TNCs 2020: A Profile of Ride-Hailing in California

Data reporting compliance and integrity is a prerequisite for effective analysis to guide the development of public policy and enforce regulations. This section examines the 2020 public TNC Annual Reports for compliance with reporting requirements and data integrity (meaning that the data is logical and internally consistent).

Are TNCs submitting the required reports?

Both companies filed the required 2020 TNC Annual Reports. In February 2022, the Transportation Authority requested 2020 public TNC Annual Reports for Uber and Lyft from the CPUC. The CPUC treated the request as a Public Records Act (PRA) request and provided the reports later that month.

Are the reports complete?

CPUC Staff prepared the 2020 public TNC Annual Reports, including its redactions.¹ A report is considered complete if all of the fields designated as public are present and not redacted.² Table 2 shows the percent completeness of each report by each company, as measured by the percent of required public fields and records that are present and unredacted. Uber’s 2020 public TNC Annual Reports are complete, with the exception of one redacted field in the Accidents & Incidents report. Lyft’s 2020 Annual Reports are not complete.

CPUC staff prepared the 2020 public TNC Annual Reports from the original reports provided by the companies. It is not clear whether Lyft’s original reports, like the public versions, are substantially incomplete. Among the redacted data are trip date, time and location, VMT data, fares, and vehicle make, model and year. Both Uber and Lyft’s reports, in some cases, include required data fields but the data itself is blank, including trip occupancy.

Complete data is important to summarize and support evaluation of the industry’s activities:

• Date and time information can be used to evaluate whether trips are taking place during the most congested times of day or whether they are providing late night or weekend service when transit runs less frequently.
• Location information can be used to evaluate whether TNCs are driving in the busiest parts of cities or near regional transit hubs.
• VMT information, combined with time and location can be used to analyze how TNCs may be contributing to congestion.
• VMT information when paired with vehicle make, model, and year can be used to evaluate emissions.
• Trip occupancy can be used to evaluate the number of passengers transported per vehicle (a measure of efficiency) and TNC’s compliance with the CO2 per-passenger-mile requirements of the Clean Miles Standard.
• The missing data from Lyft’s reports prevents these analyses for Lyft and for the industry as a whole. See Appendix B: Report Completeness Inventory for detailed accounting of each report’s completeness.

A closer look at the data can reveal other issues. For example, Figure 4 shows the daily total number of completed trips from Uber’s Requests Accepted report, revealing that the first two weeks of March 2020 are missing. This two-week period does not correspond with local COVID Shelter-in-Place (SIP) orders, which went into effect the week following the missing data. It is unclear whether any other Uber reports are also missing data from these two weeks. The redactions and omissions in Lyft’s incomplete Requests Accepted report hides these kinds of gaps and irregularities, hampering analysis and hindering regulatory oversight.

The 2021 public TNC Annual Reports, available on the CPUC website since October 2022, are even more heavily redacted. Table 3 compares the overall completeness of Uber’s and Lyft’s 2020 and 2021 public TNC Annual Reports, as measured by the percent of required public fields and records that are present and unredacted. Lyft’s 2020 and 2021 reports were both heavily redacted, but while Uber’s 2020 reports were nearly complete, their 2021 reports were redacted similarly to Lyft’s. When the CPUC releases the 2021 public TNC Annual Reports with only properly reacted data, the Transportation Authority will produce a follow-up report documenting findings.

Is the data reported internally consistent?

Internal consistency means that the data in one part of a company’s reports does not contradict data in another part. Contradictory or internally inconsistent data prevents monitoring and evaluation, informed policy-making, and effective regulatory oversight. For a subset of metrics, the TNC Annual Reports contain multiple sources of information from different reports, and each company’s reports should produce consistent metrics across all the sources. This section evaluates the internal consistency of the following metrics reported or derived from the 2020 public TNC Annual Reports. These are the most basic descriptors of TNC activity.

• Trip requests
• Completed trips
• Incomplete trip requests
• Vehicle miles traveled (VMT)
• Driver days
• Driver hours

The Traffic Congestion Mitigation Tax is a tax on all ride-hail trips originating in San Francisco, which began collections in 2020. San Francisco’s revenues from the tax have been highly irregular. Redactions of fare data in the TNC Annual Reports prevent independent validation of tax revenues, and the inconsistencies in the 2020 Annual Reports documented in this report raise questions about whether the 2020 TNC Annual Report data would be sufficient for independent validation even if fare data weren’t redacted. However, consistent, unredacted data from the TNC Annual Reports would support independent validation of tax revenues.

TOTAL TRIP REQUESTS

The total number of trip requests is a measure of TNC demand. It can be calculated 3 ways using data found in 5 reports:

1. By adding the counts of the number of records in the Requests Accepted and Requests Not Accepted reports,
2. By adding the number of requests in the Aggregated Requests Accepted and Aggregated Requests Not Accepted report, and
3. By adding the total trip requests in the Accessibility Report (Confidential).³

Table 4 and Table 5 show total trip requests by source. In the 2020 public TNC Annual Reports, Uber’s reported trip requests are internally inconsistent, differing by nearly 20 million trips, or 12%. Lyft’s reported trip requests are also internally inconsistent, differing by almost 50 million, or 75%. Lyft’s internal inconsistencies are up to 13 times greater than Uber’s internal inconsistencies.

