 10, 2019 (6-2, RG 18/08357), the Paris Court of Appeal found that despite being classified as self-employed or independent contractors, the nature of the relationship between Uber and one of its drivers was akin to an employment contract, given the degree of control Uber had over the driver’s work organization, such as fixing rates, terms of service, and performance evaluations, which are typical characteristics of an employment relationship. This judgement challenged the previous classification of drivers, but did not automatically reclassify them all as employees. If a driver seeks to be reclassified as an employee, each case needs to be assessed in court. With the help of lawyers, drivers individually challenge their status with no real knowledge of Uber’s operations or means of calculating working time, which is one of the main instruments to define an employment relationship. One study in Paris states that drivers are typically segmented into two temporal regimes: some work intensively for the platform to achieve the revenue thresholds they set for themselves, while others work only enough to balance their professional, family, and personal lives (Benvegnù and Bernard 2024). This reflects a broader trend in France, where “the majority of male platform workers engage in secondary or marginal work intensity in digital labor platforms” (Rodríguez-Modroño et al. 2022). However, the studies conducted do not provide a precise calculation of working time. We aim to address this gap.
11Our study adopts a “citizen social science” approach (Albert et al. 2021) to produce scientific knowledge related to the concerns of social actors and in cooperation with them. Hence, in our study drivers were actively involved in various stages of the research process, i.e. data collection, analysis, and interpretation in accordance with their own concerns, i.e. regulating their employment relationship with Uber. This process occurred outside academic contexts, e.g. in parking lots and at the airport where drivers were waiting for trips. We referred to this participatory methodology, which focuses on analyzing personal data, as “digipower” (Pidoux et al. 2022). It is based on recovering personal data via a Subject Access Request (SAR) and situating the data in social actors’ experiences through continuous interactions. A SAR means that an individual can request—for themselves or on behalf of another individual—the information to which they are entitled under Art. 15 of the EU’s General Data Protection Regulation (GDPR).3 The approach taken in data protection law, “empowers individuals to obtain access to data that is of specific interest to them” (Thouvenin and Tamò-Larrieux 2021). This provided drivers a means to gain transparency on how Uber organises their work. Although Uber’s algorithms are opaque, to account for their operations, one can access the variables and data inputs that feed algorithms (Eyert et al. 2022). PersonalData. IO, Brahim Ben Ali (general secretary of trade union Intersyndicale Nationale VTC (INV) and founder of the cooperative Maze4), and Hestia. AI organised two “digital residences” in Paris (December 2022, January 2023) and contributed to developing tools5 to collectively understand the data. 120 drivers exercised their data access rights, and we discussed the results of this study with 10 of them. Here, we present preliminary results from a data analysis of a select sample of drivers’ data sets obtained from Uber, as our study is ongoing. We are expanding our data sample and coding our ethnographic observations for further qualitative analysis.
12From the subject access requests made during the digital residences, 15 drivers agreed to share their data with us. To conduct a comparative analysis, we selected drivers working actively in the 4-year period from January 2019 to December 2022, resulting in a reduced data sample of 12 male drivers (table 1). The article does not claim to be representative; to date there are no official statistics about the total number of Uber drivers in EU markets (see Rodríguez-Modroño et al. 2022). One article6 estimates there are 2,480 drivers who participated in a trial against Uber in France, representing 8% of the total number of taxis in the country. This study focused on activity duration and other time-related elements, as age or ethnicity were outside the scope of this analysis.
13The drivers in our sample had between four and seven years of experience with Uber. It is important to note that during the COVID-19 period (2020-2021) the working time and kilometers (kms) driven were zero for some drivers, causing a great variation in the statistics. kms driven per month: mean= 1.780,98; median= 1.562,33 kms; hours worked per month: mean= 79,6; median= 66,87; with a minimum of 0 and a maximum of 267,96 hours. This means that, on average, drivers in our sample are mainly working part-time, or 56% of the standard working week in France (35 hours)7, which is about 19.9 hours per week, with one driver working up to 67 hours per week. This represents a 191,42% occupancy with respect to the standard working week. Although our sample is small, this is the first quantitative study confirming previous qualitative studies (Bernard, 2023; Benvegnù and Bernard, 2024) and providing additional information on the fragmentation of working time. In that sense, our data analysis provides a valuable contribution to further qualitative research and should not be considered purely a quantitative analysis.
