UNIVERSITY OF AMSTERDAM – Amsterdam Business School MSc BUSINESS ADMINISTRATION
Track: Leadership & Management Master Thesis
First Reader: Eloisa Federici Second Reader: Corine Boon
‘Gig Work’: Empowerment or Exploitation?
The effects of the Gig Economy on Work-Life Balance
Author: Bibianne Bielschowskij Student Number: 11406348
Date: 21 June 2018
Statement of Originality
This document is written by Bibianne Bielschowskij, who declares to take full responsibility for its contents.
I declare that the text and the work presented in his document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economic and Business is responsible solely for the supervision of completion of the work, not for the contents.
Acknowledgements
In this section I would like to express my gratitude towards Eloisa Federici. Her critical feedback and experienced guidance helped lift this thesis to a higher level. I would also like to thank my fellow students involved in this thesis project, that helped collect the data for this study.
Table of Contents
ABSTRACT... 4
INTRODUCTION ... 5
THEORETICAL BACKGROUND AND HYPOTHESES ... 9
GIG WORK AND WORK-ON-DEMAND ... 10
GIG WORK AND WORK-LIFE BALANCE ... 13
JOB STRESS AS A MEDIATOR... 17
EMPLOYABILITY AS A MODERATOR ... 20
METHODOLOGY ... 23
METHOD ... 23
SAMPLE AND PROCEDURE ... 24
MEASURES ... 27
RESULTS ... 29
DESCRIPTIVE STATISTICS... 29
PROCEDURE AND ASSUMPTIONS ... 32
REGRESSIONS ... 33 DISCUSSION ... 38 SUMMARY OF FINDINGS ... 38 PRACTICAL IMPLICATIONS ... 44 LIMITATIONS ... 45 FUTURE RESEARCH ... 46 CONCLUSION ... 48 REFERENCES ... 49 APPENDIX ... 59
1. EXPLORATORY FACTOR ANALYSES ... 59
2. INTERVIEWS ... 60
2.1. Interview Foodora Rider ... 60
2.2. Interview Uber Driver ... 72
Abstract
This study examines how gig working affects the gig workforce, specifically their work-life balance. Based on literature on the gig economy and empirical evidence on work-life balance, a negative relationship between hours per month spent gig working and work-life balance is hypothesized. The study proposes job stress as a mechanism to explain this negative
relationship. Additionally, based on the literature, it is hypothesized that gig workers with high self-rated external employability are less likely to experience job stress and, as a consequence, a work-life conflict due to their gig work activities. The research model is tested by means of path analyses, analyzing data collected among workers working through on-demand gig platforms in The Netherlands, United Kingdom and United States (N = 171). Contrary to the hypothesized, the results indicate hours per month spent gig working is weakly positively, rather than negatively, related to work-life balance. Evidence is not found for job stress as a mediator, but some evidence is found indicating that when working more through an on-demand platform, slightly lower levels of stress may be anticipated for highly employable gig workers. Practical implications and avenues for future research are discussed.
Key words: gig work, platform economy, on-demand platforms, work-life balance, job stress, self-rated external employability.
Introduction
“Before the Internet, it would be really difficult to find someone, sit them down for ten minutes and get them to work for you, and then fire them after those ten minutes. But with technology, you can actually find them, pay them the tiny amount of money, and then get rid
of them when you don’t need them anymore.”
—Lukas Biewald, CEO of CrowdFlower (Marvit, 2014 p.20)
“Gig working”, a relatively new employment form, is reflected by the exponential growth of online platforms and its current increased garnering of public attention (Huws, Spencer, Syrdal & Holts, 2017; Smith & Leberstein, 2015). Discussions surrounding the need for regulation and the denial by some platforms of their role as employers reflect just some of this recent attention. Indeed, a review of the existing literature on gig work shows that most research, although valuable, focusses on macro-level constructs, such as labor law and regulation (e.g. Corujo, 2017; Fabo, Karanovic & Dukova, 2017; Smith & Leberstein, 2015; Todolí-Signes, 2017). While the benefits of supplying on-demand work are an abundance for platforms, little empirical evidence exists that investigates how gig working affects its
workers; on-the-job and personally. The employment form is claimed to contrast permanent and traditional labor contracts, being characterized by more flexible labor relationships, and feeding traditional workers’ aspirations to become more independent and entrepreneurial (De Stefano, 2015; Eurofound, 2015; Manyika et al., 2016). However, what are the realities of gig work? “Is it a liberating new form of self-employment or a new form of exploitation?” (Huws et al., 2017 p.10)
In the technology-driven gig work context, workers are increasingly and intentionally bringing their work home (Huws et al., 2017). This finding is a main motivation for this study to test the effect of gig work, also defined as on-demand work, on workers’ work-life balance
(WLB). This study therefore applies existing organizational behavior theories to investigate how this boundary-crossing new form of work affects gig workers’ jobs and personal life (Deng & Joshi, 2013).
Our current understanding leads us to believe workers’ main motivation for a gig work career is due to job flexibility, considering workers’ freedom to determine their own schedule (Berg, 2016; De Groen & Maselli, 2016; Deng & Joshi, 2013; Huws et al., 2017). Previous research suggests job flexibility benefits WLB, because flexible work options are considered a resource of coping with job demands and ensure greater perceived control in the workplace (Hill, Hawkins, Ferris & Weitzman, 2001; Peters, den Dulk & van der Lippe, 2009; Valcour, Ollier-Malaterre, Matz-Costa, Pitt-Catsouphes & Brown, 2011). However, although
platforms position themselves as offering entrepreneurial possibilities and job flexibility, the reality of gig working may be quite different. This is illustrated by some platforms rewarding workers who make themselves available and penalize those who do not (Smith & Leberstein, 2015). Even when flexibility is granted, empirical evidence suggests that in the gig context this likely does not benefit, and may even harm WLB. For example, due to the unspecified start and end times of most gig work, workers are more likely to self-exploit through
overworking (Corujo, 2017; Deng & Joshi, 2016). Also, the central role mobile technologies play in this work context haven been proven to blur the boundary between work and home (Duxbury & Smart, 2011; Guest, 2002; Nam, 2014; Sarker, Xiao, Sarker & Ahuja, 2012). This study therefore expects that while in the normal working context significant proof exists that job flexibility leads to better WLB, in this gig working context the opposite may be true (Nam, 2014). Therefore, the first aim of this study is to propose, as a main effect, that the amount of time spent gig working is negatively related to worker WLB.
The study also aims to investigate the mechanisms that explain this negative relationship. Gig work is expected to go hand-in-hand with job stress, caused by a variety of stressors.
These include (1) job insecurity, most importantly caused by the limited contractual security and benefits made available to gig workers (Findlay & Thompson, 2017; Fleming, 2017; Graham et al., 2017b; Herbert, 2018; Purcell, Hogarth & Simm, 1999; Smith & Leberstein, 2015), (2) increased time- and performance pressures created by the technology embedded nature of gig working (Deng & Joshi, 2013; Duxbury & Smart, 2011; Guest, 2002; Nam, 2014; Sarker et al., 2012), and (3) lacking leadership relations caused by difficulty in
communicating with platform personnel (Graham et al., 2017b; Huws et al., 2017). As these stressors are, in turn, all expected to increase work-family conflict (Ahuja, Chudoba, Kacmar, McKnight, & George, 2007; Hill, Ferris & Märtinson, 2003; Morganson, Major, Oborn, Verive & Heelan, 2010; Neal, Chapman, Ingersoll-Dayton & Emlen, 1993), job stress is expected to provide explanatory power in the negative relationship between the amount of time spent gig working and worker WLB. Job stress is therefore proposed as a mediator. The second aim of this study is therefore to prove that the amount of time spent gig working is positively related to job stress, and job stress to, in turn, be negatively related to WLB.
