The study described here was part of a program of research on wellbeing at work. This started with research that aimed to address “What is a good job?” (Smith et al., 2011) . A key finding from that project was that one has to examine both positive and negative outcomes rather than inferring one from the other. This view fits with research showing that positive and negative emotions involved different brain processes (Watson, Clark, & Tellegen, 1988) . Much of our previous research on wellbeing at work has focused on negative outcomes such as stress and fatigue ( Smith et al., 2000 ; Smith, Allen &Wadsworth, 2006). In contrast, research on wellbeing largely deals with life satisfaction, happiness and positive affect (Diener et al., 1999) .
Another feature of the present research was that it used a process model similar to those developed in stress (Mark & Smith, 2008, 2011, 2012) and fatigue research (Smith, 2010) . Such models start with job characteristics that may influence wellbeing (e.g. negative characteristics such as job demands; positive characteristics such as control and support), appraisals (e.g. perceived stress; job satisfaction) and outcomes (health outcomes such as anxiety and depression, musculoskeletal disorders and illnesses caused or made worse by work; absenteeism; accidents; presenteeism and performance efficiency). In addition to work-related factors it is important to measure individual characteristics (e.g. positive personality and healthy lifestyle). It is also important to assess life outside of work and work-life balance.
In order to measure the multiple constructs described above it is necessary to use short scales. This approach has been used by Williams (2012) and many of the items in the present survey were taken from the Wellbeing Process Questionnaire (WPQ). These items have been shown to be highly correlated with the longer scales from which they were derived and to show the predicted associations between job characteristics, appraisals and outcomes (Williams & Smith, 2013, 2016) . In other words, the survey measured both positive and negative aspects of well-being (job characteristics; appraisals; and outcomes) using single items which had been shown to correlate highly with longer scales. These single items also showed the same predictive validity as the longer scales (Williams & Smith, 2013, 2016) .
A final aim of the project was to extend the sample to consider a very different occupation in several countries. Much of the previous research on wellbeing has used public sector employees (e.g. nurses; teachers and university staff). Business process outsourcing is a growing industry and now covers many areas traditionally done in-house by the financial sector or even the public sector. These companies often recruit staff from the host organization and are under pressure to demonstrate cost and efficiency changes. The workers are usually in teams and this hierarchical structure is known to generate high levels of stress. The present online survey was the first to examine wellbeing in this sector and the sample were working in the UK, the USA and South Africa.
2.1. Ethical Approval
The study was carried out with the approval of the School of Psychology ethics committee, Cardiff University and the informed consent of the participants.
The sample were recruited from two companies who were concerned with financial outsourcing. The staff of the two companies were sent an e-mail link to the online survey. The participants carried out a range of jobs (e.g. operations manager; team leader; administrator; and financial controller). One hundred and fifty one participants completed the survey (mean age = 34.1 years; 65.8% male).
2.3. The Survey
The online survey was conducted using Qualtrics software. It started with information about the study and was followed by a consent form which stated that participation was voluntary, questions could be avoided and that the volunteer could withdraw at any time. It was also stated that the data would be held anonymously and that it would be impossible to identify participants. The volunteers then completed the survey (21 questions; most with a 10 point response scale) which is shown in Table 1. On completion the volunteers were shown a debrief statement which repeated the aims of the study and thanked them for their participation.
2.4. Statistical Analyses
Initial factor analyses examined the structure of the questionnaire. Descriptive analyses then examined the frequencies in the different response categories. Following this the response scales were dichotomised into high/low groups (usually at a threshold of <7 versus 7 and above). Regression analyses were then conducted to determine the predictors of the appraisals and outcomes.
3.1. Factor Structure and Reliability
Factor analysis with varimax rotation revealed two factors representing positive wellbeing (e.g. job control/support; job satisfaction and happiness at work − Cronbach alpha = 0.81) and negative wellbeing (e.g. job demands; stress at work; and anxiety/depression due to work − Cronbach alpha = 0.65).
3.2. Descriptive Statistics
The frequencies for each response category for each question are shown in Table 1. The descriptive statistics shown in Table 1 suggest that the sample had high stress levels, and reported mental health problems. However, levels of job satisfaction and happiness were also high. These issues were examined by creating high and low categories and these are described in the next section.
3.3. Dichotomised Variables
The scales shown above are often dichotomized to produce high and low categories. For example, stress at work would be split at 7 and above (high stress) and 6 and below (low stress). Using this approach one finds that 38.8% of the sample reported high stress and 18.4 % high anxiety/depression due to work. These results confirm that the staff who took part in the survey have high levels of stress and mental health problems. If one examines positive wellbeing at work using a similar approach, 87.5% report high job satisfaction and 88.8% report high levels of happiness at work.
3.4. Predictors of Appraisals and Outcomes
Stress at work was found to be correlated with noise, demands, lack of control/ support and negative health-related behaviors. Multiple regression showed that high job demands and lack of control/support were the only significant predic-
Table 1. Questions in the survey and frequencies (%) in response categories.
tors of stress at work. Job satisfaction was correlated with positive personality, positive health-related behaviors, low noise and high control/support. Multiple regression showed that only high control/support remained as a significant predictor of job satisfaction. A similar procedure was applied to the negative (anxiety/depression due to work) and positive outcomes (happiness at work). Anxiety/Depression were predicted by lack of control/support and stress at work, whereas happiness reflected job satisfaction.
4. Discussion and Conclusion
This is the first international survey to investigate the wellbeing of business outsourcing staff. This was done with a short online survey (SWELL) which measured positive and negative aspects of wellbeing and examined the wellbeing process. The results showed that this group of workers reported high levels of stress and mental health problems but also had high levels of job satisfaction and happiness at work. Stress was predicted by job demands and lack of control/ support. Anxiety/Depression were also predicted by lack of control/support and by stress at work. This suggests that the key dimensions to assess are perceived stress at work, which will be predicted by job demands, and control/support. The presence of control and support increases job satisfaction which is then the key predictor of happiness at work.
The present methodology had a number of important features. The survey was short and multi-variate analysis allowed identification of key predictor variables. Further analyses can easily be conducted on this type of dataset. Other outcomes (e.g. absenteeism; presenteeism; performance efficiency; and musculo- skeletal disorders) could have been analysed using the same approach. In addition, a “combined effects” approach (Smith, McNamara, & Wellens, 2004) could have been adopted to examine the additive effects of risk factors. It is also possible to use the data to examine a case definition of occupational stress (Cox, Griffiths, & Houdmont, 2006) . Case definition of stress requires (1) reporting of a high level of stress, (2) reporting of a significant outcome such as mental health problem due to work, and (3) the absence of confounding factors that could account for the stress at work (e.g. stress outside of work). The same approach can be used to look at cases of high positive wellbeing.
Although the present survey has many positives there are also some future changes needed. Most of these will be very easy to achieve and will not lead to a large increase in the length of the questionnaire. For example, it would be better to ask about physical fatigue, mental fatigue and emotional fatigue separately rather than using a general fatigue question. Work-life balance could also be sub-divided into work interfering with life and, secondly, life interfering with work.
This article describes a new measure of wellbeing at work, the Smith Wellbeing Questionnaire (SWELL). This questionnaire is free to use and is shown in the paper. It has good psychometric properties, measures positive and negative aspects of wellbeing and is based on a simple model of the wellbeing process. It takes less than 10 minutes to complete and it can be combined with collection of sample specific information. The study was the first to examine the wellbeing of business outsourcing staff and the findings confirmed that they are another sector at risk of high levels of stress. Prevention and management of this stress is now a key issue for practitioners.
We would like to thank ActiveOps for assistance with recruitment of the participants.