Worker’s well-being. Evidence from the apparel industry in Mexico

ABSTRACT

A growing body of research links working environment with employee health and well-being. We argue that greater employee well-being is associated with positive work outcomes, benefits likely vital to business performance. We have expanded the literature on this topic by evaluating these questions within a population of factory workers. To date, studies have examined well-being in the workplace in middle- and low-income countries mostly through the lens of disease and disability or have been restricted to human rights, rather than human well-being in general. This paper connects (1) job resources/work conditions and (2) worker well-being with (3) work outcomes in Mexican apparel factories belonging to the apparel supply chain. Our analysis builds on the first wave of the SHINE Worker Well-Being Survey. We examined links between working conditions, job resources, individual well-being and work outcomes using path analysis within a cross-sectional study comprising approximately 2200 Mexican factory workers. We report that job satisfaction and self-assessed work performance are positively and directly associated with worker well-being. We found that job control, trust, respect and recognition were significant correlates of all examined work outcomes (job satisfaction, work engagement and self-assessed work performance), with significant indirect effects on well-being.

KEYWORDS:

• Well-being
• working conditions
• job resources
• work attitudes
• apparel industry
• supply chain
• Mexico

Introduction

Global supply chains did not only lead to a transformation of the structure of production and employment relationships but also introduced concerns about the social and environmental conditions in which goods and services are designed, produced and distributed (Barrientos, Gereffi, and Rossi 2011; Barrientos, Mayer et al. 2011). It is beyond doubt that demand for global production from the labor-intensive industries of emerging economies increased job opportunities, especially for women and migrant workers. It is clear, however, that appropriate terms and conditions for employment and relevant labor regulations did not follow and the safety and welfare of factory personnel is not always assured (Egels-Zandén and Lindholm 2015).

Following the reluctance and inability of local governments to address work safety challenges, global brands instigated moves to improve worker welfare in their supply chain (Collingsworth 2006; Locke et al. 2007; Egels-Zandén and Lindholm 2015). Weak rule of law and flawed legal systems provided an additional boost for this initiative to grow internally. Because of repeated multiple fatalities due to deficiencies in safety protocols and working conditions (e.g. Rana Plaza and Tazreen Fashions – Bangladesh), voluntary private regulatory systems, offering basic conditions for safe, healthy, fair and respectful working conditions were often encoded in factory audits to ensure compliance and limit risk (Barrientos, Mayer et al. 2011).

In fact, many companies now report their environmental footprints and social impacts in their annual sustainability reports. Thus, they demonstrate, to investors and to the public, their commitment to social responsibility and their company’s values and actions to improve quality of life of (1) their workforce and their dependents and (2) the greater community and society as a whole (Holme and Watts 2000). Reporting standards for environmental impacts are generally well defined by governments or non-governmental bodies such as the UN Global Compact and the Global Reporting Initiative. Standards for reporting social impacts of businesses are less clear-cut. Despite growing recognition that multinational corporations exert major influence on worker health and well-being in their supply chains and that improved employee well-being could translate to improvements in human performance and productivity (ILO and IFC 2016; World Bank 2015), an integrated reporting system that reflects the mutual dependency of employee and business needs is still missing.

This paper presents first findings from the SHINE Worker Well-Being study conducted among over 2,200 workers from three apparel factories in Mexico. This study aims to measure worker well-being with an emphasis on the role of well-being at work (health, recognition, and meaning) in promoting better work outcomes.

More specifically, our current study aims to identify drivers for worker well-being (such as job satisfaction, worker engagement, and self-assessed work performance) among Mexican apparel factory workers and their association with work outcomes, controlling for demographic, socioeconomic, and job-related variables. We hypothesized that job resources and resources in life (which can be provided by employers) contribute to work outcomes and worker well-being, and that the influence of job or life resources on well-being is mediated by work outcomes.

Background

Theoretical model

We employed the job demands-resources (JD-R) model to identify factors influencing worker well-being and work outcomes (Bakker and Demerouti 2007; Demerouti and Bakker 2011). The JDR model was developed as a modification of the job strain /demand-control-support (JDCS) (Karasek 1979; Karasek et al. 2007; Brauchli et al. 2015) and effort-reward imbalance (ERI) models (Siegrist 1996). It is implicit to these three models that employee health and well-being at work originate from a balance between positive and negative job characteristics (i.e. resources vs. demands). While the ERI and JDCS models link precisely defined sets of concepts, the JD-R model is a heuristic, representing a conceptual approach to the influence of job (and in more recent approaches – individual) characteristics on health, well-being, and motivation (Schaufeli and Taris 2014). Additionally, the JD-R incorporates numerous possible working conditions, and focuses on both negative and positive indicators of employee well-being and performance. Despite specific risk factors related to job strain, the JD-R model can be applied to a wide range of occupations because it classifies work factors into two general groups: job demands and job resources (Bakker and Demerouti 2007; Demerouti and Bakker 2011). Both refer to physical, psychological, social, or organizational aspects of the job. However, job demands concentrate on efforts and skills required and are thus related to physiological and psychological costs. Job resources include work characteristics that reduce job demands, help people achieve work goals and contribute to personal development. In other words, job demands, which are usually reflected by high work pressure, emotional demands, and role ambiguity, are theoretically believed and empirically shown to cause strain. Job resources, which comprise social support, performance feedback, and job autonomy, are instrumental for motivation (Bakker and Demerouti 2007).

