This paper investigates the effect of training on low-pay mobility in the UK and the Netherlands. We contribute to the literature by estimating the 'true' effect of training correcting for measurement error and transitory fluctuations--random shocks--of earnings. This is accomplished by using a random-effects multinomial logit model with a latent structure to take account of the measurement error. Our results indicate that although the countries have rather different training practices, training increases the likelihood for moving from low to higher pay and reduces the likelihood for a transition from higher pay to low pay. However, in the UK, contrary to what we expected, work-related or firm-specific training programmes but not general training programmes pay off better for the intermediate- and the higher-educated workers. No effect of training is found for the low-educated workers. The lower-skilled seem to gain less than the high-skilled from firms' investments in human capital in the UK.