The manager is not where your engagement problem lives

Engagement comes back down. Somebody pulls the report, the index is off three points, and by the end of the week there is a plan. The plan is manager training.

It is always manager training. Not because anyone has established that managers are the problem, but because the manager is the only thing in the report that looks like something you can do. Items about the leader are on the survey. Items about the work are on the survey. Items about the person answering are almost never on the survey at all, because they got averaged into the score and disappeared.

So here is the claim, and it is not a comfortable one: what a person brings with them to the job predicts whether they engage better than who manages them does. It is not close, it replicates, and it has been sitting in the literature for fifteen years.

The numbers

Christian, Garza and Slaughter pooled 90 samples covering 63,813 people in Personnel Psychology in 2011. Their table of antecedents is the cleanest statement of the problem anyone has published, because it puts the person and the situation in the same column.

Proactive personality correlates with engagement at .44. Conscientiousness at .42. Positive affectivity at .43.

Transformational leadership: .27. Leader-member exchange: .31.

Between those two groups sits the work itself. Task variety comes in at .53, task significance at .51, autonomy at .39, feedback at .33. Which is to say the design of the job beats the person on its best items, the person beats the manager on almost all of them, and the manager finishes third.

Mazzetti and colleagues ran the same question through a different frame in 2023: 533 correlations, 94 studies, 119,420 people, organized by the job demands-resources model. Personal resources came first at .48. Development resources at .45. Job resources at .37. Social resources, the category that holds supervisor support, at .36.

Both of those are cross-sectional, which is the standard and entirely fair objection. Correlation at one point in time cannot tell you which way the arrow runs, and an engaged person might well describe themselves as more proactive because they are engaged.

Lesener, Gusy, Jochmann and Wolter answered that one in 2019 by restricting themselves to the 55 studies that measured people twice. Over time, resources at the organizational level predict engagement at β = .13. At the group level, .06. At the leader level, .05.

What this does not say

It does not say managers are irrelevant. A .27 correlation is a real effect, it is bigger than most things anyone can buy, and a bad manager can wreck a good job faster than a good manager can rescue a bad one. The asymmetry matters.

It does not say you can hire your way out. Proactive personality predicting engagement at .44 leaves about 80 percent of the variance somewhere else, and any selection system built on the assumption that engagement is a trait will produce a workforce that is disappointing in a new way.

And it does not say the evidence is settled. Bailey, Madden, Alfes and Fletcher walked through 214 engagement studies in 2017 and their own conclusion was that the field leans too hard on cross-sectional self-report to support causal claims. That is a criticism of three of the four papers on this page, including two I have just quoted at you. I am citing the review that undercuts my own citations because leaving it out would make this piece worse and easier to dismiss.

What survives all of that is the ranking. Four independent syntheses, three research traditions, one consistent order: the person, the work, then the leader.

Why the plan always comes out backwards

Because of what the instrument can see.

A standard engagement survey asks about the manager, about clarity, about workload and about recognition, and then it reports averages. Individual differences are not a category on the report. They are the noise the average was computed to remove.

So the survey produces a picture of the environment, correctly, and then a room full of capable people looks at that picture and picks the biggest lever visible in it. The manager is the biggest lever visible in it. The largest predictor is not in the frame, and no amount of staring at the frame will put it there.

This is also why the same three-point dip produces the same workshop in year one, year three and year five. Nothing in the measurement was ever going to name a different problem.

What changes on Monday

Three questions, and your current survey can probably answer two of them.

Does engagement track with who you brought in? Cut the score by hiring cohort, by source, by tenure band. If the people who came through one channel are consistently three points lower two years in, you have a selection finding, and no manager workshop is going to touch it.

Does it track with how the work is built? Cut it by role, span of control, shift pattern, and how many systems somebody has to touch to finish a task. This is where Lesener says the returns are, and it is the cut that almost nobody runs because it implicates operating design rather than people.

Does it track with who is running it? Cut it by manager. Do this one third rather than first, and hold it against the other two, because a manager effect that vanishes once you control for role design was never a manager effect.

The third question is the only one most organizations ask. It is also the one the evidence ranks last.

If your instrument cannot answer the first, that is what to fix before the next cycle. The action plan can wait. A survey that measures the environment and calls it the whole picture will keep recommending the manager, forever, and it will keep being partly right, which is the most expensive kind of wrong.