Computer vision researchers prefer to estimate the age from face images due to informative facial features. Estimating the age from face images becomes challenging when people are far away from camcorders or occluded. As the unique biometric feature that can be perceived efficiently at a distance, gait can be an alternative way to predict the age in case that face images are not available. However, existing gait-based classification or regression methods ignore the ordinal relationship of different ages, which is an important clue to the age estimation. In this paper, we proposes an ordinal distribution regression with a global and local convolutional neural network for gait-based age estimation. Specifically, we decompose the gait-based age regression into a series of binary classifications to incorporate the ordinal information of the age. Then an ordinal distribution loss is proposed to take inner relationship among these classifications into account by penalizing the distribution discrepancy between the estimated and the ground-truth. In addition, our neural network consists of a global and three local sub-networks, which is capable of learning the global structure and more local details from head, body and feet of gait, respectively. By comparing with the state-of-the-art methods of gait-based age estimation, this paper highlights, experimentally, that the proposed approach has a better predictive performance on the OULP-Age dataset.