COMPLETED TRIPS

Completed trips are a measure of total travel and can be used to evaluate a company’s share of the TNC market and the TNC share of the total travel market. It is the most basic statistic describing TNC services provided. Completed trips are reported in the Requests Accepted report as a list where each record represents a completed trip, and in the Aggregated Requests Accepted report which contains annual completed trip totals for the reporting period by zip code.⁴

Table 6 and Table 7 show the number of completed trips reported by Uber and Lyft in each report. Uber’s reported completed trips are internally inconsistent, differing by 9.3 million, or 6%. Lyft’s reported completed trips are also internally inconsistent, differing by 49.7 million, or 81%. Lyft’s internal inconsistencies are 14 times greater than Uber’s internal inconsistencies.

INCOMPLETE TRIP REQUESTS

Incomplete trip requests are a measure of unserved demand and can be used to calculate completion rates. Incomplete trip requests are reported in Requests Not Accepted as a list and in Aggregated Requests Not Accepted as annual totals aggregated by zip code.

Table 8 and Table 9 show the total requests that were not accepted reported by Uber and Lyft in each report. Uber’s incomplete trip requests are internally consistent (numbers match exactly) in each report. Lyft’s incomplete trip requests are internally consistent in each report.

VEHICLE MILES TRAVELED (VMT)

VMT is a measure of the total amount of travel. It is used in many system performance metrics, including in environmental analysis to calculate emissions, and is a key indicator of demand and congestion. It is reported by trip in Requests Accepted and aggregated by driver-day in Number of Miles.⁵

Table 10 and Table 11 show VMT reported by Uber and Lyft in each report. Uber’s reported VMT is internally inconsistent, differing by nearly 1 billion VMT, or 59%. Lyft’s Requests Accepted report is incomplete and cannot be assessed for consistency of reported VMT.

DRIVER DAYS

Driver days are used to measure labor conditions and can be used to evaluate compliance with labor laws. Each record in the Number of Miles and the Number of Hours reports represents a driver day.

Table 12 and Table 13 show the total driver days reported by Uber and Lyft in each report. Uber’s reported driver days are internally inconsistent, differing by 1.4 million, or 15%. Lyft’s reported driver days are also internally inconsistent, differing by 100,000, or 1%. Uber’s internal inconsistency is 22 times higher than Lyft’s.

DRIVER HOURS

Driver hours are also used to measure labor conditions and can support evaluation of compliance with labor laws. Number of Miles reports total driver hours by driver day. Driver hours by trip for Period 2 (when a driver is en-route to pick up a passenger) and Period 3 (when the passenger is in the vehicle) can be derived from the Requests Accepted reports, but Period 1 (when a driver is waiting for a ride request) cannot be derived. Therefore, the total of Period 2 and Period 3 hours in Requests Accepted should be strictly less than the total hours in Number of Hours.

Table 14 and Table 15 show driver hours reported by Uber and Lyft in each report. Uber’s Requests Accepted, which only includes hours for Periods 2 and 3, reports 59 million driver hours, higher than the 47 million driver hours reported in Number of Miles which includes hours for Periods 1, 2 and 3. Lyft’s driver hours cannot be evaluated for consistency due to redactions of date and time information from Lyft’s Requests Accepted report.

SUMMARY OF INTERNAL CONSISTENCY

Table 16 summarizes the internal consistency findings for the 6 metrics for which consistency was evaluated for each company. The only metric Uber and Lyft reported in an internally consistent manner was incomplete requests. Uber’s reports were internally inconsistent for the remaining 5 metrics. Of the remaining metrics, Lyft’s reports were internally inconsistent for 3 and could not be evaluated for 2 because the required data is missing.

The 2020 public TNC Annual Reports for both Uber and Lyft are internally inconsistent for many of the most basic metrics. In two of the cases evaluated, Lyft’s reports are incomplete and their internal consistency cannot be evaluated.

The extent and scale of these inconsistencies prevent a sound understating of the state of the industry, and hinders the development of informed policy-making and effective regulatory oversight of TNCs. For example, whether Lyft completed 61 million trips, or 110 million trips, is critical to understanding the overall TNC market size. The discrepancy of one billion VMT in Uber’s Annual Reports is highly relevant for understanding California’s progress in meeting emission reduction goals.

Table 17 summarizes the consistency of the 2021 public TNC Annual Reports. Due to more extensive redactions in the 2021 public Annual Reports, a less extensive evaluation of consistency is possible. However, where consistency can be evaluated, inconsistencies are reduced in some instances. For example, Uber’s number of completed trips in the Requests Accepted and Aggregated Requests Accepted in their 2021 Annual Reports are perfectly consistent, and Lyft’s number of completed trips in these reports are nearly perfect, differing by 0.004%. But in many cases it is not possible to assess consistency because of the increased level of redaction in the 2021 Public Annual Reports.

< 1. Introduction and purpose 3. General Characteristics >