ACS Data Users Group

I am attempting to aggregate the median household value for a set of tracts using the 2014-2018 ACS 5-year. The base data is from table B19013, which reports the median household values in dollars for each census tract.

I'm doing my analysis in R.

What is the accepted methodology for aggregating median values? How should one calculate the MOEs for this aggregated medians?

I found this reference for re-calculating a median, but it is based on the range of household values:

http://www.dof.ca.gov/Forecasting/Demographics/Census_Data_Center_Network/documents/How_to_Recalculate_a_Median.pdf

Is there an approach to aggregating the median using the median values?

I'll post the code once I find a solution.

Thanks

Adam B.

Sorry for my confusion, I am focusing on Median Household INCOME, not value. My question si still what is the best way to do this. Thanks

The LA Times Data Desk includes aggregating medians in this python library github.com/.../census-data-aggregator

Perhaps it's what you need?

Hi Adam -
The best approach is to go back to the range data, aggregate that, and calculate new medians based on the aggregated data by income range.
-Beth

hello! could you specify what you mean by the "range data"? Do you mean specifically the range of samples taken from survey responses, or the stated income ranges ACS groups by?

Hello ex-neighbor,

The approach we take in our MCDC web apps is to estimate the medians by taking their weighted averages. The result is a weighted average of the medians of the components, but it is not the median of the aggregate. It may not be the actual median, but it is still a not bad measure of the typical, middle value of the distribution.

If you need the actual median, I think Beth's comment below is the best way to go.