Carcass counts notably underestimate avian collision rates due to three main bias sources: imperfect detection, carcass removal by scavengers and carcass dispersion in unsearched areas. We assessed these sources of bias at electric lines of two Canary Islands, Lanzarote and Fuerteventura, quantifying the factors influencing them. We also carried out a cost-effectiveness assessment of carcass search done perpendicularly to electric line axis. We surveyed 230km of three types of electric lines (high-voltage, medium voltage and telephone lines) during three periods (July 2015, November-December 2015 and March 2016) searching for collision fatalities (N = 431), recording the species, the carcass distance from the electric line, mean cable height, carcass detection distance and decomposition state. In addition, we carried out a disappearance rate experiment to estimate carcass removal by scavengers. A generalised least squares model was used to analyse dispersion distance of carcass from electric lines, in relation to species body mass, mean cable height and line typology. Detection probability functions were fitted to estimate carcass detectability, incorporating body mass, decomposition state and habitat structure as covariates. A Generalised Mixed-Effects model was carried out to analyse carcass disappearance in relation to time elapsed since carcass placement, carcass size, season and island. Dispersion distance decreased with body mass and increased with cable height, being further at high-voltage lines. Overall, detection probability was 0.134, increasing with carcass size, decreasing with decomposition state and being lower in rocky areas which offered a significant challenge when walking through rough terrain. Disappearance rates differed between islands probably due to differences in avian scavenger abundance, increased with time elapsed and decreased with bird size. This study provides correction factors to obtain unbiased estimates of avian mortality rates within sparsely vegetated landscapes. Moreover, it identifies a 27m threshold distance at which the cost-effectiveness of searching for carcasses is optimised.—Gómez-Catasús, J., Carrascal, L.M., Moraleda, V., Colsa, J., Garcés, F. & Schuster, C. (2021). Factors affecting differential underestimates of bird collision fatalities at electric lines: a case study in the Canary Islands. Ardeola, 68: 71-94.