“…Broadly speaking, the exposure measure is rather conceptual, and direct measurement may not be feasible in many situations. In practice, although the use of exposure measures is constrained by the availability and quality of data (Naci et al, 2009), various proxy measures have been developed and used in different crash frequency analyses, including population and fuel consumption (Amoh-Gyimah et al, 2017;Fridstrøm et al, 1995), traffic volume (Chiou and Fu, 2015;Heydari et al, 2017;Qin et al, 2004Qin et al, , 2006Wong et al, 2007), travel time (Chipman et al, 1993;Imprialou et al, 2016), vehicle-miles traveled (Li et al, 2003;Pei et al, 2016), potential conflict counts (Bie et al, 2005;Wong et al, 2006), and quasi-induced exposure (Huang and Chin, 2009;Jiang et al, 2014;Stamatiadis and Deacon, 1997).…”