Aberrant Riding Behavior of Food Delivery Riders-Model

1. Abstract

The Aberrant Riding Behavior of Food Delivery Riders-Model is a comprehensive psychometric assessment tool developed by Qian, He, and Shi (2024) to evaluate the psychological and behavioral mechanisms driving unsafe electric bicycle and motorcycle riding among on-demand food delivery workers. Operating within the booming platform-based gig economy, food delivery couriers are frequently exposed to severe road safety hazards driven by algorithmic dispatching, rigorous delivery time limits, and commercial incentives. This assessment tool systematically captures the multivariate interactions between baseline safety knowledge, multidimensional safety attitudes, multi-category risk perceptions, and concrete aberrant riding behaviors. Comprising 31 items scored on a standardized 5-point Likert-type scale, the instrument operationalizes several distinct subscales: Safety Knowledge (SA), Safety Attitudes toward Responsibility (SA-r), Safety Attitudes toward Laws (SA-l), Risk Perception across General Violations (RP-gv), Risk Perception across Distraction and Mistakes (RP-dm), and Risk Perception across Aggressive Riding (RP-ar), coupled with specific manifestations of behavioral violations and errors.

Psychometric evaluation among adult commercial riders in China established robust structural, convergent, and discriminant validity, along with high internal consistency. Exploratory factor analysis demonstrated that the measurement model accounts for greater than 60% of total variance, with Kaiser-Meyer-Olkin sampling adequacy measures reaching 0.821 and 0.912 across structural components. Confirmatory factor analysis verified an optimal model fit (χ²/df = 1.929, RMSEA = 0.064, CFI = 0.916, IFI = 0.917, TLI = 0.904). Reliability coefficients indicated exceptional internal stability, with Cronbach’s α and Composite Reliability (CR) values spanning 0.706 to 0.944 and 0.836 to 0.934, respectively, while split-half Spearman-Brown coefficients demonstrated values up to 0.821. Average Variance Extracted (AVE) consistently surpassed the 0.50 threshold, confirming convergent validity, while the Fornell-Larcker criterion verified discriminant validity across all structural factors. This scale serves as an empirically validated diagnostic instrument for traffic psychologists, urban transportation engineers, occupational health specialists, and gig platform architects seeking to mitigate occupational injury and enhance urban road safety.

2. Keywords

Aberrant Riding Behavior, Food Delivery Riders, Electric Bike Safety, Risk Perception, Safety Knowledge, Safety Attitudes toward Laws, Safety Attitudes toward Responsibility, Traffic Psychology, Commercial Couriers, Occupational Road Safety

3. Authors

The Aberrant Riding Behavior of Food Delivery Riders-Model was conceptualized, developed, and empirically validated by a specialized research team in transportation engineering and traffic safety at Tsinghua University:

• Qian Qian: Department of Civil Engineering, Tsinghua University, Beijing, China. Primary contributor to scale conceptualization, behavioral taxonomy design, and survey administration.
• Junyi He: Department of Civil Engineering, Tsinghua University, Beijing, China. Contributed to empirical data collection, statistical modeling, psychometric parameter estimation, and structural equation modeling.
• Jing Shi (Corresponding Author): Professor/Lead Investigator, Department of Civil Engineering, Tsinghua University, Beijing, China. Research focus includes transportation planning, traffic safety management, and road user behavior. Email: [email protected].

4. Purpose

The primary purpose of the Aberrant Riding Behavior of Food Delivery Riders-Model is to provide an empirically grounded, psychometrically validated diagnostic instrument capable of assessing, dissecting, and predicting unsafe riding practices among commercial food delivery workers. In recent years, urban transit networks worldwide have witnessed a monumental influx of app-based delivery platforms. In this hyper-competitive sector, delivery couriers navigate two-wheeled vehicles (predominantly electric bicycles and low-displacement motorcycles) under stringent delivery windows dictated by artificial intelligence algorithms. This operational context fosters high rates of traffic violations, including red-light running, sidewalk riding, wrong-way riding, mobile phone distraction, and extreme speeding. Consequently, commercial riders face disproportionately high morbidity and mortality rates, positioning their behavior as an urgent public safety and occupational health issue.