The third aim of this study is to analyze under which conditions the hypothesized direct and indirect effects are stronger. Perceived self-rated external employability (SRE), an employee’s own perception of his or her possibility of finding a similar or better job in the external labor market (Van der Vaart, Linde, De Beer & Cockeran, 2015), is proven to influence whether and how someone is experiencing job stress (Berntson & Marklund, 2007; De Cuyper, Baillien & De Witte, 2009a; Mohr, 2000). Feeling highly externally employable may namely provide a feeling of being in control over one’s career, making it easier to cope with work-related uncertainty, and less likely to experience strain (Berntson & Marklund, 2007; Bussing, 1999; Silla, De Cuyper, Gracia, Peiró & De Witte, 2009). Because job
insecurity is a stressor that is anticipated to be inherent in gig work, job insecurity is expected to increase when more hours are worked through a platform (Findlay & Thompson, 2017;
Fleming, 2017; Graham et al., 2017b; Purcell et al., 1999; Smith & Leberstein, 2015). Based on the Job Demand Resources (JD-R) Model, these job demands, in the form of job
insecurity, that increase when time spent gig working increases, are expected to be buffered by SRE (Demerouti, Bakker, Nachreiner & Schaufeli, 2001; Silla et al., 2009), representing a worker’s resource, consequently reducing feelings of strain and stress (Berntson & Marklund, 2007; Bussing, 1999).Therefore, in addressing the third aim of the study, it proposes the positive relationship between the amount of time spent gig working and job stress is especially true when SRE is low. Furthermore, life satisfaction is proven to be negatively related to work-life conflict (Bonebright, Clay, & Ankenmann, 2000), and employability is shown to moderate the relationship between job insecurity and life satisfaction (Silla et al., 2009). The proposed interaction between hours spent gig working and SRE is therefore not only expected to reduce feelings of stress and strain, but also to, consequently, reduce feelings of a work-life conflict. In relation of the third aim of this study, it therefore also expects SRE to moderate the direct effect, proposing the anticipated negative relationship between the amount of time spent gig working and worker WLB to be stronger for low values of SRE.
To conclude, the study aims to answer the research question “how does a worker’s gig work involvement influence this worker’s work-life balance?”, and thereby investigate in which context job flexibility may have a ‘dark side’ for workers. It does so by testing the relationship between time spent gig working and WLB, proposing job stress as a mechanism to explain this relationship, and testing how SRE influences both direct and indirect effects. The proposed conceptual model is depicted in Figure 1.
Job Stress Hours/month spent gig working Work-life balance Self-rated external employability H1 (-) H2 H3a (-) H3b (-) H2a (+) H2b (-)
Figure 1. The conceptual model
Theoretical Background and Hypotheses
“A full-time job with one employer has been considered the norm for decades, but increasingly, this fails to capture how a large share of the workforce is making a living.” (Manyika et al., 2016, p.8) Employers and workers increasingly looking for (staff to) work outside the traditional boundaries of normative ‘nine to five’ working, as well as
advancements in technology, have paved the way for new forms of employment. Gig work is an employment form evolving rapidly, as digital platforms create large-scale transparent and efficient marketplaces connecting workers to users of end-products or services (Graham, Hjorth, Lehdonvirta, 2017a; Huws et al., 2017; Manyika et al., 2016). Gig work is
characterized by (1) a high degree of autonomy, (2) payment by task, assignment, or sales, and (3) a usually short-term relationship between worker and client (Manyika et al., 2016).
The attempts to estimate the extent of the growth of the gig economy have varied widely, due to the lack of an agreement on what gig work entails exactly. This is also because most research carried out to date on work in the gig economy focuses on more well-known platforms such as Amazon Mechanical Turk (AMT), Uber, Deliveroo and Helpling, and are
leaving many other, less well-known, platforms out of the research scope (Huws et al., 2017). It is also difficult to estimate the number of workers in the gig economy, as businesses are sometimes reluctant to disclose this information, and because workers may be registered and work with several companies at the same time (De Stefano, 2015). According to Huws et al. (2017) an average of 14% of people from the United Kingdom (UK), Sweden, Germany, Austria, The Netherlands (NL), Switzerland and Italy said to ever have sold their labor via an online platform. Additionally, it is estimated that 15 percent of independent workers in Europe and The United States (US) make or made use of these online marketplaces (Maryika et al., 2016). The rapid growth of especially some of the larger platforms suggests these numbers will strongly continue rising. Research also illustrates how most people from these countries obtain less than half their personal income from gig work and that more or less half partakes in gig work at least weekly. This is because gig work often is combined with at least one other form of income generation, and serves as a supplement to other earnings (Huws et al., 2017).
The following section further addresses what this study identifies as gig work and discusses recent developments in the gig context.
Gig work and work-on-demand
Due to the novelty of the phenomenon, there is a debate surrounding terminology; “sharing economy”, “collaborative economy”, “crowdwork” and “platform economy” are just some of the labels used to address the employment form (Huws et al., 2017; Kenney & Zysman, 2016). De Stefano (2015) identifies “crowdwork” and “work on-demand via apps” as two work forms that make up gig work (De Stefano, 2015; Smith & Leberstein, 2015). The first, crowdwork, is performed through online platforms such as AMT, and is based on a short-term task or project, rather than a continuous employment relationship (Eurofound, 2015;
Saxton, Oh & Kishore, 2013). This type of work is often characterized by small human computing tasks referred to as “microtasks” (De Stefano, 2015; Irani, 2015; Kittur, et al., 2013). Crowdwork thereby allows for mass scale production by a flexible, on-demand virtual workforce, connecting clients and workers on a global scale (Deng & Joshi, 2016; Eurofound, 2015; Saxton et al., 2013). “Work on-demand via apps”, instead, is a form of gig work where traditional working activities such as cleaning and transport are channeled through mobile apps (De Stefano, 2015). These on-demand platforms set quality standards of the service and select and manage the workforce, like Uber and Deliveroo (De Stefano, 2015; Rogers, 2015). Tasks can range from for example cleaning and home-repairs, to more specialized tasks like consultancy or clerical work (De Stefano, 2015).
Some major differences therefore exist among these two forms of gig work, that have shaped the primary focus of this research towards work-on-demand platforms. Although both are new phenomena, crowdwork has been around longer than work-on-demand platforms. On-demand work has only recently taken shape and has gained the attention of many scholars, policy makers and the media. Furthermore, crowdwork is executed online and therefore connects clients and workers that could operate anywhere in the world, while work-on-demand apps match online supply and demand for activities that are executed ‘offline’. Because of this, on-demand work must occur on a much more local scale than crowdwork, and usually requires interaction between the worker and the client. Lastly, as already indicated, work-on-demand platforms actively manage their workforce by recruiting and sometimes even training workers, while crowdwork platforms usually do not. These firms also usually set and manage quality standards (De Stefano, 2015). Its novelty, required interaction between client and worker, and the platforms’ relatively higher involvement in management of the workforce have shaped the scope of this research to on-demand work. Because the scope of this research still holds similarities with crowdwork, such as both
matching supply and demand through internet technologies and providing “humans-as-a-service”, this literature review also incorporates existing research on crowdwork (De Stefano, 2015; Irani & Silberman, 2013).
