Characteristics of the ride-hailing driving job
The survey questions shed light on many aspects that expose the characteristics of the ride-hailing driving job in Chile. How does the tension between flexibility and the informal conditions that have been discussed in sharing economy jobs emerge in the drivers’ daily routines? How comfortable, satisfactory, and secure is it to drive for ride-hailing companies? First, we asked questions about drivers’ working routines, displayed in Fig. 1. We analyse the length of drivers’ working periods per week and per day, which is a relevant indicator of the quality of this job. Figure 1a shows the number of driving hours per week and its heterogeneity confirms that there are different ways of being a driver. Of the drivers surveyed, 41% work less than 30 h a week, and 20% work more than 50 h a week (as a reference, 44 is the standard number of hours worked per week in Chile for full-time jobs). Working hours depend on car ownership status, as the average, weekly ride-hailing drive time is 31 h for car owners, 33 h for drivers that borrow a car for free, and 45 h for drivers that rent a car. This information is relevant as an informal market of car rentalsFootnote 10 for ride-hailing and informal loans for car purchasesFootnote 11 have emerged.
Figure 1b shows the number of hours driven during the latest day worked, prior to responding to the survey. Because the answer in Fig. 1b can be obtained directly from the app (drivers get a daily summary when they finish a workday), this information is highly reliable as it is not dependant on perception. Besides confirming the same heterogeneity announced before, a concerning fact shows up in Fig. 1b: the extremely long working periods that some drivers endure. Almost one-third of the drivers were on route 10 or more hours during the latest day, including 4.7% of drivers who said they typically worked 16 h or more. This is way beyond a regular full day of work in Chile, and it is also a very risky situation as tired drivers are more prone to cause crashes or near misses (Fell and Black 1997; Phillips and Sagberg 2013). In the European Union, nine hours is the maximum time that drivers can work daily and 56 h is the weekly maximum (Department for Transport 2015)
Figure 1c shows the total number of hours worked per week, which includes time spent working as a ride-hailing driver and time spent at other jobs. Again, results are concerning as several drivers are working much more than the standard work time per week in Chile (44 h), although the average (46.2 h) is just slightly larger than the standard work time for a full-time employee. In our sample, 42% of respondents report working more than 50 h per week. Some respondents work as ride-hailing drivers to complement a non-work-related activity: 21% of them work less than 30 h a week, which is in line with the fact that 18% of drivers are also students
Figure 1d describes the days and periods in which respondents work as ride-hailing drivers. On regular working days (Monday through Thursday), they work mostly from 7 a.m. to midnight—the most preferred driving period is from 5 PM to midnight. This pattern changes on Fridays and Saturdays when most of the work is during the night, specifically from 8 p.m. to 6 a.m. On these two days, peak time in ride-hailing driving is 8 p.m. to midnight. This finding corresponds with actual data on Uber use in Santiago as published by local newspaper El Mercurio, which shows that the weekly peak of Uber trips in the city occurs on Fridays and Saturdays between 8 p.m. and 2 a.m. (Tirachini and del Río 2019).Footnote 12 Finally, Fig. 1d shows that on Sundays, drivers work mostly during the day and in lower numbers
To understand the relevance of having flexible working conditions, we asked drivers about their work routines. Results are shown in Fig. 1e, with each of the four categories presented as an alternative. Considering current drivers of all companies, 74% of respondents follow fixed or stable weekly routines, although 44% incorporate some flexibility in order to obtain a target income per week. The rest of the drivers (25%) work mainly in their free time, earning variable incomes (14%) or adjusting their routines to reach a target weekly income (11%). These figures should be complemented with the fact that, when asked for the reasons to work as drivers, flexibility was by far the most selected reason (as shown in “Satisfaction with the ride-hailing job” section). These numbers reveal that flexibility is indeed something that drivers value, but that the work routines are not so different from traditional jobs, and many drivers do have a weekly income target to which their work time is adjusted. Regarding job expectations, 55% of the drivers in our sample said they plan to work in this role indefinitely, whereas the rest have some expectation to switch jobs when, for example, they find a job related to their professional training.
