How Platform Economic Dependence Leads to Long Working Time: The Role of Work Pressure and Platform HRM Practices

Abstract

Drawing upon the conservation of resources (COR) theory, this paper discusses the effect of platform economic dependence on working time and the mediating role of work pressure, as well as the boundary role of platform human resource management practices, with a survey of 9576 takeaway riders. The results suggest that the greater the economic dependence of takeaway riders on the platform organization, the longer they work; work pressure plays a mediating role between platform economic dependence and working time; the amount of platform rewards has no effect on the positive relationship between platform economic dependence and working time, whereas the difficulty of obtaining platform rewards strengthens the positive relationship between platform economic dependence and working time. This paper contributes to the literature on gig work by providing a micro-individual perspective and to the literature on COR theory by enriching the studies of resource caravan passageways.

1. Introduction

Gig workers are those who provide on-demand services to organizational or individual clients through digital platforms without a standard employment relationship and are remunerated based on piecework [1,2,3]. They are often portrayed as self-employed with discretion over their working time [4,5,6], enjoying the autonomy that platforms provide. However, the current literature demonstrates that gig workers are trapped on the platform in terms of working time [6,7], leading to negative outcomes, such as loss of social time [7], higher turnover intentions [8], and negative family relationships [9]. Elucidating what contributes to overwork is favorable for the sustainable development of individuals, families, organizations, and society.

When exploring the antecedents of long working time in the platform context, the extant literature tends to focus predominantly on organizational factors, such as higher-goal stimulation and gamification design [10], algorithmic management [11], and compensation structure adjustment [12]. It is worth noting that gig workers’ individual economic situations have often been overlooked, with only a few exceptions (please refer to Schor et al. [13] and Keith et al. [14] for more details). There are heterogeneous economic circumstances among gig workers [1,6,14]: some view the platform as their primary source of income, while others view it as a supplement; some view the platform as a full-time job, while others view it as part-time; and some exclusively work for one platform, while others engage with multiple platforms. Kuhn and Maleki [1] conceptualize this phenomenon as “platform economic dependence”. Discussing gig workers’ long working time from the perspective of their individual economic circumstances can unpack the invisible but significant factor that hides behind platform organizations and provides insights into their behaviors. Although Schor et al. [13] and Keith et al. [14] have acknowledged the role of economic dependence in working time, both of them have not delved into how and under what conditions platform economic dependence contributes to working time. Thereby, it is necessary to investigate the relationship between platform economic dependence and working time, as well as its mediator and moderator.

Drawing on the conservation of resources (COR) theory [15,16,17,18], we argue that gig workers who exhibit greater economic dependence on the platform will experience a reduction in available resources, leading to an increased likelihood of perceiving resource threats. To respond to these threats, gig workers tend to conserve and invest in other resources, such as working longer hours to earn more money. Furthermore, the perceived loss of resources (i.e., loss of the platform as a source of income) may beget further resource loss (i.e., higher work pressure), which, in turn, affects working time. In addition, resource perceptions are influenced by organizational contextual factors [15]. The changes in the platform human resource management (HRM) practices directly affect the size of the resource pool and the difficulty of resource acquisition, which can be perceived by individuals. Therefore, it is proposed that the relationship between platform economic dependence and working time is mediated by work pressure and moderated by platform HRM practices. Thus, our research question is how platform economic dependence contributes to working time and its related processes and contextual factors. To test the conceptual model, we utilized secondary data, including 9576 takeaway riders from a top food delivery platform in China, which was conducted in five cities.

This research contributes to the literature on gig work and COR theory. First, it extends our understanding of why gig workers are trapped in the platform’s time regulation by providing a micro lens. To be specific, gig workers’ own economic circumstances (the extent to which their income depends on the platform) determine how long they work, and this process is mediated by work pressure. In other words, their time management decisions are not just affected by platform management, as demonstrated in the prior literature [10,11,12], but also by their economic circumstances, as displayed in our study. Second, we discuss how HRM practices, as boundary factors, may affect the perceptions of resource threats, enriching the research related to resource caravan passageways in COR theory. Specifically, takeaway riders focus on changes in the difficulty of accessing resources rather than changes in the quantity of resources, suggesting that not all resource caravan passageways [17] work.

2. Theoretical Underpinning and Framework Development

2.1. Platform Economic Dependence and Working Time

Platform economic dependence refers to the extent to which a worker’s income comes from the platform, and gig workers stick to the platform more when a greater proportion of their income comes from the platform [1]. Working time is defined as the duration of time that takeaway riders are willing to dedicate to work. According to COR theory, resources are defined as valuable physical objects, conditions, personality traits, energy, or the means of acquiring them [16]. Platform organizations serve as the source of income for takeaway riders, which are considered a conditional resource by the riders. According to the primacy of resource loss, the perception and impact of resource loss are stronger than those of equivalent resource acquisition [15]. If takeaway riders experience an increased dependence on the platform in regard to their income, it means they have limited earning opportunities [1,13]. In other words, the takeaway riders are on the weaker side compared to the platforms and cannot control resources. In this case, riders are more sensitive to resource loss. To handle the threat of resource loss, individuals will invest in new resources to recover and gain resources [15]. Thus, riders prefer spending more time working to gain more income (a new resource), through which they prepare themselves to cope with the possible loss of the platform organization as a source of income in the future. Based on this, the following hypothesis is proposed in this paper:

Hypothesis 1.

