Display settings

Pick a theme to customize the website's look.

The French mobility orientation law on transport (2019) : an initial analysis of the statistical indicators published by digital labour platforms in France

Subject to an obligation of transparency, digital labour platforms have been publishing indicators on the activity of workers on delivery platforms or ride-hailing services since 2022. PEReN has studied these new data and offers a first unpublished critical analysis.

Data about the activity of delivery platforms’ workers and ride-hailing services’ workers are very scarce. However, digital labor platforms are currently subject in France to new legal obligations of transparency about workers’ activity and have thus been publishing since 2022 statistical indicators that could lead to a better understanding of the sector. Utilizing the publicly available data aggregated by ARPE (the French Authority for Labor Relations on Digital Labor Platforms) and made available via their website, PEReN has conducted a critical analysis of these statistical indicators. This original exercise provides a glimpse into refinements that could be explored for future legal amendments of these statistical indicators.

A legal obligation of transparency for a better understanding of the sector

Under the French mobility orientation law on transport (LOM) of 24 December 2019 creating Article L. 1326-3 of the French Transportation Code, digital labour platforms (ride-hailing services and goods delivery platforms) meeting certain criteria (definition provided by Article L.1329 of the Transport Code) are subject to the obligation of publishing on their websites a list of statistical indicators relative to the waiting time, working time and earnings of workers. The publication modalities of these statistical indicators are detailed in the application decree n° 2021-501 of April 22, 2021, as well as in the order of January 12, 2022 detailing the modalities related to the presentation of these indicators and the order of January 12, 2022 detailing the method for determining non-representative values in the calculation of the indicators relative to workers’ waiting time. With these provisions, the legislator aims at improving the understanding of the ecosystem of digial labour platforms, which are frequently criticized for their lack of transparency, and for which the academic community struggles to find reliable, representative and exploitable data.

In line with these new obligations, on March 1, 2022 and 2023, digital labour platforms published their activity indicators for 2021 (September to December) and 2022 respectively. ARPE collected and aggregated them, and made them available on its website in an easily downloadable and usable format. At the same time, it commissioned PEReN to produce graphs based on the published data (see the statistical indicators for ride-hailing services and the statistical indicators for delivery platforms). Continuing this collaboration, PEReN has carried out the first succinct analysis of these new aggregate indicators. The observations that follow, although preliminary, should enable different audiences (platform workers, researchers, regulators) to better distinguish the contributions of these indicators, what they allow us to conclude, their potential limitations and the avenues that would enable us to extract even more information from them.

A series of statistical indicators with multiple components

The implementing decree no. 2021-501 of April 21, 2021 specifies three statistical indicators to be published each year by digital labour platforms, namely working time, earnings and waiting time. The later indicators should be reported with different parameters:

  • the covered period: a single service, total number of services per week or per month;
  • the number of services provided over the covered period: different brackets of number of services per week or per month;
  • different time slots within these periods: day, night, week days or weekend.

Figure 1 : Description of the data structure, including multiple sub-categories for each statistical indicator. Reading : For a given year, a digital labour platform must provide the value of average earnings in euros, as the sum of average earningss from services provided over a one month period, for workers having completed 3 to 30 services per month, between 6am and 10pm (day).

These sub-categories lead to a total of 95 expected values per year for each platform (see figures 1 and 2), including 5 values for waiting time (per service, per period), 45 values for earnings and 45 values for working time (5 values per service and 40 values per month or week). It should be noted, however, that no data on the number of workers or on the different quantiles of these statistical indicators are reported. For further details, calculation methods and any limitations to be taken into account before any economic interpretations, please refer to the appendix. ARPE has collected and compiled publicly available data for the following platforms:

  • ride-hailing services platforms: Allocab, Bolt, Caocao, Free Now, Heetch, Le Cab, Marcel and Uber;
  • delivery sector: Deliveroo, Stuart and Uber Eats.