Following Demerouti et al. (2001) and Bakker and Demerouti (2007) in their assertion that the job-demand model should be empirical, and thus flexible enough to accommodate the various job characteristics and paths that influence employee health and well-being, we grouped job resources into physical, social, mental (psychological) and organizational categories. Together with job demands and resources in life (i.e. work-family conflict and living conditions incorporated into our model), these resources are believed to influence both workers’ well-being and work outcomes.

Following the rationale of numerous reports underscoring the physical demands of work in apparel factories (ILO and IFC 2016), we also incorporated physical resources into our approach. Work in apparel factories typically demands prolonged fixed posture, repetitive movements, and physical hazards, such as improper lighting, heat and cold stress, humidity, poor air quality and circulation, and noise. These factors may significantly influence health, productivity and performance (Parsons 2000; Kjellstrom, Holmer, and Lemke 2009; Akbari et al. 2013; Xiang et al. 2014), however there is also some – though considerably limited – evidence that they may not be influential for happiness (Loscocco and Spitze 1990) and job satisfaction (Klitzman and Stellman 1989).

Social and mental job resources, labeled as psychosocial work factors, have been identified to be occupational risk factors for well-being (De Croon et al. 2004; Herr et al. 2015) and work performance (Price 2001; De Croon et al. 2004; Mossholder, Settoon, and Henagan 2005; Fuß et al. 2008; Young and Corsun 2010; Skagert, Dellve, and Ahlborg 2012). Extensive literature offers both theoretical and empirical evidence that employees exposed to adverse psychosocial factors (e.g. high work load, conflicting demands, an unsupportive or hostile working environment, work pressure) and to low levels of control (lack of decision-making authority, inflexible time schedules, lack of skill development and skill use) are more likely to report (1) unfavorable performance outcomes (e.g. lower job satisfaction and increased turnover) (Avey, Luthans, and Jensen 2009; Bambra et al. 2014; Rourke 2014; Albrecht et al. 2015), and (2) reduced well-being (Bakker and Demerouti 2007; Demerouti and Bakker 2011). On the other hand, when conceptualized from the positive perspective, support from supervisors and coworkers (Spell, Eby, and Vandenberg 2014), organizational support (Eisenberger and Huntington 1986; Rhoades and Eisenberger 2002; Eisenberger 2003) and organizational climate and culture (Zohar and Hofmann 2012) were empirically confirmed to protect against negative effects of job burden. However, there are also some reports depicting lack of impact of psycho-social working conditions either on emotional well-being (Loscocco and Spitze 1990; Bakker and Sanz-Vergel 2013) or on work engagement (Bakker and Sanz-Vergel 2013; Upadyaya, Vartiainen, and Salmela-Aro 2016).

It is known that organizational job resources contribute to workers’ health while at work. However, this perspective is rarely reflected by job models (e.g. turnover models) that are usually applied to highly skilled employees from developed countries (Holtom et al. 2008). Under these circumstances the provision of safe working conditions is more often ensured than it is in the case of less developed countries and their low-skilled work force.

Resources in life

We incorporated resources in life into our conceptual model following earlier findings showing that insufficient and unequal access to clean water, sanitation and health care (CSDH 2008; Marmot et al. 2008), inadequate nutrition (Pernia and Quibria 1999) and lack of work-family balance (Fuß et al. 2008; Bambra et al. 2014; de Neve, Krekel, and Ward 2018) lead to deterioration in health and well-being. Multinational corporations are strongly encouraged to provide sufficient resources in life for their employees, especially those in the supply chains, in order to meet their corporate social responsibility and sustainable development goals for health and well-being (Kolk 2016).

Worker well-being

Although well-being in the workplace setting has been thoroughly examined, the notion of overall employee well-being (beyond only while at work) has been established only recently (Zheng et al. 2015; Węziak-Białowolska, McNeely, and Vanderweele 2019). Further, well-being in the work setting is usually conceptualized through the lens of a single life-related measure, such as mental health/depression (Sanne et al. 2005; Stansfeld et al. 2012; Bentley et al. 2015; Saijo et al. 2015; Madsen et al. 2017), the life satisfaction–job satisfaction link (Diener and Tay 2017; Near and Sorcinelli 1986; Rice, Near, and Hunt 1980), workplace friendship (Sias and Cahill 1998; Morrison and Cooper-Thomas 2016), and work motivation and the life cycle (Kets De Vries et al. 1984), while disregarding the complex multidimensional links between well-being in the workplace, work outcomes, job resources, job demands and resources in life.

In this study, we apply a comprehensive well-being measure designed to reflect complete human flourishing, namely the Flourish Index (FI) (VanderWeele 2017; Węziak-Białowolska, McNeely, and Vanderweele 2019). VanderWeele (2017) proposed an approach based on the aggregation of five domains of human life to derive a measure of complete human flourishing. These domains include happiness and life satisfaction, physical and mental health, meaning and purpose, character and virtue, and close social relationships. While life satisfaction/happiness, meaning and purpose, and close social relationships were often included in measures of social well-being, eudaimonic or hedonic happiness (see for example: Ryff and Keyes 1995; Ryff 1995; Keyes 1998; Diener et al. 2010; Huppert and So 2013; Su, Tay, and Diener 2014), health and character and virtue are typically ignored (VanderWeele 2017). Psychometric analysis for the FI was conducted on a sample of 5565 working adults, which provided evidence on its validity, reliability and applicability (Węziak-Białowolska, McNeely, and Vanderweele 2019).

