I do not know how you did your regress. If you used size as dummy variables, the explanation of con_ would be the mean of educ for size1 (size1 is the reference group). Please check your F value and corresponding p value to see if you can or cannot reject your null hypothesis. But I guess the real interest for this linear regress/anova is to test the differences between the other groups and your reference group (assume it is size 1) such size 2 vs size 1, size 3 vs. size 1 and etc. You may also want to compare size 2 and size 3 or something like that. You can use lincom command to achieve that. Hope it helpful.