Traditional driving and riding behavior inventories, such as the classic Driver Behaviour Questionnaire (DBQ), were calibrated primarily for non-commercial passenger vehicle drivers or recreational motorcyclists. Such instruments fail to capture the unique occupational pressures, regulatory ambiguities, platform dispatch constraints, and behavioral compromises inherent to the gig economy. The model formulated by Qian, He, and Shi addresses this gap by capturing the dynamic psychological chain connecting an individual’s cognitive safety knowledge to their normative attitudes, subjective risk evaluations, and self-reported aberrant riding behaviors.

From an applied perspective, the instrument facilitates multi-tiered organizational and policy interventions:

• Platform Safety Auditing: Algorithmic food delivery enterprises can deploy the scale to audit the baseline psychological safety profile of their courier fleet, identifying cohorts exhibiting low legal compliance attitudes or diminished risk perception.
• Targeted Training Interventions: Instead of relying on generic safety orientations, fleet managers can pinpoint specific deficits, differentiating between cognitive knowledge gaps (which require factual education) and attitude-defiance patterns (which require behavioral incentives and restructuring of algorithmic penalizations).
• Urban Transportation and Law Enforcement: Municipal transit planners and traffic enforcement agencies can utilize empirical findings derived from the model to isolate high-risk violation behaviors, tailoring traffic signal timing, dedicated micro-mobility lanes, and camera-based enforcement strategies to commercial delivery corridors.
• Academic Research: The scale offers traffic safety researchers an empirically robust tool to examine mediation and moderation mechanics involving occupational stress, cognitive fatigue, platform compensation schemes, and real-world crash involvement.

5. Psychological Construct

The construct operationalized by this instrument is multidimensional, delineating the cognitive, affective, evaluative, and behavioral components that culminate in unsafe micro-mobility operations. The model is structured into four core psychological domains containing distinct subscales:

1. Safety Knowledge (SA)

The Safety Knowledge domain measures a rider’s factual, cognitive comprehension of traffic regulations, vehicle operational constraints, helmet standards, right-of-way dynamics, and safety protocols. Rather than simply functioning as rote memorization, safety knowledge within this construct reflects the rider’s intellectual grasp of how specific operational behaviors directly alter situational vulnerability. A delivery worker possessing high safety knowledge understands the empirical danger of navigating pedestrian sidewalks at high speeds or the physical braking distances required on wet asphalt.

2. Safety Attitudes (Divided into Two Sub-dimensions)

Attitudes reflect enduring evaluative orientations toward regulatory structures and moral obligations. The model delineates safety attitudes into two correlated yet distinct constructs:

• Safety Attitudes toward Responsibility (SA-r): Measures the courier’s internal moral accountability, personal ethics, and perceived social responsibility toward other road users, including pedestrians, private cyclists, and motor vehicle drivers. Couriers with elevated SA-r endorse the perspective that avoiding accidents is a shared moral duty, refusing to rationalize reckless maneuvers purely on the grounds of commercial urgency.
• Safety Attitudes toward Laws (SA-l): Captures the individual’s respect for, and subjective agreement with, established traffic codes, legal sanctions, and official transit mandates. Couriers scoring low on SA-l perceive traffic regulations as arbitrary hindrances to economic productivity, viewing violations as acceptable shortcuts whenever enforcement cameras or police officers are absent.

3. Risk Perception (RP)

Risk perception denotes the subjective appraisal of the probability, severity, and personal vulnerability associated with engaging in specific dangerous road behaviors. Unlike general risk tolerance, this construct captures domain-specific cognitive appraisals across three distinct contextual categories:

• Risk Perception of General Violations (RP-gv): Evaluates how dangerous the rider considers intentional, rule-breaking acts, such as running red lights, riding against the flow of traffic (wrong-way riding), or illegally utilizing motor vehicle express lanes.
• Risk Perception of Distraction and Mistakes (RP-dm): Assesses the courier’s subjective estimation of danger regarding cognitive and sensory distractions, such as actively navigating mobile delivery applications while in motion, answering customer phone calls, or misjudging gaps in oncoming traffic due to perceptual errors.
• Risk Perception of Aggressive Riding (RP-ar): Measures the perceived hazard of hostile or excessively assertive maneuvers, such as abruptly cutting across lanes, tailgating slower cyclists, weaving through dense pedestrian crossings, or flashing lights and sounding horns aggressively.