14The study complied with personal data protection procedures and with Sciences Po’s charter of ethics and deontology. The data protection measures and processing methodology were approved by Sciences Po’s data protection officer. As such, the drivers entrusted their personal data to this research, exclusively for aggregated and anonymized analyses that do not allow their identification.
Table 1. Data Sample
- 8 All these factors should be verified and may vary over time.
15Once received from Uber, we processed the drivers’ personal data in various ways. First, we cross-checked and verified the information present in different files. Second, we calculated the total distance driven in kms and the “fares generated” (Mishel, 2018). The fares are “net” in the sense that Uber fees and commissions are already deducted, but should not be mistaken as revenues. It does not account for social security, professional costs, or any driving-related expense, e.g. fuel, that drivers would still need to pay. Calculating drivers’ revenues falls outside the scope of our study given the complexity of Uber/driver compensation and the diverse methods used to study Uber pay (Mishel 2018). As a reference, a driver explains “there is 22,2% paid to the URSSAF (Unions de Recouvrement des Cotisations de Sécurité Sociale et d’Allocations Familiales, meaning the organizations for the collection of social security and family benefit contributions); a double VAT: 10% on 7-15% of the fares collected and 10% VAT paid on the 25% Uber commission, as Uber is exempt from VAT because it is considered as an intermediary, in addition to professional costs.”8
16Among all the data types received, we focused on (i) timestamped geolocation data per trip, as this allowed a precise calculation of the Kms driven and the hours worked; (ii) fares generated per trip, as this allowed us to calculate the gross hourly rate of drivers’ working schedules. Finally, the individual data per driver was aggregated. We added together the Kms driven, hours worked, and fares generated for all the drivers, monthly and annually.
17First, the data is mainly organized by trips, and contains information on the timestamped geolocation of the beginning and end of each trip, and the fares generated, which can be found in different files. Images 1, 2, 3 include details of one of the main files used. There are trip values including timestamps regarding when the request was made to the driver (request_timestamp), when the trip began (begintrip_timestamp) and ended (dropoff_timestamp) (image 1). The file also contains information about the fares of every trip per driver (image 2), which was used to calculate the total fares generated per month.
Image 1. Timestamps (as appears in one of the driver’s files)
Image 2. Fares (as appears in one of the driver’s files)
18We merged the time and fare data above with information from another file (image 3) containing geolocation in order to calculate the precise start and end point of each trip. This additional file contains four driver’s statuses (online/offline/en route/on trip), the geolocation (latitude and longitude) at the beginning (begin_lat, begin_lng) and end of that status (end_lat, end_lng), and the exact time spent in every status.
Image 3. Geolocation (as appears in one of the driver’s files)
19Second, Uber provides information that requires verification. Uber presents the distance in miles and kms driven from point A to point B for each trip (image 4). However, this is restricted to one driver’s status ‘on trip’ without considering the other three (see above).
Image 4. Uber’s calculation (as appears in one of the driver’s files)
20Finally, it is important to mention a data error identified: for one month, the miles recorded by Uber were unusually low, shown by the sharp drop in September 2021 (see figures 2 to 4). After confirming with several drivers, who had kms recorded on their pays lips for this period, we have decided to exclude this month from the discussion in our article.
21As Uber does not provide drivers with full calculations of their working activity, we conducted first, the total of kms driven of all trips, considering all driver’s statuses using OSRM (Open Source Routing Machine)9 built using OpenStreetMap data and used for calculating the shortest route. This is an underestimate as it does not take into account traffic or other potential road conditions but serves as the minimum base of what the drivers have worked. Second, we calculated the sum of all trips’ fares for every driver per month and per year to estimate the fares generated.
22In theory, we defined working time as “atomized and punctuated work, in continuous extension and strengthening” (Piasna 2023). In practice, drivers and lawyers need to apply national labor rights to define what is included in the calculation of paid work. This was a main topic of discussion with drivers during our digital residences in Paris. For a driver: “Working starts the moment I leave the house, put the key in the ignition and launch the app”. However, Uber claims paid work is limited to driving with a passenger in the car. In ongoing court cases elsewhere, e.g. Geneva, Switzerland, Uber now also recognises the time spent approaching the passenger prior to pick-up. A remaining concern for drivers was calculating how much time they spend waiting for trips assigned by algorithms. Collectively we achieved a granular distinction of three distinct working-time units (P1, P2, P3) computed by algorithms as follows:
- P1: searching for passengers, i.e. when the driver connects to the app and starts driving, waiting for a trip to be assigned.
