Background Within‐family designs are increasingly used to decompose genotype–trait associations into direct and indirect genetic effects. Many such designs, including trio designs or within‐sibship designs, assume an absence of sibling indirect genetic effects. Methods We expand two well‐known molecular genetic within‐family designs, one variance component (genome‐based restricted maximum likelihood) and one trait‐based (structural equation modeling with polygenic indices), to estimate sibling indirect genetic effects, along with direct genetic effects. We link the Norwegian Mother, Father, and Child Cohort Study (MoBa) to Norway's national education database to model genetic effects on national standardized testing results at ages 10, 13, and 14, and on parent‐rated attention‐deficit hyperactivity disorder (ADHD) symptoms at ages 3 and 8 in up to 15,971 genotyped and phenotyped siblings. Results Estimates of direct and indirect genetic effects from the genome‐based restricted maximum likelihood and the structural equation modeling with polygenic indices approaches converge, albeit with the variance component estimates typically an order of magnitude greater than the trait‐based estimates. We observe no indirect genetic effects of siblings on educational performance at any age, and only slightly negative indirect genetic effects of siblings on ADHD symptoms at age 3. We argue that the latter effect might reflect parental contrasting ratings. Conclusions The results suggest that within‐family models of educational performance are unlikely to be drastically biased by an assumption of absent sibling indirect genetic effects. Combining trait‐based analyses with variance component analyses can benefit understanding of indirect genetic effects, especially when the effects are not specific to a particular mechanism.