Introduction: Cancer incidence rates in Canada have changed over the past 20 years due to changes in carcinogenic exposures and detection practices. We examined whether period or cohort effects are driving trends for 24 cancer sites across sociodemographic strata. Methods: The Canadian Census Health and Environment Cohorts are data linkages of respondents of the long-form census and National Household Survey (NHS) to the Canadian Cancer Registry from 2006 to 2021. We fitted age-period-cohort models to cancer incidence; cohort effects were stratified by sex, income, education level, immigration status, and racial group; period effects were stratified by age (=50 years). Results: There were 806,510 incident primary cancers diagnosed in 17.8M individuals over 140.7M person-years. Most cancer sites have experienced an increasing incidence by period effect countered by a decreasing incidence by cohort effect for birth cohorts born over the 20th century, with a trend reversal for some cancer sites in cohorts born in the 1980s-1990s. Cohort effects differed by sex, income, education level, immigration status, and racial group. There were larger relative period-effect increases in <50 year-olds for breast (female), colorectal, kidney & renal pelvis, and uterine cancer incidence, but not other cancers. Conclusions: Declining cohort effects likely reflect ongoing efforts in tobacco control, cervical and colorectal cancer screening, and improvements in living conditions across the 20th century. The causes of period increases in cancer incidence may include changes in detection practices from screening and incidental findings. The obesity epidemic has likely contributed to both period and cohort effects.