As mentioned, in the past few years much public attention is given to on-demand work, especially in The US and Europe. Some recent developments include some platforms, like ride-sharing platforms Uber and Lyft, in The US starting to offer secondary liability insurance. This is due to recent health and safety concerns and negative media attention (Forde et al., 2017). Recent news in Europe involves the European Court of Justice ruling the same on-demand taxi service platforms to be treated as taxi operators, enforcing stricter regulation and licensing (Bowcott, 2017). In May this year, Uber announced to start providing additional protection to its Uber drivers and UberEats couriers in Europe, providing insurance against sickness and injury and granting maternity and paternity payments (Topham & Butler, 2018). In the same month, food delivery platform Deliveroo decided to equip its riders in twelve (mainly European) countries with free accident insurance (Herbert, 2018). These moves by the platforms are expected to be fueled by pressures of the united private hire drivers branch of the gig economy union IWGB on Uber and criticism on the way Deliveroo drivers are currently treated as self-employed workers (Hermanides, 2017; Topham & Butler, 2018). Still, especially in Europe and The US, criticism on these on-demand platforms continues to exist, with platforms continuing to fight worker unions in court over claims of denying its workers national living wage and holiday and sick pay (Herbert, 2018; Topham & Butler, 2018). Focusing more closely on The NL, in June 2018 the largest Dutch labor union FNV sued Deliveroo, demanding the platform company to terminate their self-employment construction and to start employing their delivery drivers. The union claims the company continuously breaks collective agreements and that there is a case of false independence (Pelgrim, 2018). These recent developments give a further
understanding of the context of the scope of this research, that focuses especially on The US and Europe. It highlights the existing turbulence of the gig economy, with major policy changes, platform management decisions and media slander occurring on a monthly to weekly basis. This turbulence is important to take into account when investigating the employment form.
Gig work and work-life balance
While benefits for businesses providing gig work are widely known, the benefits for its workers are not. On the one hand, platforms enjoy large pools of workers and thereby customers, possess the ability to harness real-time data to reach these customers, and have very limited obligations towards ‘their’ gig workers. The latter lowers firms’ risks and increases their flexibility (De Stefano, 2015; Manyika et al., 2016; Smith & Leberstein, 2015). On the other hand, the implications of gig working for workers is fairly unexplored. Business publications as well as descriptive studies highlight the opportunities this new employment form creates, and the risks it brings. Yet, to date almost no empirical evidence exists about how the gig economy, and especially on-demand work, affects its workers. Generally, working conditions and much else about platform workers’ lives remains under-researched (Forde et al., 2017). More specifically, as gig workers are constantly and
intentionally bringing work home (Huws et al., 2017), it is extremely relevant to investigate the effect of gig working by measuring its relationship with workers’ WLB.
Greenhaus, Collins and Shaw (2003, p.513) defined the WLB construct as “the extent to which an individual is equally engaged in – and equally satisfied with – his or her work role and family role.” They state that WLB consists of (1) time balance, meaning devoting an equal amount of time to work and family roles, (2) involvement balance, having an equal level of psychological involvement in both roles, and (3) satisfaction balance, indicating an
equal satisfaction with work and family roles (Greenhaus et al., 2003). As a construct, WLB involves work- and family related variables, such as working time, flexibility,
(un)employment, welfare, social security, family, fertility, migration, demographic changes, consumption and leisure time (Noor, 2003; Pichler, 2009). One is said to have a WLB when having satisfaction and good functioning at work and at home, with a minimum role conflict (Clark, 2000). When pressures from these work- and family domains are mutually
incompatible, an interrole conflict is created that is referred to as work-life conflict (Greenhaus & Beutell, 1985). Greenhaus & Beutell (1985) propose three forms of work-family conflict, including (1) time-based conflict, relating to a person’s psychological or physical impossibility to meet the demands of one’s role in one domain due to the demands of the role in the other, (2) strain-based conflict, where role-produced strain in the one domain affects the performance in the other, and (3) behavior-based conflict, referring to patterns of in-role behavior being incompatible with expectations regarding another role (Greenhaus & Beutell, 1985; Peters et al., 2009).
To comprehend the WLB construct, it is important to understand its outcomes. When a balance is experienced in work- and family roles, this is proven to lead to satisfaction in both roles (Clark, 2000; Greenhaus et al., 2003). Work-life conflict, on the other hand, is found to be related to negative outcomes such as job dissatisfaction, turnover intentions and harmed emotional and psychological well-being (Burke, 1988; Frone, Russell & Cooper, 1992; Kinman & Jones, 2008; Peters et al., 2009).
Strong supporters of gig work may argue that it benefits WLB, considering workers’ freedom to determine their own schedule, allowing control over their time invested in work as well as non-work activities. In line with this, research by Deng & Joshi (2013) suggests that job flexibility and job autonomy are major factors associated with the favorable
Groen and Maselli (2016) claim the main motivation for a crowd employment career is the combination of schedule flexibility and personal control in the form of selecting jobs, negotiating rates and having the possibility to work from any place they choose (Eurofound, 2017). Gig working is also claimed to feed traditional workers’ aspirations to become more independent and entrepreneurial (Manyika et al., 2016). Various empirical evidence indeed indicates perceived job flexibility positively affects WLB(Hill et al., 2001; Peters et al., 2009; Valcour et al., 2011). Research by Hill et al. (2001) for example shows that individuals with perceived job flexibility, given the same workload, have more favorable work-family balance. Also, Valcour et al. (2011) highlight how the fit between employees’ needs and the flexible work options made available to them are considered one of the resources positively associated with perceptions of organizational work-life support.
However, despite platforms positioning themselves as offering job flexibility and entrepreneurial possibilities, the reality can sometimes be quite different. For some on-demand platforms, no flexibility or autonomy exists, even if the work schedule is flexible in theory, as working hours are demand-dependent (Smith & Leberstein, 2015). “Many
companies reward those who make themselves available and penalize those who are not.” (Smith & Leberstein, 2015, p.6) Court documents for example even show how Uber remains the right to terminate drivers who’s “dispatch acceptance rate” is ‘too low’, and will
deactivate accounts when there are too many drivers or when business is low (Huws et al., 2017; Smith & Leberstein, 2015). Uber also uses predictive scheduling, indicating there is a high demand in the area where drivers are located, to motivate drivers to continue working at peak times (Rosenblat & Stark, 2015).
Even when gig workers are granted flexibility in their jobs, empirical evidence suggests a flipside to this flexibility: in return, gig workers must be available on-demand, and after being hired for a specific task only, they are dismissed (Eurofound, 2017; Schmidt, 2017). This is
especially the case for crowdwork. Gig workers often are required to work at short notice and at times with tight deadlines. On-demand workers specifically, performing their tasks offline, are often pressured to complete fixed-fee jobs and rapidly move on to the next to generate a living (Eurofound, 2017). These findings make us suspect the gig context may likely not benefit, but even harm WLB. Firstly, due to unspecified start and end times for work, gig workers are more likely to self-exploit through overworking. This often occurs while striving to guarantee a decent income (Corujo, 2017; Deng & Joshi, 2016). In line with this, those who are more flexible in work-at-life, are more willing to allow for the intrusion of work into life (Nam, 2014). Gig workers may have difficulties in focusing either on work or on their free time, since there is a pressure to be readily available for any potential ‘gigs’ (Eurofound, 2017; Huws, 2016). Secondly, an important contributor making on-demand work more flexible is the central role of mobile technologies. A wide variety of research highlights how technology however blurs the boundary between work and home (Duxbury & Smart, 2011; Guest, 2002; Nam, 2014; Sarker et al., 2012). Research by Hill, Miller, Weiner and Colihan (1998) proposes that the virtual office increases perceptions of productivity and flexibility, while however making work hours longer. Indeed, recent research on gig work indicates its workers bring their work home by being in contact with employer and clients whilst at home almost twice as much as non-gig workers (Huws et al., 2017). Clearly, this blurred work-home boundary and increasing work hours negatively affects WLB.
Therefore, the existing evidence makes it debatable whether platforms enhance feelings of WLB, or might sooner reduce them. As is proposed by Eurofound (2017), this influence may be dependent on the degree the gig worker is reliant on the income gained through gig work. While in the traditional working context significant proof exists that job flexibility leads to increased WLB, in this gig work context the opposite may hence be true. Based on the
evidence illustrating how gig work activities may blur the boundary between professional work and private life, this study proposes:
Hypothesis 1: Hours per month spent gig working is negatively related to worker work-life balance.
