The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence David Alzate Table of Contents Executive Summary 5 I. Introduction 6 A. Background: The rationale for protecting digital platform workers 6 B. Mapping the evidence: A framework for potential regulatory and worker protection interventions 8 II. Interventions and findings: What does the evidence say? 10 A. Tackling market power asymmetries 10 B. Tackling information asymmetries 13 C. Tackling competition barriers 15 D. Leveraging platforms to increase uptake of social insurance 17 E. Applying the evidence to low- and middle-income contexts: Key considerations 19 Appendix: Study tracker: Impact evaluations, experiments, and theoretical models estimated using data 21 References 38 Acronyms BLS Bureau of Labor Statistics CWS Contingent Worker Supplement EPRS European Parliamentary Research Service EU European Union EW Employing Workers ILO International Labour Organization LOM Mobility Orientation Law on Transport MEIC Ministry of Economy, Industry, and Commerce MOPT Ministry of Public Works and Transportation MOU Memorandum of Understanding MTESS Ministry of Labor, Employment, and Social Security NUPSAW National Union of Public Service and Allied Workers NYC New York City OECD Organization for Economic Co-operation and Development OHS Occupational Health and Safety SCJN Supreme Court of Justice of the Nation The effects of regulating platform-based work on employment outcomes: 4 A review of the empirical evidence Acknowledgment The policy brief ‘‘The effects of regulating platform-based work on employment outcomes. A review of the empirical evidence’ was prepared by David Alzate (Consultant) as part of the World Bank Labor Global Solutions Group’s initiative ‘Better Labor Regulations for the Digital Economy and Beyond’. The project, which started under the leadership of Michael Weber (Senior Economist) in FY21–22 and continued under Matteo Morgandi (Senior Economist) and Eliana Carranza (Senior Economist) in FY23–24, aims to develop context-appropriate regulatory frameworks for platform-based work in developing countries. This policy brief is part of a comprehensive series under this project, which includes: • ‘Regulating platform-based work in low- and middle-income countries: towards a context-appropriate approach’ by Matteo Morgandi and David Alzate; • ‘The regulation of platform-based work: recent regulatory initiatives and policy options for developing countries’ by Maho Hatayama and Dagmara Maj-Swistak; • ‘The effects of regulating platform-based work on employment outcomes. A review of the empirical evidence’ by David Alzate; • ‘The economic rationale to regulate platform-based work’ by Jonathan Stöterau (unpublished manuscript). The policy brief draws on insights from two-day workshops hosted by the World Bank in May 2023, titled ‘Regulating platform-based work in developing countries: How to balance job opportunities and workers’ protection’. These workshops facilitated extensive consultations with external experts, whose contributions were instrumental in shaping the recommendations presented herein. We extend our gratitude to our peer reviewers, Mark Graham (Professor, Oxford Internet Institute), and Ilsa Meidina (Senior Social Protection Specialist, World Bank), for their valuable comments. Additionally, we appreciate the support provided by Sara de Lorenzo (Consultant) in the publication and dissemination process and Agnes Mganga (Program Assistant) for her administrative assistance. As digital work platforms continue to expand globally, especially in developing economies, it is imperative to balance the potential job opportunities they offer with adequate worker protections. The findings and recommendations of this series aim to inform policymakers and stakeholders in creating effective and sustainable labor regulations that respond to the unique challenges posed by the digital economy. Through this series, we aspire to contribute to the broader discourse on labor regulation and to support the development of policies that ensure inclusive and equitable economic growth in the digital age. The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 5 Executive Summary Despite the need for knowledge on the impacts of regulating digital platform work, empirical evidence remains thin, especially in low- and middle-income country (LMIC) settings. Out of 59 studies in this brief, 18 are experiments, impact evaluations, or theoretical models estimated using data, and 14 of those 18 studies cover LMIC-based workers (see Section I.B). Effective interventions must be tailored to the realities of digital work. Digital platform markets have characteristics that may differ from other types of markets, such as power and information asymmetries between platforms and workers as well as fluid entry and exit of workers from the market (see Sections II.A and II.B). The need to tailor interventions to the characteristics of digital work is apparent in the case of minimum earnings interventions, which have had mixed effects in digital platforms. These mixed effects are in part due to the rapid entry and oversupply of new workers into digital platforms after minimum wages are imposed, resulting in limited increases in overall workers’ earnings in some cases (see Section II.A). Reputation systems—reviews and information about workers and employers—are highly valued by digital platform workers. However, they are prone to information asymmetries, suggesting that regulation or protective measures could play an important role (see Section II.B). Policy makers can leverage digital platforms to enroll workers in social insurance and social protection schemes. Social protection and insurance coverage is low among digital platform workers. Policy makers could leverage platforms’ data about workers, and their contact with workers, to target efforts to extend social insurance coverage. However, more evidence is needed to determine the best way of doing so (see Section II.D). Policy makers aiming to protect digital platform workers should consider not only labor market regulations (LMRs) but also interventions related to product market regulation (PMR). The relationship between digital workers and digital platforms is also affected by platforms’ competitive environment. However, there is lack of empirical evidence on the effects of tackling competition barriers in digital platforms (see Section II.C). Platforms and policy makers should obtain more information about what digital platform workers value and tailor regulatory and protective measures accordingly. Digital platform workers have a variety of preferences regarding which social benefits they would prioritize receiving. However, more efforts are needed to collect, expand, and incorporate this information into decision-making, particularly in LMIC settings (see Section III). The effects of regulating platform-based work on employment outcomes: 6 A review of the empirical evidence I. Introduction A. Background: The rationale for protecting digital platform workers Digital platform work is defined as task- or gig-based work that takes place through a digitally mediated marketplace that “connects ‘workers’ (providing goods or services) with ‘customers’ (who can be businesses or individuals)” (Datta et al. 2023; Stoterau 2024). Digital platform work includes location- based work through applications such as Uber and Lyft and web- based work on websites such as Amazon Mechanical Turk (MTurk) and Upwork. These different types of platforms vary widely in terms of their characteristics, but they all involve a digitally mediated relationship between a worker and a client (organization, firm, or individual) in exchange for ‘gig-’ or task-based services (Woodcock and Graham 2020). There is also wide heterogeneity in terms of the tasks that workers can engage in across web- and location- based work as well as in the skills required to complete those tasks (Stoterau 2024). This brief will indicate when certain studies and findings refer to web-based or platform-based work. Work on digital platforms constitutes “a growing and non- negligible part of the labor market,” with web-based digital platform work alone encompassing 4.4 to 12.5 percent of the global labor force (either as full- or part-time workers) (Datta et al. 2023). The global employment share of digital platform work is likely greater, as this estimate does not include workers active in location-based services such as ride-hailing and delivery platforms. The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 7 The growing availability of digital platform work tasks are often not made available to them on web- could bring about promising benefits. Digital based platforms (Berg et al. 2018), or at times, they platforms could benefit workers by giving them more experience outright discrimination and/or lower choice, flexibility, and information about available earnings due to their country of origin (Graham, jobs. In developing contexts, where digital work Hjorth, and Lehdonvirta 2017; Lehdonvirta et al. 2021 platforms are becoming increasingly popular, they in Haidar and Keune 2021). could bring new opportunities for income generation in a way that is more observable by policy makers Social insurance coverage is low among digital than non-digital informal labor (Datta et al. 2023). platform workers, in part due to the legal form Such opportunities are particularly valuable when it in which these jobs are classified (Datta et al. is difficult for individuals to find ‘traditional’ work in 2023). ILO surveys found that only “about three out local labor markets, due to either a shortage of jobs or of ten surveyed workers on crowdwork platforms factors such as discrimination (Graham et al. 2017). are covered by some form of social insurance” (Berg These platforms could also benefit micro, small, and et al. 2018 in Behrendt, Quynh, and Rani 2019), medium enterprises by widening the talent pool they and “women have less access to social insurance have access to (Datta et al. 2023). compared to men.” Digital platform workers, in turn, may value the benefits of social insurance (Ghorpade, Yet evidence suggests that digital platform work Rahman, and Jasmin 2023; Gruber 2022), but the has several features that place workers at a cost they are willing to pay for social insurance is less disadvantage in relation to their platform-enabled clear. While digital platforms could voluntarily bear employers. For example, Dube et al. (2020), through the costs of providing some form of social protection descriptive work and experimental estimates, find coverage to workers, this is rarely done on a voluntary a high degree of employer market power (that is, basis. Instead, most platforms have so far avoided monopsony) in the web-based work platform MTurk, considering their workers as dependent employees, causing workers to be paid less than they should be and their legal status is subject to significant paid based on their productivity. Online platform debate and varies according to the nature of tasks workers might also not have as much flexibility in performed. the organization of their work time and place, as platforms often impose tight deadlines (Yin, Suri, and This brief is developed as part of a series and Gray 2018) or assign tasks in a rigid manner, including provides an overview of the empirical evidence by giving tasks during irregular times of the day (for on the impacts of regulatory and worker example, nighttime tasks for digital workers serving protection interventions related to digital work clients in different time zones) (Wood, Lehdonvirta, platforms. The theoretical and economic rationales and Graham 2018). for protecting workers against the market failures that surround digital platform work are discussed Workers might also be exposed to mistreatment in Stoterau (2024). Another brief describes the or lack of pay by digital firms and clients. Almost experiences in various countries in adopting labor nine out of ten workers in an International Labour regulations or legal classifications from the legal Organisation (ILO) survey have had work rejected or standpoint (Hatayama and Swistak 2024). We bring have had payment refused (Berg et al. 2018). Digital complementary evidence and guidance to policy work platforms may not reward workers based in makers by reviewing the empirical evidence on the low- and medium-income countries (LMICs) as much effects of introducing regulations. as is warranted based on their skills and experience level (Beerepoot and Lambgrets 2014). In addition, It reviews 59 research papers, including 18 workers may not always be able to avoid ‘bad’ experiments, impact evaluations, or theoretical online jobs or firms as they often do not have good models estimated using data—out of which 14 information about the quality of tasks and the clients include workers based in low- or middle-income who assign them, even though clients are often able contexts. This review searched for experimental to pick and choose workers based on public reviews and quasi-experimental studies related to digital and ratings (Holtz, Scult, and Suri 2022; Kingsley work platforms through keyword searches in et al. 2015). Platform workers based in LMICs might Google Scholar and EconLit, by reviewing citations be exposed to additional vulnerabilities, as the best of papers found, and through a review of recently