Work outcomes

Work outcomes include objective performance measures reported by companies as well as self-reported measures. Objective performance measures include, for example, number of days off on a sick-leave, total working hours and decisions to quit. Self-reported measures are often derived from survey responses and reflect, among others, work attitudes such as job satisfaction, work engagement and job involvement. Job involvement and work engagement, despite being similar concepts, are empirically confirmed to be different (Hallberg and Schaufeli 2006). Job involvement is the willingness to exert effort on the job and is related to motivation and challenge (Price 2001; Hallberg and Schaufeli 2006). Work engagement – as measured by the Utrecht Work Engagement Scale (UWES) – was introduced as a conceptual opposite to job burnout and was conceptualized as a state of well-being, characterized by high levels of energy that are invested in work (Hallberg and Schaufeli 2006). It is defined as ‘a persistent, positive affective-motivational state of fulfilment in employees that is characterized by vigor, dedication, and absorption’ (Maslach, Schaufeli, and Leiter 2001, 417).

As reported by Bakker and Demerouti (2007) in their excellent review on the applications of the JD-R model, all these work outcomes have been successfully incorporated in the JD-R model by numerous researchers.

Conceptual research model

In this study, we tested two research hypotheses:

H1 – Job resources and resources in life contribute to both work outcomes and worker well-being;

H2 – The influence of job resources and resources in life on well-being is indirect – mediated by work outcomes.

H1 describes four groups of direct effects between (1) job resources, (2) resources in life, (3) well-being, and (4) work outcomes. Along with evaluating links between job resources and work outcomes and between resources in life and worker well-being, H1 is based on the assumptions that job resources influence well-being and that resources in life may influence work outcomes. H2 assumes that the effects of resources (in life and at work) on well-being can be indirect, impacting well-being by influencing work outcomes, which subsequently create conditions more conducive to being well. Paths corresponding to the research hypotheses are drawn in Figure 1, which presents the conceptual model of worker well-being adopted in this study.

Figure 1. Conceptual model of worker well-being. Note: H1 and H2 refer to the research hypotheses tested.

We hypothesized that an overarching category of job resources along with job demands and resources in life impacts work outcomes and worker well-being. The job resources category comprises four types of job resources: organizational, physical, social and mental. Additionally, reciprocal links between work outcomes and worker well-being are assumed.

Methods

Study design and data source

The analysis builds on the first wave of the SHINE Worker Well-Being Survey (WWBS), which was administered in February 2017. Workers of three Mexican apparel factories who had expressed interest in worker well-being programs were surveyed. Factories selected for the study had been successfully passing external compliance audits with environmental measurement tools designed to examine and measure organizational non-financial performance. These tools have been, however, recently recognized to be insufficiently efficient, flawed or biased mainly due to being based on a ‘check-box compliance’ approach (Short, Toffel, and Huguill 2014; Lebaron and Lister 2015). Hence, our study provided additional, subjective measures beyond objectively adequate working conditions confirmed by external auditors.

The WWBS is a component of the greater Worker Well-Being study, which also includes a set of monthly business metrics on the economic performance of factories (such as number of units produced, total overtime hours, and production quality), individual employee information related to voluntary turnover, working hours, absenteeism, and work injury, as well as spontaneous qualitative comments collected from workers through suggestion boxes, group discussions and interviews to both validate and complement the survey findings. The purpose of multi-mode data collection is to ensure data quality (Leeuw 2005).

During the survey administration, personnel were released from their line positions. One line at a time attended private and enclosed survey stations. Workers’ decisions whether to participate in the survey were not disclosed to management and their survey responses were kept confidential. Each respondent was asked to describe their working conditions, job resources, their own performance and job attitudes, and their health, well-being and living conditions. Additionally, the survey queried about participants’ demographic and socio-economic characteristics.

Aggregated results were shared with both participants and management on separate occasions. As a result of this feedback, factory management undertook several interventions to improve working conditions and the well-being of workers and local communities.

Participants and sample size

In total, 2,278 respondents participated in the WWBS, which accounted for 58.0% of the total eligible workforce (Table 1). In the first factory, 975 workers were surveyed (70% of the workforce), in the second, 1,243 (51% of the workforce) and in the third, 46 (46% of the workforce). In the sample, 50.1% of respondents were operators, 11.0% were supervisors, 8.4% were quality auditors, and 7.4% were assistants. Mean respondent age was 32.5 years (SD = 10.4), close to the average age of 32.8 (SD = 10.6) in the overall worker population of the three factories. The proportion of female workers in the sample was 45.9%, compared to 41.1% in the whole population. Of those employed in the factories, 39.6% had worked less than one year (40.2% in the population) and 25.9% had worked more than five years (24.6% in the population). Of the workers surveyed, 33.2% had post-secondary education, college, or vocational training, 44.1% were married, 67.4% had children, and 46.5% were responsible for elder dependents.

Table 1. Baseline characteristics (not weighted) of participants in the Worker Well-Being Survey (2017).

Characteristic	Total (n = 2278)
% of workforce surveyed	58.0
Gender (female)	45.95
Age – mean (SD)	32.49 (10.58)
below 25	29.45
25–34	28.83
35–44	28.39
45+	13.34
Marital status – married	44.10
Having children (yes)	67.44
Being the primary caretaker for a parent or elder living in the household (yes)	46.51
Education – post-secondary, college or vocational education	33.24
Job tenure
less than 1 year	39.61
1–5 years	34.45
more than 5 years	25.94
Position
operator	50.1
supervisor	11.0
quality auditors	8.4
assistants	7.4

Numbers represent percentages unless indicated differently.

