The Eötvös-type torsion balance is a well-known gravity gradient instrument for the regional or local geophysical exploration. As a key issue for the gravity gradient measurement, the estimation of the systematic effects in the instrument is rarely discussed as yet. This lack may lead to inconsistencies in the actual gravity gradient measurements. Among the systematic effects, one of the more obvious is the highorder gravitational gradient coupling, which is usually neglected in the measurement equation. This paper focuses on the estimation of such an effect. The coupling between the high-order gravitational gradient and the pendulum used in the instrument is charactered on the basis of the multipole expansion. It is presented that the coupling effect in the two versions of the commonly used Eötvös-type torsion balance can contribute a measurement bias of gravity gradient as large as approximately 20 E, approximately 10 E, respectively. The evaluation shows that the high-order gravitational gradient effect can not be ignored in the gravity gradient measurement by the Eötvös-type torsion balance. In addition, such an analysis method for the high-order gravitational gradient effect can also be applied to other types of gravity gradiometer or even gravimeters if required.