The effects of regulating platform-based work on employment outcomes: 8 A review of the empirical evidence published (2022–2023) papers in economics journals Most studies considered in this overview are and conference schedules. The search included concentrated in the LMR space. However, when keywords for experimental and quasi-experimental considering potential regulatory and worker methodologies as well as keywords related to the protection interventions for digital workers, policy topics and subtopics of this brief. In turn, the topics makers must not focus solely on LMR, but they and overarching framework of the brief were defined should also consider PMR and social protection and in consultation with World Bank staff and expanded insurance. and modified based on the availability of the evidence. Nonexperimental (that is, qualitative, descriptive, Indeed, overlap exists among these three spaces. One theoretical, or simulation based) work was also regulation might address issues that exist in both included in sections where the experimental and LMR and PMR spaces, for instance. This overview quasi-experimental work was scarce or to provide maps the studies found across four main issue areas rationale and motivation for open questions for future that cut across LMR, PMR, and social insurance research. The end result aimed to be an exhaustive and protection. These four issue areas reflect the list of empirical studies about regulations and worker key sources of vulnerability and market failures protection interventions—that either have already that surround digital workers: (i) market power been enacted or could potentially be enacted—in the asymmetries, (ii) information asymmetries, (iii) digital workspace. A full list and description of the 48 competition barriers, and (iv) an under-coverage research papers is included in the appendix. of social insurance. B. Mapping the evidence: A framework for Most of the impact evaluation evidence included in this review deals with interventions concentrated in potential regulatory and worker protection the first two areas (market power and information interventions asymmetries). The above framework is not necessarily a comprehensive map of all the possible intervention What forms of regulatory and worker protection types and vulnerabilities that affect digital platform interventions are possible in the digital workspace? workers. Yet it serves as a starting point to chart out Traditional labor protections tend to be concentrated the available evidence as of the time of writing. in the ‘labor market regulation’ (LMR) policy space. These policies grant workers with rights—such as In addition, the ‘right’ regulation or intervention type a minimum wage or standards on working hours, might also differ based on the type of digital platform dismissal procedures, and contracting—to protect work: location based versus web based. These have against power asymmetries. distinctive features that could translate to different types of worker vulnerabilities: health and safety However, the relationship between workers and hazards might differ between an individual working firms, including on digital platforms, is also affected as an Uber driver and an individual completing tasks by firms’ viability and their business decisions. on MTurk, for example. Therefore, PMRs—including competition policies, openness to trade and foreign investment, mandates, The ‘right’ intervention will also depend on the specific and exemptions based on firm size, price controls, features of the local labor market(s) surrounding preferential treatment in public procurement, and digital platform work. The final section of this brief access to finance—are also interventions that can includes a discussion of different contextual features directly affect firms’ treatment of workers and, as a specific to low- and middle-income contexts that result, job outcomes (Alzate et al. 2024). could affect the generalizability of the findings from the existing evidence base. In addition, outside of either the labor market or product market regulatory space, digital workers face a precarious lack of social insurance and protection. Interventions in this third ‘social protection and insurance’ space might still benefit digital workers, independent of changes to official or government regulation about labor and product markets. The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 9 Figure 1: A framework for potential regulatory and worker protection interventions for digital platform workers Labor market Product market Social insurance and regulation space regulation space protection space Tackling market Tackling Tackling Leveraging platforms power information competition to increase uptake asymmetries asymmetries barries of social insurance • Minimum wages • "Reputation • Portability of benefits • Identifying informal • Higher pay, type of systems" • Natural monopolies and vulnerable pay • Matching services • Price dumping workers and offering • Freedom of and intermediaries • Noncompeting clauses formalization/ association • Transparency on • Entry and licensing insurance enrollment • Job and task flexibility compensation barriers incentives • Movement across • Arbitrary exclusion • Formalizing platforms employment • Monitoring workers relationship • Firms voluntarily providing social insurance products to platform workers Source: World Bank. The effects of regulating platform-based work on employment outcomes: 10 A review of the empirical evidence II. Interventions and findings: What does the evidence say? A. Tackling market power asymmetries The monopsony power of digital work platforms might translate into situations where employers have outsized influence on working conditions. In theory, this can cause workers to be underpaid, restricted in their flexibility, and unduly monitored. Possible regulatory responses can range from the introduction of a minimum wage to interventions that strengthen workers’ bargaining power. Below are findings from 14 studies—7 of which cover workers from LMICs and 8 of which are experimental or impact evaluations—that assess the effectiveness of interventions related to these market power asymmetries. Introducing a minimum compensation or wage Findings from two experimental studies, one quasi-experimental study, and two theoretical model studies of minimum wages or earnings schemes in web-based and location-based platforms suggest these can have mixed effects in high-income country (HIC) and LMIC settings. Wages of hired workers may increase, but overall impacts on earnings are limited because the wage floor may lead to an oversupply of workers at the new wage level (Asadpour et al. 2022; Horton 2018; Nakamura and Siregar 2022; Stanton and Thomas 2021; Van Inwegen et al. 2022). However, there is room for innovation by digital platforms to address the issues of worker oversupply that might emerge from a minimum-wage style of policy. One experimental study finds that Lyft’s ‘Priority Mode’ feature for drivers in the United States solved a driver oversupply issue, leading to increased driver earnings on average as well as benefits for riders (Krishnan et al. 2022). The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 11 Study details: and employers might both be left worse off as a result of reduced hiring. • Van Inwegen et al. (2022) found that randomly assigning workers to receive one of three different • Asadpour et al. (2022) analyzed the effects of minimum wage levels in a web-based platform New York City and Seattle’s minimum earnings had heterogenous effects while increasing overall regulations for ride-hailing providers using wage equality. Treated workers who historically a theoretical model. The regulations require charged below the minimum wage reduced minimum payments to drivers for each ride their probability of employment by 14–32 based on the distance and time traveled. percent, were around 6 percent more likely to The researchers estimate that this led to an exit the platform, and their total earnings did oversupply of drivers—9 percent more drivers not increase (with earnings for some of these entered and used the platform as they were workers decreasing by 8 percent). attracted to higher payments, but riders’ demand decreased due to higher costs. The • Horton (2018) found that when a web-based oversupply of drivers resulted in many drivers platform experimentally introduced an employer- being idle and not finding work while using the level minimum wage, the wages of hired workers app, thereby limiting the regulation’s ability to (who worked from the United States, India, the increase earnings. They estimated the maximum Philippines, and Bangladesh) rose by between 4 feasible gain in net earnings for drivers in this and 9 percent per dollar increase in the minimum scenario was 3 percent. wage. However, this came at the expense of a reduction in overall hiring (ranging from 2.5 • Krishnan et al. (2022) conducted an experiment to 10 percent, depending on workers’ previous on Lyft’s ‘Priority Mode’ feature in the United earnings) and hours worked (with reductions States, which allows drivers to increase their as large as 30 percent). The reduction in hours probability of being paired to riders during worked is in part explained by how employers specific, prioritized hours. They found that this began hiring higher-skilled workers after the feature can solve the issue of worker oversupply wage policy. while safeguarding worker flexibility, increasing gains for drivers, riders, and Lyft in the process. • Nakamura and Siregar (2022) employed a Priority Mode “resulted in a generation of differences-in-differences method and synthetic system surplus equivalent to” 10 to 13 percent control methodology to evaluate the impacts of of total driver earnings. In addition, they found a federal policy on minimum fares per ride for a 60 percent positive satisfaction rating among drivers on ride-sharing apps in Indonesia. They drivers using the feature. found that, overall, the policy did not increase driver earnings or wages despite increasing Higher pay or different type of pay trip prices. This was a result of a large number of lower-earning drivers entering the platform, Even when minimum wages do not exist, platforms as reflected by a 24 percent increase in excess might decide to raise (or lower) the earnings of all ‘supply hours’ (that is, the sum of all idle hours platform workers. Two experimental studies—one from all drivers). As a result, mandating a with a location-based platform in the United States minimum fare did not seem to increase overall (Hall John, and Daniel 2023) and one with a global earnings when these minimum fares did not web-based platform (Doerrenberg, Duncan, and account for idle time (or driving distance time) Loffler 2023)—suggest that increases in platform that drivers face. workers’ task-based compensation might not lead to large or sustained benefits. This may be particularly • Stanton and Thomas (2021) used data from true when markets re-equilibrate or if workers do not transactions from a web-based platform with work more in response to a wage increase. global workers (that is 89 percent of transactions in the marketplace crossed international Digital platforms could also transition digital platform borders) to simulate the impacts of introducing workers away from a task-based, performance-pay a minimum wage. They estimated that workers model—the standard in digital work platforms— The effects of regulating platform-based work on employment outcomes: 12 A review of the empirical evidence toward a model of fixed pay resembling standard highlighting the potential benefits of flexible digital working arrangements. These platforms could also work arrangements. However, one experimental add new forms of ‘bonus’ payments on top of existing study finds that non-location-based platform work payment schemes. However, the empirical evidence (that is, on MTurk) may not always be as flexible as on these alternative types of pay schemes remains expected. In qualitative work across multiple African thin (one study, Hodor [2022], looks at the effect of countries, Anwar and Graham (2020) find some gig bonus payments for gig and permanent workers in an workers highly value autonomy but note that this online manufacturing firm). increased autonomy does not necessarily translate to improved working conditions or livelihoods. In Study details: circumstances where workers value flexibility and do not obtain it, increasing flexibility may increase the • Hall, John, and Daniel (2023) measured how quality of workers’ output. experimental increases in Uber’s base price in the United States—and payments for drivers— Study details: led to earnings increases for workers. However, these increases only lasted eight weeks, as the • Chen et al. (2017), using data on hourly earnings, demand for rides adjusted in response to the estimate that Uber drivers in the United States higher price (a 10 percent increase in a ride’s benefit significantly from real-time flexibility. base fare led to a 2.5 percent reduction in total Their theoretical model calculated Uber’s flexible transportation hours). driving arrangement led to labor surplus equal to 40 percent of total expected earnings for drivers, • Doerrenberg, Duncan, and Loffler (2023) or US$150 per week on average. For workers to evaluated an experimental increase