This is a Medium Rare template. Explore our complete range of designs here.
top
This article explains the difference between variable-centred and person-centred approaches to analysing employee survey data, with examples from research. Variable-centred analysis —such as regression— describes relationships between variables across the workforce. Person-centred analysis — such as cluster analysis — identifies distinct subgroups of employees with similar response patterns. Both approaches answer different questions about the same data.
Variable-centred analysis is the traditional and dominant approach in occupational psychology and the foundation of most employee survey reporting. Imagine your survey results laid out as a spreadsheet, with employees in the rows and questions in the columns. Variable-centred analysis works on the columns. It asks how the answers to one question relate to the answers to another. Do high scores on manager support tend to go with high scores on engagement? Which factors are most strongly associated with engagement on average? Which job characteristics protect against burnout? The unit being studied is the variable, and the findings describe relationships between things that have been measured.
Two influential papers from the engagement literature illustrate variable-centred analysis.
Crawford, LePine and Rich (2010), in the Journal of Applied Psychology, brought together findings from 64 studies to test the Job Demands-Resources model — the leading framework in the engagement literature for explaining what leads to engagement and what leads to burnout. Their analysis distinguished between two kinds of job demands. Challenge demands, such as workload pressure and time urgency, were demanding but offered an opportunity for growth. Hindrance demands, such as role ambiguity and organisational politics, were demanding without that upside. Job resources — autonomy, feedback, social support — were the third category. The finding was that resources predicted engagement positively, hindrance demands predicted burnout, and challenge demands predicted both engagement and burnout simultaneously — a pattern that resolved long-standing inconsistencies in how the JD-R literature had treated job demands as a single category.
Christian, Garza and Slaughter (2011), in Personnel Psychology, took a broader view of the same field. Drawing on a wider set of studies, they mapped the full network of factors that lead to engagement and the outcomes engagement predicts — establishing that engagement is empirically distinct from job satisfaction and organisational commitment, that the strongest predictors of engagement are task variety, autonomy, and transformational leadership, and that engagement predicts task performance and discretionary effort over and above those other measures.
Both are textbook variable-centred analyses. They ask which variables matter and how strongly, and they answer those questions by averaging relationships across thousands of respondents in the case of single studies and tens of thousands across meta-analyses.
Person-centred analysis asks a different kind of question. Rather than looking at how variables relate to each other across the workforce, it looks at how people group together based on their patterns of response. Are there subgroups of employees whose answers across the survey look similar to one another and distinctly different from other groups? If so, what do those subgroups look like, and how do they differ on important outcomes? The unit being studied is the person, and the findings describe types of respondents.
Returning to the spreadsheet image, person-centred analysis works on the rows rather than the columns. The most common techniques are Latent Profile Analysis (LPA), Latent Class Analysis (LCA) and cluster analysis.
Mäkikangas and Schaufeli (2021), in the Journal of Vocational Behavior, applied Latent Profile Analysis to job crafting — the small adjustments employees make to their own roles to better suit their preferences and strengths. Drawing on responses from 419 Finnish managers, they identified subgroups of respondents who used job crafting strategies in similar ways.
Four profiles emerged. Average crafters, who made up 47% of the sample, used all the job crafting strategies at moderate levels. Avoidance-oriented crafters, 30% of the sample, mainly tried to reduce demands they found difficult — and made little use of any of the more proactive strategies. Approach-oriented crafters, 19% of the sample, did the opposite: they actively sought out challenges and resources, and avoided the reductive strategies. A smaller fourth group of self-oriented task crafters — 4% of the sample — sought out challenges in their tasks but neglected the relational and cognitive sides of crafting.
When the authors compared these profiles on work engagement and person-job fit, the differences were substantial. Approach-oriented crafters had the highest scores on both. Avoidance-oriented and self-oriented task crafters had the lowest. The Average crafters sat in between. The same managers had answered the same questionnaire a variable-centred analysis would have used. A variable-centred analysis would have produced averages and correlations across the whole sample. The person-centred analysis produced four distinct response patterns instead.
A second example comes from work by Kanitz and colleagues (2024) in the Journal of Applied Psychology. They examined how employees respond to diversity, equity and inclusion initiatives — a question that has typically been treated in binary terms: employees either support or resist. Their analysis, drawing on more than 1,600 US employees across three studies, used Latent Profile Analysis to identify four distinct response profiles instead. Excited supporters responded positively across cognitive, emotional and behavioural dimensions. Calm compliers were neutral across all three. Torn shapers held mixed views, supporting the principles but questioning the implementation. Discontent opponents responded negatively throughout. The four profiles emerged in the first study and were replicated in the second and third — and they predicted different behaviours toward the initiatives, from active championing to silence to subtle resistance.
The variable-centred and person-centred approaches answer different questions. One asks which factors matter most across the workforce. The other asks whether the workforce contains distinct subgroups whose experience of work is qualitatively different. They sit alongside item-level analysis of low scores, comparisons to internal and external benchmarks, and comment themes — different lenses on the same data, each answering a different question.
What is the difference between variable-centred and person-centred analysis?Variable-centred analysis describes relationships between variables across the workforce — for example, how strongly manager support relates to engagement on average. Person-centred analysis identifies subgroups of employees with similar response patterns across multiple variables — for example, distinct types of employee whose experience of work is qualitatively different.
What is Latent Profile Analysis?Latent Profile Analysis (LPA) is a statistical technique that identifies subgroups within a dataset based on patterns of response across continuous variables. It is the most common person-centred technique used with Likert-style employee survey data. Related techniques include Latent Class Analysis (LCA), which is used with categorical indicators, and cluster analysis, which is the older distance-based predecessor of both.
When should I use person-centred analysis in employee surveys?Person-centred analysis adds value when there is reason to believe the workforce contains qualitatively different subgroups — for example, in analysing responses to a change initiative, a wellbeing programme, or a DEI strategy. It complements rather than replaces variable-centred analysis. Programmes that rely solely on aggregate driver analysis may miss subgroup-level findings that matter for action planning.
Does person-centred analysis replace driver analysis?No. Driver analysis is variable-centred and answers a different question — which factors are most strongly associated with engagement on average. Person-centred analysis answers whether the workforce contains distinct subgroups. Both have a role in a well-designed survey programme.
Christian, M. S., Garza, A. S., & Slaughter, J. E. (2011). Work engagement: A quantitative review and test of its relations with task and contextual performance. Personnel Psychology, 64(1), 89–136. https://doi.org/10.1111/j.1744-6570.2010.01203.x
Crawford, E. R., LePine, J. A., & Rich, B. L. (2010). Linking job demands and resources to employee engagement and burnout: A theoretical extension and meta-analytic test. Journal of Applied Psychology, 95(5), 834–848. https://doi.org/10.1037/a0019364
Kanitz, R., Reinwald, M., Gonzalez, K., Burmeister, A., Song, Y., & Hoegl, M. (2024). Supportive, resistant, or both? A person-centric view on employee responses to diversity initiatives. Journal of Applied Psychology, 109(10), 1635–1658. https://doi.org/10.1037/apl0001190
Mäkikangas, A., & Schaufeli, W. (2021). A person-centered investigation of two dominant job crafting theoretical frameworks and their work-related implications. Journal of Vocational Behavior, 131, 103658. https://doi.org/10.1016/j.jvb.2021.103658
Let’s start a conversation about how employee surveys can help you develop a workplace where people and performance grow together.