For analytical purposes, all the data was weighted with respect to gender, age, and factory to ensure representativeness. Additionally, all individuals with missing values from any well-being, work outcomes, job resources, or control variables for a given path model were excluded from the analysis. A total of 2,052 individuals provided complete data.

Measures

Table 2 presents the variables included in the analysis. They were grouped into work outcomes, worker well-being, and work-related factors, following the conceptual model presented in Figure 1.

Table 2. Variables used – questions, scale and psychometric properties.

	Question	Scale	Psychometric properties
Work performance outcomes
Job satisfaction	All in all, how satisfied would you say you are with your job? (Daley 1986; Price 2001)	0 = not satisfied all, 10 = completely satisfied	n.a.
Self-assessed performance at work	How would you rate your usual performance at work in the past year? (Cammann, Fichman, Jenkins, & Klesh 1975)	0 = poor performance, 10 = excellent performance	n.a.
Work engagement	The Utrecht work engagement scale constructed by averaging scores from nine statements: (Hallberg and Schaufeli 2006; Schaufeli, Bakker, and Salanova 2016)	1–6; higher values indicating more work engaged worker	Cronbach’s alpha = 0.889; confirmatory factor analysis model – fit statistics: CFI = 0.937, TLI = 0.916, RMSEA = 0.066
Worker well-being
Human Flourishing Measure	The Flourish Index (VanderWeele 2017; Węziak-Białowolska, McNeely, and Vanderweele 2019) – calculated by averaging the scores from each of the nine^a questions: Domain 1: Happiness and Life Satisfaction: (1) Overall, how satisfied are you with life as a whole these days? (2) In general, how happy or unhappy do you usually feel? Domain 2: Mental and Physical Health: (3) In general, how would you rate your physical health? (4) How would you rate your overall mental health? Domain 3: Meaning and Purpose: (5) Overall, to what extent do you feel the things you do in your life are worthwhile? (6) I understand my purpose in life; Domain 4: Character and Virtue: (7) I always seek to do what is good and right in all circumstances, even when it is difficult, and I always act accordingly; Domain 5: Close Social Relationships: (8) I am content with my friendships and relationships; (9) My relationships are as satisfying as I would want them to be;	0–10 – higher values indicate higher flourishing	Cronbach’s alpha = 0.838; confirmatory factor analysis model – fit statistics: CFI = 0.918, TLI = 0.890, RMSEA = 0.048
Work-related factors
Job demand	Job demand scores were calculated by averaging the scores from each of the four questions: (1) My job requires working very hard, (2) I have too much work to do to do everything well, (3) My job is very hectic, (4) My job requires lots of physical effort (Beehr, Walsh, and Taber 1976; Cammann et al. 1983); respondents provided responses on a 6-point Likert scale	1–6; the higher, the better	Cronbach’s alpha = 0.600; confirmatory factor analysis model – fit statistics: CFI = 0.985, TLI = 0.954, RMSEA = 0.055
Job resources
Physical job resources	To reflect the physical work environment, the count of reported unfavorable physical working conditions was conducted. The conditions were: temperature, humidity, lighting, air quality, air flow, noise, one element of ergonomics (uncomfortable postures or workstations) and one element of equipment use (machine or tool problem) (Morgeson and Humphrey 2006)	0–8; the higher, the worse;	n.a.
Organizational job resources	To reflect the possibility of leading a healthy life while working five statements were used: (1) My employer cares about my health, (2) I am able to eat healthy meals at work, (3) I can get clean water at work, (4) I have adequate rest time at work, (5) If I am sick, I can get medical care; respondents provided responses on a 6-point Likert scale. Organizational job resources scores were calculated by averaging scores from five statements.	1–6; the higher, the better;	Cronbach’s alpha = 0.762; confirmatory factor analysis model – fit statistics: CFI = 0.976, TLI = 0.953, RMSEA = 0.059
Social job resources	Social support scale was calculated by averaging scores from five statements: (1) My supervisor treats me fairly, (2) My supervisor is concerned about the welfare of those under him or her, (3) My supervisor is helpful in getting the job done, (4) The people I work with take a personal interest in me, (5) The people I work with can be relied on when I need help; Respondents provided responses on a 6-point Likert scale.	1–6; the higher, the better;	Cronbach’s alpha = 0.863; confirmatory factor analysis model – fit statistics: CFI = 0.993, TLI = 0.982, RMSEA = 0.046
	The trust, respect and recognition scale was calculated by averaging scores from three statements: (1) I trust management where I work, (2) I am recognized or rewarded for my work and (3) Where I work, I am treated with respect. Respondents provided responses on a 6-point Likert scale.	1–6; the higher, the better;	Cronbach’s alpha coefficient (0.802). In the confirmatory factor analytical setting, a model with three indicator variables was identified and associated with perfect fit (CFI = 1, TLI = 1, RMSEA = 0)
Mental resources	Decision autonomy was measured by a single statement: I have a lot to say about what happens at my job (Beehr 1976; Behson 2005)	a 6-point Likert scale	n.a.
Resources in life: Work-family balance	How often do demands of your job interfere with your family life?	1 = yes, demands of your job interfere with your family life sometimes or often and 0 = demands of your job never or rarely interfere with your family life	n.a.