and be indifferent between the flexible arrangement decrease (both by 20 percent) in task-based and a more restricted one, their wages would wages in Amazon MTurk. Higher wages reduced need to increase by more than 50 percent. the probability of workers quitting a labor task by 8.5 percent, while lower wages increased • Yin, Suri, and Gray (2018) find that MTurk this probability by 18.0 percent. These effects “affords workers far less flexibility than widely translate into labor supply elasticities that are believed,” with a large part of the inflexibility different when increasing versus decreasing coming from employers’ tight deadlines for wages (0.44 for the wage increase group; 0.89 tasks. They experimentally varied the amount of for the wage decrease group), suggesting that task flexibility for workers, finding this flexibility policies that decrease wages might have larger led workers to produce a larger amount of work impacts on labor supply than policies that with similar quality. They also find that “workers increase them. would give up significant compensation [at least $0.86 per hour] to control their time” and gain • Hodor (2022) estimated the effect of introducing more flexibility. bonus payment on top of regular earnings for gig workers and permanent workers in an online, • Anwar and Graham (2020) conducted a four- global manufacturing firm using a theoretical year qualitative study with 65 workers in South model. They found that the two different types Africa, Kenya, Nigeria, Ghana, and Uganda of workers responded differently to incentives. to assess how platform-based remote work The bonus payments increased productivity affected their perceived freedom, flexibility, among gig workers by 12 to 17 percent but had no precarity, and vulnerability. statistically significant impacts for permanent workers. Free movement across platforms Strengthening standards around job and task Survey data about online platform workers suggest flexibility that, while many of them work across multiple platforms, there are some restrictions to workers’ One study from the United States finds that Uber mobility. For example, the lack of portability of drivers benefit from the platform’s flexibility, reviews and ratings systems tends to lock some The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 13 workers into a single platform (the role of reputation B. Tackling information asymmetries systems is further discussed in Section B). Further research is warranted to determine the potential The power imbalance between employers and benefits and risks of increasing workers’ mobility workers in digital work platforms is also often across platforms. reflected in information asymmetries. Platform employers — including clients who assign tasks to Study details: platform workers and platform owners—often have more information about workers than workers do • Berg et al. (2018) studied responses to “an ILO about employers, tasks, and compensation. Further, survey of working conditions covering 3,500 employers can exert their informational advantage workers living in 75 countries around the world to monitor workers. These asymmetries can hinder and working on five English-speaking microtask workers’ ability to find digital work that is desirable platforms.” They find that almost half of and a good match with their profile. In turn, this respondents reported having worked on more creates opportunities for third-party firms to provide than one platform in the month preceding the matching and intermediary services that fill in the survey, in part due to an insufficient availability informational gap. of tasks. The remaining half, however, worked on only one platform, “explaining that this was due Regulation could play a role in correcting these to the high start-up and transaction costs of asymmetries. Below are findings from 12 studies, spreading oneself across platforms.” including four experimental studies, related to information transparency, reputation systems, • ILO (2021) highlighted how workers tend to be monitoring, and matching services in global web- locked into a single digital platform, in part due based platforms that could inform the design of to the “incompatibility of metrics used by the potential future regulations. major platforms” such as worker reviews and their work and financial histories. Improving transparency on work compensation and quality Freedom of association: workers’ organizations and trade unionization Increasing transparency in compensation for tasks might lead to better matches with platform workers, Digital workers might be able to counteract market based on one experimental study (Horton, Johari, and power asymmetries through collective action and Kircher 2021). organization. While this review could not find any experimental or impact evaluation evidence on the Study details: impact of unionization among digital work employees, there are some case studies exploring how digital • Horton, Johari, and Kircher (2021) revealed worker mobilization has led to gains for workers in a signal about employers’ willingness to pay the form of stronger social protection coverage. for more experienced workers to a randomly assigned set of jobseekers within an online gig Study details: platform for tasks that could be completed remotely. In response, jobseekers targeted their • Behrendt, Quynh, and Rani (2019) provided applications to employers that matched their an overview of how trade unions contributed experience level and tailored their earnings bids, to facilitating social protection for platform leading to an overall increase in hours worked of workers, looking at case studies of effective 4.6 percent. digital worker lobbying in Denmark and Germany. Regulating ‘reputation systems’ • Wood, Lehdonvirta, and Graham (2018) used survey data and interviews to highlight the role ‘Reputation systems’ refers to the existence and use of internet-based communities in facilitating of worker and employer reviews and information collective organization among digital freelancers about platform work experience. Experimental in Southeast Asia and sub-Saharan Africa. evidence from a web-based platform shows workers The effects of regulating platform-based work on employment outcomes: 14 A review of the empirical evidence place high importance on receiving positive reviews skills certificate led to a 2.1 percent increase in (Holtz et al 2022). In addition, experimental evidence earnings (that is an average gain of US$1.88). The suggests that reviews can improve workers’ likelihood returns to signaling were up to 1.5 times larger of future employment and wages in web-based for workers with zero work history compared to platforms (Pallais 2014). The effects of reviews on the average worker. workers may be transmitted (and compounded) by how a digital platform’s algorithm works. As Wood et • Argawal, Lacetera, and Lyons (2016) analyzed al. (2019) point out in qualitative work, workers with applications by workers in low-, middle-, and the best reviews in digital platforms tend to receive high-income countries for jobs posted on oDesk. more work due to the platform’s algorithmic ranking They found that lower-income applicants were of workers within search results. “only about 60 percent as likely to be hired” by contractors from high-income contexts relative Nonexperimental evidence suggests that providing to similar applicants from HICs. However, workers with a skill certificate can reduce employer workers who signaled more platform experience uncertainty and lead to higher earnings for workers (that is, had a higher number of prior jobs on (Kassi and Lehdonvitra 2019). Similarly, standardized the platform than the median worker) were and verified work history information appears to be more likely to be hired and to earn more, and particularly beneficial for workers based in lower- this was especially true for workers from lower- income contexts (Argawal, Lacetera, and Lyons 2016; income countries. These findings suggest that Lehdonvirta et al. 2018). However, worker reviews standardized and verified information about might also reproduce existing inequalities among workers can benefit LMIC-based gig jobseekers jobseekers (Lukac and Grow 2021). and address disparities in online opportunities. Research also underlines the importance of allowing • Lehdonvitra et al. (2018) analyzed data from workers to assess the reputation of employers—an a large, global web-based platform and found area where little oversight currently exists (Benson, that verifiable information which signals Sojourner, and Umyarov 2018). Future regulation about workers’ work experience—that is could consider strengthening workers’ access to the number of projects completed by each information about employers’ reputation. worker on the platform—led to an increase in task compensation. Per standard deviation Study details: unit increase in work experience, workers’ compensation per task increased by 6 to 13 • Holtz, Scult, and Suri (2022) used a survey percent, with larger gains for LMIC-based experiment to measure the value that workers workers (for example, Filipino workers saw a assigned to positive feedback on Upwork 16 percent increase in pay for writing tasks, (including workers based in HICs and LMICs), compared to 7 percent in the United States). estimating that the median freelancer valued a single positive review at around US$50. • Lukac and Grow (2021) estimated the effect that worker reputation plays in job outcomes through • Pallais (2014) experimentally hired and gave a simulation, finding that reputation systems evaluations to workers in oDesk (including “can potentially reproduce inequalities present HIC as well as LMIC-based workers). They in offline labor markets and produce unfair found providing evaluations almost tripled the outcomes that disproportionately favor already probability of inexperienced workers finding successful applicants.” employment from 12 percent to 30 percent as well as almost tripled their average earnings • Wood-Doughty (2018), using a theoretical from US$10 to US$27. model, compared the effect of information from reviews to other information about workers • Kassi and Lehdonvitra (2019), through an (such as standardized exam scores and country) event study, analyzed the effects of providing in the online labor market oDesk. They found workers with a skill certificate in a web-based that, somewhat contrary to the above findings, digital platform, finding that an additional “reviews have a relatively small effect on both The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 15 wages and attrition,” with a 1 standard deviation and Thomas 2015) in web-based platforms. However, increase in a workers’ combined review score safeguards should be put in place to ensure that reducing their probability of exit by only 1 these intermediaries do not themselves exploit percent. However, reviews did appear to reward workers or capture their earnings (Graham, Hjorth, good workers and punish bad ones. and Lehdonvirta 2017), as part of the original appeal of digital platforms may be that workers do not • Benson, Sojourner, and Umyarov (2018) studied need to pay for intermediaries to find employment the reputation of employers in MTurk. They opportunities. found that there is very little oversight for employers—no authority disciplines employers Study details: that refuse payments and workers have no contractual recourse or appeal process. In an • Graham, Hjorth, and Lehdonvirta (2017) find experiment, they found that posting employer that even though digital work platforms appear reviews on a third-party website had an impact to facilitate a more direct connection between on workers’ choices. Employers with good workers and employers, intermediaries are reputations “recruited workers about 50 percent still common. They find qualitative evidence to more quickly than otherwise-identical employers suggest these intermediaries can at times take with no ratings,” and “100 percent more quickly advantage of workers by engaging in low-pay than those with very bad reputations.” and strict working conditions. Monitoring workers • Horton (2016) finds that “algorithmically recommending workers to employers for Monitoring online platform workers might raise the purpose of recruiting can substantially privacy concerns among workers and reduce workers’ increase hiring: in an experiment conducted willingness to work, according to results from one in an online labor market, employers with experimental study on MTurk. technical job vacancies that received recruiting recommendations had a 20 percent higher fill Study details: rate compared to the control.” In addition, they find “no evidence that the treatment crowded- • Liang et al. (2022) investigated workers’ out hiring of non-recommended candidates.” responses to monitoring in MTurk along three dimensions: monitoring “intensity (how much • Stanton and Thomas (2015) study the role of information is collected), transparency (whether intermediaries in the platform oDesk. Despite the monitoring policy is disclosed to workers), the idea that online platforms allow workers and control (whether workers can remove and employers to directly engage without sensitive information).” They estimate that the need for an intermediary, they found that workers are apprehensive about monitoring intermediaries play an important role