Background: The burden of work-related injuries and illnesses in the United States is great and is accompanied by substantial direct and indirect costs. Construction workers are at higher risk of work-related fatality, injury, and illness than most other working populations. Materials and Methods: By combining administrative data sources, we identified a dynamic cohort of union carpenters who worked in Washington State from 1989 to 2008, hours worked, and their workers’ compensation claims; the data were linked on an individual basis using blinded identifiers. We calculated incidence density rates for all claims, claims that were considered Occupational Health and Safety Administration (OSHA)-recordable, and claims that involved paid lost time (PLT); the latter were further stratified by the injury mechanism groupings, “Struck by object,” “Fall to a lower level,” and “Overexertion with lifting.” Poisson regression was used to assess risk by categories of age, sex, union tenure, and calendar time. Results: The cohort of 26,591 carpenters worked 192 million hours over the 20-year period. Significant declines of reported claims occurred in all categories over the observed period with the greatest decline in “Fall to a lower level” claims that involved PLT. For all claim categories, younger workers, female workers, and workers with less union tenure were at increased risk of reporting injury when adjusting for other covariates. Comparisons of incidence rates from the cohort to Bureau of Labor Statistics (BLS) incidence rates for the construction industry revealed large absolute discrepancies (higher rates in the cohort) that converged as time proceeded. Conclusions: Declines observed in injury incidence rates may reflect the combination effect of improved construction workplace safety and changes in reporting practices, among other factors. Though the absolute differences between the cohort’s incidence rates and BLS incidence rates declined significantly overtime, the relative differences between the two remained fairly stable.