4. Aberrant Riding Behaviors

The behavioral output dimension of the model measures self-reported engagement in aberrant actions on the road. Rooted in Reason’s taxonomy of human error, the construct differentiates between intentional violations (deliberate transgressions of formal safety codes driven by time pressure or habit) and unintentional errors/slips (failures of attention, misjudgments of distance, or motor coordination lapses induced by cognitive workload and fatigue).

6. Theoretical Framework

The Aberrant Riding Behavior of Food Delivery Riders-Model is anchored in an integration of social cognitive and occupational traffic psychology paradigms, drawing heavily from the Theory of Planned Behavior (TPB; Ajzen, 1991), the Health Belief Model (HBM; Rosenstock, 1974), and Reason’s Generic Error-Modelling System (Reason, 1990).

Theoretical Foundations and Cognitive Sequencing

Traditional cognitive frameworks postulate a direct knowledge-to-behavior pipeline: increasing an individual’s knowledge base should, theoretically, induce safer conduct. However, empirical traffic safety literature consistently demonstrates that safety knowledge alone exerts a weak direct effect on behavioral compliance. To resolve this theoretical disconnect, Qian, He, and Shi synthesized a mediation paradigm wherein safety knowledge serves as a foundational cognitive antecedent that influences behavior indirectly through two central psychological channels: safety attitudes and risk perceptions.

Within this framework:

• Knowledge as a Cognitive Foundation: Safety knowledge operates as the informational architecture. Without clear knowledge of road hazards and traffic ordinances, riders cannot accurately evaluate operational risks or formulate coherent attitudes toward legal compliance.
• Attitudes as Evaluative Catalysts: In alignment with Ajzen’s Theory of Planned Behavior, attitudes toward behaviors directly shape behavioral intention. An individual’s evaluation of legal frameworks (SA-l) and personal social responsibility (SA-r) determines the internal moral acceptability of breaking a law. Under algorithmic delivery pressure, negative attitudes toward laws diminish self-regulatory control, making traffic violations an accepted strategy for meeting delivery deadlines.
• Risk Perception as a Threat Filter: Borrowing from the Health Belief Model, the perceived susceptibility to and severity of adverse outcomes (such as severe physical collisions or police fines) act as behavioral deterrents. When couriers perceive high risk in aggressive riding or mobile phone distraction, the anticipated negative utility counterbalances the external rewards of speed, suppressing aberrant actions.

Integration with Occupational and Platform Gig Pressures

The model explicitly considers the occupational ecosystem of platform gig work. Unlike general commuters, food delivery riders operate in an environment governed by algorithmic punishment systems, customer satisfaction ratings, and financial delivery bonuses. This structural reality creates intense cognitive dissonance between occupational survival (delivering orders rapidly) and physical survival (riding safely). The theoretical model demonstrates that safety attitudes and risk perceptions serve as crucial internal psychological buffers, determining whether a rider yields to commercial pressures through aberrant riding or maintains self-protective compliance.

7. Validity

The validity of the Aberrant Riding Behavior of Food Delivery Riders-Model was rigorously assessed by Qian, He, and Shi (2024) through multiple empirical procedures evaluating construct, convergent, and discriminant validity among a representative sample of commercial riders in China.

Construct and Convergent Validity

Construct validity was established through sequential exploratory and confirmatory factor analyses. The underlying operational items were adapted from validated international and regional transportation safety inventories (e.g., Wang et al., 2018; Nguyen-Phuoc et al., 2020; Jamil et al., 2021) and refined through expert review to ensure ecological relevance to food delivery work.

Convergent validity evaluates whether individual survey items reliably capture their designated theoretical latent construct. This was statistically confirmed using the Average Variance Extracted (AVE) metric and Composite Reliability (CR):

• All latent dimensions yielded AVE values exceeding the standard psychometric threshold of 0.50 (AVE > 0.50). This confirms that more than 50% of the variance observed in the indicator items is accounted for by the underlying latent construct rather than measurement error.
• Composite Reliability values ranged from 0.836 to 0.934 across subscales, well above the recommended 0.70 benchmark, confirming high internal consistency across all latent variables.

