Purpose Worker's expectations for return to working have been found to relate to return-to-work (RTW) outcomes; however, it is unclear if this varies depending upon the expected time to RTW. To advance the understanding of the relationship between expectations and RTW, we set out to answer the following research questions: Are shorter estimated times to RTW more accurate than estimates that are longer of duration? In addition, we sought to determine if there was a point in time that coincides with RTW estimates no longer being reliably related to time to RTW. Methods We utilized workers' compensation data from a large, United States-based insurance company. Injured workers' (N = 15,221) expectations for returning to work were compared with the termination of their total temporary indemnity payments. A linear regression model was used to determine if shorter lengths of expected time to RTW were more accurate. Quantile regression modelling was used to determine if there was point at which the expected time to RTW no longer reliably relates to the actual time to RTW. Results Findings indicated a positive relationship such that as the number of expected days to RTW increased, the number of days of difference (estimate error) between the actual time to RTW and the expected time to RTW also increased (β = 0.34, P < .001). The results of the quantile regression modelling indicated that for all quantiles estimated, with the exception of the quantile for estimates of 360 days, the relationship between the actual time to RTW and the expected time to RTW were statistically significant (P < .05). However, for RTW estimates of more than 14 days the relationship began decreasing in strength. Conclusion Results indicate that injured workers' expectations for RTW can be used for RTW forecasting purposes. However, it is the case that RTW events in the near future can be forecasted with higher accuracy than those that are more distant, and that in general, injured workers will underestimate how long it will take them to RTW.