Not accounting for collar loss when using observations of collared geese to estimate survival can result in an underestimation of the survival rates and a loss of precision. As part of a study of goose populations in the Western Canadian Arctic, we determined collar retention rates for “small” Canada Geese (mainly Branta canadensis parvipes) and Greater White-fronted Geese (Anser albifrons) that had been collared in one year and recaptured in a subsequent year. We recaptured 5.3% (133/2504) of the Canada Geese and 7.3% (373/5098) of the Greater White-fronted Geese that we collared. Annual retention rates for thin (0.08 cm), double-wrapped collars placed on Canada Geese were low for males (0.650 ± 0.052 SE) but significantly higher for females (0.826 ± 0.044). Retention rates of thick (0.16 cm), single-wrapped collars placed on Greater White-fronted Geese were high for males (0.982 ± 0.007) and even higher for females (1.000 ± 0). Retention rates did not vary with collar age, but our study probably did not continue long enough, or have large enough samples, to adequately assess this. Retention rates for Canada Goose collars depended on manufacturer, with rates as low as 0.282 ± 0.111 for males and 0.566 ± 0.114 for females for collars produced by one supplier. Our data suggest that collar loss was a significant problem for Canada Geese, but not for Greater White-fronted Geese. We recommend that other researchers using observations of collared individuals to calculate survival should also assess if estimates need to be adjusted because of collar loss.