Within the compiled data, we find all 95 values for delivery platforms in 2021 (Q4) and 2022. However, ride-hailing services platforms omitted an average of 19 out of these 95 indicators in 2021 (excluding Marcel and LeCab, for which no indicators were provided in 2021 Q4), and 3 in 2022. While this increase is mainly due to Caocao (80 missing values in 2021 (Q4) to zero omissions in 2022), the remaining platforms have also reported more statistical indicators in 2022 relative to 2021 (Q4), as illustrated in Figure 2. For the present publication, the code used to process the statistical indicators and generate the graphs is available via the PEReN repository platform.

Figure 2 : Number of indicators, taking into account all the different categories, published by ride-hailing services platforms in 2021 (Q4) and 2022.

Which initial findings can be drawn from these new statistical indicators?

In order to be able to compare the statistical indicators reported by platforms, we have constructed two new variables based on earnings and working time in minutes:

  • hourly earnings, including the waiting time between the end of a service and the reception of a new service proposal;
  • hourly earnings, excluding waiting time.

While these variables enrich the possible findings, it is important to remember that the waiting time indicator provides information only on the waiting time before a service is proposed, and not on the waiting time before a given proposal is actually accepted. Some platforms may have had a different interpretation of this indicator. The appendix sets out a number of important precautions and limitations with regard to these indicators and the subsequent findings. In what follows, we present a number of observations arising from the analysis of these new data. We focus primarily on ride-hailing services platforms, as the data made available by ARPE currently contains more ride-hailing services platforms than delivery platforms.

Ride-hailing services platforms: a segmentation based on hourly earnings per service (Figure 3)

Two observations emerge from the inspection of the hourly earnings of ride-hailing services platforms, as shown in Figure 3:

  • the average hourly earnings in 2022, excluding waiting time, reveal two groups of ride-hailing services platforms : the first group includes Allocab, LeCab and Marcel, for which earnings are above 60 euros per hour, the second group includes the five remaining platforms (Bolt, Caocao, Free Now, Heetch, Uber) for which earnings are below 60 euros per hour;
  • with the exception of Allocab, this segmentation into two groups no longer holds when waiting time is taken into account: all platforms have average hourly earnings of less than 50 euros. For Allocab, the reported value for waiting time is 30 seconds, a value that seems to be an error.

For LeCab and Marcel, the difference between average hourly earnings per service with and without waiting time is 27 euros and 32 euros respectively, while for the other platforms it varies between 8 euros (Uber) and 19 euros (Free Now).

Figure 3 : Average hourly earnings per service for ride-hailing services platforms, with and without waiting time. Reading : In 2022, average hourly earnings (in euros) of a worker at Le Cab was 70 euros per hour when we exclude the waiting time before receiving a service proposal, and 40 euros per hour when we include the waiting time.

Ride-hailing services platforms: difference in hourly earnings for different brackets of number of services per week (Figure 4)

Figure 4 shows the difference of average hourly earnings for ride-hailing services platforms, excluding the waiting time, as a function of workers’ activity, for different brackets of number of services (rides) performed over a week (1 to 10, 11 to 25, 26 to 40 and over 40 rides). This difference in hourly earnings is very small, except for Allocab, Caocao and Marcel. For Allocab and Caocao, hourly earnings of workers performing more than 40 services per week is approximately 30% higher than hourly earnings of those performing between 26 and 40 services per week. For Marcel, we observe that hourly earnings in the 1 to 10 services per week bracket are almost 100% higher than hourly earnings in the 40 plus services per week bracket.

It is impossible to explain with certainty the origin of these disparities based on the available data. Indeed, since the values reported are averages, it could be a volume effect. For example, for Marcel, if the population of workers performing between 1 and 10 services per week is very small and has obtained high earnings for external and non-representative reasons, the discrepancy observed may not be representative of a global phenomenon. Another explanation could be a policy for attracting new workers by paying higher fees for the first services performed. As for the other platforms, the difference in hourly earnings between the different brackets of number of services is 10% at most.

Figure 4 : Hourly earnings for different brackets of number of services per week, excluding waiting time, for ride-hailing services platforms. Reading : In 2022, for the Marcel platform, the average hourly earnings (in euros) of a worker having performed, respectively, between 1 and 10 services (more than 40 services) per week, is 175 euros per hour (86 euros per hour). It is important to note that the size of the population within each category is not indicated.

