← People analytics

People analytics · technique

Correlation: association is not a cause

Correlation summarises association between variables.

Statistical and evaluation practice; ADM HR worked application · 2026 teaching guide · Original editorial explanation
Correlation: association is not a cause — simplified model sketchTwo measures · Observed association · Other explanations. Editorial interpretation, after Statistical and evaluation practice; ADM HR worked application (2026 teaching guide).Association ≠ causation · invented patternADM HR · ORIGINAL INTERPRETATIONStatistical and evaluation practice; ADM HR worked application · 2026 teaching guide
Original simplified sketch after Statistical and evaluation practice; ADM HR worked application (2026 teaching guide). Read the explanation for assumptions and limitations. Download SVG ↗

The core idea

Correlation summarises association between variables. Pearson’s correlation describes linear association and ranges from −1 to +1; a value near zero does not rule out a nonlinear relationship. A scatter plot can expose outliers and patterns hidden by one coefficient. Association does not establish that changing one variable will change the other. Reverse direction, common causes and selection can all produce an observed relationship.

Source and attribution [1]

Using it in practice

Plot the data and check the unit of analysis before calculating a coefficient. Ask whether observations are independent and whether a small number of points dominate the result. Distinguish an employee-level association from a site-level one. Use the finding to formulate a testable explanation, then choose a design capable of addressing the causal question if that is what the decision requires.

An example, not a reported case

Worked example · illustrative

In a fictional set of sites, higher overtime and higher absence occur together. It would be premature to conclude that overtime alone causes absence: understaffing may affect both, and absence can itself create overtime. HR reviews timing, staffing changes and employee accounts before choosing an intervention. The correlation is a clue about a pattern, not an estimate of the benefit from cutting overtime.

What to watch for

Multiple exploratory correlations can produce chance findings. A statistically significant coefficient may still be practically small. Missing data, restricted ranges and aggregation can alter the result. Do not use an observed relationship to label an individual or to claim that a proposed intervention has already been validated.

Read the plot

Look for clusters, curvature and unusual observations. Check whether the axes use comparable periods and whether the population changed. A straight-line summary can be misleading when several different groups are mixed together.

Write a cautious conclusion

State which variables were associated, in which population and period, and what remains uncertain. Replace “X drives Y” with a description of the observed pattern unless the design supports a causal claim. Identify the next piece of evidence needed for a decision.

Take it into your next conversation

Three useful questions.

  1. What question can this method answer, and what can it not establish?
  2. Are the comparison, observation period and assumptions defensible?
  3. What decision follows, and how will its consequences be reviewed?

Related terms

Go to the evidence

Sources & attribution

[1] NIST: Scatter plots and association ↗

The core idea is an original summary of the cited work. Application notes, examples and sketches are our interpretations, not quotations or reproductions of the authors’ figures. Publisher records may require access to read the full original work.

Explore People Analytics →

Published 2026-09-20 · Reviewed 2026-09-20. Editorial approach

Keep exploring

A little more context.

People analytics · technique03

A/B tests and pilot evaluation in HR

A pilot explores feasibility; an impact evaluation asks what changed because of an intervention.

Statistical and evaluation practice2 min