- P2: approaching passengers, i.e. a passenger has requested a trip, the driver is on their way or might still be dropping off the previous client.
- P3: driving with passengers in the car, i.e. a trip was accepted, the client was picked up and is being taken to the destination.
23The three working-time units are used in four different analyses: work demand; comparison of hours driven; comparison of kms driven; gross hourly rate, conducted for the 13 drivers together, per month and per year (January 2019-December 2022), to avoid the possibility of them being identified individually.
24The first analysis provides a general view of the work demand in Paris, limited to the allocation of trips by Uber’s algorithms that drivers have accepted. Figure 1 considers all of the Uber’s trips’ five different statuses: completed (if the trip has been fulfilled), cancelled by driver or rider (i.e. client), fare split (the cost of the trip was divided among multiple clients) and unfulfilled (definition unclear in files, could be the trip was ended prematurely).
Figure 1. Work demand in Paris per year according to Uber’s trip statuses
25Figure 1 shows a 52,25% decrease in completed trips from 2019 to 2020, followed by a small increase (27,4%) in 2021. In 2019 drivers cancelled 3,4% of the trips, decreasing to 1,8% in 2021, meaning drivers were accepting most of the trips they received. On the clients’ side, the cancellation rate increased up to 15,6% in 2020. Overall, we observe less work demand since 2020 and a higher acceptance rate of drivers through time, impacted by the increasing cancellation rate of clients. We assume this is directly linked to the pandemic and the fact that Uber’s algorithmic management regulates supply and demand. It demonstrates that drivers are directly affected by market fluctuations and are compelled to accept almost every trip assigned (without knowing how many potential clients there are and how many drivers they are competing against) at any rate. Rates fluctuate according to the supply and demand, with the surge pricing algorithms (Eyert et al. 2022) and are not always to the drivers’ advantage (Rosenblat 2018). Thus, they have little room for manoeuvre to achieve independent contractor growth, meaning a low level of autonomy.
26The second analysis in figure 2 shows the total hours driven and compares the hours according to every working-time unit (P1, P2, P3) and their sum (P123).
Figure 2. Comparison of hours driven per month according to different working-time units (Paris)
27After the clear impact of COVID-19 in April 2020 Figure 2, there is a continuous increase in hours spent searching for passengers (P1), blue line, until Dec. 2022, except for certain months where P3 was higher. In contrast, there is no significant increase in the number of driving hours with passengers (P3), green line, nor approaching them (P2), orange line. The results indicate that the pandemic led to a shift in drivers’ work organization, characterized by low demand. Other likely factors include holiday periods and an increase in the number of new drivers registering on the app. Before April 2020, drivers’ activity was mainly driving with passengers (P3). Later, searching for passengers (P1) also became an important working-time unit. This shows an overall decrease in work demand and an increase in drivers waiting for trips.
28Finally, when the three working-time units are added (P123), red line, we observe on average that drivers worked 95,5 hours per month before April 2020: a 68% occupancy, and 60 hours in the following years: a 43% occupancy, which means also a reduced source of income. The results suggest that following a decrease in market demand mainly due to the pandemic, drivers worked fewer hours and increasingly spent time searching for passengers—a working-time unit that Uber does not compensate—thereby reducing their earnings from actual driving time with passengers (P3).
29The third analysis in figure 3 focuses on the total of kms driven according to every working-time unit (P1, P2, P3) and their sum (P123).
Figure 3. Comparison of kms driven per month according to different working-time units (Paris)
30After April 2020, in figure 3, drivers drove more kms searching for passengers (P1) or about the same as driving with passengers (P3), compared to the period before pandemic where drivers were driving more often above 1500 km per month.
31Uber fixes the rates by a combination of factors that are not clear (Rosenblat 2018) but there are three significant elements in the drivers’ payslips (Pidoux et al. 2024): base rate, compensation per km, and per minute. The combination of the two analyses of hours and Kms driven suggests that drivers might be losing compensation as they are mainly driving without remuneration for searching for passengers (P1), in addition to P2, slightly mitigated by a compensation per km increasing after June 2021 (in P3), but there is an overall loss of fares given the decreasing work demand. Moreover, this shows drivers are actively driving whether they have passengers in the car or not. This is surprising for P1 after April 2020. It means that while drivers are available for the app, they are actively driving and consuming fuel, while waiting to receive a trip from the app. The results point to Uber’s gamification strategies that prompt drivers to move around bonus zones without actually receiving trips (Rosenblat 2018). It also shows the value of this working-time unit to learn about the market.
