The guarded hot plate (GHP) method is the most popular way of measuring thermal conductivity of thermal insulations. However, there are large differences among the thermal conductivities measured by different GHP apparatuses in the temperature range above 100℃. In the previous study, the estimation method for thermal conductivity was proposed by making use of heat losses from side surfaces of the specimen. However, the evaluation of measurement error for this method was not shown satisfactorily. In this study, the measurement error was evaluated by dividing the specimen into extremely thin boards. It was assumed that the heat, which was flowing into the thin board, separated into two heats in the directions of thickness and plane of the thin board. It was derived that the main causes for the error of this method were the following three assumptions. Namely, the temperature gradients were uniform in the direction of thickness of specimen, and the heat conductances and temperature gradients were constant in the plane direction of any thin boards in the specimen. Furthermore, these causes of the error were evaluated experimentally.