in these and, on average, the compensations required markets. Around 30 percent of non-US oDesk for workers to accept monitoring are between workers are affiliated with an intermediary. In US$1.8 and US$1.6 per hour (roughly 37.5 to addition, they found “workers affiliated with an 28.6 percent of average hourly wages). agency have substantially higher job-finding probabilities and wages at the beginning of their The role of matching services and intermediaries careers compared to similar workers without an agency affiliation.” Facilitating services that match digital workers with employers—either in the form of algorithmically generated recommendations or intermediary C. Tackling competition barriers companies—may be a promising way to reduce information asymmetries. Reducing these The relationship between digital workers and digital asymmetries, in turn, might increase employment platforms is also affected by platforms’ competitive and wages, based on findings from one experiment environment. Platforms with large market power, for (Horton 2016) and one descriptive study (Stanton example, will likely treat digital workers differently— The effects of regulating platform-based work on employment outcomes: 16 A review of the empirical evidence imposing their monopsonistic power—than platforms Price dumping operating under a more competitive environment. The firms that own platforms can also subject workers Online platforms might be able to engage in ‘price to certain constraints, such as barring them from dumping’—the offering of their products and services working for competitors, to preserve their market at prices lower than their cost, sacrificing revenue power. for the sake of gaining market share. For example, a ride-sharing platform might be able to stifle the As a result, tackling competition barriers—for competition by offering ‘unnaturally’ low prices for instance, through PMR—might lead to better rides. Whether these practices indeed occur among outcomes for digital platform workers. However, digital work platforms, whether this varies for on- which exact competition barriers to tackle and how to location versus web-based platforms, and whether address them are not straightforward. Overall, there these practices are subject to existing regulation is lack of empirical evidence on whether lowering about fair competition practices are still debated competition barriers can improve digital platform questions (for example, Agrawal 2021; Bamberger workers’ outcomes. and Lobel 2017; Bostoen 2019). Below is an overview of some competition barriers In addition, the potential effects of ‘price dumping’ that could be potentially subject to regulations on digital workers’ welfare are also understudied. If and interventions in the digital platform space. The digital platforms that engage in price dumping also evidence in this section is more descriptive than it is slash workers’ salaries in the future to make up for experimental or quasi-experimental: only two quasi- losses in revenue, for example, workers’ earnings experimental studies, both from the United States, might be affected. are included. The section emphasizes knowledge gaps and key open questions for future research. There is also evidence of digital platform workers engaging in ‘wage dumping’ of their own: accepting Natural monopolies very low wages for the sake of out-competing other workers (Aleksynska, Bastrakova, and Kharchenko Whether to encourage competition in digital 2019). Whether wage dumping is connected to price platforms through regulatory or other interventions dumping and whether regulatory interventions depends on whether these platforms constitute a can effectively address their negative effects are natural monopoly—such that a “single firm will serve also questions where further empirical evidence is a market more efficiently than competing firms” warranted. (Ducci 2020). In the presence of a natural monopoly, the right regulatory approach might focus on Noncompeting clauses preventing the abuse of power while limiting (rather than promoting) competition. However, there is lack Digital platforms may restrict competition by of empirical evidence on whether digital platforms requiring workers to agree to noncompeting constitute natural monopolies, and on whether this clauses—barring them from working for competitors’ differs by platform type and location. platforms. While this might appear to be at odds with digital platforms’ offer of flexibility, there are some Study details: cases of platforms imposing these kinds of restraints on workers (McDonald, Williams, and Mayes 2020). • Ducci (2020), in a theoretical overview, describes Evidence is needed on the impacts of such clauses, the ambiguity behind classifying ride-hailing and if they are indeed enforceable, as well as on services as natural monopolies. In some cases, potential mitigating strategies through regulation. they could potentially be seen as natural monopolies due to substantial “demand-side Entry and licensing barriers economies of scale through the creation of large networks.” However, in other cases, competition Occupational licenses—which impose specific might be desirable and possible based on the “size requirements on workers who wish to perform of demand, density of population, and availability certain kinds of services—aim to guarantee quality of alternative methods of transportation” within standards and protect consumers. Evidence from two a given geographical market. quasi-experimental studies of occupational licensing The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 17 within on-location, home services digital platforms Portability of benefits in the United States suggests that it might reduce labor supply while having muted effects on customer Digital workers might value portable benefits, that satisfaction. is, social benefits, such as health insurance and pension plans, that they can keep with them as they Study details: move from one digital platform to another. However, digital platforms might have an incentive to resist • Farronato et al. (2020) studied the effects of portable benefits. This is because digital platforms occupational licensing laws within a large online might be able to extend exclusive, non-portable platform for residential home services. They benefits to keep workers from ‘moving’ and working found that more stringent licensing regulations for competitor platforms. were “associated with less competition and higher prices, but not with any improvement in Whether digital workers stand to benefit from customer satisfaction as measured by review portable benefits and whether regulation can play ratings or the propensity to use the platform a role in safeguarding this portability are questions again.” lacking in empirical evidence. • Blair and Fisher (2022) exploited two natural experiments (state variation in licensing laws D. Leveraging platforms to increase and a change in licensing laws within a state) uptake of social insurance to study the effects of occupational licensing in Angi’s HomeAdvisor—a leading platform within There is evidence to suggest that individuals the US home services market. They found that experiencing unemployment or unexpected shocks licensing reduced successful matches between can turn to online platform work as a ‘buffer’ to customers and workers for tasks by 25 percent. protect against income losses (Jackson 2022, Jones This was driven by a reduction in the labor supply and Manhique 2022; Kass 2022; Kecht and Marcolin of workers. The researchers found the elasticity 2022). of workers accepting a task with respect to licensing was −0.11. However, online platform work does not substitute the need for social protection and insurance. As Arbitrary exclusion of competitors described in Section I.A, social protection and insurance coverage is low among digital platform The concentrated market power held by some workers. Many of these workers might be able to digital platforms can, in theory, result not just smooth their consumption owing to digital platforms, in discrimination against workers but also in the but they might still require social benefits and arbitrary exclusion of competitors. For example, protections against shocks, such as health insurance Stylianou (2018) notes examples of large tech firms and retirement plans. (such as AT&T and Apple) excluding competitors through several methods. These include vertical Policy makers might be able to leverage digital integration (that is, harming competitors by blocking platforms to increase uptake of social insurance access to other markets in the supply chain) and in several ways. First, through regulation, policy blocking access to a competitor’s service (for makers can require platforms to extend insurance example, blocking access to an app on an operating to workers, that is, by mandating a formal employee system or making an app incompatible). relationship, in which the employee gets access to contributory social protection coverage. Platforms More evidence is needed on whether the digital might also choose to extend social benefits to platforms discussed in this brief engage in arbitrary workers voluntarily, without the need for regulation. exclusion against competitors, whether this varies In addition, policy makers can leverage platforms’ by location-based versus web-based platforms, and ‘data’ about workers to identify vulnerable or informal whether regulation that addresses these practices workers and accordingly target efforts to enroll these results in a more competitive environment that workers in social insurance schemes. improves workers’ outcomes. The effects of regulating platform-based work on employment outcomes: 18 A review of the empirical evidence This section summarizes findings from six studies— and Deliveroo, Glovo, Ola, Swiggy, and Uber’s including one experimental study and most of them provision of in-ride insurance of varying degrees. including workers from LMICs—related to these In Deliveroo’s case, for instance, insurance covers potential interventions. riders “against injuries and third-party liability while they are online and for one hour after they Formalizing an employment relationship have gone offline.” Deliveroo extended further benefits to workers in France in the form of paid Mandating digital work platforms to pay taxes on sick leave in response to a series of protests over their employers—a way of formalizing an employment pay dispute (Boucherak 2019). relationship that could include contributory social insurance coverage—might result in lower hiring, • Rhani and Dhir (2020) provide a descriptive according to estimates from one simulation study. overview of the impact of COVID-19 on online gig Further empirical evidence is needed on the effects platforms, highlighting how several platforms of requiring platforms to register their workers as set up emergency COVID-19 funds and “other employees (or, at least, to establish a clear legal forms of sick pay to assist workers” during the definition that constitutes some type of employment pandemic. and guarantees a certain set of benefits) and extend social benefits accordingly. Using platforms to incentivize workers to enroll in social protection and insurance Study details: Policy makers could potentially leverage digital • Stanton and Thomas (2021) used data from platforms to make it simpler for workers to enroll transactions from a web-based platform in, and contribute to, social insurance products. with global workers (that is, 89 percent of These products can range from social security transactions in the marketplace crossed to unemployment insurance (UI) and investing in international borders) to simulate the impacts of voluntary savings plans. The empirical evidence on introducing a 10 percent tax paid by buyers when these incentives remains thin: this review found only hiring jobseekers, which is meant to reflect the one related experimental study (Guerrero and Silva- costs of paying an income tax on a worker. Their Porto 2020) and one quasi-experimental study (Garin simulation found that this could lower hiring by et al. 2023). around 26 percent, in large part due to a decline in the number of jobs posted on the platform by Study details: 34 percent. • Behrendt, Quynh, and Rani (2019) gave an Firms voluntarily providing social insurance to overview of how Uber and other ride-sharing platform workers platforms have facilitated access to official social protection coverage in Uruguay, The firms that own digital work platforms have, in Malaysia, Indonesia, Estonia, Lithuania, and some instances, voluntarily provided social insurance Sweden, for example, by permitting Uber and benefits for online gig workers in high-, middle-, drivers to automatically deduct social security and low-income countries. However, empirical contributions and therefore formally contribute evidence on the impact of these firm-led initiatives to the social security system in Uruguay or by remains thin. simplifying the tax reporting of income earned from Uber in Sweden. This underlines the Study details: potential of online gig platforms to simplify the process for workers to enroll in and pay for social • ILO (2021) provides an overview of how location- insurance. based online gig platforms have provided medical coverage benefits to their workers, including • Guerrero and Silva-Porto (2020) sent out the introduction