While working on loading the recent release of the American Community Survey (ACS), 5 -year estimates, 2014-2018, in Data Planet, we landed on a great tool and want to share it with you! The US Census Bureau made available a Statistical Testing Tool in 2018 to help users easily test the statistical significance of ACS estimates.
A bit of background on the ACS – the survey collects data from the 50 states, Washington, DC, and Puerto Rico. It is a continuous survey, in which each month a sample of housing unit addresses receives a questionnaire, with approximately 3.5 million addresses surveyed each year. Results are pooled to produce estimates for geographic areas in the US and Puerto Rico, ranging from neighborhoods to congressional districts to the entire nation. The statistics reported represent the characteristics of the population for the entire period vs a specific year within that period.
It is important to understand that ACS results are estimates – meaning they are based on a sample of the full population vs an actual count of population. Because of this, there is some uncertainty associated with the accuracy of the estimates and the Census Bureau publishes a margin of error (MOE) that provides a measure of the variation between the estimate and the actual value for the whole population had the entire population been surveyed.
The uncertainty of the estimates makes it critical that a data user consider, when comparing estimates, whether the difference between two (or more) values is statistically significant, ie, is the difference larger or smaller than would be expected by chance? The cool new(ish) tool from the Census makes it super simple for users to conduct a z-test of the statistical significance of ACS estimates.
For example, you might want to compare states by counts of households that have a desktop or laptop computer. Download the tool at the link above and then download a spreadsheet of ACS values and MOEs from the ACS 2013-2017 from Data Planet by clicking on the Export option and selecting Excel.
Import the data values and geographies directly into the Census tool and it will quickly help you determine whether estimates are higher, lower, or not statistically different from one another. Download the tool at the link above and then download a spreadsheet of ACS values and MOEs from the ACS 2013-2017 from Data Planet by clicking on Export option above the chart below and selecting “Excel”! You’ll find an overview and instructions in the tool itself and you can also learn more about it on the Census Bureau website.
Watch this space for the release of the ACS, 2014-2018, results in Data Planet. Coming soon!!