Development accommodation statistics for small areas have been largely neglected in tourism literature. The challenges are high variance and/or bias of estimates. Therefore, tourism management policies and business strategies for regional tourism areas have not been well supported by good statistics. This study utilizes accommodation data of Japanese DMOs as a case study. Regression is used to compare to a traditional linear estimator. Each estimator is tested with two sampling schemes, namely simple random sampling and clustering sampling. The results demonstrate that the traditional estimation method, which has been utilized widely in tourism literature, failed in developing tourism accommodation statistics for small-scale regions. Integration of regression and clustered sampling scheme made it possible to provide a more accurate estimate, i.e., less bias and variability. The results make the policy formulation process possible for small tourism areas.