In this paper, the observed systematic differences in costs for teaching hospitals (TH) in Spain are examined. Concern has been voiced regarding the existence of a bias in the financing of TH once prospective budgets are introduced for hospital finance. Claims for adjustment to take into account the 'legitimate' extra costs of teaching on hospital expenditure may be well grounded. Focus is on the estimation of the impact of teaching status on average cost, using a version of a multi-product hospital cost function, taking into account some relevant factors from which to derive the observed differences. It is assumed that the relationship between the explanatory and the dependent variables follows a flexible form for each of the explanatory variables. Also, the underlying covariance structure of the data is modelled. Two qualitatively different sources of variation are assumed: random effects and serial correlation. Random variation refers to both general level variation (through the random intercept) and the variation specifically related to teaching status. It is postulated that the impact of the random effects is dominant over the impact of the serial correlation effects. The model is estimated by restricted maximum likelihood. The results show that costs are 9% higher (15% in the case of median costs) in teaching than in non-teaching hospitals. That is, teaching status legitimately explains no more than half of the observed difference in actual costs. The impact on costs of the teaching factor depends on the number of residents, with an increase of 51.11% per resident for hospitals with fewer than 204 residents (third quartile of the number of residents) and 41.84% for hospitals with more than 204 residents. In addition, the estimated dispersion is higher among teaching hospitals. As a result, due to the considerable observed heterogeneity, results should be interpreted with caution. From a policy making point of view, it is concluded that since a higher relative burden for medical training falls on public hospitals, an explicit adjustment for the extra costs that the teaching factor imposes on hospital finance is needed, before hospital competition for inpatient services takes place.