Work and wellbeing: measurement, intervention, and performance

I intend to successively study the following stages in the next three years.
Study 1: How are wellbeing and workplace characteristics interlinked? Fortunately, there are now more data available on wellbeing than ever before (Powdthavee, 2015). Descriptive correlations have already been established using large-scale datasets like the European Social Survey (Ward & De Neve, 2017). However, there is some more potential here with new approaches and other datasets. For example, regression discontinuity designs in Gallup Daily data promise access to causal cues. In addition, social media analyses and machine learning algorithms such as natural language processing are potential techniques to be used in this research. This would increase the chances of uncovering new causal effects. For example, there is a good chance of using non-traditional social network data like the one used by Johannes Eichstaedt at the University of Pennsylvania (Kern et al., 2016). His freely available code can be used for my purposes and is familiar to me given my background in data science including machine learning and programming. All in all, the main purpose of stage 1 is the formulation of hypotheses for the later stages that look at the noteworthy observed relationships in more detail.
Study 2: Based on the results in Study 1 and general results from the literature, the initial step in this research would be a lab study. The Nuffield Centre for Experimental Social Sciences stands out as a place in Oxford with the funding and the facilities for this research. The field has seen numerous lab studies, for example by Oswald et al. (2015): in their study, the authors increased the happiness levels of participants in an RCT-like intervention, by showing them entertaining videos. They then showed that workers in the experimental group performed relatively better on tasks than those in the control group. I have already been able to gain experience in lab studies during my time at the Haas School of Business in the lab of Juliana Schroeder (Management of Organizations group). In my role as Summer Research Assistant I took part in the design and execution of several lab, field, and online experiments. The unique advantage of the lab experiment in Study 2 is that it allows for custom situations. For example, boundary condition experiments can be used to establish more information about a relationship. Another comparatively cheap and easy way of doing testing before Study 3 would potentially be the following study design, which I got to know at the Haas School of Business: with the help of crowdsourcing sites such as Amazon Mechanical Turk one can get a large number of people to work on an experiment relatively quickly (Brooks et al., 2016).
Study 3: The last stage follows from Studies 1 and 2. It would consist of conducting a field experiment on a hopefully large scale. The observed most meaningful relationships in Studies 1 and 2 would thus go into the field to be tested. Another criterion to keep in mind is the cost-effectiveness of the implemented measures. This field experiment can be particularly effective with companies using a complete feedback cycle, such as call centres or other highly automated and digitalised businesses. It can also touch on other topics of interest, such as the aforementioned link between wellbeing and firm success. An example of a well-implemented field experiment looking at the relationship between employee recognition and performance was executed with a sample of some 300 students executing data tasks (Bradler et al., 2016).
All in all, this study would allow for a detailed dissection of various effects, e.g. exactly what types of workplace characteristics influence wellbeing in which ways.