Eliciting preferences of TNC users and drivers: Evidence from the United States

Transportation Network Companies (TNCs) have rapidly expanded in the U.S., with 2.61 billion passengers using TNCs in 2017 (a 37% increase from 2016). While proponents highlight increased mobility options and potential reductions in vehicle ownership and VMT through pooled rides, other studies point to induced travel and reduced transit ridership in some areas. Policymakers remain uncertain about net impacts on energy use, environment, and congestion because system-level outcomes derive from individual-level choices by riders and drivers. This study addresses data gaps by analyzing a 2017 survey (N = 11,902) of U.S. residents in TNC-served areas to estimate associations between socio-demographics and: a) individuals’ propensity to be TNC riders, drivers, or non-users; b) riders’ propensity to use pooled ridesourcing; and for drivers: c) inclination to switch to higher fuel-economy vehicles; and d) likelihood that driving for TNCs influences buying/renting/leasing a new vehicle. The study also provides descriptive evidence on usage frequency, mode substitution when a preferred mode is unavailable, ownership decisions, first/last-mile preferences, driver downtime activities, and pickup times.

The review covers the evolution and impacts of TNCs, user and driver characteristics, and research gaps. TNC growth was enabled by smartphones and internet connectivity; Uber and Lyft expanded widely and introduced pooling (UberPool, LyftLine; later UberPool Express). Daily/weekly use remains limited for most riders. Vehicle ownership impacts are mixed: some evidence of reduced driving and ownership among frequent users, but also increases in ownership when TNC services are suspended (e.g., Austin). Transit impacts are heterogeneous: studies report both complementary and substitutive effects depending on city size, transit coverage, timing, and competition between TNCs. User characteristics associated with higher TNC adoption include being younger, college-educated, urban residents, and living in dense, amenity-rich built environments; population and infrastructure density predict TNC demand. Driver studies document demographics (more men, many with college degrees), motivations (extra money, liking driving), and some labor aspects. Research gaps include limited revealed-preference, representative data for TNC-served areas, little attention to who pools rides, and virtually no prior work on TNC drivers’ preferences for fuel economy and vehicle acquisition decisions. This study contributes by addressing these gaps.

Data: Proprietary survey conducted by Strategic Vision Inc. in 2017 among 11,902 U.S. consumers in TNC-served areas, collected via targeted email. The dataset includes household socio-demographics (age, gender, marital status, education, income, ethnicity, residence location, commute mode, household size, vehicle ownership), TNC usage, attitudes, and preferences. TNC driver subsample: 1,541 drivers (about 25% drive daily). Weighting: To correct under-/over-representation of demographic groups, person-level weights were computed using iterative proportional fitting (IPF) to match joint distributions in population data (2016 ACS, 2010 Census) across 32 categories (age/gender, income, travel mode, residence location). Weights ranged from 0.15 to 4.65; all analyses use weighted data. Models: - Model 1: Multinomial logistic regression for preference category (TNC rider, TNC driver, TNC non-user as base). Covariates include gender, marital status, age, postgraduate indicator, income, metropolitan residence, household size 3+, total vehicle ownership, early adopter indicator. Nonlinear effects examined via predicted probabilities. - Model 2: Binary logistic regression for pooled ridesourcing use among TNC riders (user vs non-user of pooling). Covariates: gender, age, postgraduation, metropolitan residence, vehicle ownership, early adopter; interactions with age (gender×age, postgrad×age). - Model 3: Binary logistic regression for TNC drivers’ preference to switch to more fuel-efficient vehicles (prefer vs not prefer). Covariates: drive daily, single, age, postgrad, metropolitan residence, vehicle ownership, early adopter; interactions with age (single×age, postgrad×age, metro×age). - Model 4: Binary logistic regression for whether driving for TNCs was a major consideration in buying/leasing/renting a new vehicle (yes vs no). Covariates: drive daily, male, single, age, postgraduation, income, metropolitan residence, vehicle ownership, early adopter; interactions: single×age, postgrad×age, single×income. Estimation reports parameter estimates, odds ratios/relative risk ratios with 95% CIs. Predicted probabilities plotted over covariate supports with other variables held at means. Descriptive analyses include mode-use frequencies, substitution when preferred mode unavailable, impacts on vehicle ownership, first/last-mile preferences, driver downtime activities, and pickup times.

