Note(s): ***, ** and * indicate significance at 1, 5 and 10 percent, respectively. t statistics are reported in parentheses. All regressions are estimated with Pooled OLS. The dependent variable is TFP, estimated with the Olley and Pakes (1996) method in columns 1 and 3, and with the Levinsohn and Petrin (2003) method in columns 2 and 4. All regressions include country, year and quarter dummies. The Hausman test assesses the null hypothesis that random effects are preferable to fixed effects. In the Breusch–Pagan RE test the null hypothesis is that random effects are not present. In the Wooldridge serial correlation test the null hypothesis is the absence of serial correlation. In the Breusch–Pagan heteroscedasticity test the null hypothesis is the absence of heteroscedasticity. The VIF, not reported in full due to space constraints, tests the presence of multicollinearity. The AIC and BIC are the Akaike and the Bayesian Information Criterion, respectively

Source(s): Authors’ own calculations based on data from the AI Index Report 2023 (Maslej et al., 2023) and OECD – Quarterly National Accounts