Job stress as a mediator
This study also aims to investigate the mechanisms that explain the expected negative relation between gig working and WLB. Much empirical evidence relates stress to work-life imbalance. Job stressors, and hence job stress, is considered an important antecedent of work-life conflict by most accredited literature (e.g. Aryee, 1992; Bell, Rajendran, & Theiler, 2012; Burke, 1988; Byron, 2005; Frone et al., 1992; Guest, 2002; Mauno & Kinnunen, 1999; Michel, Kotrba, Mitchelson, Clark & Baltes, 2011; Zedeck, 1992). Parker and DeCotiis (1983) describe job stress, being a feeling of discomfort, as a proximal outcome of the job and the organization. In turn, job stress has distal outcomes, which are further removed and reflect cumulative experiences over time, including varying levels of satisfaction,
organizational commitment, motivation, avoidance behavior and job performance (Barnett, Gareis & Brennan, 1999; Parker & DeCotiis, 1983). Sources of job stress are grouped into six categories, being (1) characteristics of the job itself, (2) organizational characteristics such as structure and climate, (3) job role factors, (4) work relationships, (5) perceived career
development and (6) external commitments and responsibilities. Within these six categories, stressors may range from perceived limitations on the relationship between performance and pay to the (lack of) concern shown for individuals in the organization (Parker & DeCotiis, 1983). Job insecurity, time pressures, bad leadership relations and a lack of role clarity and job autonomy are some of the most frequently mentioned sources of stress (Deery & Jago, 2009; De Cuyper, Notelaers & De Witte, 2009b).
Existing research gives an indication gig work may go hand-in-hand with various sources of stress. Firstly, it is evident that the nature of non-standard work is less contractually secure, offering less to no access to paid leave, career development, training and legal protection from dismissals or occupational hazards. This insecurity partly arises because both parties have the freedom to end the contract at any time (Findlay & Thompson, 2017; Fleming, 2017; Graham et al., 2017b; Smith & Leberstein, 2015). It is also proven that those working from the virtual- or home office, as many gig workers do, generally have less benefits, job security, WLB support and workplace inclusion than their counterparts working at the main office (Hill et al., 2003). Much of recent media attention on the platform economy focuses on platforms neglecting their duties in providing its workers national living wage and holiday and sick pay (e.g. Herbert, 2018; Topham & Butler, 2018). Also, contract flexibility, that is discussed in the previous section, is linked to greater job insecurity and poor conditions of employment (Purcell et al., 1999). Increased schedule flexibility may also result in self-exploitation, leading to stress and burnout (Corujo, 2017; Deng & Joshi, 2016; Nam, 2014). Therefore, while flexibility and personal control are indicated to be a major motivation to partake in gig work (Berg, 2016; De Groen & Maselli, 2016; Deng & Joshi, 2013), “for some crowd workers these specific elements cause stress due to the need for self-organisation and the blurring of work and private life.” (Eurofound, 2015, p.115) This lack of support, benefits and employment security, make job insecurity a likely common source of stress for those active in the gig economy. This may constrain workers’ ability to make long-term life decisions, also resulting in anxiety (Chan & Tweedie, 2015). Due to much recent media attention to this lack of security, some on-demand platforms have recently decided to take certain measures. Will Shu, founder of Deliveroo, said in May 2018 with regards to
equipping riders with accident insurance: “We know riders value the flexibility of being able to fit their work around their life, but they also deserve security if they’re involved in an
accident.” (Herbert, 2018) It is however debated whether such recent measures, similarly taken by Uber, are a mere cry for positive publicity or actually succeed in making workers feel more secure.
Secondly, Guest (2002, p.257) mentions that “factors such as the advances in information technology and information load, the need for speed of response, the importance attached to quality of customer service and its implications for constant availability and the pace of change with its resultant upheavals and adjustments all demand our time and can be sources of pressure.” All mentioned factors are integral aspects of gig working via on-demand platforms. Its technological nature allows platforms to monitor response rates, availability, and manage quality standards, often based on customer ratings (De Stefano, 2015). In the on-demand work context, there generally is the need to respond swiftly when a task is received (Huws et al., 2017), and arbitrary decisions may be made based on for example cancellation rates, acceptance rates or client ratings (Rosenblat & Stark, 2015). Research also suggests a substantial amount of gig workers habitually work for very long periods, according to when work is provided (Wood, Graham & Lehdonvirta, 2017). Because of this need for a quick response and long working hours, gig work is expected to go paired with time- and performance pressures, that are considered job stressors (Deng & Joshi, 2013; Duxbury & Smart, 2011; Nam, 2014; Sarker et al., 2012).
Finally, another anticipated source of stress and anxiety includes the difficulty in communicating with platform personnel (Huws et al., 2017; Lewchuk, 2017). Research by Huws et al. (2017) indicates that for some gig workers it is very difficult to get in touch with platform personnel by phone or face to face, and that questions or problems are often not dealt with accurately or at all. Another aspect of the poor communication between platform and personnel are unilateral arbitrary deactivations, whereby on-demand workers may suddenly find themselves unable to register for work, often with no explanation or warning
(Huws et al., 2017). Also, as mentioned, a variety of on-demand platforms use user ratings and customer complaints as a management tool, which may create a sense of unfairness and the perception that the platform generally takes the side of the customer rather than of workers. These elements, that may be inherent in some on-demand work, are a source of stress in the form of lacking leadership relations.
Therefore, the more a worker partakes in gig work, the more stressors, in the form of job insecurity, time- and performance pressures and lacking leadership relations, are anticipated to be experienced by this worker. These mentioned stressors consequently are expected to increase work-family conflict (Ahuja, et al., 2007; Hill et al., 2003; Morganson et al., 2010; Neal et al., 1993), making this research propose:
Hypothesis 2: The negative relationship between hours per month spent gig working and worker work-life balance is mediated by job stress.
Hypothesis 2a: There is a positive relationship between hours per month spent gig working and job stress.
Hypothesis 2b: There is a negative relationship between job stress and worker work-life balance.
Employability as a moderator
Other than investigating the relationship between gig working and worker WLB and constructs that explain this relationship, this study also aims to analyze under which
conditions the hypothesized direct and indirect effects are stronger or weaker. Relating to the mediation effect, research indicates that whether and how an employee is experiencing job stress can be influenced by this workers’ employability (Berntson & Marklund, 2007; De Cuyper et al., 2009b; Mohr, 2000).
Perceived employability is an employee’s own perception of the possibility of finding a similar or better job with the current employer or another organization (Berntson &
Marklund, 2007; De Cuyper, Van der Heijden & De Witte, 2011; Van der Vaart et al., 2015). Fugate, Kinicki and Ashforth (2004) propose employability is work-specific adaptability, consisting of three dimensions: career identity; personal adaptability and social and human capital. This research focusses on self-rated rather than other-rated employability, as workers likely act upon own perceptions of available employment opportunities rather than any other clues (De Cuyper & De Witte, 2011; Katz & Kahn, 1978). Other than self-rated or other-rated, there is a distinction between employability in the internal and external labor market. Internal employability relates to the current employer’s provision of career prospects, job enrichment, and investment in for example training programs. Central to external
employability, conversely, is the idea that the ‘deal’ with the present employer can be replicated elsewhere (De Cuyper & De Witte, 2011; Ng & Feldman, 2008; Rothwell & Arnold, 2007). This research focusses on external employability, as in the gig working context, one can usually expect very limited internal training and career progression opportunities (Findlay & Thompson, 2017), and therefore overall internal employability is probably low. Also, it is especially self-rated external employability that may provide a feeling of being in control over one’s career, as workers are aware of other employment alternatives from which to choose (De Cuyper et al., 2009b; Fugate et al., 2004).
