Median family income question

Bill

If you are trying compare change in income on a one-to-one basis, I'd recommend adjusting any income levels you use to the same point in time. It could be 2016 or it could even be 2018. If you make this adjustment using ACS guidelines for comparing income data, which calls for making adjustments using the CPI-U-RS. You can find details on the Census ACS website. Look under "Comparing ACS Data" and pick the years for the data sets you are using.

However, your plan seems to raise another issue. When you compare 5 year ACS data that is less than 5 years apart there will be overlapping data sets underlying the statistics (eg, 2009-13 and 2011-16 share the data from 2012-13). While the larger sample dampens the effect on the margin of error it also means that you are not really comparing two independent values, and you will see more correlation than you might expect in the figures. This is not to say you should not follow this method. The answer to that question depends on what question you want to answer and who your audience will be. If I were undertaking this type of comparison I would use two non-overlapping 5 year ACS period if the data is available.

Bill

If you are trying compare change in income on a one-to-one basis, I'd recommend adjusting any income levels you use to the same point in time. It could be 2016 or it could even be 2018. If you make this adjustment using ACS guidelines for comparing income data, which calls for making adjustments using the CPI-U-RS. You can find details on the Census ACS website. Look under "Comparing ACS Data" and pick the years for the data sets you are using.

However, your plan seems to raise another issue. When you compare 5 year ACS data that is less than 5 years apart there will be overlapping data sets underlying the statistics (eg, 2009-13 and 2011-16 share the data from 2012-13). While the larger sample dampens the effect on the margin of error it also means that you are not really comparing two independent values, and you will see more correlation than you might expect in the figures. This is not to say you should not follow this method. The answer to that question depends on what question you want to answer and who your audience will be. If I were undertaking this type of comparison I would use two non-overlapping 5 year ACS period if the data is available.

Thanks for your response! This is really helpful. It seems like I'm learning something new every day about how to properly utilize ACS data. Right now I'm just interested in showing how median family income fluctuates over time. The concern about overlapping data is a valid one. After a little digging on the "Comparing ACS Data" site, it seems that comparing 1-year estimates (i.e., 2013 to 2014, 2014 to 2015, etc.) is not sound because there's some overlap there too. I wonder if I should compare 1-year data from even-numbered years to avoid overlap, and then use the consumer price index to adjust them for inflation?