32The fourth analysis shows drivers’ gross hourly rate in euros according to their fares generated and the hours worked with respect to France’s minimum wage. We took the 2021 minimum wage10 as a reference (in French: Salaire minimum interprofessionnel de croissance (SMIC): gross hourly rate €10,48, gross monthly wage €1.589,47 based on the legal 151,67 working hours per month, 35 hours per week. The gross hourly rate €10,48 is indicated by a horizontal dotted line in red in figure 4. The gross hourly rate is calculated by dividing the total fares generated for all trips (after Uber’s commission is deducted) by the total number of hours worked per month. The hourly rate in euros is analysed according to the three different working-time units to show what drivers and trade unions can claim as paid work. First, we show one working-time unit (P3), which is the initial working-time unit considered by Uber as paid work in France. Second, we add together the two working-time units (P23) that are considered as paid work elsewhere. Finally, we add together three working-time units (P123) that drivers consider as their full working time.
Figure 4. Gross hourly rate in euros per month with respect to minimum wage (SMIC) in Paris
1133Overall, there is a low gross hourly rate if we add together all working-time units (P123), especially given that this calculation does not consider social security, professional costs, nor driving-related expenses. Crucially, drivers depend on the market to generate fares, like taxis (Serafin 2019). However, Uber drivers individually shoulder the financial risk linked to the market because Uber’s algorithmic management is opaque, i.e. not showing drivers all the working-time units, the mechanism of allocating trips among the competition, nor the market variations. Uber drivers continue working even when there is low demand in case further work is available, without the capacity to control the variations of the market. In contrast, Uber has historic data on clients’ and drivers’ activities, as well as the technology to predict the market variations, with the possibility to adapt their rates and strategies accordingly.
34Thanks to the personal data of drivers in Paris, we are able to show that Uber’s algorithmic management organizes work according to three different working-time units: searching for passengers, approaching passengers, driving with passengers that are not all considered paid work. These units are tracked by geolocation via the drivers’ mobile phones, and allow Uber to calculate distances and fares. However, according to Uber’s calculations in the drivers’ files, and the experiences of drivers in court cases, Uber only compensates them for the time spent “driving with passengers”. Our study confirms Piasna’s (2023) theory that algorithmic management is changing the temporalities of work, with time extended or shortened as tasks are atomized and punctuated. By calculating the hours worked, Kms driven, and gross hourly rate, our results show a broader picture of drivers’ work organization. Uber drivers are not organized as suggested by Piasna (2023) in “a patchwork of ever-shorter units” of working time. In contrast to other platform workers who are paid per task (Casilli, 2019), Uber drivers are paid per base fare, compensation per km and minute. Therefore, using geolocation, Uber calculates the distance between a client and a driver’s position, and the client’s destination, as well as the duration of the trip. The principles of this approach—which is specific to mobility services—are similar to taxi drivers’ tachymeter (Serafin 2019). So how are Uber’s algorithms transforming the temporalities of work? Uber is calculating the drivers’ compensation automatically, and in advance, while forcing its drivers to assume the whole financial risk of the market fluctuations, which are in turn controlled by Uber. We can refer to an “invisibiliation effect” of algorithms (Tighanimine 2023). At Uber, a large part of workers’ time is made invisible in spatial and temporal calculations that are partly accessible from the app, in contrast to a tachymeter, which undermines their possible gains.
35Algorithmic management is said to demand constant availability from drivers given the atomization of work (Piasna 2023). However, constant availability is common to taxi drivers (Serafin 2019) and other sectors like cleaning that existed long before algorithmic systems. This availability can be defined according to the labor conventions set out by Salais (1991: 28). One of the conventions is “quality-free” work based on standardization, which means that the personal qualities of employees are disregarded. The occupation knowledge is incorporated into the technology, their work is precise, limited and uncertain, and depends on the specific demands of current and potential customers, with no possibility of acquiring useful information about demand. Individuals are then asked to make their time fully available, as if they were a free, natural resource.
