I am not able to find a clear description of how to interpret 5-Year Estimates for ACS data. In particular, I am not clear on what the values in question, in my case, population data (e.g., Population 5 to 9 years and 5 to 9 year olds enrolled in school) represent. I am seeing conflicting information on whether the value (in this case, the population number) is an average over the 60 months, or what exactly it represents.
For example, in the document "Interpretation and Use of American Community Survey Multiyear Estimates" (https://www.census.gov/content/dam/Census/library/working-papers/2012/adrm/rrs2012-03.pdf) , it states, "An important property of ACS...estimates is that they do not represent a single point in time but an average [emphasis added] of the characteristics of a geography over a one-year, three-year, or five-year period, so they are referred to as period estimates. Data collected during the 60 months of five calendar years are combined together to produce estimates for the same levels of geography as did the Census 2000 long form." So, this makes it sound like it is an average.
Meanwhile, in the document "A Compass for Understanding and Using American Community Survey Data: What Researchers Need to Know " (https://www.census.gov/content/dam/Census/library/publications/2009/acs/ACSResearch.pdf), it states, "While one may think of these estimates as representing average characteristics over a single calendar year or multiple calendar years, it must be remembered that the 1-year estimates are not calculated as an average of 12 monthly values and the multiyear estimates are not calculated as the average of either 36 or 60 monthly values. Nor are the multiyear estimates calculated as the average of 3 or 5 single-year estimates [emphasis added]. Rather, the ACS collects survey information continuously nearly every day of the year and then aggregates the results [emphasis added] over a specific time period—1 year, 3 years, or 5 years. The data collection is spread evenly across the entire period represented so as not to over-represent any particular month or year within the period." This is pretty clear that it is not an average, but is an aggregate (whatever that means).
The most recent version (2020) of the document "Understanding and Using American Community Survey Data: What All Data Users Need to Know" (https://www.census.gov/content/dam/Census/library/publications/2020/acs/acs_general_handbook_2020.pdf), only says that, "While an ACS 1-year estimate includes information collected over a 12-month period, an ACS 5-year estimate includes data collected over a 60-month period.", but does not give any more detail on what that means.
Any clarification on how to interpret and write up the 5-Year estimate data would be appreciated.
My interpretation of the literature provided is they don't take the average of monthly percents for single year or multiyear estimates, nor do they take the average of yearly percents for multiyear estimates…
Hi Todd,
Others may be able to explain it better, but the way I usually explain the 5-year estimates is that they are period estimates. They are estimates created from surveys collected throughout the 5-year period. So basically you could have responded the survey in the 1st year (say 2015 for the 2015-2019 estimates) and I responded in the 4th year (2018). Both of our responses would be used as samples to come up with the estimates for that 5-year period. Following along that logic, the next year for creating 2016-2020 estimates, your survey responses would not be used but mine would as they'd still be part of the 5-year estimate period (2016-2020) - yours (and all those from 2015) would drop off and responses from 2020 would be then be added to the sample. This is also why you don't want to compare 2015-19 estimates with 2016-2020 because the sample surveys overlap. Rather, for comparisons, you'd use 2010-2014 and 2015-2019. So it's all about the survey responses used to create the estimates. Taking 5 years worth of surveys gives a larger sample of responses so that they can develop estimates for smaller geographic areas with a decent enough margin of error (MOE) to make them reliable as estimates.-Jami
Thank you for your response, Jami.
My interpretation of the literature provided is they don't take the average of monthly percents for single year or multiyear estimates, nor do they take the average of yearly percents for multiyear estimates. Doing so would give a different (and incorrect) result than tallying (aggregating) all responses in the positive for a variable over the entire period and determining what percent that is of the total.
I provide an example (Building Permits Survey data, not ACS) in the table below that shows housing units permitted in Raleigh by year during the 5-year period 2018-2022. Averaging the respective yearly percents gives the incorrect answer (27.7) if one wants to know the percentage of all permitted units in the 5 year period that were allowed to be single-family units. The correct answer (25.3) comes from dividing the aggregate (sum) of single-family units in the period by the aggregate of total housing units in the period, and converting that to a percentage by multiplying by 100.
Yes, I agree with both JamiRae & Jeramiah's response. Nice example, Jeramiah! Whenever I am asked to explain this, I use the phrase in the text that Todd posted: the entire 60-month period.