Ride-hailing services platforms: characterization of hourly earnings by activity period (Figure 5)

Figure 5 shows the difference in average hourly earnings, excluding waiting time, between the different periods of activity (day, night, weekend, week, all periods) for different ride-hailing services platforms. Three groups of platforms stand out with regard to the difference in earnings between night and day periods :

  • for LeCab and Marcel, workers have average hourly earnings for the “night” period, respectively 33% and 31% higher than the average hourly earnings for the “day” period;
  • for Allocab, the difference of average hourly earnings between the “day” and “night” periods is 6%;
  • for the rest of the platforms, the difference varies between 16% and 24%.

It should be noted that the weekly and monthly indicators (calculated as a sum rather than an average per service), could allow more detailed analyses of the distribution of working time between day and night. However, differences of interpretation in the definition of these values prevent a reliable analysis of the resulting data (a discussion about the differences of interpretation can be found in the appendix).

Figure 5 : Hourly earnings for different periods of activity for ride-hailing services platforms in 2022, excluding waiting time. Reading : In 2022, average hourly earnings (in euros) of a worker at Marcel is respectively 90 (70) euros per hour when the activity period considered is night (day) from 10pm to 6am (6am to 10pm).

Delivery platforms: difference in hourly earnings between 2021 (Q4) and 2022 (Figure 6)

Figure 6 illustrates, for delivery platforms, the differences in hourly earnings (per service) between 2021 (Q4) and 2022. With the exception of Uber Eats, hourly earnings per service seem to have increased between 2021 (Q4) and 2022. This finding should be read with caution for three reasons: firstly, waiting time is not taken into account in the considered statistical indicator; secondly, 2021 is not a complete year, as the data was made available by platforms correspond only to the last quarter of 2021; and finally, seasonal effects may have played a role, such as the gradual end of Covid-19-related lockdowns in 2021, inflation in 2022, etc.

In the case of Uber Eats, in addition to a voluntary reduction in the earnings of workers per service, changes in calculation modalities could also explain the drop in hourly earnings: for example, if the platform began to take approach time into account when computing earnings in 2022 without having done so in 2021, this would add this time to the working time indicator, which would induce an “artificial” drop in hourly earnings (the data not being comparable between 2021 and 2022). These observations will be interesting to analyze over longer timeframes as the years go by.

Figure 6 : Average hourly earnings per service for delivery platforms in 2021 (Q4) and 2022, excluding waiting time before receiving a service proposal. Reading : Average hourly earnings at Uber Eats were respectively between 27 and 28 (23 and 24) euros per hour in the last quarter of 2021 (over the four quarters of 2022).

Can we take these initial findings further?

The analyses produced here are still very preliminary, but it is important to emphasize the specificity of these data which, despite certain shortcomings, constitute one of the rare sources of quantified information on the activity of workers within this sector. In addition to this first set of findings, the exploration of this data can still be taken further.

These indicators could be used to support workers, both in terms of quantifying their claims and in terms of gaining greater visibility of the differences in earnings between digital labour platforms, without having to test them all. These indicators also represent an opportunity for public authorities to gain a better understanding of this sector of activity and design appropriate public policies. By taking the appropriate precautions in terms of anonymization, these new sources of data can be put into perspective by coupling them with other sources (surveys, the right of access to personal data collected by platforms, etc.) and could thus provide rich analyses that take into account the various relevant contexts.