The platform economic dependence of gig workers is positively related to their working time, such that the more dependent gig workers’ income is on the platform, the longer they work.

2.2. The Mediating Role of Work Pressure

We believe that there is a degree to which riders’ economic dependence on the platform impacts their work pressure. Work pressure is defined as the subjective perceptions of job demands that may be elicited by forces within or outside organizations [19]. In this light, the economic circumstances of takeaway riders can be an external stressor that contributes to work pressure. First, COR theory states that individuals will experience psychological pressure when they perceive a threat of resource loss [15,16]. The more economically dependent takeaway riders are on the platform, the more crucial it is for them. Thus, losing the platform, which is a conditional resource, will be a stronger threat. Additionally, the riders hold a lower status and possess less influence compared to the platform, which hinders their ability to safeguard resources and then leads to a sense of resource threats. Consequently, those takeaway riders who depend more on the platform will perceive stronger threats, resulting in heavier work pressure. Second, drawing on COR theory, those riders lacking resources are more vulnerable to resource loss [17,18], thereby the takeaway riders who depend more on the platform will feel higher levels of work pressure. On the one hand, takeaway riders, due to their non-standard employment, are required to be self-responsible. Specifically, the self-employed gig workers construct a new psychological contract [3] wherein limited organizational resources can be supported by the platform [20]. This exacerbates the plight of resource-poor individuals, especially those riders who heavily depend on the platform, incurring them more sensitive to resource loss. For instance, platforms do not pay riders if they fail to deliver, but the compensation is quite critical for riders who regard the platform as their main source of income. As a result, they will encounter tremendous work pressure. On the other hand, the riders are paid on a “piece-work” basis [21,22], while market demand is volatile (an unstable conditional resource). This instability leads to a precarious income, heightening the resource threat for gig workers, particularly those who are more financially dependent on the platform, which, in turn, further intensifies work pressure. In summary, the riders who rely more heavily on the platform are more susceptible to mistakes at work and market unpredictability, engendering greater work pressure.

Furthermore, we argue that takeaway riders are inclined to work longer when faced with work pressure. First, COR theory posits that “people employ key resources not only to respond to stress but also to build a reservoir of sustaining resources for times of future need” (p. 104) [18]. In this regard, takeaway riders will utilize their key resource, i.e., time to work longer in exchange for money that can not only handle current work pressure but also prepare for the future. Second, it is proposed by COR theory that people may shift their focus from what they might lose to what they might gain when investing resources [16]. When confronted with work pressure, riders pay more attention to how much money they can obtain from the platform for the moment and thereby work longer hours. Based on the above analyses, we propose the following hypothesis:

Hypothesis 2.

Work pressure mediates the relationship between platform economic dependence and working time.

2.3. The Moderating Role of Platform HRM Practices

The COR theory suggests that social and cultural factors can influence the perceptions of resource threats and subsequent response behaviors [15]. HRM practices refer to the “daily enactment of HR philosophies and policies” (p. 3) [23], which act as social cues that influence individual perceptions [24]. As the platform rewards increase, the riders reasonably expect that working the same hours will earn them more money. Such an optimistic expectation diminishes the perception of resource loss, leading to a decrease in working hours, i.e., the amount of platform rewards weakens the positive relationship between platform economic dependence and working time. Similarly, when the difficulty of gaining platform rewards increases, fewer resources (i.e., money) can be obtained for the same amount of working time. This pessimistic anticipation, in turn, enhances riders’ perception of resource loss and then motivates riders to invest more resources by extending working hours, i.e., the difficulty of obtaining platform rewards strengthens the positive relationship between platform economic dependence and working time. Based on the above analyses, the following hypotheses are proposed in this paper:

Hypothesis 3.

The amount of platform rewards weakens the positive relationship between the platform economic dependence of the gig workers and their working time, such that the more rewards the platform provides, the weaker the positive relationship is between the platform economic dependence of gig workers and their working time.

Hypothesis 4.

The difficulty of obtaining platform rewards strengthens the positive relationship between the platform economic dependence of the gig workers and their working time, such that the more difficult to gain platform rewards, the stronger the positive relationship is between the platform economic dependence of gig workers and their working time.

The conceptual model is shown in Figure 1.

3. Research Methods and Variables

3.1. Data Source and Collection

The data in this study were obtained second-hand from a research project that was conducted in 2020, encompassing five cities: Shenzhen, Beijing, Chengdu, Hangzhou, and Harbin. Among these five cities, Shenzhen and Beijing represented the first-tier cities; Chengdu and Hangzhou stood for the second-tier cities; and Harbin was on behalf of the third-tier cities. In this way, we tried to improve the credibility and generalizability of the findings. Our survey targeted takeaway riders from a leading food delivery platform in China, with the goal of understanding the lived experience of gig workers. Due to the COVID-19 outbreak, this survey took a total of 4 months to complete. The questionnaire covered various aspects, such as personal background, employment relationships, working conditions, and other vital information. Initially, there were 10,000 cases, but after eliminating the missing and obviously false ones, 9576 cases remained. Based on our research question, we selected 12 variables from the data.

