Abstract: Decision-making in the‘newsvendor problem' using laboratory experiment is studied. In comparison with the literatures which suppose uniform customer demand conditions, the orders still show a "pull-to-center" effect when the demand for the product follows a normal distribution. The participants' order quantities are also affected by demand samples. The order quantities are higher and the participants attain higher profits when the mean of the demand sample is higher. Keeping the same demand samples, it can improve performance significantly in the low cost condition if providing history profits of all the feasible choices for the participants. The order quantities move towards the optimal value only with the decision-maker's experience improving in the high cost condition, which shows decision-makers have a learning capability to the giving information. A dynamic panel data model with ordering data is constructed. The GMM results from the dynamic panel model indicate that the parameters of anchoring on the previous order quantity and adjusting ordering deviation are significantly positive, which show that the order quantities of adjacent two periods are closely related and ordering is a dynamic adjustment process. The results also show that subjects have a strong demand chasing tendency which is unrelated to cost but related to whether feedback about history profit statistics is provided.