Table 2 synthesizes drivers’ perceptions regarding the distribution of their time while working in ride-hailing. Waiting for a request can be done while either parked or cruising around, and drivers have different strategies about what to do while waiting for a request in order to maximize their income, depending on personal preferences and location. On average, drivers report spending 53% of their work time travelling with passengers, 18% driving while waiting for a new request, 15% driving on their way to pick up a new passenger, and 14% parked. Time spent driving with passengers coincides with the calculations by Cramer and Krueger (2016), who estimate that the time with passengers for Uber drivers in Boston, Los Angeles, New York, San Francisco, and Seattle was between 43 and 55% of their total driving time. The analysis of the efficiency in the use of time by drivers is relevant from a traffic sustainability viewpoint, as the time in which ride-hailing vehicles drive without passengers (on average, 33% as shown in our sample) has been analysed as a significant contributor to traffic and congestion in cities (Henao and Marshall 2019b; Tirachini and Gomez-Lobo 2020)
Table 2 Distribution of time while driving (n=272) Figure 2 deals with risky situations that drivers face when working in ride-hailing. Although some of these situations can occur in any job, analysing them in a sharing economy scheme is relevant because workers are less protected (or not protected at all if their jobs are not legally regulated). Taking this into account, the results are disturbing: less than 3% of the drivers report having experienced no risky situation. The most common dangerous situation is driving in places perceived as insecure. The lack of regulation has enhanced the tension between taxi and ride-hailing drivers, which is also dangerous, with two out of five drivers in our sample report having been victims of attacks or threats from taxi drivers. The rest of the risky situations (e.g., assault, harassment from passengers, and crashes) are not so common, but the figures are still high.Footnote 13 We cannot rule out having a selection bias in our sample: drivers that have faced risks might be more likely to respond to this type of survey if they want their situation to be known. But even assuming the possibility of such bias, these figures are very concerning and reveal the need for increased job security. It is worth mentioning that being controlled by the police was reported by several of the open answers classified in “Other”, but it was not an explicit alternative, so it is likely to be another relevant risky situation given that ride-hailing remains an illegal activity in Chile. This situation that can lead to temporary car retention should be solved with proper regulation of the ride-hailing activity. Currently, ride-hailing companies pay the fine needed to get the car back but do not compensate for the days in which the driver was unable to work
Estimation of driver income and expenses
One of the most relevant aspects discussed in the ride-hailing literature and media involves driver earnings (Henao 2017; Wells et al. 2018). In this section, we estimate income and expense focusing only on the largest group in our sample: Uber drivers in Santiago.
Estimation of Uber trip fares and driver wage (by the authors)
In order to estimate how much money drivers earn by working in ride-hailing, we performed 160 “virtual trips” in Uber. This sample size is chosen because it provides a narrow width for the 95% confidence interval (C.I.) of the mean income value (per kilometre or per minute), as seen below (the C.I. width is between 5% and 8% of the mean income value). A virtual trip consists of defining an origin and destination at some moment of the day, and registering the fare for the users, the distance, and the time needed to complete that trip at that moment using two online tools: Uber’s fare estimatorFootnote 14 and Google Maps. Origins, destinations, time, and day were chosen to replicate as well as possible the ride-hailing usage information collected by the CNP (2019) and used by Tirachini and del Río (2019), which provided a distribution of ride-hailing trips in Santiago according to length, zone of origin, zone of destination, time of day, and day of the week. From this database, approximately 64% of riders use ride-hailing on Friday or Saturday and 59% travel between 5:01 pm and 11:59 pm. Trip lengths (estimated using Google Maps) were chosen to replicate the trip length distribution of 1,474 surveyed Uber trips in Santiago as presented in Tirachini and Gomez-Lobo (2020): 18% of trips are shorter than 3 km, 40% of trips are between 3 and 6 km, 26% of trips are between 6 and 10 km, and 16% of trips are longer than 10 km. For all 160 virtual trips, the fare \(P\) per trip was calculated in two ways, in both cases using Uber’s fare estimator. None of the methods include surge pricing.