3.2. Variables and Measures

3.2.1. Working Time

The dependent variable is working time, which was operationalized as the time spent receiving and delivering orders per day. We categorized working time into three groups: those working less than 4 h were coded as 1, those working between 4 and 8 h were coded as 2, and those working more than 8 h were coded as 3. Additionally, in the robustness test, we measured working time by asking them how many days per week they engaged in the work.

3.2.2. Platform Economic Dependence

The core independent variable is platform economic dependence. Kuhn and Maleki [1] were the first to introduce the concept of “platform economic dependence”, which can be captured from the perspectives of income source (i.e., primary/supplementary income, full-time/part-time job, single/multiple platforms), capital investment, and site-based reputation. Other studies [7,13,14,25], also from the aspect of income source, have suggested similar ways to measure platform economic dependence. Among them, primary/supplementary income is the most popular measure, followed by full-time/part-time job and single/multiple platforms. Based on the aforementioned literature, we adopted a similar method to measure platform economic dependence. First, we tried to capture whether the platform income was primary or supplementary for the gig workers by asking “What is the percentage of platform income out of individual total income each month?” This measure was used in Section 4.2, and its coding method is shown in Appendix A (

Table A1. Definition, measurements, and statistical description of variables.

Variables	Definition of the Variables	N	Percentage
Dependent variable
Working time	Less than 4 h = 1	1220	12.74%
	4–8 h = 2	2449	25.57%
	More than 8 h = 3	5907	61.69%
Core independent variable
Platform economic dependence	Minimal dependence (<25%) = 1	1611	16.82%
	Mild dependence (25~50%) = 2	1144	11.95%
	Normal dependence (50~75% = 3)	849	8.87%
	High dependence (75~100% = 4)	1398	14.60%
	Full dependence (100%) = 5	4574	47.77%
Mediator
Work pressure	No pressure = 1	499	5.21%
	Low pressure = 2	288	3.01%
	Normal pressure = 3	3584	37.43%
	High pressure = 4	2125	22.19%
	Great pressure = 5	3080	32.16%
Moderators
Platform reward amount	Decreased = 1	7857	82.05%
	No change = 2	1036	10.82%
	Increased = 3	683	7.13%
Platform reward difficulty	Decreased = 1	4031	42.09%
	No change = 2	1205	12.58%
	Increased = 3	4340	45.32%
Control variables
Gender	Female = 0	302	3.15%
	Male = 1	9274	96.85%
Age	Continuous variable	9576	—
Marital status	Unmarried/divorce/widowed = 0	3749	39.15%
	Married = 1	5827	60.85%
Household registration	Rural = 0	7590	79.26%
	Urban = 1	1986	20.74%
Educational background	Middle school and below = 1	3921	40.95%
	High/junior high/vocational high school = 2	4430	46.26%
	Junior college = 3	948	9.90%
	Undergraduate = 4	243	2.54%
	Master and above = 5	34	0.36%
Workplace	Shenzhen = 1	2569	26.83%
	Beijing = 2	3828	39.97%
	Chengdu = 3	1695	17.70%
	Hangzhou = 4	1031	10.77%
	Harbin = 5	453	4.73%
Platform work experience	6 months and below = 1	2842	29.68%
	7–12 months = 2	1484	15.50%
	1–2 years = 3	2325	24.28%
	2 years and above = 4	2925	30.55%

). In addition, we attempted to capture whether the gig work was full-time or part-time by asking “What is your current employment status?” (coded as category 0–1), and whether the gig workers earned income exclusively from the platform by asking “How many platforms do you currently receive and deliver orders from?” (coded as category 0–1). These two measurements were used in the robustness test (Section 4.3).

3.2.3. Work Pressure

The mediator is work pressure, and its measurement is followed by Russell et al. [26]. Given that our focus is on the overall experience of work pressure, the item “I work under a great deal of pressure” was selected. The question was further concretized as “Are you under a great deal of pressure when accepting and delivering orders through the platform now?” with a 5-Likert scale in our study.

3.2.4. Platform HRM Practices

The moderating variables are platform HRM practices. Changing performance practices is a common strategy to guide gig workers’ perceptions, attitudes, and behaviors in the platform context. Thus, we particularly focused on changes in the amount and difficulty of platform rewards.

3.2.5. Control Variables

Gender and age are two key demographic factors that have implications for working time [27]. People with different marital statuses anticipate varying after-work activities [28], which, in turn, can influence their working time. The household registration (i.e., hukou) system, a Chinese feature, serves as an instrument to access municipal government grants, such as certain public services and welfare [29], which indicate the cost of living in a city [30]. Consequently, household registration impacts working time due to living costs. Furthermore, people with higher education have more opportunities and choices in the labor market and do not necessarily spend a lot of time on gig work. Work location reflects not only employment opportunities but also living costs, exerting impacts on working time. Work experience on the platform implies the degree to which one acknowledges it, affecting how much time he would like to invest. Thus, we included gender, age, marital status, household registration, educational background, workplace, and platform work experience in our controls. All definitions and measures of variables are shown in Appendix A (Table A1).

