Small domain estimation of business statistics by using multivariate skew normal models

Small domain business statistics are becoming important for better planning business policies. We focus on the estimation of the averages of value added and labour cost in small domains. To take into account the positive skewness in the distribution of outcomes and the correlation between them, we propose a bivariate skew normal small area model. Estimates are obtained from real survey data. The performance of the estimator proposed is evaluated on the basis of both survey data and a synthetic firm population. Results show that the model proposed increases the estimates’ reliability and that the estimates obtained make it possible to perform detailed regional economic studies.