-
Method 1 This method uses Uber’s fare estimator exclusively. Once an origin and destination for a trip are provided, Uber’s fare estimator does not provide a unique fare estimation, but rather it provides a range \(\left[{P}_{min},{P}_{max}\right]\) for the final fare. We then compute the fare as the average value of that range, \(P=\left({P}_{max}-{P}_{min}\right)/2\).
-
Method 2 We use Uber’s formula for fare calculation, which is also provided on Uber’s website:
$$P=\text{max} \left\{{P}_{base}, {a}_{0}+{a}_{1}T+{a}_{2}D\right\}$$
(1)
where \({P}_{base}\) is the minimum fare as set by Uber, T is the duration of the trip in minutes, D is the length of the trip in kilometres, and \({a}_{0}\), \({a}_{1}\), and \({a}_{2}\) are parameters. At the time of the survey, parameters were \({P}_{base}\) = 1200 CLP, \({a}_{0}\) = 450 CLP, \({a}_{1}\) = 80 CLP/min, and \({a}_{2}\) = 220 CLP/km (CLP: Chilean Peso), which are equivalent to $ (USD) 1.93, 0.72, 0.13, and 0.35, respectively. The $ symbol is used for US Dollar (USD) in the paper.Footnote 15 In order to use (1), distance and travel time per trip are estimated using Google Maps simultaneously to the request of price range \(\left[{P}_{min},{P}_{max}\right]\) from Uber’s fare estimator in Method 1.
Results indicate that average fares per minute are ($ cent) 38.8 with Method 1 (95% Confidence Interval C.I. 37.6, 40.0) and 34.9 with Method 2 (95% C.I. 34.0, 35.9). Figure 3 presents the fare for all 160 trips, in $ cents per kilometre, as calculated with both methods; trips are sorted from largest to smallest fare per kilometre, using Method (1) Average fares are 86.6 (95% C.I. 83.3,89.9) and 78.2 (95% C.I. 75.3, 81.2) for Methods 1 and 2, respectively. Thus, we find that there is a statistically significant difference in the mean fare estimated with both methods; on average, Method 1 estimates a fare that is 10% larger than Method (2) Moreover, in 91% of cases, the fare from Method 1 is larger than the fare from Method 2 (Fig. 3). Longer trips have a lower price per kilometre (Eq. 1). A number of reasons might explain the discrepancies observed in Fig. 3. First, for Method 1, the average value between \({P}_{min}\) and \({P}_{max}\) was chosen, however, we do not know how Uber estimates the range \(\left[{P}_{min},{P}_{max}\right]\) and, in particular, how close to reality the average value is within that range. Second, the optimal route suggested by Google Maps and the Uber app may differ and, in reality, we do not know which route drivers actually follow. Given these uncertainties, we will provide estimations of earnings using these two alternative methods.
Estimating the drivers’ hourly income requires assuming a value for the drive time with passengers. We assume that drivers travel with riders on average 53% of their working time (Table 2) and that the commission discounted by Uber is 25% (as reported by drivers). With these assumptions, we obtain an average hourly income of $9.30 with Method 1 and $8.30 with Method 2.
Estimation of driver costs (by the authors)
First, we estimate average gasoline consumption, which depends on the consumption efficiency of cars. In a previous cost study for taxis and ride-hailing in Santiago, Bennett and Zahler (2018) use an empirical U-shape function that relates gasoline consumption to average car speed based on research from the Department for Transport from the United Kingdom. Computing Bennet and Zahler’s curve with the average speed of our sample of Uber trips as estimated with Google Maps (26 km/h), we obtain an average fuel efficiency of 12.5 km/litre, which we assume as representative of the fleet used by Uber drivers in Santiago.Footnote 16 Gasoline price during the period observed is $1.2 per litre; therefore, we estimate fuel cost to be $2.4 per hour for an average hour of driving. For the estimation of other costs associated with maintenance (e.g., periodic car controls, change of parts), we rely on Bennett and Zahler (2018), who estimate these costs to be 3 cents per kilometre, equivalent to 81 cents per hour, which is 34% of average fuel cost. Therefore, we estimate the total driving cost to be $3.2 per hour. This calculation assumes that drivers are always driving, however, it is common that drivers also spend time parked while waiting for a new request. Average parking time, as reported by drivers, is 14% (Table 2). Using this value, we obtain the adjusted cost to be $2.80 per hour.
