When Teacher Training Doesn't Work

Every year, governments spend millions trying to make their teachers more effective. The logic is intuitive: better-trained teachers deliver better lessons, and better lessons produce better students. In developing countries, where teachers often have limited subject expertise and schools are under-resourced, professional development programs are widespread (Popova et al., 2022).

Vocational schools are a natural place to test whether that logic holds, and where the stakes are especially high. Unlike general academic tracks, vocational schools are explicitly meant to prepare students for specific jobs—so if teachers’ knowledge lags behind what employers actually need, the mismatch shows up directly in whether graduates can find work in their field.

Indonesia is a good test case. Globally, vocational high schools serve more than 48 million students across low- and middle-income countries (EdStats, 2022). Indonesia alone has over five million students enrolled in vocational secondary schools—half of all secondary enrollment. Vocational graduates, however, often struggle in the labor market: they are more likely to be unemployed than graduates from academic tracks. Closing that gap has made upgrading teacher skills a top priority for the government.

A promising design

The program studied in our paper had an appealing design. Rather than relying on the government itself to retrain teachers, the government contracted private-sector firms—the companies that actually hire vocational graduates—to deliver intensive, field-specific training to teachers. Lasting six to eight weeks on average, the courses were substantially longer than the typical teacher professional development program worldwide, which runs just two and a half weeks (Popova et al., 2022).

The intuition was compelling: who better to teach welding, nursing, or electronics than the firms that use these skills every day? Teachers would spend time learning up-to-date, industry-relevant techniques and bring that knowledge back into the classroom. At scale, this could make vocational schooling genuinely useful for the labor market.

We used a randomized evaluation to test this idea: some teachers were selected to participate in the program, others were not—similar to a clinical trial for a new medicine. This setup allows a clean comparison of what happened in program schools versus control schools.

The results were discouraging

The program failed to deliver on its promise:

• No increase in overall training. Teachers in program schools were nearly twice as likely to participate in the new program as those in control schools—and their exposure to the private sector increased by 16 percentage points. Yet they received no more training overall. The new program simply replaced professional development they were already doing elsewhere. Economists call this crowding out: a new intervention displaces existing activity rather than adding to it.
• No improvement in teacher knowledge. Despite intensive delivery by industry professionals, teachers’ content knowledge in their vocational fields did not improve. Our estimates rule out effects larger than 0.15 standard deviations—far below the 0.2 to 0.4 standard deviation gains typical of intensive teacher training programs in other contexts (Fryer, 2017).
• No improvement in school quality. Broader measures of school performance—including school accreditation scores—showed no meaningful change.

The program changed what type of training teachers received, not how much they learned or taught.

Why didn’t it work?

Three patterns from our survey help explain the null results. First, the training was not tailored to actual skill gaps: approximately 80% of attendees reported already being familiar with the materials covered. Second, teachers had limited support for turning training into lasting classroom change. Only 26% reported any follow-up sessions after the program ended, and more than half said they needed stronger support from school leadership. Third, control-group teachers had access to alternative, vocational-specific training elsewhere, making the new program less of an addition and more of a substitution.

One exception: more ICT use

There was a single exception to the null results: teachers in program schools showed some evidence of increased use of Information and Communication Technologies in the classroom. The ICT-focused elements of the training appear to have stuck where the vocational content did not—perhaps because ICT skills were genuinely new for most teachers, leaving less room for crowding out.

What this means

The findings carry lessons well beyond Indonesia:

• Crowding out is a hidden risk. When a new program replaces existing training rather than supplementing it, even a well-designed intervention can have no net effect. Policymakers should measure total training participation—not just take-up of the new program—before declaring success.
• Target skill gaps, don’t assume them. The fact that 80% of attendees already knew the training materials suggests the program was not matched to what teachers actually needed. Future programs should begin with careful needs assessments.
• Industry delivery is not a magic ingredient. Firms are good at doing their own jobs. That doesn’t automatically make them good at teaching. The mechanism linking industry involvement to teacher improvement needs to be understood and tested, not assumed.
• Follow-through matters as much as training. Without sustained support after the course ends, even good training rarely translates into lasting changes in classroom practice.

References

• EdStats. 2022. “Education Statistics.” World Bank. https://datatopics.worldbank.org/education.
• Fryer, R G. 2017. “The production of human capital in developed countries: Evidence from 196 randomized field experiments.” In Handbook of Economic Field Experiments, Vol. 2, 95–322. Elsevier.
• Popova, A, D K Evans, M E Breeding, and V Arancibia. 2022. “Teacher Professional Development around the World: The Gap between Evidence and Practice.” The World Bank Research Observer, 37(1): 107–136.