Status:Closed    Asked:Apr 22, 2016 - 11:48 AM

ACS hhwt for 2001-2014 totals too high - 4 billion

Using ACS 2001-2014 in the online query, the HH variable ownershp (tenure) and hhwt (Household weight) I get a table with total weighted N of 4 billion, rather than the expected 116 million. How do I get a table with the correct N?

Do you have the same question? Follow this Question

Staff Answer




Using the online analysis system, a weighted tabulation of ownershp will produce a cumulative total for all ACS years (2001-2014) and will duplicate households with more than one respondent. If you want to see tenure responses by year, you will need to enter "year" as the column. Further, since you are observing a household variable, filtering your results by pernum(1) will make it so each household is counted once. These two changes should help create results more consistent with your anticipated N.

I hope this helps.


Apr 22, 2016 - 02:09 PM

Report it


There is no official data product ACS 2001-2014. If anything, ACS only started in 2005. Do you mean 2010?

At any rate, some of the reasons to get the wrong population totals are:

1. You are using person-level data, so your HH weights are effectively multiplied by the HH size. (You may want to subset to relate==1 or something like that).

2. You are using annual weights instead of multiple year weights. The multiple year weights are carefully developed by the Census statisticians; there is no guidance on how to use the weights if you combine the years on your own. If you have 10 years worth of combined data, then your annual weights will produce a population total that is about 10 times off.

Finally, without knowing what you did, there is no way for us to tell "how to get the right table". What software are you using? How did you specify the weights? What were the commands that you typed? Provide a reproducible example.


Apr 22, 2016 - 12:19 PM

Report it


Weighted N for Federal Government Workers Subsample is Bigger than that Group...
As a new user, can someone check my analysis that used HHWT field?
hhwt to calculate weighted means and repwt 1-80 for standard error across 201...
I have an extract 2001 to 2015. Total number of people: 4 million, but in bir...
Login   |   Register

Recently Active Members

View More »

Share |