All these estimations assume that drivers own their vehicles, which is the case for approximately 80% of the drivers, according to our survey and the CNP (2019). For drivers who want to work in ride-hailing and do not own a car, there is already an informal market in Chile in which drivers rent cars on a weekly basis. In this case, drivers are only responsible for fuel costs because the cost of maintenance (81 cents per hour in our estimation) is assumed by the owner. Renting a car in this informal market costs between $130 and $190 USD per week (based on a Google search of online ads), depending on the car type.
Driver wage estimation
In this section, we present an estimation of income and cost as reported by the drivers we surveyed. Three aspects of the survey have been analysed so far: (1) how much money drivers make monthly (e.g., how much money they receive from the ride-hailing company, after discounting the commission charged by the company but without considering their spending in terms of costs), (2) how many hours drivers work weekly, and (3) the percentage of total income drivers spend on fuel, maintenance, and all other relevant expenses. Although these figures should provide a direct estimation of the hourly wage of the drivers, some basic statistical processing was needed. On the one hand, answers were given in intervals: intermediate values were assumed but in the extreme cases (intervals containing zero or infinity) in which the positive finite value was assumed. On the other hand, the “large extreme” interval for the number of hours worked in a week (60 h or more) had almost 10% of the answers, which is why we subdivided it into two categories: between 60 and 70 h, and 70 h or more. Interpolating with a normal distribution from the previous time brackets, we estimate 17 answers to be in the “60 to 70” category and nine answers in the “70 or more” one. Under these assumptions, we estimate an average income of $5.90 per hour and average costs of $1.70 per hour.
Uber’s estimation
Finally, we compare our figures with the income promise made by Uber. In 2016, Uber’s general manager in Chile publicly assured drivers that if they work between seven and 10 h per day, they can make 450,000 CLP (or $723 USD) per week.Footnote 17 Using this figure, and assuming one free day per week (therefore, a work time that’s somewhere between 42 and 60 h per week), we obtain income rates between $12.10 and $17.20 per week (discounting the fee kept by Uber but without discounting any operational cost).
Synthesis and comparison of the estimations
Table 2 synthesizes income and expenses according to the alternative estimation methods presented. For the Uber’s estimation column, our own estimation for the expenses is used. First, our estimations (Methods 1 and 2) produce estimate wages between $5.10 and $6.50 per hour, which is larger than the estimation based on the drivers’ own assessment: $4.20 per hour. At the same time, we observe that Uber’s promise is too far from the other estimations, as the estimation of wage using Uber’s promise is (on average) roughly double that of our estimations and more than double that what drivers said in the survey. In the United States, the Federal Trade Commission noted Uber’s exaggeration of drivers’ earnings as the dissemination of “advertisements that overestimated the likely hours and yearly income of drive” and announced a $20 million settlement with Uber for these actions (Calo and Rosenblat 2017). As stated earlier in this paper, in Denver, Henao and Marshall (2019a) estimate that drivers earn between $5.70 and $10.50 per hour—figures that are below the minimum wage in the State of Colorado in several cases. In the case of Santiago, all of the estimates show that drivers earned more than the minimum legal wage in Chile in 2018 (which is CLP 288,000 per month, or around $2.50 USD per hour). In Chile, median and mean monthly income during 2018 were CLP 400,000 (about $3.50 USD per hour) and CLP 574,000 (about $5 USD per hour). Interestingly, Uber drivers are likely to make a wage that is larger than the median wage in Chile.
It is also interesting to observe that, compared to our estimations, drivers tend to underestimate both their income and their expenses. Their estimation of expenses is similar to our estimation of the expenses in fuel only ($1.70 verses $2.10). This is in line with the findings of Ivehammar and Holmgren (2015), who show that car commuters might underestimate their private monetary cost by considering only those expenses that come “out-of-pocket.”