^aThe final approach to human flourishing comprises 10 questions with two questions in the Character and virtue domain. At the time of the survey administration, the flourish index was still in the development stage and included only one character and virtue question.

n.a. – not applicable.

Control variables

Controls for demographic factors were introduced to each model. These include gender, age (below 25, 25–34, 35–44, 45+), and business-related characteristics, such as factory of origin, job tenure (up to one year, 1–3 years, above 3 years) and work shift (day only, night only, day and night). When evaluating each work outcome, we controlled for gender, age, factory of origin, job tenure and work shift, and we controlled only for gender and age while evaluating the well-being outcome. These variables were proven to be essential predictors of workers’ treatment beyond working conditions (Węziak-Białowolska et al. 2017).

Additionally, the analysis took into account workers’ living conditions. The living conditions scale was constructed from a question on income poverty (i.e. worrying about being able to meet monthly living expenses) and three questions reflecting material deprivation with regard to how often participants worried about (1) safety, (2) food, and (3) housing (responses: 0 = never, 1 = occasionally, 2 = often, 3 = all the time). The score was obtained as the average of responses and the scale ranged from 0 to 3. Psychometric properties of the scale in the population of interest were satisfactory (Cronbach’s alpha = 0.858; confirmatory factor analysis model – fit statistics: CFI = 0.998, TLI = 0.995, RMSEA = 0.038).

Analytical approach

Satisfactory measurement properties for all scales used in our model were assured by high Cronbach’s alpha coefficients (Bland and Altman 1997; Tavakol and Dennick 2011) and good fit of the confirmatory factor analysis (CFA) indicating high reliability and validity, respectively (Brown 2006; Jackson, Gillaspy, and Purc-Stephenson 2009; Kline 2011).

Subsequently, we ran three path analyses to establish the role of work-related factors for worker well-being and work outcomes (Bielby and Hauser 1977; Pearl 1998; Kline 2011). All models assumed the same set of mediating and control variables. The only difference was the choice of the outcome variable: job satisfaction (model 1), self-assessed work performance (model 2), and work engagement (model 3). Following the recommendations of MacKinnon et al. (2002, 2007) and Aguinis, Edwards, and Bradley (2017), we tested the joint significance of the association between independent and the dependent variable (direct effect) and associations between independent variables, the mediator, and the dependent variable (indirect effect). The relationship between work-related factors and well-being/work outcomes was subsequently modeled using three specifications. Each model consisted of two linear equations:(1) $W{B}_i={\gamma }_{1}WR{F}_i+{\gamma }_{2}W{O}_{i,k}+{\text{β}}_1{X}_{i,}+{\eta }_{i,k}$(1) (2) $W{O}_{i,k}={\gamma }_{3}WR{F}_i+{\gamma }_{4}W{B}_i+{\delta }_1{Z}_i+{\epsilon }_{i,k},i=1,\dots ,N,k=1,\dots ,3$(2) Here, subscript i represents an individual, the variable WRF indicates exposure to a set of work-related factors, WB is the well-being indicator, WO is one of the three (k = 1,2,3) work outcomes. X is a vector of control variables (gender and age) and Z is the vector of work-related control variables (gender, age, factory, job tenure and work shift). β₁ shows the association between control variables and the well-being outcome, δ₁ shows the association between control variables and a work outcome and γ₁ and γ₃ reflect effects of work-related factors on well-being or a work outcome. γ₂ shows the effects of a work outcome on well-being, while γ₄ shows the effects of well-being on a work outcome. η_i and ε_i are disturbance terms. In this specification, work outcomes and well-being outcome are both mediators and outcomes.

To examine direct and indirect effects, we applied the effects decomposition (Kline 2011). The significance of indirect and total effects was tested with the bootstrapping method (Aguinis, Edwards, and Bradley 2017). To examine a potential effect of mediation, we adopted the strategy suggested by Aguinis, Edwards, and Bradley (2017), namely that the existence of mediation is corroborated when an indirect effect is significant, regardless of the presence or absence of the direct effect.

To evaluate model fit (for both CFA and path analysis), a set of commonly accepted measures was applied, i.e. Root Mean Square Error of Approximation (RMSEA), Standardized Root Mean Square Residual (SRMSR), Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI). For RMSEA and SRMSR, the range of acceptable values is 0.00–0.08 (Browne and Cudeck 1992). With respect to CFI and TLI, values are expected to be above 0.9 (or even 0.95).

Following the results of Rhemtulla, Brosseau-Liard, and Savalei (2012) and suggestions of Norman (2010), all variables measured on a response scale of at least 6 points were treated as continuous data. Results are presented as standardized estimates to present the effect sizes. Analyses were performed using Stata 15 and Mplus 8.

Robustness of the results was assured by controlling for common method bias through the design of the study procedure (Podsakoff et al. 2003). Although it was not feasible to account for rater bias, common measurement context and time biases (as we needed to obtain data from the same individuals in the same measurement context to test the research hypotheses), measures of predictors, mediators and outcome variables were separated proximally and methodologically. In particular, measures of interest were located in different sections of the questionnaire and different response scales were used, such as 6-point Likert scales, 0–10 rating scales, and intensity scales, along with different scale endpoints and different labels.

Results

Statistical analysis

Descriptive statistics for all study variables are presented in Table 3.

