2.1. Data and Sample
The Spanish Survey of Working Conditions (Encuesta Nacional de Condiciones de Trabajo—ENCT) is a national cohort study designed and conducted by the Ministry of Employment (Spain) to analyze exposure to various occupational hazards and the factors in the work environment that influence the health of workers. The Spanish Survey of Working Conditions is included in the National Statistical Plan, in accordance with the Laws 4/1990 and 13/1996 of the Kingdom of Spain, ensuring ethical data collection, storage and handling of data. The data confidentiality is assured by the Law 12/89 of the Government Statistic Act that guarantees that the data provided are covered by statistical confidentiality, avoiding its misuse in all cases. In our study, we use data from the 7th Spanish Survey of Working Conditions (the most recent version available to date), which encompasses data of the fieldwork performed between October 2011 and February 2012. The geographical scope of the survey was the entire Spanish territory.
The ENCT is based on population register data as a directory to extract the sample of dwellings to be used for the interviews, combining probabilistic with other characteristics from quota sampling. The population scope is defined as the employed population aged 16 and over, from all economic activities, residing in family housing. The collection of data was carried out through face-to-face structured interviews in family residences. In total, 22,312 homes were visited with a decline rate of 60.15%. The absence of the residents at their homes (24.3%), the lack of workers in the household (17.7%), refusal to cooperate (4.9%) and empty dwellings (4.3%) were among the main reasons for an initial decline to participate. The obtained sample was made up of 8892 people (8070 Spanish vs. 814 from other nationalities), aged between 16 and 64, from all economic activities. After excluding 384 non-responses (4.32%), our final study sample comprised of 8508 respondents (90.87% native workers: n = 7731 and 9.13% migrant/foreign workers: n = 777). These data are close to the population distribution in 2011, when there were 5,751,487 foreign nationals registered in Spain (12.29% of the Spanish population) [6
This study uses the term “migrant worker” to refer to people whose country of origin is different to the country they are residing and working in. We therefore dichotomized the question “nationality” (Q60) into two answers (Spanish/Other nationality). A “Spanish worker” is therefore synonymous with a “native worker”, while a “migrant worker” is synonymous with a “foreign worker” for the purposes of this study. A descriptive analysis of nationalities indicated that the majority of the foreign workers came from America (43.3%) and other European countries (37.6%), whereas only 16.1% and 2.9% came from Africa and Asia, respectively. By country, the largest groups were born in Romania (13%), Morocco (12.9%), Ecuador (10.5%), Colombia (6.3%), Argentina (4.7%) and Peru (4.4%), representing 51.8% of the subsample.
The 7th Spanish Survey of Working Conditions (ENCT) has 62 questions classified into 14 categories: (a) employment status (questions 1 to 7); (b) information about the workplace (questions 8 and 9); (c) type of job (questions 10 to 14); (d) physical agents (questions 15 to 18); (e) chemical and biological contaminants (questions 19 to 25); (f) safety conditions (questions 26 and 27); (g) workplace design, workload and psychosocial factors (questions 28 to 33); (h) prevention organization (question 34); (i) working hours (questions 35 to 40); (j) preventive activities (questions 41 to 47); (k) information/training (questions 48 and 49); (l) violent behaviors at the workplace (question 50); (m) health hazards (questions 51 to 55); and (n) sociodemographic data (questions 56 to 62).
For the purpose of this study, the following items were selected:
Organizational factors: “Hired by company or subcontractor?” (Q5), “Main activity of your company?” (Q6) and “How many people work in your company?”(Q7).
Work-related factors: “Type of employment contract?” (Q3), “Which type of job do you have?”(Q11), “In which conditions do you work (isolation/cooperation with other workers)?” (Q12), “How long have you been working in your workplace?” (Q13) and “On average, how many hours a week do you work (excluding lunch time)?” (Q35).
Psychosocial working conditions
: “To what extent are you annoyed or worried about the risk of losing your current work?” (Q55_18) (job insecurity was conceptualized as a health hazard in line with Probst and Jiang [20
]). Other items selected for inclusion were: “In your workplace, how often can you receive help from your colleagues if you ask for it?” (Q30_1), “…Can you receive help from your superiors/bosses if you ask for it working at very high speeds?” (Q30_2), “…Do you have the chance of doing what you do best?” (Q30_3), “…Can you put your own ideas into practice?” (Q30_4), “…Do you have the feeling of doing something useful?” (customers, passengers, students, patients, etc.) (Q30_5), “…Can you learn new things?” (Q30_6), “…Do you have much work and feel overwhelmed?” (Q30_7), “In your workplace, how often can you receive help from your colleagues if you ask for it?” (Q31_1), “… Can you receive help from your superiors/bosses if you ask for it?” (Q31_2), “…Do you have the chance of doing what you do best?”(Q31_3), “…Can you put your own ideas into practice?” (Q31_4), “In your workplace, how often can you choose or change the order of the tasks?” (Q32_1), “….The method of work?” (Q32_2), “…The rhythm of work?” (Q32_3) and “…The distribution and/or length of your breaks?” (Q32_4).
Responses for questions on job design and psychosocial factors Q30 (7 items), Q31 (4 items) and Q32 (4 items) were collected on a Likert-type scale, coded (from 1 to 5) for Q30, where 1 means “always” and 5 “never” (except Q31 and Q32, which were reverse-coded). For these questions (Q30, Q31 and Q32), a factor analysis was performed to estimate the latent factors by entering all scale items into a principal component analysis and examining the unrotated factor solution (Harman’s single-factor test) in order to identify if common method variance (CVM) is a problem within the data. This analysis did not produce a single assigned factor, since the main factor only explained 32.2% of the total variance [32
]. The use of latent factors, instead of individual measures, is more parsimonious, as it allows a more accurate modeling of the measurement error and better explanation of the contribution of each measure [33
]. Standardized factor loadings for each of these measures also ranged from 0.48 to 0.86, and all were statistically significant at the minimum probability level of 0.001 [34
]. Table 1
shows the factor analysis and latent factors obtained.
Well-being: “Could you tell me if you have any of the following health conditions? Stress, anxiety or nervousness: Stress, anxiety or nervousness” (Q54_B13) and “…Tiredness or exhaustion?” (Q54_B15).
Sociodemographic data: Age (Q56), gender (Q58), educational level, “Which is the highest official education level you have?” (Q59) and nationality (Q60).
The results show three main psychosocial factors: factor 1, labeled as “autonomy”, factor 2, labeled as “workload” and factor 3, labeled as “social support”. Factor 4 was excluded from the analysis, as it comprised of only one item (Q30_1) and, therefore, was not reliable in large samples [35
], while item Q30_5 was excluded because it had low loading scores in each factor [36
2.3. Data Analysis
To analyze the relationships between these variables, the first step in the analysis was to describe the main sociodemographic characteristics, both in the general sample and in our two subsamples (native and migrant workers). Secondly, and in order to examine the differences between native and foreign workers, we carried out a chi-square analysis of all organizational and work-related factors described above using the Statistical Package for the Social Sciences (SPSS) version 23 (IBM, Armonk, NY, USA).
After selecting the organizational and work-related variables, and for the purpose of estimating the associations between these variables and the latent factors identified in our conceptual framework (see Figure 1
), a linear regression within a structural equation modeling (SEM) framework was conducted with the “Lavaan R Package” [37
]. Conventional levels of acceptable model fit (goodness-of-fit (GFI) and comparative fit index (CFI) values over 0.85; root mean square error approximation (RMSEA) values >0.05) and a statistically significant minimum probability level of 0.001 were taken into consideration [38