Discriminant Validity

Discriminant validity verifies that theoretical dimensions remain statistically distinct from one another. This was tested using the rigorous Fornell-Larcker criterion:

• The square root of the AVE for each latent variable was systematically compared against the bivariate correlations between that construct and all other latent constructs in the structural model.
• Across all dimensions, the diagonal values (representing the square root of the AVE) significantly exceeded the corresponding inter-construct correlation coefficients in both rows and columns.
• These results confirm that while constructs like Safety Attitudes toward Responsibility and Safety Attitudes toward Laws share common variance, they represent distinct psychological dimensions.

8. Reliability

The psychometric reliability of the 31-item instrument was demonstrated across multiple indices of internal consistency and split-half stability:

Internal Consistency Reliability

The scale demonstrated high internal consistency across all operational subscales:

• Cronbach’s Alpha (α): Alpha coefficients across the individual subscales ranged from a minimum of 0.706 to a maximum of 0.944. These values exceed the conventional 0.70 threshold for research instruments, indicating robust internal consistency without excessive item redundancy.
• Composite Reliability (CR): CR values across all latent factors spanned from 0.836 to 0.934, indicating uniform measurement precision across constructs.

Split-Half Reliability

To confirm stability beyond item-total correlations, split-half reliability testing was conducted using the Spearman-Brown prophecy formula:

• The Spearman-Brown split-half coefficient for Safety Attitudes toward Laws (SA-l) reached 0.804.
• The Spearman-Brown coefficient for Risk Perception dimensions (specifically RP-p/RP categories) reached 0.821.

These coefficients verify that the scale possesses strong internal stability, resisting measurement error stemming from fatigue, item placement, or scale length.

9. Factor Analysis

The structural dimensionality of the instrument was evaluated through a two-stage factor-analytic workflow combining Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).

Exploratory Factor Analysis (EFA)

The initial structural extraction evaluated the multidimensional configuration of safety knowledge, safety attitudes, risk perceptions, and aberrant behaviors. Test assumptions and sampling adequacy were verified through the Kaiser-Meyer-Olkin (KMO) measure and Bartlett’s Test of Sphericity:

• KMO values of 0.821 and 0.912 were documented across operational sub-matrices, indicating excellent sampling adequacy.
• Bartlett’s Test of Sphericity yielded statistically significant results (p < 0.001), rejecting the identity matrix hypothesis and confirming appropriate data structure for factor extraction.
• Principal axis factoring with oblique rotation extracted factor components accounting for more than 60% of the total cumulative variance, confirming substantial explanatory power across the underlying dimensions.

Confirmatory Factor Analysis (CFA)

A structural equation modeling framework was utilized to conduct Confirmatory Factor Analysis to validate the multi-factor structural model. The validated CFA measurement model demonstrated an excellent fit to the empirical data across multiple goodness-of-fit indices:

• Normed Chi-Square (χ²/df): The model yielded a χ²/df ratio of 1.929, well below the conservative maximum threshold of 3.0, indicating acceptable discrepancy between the sample and model-implied covariance matrices.
• Root Mean Square Error of Approximation (RMSEA): The RMSEA was calculated at 0.064, well below the standard 0.08 cutoff value, confirming acceptable approximation error in the population covariance structure.
• Comparative Fit Index (CFI): CFI reached 0.916, exceeding the 0.90 baseline for good model fit.
• Incremental Fit Index (IFI): IFI was documented at 0.917, exceeding the 0.90 threshold.
• Tucker-Lewis Index (TLI): TLI was observed at 0.904, confirming solid baseline relative fit.

All standardized factor loadings on designated latent constructs were statistically significant (p < 0.001), confirming clear item assignment to targeted psychological dimensions.