Table 3. Descriptive statistics for the study variables.

	Mean	SD	Range
Job satisfaction	8.38	2.37	0–10
Performance quality	7.79	2.62	0–10
Work engagement	4.72	1.26	1–6
Flourish Index	8.93	1.31	0–10
Job demand	3.16	0.65	1–6
Job resources
Physical job resources	1.68	0.65	1–4
Organizational job resources	4.09	1.19	1–6
Social job resources
Social support	4.26	1.32	1–6
Trust, respect and recognition	4.05	1.44	1–6
Mental resources – job control	4.34	1.67	1–6
Resources in life
Experience of work-family conflict	43.1	–	%
Living conditions	1.93	0.95	0–3

SD = standard deviation.

All three models proved to be well-fitted (Table 4).

Table 4. Fit indices – model 1–3.

Model	Chi-square (p-value)	CFI	TLI	RMSEA (95% CI)	SRMSR
Model 1 (job satisfaction)	2.555 (0.465)	1.000	1.006	0.000 (0; 0.039)	0.002
Model 2 (work quality)	1.105 (0.776)	1.000	1.038	0.000 (0; 0.027)	0.002
Model 3 (work engagement)	15.431 (0.001)	0.988	0.871	0.050 (0.027; 0.076)	0.003

Two work outcomes (Table 5) – job satisfaction and self-reported performance at work – were significantly and positively correlated with well-being. These correlations were also characterized by considerable effect sizes; however, the association between job satisfaction and well-being was stronger than the link between self-reported work performance and well-being (standardized estimates of 1.186 vs. 0.911, respectively). No evidence of association between work engagement and well-being was found. Direct associations between work-related factors (either physical or psychosocial) and well-being were also not evident.

Table 5 Path analysis – standardized estimates and standard errors in parenthesis.

Contributing factor (range)	Model 1		Model 2		Model 3
	Outcome: job satisfaction (0–10)	Outcome: flourishing (0–10)	Outcome: work quality	Outcome: flourishing (0–10)	Outcome: work engagement (0–6)	Outcome: flourishing (0–10)
Job satisfaction (0–10)	na	1.186* (0.482)	na	na	na	na
Work quality (0–10)	na	na	na	0.911** (0.308)	na	na
Work engagement (0–6)	na	na	na	na	na	1.311 (1.051)
Flourish index (0–10)	−1.051 (0.877)	na	−0.892 (0.551)	na	−0.930 (1.265)	na
Healthy work environment (1–6)	0.282 (0.184)	0.119 (0.064)	0.375** (0.120)	0.013 (0.082)	0.319 (0.255)	0.026 (0.148)
Physical work environment scale (0–1)	−0.033 (0.049)	0.070 (0.044)	−0.016 (0.048)	0.040 (0.034)	0.009 (0.044)	0.017 (0.039)
Social support (1–6)	0.171 (0.095)	−0.009 (0.058)	0.097 (0.073)	0.073 (0.047)	0.167 (0.121)	−0.024 (0.100)
Trust, respect and recognition (1–6)	0.399*** (0.119)	−0.281 (0.162)	0.219** (0.092)	−0.028 (0.065)	0.356** (0.140)	−0.279 (0.304)
Decision authority (1–6)	0.174* (0.079)	−0.046 (0.065)	0.122* (0.059)	0.018 (0.042)	0.225** (0.098)	−0.136 (0.178)
Job demand (1–6)	−0.157* (0.079)	0.043 (0.058)	−0.085 (0.061)	−0.045 (0.038)	−0.175 (0.098)	0.085 (0.122)
Work family conflict (ref. = no)	−0.039 (0.047)	−0.021 (0.030)	−0.005 (0.041)	−0.045 (0.029)	−0.069 (0.050)	0.025 (0.050)
Controls
Gender (ref. = male)	−0.010 (0.041)	−0.036 (0.028)	−0.036 (0.044)	−0.003 (0.028)	−0.018 (0.040)	−0.018 (0.030)
Age (ref. = below 25)
25–34	0.062 (0.048)	−0.064 (0.047)	0.012 (0.046)	−0.006 (0.034)	0.127** (0.041)	−0.161 (0.139)
35–44	0.147* (0.061)	−0.079 (0.063)	0.082 (0.049)	−0.004 (0.033)	0.241*** (0.067)	−0.227 (0.223)
45+	0.121* (0.049)	−0.090 (0.059)	0.036 (0.044)	−0.004 (0.040)	0.207*** (0.049)	−0.226 (0.202)
Job tenure (ref. = up to one year)
1–3 years	0.034 (0.064)	na	0.056 (0.055)	na	0.036 (0.070)	na
More than 3 years	0.045 (0.060)	na	0.071 (0.051)	na	0.049 (0.056)	na
Factory (ref. = Factory 2)
Factory 1	−0.083* (0.040)	na	−0.060 (0.038)	na	−0.093** (0.039)	na
Work shift (ref. = both shifts or only night shift only)	0.101 (0.076)	na	0.125* (0.055)	na	0.061 (0.103)	na
Living conditions (0-3)	−0.004 (0.040)	−0.033 (0.028)	0.009 (0.038)	−0.035 (0.027)	0.034 (0.037)	−0.084 (0.058)
N	1663		1663		1658

*p < 0.05 **p < 0.01 ***p < 0.001; na-not applicable; standardized coefficients are estimated coefficients in the completely standardized solution (all variables are standardized).