Finally, we highlight two elements that are missing in this discussion: (1) Uber’s surge pricing, which is applied in periods of low driving supply relative to trip requests (actual data of surge pricing application is presented by Castillo et al. 2018), and (2) car depreciation costs. In Table 3, surge pricing is included in the drivers´ estimation and Uber´s estimation of income, but it is not included in Methods 1 and 2. As depreciation increases, cost and surge pricing increases income. These two elements push in different directions and (at least partially) cancel out in the calculation of wages with Methods 1 and 2, however, the result of their sum is highly uncertain. On the one hand, dynamic surge pricing applied by ride-hailing companies depends heavily on the city and time period. For instance, data collected by Chen et al. (2015) show that the average surge multiplier is around 1.4 in San Francisco and around 1.1 in Manhattan, while in the database compiled by Henao and Marshall (2019a) in Denver, only 7.2% of ride-hailing trips were subjected to surge pricing, with multipliers between 1.25 and 2.0. On the other hand, depreciation is not included in the estimation of costs as there is no reliable data on depreciation costs per kilometre for cars in Chile, let alone for vehicles used for commercial purposes such as taxis and ride-hailing vehicles. Depreciation varies a lot from driver to driver depending on the value of the car and how many kilometres the car has driven already. In other countries, depreciation costs have been estimated to be substantial. For private cars in the United States, for example, average depreciation cost is estimated to be in the same order of the costs of fuel and maintenance combined on a per kilometre basis (AAA 2019). Such large depreciation cost would result in an overestimation of earnings if the application of surge pricing is not large enough to neutralise it.
Table 3 Hourly income, expenses, and wage of Uber drivers in Santiago Drivers’ satisfaction and opinions
Satisfaction with the ride-hailing job
Regarding questions about satisfaction with the job, drivers were first asked to cite specific reasons why they chose to work as a ride-hailing driver. Respondents had to express their assigned level of relevance based on a 1-to-7 ordinal scale, where 1 and 7 are the lowest and highest relevance levels, respectively (1 to 7 is the scale for grading in schools and other educational institutions in Chile, where 4 is the pass mark). Results for the job attributes included in the survey are depicted in Fig. 4a. Flexibility to choose work times stands out as the most important feature of ride-hailing driving, as 76% assign 6 or 7 points to this attribute. The average score for “flexibility” is 6.1 (85% as a rate of the maximum score). The second reason respondants chose to work as ride-hailing drivers is because they say they “enjoy driving”, which has 5.3 (72%) as the average mark. The third reason is that it offers “better conditions than alternative jobs”, with an average score of 5.0 (67%), and finally, “wage level” with an average score of 4.9 (65%).
Drivers were then asked about their level of satisfaction with specific elements of their job and with the ride-hailing driving job in general, using the same 1-to-7 scale (where 1 and 7 are the lowest and largest level of satisfaction, respectively). Results are shown in Fig. 4b. The average global evaluation is 4.5. The most common global evaluation is 5.0 (30% of the drivers), while 24% of the drivers evaluate with a very good mark (6 or 7), and 23% with a “fail” mark (3 or lower). Concerning specific attributes, transparency in the location of the passengers, the evaluation system by riders through stars given to drivers, and the efficiency in assigning passengers are all evaluated with 4.4 as average score (just above the pass mark 4.0). The wage level is evaluated with 3.8 and the transparency regarding wages with 3.4. Note that the global evaluation is better than each of the specific aspects, suggesting that there are other positive characteristics that were not included in the questionnaire. In total, two out of five respondents are not satisfied with the amount of money that they make as drivers (score 1 to 3), and 53% of drivers are not satisfied with the way their income is calculated by their ride-hailing app, as they do not feel it as “transparent”. The latter aspect is relevant as it deals with the core of the discussion regarding sharing economies: that drivers are partners of ride-hailing companies rather than employees. This stated lack of transparency regarding wages takes place in a situation in which drivers have no bargaining power.