3.3. Econometric Estimation

Since working time is an ordered categorical variable, an ordinal logistic (i.e., ologit) model is tentatively constructed to examine the main effect between platform economic dependence and working time. Model (1) includes only dependent variable and independent variable, while Model (2) incorporates control variables in addition to Model (1):

worktime = β₁₁ + β₁₂ ecodependence + ε₁

(1)

worktime = β₂₁ + β₂₂ ecodependence + Σβ₂₃ control + ε₂

(2)

To test the mediating role of work pressure between platform economic dependence and working time, we employed the stepwise method [31] to construct Model (3) and Model (4). Furthermore, we utilized bootstrapping methods to further test the mediation effect:

workpressure = β₃₁ + β₃₂ ecodependence + Σβ₃₃ control + ε₃

(3)

worktime = β₄₁ + β₄₂ ecodependence + β₄₃ workpressure + Σβ₄₄ control + ε₄

(4)

To test the moderating role of platform HRM practices, we constructed Model (5) and Model (6):

worktime = β₅₁ + β₅₂ ecodependence + β₅₃ rewardamount
+β₅₄ ecodependence * rewardamount + Σβ₅₅ control + ε₅

(5)

worktime = β₆₁ + β₆₂ ecodependence + β₆₃ rewarddifficulty
+β₆₄ ecodependence * rewarddifficulty + Σβ₆₅ control + ε₆

(6)

Worktime is the dependent variable, representing the amount of time the takeaway riders dedicate to their work per day; ecodependence is the core independent variable, representing the extent of economic dependence of takeaway riders on the platform; workpressure is the mediating variable, representing how much pressure the takeaway riders bear at work; rewardamount is moderator 1, representing changes in the number of platform rewards; rewarddifficulty is moderator 2, representing changes in the difficulty of obtaining platform rewards; control represents all the control variables based on the existing literature, including individual characteristics, human capital characteristics, and work characteristics. β₂₂ is the main (total) effect of platform economic dependence on working time; β₃₂ is the effect of platform economic dependence on work pressure; β₄₃ is the effect of work pressure on working time after controlling for platform economic dependence; β₃₂ * β₄₃ is the indirect effect of platform economic dependence on working time through work pressure; β₄₂ is the direct effect of platform economic dependence on working time; β₅₄ is the moderating effect of platform reward amount on the relationship between platform economic dependence and working time; β₆₄ is the moderating effect of platform reward difficulty on the relationship between platform economic dependence and working time; and ε₁–ε₆ are the random errors associated with the models.

4. Results

4.1. Descriptive Analyses

According to the survey, more than 60% of takeaway riders reported that they worked above 8 h per day, nearly half (47.77%) of them had full economic dependence on the platform, and only 8.22% of them believed that their work pressure was lower than normal pressure. Most takeaway riders (82.05%) expressed that the platform had decreased the number of rewards; regarding the difficulty of obtaining platform rewards, approximately 42.09% of takeaway riders felt that it had decreased, while 45.32% felt that it had increased. In terms of demographic characteristics, the majority (96.85%) were male, over 60% were married, and almost 80% held a rural household registration. Regarding education, approximately 87% of the takeaway riders had completed less than a junior college degree. Furthermore, a substantial proportion of riders worked in first-tier cities (i.e., Beijing and Shenzhen), accounting for about 66.8% of the total sample. About one-third (30.55%) of riders had been working on the platform for more than 2 years, while a similar proportion (29.68%) had been on the platform for less than half a year. More details about the sample are shown in Appendix A (Table A1).

In this paper, Spearman’s rank-order correlations were employed to examine the relationships between the levels of the variables using Stata 17, as not all the variables were continuous. As indicated in

Table 1. Correlations between variables.

Variable	1	2	3	4	5	6	7	8	9	10	11	12
1. Working time	1
2. Platform economic dependence	0.536 ***	1
3. Work pressure	0.183 ***	0.196 ***	1
4. Platform reward amount	−0.114 ***	−0.144 ***	−0.242 ***	1
5. Platform reward difficulty	0.004	0.058 ***	0.001	0.119 ***	1
6. Gender	0.047 ***	0.034 ***	0.037 ***	−0.020 *	0.010	1
7. Age	−0.033 **	−0.094 ***	0.007	0.065 ***	0.021 *	−0.051 ***	1
8. Marriage	0.004	−0.067 ***	0.045 ***	−0.028 **	−0.003	−0.038 ***	0.411 ***	1
9. Household	−0.060 ***	−0.046 ***	−0.045 ***	0.030 **	0.017	−0.055 ***	0.100 ***	−0.022 *	1
10. Educational background	−0.032 **	−0.024 *	−0.010	−0.061 ***	0.041 ***	0.020 *	−0.119 ***	−0.063 ***	0.193 ***	1
11. Workplace	0.003	0.038 ***	−0.007	−0.041 ***	0.018	−0.033 **	0.044 ***	0.020 *	0.084 ***	0.036 ***	1
12. Platform work experience	0.185 ***	0.180 ***	0.186 ***	−0.156 ***	−0.018	0.027 **	0.197 ***	0.129 ***	−0.039 ***	−0.082 ***	0.072 ***	1

