A statistical model of the population of first-order olfactory receptor neurons (ORNs) is proposed and analysed. It describes the relationship between stimulus intensity (odour concentration) and coding variables such as rate and latency of the population of several thousand sex-pheromone sensitive ORNs in male moths. Although these neurons likely express the same olfactory receptor, they exhibit, at any concentration, a relatively large heterogeneity of responses in both peak firing frequency and latency of the first action potential fired after stimulus onset. The stochastic model is defined by a multivariate distribution of six model parameters that describe the dependence of the peak firing rate and the latency on the stimulus dose. These six parameters and their mutual linear correlations were estimated from experiments in single ORNs and included in the multidimensional model distribution. The model is utilized to reconstruct the peak firing rate and latency of the message sent to the brain by the whole ORN population at different stimulus intensities and to establish their main qualitative and quantitative properties. Finally, these properties are shown to be in agreement with those found previously in a vertebrate ORN population. This article is part of a Special Issue entitled: Neural Coding.