The Passenger Charter-party Carriers’ Act, enacted in 1961, authorizes the CPUC to regulate “[t]he use of public highways for the transportation of passengers for compensation ... and to promote carrier and public safety through its safety enforcement regulations.”1 The CPUC requires TNCs submit a number of annual reports relevant to passenger and public safety:
- Accidents & Incidents documents vehicle collisions
- Assaults & Harassments documents reports of assault and harassment
- Law Enforcement Citations documents citations issued by law enforcement officers
- Zero Tolerance documents reports of driving under the influence
This section presents an analysis of public safety incidents from September 2019 to August 2020 from the 2020 public TNC Annual Reports. It includes incident totals, rates per square mile, and rates per 100,000 trips. Areal (per square mile) rates are useful for understanding incidents that may impact the general public. Trip-based rates are useful for understanding risks to TNC users. VMT-based rates (which are preferable over trip-based rates) are useful in assessing risks to passengers and to the general public relative to the total amount of driving, but cannot be included because Lyft’s reports are redacted to remove VMT information.
How many TNC public safety incidents were reported?
Figure 13 shows the number of incidents reported by each company within the categories of collisions, assaults and harassments, DUI complaints, and citations. Uber reported 30,000 public safety incidents, while Lyft reported almost 45,000 public safety incidents. There were nearly 27,000 collisions, approximately 14,800 reported by Uber and 11,200 reported by Lyft. In addition, over 20,000 assaults and harassments (almost all of them reported by Lyft), 15,000 DUI complaints, and 14,000 citations were also reported.
Figure 14 shows the rates of incidents per 100,000 trips. Lyft reported total public safety incidents rates that were more than 3 times higher than Uber. Lyft’s collisions rates were twice Uber’s. Lyft’s assaults and harassment rates were more than 30 times Uber’s, Lyft’s DUI complaints were over 2.5 times Uber’s, and Lyft’s citations were twice Uber’s. These figures suggest that the companies may be reporting public safety incidents differently, pointing to the need for increased review by regulators.
Where did public safety incidents occur?
Assaults and harassments, collisions, and DUI complaints happened everywhere that TNC trips happened. Figure 15 shows incident totals and rates per 100,000 trips by category for the top 10 counties by number of trips. Incident rates vary by county and by company. San Francisco, which has the highest density of trips, has among the lowest public safety incident rates. By contrast, Sacramento, which has a low trip density, has the highest rate of public safety incidents. This may be partly explained by trip lengths, as denser counties with shorter average trip lengths may be less likely to be involved in a public safety incident on any particular trip. However, Lyft’s incomplete reports prevent an analysis of the relationship between public safety events and trip lengths.
Lyft’s public safety incident rates were much higher than Uber’s in each of the top 10 counties with the most TNC trips. The percent difference in incident rates between the companies was closest in Los Angeles County, where Lyft’s rate is 122% higher than Uber's, and furthest in Santa Clara County, where Lyft’s rate is 268% higher than Uber’s.
Assaults and harassments, collisions, and DUI complaints, like trip requests were reported by zip code. Citations, per the Law Enforcement Citations template were reported with a citation location. However, the CPUC did not provide explicit requirements for how the location should be reported and as a consequence each company reported it differently.
Table 21 shows the total number of citations by location. It is not clear why nearly all the reported citations were at airports. Uber reported 7,711 citations, all at airports. Lyft reported 6,259 citations, 6,038 (96%) of which were at airports, while 214 were in cities, and the remaining 7 were in other locations like an unincorporated neighborhood.
|Citations at Airports||7,711||6,038||13,749|
|Citations in Cities||0||214||214|
|Citations at Other Locations||0||7||7|
CPUC has not provided guidance to report citations solely at airports. The almost complete absence of TNC citations in locations other than airports suggests inconsistent or incomplete reporting and prevents the CPUC from assessing a key indicator of public safety and compliance with laws and regulations.
What types of public safety incidents occurred?
It is difficult to provide a clear depiction of the types of public safety incidents because the CPUC has not standardized reporting requirements for collisions, assaults and harassments, DUI complaints, or law enforcement citations. The lone exception, Decision 22-06-029, issued on June 24, 2022, adopted taxonomies for sexual assault and sexual harassment. These taxonomies only apply to a subset of the events reported in the Assaults and Harassments reports and had not been adopted when the 2020 Annual Reports were filed. In any event, the type of assault and/or harassment has been removed from the 2020 Public Annual Reports entirely.
In the absence of clear and consistent requirements, each company decides themselves how they report public safety data. As a result, this report is limited to only summarizing the overall number of incidents and cannot provide a more detailed analysis of types of public safety incidents. Similarly, it is difficult to develop public policy or exercise any meaningful or consistent regulatory oversight with respect to these public safety concerns.
Table 22 shows how each company reports the types of public safety incident for collisions, DUI complaints, and citations. Note that the public version of Assaults & Harassments does not contain any incident descriptions or categorizations. The table that each company uses reflects a different taxonomy to categorize the type of collision. Uber uses 7 response codes briefly describing the collision type. Lyft uses 5 response codes that describe, not the type of collision, but a qualitative description of the extent of damage. For DUI complaints, Uber uses 8 response codes, all describing allegations against the driver. Lyft used 4 response codes, 2 for describing allegations against the driver and 2 describing allegations against the passenger. One of the response codes is qualified as a first occurrence, “alleged_marijuana_ smell__first_instance”, but no other response codes for further occurrences. The other codes Lyft uses are not qualified in this way. For citations, Uber used 657 unique response codes and Lyft used 347 unique response codes.
|Citations||CitationReason||Unique incident description||Unique incident description|
Table 23 shows the consequences to the driver resulting from public safety incidents. As with incident classifications, the CPUC has in most cases not provided clear guidance for how to report consequences to the driver, leaving companies to determine themselves how to report driver consequences.2 Some classification of consequences to the driver is reported for assaults and harassments, DUI complaints, and citations, but not for collisions. Additionally, a binary indicator of whether the involved driver is currently authorized to drive is available for assaults and harassments and DUI complaints.
|Assaults & Harassments||DriverConsequence|
|Assaults & Harassments||DriverCurrentAuth|
|Collisions||IncidentAccidentGuiltyParty||not reported||not reported|
|Collisions||Liability||not reported||not reported|
How many drivers were suspended or deactivated?