Concerning the format of the data, it would also seem worthwhile to continue collecting and further enriching these data sources with respect to various aspects:

  • the collection of quarterly data would be interesting in order to spot seasonal effects, or the impact of cyclical events on the activity of this sector;
  • waiting time between service proposals could be provided for monthly and weekly modalities, similarly to earnings and working hours. Moreover, the addition of waiting time between services actually provided would greatly enrich these indicators..
  • A standardization of platforms’ reporting practices across the various indicator parameters would also be desirable to reinforce the reliability of the analyses produced. In particular, this concerns :
    • the specific ranges detailed in the appendix;
    • the inclusion of bonuses and premiums, which in some cases do not comply with application of decree no. 2021-501 of April 21, 2021;
    • the inclusion of approach time in the working time, which, according to the same decree, depends on the way in which the price of the service is calculated, and can therefore vary from one platform to the platform.
  • The specification of an expected data layout and data upload on data.gouv.fr by the platforms could also simplify the data collection process, interpretation and storage. (example: specification of the exchange file relating to the share of low-emission vehicles in the fleets of vehicle rental, hire-purchase and leasing companies, delivery platforms, cab and ride-hailing services reservation centers)
  • information on the number of workers represented by each indicator, via aggregate numbers or, where appropriate, quantiles, would also significantly enrich the analyses and enhance their reliability.*

Appendix - Details about the data

The reporting year is the calendar year in which the indicators have been measured: 2021 (Q4) and 2022 at the date of this publication. Nevertheless, we feel it is important to highlight a few specific features. Firstly, the indicators for 2021 do not take into account the full year, but the year from September onwards. Secondly, 2021 is still a year strongly marked by COVID-19. Comparisons between 2021 (Q4) and 2022 should therefore be made with great caution.

The three statistical indicators reported by platforms are: earnings, working time and waiting time.

  • Earnings is "the price actually paid by the platform to the worker in return for his or her service, minus any commission fees that the platform charges". Bonuses are normally included, but not tips. According to the information provided on platforms' websites where these statistical indicators are provided, it appears that some of them do not comply or do not fully comply with application decree n°2021-501 of April 21, 2021.
  • The working time is the time period "between taking in charge the person or the goods to be transported in the vehicle, and handing over the goods to their recipient, or the departure of the person transported from the vehicle for good". The decree indicates that the time between the acceptance of the service and the pick-up of the person or goods must only be taken into account if it is included in the calculation of the price of the service. Therefore, as platforms may have different rules regarding this option while still being compliant, appropriate comparisons of their data may not always be possible.
  • Waiting time is counted between two service proposals, not between two services actually accepted. In view of the times displayed, which sometimes appear abnormally high for certain platforms, it is not certain that all platforms comply with this definition. As a result, waiting times are not always comparable.

Measurement can be carried out in three ways: as an average over the year for one service (or, for 2021, over four months), as the total sum of services per week, or as the total sum of services per month. It should be noted, however, that waiting time is required only for annual averages per service, and is not broken down on a weekly or monthly basis, unlike earnings or working time.

For weekly or monthly periods, a breakdown is used by brackets of number of services over the period. Thus, weekly statistical indicators are required separately for workers performing between 1 and 10 services, between 11 and 25, between 25 and 40 or more than 40 services. Similarly, monthly statistical indicators are broken down according to 3 to 30 monthly services, 31 to 75, 76 to 120, and more than 120 services per month. However, no quantile or information is provided on the share of workers belonging to these different brackets of services.

Finally, all these indicators are broken down according to specific time periods: weekdays (Monday to Friday), weekends, day (between 6am and 10pm), night (between 10pm and 6am), and all these time periods combined. While these indicators are relatively clear when using the service as a time period (for example "a service performed at night (between 10pm and 6am) earns…"), aggregating them over the week or month produces indicators that are more complex to interpret, a fortiori with regard to the number of services per period. For example, weekly night-time earnings for workers performing between 1 and 10 services per week can be interpreted in at least three different ways:

  • earnings from services performed at night (between 10pm and 6am) for workers performing between 1 and 10 services per week in total;
  • Total earnings (for both day and night) for workers performing between 1 and 10 services per week only during night-time (between 10pm and 6am) ;
  • earnings from night services for workers performing between 1 and 10 night services per week.

Not knowing which interpretation the platforms have adopted, or even whether this is constant between platforms or between years for the same platform, the interpretation of these indicators should be read with caution.


Find out more…

Legal references regarding the publication modalities of the statistical indicators by digital labour platforms:

ARPE refrences regarding the statistical indicators published by digital labour platforms in 2022 and 2023:

The code developed by PEReN to process the indicators and generate the figures in the present article