Well, this still makes sense to me. Couldn't this just be due to the reduction of the sample set as people go up in age(I.E. they die)? More people are going to make it to 20 years old than 40 and so on. This will also be amplified by population growth so it should be increasingly skewed towards younger people having more stable data.<p>Also, the older generation is more likely to have the family structure where the man works and the wife stays at home. I would think there would be financial incentives for a wife to over report her age for senior benefits and her husband to under report so as not to be forced into retirement.