When Innovation Meets Resistance: Information Provision and Consumer Decisions on Autonomous Taxis

We will conduct a randomized information provision experiment examining how different framings of autonomous taxi (AT) information affect consumer adoption decisions. A total of 1,200 participants will be randomly assigned to four groups of 300 participants each. The Control Group will receive neutral, non-AT-related scientific information. The Positive Information Group will receive information about economic efficiency gains of ATs. The Negative Information Group will receive information about potential socioeconomic impacts regarding job replacement for taxi drivers. The Combined Information Group will receive both positive and negative information.
Our experiment employs a two-stage design to examine both immediate responses and longer-term behavioral changes in autonomous taxi adoption.
Stage 1 encompasses three sequential components: First, participants complete a pre-intervention survey measuring baseline preferences and demographic characteristics. Second, they receive their assigned information treatment, with content varying by experimental group but maintaining consistent length and format. Third, participants receive a promotional coupon for the Robotaxi APP, enabling one free autonomous taxi ride within a specified validity period.
Stage 2 captures sustained behavioral changes through dual data collection channels. Following the coupon expiration date, we administer a comprehensive follow-up survey to measure post-intervention attitudes and document actual usage experiences. Simultaneously, through our partnership with the platform provider, we continuously monitor participants' APP usage patterns to track revealed preferences in AT adoption. This two-stage design enables us to measure both stated preferences through surveys and revealed preferences through actual behavioral data.

2024-12-06

2025-01-01

1. Immediate AT Preference: 10-point scale measuring willingness to use AT services, collected immediately after the information intervention
2. Short-term AT Adoption: Binary indicator (0/1) of whether the participant uses the free AT coupon during its validity period
3. Long-term AT Engagement: Continuous measure of APP usage frequency in the three months following the initial intervention
4. Long-term AT Attitude: 10-point scale measuring willingness to use AT services in the follow-up survey

Our primary outcomes track both stated and revealed preferences for AT adoption across three distinct time horizons. Immediate preferences are measured through a 10-point willingness-to-use scale collected directly after the information intervention in Stage 1, providing a baseline measure of the intervention's immediate impact on stated preferences. Short-term adoption is measured through actual coupon redemption behavior during the one-month validity period, offering our first revealed preference measure. Long-term engagement is tracked through APP usage data provided by our platform partner, including both APP access frequency and completed rides over the three-month post-intervention period. Finally, the long-term attitude measure from our follow-up survey enables us to compare changes in stated preferences over time.

1. Economic Perception: Five-category measure of perceived price differences between ATs and traditional taxis
2. Labor Impact Perception: Five-category measure of expected driver displacement rate
3. Safety Perception: Five-category measure of expected change in accident rates
4. Environmental Impact Perception: Five-category measure of expected change in PM2.5 emissions

Our secondary outcomes examine the mechanisms through which information interventions affect AT adoption decisions. All measures are collected in the follow-up survey using multiple-choice questions with precisely defined categories:
1. Economic Perception measures expected price differences:
Categories range from "More than 50% cheaper" to "More than 50% more expensive".

2. Labor Impact Perception assesses expected employment effects:
Categories range from "No impact" to "Will replace over 75% of drivers".

3. Safety Perception captures expected changes in accident rates:
Categories range from "60-90% reduction" to "60-90% increase".

4. Environmental Impact Perception measures expected pollution effects:
Categories range from "20-30% reduction" to "30-40% increase" in PM2.5.

We conduct a randomized field experiment examining consumer adoption of autonomous taxis (ATs) in Wuhan, China. We recruit 1,200 participants from high-traffic urban locations and randomly assign them to four treatment groups. Each participant completes a baseline survey, receives randomly assigned information about ATs, and is given a promotional coupon for one free AT ride. We then track both their short-term coupon usage and long-term app engagement. A follow-up survey is conducted after the coupon expiration to measure sustained attitudes and behaviors.

Not available

Randomization is implemented through a built-in computerized randomization algorithm in our survey platform. When a participant begins the survey, the platform automatically assigns them to one of the four treatment groups based on a pre-set randomization sequence. This ensures immediate and unbiased treatment assignment during field implementation.

The unit of randomization is the individual participant. Each person who agrees to participate in our study is independently randomized into one of the four treatment conditions: Control, Positive Information, Negative Information, or Combined Information. All treatments are administered at the individual level, and there is no clustering or group-level randomization in our design. This individual-level randomization maximizes statistical power and allows us to control for location-specific fixed effects in our analysis.

No

Not applicable - our design uses individual-level randomization without clustering.

1,200 individual participants from urban areas in Wuhan.

Control Group (T1): 300 participants
Positive Information (T2): 300 participants
Negative Information (T3): 300 participants
Combined Information (T4): 300 participants
Total: 1,200 participants

With 300 participants per treatment arm (1,200 total), assuming 80% power and 5% significance level, we can detect, across treatments: Immediate preferences: 0.46 point difference on 10-point scale. Short-term adoption: 10.7 percentage point difference in coupon usage. Long-term engagement: 0.77 ride difference in monthly usage. Long-term attitudes: 0.56 point difference on 10-point scale. These calculations account for differential attrition across time periods (0% immediate, 5% short-term, 20% long-term) and assume independent observations without clustering.