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

, all the correlation coefficients between the variables were below 0.6, with none exceeding the threshold value of 0.8. This suggests the absence of serious multicollinearity among the independent variables. In addition, a variance inflation factor (VIF) test was conducted in this study. It is generally believed that multicollinearity exists when VIF > 10. The VIF for each explanatory variable was less than 10, and the average VIF was 1.10, which further confirmed that there was no serious multicollinearity.

The results of Spearman’s rho show that platform economic dependence was positively correlated with work pressure (r_s = 0.196, p < 0.001) and working time (r_s = 0.536, p < 0.001). There was a positive relationship between work pressure and working time (r_s = 0.183, p < 0.001). The results preliminarily support the theoretical hypotheses and allow for further testing of the regression models.

4.2. Tests of Hypotheses

4.2.1. Test for the Main Effect between Platform Economic Dependence and Working Time

The results of the ologit regression are shown in

Table 2. Main effect of platform economic dependence and working time.

Variable	Model (1)		Model (2)
	Working Time		Working Time
	β	OR	β	OR
Platform economic dependence	0.814 ***	2.257 ***	0.792 ***	2.208 ***
	(51.948)		(48.905)
Controls	No		Yes
N	9576		9576
Prob > chi2	0.0000		0.0000
LR chi2	3251.25		3416.43
Pseudo R2	0.1867		0.1962

Note: OR = odds ratio; *** p < 0.001.

. Model (1) did not control for individual characteristics, human capital characteristics, and work characteristics, whereas Model (2) did. The Prob > chi² for both Model (1) and Model (2) were 0.000, indicating that both models were statistically significant; both pseudo R² were close to 0.2, indicating that both models exhibited a substantial level of explanatory power. The results of both models were similar, and we focused on Model (2). The coefficient (β = 0.792) estimate for platform economic dependence was found to be significant at the 0.1% level. The corresponding odds ratio of platform economic dependence was 2.208, suggesting that, for the takeaway riders, the odds of working longer in a higher category will increase by about 120% as platform economic dependence increases to the next level. Hypothesis 1, which posited a positive relationship between platform economic dependence and working time, is supported.

To investigate the nuanced impact of platform economic dependence on working time, we conducted an alternative ologit regression, treating platform economic dependence as a dummy variable. The results can be found in Appendix B (

Table A2. Main effect between different levels of platform economic dependence and working time.

Variable	Model (1)		Model (2)
	Working Time		Working Time
	β	OR	β	OR
Platform economic dependence (minimal = reference)
Platform economic dependence: mild	0.798 ***	2.221 ***	0.789 ***	2.201 ***
	(10.893)		(10.683)
Platform economic dependence: normal	1.994 ***	7.345 ***	1.937 ***	6.938 ***
	(23.253)		(22.278)
Platform economic dependence: high	2.694 ***	14.791 ***	2.640 ***	14.013 ***
	(33.413)		(32.217)
Platform economic dependence: full	3.234 ***	25.381 ***	3.147 ***	23.266 ***
	(48.236)		(45.585)
Controls	No		Yes
N	9576		9576
Prob > chi2	0.0000		0.0000
LR chi2	3289.91		3453.25
Pseudo R2	0.1889		0.1983

Note: *** p < 0.001.

), which are similar to those in Table 2.

4.2.2. Test for the Mediating Effect of Work Pressure

As shown in

Table 3. Mediation effect of work pressure by causal steps method.

Variable	Model (2)	Model (3)	Model (4)
	Working Time	Work Pressure	Working Time
	β	β	β
Platform economic dependence	0.792 ***	0.114 ***	0.779 ***
	(48.905)	(15.642)	(47.771)
Work pressure			0.146 ***
			(6.866)
Controls	Yes	Yes	Yes
N	9576	9576	9576
Prob > chi2	0.000		0.000
LR chi2	3416.43		3463.36
Pseudo R2	0.1962		0.1989
Prob > F		0.0000
Adjusted R2		0.0605
F		39.52

Note: *** p < 0.001.

, the Prob > chi² for both Model (2) and Model (4) were 0.000, indicating that the models were statistically significant; both pseudo R² were close to 0.2, indicating that the models had a considerable level of explanatory power. Model (3) exhibited a Prob > F of 0.000, indicating that the model was statistically significant; the adjusted R² was 0.0605, indicating that the model had explanatory power.

As presented in Model (2), platform economic dependence and working time were significantly positively related (β = 0.792, p < 0.001). Model (3) shows that platform economic dependence and work pressure were significantly positively related (β = 0.114, p < 0.001). As illustrated in Model (4), work pressure was significantly positively related to working time (β = 0.146, p < 0.001), and platform economic dependence and working time were significantly positively related (β = 0.779, p < 0.001). Thus, work pressure partially mediated platform economic dependence and working time.