While suspending a driver can adversely affect drivers’ livelihood by cutting off an income stream, suspending a driver is one of the actions a TNC company can take to protect its customers. Though each company used their own taxonomy for reporting driver consequences, both identified whether a driver was temporarily suspended or permanently deactivated. Table 24 shows the consequences to drivers resulting from assaults and harassments.
For this analysis, temporary suspensions are those that Uber classified as “Waitlisted” and Lyft classified as “Driver was temporarily suspended”, and permanent deactivations are those that Uber classified as “Deactivated” and Lyft classified as “Driver was permanently deactivated”. The table shows that 76% of Uber’s reported assaults and harassment resulted in a temporary suspension, and 24% resulted in a permanent deactivation, while 3% of Lyft’s reported assaults and incidents of harassment resulted in a temporary suspension, 2% resulted in a permanent deactivation, and 95% were neither temporarily suspended nor deactivated. The data suggests that Uber more aggressively suspends or deactivated drivers than Lyft does. It also suggests that the companies use different standards for reporting assaults and harassments.
|Not temporarily suspended or permanently deactivated||0||17,299||17,299|
|Percent temporarily suspended||76%||3%||9%|
|Percent permanently deactivated||24%||2%||3%|
|Percent neither temporarily suspended nor decativated||0%||95%||88%|
The CPUC requires that “[p]romptly after a zero-tolerance complaint is filed, the TNC shall suspend the driver for further investigation.”3 As with assaults and harassments, driver consequences of DUI complaints are reported with different taxonomies by each company, but each identifies temporary suspensions and permanent deactivations. Table 25 shows the driver consequences resulting from DUI complaints for each company. In this analysis, permanent deactivations are those Uber classified as “Driver Deactivated — Confirmed Allegation”, “Driver Deactivated — Third Unconfirmed Allegation”, and “Driver Previously Deactivated”, and Lyft classified as “Driver was permanently deactivated”. Temporary suspensions are those Uber classified as “Driver Reactivated — Unconfirmed Allegation” and Lyft classified as “Driver was temporarily suspended”. The remaining record records are those which Lyft classified as “Driver provided with warning and/or education,” which implies neither a temporary suspension nor permanent deactivation. The table suggests that Lyft frequently fails to comply with the CPUC’s requirement to suspend drivers following DUI complaints, only suspending or deactivating drivers in 6% of cases. By contrast, 94% of DUI complaints against Uber drivers resulted in a temporary suspension, and 6% resulted in a permanent deactivation.
|Not temporarily suspended or permanently deactivated||0||7,240||7,240|
|Percent temporarily suspended||94%||6%||49%|
|Percent permanently deactivated||6%||< 1%||3%|
|Percent neither temporarily suspended nor decativated||0%||93%||48%|
Driver suspensions are also reported in the Suspended Drivers report. These suspensions, unlike the ones reported above, are not linked to a specific type of incident. Figure 16 shows the total driver suspensions for each company. Lyft suspended nearly 5 times the number of drivers as Uber. Lyft also permanently suspended 50% more drivers than Uber.
Figure 17 shows driver suspension rates by company. Lyft suspended drivers at more than 11 times the rate of Uber and permanently suspended drivers at 15 times the rate of Uber.
These two figures reveal either that each company has significantly different approaches to driver suspensions or that they report driver suspensions differently.
For each driver suspension, the companies report whether drivers were permanently deactivated, and whether they have been reactivated. Driver suspensions by suspension type and reactivation status for Uber are shown in Table 26 and for Lyft in Table 27. Presumably, a driver that is permanently deactivated cannot be reactivated. As expected, none of Lyft’s permanently suspended drivers are reported to be reactivated. But Uber data shows that 1,250 (30%) of the 4,162 drivers classified as permanently suspended are also classified as reactivated. It is unclear whether these drivers are permanently deactivated or not. If they were reactivated, it is not clear why their permanent suspension was overturned, or the potential impacts to the safety of passengers and the general public.
|Not Permanently Suspended||110||9,505||9,615|
|Not Permanently Suspended||9,974||49,322||59,296|
California Public Utilities Code § 5352(a). https://leginfo.legislature.ca.gov/faces/codes_displayText. xhtml?lawCode=PUC&division=2.&title=&part=&chapter=8.&article=1. ↩︎
To describe the driver consequences of assaults and harassments, Uber uses 2 response codes and Lyft uses 3 response codes. For DUI complaints, a description of the resolution and a driver consequence are reported. Uber uses 4 response codes to describe the DUI complaint, and the same 4 response codes to classify the driver consequence. Lyft used 3 response codes to describe the DUI complaint resolution and 3 different response codes to describe the driver consequences. The only consequence reported for citations is the payor of the citation. Lyft’s responses include both “LYFT” and “DRIVER”, while Uber’s only include “Uber”. ↩︎
D. 13-09-045, p. 27. CPUC Rulemaking R12-12-011. 9/19/2013. ↩︎