Self-rated employability is indicted to be positively related to career satisfaction and perceived marketability, proving employability to be a predictor of career success (De Vos, De Hauw & Van der Heijden, 2011). Also, high perceived employability is associated with lower levels of job exhaustion and psychological symptoms and high self-rated job
performance and life satisfaction (De Cuyper et al., 2011; Kinnunen, Mäkikangas, Mauno, Siponen & Nätti, 2011). Research indicates that workers who feel that they have alternative
employment opportunities may find it easier to cope with uncertainty, and are less likely to perceive their jobs as threatening (De Cuyper et al., 2009a; De Cuyper, Bernhard-Oettel, Berntson, De Witte & Alarco, 2008; Silla et al., 2009). Alternatives in the labor market are therefore mobilized as a coping mechanism at insecure workplaces (Bussing, 1999). Silla et al. (2009) highlighted how employability moderates the relationship between job insecurity and life satisfaction. Specifically, the relationship between job insecurity and well-being is less negative when employees perceived a wide variety of outside employment opportunities (Silla et al., 2009). This interaction between job insecurity and employability may be
explained by means of the JD-R Model, proposing resources buffer unfavorable effects of job demands (Demerouti et al., 2001). As resources promote control, development and personal growth, a worker that perceives him- or herself as employable, likely perceives to dispose of resources. As is mentioned in the previous section, the more a worker partakes in gig work, the more job stressors in the form of job insecurity this worker is anticipated to experience (Findlay & Thompson, 2017; Fleming, 2017; Graham et al., 2017b; Purcell et al., 1999; Smith & Leberstein, 2015). This study hence expects job insecurity to increase with more hours worked through a platform. This job insecurity is regarded as a job demand, as it is an aspect of the job that creates a burden (Demerouti et al., 2001; Silla et al., 2009). Based on the JD-R Model, these job demands, in the form of job insecurity, that are inherent to gig work, are expected to be buffered by SRE, representing a worker’s resource. Hence, the interaction of hours spent gig working and SRE is, like in Silla et al.’s (2009) study, expected to buffer the negative consequences of feelings of insecurity that go paired with gig work activities, consequently reducing feelings of strain and stress (Berntson & Marklund, 2007; Bussing, 1999). This research therefore proposes:
Hypothesis 3a: The positive relationship between hours per month spent gig working and job stress is moderated by self-rated external employability, so that this relationship is stronger for lower values of self-rated external employability.
Furthermore, life satisfaction has been proven to be negatively related to work-life conflict (e.g. Bonebright et al., 2000), and, as indicated, employability is shown to moderate the relationship between job insecurity and life satisfaction (Silla et al., 2009). The earlier proposed interaction between hours spent gig working and SRE is therefore not only
expected to reduce feelings of stress and strain (Berntson & Marklund, 2007; Bussing, 1999), but also to, in turn, reduce feelings of a work-life conflict. SRE is therefore also expected to buffer negative consequences of job demands in the form of job insecurity, caused by gig work activities, for a workers’ WLB:
Hypothesis 3b: The negative relationship between hours per month spent gig working and worker work-life balance is moderated by self-rated external employability, so that this relationship is stronger for lower values of self-rated external employability.
Methodology Method
This research, that is exploratory by nature, tests the previously stated hypotheses by means of a cross-sectional and correlational quantitative study using surveys. This method makes it feasible to gather as complete as possible insights from the difficult to reach population. Three interviews have also been performed to gather additional in-depth qualitative insights. These insights are discussed in the discussion section.
Sample and procedure
The population of this study is on-demand workers, a specific type of gig worker, active in The NL, UK and US. Initially, the focus of this study was on The NL, but during the data collection it became evident that the population is difficult to reach, directly, as well as through platforms. Therefore, the questionnaire was not only administered digitally and by means of hard copies by the four students in this thesis project, but the services of experience management company Qualtrics were also used to gather additional respondents. Upon requesting these services of Qualtrics, the scope was expanded to The UK and US, as these are native English-speaking Western countries, that are at least as advanced in on-demand work as The NL.
Despite overall similarities between the state of the gig economy in these three countries, this study further defines the context by discussing this more in depth. Due to local economic conditions, institutional arrangements and regulatory mechanisms, it is very difficult to make comparisons between The US and Europe regarding developments of the gig economy to date. Especially in The US, considerable variation exists among different states and regions within the country, with certain tech ‘hot spots’ like San Francisco and New York for example being a lot more developed compared to other cities and states. Generally, discussions surrounding benefits and disadvantages of work in the gig economy are very much aligned in The US and Europe (Forde et al., 2017). With regards to social protection of gig workers, The NL enforces separate protection schemes for the self-employed, while in The UK there are more limited social protection rights (Eurofound, 2015; Forde et al., 2017). In The US, calls have been made to create a hybrid worker category due to concerns of the contractual status of platform workers (Forde et al., 2017). In terms of pay, in The UK the median pay for gig work is 47% lower than the national hourly minimum wage, which is the lowest in Europe after France. In The US, pay levels are closest to the national minimum
wage, however, it should be noted that in minimum pay levels vary across US states. Here, litigation has emerged regarding unpaid wages and expenses. Based on this information it can be concluded that the decisions of policy makers rather than technological advancements are likely to drive change in the platform economy (Forde et al., 2017). An example of this are taxi service platforms dealing with the European Court of Justice enforcing stricter regulation and licensing within The EU, including The UK (Bowcott, 2017). Nevertheless, perhaps the most important differences in developments in the gig economy exist between platforms rather than between countries or regions (Findlay & Thompson, 2017; Forde et al., 2017; Kalleberg & Dunn, 2016). To illustrate, due to health and safety concerns and negative media attention, some platforms such as Uber and Lyft have begun to offer secondary liability insurance (Forde et al., 2017). These are decisions that are made on platform level.
Because not all cases in the population were known, no sampling frame could be
identified. Respondents were gathered first of all by contacting on-demand platforms in The NL to distribute the questionnaire internally. In total, 18 on-demand platforms were
contacted, by email and telephone, of which only Temper, an on-demand platform providing hospitality jobs, agreed to participate. Some support was also provided by Foodora. Most platforms that declined the request did so because they were already performing an internal research of their own, were concerned for the privacy or peace of their on-demand workers, or because they were still very young and did not yet have a broad network. On-demand workers were also approached directly, especially in areas where many were expected to be active, such as at delivery drivers’ meeting points. Additionally, Qualtrics approached their network of panels consisting of self-employed people. Clear instructions were provided to Qualtrics regarding the scope of the research and examples of platforms in this scope were discussed. To ensure quality in the responses gathered by Qualtrics, three quality check items in the form of instructed responses were included. Overall, a non-probability convenience
sample was therefore used in this study. Workers were also asked to distribute the
questionnaire in WhatsApp groups with colleagues, representing volunteer sampling in the form of self-selection. Strict confidentiality was ensured in the introduction to the survey, and, when applicable, in accompanying e-mails.
Perceived relevance for participants in this research was expected to be high, as the perspective of gig workers has received very limited attention in past research on the gig economy. As high perceived relevance is one of the best incentives of participation, this relevance was stressed in the survey introduction and accompanying emails. Also, the chance of winning a voucher worth €50,- served as an additional incentive for completing the
questionnaire. This incentive was not offered to the respondents collected through Qualtrics. Platforms, on the other hand, could receive the research results in turn for their distribution of the questionnaire.