of a medical insurance plan by invitations to 5,022 Cabify drivers in Peru to join the ride-share scheme DiDI Chuxing in China one of two voluntary savings plans: an emergency savings plan “in which drivers could save 2% of The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 19 their weekly earnings to cover emergencies” and • What regulatory instruments are feasible a more flexible plan “that offered the driver the and enforceable in LMIC contexts? The option to save 3% of their weekly earnings each effectiveness of any regulation will depend time they exceed a threshold, which the driver on the ability not only to craft and legislate themselves determined.” They found that 18 the regulation but also to ensure its de facto percent of drivers signed up to one of the two implementation. Yet the enforceability of schemes after 8 weeks, with the emergency regulations specific to digital work platforms savings plan having a higher take-up (20 percent in LMICs is not guaranteed. Indeed, drivers of versus 16 percent). They also found that “after location-based ride-hailing platforms often four months, the average savings generated by circumvent regulatory requirements, such as drivers on the platform was USD 29.” having a taxi license, in some LMICs. • Garin et al. (2023) examined the impact of • What do LMIC-based workers want? expanding UI to self-employed workers, including Digital platform workers in low- and middle- digital platform workers, in the United States income contexts likely have different levels as a part of the Pandemic Unemployment of social insurance and protection coverage Assistance program during COVID-19. Employing than digital platform workers in high-income a multivariable and instrumental variable contexts, especially if they are not employed regression analysis, they found the expansion of elsewhere in the formal sector. Yet there is lack UI potentially led to a decrease in work: for each of experimental evidence on the social benefit dollar increase in UI, reported self-employed preferences of LMIC-based digital workers, with income receipts fell by US$0.5–0.6 for platform only one (survey experiment) study from Malysia workers. found (Ghorpade, Rahman, and Jasmin 2023).1 In that context, web-based and location-based E. Applying the evidence to low- digital platform workers expressed a high level of willingness to pay for UI, retirement savings, and and middle-income contexts: Key accidental and injury insurance. However, the considerations exact type of insurance workers preferred varied depending on whether they already had access While the overall empirical evidence on regulations to some type of social protection. A one-size-fits and worker protection interventions in digital work all policy is unlikely to work in providing social platforms is scarce, this is particularly true for LMIC insurance coverage to digital platform workers in contexts. Indeed, the above overview only found LMICs, given the large amount of heterogeneous 18 studies that included workers based in low- or preferences these workers have. middle-income contexts. • Similarly, there is lack of evidence on the Below are four main features and considerations about preferences of LMIC-based platform workers labor markets in LMICs that suggest interventions regarding earnings regulations such as a might work differently than in high-income settings. digital platform minimum wage. Heeks (2017) Of course, not all LMIC contexts are the same, and summarizes findings from studies that suggest digital workers’ experiences might vary widely from that LMIC-based platform workers can earn one LMIC to the next. These four considerations are “typically 10–20 times the local minimum wage.” instead grounded on a few key similarities across It might be the case that LMIC-based workers most LMIC contexts: the challenges of enforceability, prefer regulation that focuses on improving other a relative lack of evidence about digital worker and elements of platform work besides minimum consumer preferences and responses, and a large earnings schemes, especially if imposing informal sector: minimum wages results in some workers being 1  Some evidence on eliciting digital workers’ preferences for social benefit coverage also exists from high-income contexts. Gruber (2022) finds that US Uber drivers value retirement savings, health savings, and sick leave benefits almost as much as equivalent cash payments. The effects of regulating platform-based work on employment outcomes: 20 A review of the empirical evidence made worse off. However, evidence on these As these platforms mature, however, potential issues preferences remains thin. around profitability and/or worker well-being may begin to emerge. At this point, policy makers might • How do LMIC-based workers respond? be more tempted to regulate digital platforms. Similarly, workers in LMICs have different Yet regulating them at an advanced stage may be employment options outside of the digital work difficult if they are already powerful actors in the local sector than their HIC counterparts, given the economy.2 large size of the informal economy. If changes in regulation lead to changes in the costs and The empirical evidence does not (yet) offer clear benefits of engaging in digital platform work guidance on which is the best timing for regulating (for example, if digital platform work becomes digital platforms. Policy makers who wish to adopt less flexible or more expensive as a result of recommendations from the available evidence regulation), LMIC-based workers might respond should keep in mind when other regulatory reforms differently, given they might be able to find took place, the current stage of ‘maturity’ of the alternative gig work in the informal sector. digital platform market in their own economy, and whether these two answers align with one another to • How do consumers respond to digital work determine the generalizability of other findings. regulations in LMICs? Customers in LMIC contexts might also respond differently to changes in online platforms, potentially leading to impacts on earnings and opportunities for workers. For example, if the introduction of a minimum wage increases the cost per ride of a ride-sharing app, will consumers in an LMIC context decrease their demand by more or less than their counterparts in an HIC context, and how will this affect driver earnings? In addition to these four considerations, policy makers, particularly in LMICs as well as in HICs, should keep in mind that the ‘right’ regulatory response may also vary based on the level of ‘maturity’ of the digital platform economy in their context. In their initial phase, digital platforms may dedicate large amounts of resources into marketing and subsidizing their own services (for example, Uber or Lyft subsidizing trips, even if doing so generates a loss for the company). In these early stages, the relative attractiveness of digital platforms due to their growing size may tempt policy makers to avoid imposing stringent regulations and instead see them as an alternate solution to high unemployment or informality rates. 2  Thank you to Ilsa Medina for this point. Appendix: Study tracker: Impact evaluations, experiments, and theoretical models estimated using data Study Country Digital Study type Intervention Study sample What is the Main findings Impact on Impact on Impact on reference platform type regulation or labor supply earnings occupation/ intervention (probability of job quality/ studied? working or hours formality worked) Asadpour et United Ride-hailing Theoretical Introducing n.a. Analyzing The researchers estimate Estimate labor The n.a. al. 2022 States platforms model a minimum the effects of this led to an oversupply supply "increases oversupply (for example estimated compensation New York City of drivers: more drivers by 9% in response of workers Uber, Lyft) using data or wage and Seattle’s enter and use the platform to higher in response minimum as they are attracted to earnings." to the earnings higher payments, but policy may regulations riders’ demand decreases limit their for ride-hailing due to higher costs. The earnings: providers oversupply of drivers estimate "the The regulations results in many drivers maximum require being idle and not finding feasible minimum work while using the gain in net payments to app, therefore limiting earnings is drivers for each the regulation’s ability to about 3%." ride based on increase earnings. the distance and time traveled Ghorpade, Malaysia Not specified Survey n.a. 1,038 gig Vignette-based "The analysis finds overall n.a. n.a. n.a. Rahman, and experiment workers, experiment a large unmet need for Jasmin 2023 including to ascertain social insurance among digital gig workers’ gig workers, as well as a freelancers willingness to high level of willingness and location- pay for social to pay for (especially) based workers insurance unemployment insurance, coverage retirement savings, and accidental and injury insurance." A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 21 22 Study Country Digital Study type Intervention Study sample What is the Main findings Impact on Impact on Impact on occua reference platform type regulation or labor supply earnings Impact on A review of the empirical evidence intervention (probability occupation/ job studied? of working or quality/ formality hours worked) pation/ Benson, Multiple Amazon Randomized Regulating 36 employers The effect Posting employer "Employers "Estimate that n.a. Sojourner, (HIC as well MTurk Controlled “reputation on MTurk of providing reviews on a third- with good posted wages and Umyarov as LMIC) Trial (RCT) systems” reviews about party website has an reputations would need to 2018 digital work impact on workers’ recruit workers be almost 200 platform choices: employers about 50 percent greater The effects of regulating platform-based work on employment outcomes: employers with good reputations percent more for bad-reputation “recruited workers quickly than employers and about 50 percent more our otherwise- 100 percent quickly than otherwise- identical greater for identical employers employers with no-reputation with no ratings,” and no ratings and employers to “100 percent more 100 percent attract workers at quickly than those with more quickly the same rate as very bad reputations.” than those good-reputation with very bad employers." reputations." Blair and United Angi’s Natural Entry and A large online What is the Licensing reduces the Elasticity n.a. n.a. Fisher 2022 States HomeAdvisor: experiment licensing marketplace impact of success rate of of workers Industry barriers in the US$500 occupational customer search on accepting leading billion home licensing on the platform by 25 a task with platform in services the likelihood percent. The reduction respect to home services industry that a customer in the success rate of licensing is market where we engaged in customer search in the −0.11. observe search on a presence of licensing task-level digital platform is fully explained by a variation in finds at reduction in the labor occupational least one worker supply of workers licensing for who is legally on the platform and 21 million permitted to do not by an increase in transactions. the work? customer search. Study Country Digital Study type Intervention Study What is the Main findings Impact on labor supply Impact on Impact on reference platform type sample regulation or (probability of working earnings occupation/ intervention or hours worked) job quality/ studied? formality Chen et al. United Uber Theoretical Strengthening 260,605 The value of work Using data on hourly "Estimate large labor "We compute Estimate that 2017 States model standards active Uber flexibility provided earnings, estimate supply elasticities driver labor wages would estimated around job and drivers by Uber that Uber drivers in the exceeding 1.5 for most surplus— need to increase using data task flexibility United States benefit drivers and on the accounting by more significantly from aggregate level." for 40% than 50% to real-time flexibility, of total "make drivers earning more than expected indifferent twice the surplus they earnings, between would in less flexible or $150 per the highly arrangements. week on adaptable Uber average— arrangement and under the more restricted existing Uber arrangements." arrangement." Doerrenberg, Not MTurk RCT Higher pay or 1,168 MTurk Experimental Wage increases have a Workers in the wage n.a. n.a. Duncan, and specified different type workers increase/decrease positive effect on labor decrease group Loffler 2023 of pay with a US IP in task-based wages supply whereas wage have a 18%p higher address (We announce a decreases reduce labor probability to quit the piece rate of $0.15 supply in our task. labor task (compared per transcribed labor supply responses to the control group), picture and workers to wage increases while workers in the complete a batch and decreases are wage increase group of six transcriptions asymmetric; workers are 8.5%p less likely for the announced react more strongly to quit (relative to the wage. Workers are to wage decreases control group). The randomly assigned than wage increases treatment effects in to one of three of equal magnitude (in Panel A translate into groups: (i) the wage absolute terms). labor supply elasticities increases by 20%, (ii) of 