Work-related factors likely to contribute to workers’ performance were identified:

1. Trust, respect and recognition were moderately correlated with each of three work outcomes; the lowest effect size (0.219) was noted for self-reported work performance, while the highest (0.399) was observed for job satisfaction;

2. Job autonomy was associated, though with a rather low effect size, with all three work outcomes; the effect sizes ranged between 0.122 for self-reported work performance and 0.225 for work engagement.

Additionally, we found that two other work-related factors were related to specific work outcomes. A healthy work environment was positively associated with self-reported work performance (moderate effect size of 0.375) and job demand was negatively linked to job satisfaction (small effect size −0.157). All significant paths are presented in Figures 2^{Figure 3}–4.

Figure 2. Path analysis for self-reported work performance (only significant standardized estimates are presented).

Figure 3. Path analysis for work engagement (only significant standardized estimates are presented).

Figure 4. Path analysis for job satisfaction (only significant standardized estimates are presented).

We did not find any evidence for direct associations between work performance and physical working conditions, social support, work-family conflict, or living conditions. This finding is based on the model in which all work-related factors are modeled simultaneously, under the assumption of their coexistence in a real work environment.

These findings do not preclude an indirect influence from work-related factors on well-being (which could be moderated by work outcomes). Therefore, we also examined indirect effects of work-related factors (Table 6).

Table 6. Total indirect effects of work and life resources on well-being – standardized estimates and 95% bootstrap confidence intervals in parentheses.

Contributing factor	Indirect effects of the contributing factor to flourishing via:
	Job satisfaction	Work quality	Work engagement
Healthy work environment	0.037 (0.007;0.098)	0.101 (0.055;0.135)	0.078 (0.039;0.130)
Physical work environment	−0.025 (−0.061;−0.001)	−0.014 (−0.048;0.009)	−0.002 (−0.217;0.030)
Social support	0.043 (0.009;0.085)	0.009 (−0.020;0.042)	0.051 (0.012;0.099)
Trust, respect and recognition	0.163 (0.074;0.226)	0.068 (0.033;116)	0.164 (0.038;0.198)
Decision authority	0.052 (0.020;0.090)	0.029 (0.009;0.064)	0.094 (0.027;0.119)
Job demand	−0.047 (−0.093;−0.014)	−0.013 (−0.055;0.005)	−0.068 (−0.212;−0.061)
Work family conflict (ref. = no)	−0.004 (−0.026;0.022)	0.010 (−0.009;0.035)	−0.024 (−0.141;−0.009)
Living conditions	0.007 (−0.007;0.031)	0.011 (−0.002;0.038)	0.030 (0.014;0.077)

Standardized coefficients are estimated coefficients in the completely standardized solution in which all variables are standardized; 95% bootstrap confidence intervals (B = 500 replications).

Our analyses indicated that some work-related factors are likely to contribute indirectly to worker well-being. Significant indirect effects for all work outcomes were found for: (1) a healthy work environment (positive effect), (2) trust, respect and recognition through all performance outcomes (positive effect) and (3) job autonomy (positive effect). The physical work environment, which is negatively oriented (the higher the score, the worse the physical environment), was found to have a detrimental effect on well-being. The impact was, however, mediated by job satisfaction. This implies that changes in the physical work environment may not have direct effects on well-being in life as they may be confined temporarily and spatially to the work environment. However, it is also possible that declining job satisfaction may subsequently translate to reduced overall well-being. Social support (positive impact) and job demand (negative impact) were found to influence well-being through effects on job satisfaction and work engagement. It is an interesting finding pointing to the fact that job-related issues and support required to resolve problems at work do not affect well-being outside of work unless they have a bearing on job satisfaction and work engagement. Work-family conflict (negative impact) and living conditions (positive impact) were found to be influential on well-being through effects on work engagement. Especially for the work-family conflict, one can imply that if a person was not able to keep high engagement in work due to work-family conflict, well-being was likely negatively affected. Consequently, all relations predicted by our initial hypotheses were confirmed.

Discussion

This study sought to reveal the interlink between working conditions, work outcomes and worker well-being among workers in Mexican apparel factories. Our results indicate that job satisfaction and self-reported work performance are positively associated with worker well-being, pointing to a possible direct impact of work outcomes on well-being. This finding is in line with the part-whole theory (Rice, Near, and Hunt 1980; Near and Sorcinelli 1986) which posits that job satisfaction influences subjective well-being.

We also found that trust, respect and recognition, in addition to job autonomy, were significant contributors to all examined work outcomes (i.e. job satisfaction, work engagement and work quality), with significant indirect effects on well-being. This finding supports the importance of societal or relational well-being (Prilleltensky 2005; White 2017), known as the caring climate (Fu and Deshpande 2014) or organizational climate and culture (Zohar and Hofmann 2012) in the field of organizational psychology, for both work performance and worker psychological well-being. It also corroborates the importance of organizational rewards, which is one of the elements of organizational support theory (Eisenberger and Huntington 1986; Rhoades and Eisenberger 2002; Eisenberger 2003), in shaping work performance outcomes.

Organizational resources reflected by healthy working conditions were found to contribute significantly to work quality and indirectly to worker well-being – via job satisfaction, work quality and work engagement. This finding was supported by spontaneous comments provided by workers who complained about very limited access to drinking water and its quality, as well as heat and work load – elements of the organizational resources scale we used – often indicating how these conditions affected their health and well-being.