A total of 280 responses were collected, which were later reduced to 181 completed responses. It is very difficult to estimate the full reach of the questionnaire, and hence
determine the response rate, as the questionnaire was distributed online in various WhatsApp groups, via Qualtrics’ panels, and so on. The cleaning of the responses was done based on partial responses due to answering ‘no’ to the first check question whether the respondent is working through an on-demand platform, being 36 cases. Cases were also deleted due to completing less than half (until 44%) of the questionnaire (57 cases), completing the survey in under five minutes (1 case), and because the platform the respondent worked through most did not fall within the research scope, such as crowdsourcing platforms (5 cases, including Rev.com, Textbroker, Appen, Bookscouter and Twitch). Missing values were coded 999, and four typos (e.g. one indicating to work 125 years through the platform) were identified by means of a frequency analysis, and were treated as missing data.
Of the 181 responses, 65 were collected by the four students, and 116 by Qualtrics. The average age of the respondents was 32.6 years and 63.5% was male, meaning 36.5% was female. 35.9% executed their on-demand work in The NL, 47% in The US and 17.1% in The UK. For 69.1% of the respondents, on-demand work is their primary source of income. This is substantially more than most descriptive research on the gig economy (e.g. Huws et al., 2017). This is expected to be the case due to these studies often including crowd workers. The “microtasks” in crowdwork are expected to serve as a supplement to other earnings more often than on-demand work activities (Huws et al., 2017). With regards to the on-demand platform respondents work through most, 44.2% indicated to work most through Uber, 24.9% through Foodora, 13.3% Deliveroo, 8.3% UberEats, 2.8% Temper, 1.1% Helpling, and 5.5% indicated to work most through another on-demand platform. On average, respondents work 3.1 years (SD=4.80) through this platform. Their average work hours through a platform per week is 24 hours (SD=13.69), and 32.8 hours (SD=18.27) per week in total (i.e. considering their work through a platform as well as any other paid work activities).
Measures
The questionnaire consisted of a wide variety of survey items, including descriptive measures and items that were relevant for others involved in this thesis project. The following measures are used to test the model’s independent variable (hours per month spent gig working),
dependent variable (work-life balance), mediating variable (job stress) and moderating variable (self-rated external employability). Unless otherwise indicated, on-demand workers responded to existing and validated survey items using a five-point Likert scale (strongly disagree – strongly agree). The survey was administered in Dutch and English, using the translation-back-translation to translate the items (Brislin, 1980).
Hours per month spent gig working (HrsGig) was assessed by the question “How many hours (on average) do you spend per month working through (a) platform(s)?”. At the start of the survey, a brief definition was given of a platform. This definition was “an online
environment (website or app) through which a worker is linked to a client to perform tasks/services (e.g. Uber, UberEats, Foodora, Deliveroo, Temper, etc.)”.
Work-life balance was measured using Hill, Hawkins, Ferris & Weitzman’s (2001) five-item measure, including “I have sufficient time away from my job(s) to maintain adequate work and personal/family life balance”. Scales were adapted to range from ‘not at all’ to ‘a great deal’. Based on a performed reliability test and exploratory factor analysis (EFA) one reverse-coded item was removed, of which the corrected item-total correlation was .215, which indicates the item does not have a strong correlation with the total score of the scale. In the EFA, presented in Appendix 1, this item also did not load (.013) onto the same factor as the other four items. After removal of this item, the Cronbach’s alpha of this construct was .83. Most literature adopts 0.7 as a threshold for an acceptable reliability coefficient
(Nunnally, 1978), so this alpha indicates internal consistency among the four remaining items.
Job stress was measured based on the four-item measure of Motowidlo, Packard and Manning (1986). An example item is “I feel a great deal of stress because of my (total) work.” The reliability analysis revealed an increase of almost .15 of the Cronbach’s alpha when deleting the second item of the construct, which is one of the two reversed items. The corrected item-total correlation of this item was .029, and this item loaded -.120 on the factor the other items loaded onto. After removal of this second item, the remaining three items had a Cronbach’s alpha of .70, and all three loaded onto the same factor (>.55) based on principal component analysis, as shown in Table A.2 in the Appendix.
Self-rated external employability (SRE) was assessed by the five-item measure by Berntson and Marklund (2007). The measure includes items related to the respondents’ perceived skills, experience, network, personal traits, and knowledge of the labor market, e.g.: “My experience is in demand on the labor market”. The reliability test displayed a high Cronbach’s alpha and the corrected item-total correlation of all five items to be >0.3. The EFA revealed that all five items loaded onto the same factor (>.65)(Appendix 1). Therefore, all five items were included, with a Cronbach’s alpha of .84.
Control variables. It was decided to control for age (in years), gender, type of contract, payment form, gig work location, gig platform and source of income in the analyses. The first two were controlled for because demographic variables, especially age and gender, are often used in WLB research (e.g. Hill et al., 2001, Peters et al., 2009), as demographic
characteristics can be assumed to affect perceived work-home interference (Peters et al., 2009). The other adopted control variables are expected to be of relevance in the gig working context.
Results
SPSS Statistics version 24 was used for statistical analysis. Also, PROCESS, a regression path analysis modeling tool developed by Hayes, was used within SPSS to test the complete model presented in this study.
Descriptive Statistics
A correlation matrix (N = 171) is presented below in Table 1, displaying means, standard deviations, Cronbach’s alpha’s, and correlations between all variables used in the study. Despite the hypotheses of this thesis, no significant correlation is found between HrsGig and job stress. HrsGig and WLB are correlated (r = .23, p <.01), however, contrary to the
theoretical background, positively rather than negatively. The correlation matrix also shows that SRE is weak to moderately positively correlated with HrsGig (r = .17, p <.05), and WLB (r = .35, p <.01). As anticipated, WLB and job stress are negatively related (r = -.49, p <.01).
Based on the correlation matrix, control variables were selected. As gender and age were not related to the outcome variables of the model, these were excluded. Furthermore, type of contract, payment form, gig work location, type of platform (specifically Uber) and source of income are used as control variables in the analysis of the model due to various correlations with especially the outcome variables of this study.
Finally, normality was assessed for the variables in the model by means of testing for skewness and kurtosis, as displayed in Table 2. Although slight skewness is detected for some of the variables, the detected amounts are controllable in the regression by PROCESS, that overcomes non-normality by bootstrapping.
