Does simulation training work for school leaders?

Most new leaders start underprepared

In New Zealand, the Education Review Office surveyed 596 new principals. Only 27% felt prepared or very prepared when they started (ERO, 2023).

The pattern reaches well beyond one country. UNESCO reports that "almost half of principals in richer countries do not receive any training before appointment" (UNESCO, 2024). Only 31% of countries regulate the induction of new principals.

In the UK, "most headteachers do not feel fully prepared when they first take on the role" (Greany et al., 2025). In the US, only 46% of principals reported an internship that gave them real leadership responsibility (Darling-Hammond et al., 2022).

New leaders describe a sudden jump to sole responsibility. Spillane and Lee interviewed 17 new Chicago principals three months into the job (2013, working-paper draft). The volume and the unpredictability of problems were each salient for 59% of them, and 47% reported early loneliness.

Support makes a measurable difference to readiness. New Zealand principals who had coaching or mentoring, or a postgraduate leadership qualification, were twice as likely to feel prepared (ERO, 2023).

Simulation works, and the comparison sets the size

The largest evidence base comes from the health professions. Cook and colleagues pooled 609 studies with 35,226 trainees (Cook et al., 2011).

Against no intervention, effect sizes were large: 1.20 for knowledge and 1.09 for process skills. They fell to 0.81 for behaviour with real patients and 0.50 for direct effects on patients.

Those figures carry bias. Adjusting for missing studies cut the knowledge effect from 1.20 to 0.86, and two-thirds of the studies tested one group before and after training, with no comparison group.

Against other teaching, the advantage narrows. In emergency medicine, effects against other instruction were 0.33 or less and not significant (Ilgen et al., 2013).

Outside medicine, the evidence thins. Across 145 higher-education studies, 126 of them in medicine and seven in teacher education, simulation gave g = 0.85 for complex skills (Chernikova et al., 2020). Role play gave 0.63, and communication and teamwork gained least, at 0.44 and 0.50.

In a US nursing trial, ten programmes randomised 847 students (Hayden et al., 2014). Replacing up to half of clinical hours with simulation made no significant difference to clinical competence, knowledge or licensing pass rates, up to six months into practice.

What makes simulation work

The evidence for debriefs is robust. Across 46 samples, debriefs improved performance by about 25%, an effect of d = 0.67 (Tannenbaum and Cerasoli, 2013).

A later meta-analysis of 61 studies found d = 0.79 for after-action reviews, read here through a reproduced abstract (Keiser and Arthur, 2021). A practitioner research summary drawing mainly on the same meta-analysis says reviews are "particularly effective when the review includes objective records of the event" (Paul, 2021).

Realism matters less than people expect. Across 24 health studies, high- and low-fidelity simulators differed by 1% to 2% on average (Norman et al., 2012). In Cook's words, "the appearance of the simulator is less important than a proper representation of the key steps in the task" (Cook, 2013).

Teacher education has little rigorous causal evidence, but one series of randomised trials tests coaching between sessions. Five randomised trials at one US university compared five minutes of directive coaching between mixed-reality sessions with self-reflection. Coaching gave an average 1.44 SD on blinded ratings of teaching quality (Cohen et al., 2021, working paper).

Those gains were measured inside the simulator. A 2026 scoping review repeats an earlier warning about mixed-reality simulation, or MRS. The warning says "one cannot claim that skills developed in MRS generalize to real classrooms" (Mikeska and colleagues, 2019, as cited in Anderson and Calandra, 2026).

School leadership simulations: an old idea, little outcome data

Simulation entered school leadership early. In 1958–59, 232 US elementary principals each spent a week running a simulated school called Whitman. The project set out to develop criteria of success in school administration, and was also expected to improve principal selection (Shin, 1989; Hemphill et al., 1962).

Selection research moved on to principal assessment centres. A 1992 meta-analysis of 34 validity studies of one such process, the NASSP Assessment Center, found its criterion-related validity "very low" (Pantili, 1992).

Teaching simulations followed. Virginia Commonwealth University's Project ALL stages "a year in the life of a school principal" through more than 400 story elements (Mann and Shakeshaft, 2013). The abstract of its study reports no outcome data (Mann et al., 2011).

A 2025 study followed branching simulations through one US university-district partnership, where 27 interns completed 15 simulations each (Azukas et al., 2025). Its authors name fixed choices, no fit to district policy or state law and little practice in dialogue as limits.

Reviews of the field are blunt. Research on problem-based learning for school leaders is "immature" (Hallinger and Bridges, 2017). Studies of digital simulations and serious games are limited by "weak research designs and methods" (Hallinger et al., 2025).

Role play with AI, from language-model students to AI avatars and chatbots, now runs in pilots and projects in the US, Australia, Norway and Switzerland. No randomised trial of it has yet measured real teaching or leadership outcomes.

Where the evidence runs out

• No study links simulation-based leadership preparation to leaders' practice or students' outcomes.
• Most gains in teacher and leader simulation are measured inside the simulator. Reviews of leadership simulation find weak research designs.
• No study has tested replaying a scenario from a save point. Rapid cycle deliberate practice, which pauses and rewinds a scenario for coaching, has not proved superior to traditional simulation (Ng et al., 2021).
• No study compares AI-led debriefs with human-led ones.
• Most of the evidence comes from the United States.

Even the size of the prize is debated. A 2021 synthesis of six US studies found that a principal one standard deviation more effective added 0.13 SD a year in maths (Grissom et al., 2021). A 2024 re-estimate puts persistent effects nearer 0.03 SD (Bartanen et al., 2024, read here through a reproduced abstract).

What this means for programme designers

1. Pitch expectations at the smaller effects found against active teaching, and for skills such as communication and teamwork.
2. Put design effort into the debrief, and give it an objective record of what the trainee did.
3. Pair simulation with coaching or mentoring, which New Zealand's survey links to readiness.
4. Judge learning on new scenarios, because repeating one scenario lets trainees anticipate its events.
5. Keep simulation scores out of selection decisions, given how weakly assessment centres predicted performance.
6. Build around one system's procedures first. Adapting a change-management simulation for Chinese school leaders followed an eight-stage research-and-development cycle (Tang, 2015).