Comparing single year to multiyear

I am working with single year ACS files and have a question regarding comparison with multiyear files.

Using the “RELATED” and “AGE” variables I am trying to figure out the number of adopted children under the age of 3 (meaning those under 1 year old, 1 year old, and 2 years old). I find the following:

2009 - 148,860

2010 - 140,401

2011 - 137,496

A report from the Census uses 2009-2011 multiyear data. It has the number of adoptees under the age of 3 as 137,088. I am trying to figure out if this number is consistent with what I have found and am struggling to grasp my mind around why I would come up with a similar number in 2011 to what ACS has for over a 3 year period. Any explanations will help.

There are three reasons why your numbers do not quite match the official Census numbers. First, the 2009-2011 multi-year data is essentially reporting an annual average number of adoptees under the age of 3 over that time period. It would be more accurate to compare the average of your three numbers (i.e. 142,252) with the Census number of 137,088. Second, the Census Bureau adjusts the weight values when creating the multi-year samples. Instead of using each single-year file, it would be more accurate to use the 2009-2011 3-year ACS sample in IPUMS instead. Using the multi-year sample produces an estimated 140,167 adoptees, even closer to the official Census estimate. Third, the IPUMS data is a subsample of the full ACS data. As a result, the IPUMS estimates can differ slightly from the official Census figures; however, they should still fall within the margin of error.

Hope this helps.