0.44 for the wage the wage decreases increase group, and 0.89 by 20%, or (iii) the for the wage decrease wage remains group. constant (control group)) A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 23 24 Study Country Digital Study type Intervention Study sample What is the Main findings Impact on Impact on Impact on reference platform type regulation or labor supply earnings occupation/ job A review of the empirical evidence intervention (probability of quality/ formality studied? working or hours worked) Dube et al. Multiple MTurk Descriptive n.a. Mixed (from None - descriptive Find substantial Estimate n.a. n.a. 2020 analysis different study monopsony power in remarkably (elasticities) studies) MTurk, as measured consistent using double by the elasticity estimates of machine of labor supply the labor supply The effects of regulating platform-based work on employment outcomes: learning and facing the requester elasticity facing experimental (employer). They MTurk requesters. estimates estimate low labor supply elasticities, around 0.1, with little heterogeneity. Farronato et United Large online Multivariable Entry and Over 1 million The role of The platform-verified n.a. n.a. More stringent al. 2020 States platform for regression licensing requests by occupational licensing status licensing regimes residential plus event barriers consumers licensing laws on of a professional do not improve home study in hundreds individual choices is unimportant for transaction quality services of distinct and market consumer decisions as measured by service outcomes relative to review review ratings or categories ratings and prices. the propensity throughout More stringent of consumers to the United licensing regulations use the platform States for are associated with again. On the cost over eight less competition side, we find that months. and higher prices more stringent but not with any licensing regimes improvement in result in less customer satisfaction competition and as measured by higher prices. review ratings or the propensity to use the platform again. Study Country Digital Study type Intervention Study sample What is the Main findings Impact on Impact on earnings Impact on reference platform type regulation or labor supply occupation/ job intervention (probability of quality/ formality studied? working or hours worked) Garin et United Platform work Multivariable Using IRS filers from Examine the We present n.a. For each dollar n.a. al. 2023 States observed in regression platforms to Vermont and impact of UI evidence that increase in UI tax data (90 plus incentivize Massachusetts expansions the availability of given by the state, percent is instrumental workers to 2012–2021 to include new PUA benefits reported receipts transportation- variables enroll in social self-employed resulted in many fall by between related) protection and workers, in a individuals who US$0.5 and insurance program known were platform US$0.6, and self- as Pandemic workers in 2019 employment profits Unemployment not reporting any fall by US$0.22 for Assistance (PUA) self-employment primary platform during COVID-19. income in 2020 and workers. Total 2021. We show that earnings, the sum of this appears to be profits plus wages, a real labor supply falls only slightly response rather more, around than more activity US$0.24, suggesting falling below 1099-K that most of the reporting gaps. impact comes from reductions in reported profits. Glasner United Uber Quasi- Introducing “Data from Local minimum Studied “the effect “When the analysis “Among “Among 2023 States experimental a minimum 2000 to 2018 wages outside of of minimum wage is restricted transportation transportation analysis compensation on nonemployer the digital work increases on work exclusively to and warehousing and warehousing (differences- or wage establishments, sector that is not covered transportation services, a 10% services, a 10% in- a category by minimum wage and warehousing increase in the increase in the differences; of workers laws” in the United services—the minimum wage minimum wage two-way primarily States, including industry that results in a 3% results in 9% more fixed effects; composed work as an Uber captures the reduction in average nonemployer synthetic of the driver, finding that expansion in Uber receipts, among establishments control) unincorporated a “10% increase and Lyft over counties with classified as self-employed, in the minimum this period—a low labor market transportation which includes wage resulted in 10% increase in concentration and and warehousing independent a 2.7% increase the minimum with Uber active.” services, among contractors and in the number of wage results in counties with participants in participants in the a 2.7% increase low labor market the online gig uncovered labor in the number of concentration and economy” market.” participants in the with Uber active.” uncovered labor A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: market.” 25 26 Study Country Digital Study type Intervention Study sample What is the regulation or Main findings Impact on Impact on Impact on reference platform type intervention studied? labor supply earnings occupation/ A review of the empirical evidence (probability job quality/ of working or formality hours worked) Gruber United Uber Survey n.a. 1,063 Uber drivers None - descriptive study. After accounting for the “tax n.a. n.a. n.a. 2022 States experiment (from outside Are workers willing to advantage of benefits, workers California) trade off additional income are roughly indifferent on for benefits? average between the two [additional income vs. benefits] The effects of regulating platform-based work on employment outcomes: (…) While there are some trends in valuation, such as higher valuation for pension than for health contributions, the most notable feature of the data is the wide variation across workers in their preferences across benefit types and relative to income. Workers also show a preference for benefits that can help them commit to increase savings in the future.” Guerrero Peru Cabify RCT Using 5,022 drivers Invitations to 5,022 Cabify Found that 18 percent of n.a. n.a. n.a. and Silva- platforms to affiliated with drivers in Peru to join one drivers signed up to one of Porto incentivize the Cabify Peru of two voluntary savings the two schemes after 8 2020 workers platform plans: an emergency weeks, with the emergency to enroll savings plan “in which savings plan having a higher in social drivers could save 2% of take-up (20 percent versus 16 protection their weekly earnings to percent). They also found that and insurance cover emergencies” and “after four months, the average a more flexible plan “that savings generated by drivers offered the driver the on the platform was USD 29.” option to save 3% of their weekly earnings each time they exceed a threshold, which the driver themself determined.” Study Country Digital Study type Intervention Study sample What is the regulation or Main findings Impact on Impact on earnings Impact on reference platform type intervention studied? labor supply occupation/ job (probability of quality/ formality working or hours worked) Hall, John, United Uber Theoretical Higher pay or Panel consisting Uber-initiated fare Increases in Uber’s n.a. “On the driver side, n.a. and Daniel States model different type of Uber rides increases base price in the with a higher base 2023 estimated of pay in 36 US cities United States—and fare, the driver’s using data over 138 weeks, payments for hourly earnings rate beginning with drivers—led to rises immediately the week of earnings increases as drivers make 2014-06-02 and for workers. more money per trip. ending with the However, these However, the hourly week of 2017- increases only earnings rate begins 01-16 lasted eight weeks, to decline shortly as the demand for thereafter. After about rides adjusted in 8 weeks, there is no response to the clear difference in the higher price. driver’s gross average hourly earnings rate compared to before the fare increase.” Holtz, Multiple Upwork Survey Regulating 520 Upwork Offering workers the The median n.a. n.a. n.a. Scult, and (HIC as experiment “reputation freelancers choice “between a 5-star freelancer valued Suri 2022 well as systems” rating with a positive a single positive LMIC) textual review, or a review at around monetary bonus that was US$50. randomly chosen from a set of values ranging from $25 USD to $175 USD.” A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 27 28 Study Country Digital Study type Intervention Study sample What is the regulation or Main findings Impact on labor Impact on Impact on reference platform type intervention studied? supply (probability earnings occupation/ job A review of the empirical evidence of working or hours quality/ formality worked) Horton, Not Not RCT Improving 50,877 Employers allocated to In terms of match Increase in hours Revelation n.a. Johari, specified specified transparency employers one of two treatment outcomes, the worked. This increase increased the and on work were allocated arms: ‘explicit’ arm, revelation of employer in hours worked even wage bill by Kircher compensation to the where jobseekers see preferences increased occurred among 2.7%. 2021 and quality experiment. employers’ pay scale total transaction employers selecting These and skill preferences, volume on the ‘high’ vertical The effects of regulating platform-based work on employment outcomes: employers or an ‘ambiguous’ arm, platform by about 3%. preferences who hired collectively where jobseekers still This increase came workers at higher posted 220,510 see this information but from an increase in wages. As employers job openings. employers do not know the quality of matches decide on hours whether the signal will be (but not the quantity), worked, this is strong revealed leading to larger evidence of match within-relationship quality improvements. expenditure and hours Pooled across tiers, worked. revelation increased hours worked by 4.6%, with increases of 2.9% in the high tier and 5% in the low tier. Horton Multiple Not RCT Introducing Around “Minimum hourly wages Wages of hired A decrease in hiring Each US$1 n.a. 2018 (HIC as specified a minimum 160,000 job were randomly imposed workers (who worked ranging from 2.5 increase in the well as compensation openings on firms posting job from the United States, percent to 10 percent, minimum wage LMIC) or wage: openings in an online India, the Philippines, depending on workers’ was associated labor market.” and Bangladesh) rose pervious earnings. with an increase at the expense of a In addition, “hours- in wages reduction in hiring worked fell sharply, ranging from 4% and hours worked. with reductions as to 9%. The reduction in large as 30% in some hours worked is in sub-populations of job part explained by how openings expected to employers began pay low wages.” hiring higher-skilled workers after the wage policy. Study Country Digital Study type Intervention Study sample What is the regulation Main findings Impact on labor Impact on Impact on reference platform type or intervention supply (probability earnings occupation/ job studied? of working or hours quality/ formality worked) Horton Not oDesk RCT The role of 6,209 job Algorithmically Employers with Increases in hiring: n.a. n.a. 2016 specified matching openings recommending technical job “employers with services and workers to employers vacancies that technical job intermediaries for the purpose of received recruiting vacancies that recruiting recommendations had received recruiting a 20 percent higher fill recommendations had rate compared to the a 20% higher fill rate control.” In addition, compared to they find “no evidence the control. “ that the treatment crowded-out hiring of non-recommended candidates.” Jackson United Various Quasi- n.a. “The universe Uses “US Finds “an increase in ‘High-gig-propensity’ “The 10.16 Overall record 2022 States experimental of individual administrative tax gig work following an individuals (that is, percentage a shift toward analysis income tax records to measure unemployment spell” Individuals similar to points increase gig employment (triple returns fled take-up of gig and that individuals those who had already in gig work as a result of differences- in the United employment following are “correspondingly been previously corresponds to a unemployment in- States from unemployment spells better able to employed in gig roughly US$588 differences) 2005–2017” and to evaluate the smooth the resulting work) become 10.16 increase in gig effect of working in drop in income.” percentage points earnings in the the gig economy on However, individuals more likely to enter gig year of UI receipt,” workers’ overall labor who entered gig work in the year of UI although there supply and earnings work following an receipt relative to two are decreasing trajectory” unemployment spell years prior, subject impacts on were more likely to gig work being earnings over the to “stay in these available in an area. long run positions and are less likely to return to traditional wage jobs.” A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 29 30 Study Country Digital Study type Intervention Study sample What is the Main findings Impact on labor Impact on Impact on reference platform type regulation or supply (probability earnings occupation/ job A review of the empirical evidence intervention of working or hours quality/ formality studied? worked) Jones and Mozambique Biscate, Event study n.a. A panel Effect of COVID-19 The COVID-19 Our most simple n.a. n.a. Manhique a digital constructed as an income pandemic “was specification 2019 platform in at the shock on workers’ associated with a shows the task Mozambique profession × decisions to net increase in tasks agreement rate for province level, engage in digital demanded per worker, increased by between contracting yielding 198 platform work. but no clear change approximately 20 and The effects of regulating platform-based work on employment outcomes: manual distinct units “Based on the in supply growth (new 40 percent in the first freelancers observed over universe of registrations). These year of the pandemic such as a maximum of records from a trends are evident (for example, from plumbers, 150 weeks matching platform across multiple market around 1.6 tasks per carpenters, for informal segments, including week per 100 workers and sector manual female-dominated to around 2.3 tasks). hairdressers freelancers in professions, Mozambique, suggesting digital we analyze how labour markets can task supply and help workers adjust to demand altered economic shocks in with the onset of low-income contexts.” COVID-19.” Kass 2022 United None Theoretical n.a. Respondents Estimating Finds “the recent The availability of n.a. n.a. States model of National whether the development of apps ‘contingent work’ estimated Longitudinal availability of (such as Upwork) lowers unemployment using data Survey of ‘contingent’ jobs, that make contingent (that is, lowers Youth 1979 such as online gig work easy to find the optimal UI (NLSY79), work, provides lowered the optimal UI replacement rate for which is a an alternative to replacement rate for traditional employees national panel unemployment traditional employees from 48 percent to 41 survey of from 48 percent to percent). the cohort of 41 percent, which individuals shows that contingent born in work provides the years valuable insurance 1957–1964 to all workers in the economy.” Study Country Digital Study type Intervention Study sample What is the Main findings Impact on labor Impact on Impact on reference platform type regulation or supply (probability earnings occupation/ job intervention of working or hours quality/ formality studied? worked) Kassi and Multiple “One of Event study Regulating 422,199 Giving workers We show that Impact on productivity An additional n.a. Lehdonvirta the largest “reputation freelancer in a web-based obtaining skill not statistically skills certificate 2019 online labour systems” projects digital platform a certificates increases significant leads to a 2.1% platforms” skill certificate freelancers’ earnings. increase in This effect is not earnings (average driven by increased gain of US$1.88). freelancer productivity Returns to but by decreased signaling were employer uncertainty. smaller for more The increase in experienced freelancer earnings freelancers: is mostly realized returns to through an increase completing skill in the value of the certificates are projects won rather up to 1.5 times than an increase in the larger for workers number of projects with zero work won. history compared to average. Kecht and Germany Various Quasi- n.a. 101,248 Expansion of gig Find “recently “We demonstrate n.a. n.a. Marcolin (Deliveroo, experimental individuals delivery platforms unemployed workers that, following a job 2022 Foodora and analysis from a are more likely to separation, individuals Lieferando) (differences- nationally take up gig work and are 11.6% more in- representative less likely to receive likely to be employed differences) sample, with unemployment in mini-jobs when a total of benefits (UB) when a gig platforms are 113,493 job gig delivery platform is available. Moreover, separation available to them.” recently unemployed events workers with access to these platforms are 8.19% less likely to receive traditional unemployment insurance and, overall, receive 3.44% lower benefits.” A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 31 32 Study Country Digital Study Intervention Study sample What is the regulation Main findings Impact on labor Impact on Impact on reference platform type type or intervention supply (probability earnings occupation/ A review of the empirical evidence studied? of working or hours job quality/ worked) formality Krishnan et United Lyft RCT Introducing Lyft drivers Two experiments to Found this feature Results from the first Find the Find a 60% al. 2022 States a minimum in “major, understand the impact can solve the issue experiment “show a introduction positive compensation oversupplied of Lyft’s Priority Mode of worker oversupply 5% increase in driving of priority satisfaction or wage: urban markets” feature on worker while safeguarding hours from this subset mode rating among in North oversupply: (1) a partial worker flexibility, treatment group, “resulted in drivers using America rollout of Priority Mode increasing gains for indicating strong a generation Priority Mode The effects of regulating platform-based work on employment outcomes: to a random subset drivers, riders, and Lyft driver preference for of system (that is, saying of drivers, and (2) a in the process. Priority Mode.” surplus it positively randomized rollout equivalent affected overall of modifications of to” between experience as a Priority Mode that 10 to 13 driver). varied how and when percent of drivers could activate total driver the mode earnings. Liang et al. United MTurk; RCT Monitoring Workers who Randomly assigning Workers are n.a. n.a. n.a. 2022 States Prolific workers have finished gig workers apprehensive about more than participating in tasks monitoring and 1,000 tasks on MTurk and Prolific that, on average, on MTurk to four groups with the compensations and have an different monitoring required for workers approval rate policies. Monitoring to accept monitoring higher than was varied along are between US$1.8 98% and are three dimensions: and US$1.6 an hour from the United “monitoring intensity (roughly 37.5 to 28.6 States (how much information percent of average is collected), hourly wages). transparency (whether the monitoring policy is disclosed to workers), and control (whether workers can remove sensitive information).” Study Country Digital Study type Intervention Study sample What is the regulation Main findings Impact on labor supply Impact on Impact on reference platform type or intervention (probability of working or earnings occupation/ studied? hours worked) job quality/ formality Lukac and Not Not Theoretical Regulating Scraped Investigated the “extent Reputation systems n.a. n.a. n.a. Grow 2021 specified specified model “reputation data on two to which reputation “can potentially estimated systems” datasets, systems can create reproduce inequalities using data containing segmented hiring present in offline labor 3,434 and patterns that are biased markets and produce 1,454 work toward freelancers with unfair outcomes that bids each good reputation.” disproportionately favor already successful applicants.” Nakamura Indonesia Not Differences- Introducing Drivers from a We study the market- We find that, on Minimum wage leads to a Minimum n.a. and Siregar specified in- a minimum “collaborating wide implications of average, the policy higher excess labor supply: wage does 2022 differences compensation platform” a federal policy on increases the trip statistically significant 24.3% not lead to and or wage We restrict minimum fares for price but does not increase in excess supply increased synthetic our sample to drivers on ride-sharing significantly affect the hours, that is, the sum of all earnings: control all completed apps. overall transaction idle hours statistically methods motorcycle volume nor increase on the app from all drivers. insignificant trips that driver earnings or Mainly driven by drivers that 1.7% had non-zero wages. These effects made less before the policy reduction payments to are driven by a higher change (the policy increases in daily associated excess labor supply, the total labor supply of earnings drivers. Our reducing the number workers from driver data set of transactions per in the bottom 3 to 4 deciles fare and a contains driver. The excess of pre-policy earnings by 20 statistically trips from 64 labor supply comes to 40%.) insignificant Indonesian from lower-earning Also lowers productivity, 6.7% cities, 55 of drivers but does not driven by two margins: an reduction which are in lead to their increased increased share of in wage, the data for earnings. less productive drivers in that is, daily our analysis. the workforce and reduced earnings The data cover individual productivity due to divided by the period of crowding on the supply side. supply hour. January 1 to There is an 8 to 10% reduction August 8, 2019. in average driver productivity due to the policy, statistically significant at the 10% level. A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 33 34 Study Country Digital Study type Intervention Study sample What is the Main findings Impact on labor supply Impact on Impact on reference platform type regulation or (probability of working or earnings occupation/ intervention hours worked) job quality/ studied? formality Pallais 2014 Multiple oDesk RCT Regulating 952 randomly Evaluated the effects “Both hiring workers Providing coarse Providing coarse n.a. A review of the empirical evidence (HIC as “reputation selected of hiring workers and providing more evaluations “almost evaluations well as systems” workers on and “revealing more detailed evaluations tripled the fraction of “almost tripled the LMIC) oDesk information about substantially improved [inexperienced] workers average earnings their abilities” by workers’ subsequent with any employment of [inexperienced] giving them “either employment from 12 percent to 30 workers from detailed or coarse outcomes.“ percent.” The coarse $10 to $27” and public evaluations.” evaluation treatment did also increased not significantly improve the wage these employment outcomes workers posted The effects of regulating platform-based work on employment outcomes: for experienced workers. on their profiles Providing detailed by approximately evaluations increased 10 percent. “the fraction of workers Providing detailed with any subsequent evaluations employment from 53 increased percent to 69 percent,” but experienced did not improve average workers’ average employment outcomes earnings from of inexperienced workers US$101 to US$187 relative to the coarse and average evaluation treatment. posted wags by 15 percent. Stanton and Multiple oDesk Theoretical The role of 16 months Studies the role of “Workers affiliated “Compared to non- Estimate that n.a. Thomas (HIC as model matching of data intermediaries, and with an agency have affiliates, agency “agency affiliation 2015 well as estimated services and from oDesk, workers’ affiliations substantially higher switchers are estimated to is associated with LMIC) using data intermediaries including with intermediary job-finding probabilities find jobs with” between 22 an 8.6-percent 1,126 separate agencies, on and wages at the and 35 percent fewer job increase in the agencies with their job-finding beginning of their applications. wage at which at least one probabilities and careers compared to an inexperienced new non-US wages similar workers without worker offers to affiliate active an agency affiliation. work.” This advantage declines after high- quality non-affiliated workers receive good public feedback scores” Study Country Digital Study type Intervention Study sample What is the regulation Main findings Impact on labor supply Impact on Impact on reference platform type or intervention (probability of working earnings occupation/ studied? or hours worked) job quality/ formality Stanton and Multiple Not Theoretical Introducing Data on Simulating impact of Find introducing a Find introducing a n.a. n.a. Thomas (HIC as specified model a minimum 169,578 policies that resemble minimum wage might minimum wage might 2021 well as estimated compensation jobs posted traditional employment lower overall hiring, lower overall hiring, LMIC) using data or wage by 67,292 regulation: introducing leaving employers and leaving employers and potential a 10 percent tax workers both worse workers both worse off. buyers paid by buyers when off. Find introducing a Find introducing a 10 between hiring jobseekers and 10 percent tax might percent tax might also January 2008 introducing a minimum also lower hiring by lower hiring by about and June 2010 wage about 26 percent 26 percent due to a due to a decrease in decrease in the jobs the jobs posted on posted on the platform the platform of 34 of 34 percent. percent. Stylianou United n.a. Theoretical Arbitrary n.a. Notes examples of Methods of exclusion n.a. n.a. n.a. 2018 States frameworks exclusion of large tech firms (such include vertical and case competitors as AT&T and Apple) integration (that is, studies excluding competitors harming competitors through several by blocking access methods to other markets in the supply chain) and blocking access to a competitor’s service (for example, blocking access to an app on an operating system, or making an app