GRAND RAPIDS, Mich. (WOOD) — According to a survey released by the U.S. Census Bureau, while 36 states and the District of Columbia’s counts were fairly accurate, eight states were overcounted and six were undercounted.

According to the Bureau, Michigan’s count was one of the most statistically accurate.

The data comes from the Bureau’s Post-Enumeration Survey, which measures the coverage of the household population, not places where people live in a group setting, like nursing homes or college dorms.

The survey takes a sample of people from across the country and compares its results with the census. The Bureau says they sampled 161,000 households and collected 114,000 data points.

Inforgraphic by Matt Jaworowski/WOODTV8

The biggest outliers were mostly on the positive end. Hawaii (6.79%), Delaware (5.45%) and Rhode Island (5.05%) were the three largest, all showing that the census overcounted for people that live in those states. Arkansas (-5.04%), Tennessee (-4.78%) and Mississippi (-4.11%) were the three most undercounted states.

At 0.14%, Michigan’s residents were slightly overcounted, but the census was extremely close. Michigan’s count had the fourth-lowest error rate, only behind New Mexico (0.10%), New Jersey (-0.12%) and Connecticut (-0.13%). In all, the survey found the census likely undercounted the entire nation by 0.24%.

The Census Bureau says while some of the numbers appear big, it’s important to take them in context. In a release, officials explained why smaller states are more likely to show bigger swings in results.

“Standard errors quantify the amount of uncertainty in the estimates because they are based on a sample of people rather than a census of everyone. The amount of sample in each state depends on the number of people in the state and other more technical considerations,” the release stated. “States with smaller sample sizes generally have larger standard errors.”

According to the Bureau, the Post-Enumeration Survey has no impact on the actual census results. The Bureau cannot retroactively make changes based on that data. Instead, it is used to target problem spots and better tactics for the next census.