The state of the practice in crash blackspot identification (BSI) has largely been driven by empirical research without much explicit attention paid to the underlying theoretical assumptions. These embedded assumptions have shaped the science of blackspot identification methodologies and developments over time. Despite the fairly extensive methodological enhancements made during the past five decades, little attention has been paid to reviewing, questioning and possibly revising these underlying theoretical assumptions. The theoretical assumptions underlying blackspot identification include: 1) crash risk can be adequately captured by the total number of crashes at a site divided by the amount of sites’ exposure to potential crashes, 2) designed roads pose differential crash risk to motorists arising from observed (operational) features of the transport network, and 3) crashes are the outcomes of a single source of risk at a site. This doctoral dissertation first reviews the theoretical assumptions underlying blackspot identification, raises fundamental questions about these theoretical assumptions and presents the associated gaps in the blackspot identification literature. These gaps include: 1) non-operational crash contributing factors and their unobserved effects have not been explicitly incorporated into the BSI, and 2) crashes may not be the outcomes of a single source of risk, but rather may be the outcomes of multiple sources of risk at a site. This focus on the underlying theory evolution, its influence on empirical work, and its reflection on remaining theory gaps serves as one of the unique contributions of this research to the literature. A more accurate underlying mechanism for explaining motor vehicle crash causation is then hypothesized as a potential solution to address the research gaps. Stated succinctly, the current theoretical assumption underlying BSI is that crashes are well-approximated by a single source of risk, wherein several contributing factors exert their collective, non-independent influences on the occurrence of crashes via a linear predictor. This PhD study first postulates, and then demonstrates empirically, that crash occurrence may be more complex than can be adequately captured by a single source of risk. It is hypothesized that the total observed crash count at a transport network location is generated by multiple underlying, simultaneous and inter-dependent sources of risk, rather than one. Each of these sources may uniquely contribute to the total observed crash count. For instance, a site’s crash occurrence may be dominated by contributions from driver behaviour issues (e.g. speeding, impaired driving), while another site’s crashes might arise predominately from design and operational deficiencies such as deteriorating pavements and worn lane markings. A multiple risk source methodology is developed to correspond with and empirically test this hypothesis. Two modelling approaches are then used to show the applicability of the multiple risk source methodology: 1) Bayesian latent mixture model, and 2) joint econometric model with random parameters and instrumental variables. Finally, the severity of crashes is explicitly incorporated into the multiple risk source methodology by extending the multiple risk source model to a joint model of crash count and crash severity. To test the viability of the methodological framework, all models are applied to a comprehensive dataset for the state controlled roads in Queensland, Australia and the results are compared with the traditional approaches. The results show that the new multiple risk source models outperform the traditional single risk source models in terms of prediction performance and goodness of fit. In addition, the multiple risk source models are able to provide more insight into crash contributing factors, their impact on the total crash count and their impact on the crash count proportions generated by each risk source. It is found that the parameters of the joint model of crash count and crash severity are moderated by the correlation between these two models and therefore, the total risk at a site can be adequately recognized by crash count and severity, simultaneously. Over all, the findings of this research indicate that decomposing the total crash count into its constituent components, separating the risk sources and incorporating crash severity into the overall framework leads to efficient, cost-effective identification of crash blackspots.