How to Use the American Community Survey Estimate Analyst Prompt to Report Neighborhood Numbers With Honest Margins of Error

American Community Survey data is the go to source for neighborhood level facts about income, rent, commuting, and more. It is also easy to misuse. Every ACS figure is an estimate with a margin of error, small areas have wide margins, and adding tracts together or calculating a share means recalculating that margin. The American Community Survey Estimate Analyst: Table and Dataset Choice, Margins of Error for Derived Estimates, Coefficient of Variation Flags, and Statistical Difference Tests for Small Areas prompt walks through the Census Bureau's approximation formulas step by step, flags unreliable numbers, tests whether a difference is real, and writes a plain language summary that keeps the caveats.

What the prompt produces

1. A dataset check confirming whether 1 year or 5 year data fits your geographies and whether the table's universe matches your question.
2. Derived estimates for sums and differences, with margins of error combined using the square root of summed squares.
3. Shares calculated with the proportion formula, and the ratio formula as a fallback when the proportion formula fails.
4. Reliability flags using the standard error and the coefficient of variation.
5. A difference test that shows whether two estimates are statistically different at 90 percent confidence.
6. Cautions that apply to your case, such as not adding medians.
7. A plain language write up for your audience.

How to fill the inputs

ResearchQuestion is the question in everyday words, such as whether renters in two tracts are more rent burdened than renters countywide.

Geographies lists the areas and their approximate population. Areas under 65,000 people only have 5 year estimates.

SurveyPeriod names the dataset and years exactly as data.census.gov shows them.

TableEstimates is the most important input. Paste each estimate with its margin of error from the table, by geography. The prompt uses only these numbers.

ComparisonTarget is what you are comparing against, such as the county or the state.

Audience shapes the write up. A council memo and a newsletter can use the same numbers with different framing.

Reading the example output

The example combines two census tracts and compares their rent burden share with the county:

• The denominator is cleaned first. Renter households whose rent burden could not be computed are removed from the total before the share is calculated.
• Each margin is rebuilt. The sum of eight burden categories across two tracts gets its margin from the square root of the summed squares, and every step is shown.
• The share uses the proportion formula. Because burdened renters are a subset of all renters, the proportion formula applies, and the output confirms the value under the root was positive.
• Reliability is checked. The combined estimate has a modest coefficient of variation. A single tract has a wider margin, so the output recommends reporting the tracts together.
• The difference is tested, not assumed. The neighborhood share is higher than the county share, but the Z score falls below the threshold, so the difference is not statistically significant.
• The write up says so plainly. It gives every estimate with its margin and explains that the gap is within the margin of error.

Tips for better results

• Copy estimates and margins directly from the table export rather than retyping them.
• Combine small areas when a single tract is too noisy, and say that you did.
• Keep the worked math in an appendix so anyone can check it.
• Use the same table and period for every geography you compare.
• Ask the prompt to rerun the test for another comparison, such as severe burden only.

Mistakes to avoid

• Do not add or average medians across geographies. Medians need the underlying distribution.
• Do not compare overlapping 5 year periods, such as one period with another that shares four years.
• Do not drop margins of error from charts or headlines.
• Do not call a difference real when the test says it is not significant. Say the data does not show a difference.

Who it is for

Housing and community nonprofits, city and regional planners, local reporters, public health analysts, grant writers who need neighborhood statistics, and students working with census data.

Related PromptDig links

Open the American Community Survey Estimate Analyst: Table and Dataset Choice, Margins of Error for Derived Estimates, Coefficient of Variation Flags, and Statistical Difference Tests for Small Areas prompt and paste your table estimates with their margins. For more research and data prompts, Browse more prompts. If you have a data analysis prompt that keeps the statistics honest, Share a prompt.