incompatible). A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 35 36 Study Country Digital Study type Intervention Study sample What is the Main findings Impact on labor supply Impact on Impact on reference platform type regulation or (probability of working earnings occupation/ job A review of the empirical evidence intervention studied? or hours worked) quality/ formality Van Not Not RCT Introducing 124,945 digital Digital platform Treated workers The number of hourly Workers who Treated workers Inwegen et specified specified a minimum platform workers were who historically jobs won by low- previously had “did not move to al 2022 compensation workers randomly assigned to charged below wage workers (that average hourly the uncovered or wage a control group or one the minimum is, those with average wages or wage sector - jobs with of three treatments wage reduced hourly wages or wage bids of less than a fixed price rather involving a minimum their probability bids below US$5) US$5 an hour than an hourly wage (one minimum of employment decreased by either saw their hourly wage- nor did they The effects of regulating platform-based work on employment outcomes: wage of US$2, one and increased 14, 18, or 32 percent wages increase direct their search of US$3, and one of their probability (for each of the three (by US$22 in the to better fitting US$4). of exiting the treatment groups). US$4 minimum jobs. They were platform; yet the Low-wage workers in wage treatment also more likely to treated workers the US$4 minimum group). However, exit the platform.” who stayed on wage group sent out 53 “workers total Specifically, low- the platform percent fewer hourly earnings did wage workers who and found a job applications than low- not follow suit.” received the US$4 increased their wage workers in the Workers who minimum wage earnings. control group. received a US$4 treatment were minimum wage 6% more likely to “experienced exit the platform an 8% decrease than workers in the in their total control group. earnings.” Wood- Multiple oDesk Theoretical Regulating 19,598 digital Estimates the effect “Reviews have n.a. “The wage “A one standard Doughty (HIC as model “reputation platform of online reviews as a relatively effect of reviews deviation increase 2018 well as estimated systems” workers well as “other sources small effect increases over in a worker’s LMIC) using data of information on both wages time, from only a combined review about worker and attrition, 3% increase for score reduces their ability, including the however, I am a one standard probability of exit review comments, able to separate deviation increase by 1%.” standardized exam out the dual in the first job, to scores, and the role of reviews: a 9% bonus in the worker’s country” ok rewarding good 20th.” worker wages and workers and attrition. punishing bad ones.” Study Country Digital Study Intervention Study sample What is the regulation Main findings Impact on labor Impact on earnings Impact on reference platform type type or intervention supply (probability occupation/ studied? of working or hours job quality/ worked) formality Yin, Suri, Not MTurk RCT Strengthening 399 MTurk Study how granting Find “workers Find “granting “We estimate the n.a. and Gray specified standards workers workers with more would give higher “in- compensating differential 2018 around job in-task flexibility up significant task flexibility” to be at least $0.86/hour, and task influences worker compensation to dramatically which means that on flexibility behavior as well as control their time, affected the average, workers equate work quantity and indicating workers temporal dynamics the ability to control quality attach substantial of worker behavior scheduling their work with value to in-task and produced a a financial compensation of flexibility” larger amount of at least $0.86/hour” work with similar quality.” A review of the empirical evidence The effects of regulating platform-based work on employment outcomes: 37 The effects of regulating platform-based work on employment outcomes: 38 A review of the empirical evidence References Agrawal, A. 2021. “Predatory Pricing and Platform Competition in India.” World Competition 44 (1): 109-120. Aleksynska, M., Bastrakova, and N. Kharchenko. 2019. “Working Conditions on Digital Labour Platforms: Evidence from a Leading Labour Supply Economy.” IZA Discussion Paper No. 12245. Alzate, D., E. Carranza, J. Duran-Franch, T. Packard, and C. Proffen. 2024. “How Regulations Matter for Labor Market Outcomes: A Joint Review of Product and Labor Market Regulations.” Working Paper. Anwar, A. M. and M. Graham. (2020). “Between a rock and a hard place: Freedom, flexibility, precarity and vulnerability in the gig economy in Africa”. Competition & Change 25(2). Argawal, A., N. Lacetera, and E. Lyons. 2016. “Does Standardized Information in Online Markets Disproportionately Benefit Job Applicants from Less Developed Countries?” Journal of International Economics (103): 1–12. Asadpour, A., I. Lobel, and G. van Ryzin. 2022. “Minimum Earnings Regulation and the Stability of Marketplaces.” Working Paper. Bamberger, A. K., and O. Lobel. 2017. “Platform Market Power.” Berkeley Technology Law Journal 1051. Beerepoot, N., and B. Lambgrets. 2014. “Competition in Online Job Marketplaces: Towards a Global Labour Market for Outsourcing Services?” Global Networks 15 (2): 236–255. Behrendt, C., A. N. Quynh, and U. Rani. 2019. “Social Protection Systems and the Future of Work: Ensuring Social Security for Digital Platform Workers.” International Social Security Review 72. Benson, A., A. Sojourner, and A. Umyarov. 2018. “Can Reputation Discipline the Gig Economy? Experimental Evidence from an Online Labor Market.” Opportunity and Inclusive Growth Institute Working Papers 16, Federal Reserve Bank of Minneapolis. Berg, J., M.Furrer, E. Harmon, U. Rani, and M. Six Silberman. 2018. Digital Labour Platforms and the Future of Work: Towards Decent Work in the Online World. Geneva, International Labour Office. Blair, P. Q., and M. Fischer. 2022. “Does Occupational Licensing Reduce the Effectiveness of Customer Search on Digital Platforms?” Working Paper. Bostoen, F. 2019. “Online Platforms and Pricing: Adapting Abuse of Dominance Assessments to the Economic Reality of Free Products.” Computer Law and Security Review (35) 3: 262–280. Boucherak, Farah. 2019. “Deliveroo Promises Sick Pay for Drivers.” France24, October 18, 2019. https://www. france24.com/en/business/20191018-deliveroo-promises-sick-pay-for-riders-1. Chen, K. M., J. Chevalier, P. Rossi, and E. Oehlsen. 2017. “The Value of Flexible Work: Evidence from Uber Drivers.” Working Paper. Datta, N., C. Rong, S. Singh, C. Stinshoff, N. Iacob, N. S. Nigatu, M. Nxumalo, and L. Klimaviciute. 2023. Working without Borders: The Promise and Peril of Online Gig Work. World Bank, Washington, DC.  Doerrenberg, P., D. Duncan, and M. Loffler. 2023. “Asymmetric Labor-Supply Responses to Wage Changes: Experimental Evidence from an Online Labor Market.” Labor Economics 81. Dube, A., J. Jacobs, S. Naidu, and S. Suri. 2020. “Monopsony in Online Labor Markets.” AER: Insights 2 (1): 33–46. Ducci, F. 2020. Natural Monopolies in Digital Platform Markets. Cambridge University Press, UK. Farronato, C., A. Fradkin, B. Larsen, and E. Brynjolfsson. 2020. “Consumer Protection in an Online World: An Analysis of Occupational Licensing.” Working Paper. The effects of regulating platform-based work on employment outcomes: A review of the empirical evidence 39 Garin, A., E. Jackson, D. Koustas, and A. Miller et al. 2023. “The Evolution of Platform Gig Work, 2012–2021.” Working Paper. Ghorpade, Y., and A. A. Rahman, and A. Jasmin. 2023. “Social Insurance for Gig Workers: Insights from a Discrete Choice Experiment in Malaysia.” World Bank Policy Research Working Paper 10629. Glasner, B. 2023. “The Minimum Wage, Self-Employment, and the Online Gig Economy.” Journal of Labor Economics (41) 1: 1–289. Graham, M., I. Hjorth, and V. Lehdonvirta. 2017. “Digital Labour and Development: Impacts of Global Digital Labour Platforms and the Gig Economy on Worker Livelihoods.” Transfer: European Review of Labour and Research 23 (2): 135–162. Gruber, J. 2022. “Designing Benefits for Platform Workers.” NBER Working Paper No. 29736. Guerrero and Silva-Porto 2020. Digital Savings for On-Demand Platform Workers. Retirement Savings Laboratory, Inter-American Development Bank. Hall, J. V., J. H. John, and T. K. Daniel. 2023. “Ride-Sharing Markets Re-Equilibrate.” Working Paper. Hatayama, M., and D. Swiztak. 2024. “Regulating Platform-Based Work in Low- and Middle-Income Countries: Insights from Recent Regulatory Approaches.” World Bank Policy Brief. Heeks, R. 2017. “Decent Work and the Digital Gig Economy: A Developing Country Perspective on Employment Impacts and Standards in Online Outsourcing, Crowdwork, Etc.” Development Informatics Working Paper no. 71, 2017. Hodor, M. 2022. “Gig Workers and Performance Pay: A Dynamic Equilibrium Analysis of an On-Demand Industry.” Working Paper. Holtz, D., L. Scult, and S. Suri. 2022. “How Much Do Platform Workers Value Reviews? An Experimental Method.” CHI Conference on Human Factors in Computing Systems (CHI ’22). https://doi.org/10.1145/3491102.3501900. Horton, J. J., R. Johari, and P. Kircher. 2021. “Cheap Talk Messages for Market Design: Theory and Evidence from a Labor Market with Directed Search.” Working Paper. Horton, J. J. 2016. “The Effects of Algorithmic Labor Market Recommendations: Evidence from a Field Experiment.” Working Paper. Horton, J. J. 2018. “Price Floors and Employer Preferences: Evidence from a Minimum Wage Experiment.” Working Paper. ILO (International Labour Organization). 2021. World Employment and Social Outlook 2021 (1). Jackson, E. 2022. “Availability of the Gig Economy and Long Run Labor Supply Effects for the Unemployed.” Working Paper. Jones, S., and I. Manhique. 2022. “Digital Labour Platforms as Shock Absorbers: Evidence from COVID-19.” WIDER Working Paper Series wp-2022-108. Kass, T. 2022. “The Value of Contingent Work.” Working Paper. Kassi, O., and V. Lehdonvirta. 2019. “Do Digital Skill Certificates Help New Workers Enter the Market? Evidence from an Online Labour Platform.” CESifo Working Paper No 7810. Kecht, V., and G. Marcolin. 2022. “Gig Work as Income Insurance during Unemployment: Evidence from Germany.” Working Paper. Kingsley, S. C., M. L. Gray, and S. Suri. 2015. “Accounting for Market Frictions and Power Asymmetries in Online Labor Markets.” Policy and Internet 7 (4). Krishnan, V., R. Iglesias, S. Martin, V. Pattabhiraman, S. Wang, and G. van Ryzin . 2022. “Solving the Ride- Sharing Productivity Paradox: Priority Dispatch and Optimal Priority Sets.” Working Paper. The effects of regulating platform-based work on employment outcomes: 40 A review of the empirical evidence Lehdonvirta, V., O. Kassi, I. Hjorth, H. Barnard, and M. Graham. 2018. “The Global Platform Economy: A New Offshoring Institution Enabling Emerging-Economy Microproviders.” Journal of Management 45 (2): 567–599. Lehdonvirta, V., I. Hjorth, H. Barnard, and M. Graham. 2021. “Global Earnings Disparities in Remote Platform Work: Liabilities of Origin?” In Work and Labour Relations in Global Platform Capitalism. Edward Elgar, Cheltenham, edited by J. Haidar, and M. Keune, 111–133. Liang, C., J. Peng, Y. Hong, and B. Gu . 2022. “The Hidden Costs and Benefits of Monitoring in the Gig Economy.” Information Systems Research 34 (1). Lukac, M., and A. Grow. 2021. “Reputation Systems and Recruitment in Online Labor Markets: Insights from an Agent-Based Model.” J Comput Soc Sc 4: 207–229. https://doi.org/10.1007/s42001-020-00072-x. McDonald, P., P. Williams, and R. Mayes. 2020. “Means of Control in the Organization of Digitally Intermediated Care Work.” Work, Employment and Society 35 (5): 872–890. https://doi.org/10.1177/0950017020969107 Nakamura, S., and R. Siregar. 2022. “Distributional and Productivity Implications of Regulating Casual Labor: Evidence from Ridesharing in Indonesia.” Working Paper. Pallais, A. 2014. “Inefficient Hiring in Entry-Level Labor Markets.” American Economic Review 104 (11): 3565– 3599. Rhani, U., and R. K. Dhir. 2020. “Platform Work and the COVID-19 Pandemic.” Indian J Labour Econ 63 (1): 163– 171. Stanton, C. T., and C. Thomas. 2015. “Landing the First Job: The Value of Intermediaries in Online Hiring.” Working Paper. Stanton, C. T., and C. Thomas. 2021. “Who Benefits from Online Gig Economy Platforms?” Working Paper. Stoterau, D. 2024. “The Economic Rationale to Regulate Platform-Based Work.” World Bank Policy Brief.. Stylianou, K. 2018. “Exclusion in Digital Markets.” Michigan Telecommunications and Law Review 24 (2). Van Inwegen, E., A. Filippas, and J. J. Horton. 2022. “The Experimental Imposition of a Worker Level Minimum Wage: Evidence from an Online Platform.” Working Paper. Wood, A. J., V. Lehdonvirta, and M. Graham. 2018. “Workers of the Internet Unite? Online Freelancer Organisation among Remote Gig Economy Workers in Six Asian and African Countries.” New Technology, Work and Employment 33 (2): 95–112. Woodcock, J., and M. Graham. 2020. The Gig Economy: A Critical Introduction. Cambridge: Polity Press. Wood-Doughty, A. 2018. “The Role of Reputation Systems in an Online Labor Market.” Working Paper. Yin, M., S. Suri, and M. L. Gray. 2018. Running Out of Time: The Impact and Value of Flexibility in On-Demand Crowdwork. Microsoft Research.