By Ari Lamstein
Today I will demonstrate how the margin of error in American Community Survey (ACS) estimates grow as the size of the geography decreases. The final chart that we’ll create is this:
The way I interpret the above chart is this: The ACS is very confident about its state-level estimates. It’s a bit less confident about county-level estimates. But when you get down to tract and ZIP Code (ZCTA) level estimates, though, the margin of error sometimes increases sharply.
Why is the American Community Survey (ACS) Important?
I spend quite a bit of time on this blog talking about ACS data. The reasons are:
- Availability: The census bureau has an API for the data, and Ezra Glenn has written an R package (acs) for getting the data into R.
- Interest: The survey tells you all sorts of interesting information about Americans (race, income, education, etc.).
- Importance: Over $400 billion in federal and state funds are allocated each year based on the survey (link).
So if you want to analyze an important, contemporary dataset for free, I think that exploring ACS data in R is a good place to start. Also, I have my own mapping project in R (choroplethr). And because ACS data has a geographic component, I can easily use it in my examples.
What does Margin of Error Mean?
Wikipedia does a better job of explaining it than me (link). But the gist is this: The ACS only interviews about 1% of US households each year. Because they are sampling the population, they do not have complete knowledge about the population. As a result, each reported value has both an estimated value and a margin of error. In this case, they are 90% sure that the estimated value lies within the reported estimate +- the marign of error.
The above boxplot shows that the margins of error …read more
Source:: r-bloggers.com