To further examine the mediation effect of work pressure, this paper conducted bootstrapping methods using Stata’s external command “sgmediation”. We set the random sample to 1000 times with a 95% confidence interval. The results of bootstrapping are presented in

Table 4. Mediation effect of work pressure by bootstrapping methods.

Variable	β	Boot SE	95% Confidence Interval
Indirect effect (a * b)	0.0041	0.0007	(0.0027, 0.0055)
Direct effect (c’)	0.2491	0.0044	(0.2404, 0.2577)

; the indirect effect of work pressure did not contain zero (0.0027, 0.0055) at the 95% confidence interval. This finding implies that the mediating effect of work pressure was significant, with an effect value of 0.0041. After controlling for the mediating variable work pressure, the direct effect of platform economic dependence on working time did not contain zero (0.2404, 0.2577) at the 95% confidence interval, indicating that the direct effect was significant, with an effect value of 0.2491. The mediating effect accounts for 1.62%. Thus, Hypothesis 2 is further confirmed.

4.2.3. Test for the Moderating Effect of Platform Reward Amount and Difficulty

As displayed in

Table 5. Moderation effect of platform reward amount and difficulty.

Variable	Model (5)	Model (6)
	Working Time	Working Time
	β	β
Platform economic dependence	0.794 ***	0.695 ***
	(22.434)	(19.634)
Platform reward amount	−0.087
	(−1.064)
Platform economic dependence * platform reward amount	−0.004
	(−0.148)
Platform reward difficulty		−0.217 ***
		(−3.844)
Platform economic dependence * platform reward difficulty		0.049 **
		(3.138)
Controls	Yes	Yes
N	9576	9576
Prob > chi2	0.000	0.000
LR chi2	3422.96	3431.74
Pseudo R2	0.1966	0.1971

Note: ** p < 0.01, *** p < 0.001.

, the Prob > chi² for both Model (5) and Model (6) were 0.000, indicating that the models were statistically significant; both pseudo R² were close to two, indicating that the models exhibited a considerable level of explanatory power. As reported in Model (5), the interaction term between platform economic dependence and platform reward amount was not significant (β = −0,004, p > 0.05), indicating that platform reward amount had no effect on the positive relationship between platform economic dependence and working time of riders. Model (6) shows the interaction term of platform economic dependence and platform reward difficulty was significantly positively related (β = 0.049, p < 0.01), and platform economic dependence was significantly positively related to working time (β = 0.695, p < 0.001). These indicate that platform reward difficulty had a reinforcing effect on the positive relationship between platform economic dependence and working time, and its interaction plot is displayed in Figure 2. Hypothesis 3 is not supported, while Hypothesis 4 is supported.

4.3. Robustness Test

To ensure the reliability of the estimated results, robustness tests were conducted. First, we redefined the core explanatory variable. To be specific, we remeasured platform economic dependence by asking “What is your current status as a rider?”, which was coded as one if the response was a full-time job and coded as zero if the response was a part-time job. A full-time job denotes higher platform economic dependence, whereas a part-time job indicates lower platform economic dependence. The results are displayed in

Table A3. Robustness test for main effect between platform economic dependence and working time.

Variable	Model (1)		Model (2)
	Working Time		Working Time
	β	OR	β	OR
Full-time/part-time job	2.817 ***	16.727 ***	2.791 ***	16.297 ***
	(54.021)		(51.461)
Controls	No		Yes
N	9576		9576
Prob > chi2	0.0000		0.0000
LR chi2	3534.84		3763.57
Pseudo R2	0.2030		0.2161

Note: OR = odds ratio; *** p < 0.001.

,

Table A4. Robustness test for mediation effect of work pressure by causal steps method.

Variable	Model (2)	Model (3)	Model (4)
	Working Time	Work Pressure	Working Time
	β	β	β
Full-time/part-time job	2.791 ***	0.434 ***	2.749 ***
	(51.461)	(17.568)	(50.182)
Work pressure			0.105 ***
			(4.874)
Controls	Yes	Yes	Yes
N	9576	9576	9576
Prob > chi2	0.000		0.000
LR chi2	0.2161		3787.20
Pseudo R2	3763.57		0.2175
Prob > F		0.000
Adjusted R2		0.0666
F		43.67

Note: *** p < 0.001.

and

Table A5. Robustness test for moderation effect of platform reward amount and difficulty.

Variable	Model (5)	Model (6)
	Working Time	Working Time
	β	β
Full-time/part-time job	2.927 ***	2.805 ***
	(25.784)	(24.159)
Platform reward amount	0.012
	(0.232)
Full-time/part-time job * platform reward amount	−0.109
	(−1.425)
Platform reward difficulty		0.015
		(0.381)
Full-time/part-time job * platform reward difficulty		−0.007
		(−0.143)
Controls	Yes	Yes
N	9576	9576
Prob > chi2	0.000	0.000
LR chi2	3766.46	3763.77
Pseudo R2	0.2163	0.2161

Note: *** p < 0.001.

