Top-down vs. bottom-up change

In plain terms

Top-down change is designed and directed by senior leaders and handed to the organization to carry out. Bottom-up change starts with the people closest to the work, who identify the problem, shape the fix or build it before leaders adopt it.

Neither one wins in the research. The Atlas of the Middle's answer to this question finds that about half the studies support starting change close to the work and nearly as many find that it depends. The more useful question is which parts of a change need direction from the top, which need the people doing the work, and who carries the change between them.

Where the debate came from

Participation, 1948. Lester Coch and John French's 1948 study of worker participation, "Overcoming Resistance to Change," is the classic early case for involving workers in planning a change (Coch & French, 1948). Its conclusions have been questioned. Bartlem and Locke (1981) argued that differences in how the job changes were explained, how time studies were done and how much extra training each group got could explain the results. They suggested that perceived fairness of pay was probably the key factor. See resistance to change for the full story.

Was Lewin top-down? A common criticism of Kurt Lewin's planned approach to change is that it was top-down and management-driven. Burnes (2004) lists that among the main criticisms and argues against it, describing Lewin's commitment to resolving social conflict and the place of group dynamics and action research in his approach. See Kurt Lewin and action research.

Theory E and Theory O, 2000. In the Harvard Business Review, Michael Beer and Nitin Nohria named two theories of change (Beer & Nohria, 2000). Theory E is change based on economic value, managed from the top down. Theory O is change based on organizational capability, which encourages participation from the bottom up. They argued that companies seeking long-run results should combine the two: set direction from the top and engage the people below. In a 2001 excerpt from their book on the subject, they also warned that arbitrary, halfhearted mixing of E and O is confusing and debilitating (Beer & Nohria, 2001).

A caution on that article: it states early on that about 70% of all change initiatives fail. Hughes (2011) reviewed five published sources of the 70% figure and found no valid and reliable empirical evidence for it. The Theory E and O distinction does not depend on the figure, but the figure should not be repeated as fact.

How the two compare

The table is my own plain-language summary of the usual arguments, not a finding of the studies above, and not a finding that either column is right.

What the research says

In the Atlas of the Middle. The Atlas holds 9 studies on this question, none rated high strength, 6 medium and 3 low.

• Who starts it matters. In a survey of 1,795 respondents in 468 Dutch organizations in the early stages of a change, employees backed change most when middle managers started it and top managers carried it out. That pattern was the rarest, about 3% of cases. Change started at the top drew no more than average support, whoever carried it out (Heyden et al., 2017, rated Medium).
• Empowerment needs backing from above. At a large trucking company, a randomized empowerment program (38 treated and 30 control managers) raised unit managers' sense of control only where they felt their own bosses supported them, and it also lifted their units' measured performance (Logan & Ganster, 2007, rated Medium).
• The two can work together. At an Italian power utility's change program, top-down direction and a middle-manager change team's shared leadership worked together in every phase, with the balance shifting as the project's needs changed (Binci et al., 2016, rated Low).
• More freedom is not always better. Among 383 middle managers in a multinational, the tie between taking initiative and job performance was weaker for risk-takers, a gap that widened with high autonomy and narrowed in units with strong performance management (Glaser et al., 2016, rated Medium).

In the Evidence Library. Several syntheses bear on participation, the core of the bottom-up case.

• Participation and performance. A moderator meta-analysis of 70 field and laboratory samples found participation more strongly related to performance in group rather than leader settings, among other conditions (Sagie, 1994, rated Medium from its abstract). The abstract does not report an overall effect size.
• Involving employees in change. A systematic review published by the UK's National Institute for Health and Care Research found that a small number of studies supported involving employees in the design and implementation of changes that affect their work (Patterson et al., 2010, rated Medium after a full-text read). Its search ran in 2006, so the evidence is dated.
• Bottom-up engagement programs. Of 40 work engagement interventions, half improved engagement, two worsened it and 18 had no effect. Bottom-up approaches such as job crafting and mindfulness did best, a narrative comparison rather than a pooled test (Knight et al., 2019, rated Medium after a full-text read).
• What makes bottom-up change happen. A meta-analysis of 105 studies found that a change agent's human capital and social capital both went with endorsement of bottom-up change (Lam et al., 2026, rated Medium from its abstract). Social capital was the dominant predictor of whether the change was actually implemented. Rank, education and work experience predicted less than human or social capital themselves.

Participation is not a guaranteed win. A participatory action intervention with nurses had no effect on work engagement, and relatedness fell compared with matched controls (Knight et al., 2017, rated Medium; 45 nurses on 8 wards, not randomized).

Where the evidence is thin

The direct comparison is missing. In the studies reviewed here, no controlled trial assigns the same change to be designed by the front line or by managers and compares the results. The Loop page records this gap for Start it close to the work. Most of what exists is surveys, case studies and syntheses of correlational work, so it shows what goes with support for a change more than what causes it.

How it maps to the Unfrozen Loop

1. Week 0
   Before you start
2. 1
   Surface both ways

   Ideas from the front line are more likely to be carried out when the person raising them has strong ties. In a meta-analysis of bottom-up change, the change agent's social capital was the dominant predictor of whether the change was implemented.

3. 2
   Make sense together, top included

   The people above say out loud what the change costs. In the Atlas, a randomized empowerment program raised managers' sense of control only where they felt backed by their own bosses.

4. 3
   Start it close to the work

   Start it close to the work is the Loop's answer to this debate. The manager leads the rollout with the team, and the Atlas finds this the most studied and most contested part.

5. 4
   Review on a rhythm, then hold the pace
6. 5
   The middle’s own weather
7. Ring
   Safe to speak, and speaking changes something

Highlighted parts are the ones this idea speaks to. The mapping is my reading, not a tested finding. See the whole Loop and the evidence for each part.

Common misreadings

"Bottom-up means leaderless." In two Atlas studies, starting change close to the work went with backing from above: a randomized empowerment program raised managers' sense of control where they felt supported by their bosses (Logan & Ganster, 2007), and employee support was highest when the middle started a change and the top carried it out (Heyden et al., 2017).

"Top-down is faster, so it is better." Speed of announcement is not speed of adoption. In the Heyden et al. (2017) survey, change started at the top drew no more than average employee support.

"Participation always reduces resistance." The classic 1948 evidence is contested (Bartlem & Locke, 1981), and one participatory intervention made some outcomes worse (Knight et al., 2017). Participation is a design choice to make carefully, not a ritual.

Using it well

Before your next change, split it into three lists. First, what has to be decided at the top: the goal, the budget, the deadline. Second, what the people doing the work should design: how it runs day to day. Third, who will carry ideas between the two. Then check the third list. Lam and colleagues' (2026) meta-analysis found that a change agent's social capital was the dominant predictor of whether bottom-up change was implemented. That is correlational, but it is a reason to check the third list early.