Automation, recessions, and jobs: What Okun’s Law tells us

How does automation shape the short-term relationship between economic performance and employment? This question matters not only to researchers but also to policymakers and employers who worry that rising robot adoption may lead to job losses - especially during economic downturns. A classic tool for thinking about this link is Okun’s Law, which describes how changes in output usually translate into changes in unemployment: when GDP falls, joblessness tends to rise, and when GDP grows, unemployment tends to decline.

Our recent study revisits this fundamental relationship in a world where automation technologies are spreading rapidly. The surprising conclusion is that automation does not necessarily make recessions more harmful for workers. In fact, we find that higher robot adoption can cushion the rise in unemployment during economic contractions.

What we examined

Okun’s Law has long served as a benchmark for assessing labour market responses to economic cycles. But much of the concern surrounding automation assumes that new technologies weaken the connection between output and employment, contributing to what some call “jobless recoveries.” We asked whether there is evidence for this assumption in the data.

To explore this question, we analysed how Okun’s Law behaves in the presence of growing robotisation. We assembled a panel dataset covering 35 OECD countries from 1996 to 2020—a period including several distinct business cycle phases, the Global Financial Crisis, and the early months of the COVID-19 pandemic. Using various model specifications, fixed-effects estimators, and a series of robustness checks, we estimated how unemployment responds to changes in output depending on a country’s degree of automation.

What we found

The results point to a consistent pattern:

• Countries with higher robot adoption experience a smaller rise in unemployment when output declines. In other words, automation appears to make recessions less damaging to employment.
• Automation does not slow employment recovery when growth resumes. We find no evidence that economies with more robots suffer from jobless recoveries.
• These effects are visible across different types of economic shocks and survive a wide range of statistical tests.

Put simply, the usual “output–employment” relationship behaves differently in economies that have invested more in automation. When downturns occur, workers are not disproportionately pushed out of jobs in more robotised labour markets. And when the economy begins to expand again, employment recovers in line with growth rather than lagging.

Why would automation help?

The finding may sound counterintuitive because automation is still widely viewed as a job-destroying force. But the picture is more nuanced. Robot adoption can promote productivity and help firms remain competitive even in difficult times. This may reduce the pressure on employers to cut jobs when output falls. Automation also complements many forms of human labour, enabling workers to shift toward tasks where their skills matter more. These mechanisms can help retain employment during recessions, rather than replace workers.

Our analysis does not deny that specific jobs may be displaced by technology or that transitions can be painful for some workers and communities. Instead, it shows that the net effect of automation on the business-cycle dynamics of employment has been more protective than harmful in OECD countries during the period studied.

Why this matters for policy

There are three key implications for ongoing debates about technology and the future of work:

1. Automation should not be assumed to worsen job losses during recessions. The evidence suggests the opposite: it can increase labour market resilience by dampening the usual spike in unemployment when output declines.
2. The data do not support the concern about jobless recoveries. Employment growth continues to track output growth even in highly robotised economies.
3. Technological investment and labour policy need to work together. To fully realise the potential benefits of automation, complementary policies are essential - such as reskilling initiatives, education systems that update workers’ competences, and labour market institutions that support smooth transitions.

These lessons are especially relevant at a time when many governments are considering how to strengthen their economies against future shocks. If confirmed by further research, our findings suggest that supporting innovative technology adoption could be one way to protect employment during downturns - provided that workers are equipped to adapt to changing skill demands.

Looking ahead

The relationship between technology and labour markets is evolving, and Okun’s Law provides a valuable lens for observing those shifts. Our results show that automation’s effects are not predetermined: they depend on how technologies are adopted and managed.

Rather than assuming robots will eliminate jobs, we should ask how they can help create more robust labour markets—ones that can better withstand economic cycles and support both firms and workers through change.

For reference:

• Exploring the impact of automation on employment during expansions and contractions: An examination of Okun’s Law
• Just reallocated? Robots displacement, and job quality
• Okun's Law: Fit at 50?
• Robots and Firms
• Jobs, Productivity and the Great Decoupling
• Job Polarization and Jobless Recoveries
• Productivity growth over the business cycle: cleansing effects of recessions
• Skills-displacing technological change and its impact on jobs: challenging technological alarmism?
• How the G20 Can Build on the World Economy’s Recent Resilience