Falling # of obs. with nonmissing education with time

USA
1

Hello!

I have been using IPUMS USA extract from 1940 to 2013 using default sample on the website.

So I’ve been trying to detect discontinuities in the data for education attainment at or following the year the changes to compulsory schooling laws were effected.

What I think easy to interpret way of looking at the data is plotting or tabulating the share/number of people against the year a certain cohort turned 16 (yearat16) by each category of educd variable for a particular state. The year aged 16 variable is created simlply by adding 16 to the birthyr of oservations. I have also restricted the sample to age in range 18-64, nonmissing education report (educd >= 2 & educd <= 116) and us born citizens.

I do not think that this way of presenting the data is somehow fundamentally wrong or alters the data in weird way. But what I find is very strange to my eyes. I see that number of total observations with nonmissing education report (and by each category of education) falls as yearat16 (or equivalently birth year) rises.

So what I wanted to ask if any of you know a specific reason for that? Or if you spot any fault in my method?

I would appreciate your help very much!

2

Your focal variable seems to be those aged 16, but you include those only aged 18-64. I am confused by this, but wonder whether this is causing the problem you face.

3

If we look at respondents with yearat16=1978, then we will have respondents from the 1980 Census through the 2013 ACS. Anyone from the 1970 Census would have been too young at the time of the Census to be included. If we then look at yearat16=1979 respondents, you will have only respondents from the 1990 Census through the 2013 ACS. In other words, you have lost the 1980 Census respondents, because anyone that turned 16 in 1979 would not yet have turned 18 when the 1980 Census was administered. Similar drops in sample size will occur every 10 (birth) years for decennial census samples and every (birth) year for ACS samples.

Hope this helps.