. In addition, platform economic dependence was also measured by asking “How many platforms do you currently work for?”, with responses equaled to one for individuals working on only one platform and zero for those working on multiple platforms. When takeaway riders work on multiple platforms, it means that they have more options and thus reduce their financial dependence on one particular platform. The results are presented in

Table A6. Robustness test for main effect between platform economic dependence and working time.

Variable	Model (1)		Model (2)
	Working Time		Working Time
	β	OR	β	OR
Single/multi platforms	0.459 ***	1.582 ***	0.509 ***	1.664 ***
	(5.938)		(6.428)
Controls	No		Yes
N	9576		9576
Prob > chi2	0.0000		0.0000
LR chi2	34.27		626.00
Pseudo R2	0.0020		0.0359

Note: OR = odds ratio; *** p < 0.001.

,

Table A7. Robustness test for mediation effect of work pressure by causal steps method.

Variable	Model (2)	Model (3)	Model (4)
	Working Time	Work Pressure	Working Time
	β	β	β
Single/multi platforms	0.509 ***	−0.182 ***	0.563 ***
	(6.428)	(−4.089)	(7.051)
Work pressure			0.282 ***
			(14.531)
Controls	Yes	Yes	Yes
N	9576	9576	9576
Prob > chi2	0.000		0.000
LR chi2	626.00		838.39
Pseudo R2	0.0359		0.0481
Prob > F		0.0000
Adjusted R2		0.0381
F		24.71

Note: *** p < 0.001.

and

Table A8. Robustness test for moderation effect of platform reward amount and difficulty.

Variable	Model (5)	Model (6)
	Working Time	Working Time
	β	β
Single/multi platforms	0.258	0.279
	(1.210)	(1.405)
Platform reward amount	−0.479 **
	(−2.872)
Single/multi platforms * platform reward amount	0.240
	(1.410)
Platform reward difficulty		−0.075
		(−0.928)
Single/multi platforms * platform reward difficulty		0.108
		(1.287)
Controls	Yes	Yes
N	9576	9576
Prob > chi2	0.000	0.000
LR chi2	677.40	628.88
Pseudo R2	0.0389	0.0361

Note: ** p < 0.01.

. The significant positive relationship between platform economic dependence and working time was confirmed by Table A3 and Table A6, which was partially mediated by work pressure (shown in Table A4 and Table A7). However, the moderating role of both platform reward amount and difficulty was not verified, as suggested by Table A5 and Table A8.

Second, we replaced the measure of the dependent variable (i.e., working time) with the number of days worked per week. How many days a takeaway rider works can also reflect one’s working time. In terms of the main effect, mediation effect, and moderation effect, similar conclusions to those presented in Section 4.2 are drawn from

Table A9. Robustness test for main effect between platform economic dependence and working time.

Variable	Model (1)		Model (2)
	Days per Week		Days per Week
	β	OR	β	OR
Platform economic dependence	0.550 ***	1.733 ***	0.555 ***	1.742 ***
	(38.339)		(36.739)
Controls	No		Yes
N	9576		9576
Prob > chi2	0.0000		0.0000
LR chi2	1551.42		1772.13
Pseudo R2	0.0721		0.0823

Note: OR = odds ratio; *** p < 0.001.

,

Table A10. Robustness test for mediation effect of work pressure by causal steps method.

Variable	Model (2)	Model (3)	Model (4)
	Days per Week	Work Pressure	Days per Week
	β	β	β
Platform economic dependence	0.555 ***	0.114 ***	0.546 ***
	(36.739)	(15.642)	(35.772)
Work pressure			0.088 ***
			(4.262)
Controls	Yes	Yes	Yes
N	9576	9576	9576
Prob > chi2	0.000		0.000
LR chi2	1772.13		1790.21
Pseudo R2	0.0823		0.0832
Prob > F		0.000
Adjusted R2		0.0605
F		39.52

Note: *** p < 0.001.

and

Table A11. Robustness test for moderation effect of platform reward amount and difficulty.

Variable	Model (5)	Model (6)
	Days per Week	Days per Week
	β	β
Platform economic dependence	0.520 ***	0.483 ***
	(15.328)	(14.171)
Platform reward amount	−0.225 **
	(−2.795)
Platform economic dependence * platform reward amount	0.025
	(1.016)
Platform reward difficulty		−0.208 ***
		(−3.714)
Platform economic dependence * platform reward difficulty		0.037 *
		(2.458)
Controls	Yes	Yes
N	9576	9576
Prob > chi2	0.000	0.000
LR chi2	1789.68	1790.35
Pseudo R2	0.0832	0.0832

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

.

5. Discussion

5.1. Theoretical Implications

Our findings suggest that platform economic dependence leads to working time, which is mediated by work pressure and moderated by changes in the difficulty of the platform rewards. This research makes contributions to the literature on both gig workers and COR theory in two main ways.

5.1.1. Provide a Micro Lens to Understand Working Time of Gig Workers

First, existing studies place more attention on how workplace context contributes to gig workers’ working time [10,11,12], neglecting individuals’ own situations. The lens that focuses on the economic circumstances of gig workers is essential, as many of them face viability challenges [20]. It is beneficial for us to institute and implement practices if we understand their work behaviors based on their individual situations.

