In the Burwash area of north‐central Ontario, Canada, expansion of the Trans‐Canada highway from 2 to 4 lanes was accompanied by installation of a range of wildlife collision‐mitigation infrastructure (e.g., exclusion fencing, underpasses). To assess the overall effectiveness of these measures, we monitored the spatial distribution and mortality rates of elk (Cervus canadensis) prior to and following highway expansion, distinguished by season (winter, snowfree) and corridor‐type (highway, railway). We measured herd‐level risk by the proportion of positions falling within 200‐m railway and highway buffer zones using Bayesian methods. Spatial analysis confirmed that there was a distinct northward shift in the winter distribution of elk following construction, situating the elk past the north end of the exclusion fence. This increased the herd's exposure to highway traffic by 3.6 times (proportion of points before = 0.0041 ± 0.002 [SE], after = 0.0147 ± 0.003, P = 0.005), and resulted in a more than 2‐fold increase in elk road mortality from 0.6 elk/yr/20 km during 8 years prior to implementation to 1.5 elk/yr/20 km during 8 years after implementation. Exposure to railways remained unchanged and consistently higher than highway exposure regardless of season (e.g., post‐mitigation, winter proportion of points = 0.0453 ± 0.005), matched by consistently high mortality counts (proportion of points before = 6.4 elk/yr/20 km, after = 6.6 elk/yr/20 km). Our results demonstrate that while wildlife‐vehicle collision mitigation is generally beneficial to wildlife and humans, failure to account for the local characteristics of wildlife populations can lead to suboptimal mitigation designs that reduce their effectiveness and lead to unintended wildlife impacts.