- Sample composition (weighted): 29% TNC users, 29% TNC drivers, 42% non-users. Among users, 13% had used pooled ridesourcing; among drivers, 53% would switch to more fuel-efficient vehicles; 47% reported high propensity to buy/lease a new vehicle with TNC driving a major factor. - Mode use and substitution: Frequent TNC users most often use personal vehicles (53–61%) or TNCs (22–32%); infrequent users: personal vehicles 79–87%, TNCs 3–5%. If TNCs are unavailable, 66% of those who mostly use TNCs would drive personal vehicles and 14% would use transit. If personal vehicles are unavailable, 31% would use TNCs and 46% carpooling/carsharing. Only 0.45% of frequent TNC users would not make the trip absent TNCs, suggesting limited induced demand in this sample. Findings imply TNCs and carsharing mainly capture personal driving demand with marginal effect on transit demand. - Ownership and trip purposes: About 10% of TNC users postponed purchasing a new car due to TNC availability. TNCs/carsharing/carpooling are often chosen for social/recreational trips (~46%). Top reasons for using TNCs: convenience (~24%) and avoiding driving after drinking (~21%). - Pooling awareness and behavior: Only 13% of TNC users had pooled; among these, ~34% of their past TNC trips were pooled. About 50% of users unaware of pooling options; 22% prefer private rides. - Non-users: ~36% prefer to drive themselves; ~21% reported no need for taxi/TNC previously; ~5% cite cost. Among non-users who drive for first/last-mile, ~41% would switch to TNCs for last-mile. - Driver experience and operations: 54% rated driving experience excellent; 28% neutral; 18% unsatisfactory. Average weekly miles rise with driving frequency (daily: 42 miles; every other day: 31; less than once per month: 16). 65% of daily drivers considered TNC driving in new vehicle decisions (declines with lower frequency). 93% use their primary vehicle for TNC driving. 29% drive to busier areas during downtime (potentially increasing VMT). Average pickup times: 9 minutes (peak) and 10 minutes (off-peak), adding ~2–3 miles VMT per trip. About 26% of drivers working >20 hours/week prefer diesel; 8–11% among those driving less; ~25% prefer hybrid electric vehicles for future. - Model 1 (multinomial logit): Likelihood of being a TNC user increases with age up to ~44 years, then declines; propensity to be a driver decreases with age; non-user likelihood increases with age. Higher income increases likelihood to ride but reduces likelihood to drive. Greater vehicle ownership slightly reduces likelihood to use TNCs in any form. Metropolitan residence and early-adopter status strongly increase odds of being a rider (OR ~1.53 and 1.36, respectively) and driver (RR ~1.94 and 4.81). Postgraduates and singles more likely to be riders; males less likely to be riders and drivers compared to non-users. - Model 2 (pooling): Probability of pooled ridesourcing decreases with age; males less inclined to pool; metropolitan residents more inclined (OR ~1.73); higher vehicle ownership reduces pooling odds (OR ~0.87 per additional vehicle). Interactions: age effects differ by gender and education—young females (and below-postgrad) have higher pooling propensity than males (and postgrads), with crossover ages around 54 (gender) and ~34 (education). - Model 3 (drivers switching to fuel-efficient): Daily drivers (OR ~1.41), early adopters (OR ~1.47), postgraduates (OR ~4.25), and metropolitan residents (OR ~4.02) are more inclined to switch; age negatively associated. Interactions show that pro-fuel-efficiency propensity of postgraduates and metro residents declines more steeply with age, with crossover around age ~48. Married drivers are more inclined than singles below age ~60. - Model 4 (drivers’ new vehicle decisions): Likelihood that TNC driving was a major consideration decreases with age; greater for daily drivers (OR ~2.09), males (OR ~1.58), metropolitan residents (OR ~1.90), early adopters (OR ~2.59), and those with more vehicles (OR ~1.18 per vehicle). Overall, higher income associates with lower inclination, but interaction shows income raises propensity for married drivers while reducing it for singles; e.g., at $100,000 income, predicted probability ~0.51 (married) vs ~0.38 (single), all else equal.

The findings address the study’s research questions by quantifying how socio-demographics relate to being a TNC rider, driver, or non-user; riders’ propensity to pool; and drivers’ preferences for fuel economy and vehicle acquisition decisions. Results indicate TNCs primarily displace personal driving rather than transit, and pooling adoption is constrained by age, gender, education, household vehicle availability, and urban residence, with awareness gaps being substantial. The non-linear age effects and interactions (e.g., age×gender, age×education) add nuance beyond prior linear findings. For drivers, metropolitan residence, higher education, early-adopter status, and high engagement (daily driving) correlate with stronger interest in fuel-efficient vehicles and greater likelihood that TNC driving influences vehicle purchases. These insights are relevant for planners seeking to reduce VMT and emissions via targeted incentives for pooled rides and for promoting higher fuel economy within TNC fleets. They also help TNCs tailor outreach (e.g., pooling awareness campaigns) and partner with automakers or lessors to offer attractive deals to likely adopters of efficient vehicles, potentially supporting green service options and regulatory targets.

This study uses a large, weighted, revealed-preference survey of U.S. residents in TNC-served areas to estimate associations between demographics and: (1) the propensity to be a TNC rider, driver, or non-user; (2) riders’ inclination to use pooled ridesourcing; and, uniquely, (3) TNC drivers’ willingness to switch to fuel-efficient vehicles and (4) whether TNC driving influences new vehicle acquisition decisions. Key contributions include uncovering non-linear age effects and interactions for pooling, and identifying driver segments most receptive to fuel-efficient vehicles and TNC-related vehicle purchases. Policy implications include targeted pooling promotion (especially among younger females and urban residents), campaigns to increase pooling awareness, and OEM–TNC partnerships or leasing programs to accelerate adoption of high fuel economy vehicles among frequent, metropolitan, and technologically oriented drivers. Future research should pursue causal identification (e.g., randomized experiments with TNCs), assess geographic heterogeneity, and evaluate system-level impacts of targeted interventions on VMT, emissions, and transit integration.

The study is observational and identifies associations rather than causal effects. The survey sample, while weighted via IPF to reflect population distributions across key categories, initially under-/over-represented some demographic groups and is limited to residents of TNC-served areas in 2017, which may affect generalizability. Self-reported behaviors and preferences may be subject to recall or desirability biases. Model specifications, while including interactions and nonlinear checks, may omit unobserved factors (e.g., detailed built environment measures for all respondents), and results may vary across regions and over time.