Characterising drivers’ satisfaction with ordered models
In this section, the satisfaction scores from drivers are used to estimate the specific variables that are statistically significant in determining drivers’ satisfaction with their ride-hailing job. An understanding of job satisfaction relies on the relationship between workers’ well-being (Green 2010), labour productivity (Böckerman and Ilmakunnas 2012), and the propensity to quit (Freeman 1978; Clark 2001; Green 2010), among other variables. In the case of the transport sector, studies have shown that self-reported levels of driving stress and job insecurity can be positively correlated with traffic crashes from bus and taxi drivers (Useche et al. 2018) and that an imbalance between effort and reward (measured as a combination of salary level, job promotion opportunities, and being treated with respect) increases job strain and health problems of bus drivers (Chung and Wu 2013).
We estimate ordered models to explain ride-hailing job satisfaction. The ordered logit and ordered probit models are the two most common specifications used in the transport literature for analysing ordinal variables. Even though the parallel regressions assumption applies for both ordered logit and probit models (Greene and Hensher 2010), the ordered probit model does not require to meet the proportional odds assumption (Williams 2016). We specify two ordered probit models to explain the general level of satisfaction: one concerning different explanatory variables that describe the drivers (Table 4) and the other concerning the drivers´ evaluations of specific features of this job, as explained in “Satisfaction with the ride-hailing job” section (Table 5). In other words, the former model encompasses objective facts, while the latter model uses subjective opinions as explanatory variables, which is why we split them into two different models—to obtain neater interpretations.Footnote 18
Table 4 Ordered probit model that explains drivers’ general satisfaction with respect to specific drivers’ attributes
Table 5 Ordered probit model that explains drivers’ general satisfaction with respect to their evaluations of specific aspects of these jobs
Probit estimators are frequently used to model job satisfaction in the labour economics literature (e.g., Gazioglu and Tansel 2006; Green 2010). The general form of the model is presented in Eq. (2), where \(X\) is the global evaluation (job satisfaction) and \({\Phi }\)are the accumulative probability function for a standard normal distribution,\({z}_{k}\) are the explanatory variables, and \({\beta }_{k}\) the estimated parameters, shown in Tables 4 and 5.
$$P\left(X\le i\right)={\Phi }\left({\gamma }_{i}+\sum _{k}{\beta }_{k}{z}_{k}\right)$$
(2)
We estimate the models using mnrfit function of the Matlab software package.Footnote 19 Note that
$$\frac{\partial P(X\le i)}{\partial {\beta }_{k}}={z}_{k}{\Phi }^{\prime }\left(\gamma _{i}+\sum _{k}{\beta }_{k}{z}_{k}\right)\ge 0$$
(3)
where the inequality holds because all explanatory variables \({z}_{k}\) are positive and \({\Phi }\) is a strictly increasing function; therefore, a positive value for \({\beta }_{k}\) means that the larger the value of the explanatory variable \({z}_{k}\), the higher the probability of obtaining a low \(X\)(or, put plainly, that the correlation between \(X\) and \({\beta }_{k}\) is negative). In particular, using the well-known expression for the expected value of a random variable \(X\) that takes natural numbers
$$E\left(X\right)=\sum _{i\ge 1}P(X\ge i)$$
(4)
we can deduce that
$$\frac{\partial E\left(X\right)}{\partial {\beta }_{k}}<0$$
(5)
The intercepts \({\gamma }_{i}\) deal with the cumulative probabilities of each category \(i\) (Greene and Hensher 2010) and say nothing about specific explanatory variables \({z}_{k}\) .
