Small cell networks have recently been proposed as an important evolution path for the next-generation cellular networks. While such approach has the potential of meeting the growing network throughput requirement, the energy efficiency of small cell networks is of great concern as the base station (BS) density will be significantly increased. The objective of this paper is to analyze the energy efficiency in small cell networks. To do so, we adopt a random spatial network model, where BSs and users are modeled as two independent spatial Poisson point processes (PPPs). We shall derive analytical results for the network energy efficiency, which show that the BS power consumption model plays a critical role. In particular, it will be shown that increasing the BS density can actually improve the energy efficiency if the BS power consumption that is not related to signal transmission is less than a certain threshold. By comparing the cases between single-antenna and multi-antenna BSs, we find that single-antenna BSs provide a higher energy efficiency if the circuit power is larger than a threshold. Simulation results will demonstrate that our conclusions which are based on the random network model also hold in a regular grid-based model.