T abl e 1: M ea ns , S ta nda rd D evi at ions , C o rr e la ti ons . V ar iabl es M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 1. H o u rs gi g w o rk 24.30 13.65 -2. W o rk -l ife ba la nc e 3.92 0.82 .23** (. 8 3 ) 3. J ob s tre ss 2.48 1.02 -. 0 1 -. 4 9 * * (. 7 0 ) 4. S el f-ra te d e xt ern a l e m pl oya bi li ty 3.74 0.82 .17* .35** -. 0 1 (. 8 4 ) 5. A ge 32.84 10.44 .18* -. 0 3 -. 0 1 .13 -6. G ende r 1.39 0.49 -. 0 3 -. 0 4 .05 -. 0 2 .05 -7. Cont ra ct - Z ero -H our 0.3 0.46 -. 0 8 -. 0 7 -. 0 2 -. 2 1 * * -. 1 5 -. 0 3 -8. Cont ra ct - S et H o u rs 0.34 0.48 .15 .08 -. 0 3 .05 .01 .07 -. 4 7 * * -9. Cont ra ct - Z Z P 0.36 0.48 -. 0 8 -. 0 1 .04 .15* .13 -. 0 4 -. 4 9 * * -. 5 3 * * -10. P aym ent F o rm - T as k 0.54 0.5 -. 0 0 -. 0 4 .24** .17* .13 .05 -. 1 6 * -. 1 6 * .31** -11. P aym ent F o rm - B a se d on hour s 0.42 0.49 -. 0 4 -. 0 1 -. 2 3 * * -. 1 8 * -. 1 6 * -. 1 3 .19* .17* -. 3 6 * * -. 9 2 * * -12. P aym ent F o rm - S et a m ount 0.04 0.19 .10 .10 -. 0 3 .06 0.10 .18* -. 0 6 -. 0 7 .12 -. 2 1 * * -. 1 6 * -13. L oc at io n - U S A 0.5 0.50 .25** .09 .10 .29** .44** .15 -. 2 3 * * -. 1 0 .31** .37** -. 4 1 * * .13 -14. L oc at io n - U K 0.18 0.38 -. 0 3 -. 1 5 * .19* -. 0 1 -. 0 5 .05 .06 0.12 -. 1 8 * .11 -. 1 1 -. 0 0 -. 4 6 * * -15. L oc at io n - N L 0.33 0.47 -. 2 5 * * .02 -. 2 6 * * -. 2 3 * * -. 4 3 * * -. 2 0 * .19* 0 -. 1 8 * -. 4 9 * * .53** -. 1 3 -. 6 9 * * -. 3 2 * * -16. P la tfo rm - U be r 0.47 0.5 .25** .05 .19* .35** .23** .00 -. 1 9 * -. 0 0 .18* .46** -. 4 1 * * -. 1 2 .57** -. 0 0 -. 6 0 * * -17. P la tfo rm - D el ive ro o 0.12 0.33 -. 1 9 * -. 1 8 * .02 -. 1 0 -. 1 1 -. 0 0 .06 -. 0 4 -. 0 2 .02 -. 0 3 .03 -. 3 7 * * .48** .01 -. 3 5 * * -18. P la tfo rm - F o o d o ra 0.24 0.43 -. 1 3 .04 -. 2 2 * * -. 2 6 * * -. 2 1 * * -. 1 1 .19* .18* -. 3 6 * * -. 5 9 * * .61** -. 1 1 -. 4 8 * * -. 2 6 * * .72** -. 5 3 * * -. 2 1 * * -19. P la tfo rm - T em pe r 0.02 0.15 -. 1 1 .04 -. 0 1 .12 -. 1 1 .04 -. 1 0 -. 1 1 .21** -. 0 9 .11 -. 0 3 -. 1 5 * -. 0 7 .22** -. 1 5 -. 0 6 -. 0 9 -20. P la tfo rm - U be rE at s 0.08 0.28 .10 .04 -. 0 1 -. 1 8 * -. 0 8 .16* -. 0 1 -0 .0 8 .09 .06 -. 1 2 .18* .13 -. 0 3 -. 1 2 -. 2 8 * * -. 1 1 -. 1 7 * -. 0 5 -21. P la tfo rm - H el pl ing 0.01 0.11 -. 0 1 .01 -. 0 5 .09 .14 .03 -. 0 7 .15* -. 0 8 -. 0 1 -. 0 9 .28** .11 -. 0 5 -. 0 8 -. 1 0 -. 0 4 -. 0 6 -. 0 2 -. 0 3 -22. S o u rc e o f In c om e 1.31 0.46 -. 3 3 * * -. 1 7 * .09 .01 .03 -. 0 1 .08 -. 2 7 * * .19* .13 -. 1 3 .01 .14 -. 1 4 -. 0 4 .08 -. 0 6 -. 1 4 -. 0 2 .03 .05 -Not e. L ist wi se N =1 7 1 * p < .0 5 . * * p < .0 1
Table 2: Normality check.
Skewness Kurtosis
Variables Statistic Std. Err. Lower Upper Statistic Std. Err. Lower Upper
HrsGig .387 .181 .032 .742 -.431 .359 -1.135 .273
WLB -.777 .184 -1.138 -.416 .247 .365 -.468 .962
JS .187 .184 -.174 .548 -.826 .365 -1.541 -.111
SRE -.554 .185 -.917 -.191 -.229 .368 -.950 .492
Procedure and assumptions
The direct- and indirect effect were firstly tested with model 4 of Hayes (2013) by means of PROCESS in SPSS. This regression model estimates the extent to which an indirect effect of X on Y through mediator M takes place. Note that due to less missing data for these three variables compared to the full model including the moderator, the sample size for this
analysis was 174, instead of the N = 171 of the correlation table. After, the conditional effect was tested with Hayes’ model 8. Through this regression model, the conditional direct and indirect effects of X, moderated by W, are estimated and interpreted (Hayes, 2012). Here, N = 171. In the models, X is hours per month spent gig working, M is job stress, Y is work-life balance, and model 8 adds W, as self-rated external employability.
PROCESS generates 95% bias-corrected bootstrap confidence intervals (CIs) for inference about the conditional indirect effects, by means of requesting 5,000 bootstrap samples. This approach respects the non-normality of the sampling distribution, thereby dealing with the slight skewness in the data mentioned in earlier (Hayes, 2012). The predictor and moderator were mean centered by means of PROCESS. Because of significant correlations with the outcome variables, type of contract (specifically zero-hour contracts), payment form (based on task and based on hours), gig work location (The UK and The NL), type of platform (Uber vs. all other platforms) and source of income were controlled for in the analyses. Tables are presented below.
Regressions
Firstly, model 4 explains 33% of the variance of WLB (R² = 0.329; F(9,174) = 8.95), which is statistically significant (p <0.001). Please refer to Table 3. Model 8 explains 42% of the variance of WLB (R² = 0.418; (F(11,171) = 10.38), which is statistically significant (p <0.001). Please refer to Table 5.
Hypothesis 1 proposed HrsGig is negatively related to worker WLB. Based on model 4, the direct effect of HrsGig (c’1 = 0.01, SE = 0.005) is the difference in WLB between two employees that differ by one unit in hours they spent per month gig working, and who experience the same level of job stress. This direct effect was proven to be significantly different from zero (p <0.05), with a CI excluding zero (0.000 to 0.018). Please refer to Table 3 and 4. This means that two employees that differ by one unit on HrsGig and who
experience the same level of job stress, are estimated to differ by 0.01 units in their WLB. Therefore, the two are very weakly positively, rather than negatively, related. Therefore, Hypothesis 1 is rejected. As shown in Table 3, the total effect (c1 = 0.01, SE = 0.005) is however not significant (p = 0.065, CI [-0.0006 to 0.020]). This means it is not proven that, irrespective of job stress, a person working more hours per month gig working reports better WLB.
Hypothesis 2a indicated there is a positive relationship between HrsGig and job stress. The effect of HrsGig on job stress (a1 = -0.002, SE = 0.01) was not significant (p = 0.771, CI [-0.014 to 0.011]), as shown in Table 3. Hypothesis 2a is hence rejected. Hypothesis 2b, in turn, proposed there is a negative relationship between job stress and WLB. An effect of b1 = -0.41 (SE = 0.06) was found, indicating that two employees who spend the same hours gig working per month but that differ by one unit in their level of job stress are estimated to differ -0.41 units in WLB. As this b is negative, those relatively higher in job stress are estimated to
have a lower WLB. This effect is significant (p <.001), with a 95% CI from -0.523 to -0.305 (Table 3). Therefore, the results support Hypothesis 2b.
The mediation was further analyzed by testing the indirect effect, a1b1, as shown in Table 4. This effect (a1b1 = 0.001, SE = 0.003) is not significantly different from zero, with CI [-0.005 to 0.006]. Therefore, based on the simple mediation analysis conducted using ordinary least squares path analysis, HrsGig does not indirectly influence WLB through its effect on job stress (ab = 0.001, n.s.). A bias-corrected bootstrap CI for the indirect effect based on 5,000 bootstrap samples was not significantly different from zero (-0.005 to 0.006). This means that it cannot be concluded that those who spend more hours per month gig working experience more job stress, which in turn would translate into worse WLB. Although participants relatively higher in HrsGig are not estimated to be higher in job stress (a1 = -0.002, p = .771), participants relatively higher in job stress are estimated to have lower WLB (b1 = -0.41, p <0.001). Overall, the hypothesized indirect effect is rejected.
