Population data and apparent prevalence data from the Salmonella surveillance and control program in pigs (NSSCP) from 1998 and 1999 were used in a simulation model to evaluate the efficacy of the program. The model consists of three parts: modelling of individual prevalence at the abattoir (abattoir part), modelling of the number of sampled herds of different sizes when carcasses are randomly sampled at the abattoir (sampling strategy part) and finally, modelling of the within herd prevalence (within herd part). A total of 136,550 sows and 2,866,550 finishing pigs slaughtered, 4446 herds and 11 herds positive for Salmonella in 1994/1995-2000 were included in the abattoir part, sampling strategy part and the within herd part of the model, respectively. The abattoir part showed an average estimated prevalence of Salmonella in sows and finishing pigs of (median) 0.4% (5-95 percentiles = 0.03-2%) and 0.1% (0.04-0.2%) respectively. The estimated number of infected sow carcasses that entered the market was 502 (37-2157) while the estimated number of finishing pig carcasses was 2919 (1218-5771). The probability of being sampled for the 10% smallest herds was (mean) 1.9% (1.6-2.2), to 25% (24.7-26.5%) for the 10% largest herds. The within herd prevalence was estimated to be from 1% to 4% for Norwegian pig herds. The conclusions drawn from this evaluation are that the NSSCP does not have any significant consumer protection effect, and that the documentation could be done more effectively using the herd rather than the individual animal as the unit of sampling. Sampling should focus on the larger herds supplying most of the meat on the market and on herds that produce breeding sows and piglets and thus can contribute to the spread of Salmonella among herds.