Consistent with the fact that level of job demand is a source of significant stress (Demerouti and Bakker 2011), we found that Mexican workers were susceptible to decreased job satisfaction and lower well-being (indirect effect through job satisfaction and work engagement) due to high workload. These conclusions corroborate both of our research hypotheses.

Our study also revealed that workers experiencing work-family conflict and unfavorable physical working conditions might experience depleted well-being, although the effect is indirect and mediated by effects on either work engagement or job satisfaction. Social support from supervisors and co-workers was also likely to indirectly stimulate well-being through impacts on job satisfaction and work engagement. These indirect effects corroborate the important role of perceptions of developmental support from co-workers and supervisory mentors for the promotion of work engagement (Spell, Eby, and Vandenberg 2014).

In contrast to previous findings (Parsons 2000; Bakker and Demerouti 2007; Fuß et al. 2008; Avey, Luthans, and Jensen 2009; Kjellstrom, Holmer, and Lemke 2009; Demerouti and Bakker 2011; Akbari et al. 2013; Xiang et al. 2014; Bambra et al. 2014; Rourke 2014; Albrecht et al. 2015; de Neve, Krekel, and Ward 2018), we did not observe direct impacts from physical working conditions, social support provided by supervisors and colleagues, or work-family conflict on either well-being or work outcomes.

Our findings, while not what we expected, are supported by findings of other scholars. For example, Loscocco and Spitze (1990) reported no association between physical working conditions and happiness among female factory workers. Klitzman and Stellman (1989) found no evidence of associations between workplace lighting conditions and job satisfaction among office workers. Similarly, other studies report null associations between level of job demand and distress and happiness among factory workers (Loscocco and Spitze 1990), human well-being (Bakker and Sanz-Vergel 2013), and work engagement (Bakker and Sanz-Vergel 2013; Upadyaya, Vartiainen, and Salmela-Aro 2016). Insignificant association between job autonomy and distress among female factory workers was found by Loscocco and Spitze (1990). These authors also found no significant associations between (1) supervisor support and both distress and happiness (however, only among female factory workers) and (2) satisfaction with co-workers and distress, only among male factory workers. We also note that although negative or counterintuitive findings are not numerous in the literature, this might result from the well-known difficulty and publication bias in communicating and publishing negative results (Dirnagl and Lauritzen 2010; Fanelli 2012; Matosin et al. 2014).

Although, with respect to the above work-related factors, our findings led us to reject hypothesis H1, it should be stressed that these findings do not preclude this type of impact in other factory settings. Further research is needed to clarify these associations. Longitudinal studies, in particular, would control for baseline effects and so, improve causative interpretation of results. However, the study of Bakker and Sanz-Vergel (2013), who reported no association between job demand, flourishing, and work engagement, was conducted using longitudinal data.

Limitations of our study include use of cross-sectional data, which precludes inferences about the direction of causality for observed associations. Hence, associations that we observed can be interpreted as likely drivers of effects, but we cannot rule out the possibility of reverse causality or correlation without causation. Nevertheless, to circumvent this limitation, reciprocal relations were examined. Future studies, however, should include in-depth analysis of longitudinal panel data to confirm causal influence, after successful identification of relevant correlates. Our study also focuses on Mexican factory workers in the apparel industry, which limits the generalizability of our findings for other professions, work settings, and countries/culture. Future research may benefit from examining other sectors of industry as well as evaluating workers in different countries.

Conclusions

Our study shows that in order to improve working conditions and well-being of workers, specific interventions by both factory management and global brands could be beneficial – recommendations already formulated by Locke, Amengual, and Mangla (2009) and Lund-Thomsen and Lindgreen (2014), among others. Cultivating appreciation of the importance of a worker’s value as a resource and not only perceiving them through the lens of cost is one such example. A second example is allowing greater decision autonomy of workers (Locke, Amengual, and Mangla 2009). Our findings indicate likely rewards to business from this type of worker empowerment.

Acknowledgement

The paper benefited substantially from comments and suggestions made by Irina Mordukhovich to whom authors are truly grateful.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Dorota Węziak-Białowolska, Ph.D., is the research scientist at the Department of Environmental Health, Harvard T.H. Chan School of Public Health. She is also affiliated at the Sustainability and Health Initiative for NetPositive Enterprises (SHINE) and the Human Flourishing Program at the Harvard University. Her research spans organizational psychology, psychometrics, economics, health, and wellbeing.

Piotr Białowolski, Ph.D., is the research scientist at the Department of Environmental Health, Harvard T.H. Chan School of Public Health. He is also affiliated at the Sustainability and Health Initiative for NetPositive Enterprises (SHINE), Harvard University. He specializes in applied social science and quantitative methods for applied social and economic research. His empirical research is mainly on household economic and financial behavior and their association with well-being.

Eileen McNeely, Ph.D., conducts research and teaches in the Environmental Occupational Medicine and Epidemiology Program at the Harvard T.H. Chan School of Public Health. She has worked as a consultant, researcher, clinician, and educator in the field for over twenty years. She is the founder and director of the SHINE.

ORCID

Dorota Węziak-Białowolska http://orcid.org/0000-0003-2711-2283

Piotr Białowolski http://orcid.org/0000-0003-4102-0107

Additional information

Funding

This work was supported by Robert Wood Johnson Foundation: [Grant Number 74275 Building a Culture of Health: A Business Lea]; Levi Strauss Foundation.