Hypothesis 3a, analyzed by means of model 8, predicted the anticipated positive
relationship between HrsGig and job stress to be moderated by SRE, so that this relationship would be stronger for lower values of SRE. While the effect of HrsGig (Hypothesis 2a) and SRE on job stress (a2 = -0.12, SE = 0.10) both were found to be not significant (p = 0.21, CI [-0.320 – 0.073]), the interaction term of these two variables did show a significant effect (a3 = 0.02, p <0.05). Please refer to Table 5. Its 95% confidence interval ranges from 0.032 to -0.002. This interaction term indicates that the effect of hours spent gig working on job stress depends on the value for SRE. Here it is significant and negative, meaning that with one unit increase in hours spent gig working, the difference between high or low levels of SRE in job stress is 0.02 units. This means that when working more through an on-demand platform,
slightly lower levels of stress may be anticipated for highly employable people.An index of
moderation of the indirect effect (Hayes, 2016). The index (0.0062, SE = 0.0034) is significantly different from zero, with a bootstrap CI ranging from 0.0002 to 0.0136,
indicating moderated mediation takes place in the model (Table 8). However, when looking more closely to the conditional indirect effect of HrsGig on job stress at values of SRE per 10th, 25th, 50th, 75th and 90th percentile, these bootstrap CIs are all not different from zero. The
10th percentile, indicating very low SRE, for example ranges from -0.016 to 0.001, and the
50th percentile, representing moderate SRE, ranges from -0.005 to 0.007. For high values of
SRE, the 90th percentile, the bootstrap CI ranges from -0.002 and 0.016. Please refer to Table
7. Therefore, among those very low, moderate or high on SRE, the indirect effect is not different from zero as evidenced by different bootstrap CIs all straddling zero. Thus, although a significant interaction term of -0.02 is found, that would indicate that for high values of SRE HrsGig is weakly negatively related to job stress, this evidence is not supported in all outputs, and hence not strong enough to fully support Hypothesis 3a based on bootstrap CIs.
Hypothesis 3b proposed the negative relationship between HrsGig and WLB to be moderated by SRE, so that this relationship would be stronger for lower values of SRE. As already indicated when discussing Hypothesis 1, the direct effect (c’1 = 0.01) was found to be significant and positive. The effect of the SRE directly on WLB is proven to be c’2 = 0.34 (SE = 0.07), which is significant (p <0.001, CI [0.204 to 0.469]) (Table 5). This means that at the mean level of HrsGig, two employees that differ by one unit in their level of SRE are estimated to differ 0.34 units in WLB, meaning those relatively higher in SRE are estimated to have a higher WLB. The interaction term of HrsGig and SRE, c’3 = 0.01 (SE = 0.01), is not significant, with p = 0.320 and CI [-0.005 to 0.015], as shown in Table 5 and 6. This means that the effect of HrsGig on WLB, which is found to be significant and positive, does not depend on the level of SRE. Therefore, Hypothesis 3b is rejected.
Finally, as can be concluded from Table 3 and 5, none of the included control variables were significantly related to WLB or job stress in the models. This means that, when simultaneously included in the regression equation, there is no association between
WLB/stress and type of contract, payment form, gig work location, type of gig work platform (specifically Uber vs. all other platforms) or source of income.
Table 3: PROCESS results: direct and indirect effects of HrsGig on WLB (model 4).
Antecedent
Consequent
Job stress (M) Work-life balance (Y)
Coeff. SE p Coeff. SE p HrsGig (X) a1 -0.002 0.01 0.771 c'1 0.01 0.005 <0.05 Job stress (M) - - - b1 -0.41 0.06 <0.001 constant i1 2.16 0.52 <0.001 i2 5.33 0.39 <0.001 Contract: Zero-hour 0.08 0.17 0.633 -0.05 0.12 0.704 Payment: Task 0.20 0.41 0.626 -0.39 0.29 0.180 Payment: Hours -0.04 0.41 0.931 -0.54 0.29 0.069 Location: UK 0.38 0.22 0.092 -0.14 0.16 0.393 Location: NL -0.29 0.24 0.230 0.09 0.17 0.587 Platform: Uber 0.16 0.21 0.456 0.18 0.15 0.219 Source of Income 0.14 0.18 0.444 -0.18 0.13 0.167 R² = .112 R² = .329 F(8,174) = 2.592 , p<.05 F(9,174) = 8.951 , p<.001
Table 4: PROCESS results: direct and indirect effects of HrsGig on WLB (model 4).
Effect SE p LLCI ULCI
Direct effect c'1 0.010 0.005 <0.05 0.000 0.018
Total effect c1 0.010 0.005 0.065 -0.0006 0.020
Boot SE BootLLCI BootULCI
Table 5: PROCESS results: conditional effect on job stress and WLB (model 8).
Consequent
Job stress (M) Work-life balance (Y)
Coeff. SE p Coeff. SE p HrsGig (X) a1 -0.002 0.01 0.803 c'1 0.01 0.004 0.077 Job stress (M) - - - b1 -0.37 0.05 <0.001 Self-rated external employability (W) a2 -0.12 0.10 0.21 c'2 0.34 0.07 <0.001 HrsGig x self-rated external employability (XW) a3 -0.02 0.01 <0.05 c'3 0.01 0.01 0.320 constant i1 2.29 0.46 <0.001 i2 5.40 0.33 <0.001 Contract: Zero-hour 0.08 0.17 0.655 0.04 0.12 0.700 Payment: Task 0.15 0.40 0.703 -0.36 0.27 0.181 Payment: Hours -0.12 0.41 0.777 -0.50 0.27 0.069 Location: UK 0.35 0.22 0.119 -0.13 0.15 0.405 Location: NL -0.31 0.24 0.202 0.15 0.16 0.363 Platform: Uber 0.22 0.21 0.304 0.04 0.14 0.760 Source of Income 0.08 0.18 0.677 -0.17 0.12 0.159 R² = .139 R² = .418 F(10,171) = 2.588 , p<.01 F(11,171) = 10.383 , p<.001
Table 6: Conditional direct effect of HrsGig on WLB at values of SRE (model 8).
Percentile Boot
Effect SE BootLLCI BootULCI
10th 0.002 0.007 -0.013 0.016
25th 0.005 0.005 -0.005 0.015
50th 0.008 0.004 -0.001 0.016
75th 0.011 0.005 0.0003 0.021
Table 7: Conditional indirect effect of HrsGig on job stress at values of SRE (model 8).
Percentile Boot
Effect SE BootLLCI BootULCI
10th -0.007 0.004 -0.016 0.001
25th -0.003 0.003 -0.010 0.002
50th 0.001 0.003 -0.005 0.007
75th 0.005 0.004 -0.002 0.014
90th 0.006 0.004 -0.002 0.016
Table 8: Index of moderated mediation (model 8).
Boot
Index SE BootLLCI BootULCI
Job stress 0.0062 0.0034 0.0002 0.0136
Discussion Summary of findings
The aim of this study was to examine the relatively unexplored worker perspective to gig working, specifically through on-demand platforms, thereby investigating how the new employment form affects its workers on-the-job and personally. It does so, firstly by examining the effect of gig working on worker work-life balance. While the effects of teleworking and technology use (e.g. Ahuja et al., 2007; Hill et al., 1998; Nam, 2014), flex working (e.g. Peters et al., 2009) and precarious work (e.g. Chan & Tweedie, 2015) on WLB have been widely documented, the construct is still empirically relatively unexplored in the gig working context. The second aim of this study was to explore the effect of gig working on its workers by investigating the role of job stress as a mediator in the relationship between gig working and WLB. The third and final aim of the study was to analyze under which
conditions the direct and indirect effects would be stronger, exploring self-rated external employability as a moderator. In order to address these research aims in gaining further