The descriptive characteristics considered in the first of these models include both personal characteristics (such as gender and age) and attributes of the job that are driver-specific (such as vehicle ownership status and exposure to risk situations). For a significance level \(\alpha =0.05\), Table 4 shows that having experienced a risky situation (either high-risk or mid-risk) is statistically significant to explain job satisfaction; high-risk situations (defined as having been a victim of assaults or having car accidents with injuries) are more relevant as the mean global job satisfaction is 4.1 in the sub-sample of drivers that have faced them, while this value is 4.4 for drivers that have experienced mid-risk situations only. In parallel, those who drive in ride-hailing as a way to complement another part-time job show a statistically significant larger satisfaction than the rest (while having a full-time job is significant only for \(\alpha =0.1\)). This is an interesting finding, as the flexibility of driving seems to be more important and satisfying for drivers when it adjusts to another job. As also found in other job satisfaction studies (Gazioglu and Tansel 2006), the level of weekly earnings as a driver is significant in explaining job satisfaction. After controlling for income, job status, and risk situations, respondents that own the vehicles they drive have a lower satisfaction level than others (those who borrow a vehicle for free and those who pay a weekly rent for the vehicle).Footnote 20 All the other explanatory variables are not significant, including hourly wages, gender, age, and number of hours driven. Other studies on general job satisfaction across different sectors have found that women experience greater job satisfaction than men (Clark 1997; Gazioglu and Tansel 2006), a finding that is not replicated with our sample of ride-hailing drivers in Chile.
The second model (Table 5) attempts to estimate which specific aspects of the ride-hailing driving job are relevant to explain general job satisfaction. Satisfaction level with each of the attributes shown in Fig. 4b is used as the explanatory variable. By far, given its p-value, satisfaction with the wage level is the most relevant factor. Wage transparency and efficiency assigning passengers to users are also significant for \(\alpha =0.05\). The system of driver evaluation through stars from users and the transparency related to passenger locations are not significant.
Reasons to quit
The survey was completed by 36 former ride-hailing drivers (people who used to be ride-hailing drivers but quit this job), for whom there was a specific question on their reasons to quit. Respondents could select one or more alternative options, results are shown in Fig. 5. The most common reason to quit (more than 25% of choices) is that the money earned was not enough, which is consistent with the discussion from “Characterising drivers’ satisfaction with ordered models” section about the relevance of earnings for drivers’ satisfaction. Some drivers are not happy with their earnings and the satisfaction (or lack of it) with their wages is the most relevant characteristic that explains general satisfaction with this job. This is consistent with other studies that have found a lack of satisfaction with one’s pay and job security as having the strongest influence on the decision to quit a job (Clark 2001). Finding another job is the second most common reason to quit. This suggests that ride-hailing has the potential to be a temporary job alternative during unemployment situations. More than 15% of these drivers said insecurity (e.g., feeling scared of having problems with passengers or taxi drivers) was a reason they quit. The rest of the suggested possible explanations (including stress and problems with the platform or with the police) were selected by less than 10% of respondents.
Automated vehicles for ride-hailing
The analysis of ride-hailing and the job market is conducted in the context of rapid technological changes, as this market is likely to be subject to profound changes in the future brought on by automation and the development of driverless vehicles. The large cost savings that are anticipated due to automation has prompted ride-hailing companies such as Uber, Lyft, and Didi Chuxing to form alliances with car manufacturers for the deployment of automated ride-hailing services, such as the pilot started by Waymo (Google´s company for automated vehicles) and Lyft in Phoenix, Arizona, in May 2019.Footnote 21 In this context, we asked drivers a question on how they feel about automated vehicles, in particular, the use of driverless vehicles for ride-hailing. Results in Fig. 6 show that most drivers do not worry much about this situation. More than 80% of respondents either do not think that vehicle automation represents a risk to their jobs (either because they think it is too far in the distant future or because they think it is never going to happen in Chile) or because they think that having automated vehicles is good even if they lose their jobs. Only 10% of current drivers show some level of concern about the possibility of losing their jobs due to automation.
Regulation of ride-hailing services
Drivers were also asked about the tension between flexibility and precarity of these jobs as, at the time of the survey and of this writing, ride-hailing remains unregulated in Chile. We asked respondents to choose between a set of options dealing with how to regulate their jobs. Fifty-six percent said they would prefer for the ride-hailing system to be regulated (e.g., ensuring in a contract labour rights such as social security, vacations, and transparent salary, among others) even if that requires drivers to obtain a professional driving license and pay taxes (Fig. 7). Only 10% of the drivers are satisfied with the current unregulated situation. Therefore, even though drivers highly value the time flexibility of ride-hailing as a source of income, there seems to be a general dissatisfaction with the current situation and a clear choice for more regulation and security in their labour situation.