36Uber’s algorithmic management depends on the market behavior, much like other platforms in the gig economy. This “embeds work within an efficient-economy approach” (Cousin 2016). In this model, workers are compensated “in irregular and discontinuous patterns according to business demand” (Piasna 2023). Indeed, this is made explicit by Uber itself and expressed in the principles regulating Uber’s up-front and surge pricing algorithms. However, Uber’s algorithms can understand the market behavior in a situation where the platform collects and owns the data, allowing it to know and control the distribution of the offer (i.e., drivers) with respect to the demand (i.e. clients). The algorithms allow Uber to optimize their profit, but not the drivers, even though they are considered independent contractors. The data allows Uber to predict the distances between drivers and clients, to assign trips to one driver instead of another, and to predict the fares for each trip in advance. The data can also be used to make drivers remain available for longer, to balance supply and demand. Critically, the constant and increasing time drivers spend searching for passengers is essential for Uber’s algorithms to learn in which areas and at what times potential clients can be attracted but all these operations are opaque to drivers.
37Despite Uber’s algorithmic opacity, this study has provided transparency over certain data types and dynamics controlled by Uber, but the drivers’ personal data required verifying, correcting and computing new calculations given the poor quality and incompleteness of Uber’s data. Together with the drivers, we had to build a new “calculative power” (Van Doorn 2020) so that, through personal data, we could understand empirically the work demand in Paris and show how market fluctuations directly impact drivers, forcing them to accept nearly every trip during periods of low demand, thereby limiting their autonomy. In addition, our analysis of drivers’ gross hourly rate compared to France’s minimum wage indicates that drivers often earned significantly above the minimum wage when only driving time with passengers (P3) is considered. This rate saw a substantial increase after October 2021. However, when all working-time units (P123) are considered, reflecting the full definition of drivers’ work, the gross hourly rate drops to €10-€34. The different units of work, to Uber or to drivers, reveal a significant variance in fares generated, depending on the market behavior. In certain periods, drivers have the potential to earn more than the minimum wage, but the instability of these rates and the inclusion of non-compensated working-time units (such as searching for passengers) create economic challenges that drivers face alone in the gig economy.
38As self-employed or independent contractors, in theory drivers must assume the financial risks of the market fluctuations, i.e. losses during the pandemic lockdowns, and can profit from the gains afterwards. However, the design of the app, the algorithms’ calculations and the way they set rates are all decided by Uber. While Uber has full visibility on supply and demand, the drivers do not. Therefore, their autonomy and economic growth depend on the terms set by Uber, and on the unpredictability of the market, with no information on how to control its fluctuations where a new kind of risk is passed on drivers. We call this an “algorithmic predictability risk”. Uber optimizes work organization, which means collecting geolocation to “dispatch” drivers to different trips according to the predicted distance/duration between the client’s and driver’s position, and the client’s destination. These algorithmic predictions made by Uber are later updated according to the trip’s real conditions, e.g. traffic, or following a different route. All this information is possible to know through drivers’ three working-time units. However, drivers that do not have calculative power and autonomy over their data, nor these algorithmic predictions undertake the financial risks caused by the market fluctuation and the way Uber controls it.
39Our study empirically demonstrated that Uber’s algorithmic management, and consequently, drivers’ working time are dynamically influenced by fluctuating demand, with drivers lacking visibility over the market and the stability of a minimum wage, given their self-employed / independent contractor status. Consequently, drivers may work significantly more or fewer hours per month, and drive more or fewer kms, resulting in a variable low-gross hourly rate, that creates precarious work, particularly when calculating the time spent searching for passengers that is not paid by Uber, and as professional costs and driving-related expenses have to be deducted afterwards. The study offers researchers and drivers the first view of the market based on a novel participatory methodology, as this data was previously only in Uber’s possession and never aggregated between drivers.
40For further research, a broader sample of drivers would consider their sociodemographic profiles and different working time practices in relation to their family and personal lives, as these might influence their motivations and the time they commit to the platform. As Benvegnù and Bernard (2023) showed, Uber drivers have different migration trajectories and have two options: to be available to work at almost all hours, or to preserve their personal lives (and their health), but accept a significant drop in their level of income. Other studies could investigate in-depth the market fluctuation with a new indicator, e.g. a monthly or annual rate of change and also consider the calculation of revenues, including professional costs and social security, to better estimate drivers’ gains in the national labor market—as did one study in the United States, which accessed Uber’s “administrative data” (Mishel, 2018). Finally, the empirical evidence we provided can serve to expand an inter-disciplinary discussion on working-time, especially as related to national labor rights.
Flux RSS