In this light, this paper elucidates that riders extend their working time due to their own economic situation, i.e., the stronger their economic dependence is on the platform, the longer their working hours. Our finding is consistent with two other studies [13,14], stimulating reflection on the paradoxical relationship between autonomy and control. Platform organizations enable gig workers to decide how much time to spend on work, whereas riders are stuck on the platform for long hours. This time regulation, on the one hand, is related to platforms’ gamification design [10], algorithmic management [11], and compensation structure adjustment [12]. On the other hand, it is relevant to gig workers’ own situations, as denoted by our study. It is suggested that economic dependence offers an alternative explanation for platform labor control (i.e., time regulation in our study), which echoes the finding of Schor et al. [13].

Furthermore, our findings reveal that work pressure plays a mediating role, unpacking the psychological process mechanism from platform economic dependence to working time. Thus, platform organizations should pay attention to the psychological state of riders when managing them, such as by conducting pressure management training.

5.1.2. Enrich Research on Resource Caravan Passageways by Examining the Role of Platform HRM Practices

Second, the resource caravan passageways in COR theory are underlined compared to its basic tenet, principles, and corollaries, thus, our research contributes to the advancement of this line. The resource caravan passageways are defined as the external conditions that are beneficial or harmful to resources [17]. Consistent with previous studies [32,33], we find that not all resource caravan passageways work: riders are only sensitive to the difficulty of accessing resources (i.e., platform reward difficulty), but not to changes in the resource pool (i.e., platform reward amount). This finding may be explained by the characteristics of gig workers. Most of them come from a lower level of society [34,35] with limited resources and struggle to survive. Thus, they are the realists who care more about whether they can obtain the resources (i.e., difficulty) than the size of the resources (i.e., amount) because the resource pool means nothing without possession. Furthermore, the finding that the platform reward difficulty reinforces the positive relationship between platform economic dependence and working time supports the view of Halbesleben et al. [36], that is, resource caravan passageways may heighten resource burden and subsequently lead to negative outcomes. During the initiation and implementation phases of HRM practices, it is advisable to evaluate the impact of HRM practices on resources.

5.2. Practical Implications

The platform economic dependence of the takeaway riders has significant implications for both the platforms and society. Excessive platform economic dependence creates a potential for the extreme exploitation of gig workers by the platforms, leading to what is referred to as “a race to the bottom” [10,11,13]. This may be detrimental to platforms and society. On the one hand, a long working time may cause gig workers to leave the platform, which, in turn, creates a setback for the platforms. Platform workers are essential for the functioning of platform organizations. In this regard, our research offers a micro lens for platform organizations, enabling platforms to gain a deeper understanding of platform workers and their needs. By understanding the experiences and challenges faced by platform workers, platform organizations can develop more humane and effective management strategies to support and improve the working conditions of these gig workers. For instance, platforms can utilize digitalization to enhance participation in goal setting and development feedback [37] concerning working time, through which platforms offer suggestions for gig workers on how to balance their work and rest time.

On the other hand, the well-being of platform workers will decline due to overwork, which may have adverse effects on societal sustainability. HRM policies can play a role in contributing to societal well-being [38]. From this point, our study provides insights into the selection of HRM practices that can improve the well-being of platform workers and, thus, enhance societal well-being. What is more, as Schor et al. [13] advised, providing more job opportunities in the labor market may be a potential solution to mitigate the “race to the bottom” issue.

5.3. Limitations and Directions

This paper acknowledges several limitations. First, the data in this paper are cross-sectional, making it challenging to establish causal relationships. However, the large samples of the data may partially compensate for this shortcoming. Second, all the variables are measured by a single question, which may affect the reliability of the results. To address this concern, we conducted several robustness tests to improve the reliability and credibility of our conclusions. It should be noted that the moderating effect of HRM practices fails the robustness tests. This indicates that the impact of HRM practices on the relationships between economic platform dependence and working time may vary. Given this instability, further research is required to explore the specific conditions under which HRM practices may act as moderators in the relationships between platform economic dependence and working time among gig workers. Third, we assume platform economic dependence is a challenge, only focusing on its negative effects. However, challenges at work may lead to favorable outcomes, such as improving one’s resilience [39]. Thus, future work can focus on the bright side of platform economic dependence.

6. Conclusions

Our research demonstrates that the takeaway riders’ own economic situation (i.e., platform economic dependence) is positively related to their working time. This offers an alternative approach to explaining why the riders accept time regulation from the platform. This finding is consistent with previous research [13] and further reveals the lived experience of gig workers. What’s more, our paper supports the mediating role of work pressure between platform economic dependence and working time, unpacking the psychological processes of gig workers. Last, we discussed how HRM practices affect the relationship between platform economic dependence and working time, clarifying that resource caravan passageways are sometimes inefficient for the resource-poor. Understanding the real situation of gig workers is essential for policymakers. It is expected that this research could provide a micro perspective for the literature on gig work and enrich the studies of resource caravan passageways in COR theory.