Access the report summary here.

Read the full report here.

Executive summary

Background and method

Food delivery workers (FDWs) in the gig economy are at risk of illness or injury while working. While the number of incidents involving FDWs that have been reported to the health and safety regulator in NSW is low, there has been an exponential increase over the past three years, from one incident reported in 2017 to 19 in the first half of 2020 alone. Tragically, a further five FDWs died in traffic accidents on Australian roads between September and November 2020, underscoring the pressing need to address this problem.

FDWs are commonly defined as independent contractors and are thus primarily responsible for their own work health and safety (WHS). However, food delivery platforms (FDPs) have significant influence over the work of their FDWs. FDPs and FDWs thus share a duty of care toward each other and anyone involved in or impacted by their work. However, lack of direct oversight of FDWs, in combination with workforce demographic factors and limited experience, means that FDWs may lack the capacity to fully and effectively manage WHS risks on their own.

The Behavioural Insights Team, in collaboration with Macquarie University and the Centre for WHS, undertook a four-phase project that aimed to improve the WHS of FDWs in the gig economy. Phase 1 aimed to describe the characteristics of FDWs and their WHS behaviours, knowledge, and concerns, while Phase 2 focused on exploring the WHS behaviours, knowledge and concerns of FDPs. The primary WHS concern nominated by FDWs and FDPs was traffic accidents. FDWs also identified verbal abuse and robbery or vandalism of delivery equipment as key concerns. Risk exacerbation behaviours among FDWs included using mobile phones while riding or driving, working while fatigued, wearing dark clothing at night, cycling on footpaths and in other pedestrian-only areas, failing to follow COVID-safe guidelines on social distancing and mask-wearing, and speeding or rushing. These behaviours are extremely widespread - only 5% of the FDWs we surveyed reported “never” rushing to deliver orders - and are driven largely by the desire to achieve and maintain high in-app ratings, satisfy customers, and maximise income. Detailed results of Phases 1 and 2 can be found on the Centre for WHS' Knowledge Hub.

This report details the results of Phases 3 and 4. In Phase 3, we worked closely and extensively with FDWs, FDPs, and an e-bike rental company to co-design a proactive risk reduction intervention to improve FDWs' safety on the road. In Phase 4, we evaluated the intervention in a randomised controlled trial (RCT) with a major FDP, who rolled out the intervention to their fleet in Greater Sydney and Greater Melbourne for a one-month trial period.

Results and discussion

An iterative co-design process with five FDWs, representatives of four major FDPs, and the owner of an e-bike rental shop was conducted to create the intervention. The co-design process involved 1:1 interviews, a workshop, and co-creation and iteration of the intervention and trial design. A key issue that emerged during co-design was that many FDPs intend delivery times to be an estimation, while FDWs perceive them as an expectation, beyond which they could incur penalties for failing to consistently meet the delivery times shown in the app. As a result, we co-designed an intervention with our stakeholders that aimed to reframe FDWs’ perception of delivery times as estimations rather than expectations. The intervention was a suite of four messages, each underpinned by a different behavioural science principle: beliefs about consequences, anchoring, descriptive social norms, and implementation intentions. The messages were then evaluated in a randomised controlled trial (RCT) that was conducted in partnership with a major FDP in Greater Melbourne and Greater Sydney. In the trial, the FDP sent one message per week via the FDP’s in-app messaging function to 7,066 of their active FDWs, with the remaining 7,215 FDWs acting as the control group.
FDWs who opened at least one message showed 3% slower average travel speeds relative to the control group, which was a statistically significant difference. This finding suggests that the intervention was modestly effective. For the fleet of 14,281 FDWs as a whole, however, the messages had no significant effect on average travel speed, the subjective belief that the FDP penalises FDWs for late deliveries, or FDWs’ perception of safety on the road while delivering for the FDP. This can be at least partially explained by low engagement with the messages and the post-trial survey, which is consistent with past challenges that FDPs and researchers have faced when attempting to engage FDWs as a singular cohort. The findings of this trial suggests that the greater challenge for implementation and scaling of messaging interventions is first determining how best to ensure these interventions are received, read, and acted upon by FDWs.