Background: The Nordic countries’ welfare state is under increased pressure. Health care expenditures are rising and expected to do so in the coming decades. It is therefore interesting to find out if the Norwegian specialist health care is cost efficient compared to its neighboring countries, or whether there are resources which could be allocated differently. The SINTEF report A12200 (Kittelsen et al., 2009a) was published in 2009. It found that Finland executes its specialist health care with higher productivity compared to Norway. This thesis builds upon the SINTEF report and extends it.The objective: To find out if the differences in hospital costs between the countries, over time, are due to differences in the productivity of the best performing hospitals and periods, which defines the efficiency frontier, or due to the distribution of efficiency behind the frontier.
Method: The parametric method of Stochastic Frontier Analysis is chosen in order to decompose the country specific frontiers. The dataset originates from the SINTEF report (Kittelsen et al., 2009a).
Results: Compared to Norway; the frontiers of Finland and Denmark are significantly different from the Norwegian frontier. The cost penalty for providing specialist health care in Norway as compared to Finland and Denmark is estimated to 26.9 % and 9 %. There is found no statistical significant difference between Sweden and Norway. There are differences between country specific frontiers but not the efficiency behind each frontier. In addition, significant differences are found within Norway which implies regional level frontiers.
Conclusion: Norway could reduce its expenses by learning from the way Finland has organized their specialist health care. The results in this thesis makes it evident that the differences in hospital costs are due to differences in country specific productivity, and not the distribution of efficiency behind the frontier. It is recommended that future research attempts to develop a generic tool for assessing the best deterministic function as well as probability density function among several competing models. It would be interesting to examine allocative efficiency by shadow price models applied to this dataset. Quality outcomes ought to be incorporated in future efficiency evaluation.