10. Instrument / Measurement Tool

The technical parameters and administration protocol of the Aberrant Riding Behavior of Food Delivery Riders-Model are summarized below:

• Instrument Name: Aberrant Riding Behavior of Food Delivery Riders-Model
• Primary Reference Authors: Qian Qian, Junyi He, and Jing Shi (2024)
• Test Type: Psychometric Self-Report Questionnaire / Multi-Factor Behavioral Inventory
• Administration Modality: Electronic administration via digital survey platforms, mobile web interfaces, or platform-integrated terminal apps
• Target Population: Commercial food delivery riders, courier platform couriers, micro-mobility two-wheeled gig workers (Adults 18+ years)
• Total Number of Items: 31 items
• Response Format: Standardized 5-point Likert-type scale, typically structured as:
  • 1 = Strongly Disagree / Never / Extremely Low Risk
  • 2 = Disagree / Rarely / Low Risk
  • 3 = Neutral / Sometimes / Moderate Risk
  • 4 = Agree / Often / High Risk
  • 5 = Strongly Agree / Frequently / Extremely High Risk
• Construct Subscales:
  • Safety Knowledge (SA)
  • Safety Attitudes toward Responsibility (SA-r)
  • Safety Attitudes toward Laws (SA-l)
  • Risk Perception of General Violations (RP-gv)
  • Risk Perception of Distraction and Mistakes (RP-dm)
  • Risk Perception of Aggressive Riding (RP-ar)
  • Aberrant Riding Behaviors (Violations, Errors, Distractions)
• Scoring Procedures: Dimension-specific scores are derived by calculating the arithmetic mean or summation of items within each subscale. Higher scores on Safety Knowledge, Safety Attitudes, and Risk Perception reflect stronger cognitive awareness, positive safety norms, and higher danger awareness, respectively. Conversely, higher scores on the Aberrant Riding Behavior subscale indicate more frequent engagement in hazardous riding behaviors.

11. Permissions & Fee and Test Year

The Aberrant Riding Behavior of Food Delivery Riders-Model was developed and published in 2024. The empirical study and underlying measurement framework appeared in Transportation Research Part F: Traffic Psychology and Behaviour (Volume 100, January 2024, Pages 273–288).

Copyright and Usage Terms:

• The conceptual framework, psychometric properties, and structural findings are published by Elsevier Ltd. on behalf of the authors.
• The instrument is intended for academic research, public health investigations, and non-commercial educational evaluations without licensing fees, provided proper formal academic citation is rendered to Qian, He, and Shi (2024).
• Commercial entities, digital delivery platform corporations, and proprietary fleet software developers wishing to integrate the proprietary item bank into corporate safety auditing applications or algorithmic dispatch tracking systems should obtain prior formal permissions from the corresponding author, Professor Jing Shi ([email protected]), and the publisher.

12. References

The following scientific literature provides the theoretical, empirical, and methodological foundations for the Aberrant Riding Behavior of Food Delivery Riders-Model:

• Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
• Jamil, H. M., Cheng, L., & Farooq, D. (2021). Assessing motorcyclists’ risky riding behaviors: A cross-sectional comparative study using an extended motorcycle rider behavior questionnaire. Safety Science, 142, 105386. https://doi.org/10.1016/j.ssci.2021.105386
• Nguyen-Phuoc, D. Q., De Gruyter, C., Nguyen, H. A., & Mark, P. (2020). Risk perception and behavioral intentions of ride-sourcing motorcycle drivers in Vietnam. Transportation Research Part F: Traffic Psychology and Behaviour, 74, 303–315. https://doi.org/10.1016/j.trf.2020.08.031
• Qian, Q., He, J., & Shi, J. (2024). Analysis of factors influencing aberrant riding behavior of food delivery riders: A perspective on safety attitude and risk perception. Transportation Research Part F: Traffic Psychology and Behaviour, 100, 273–288. https://doi.org/10.1016/j.trf.2023.12.007
• Reason, J. (1990). Human error. Cambridge University Press. https://doi.org/10.1017/CBO9781139062366
• Rosenstock, I. M. (1974). Historical origins of the health belief model. Health Education Monographs, 2(4), 328–335. https://doi.org/10.1177/109019817400200403
• Wang, Y., Qu, W., Ge, Y., Sun, X., & Zhang, K. (2018). The driver behavior questionnaire: A meta-analysis. Accident Analysis & Prevention, 118, 186–198. https://doi.org/10.1016/j.aap.2018.06.015

13. Items of the Scale