This is the most important survey in America, but it keeps getting defunded
The US Census isn't just a bureau that counts people.
Sure, the census asks where you live and how many people are in your family. But it also asks what type of toilet you use, when you came to the US, and how much money you earn. These surveys are a way to gauge how the country is doing and a way to make decisions.
But since 2010, the Census Bureau hasn't just asked questions once every 10 years. It started talking to 295,000 people every month.
Today the bureau released important results from that questionnaire — data sets that help us get an accurate bird's-eye view of America.
It's called American Community Survey, and it's arguably one of the most controversial government surveys — and the most important. Republicans have tried to defund the survey time and time again. But researchers argue it's important to count Americans this way because the ACS is the only survey of its kind. When agencies and nonprofits figure out how to spend their resources, the survey is about the only way they can get a bigger-picture view of the population they want to serve. If you want to find pockets of Korean-speaking populations across the US, the ACS is about the only way you can do that. If you want to develop a strategy to help families move off welfare in Michigan, you need the ACS.
So why do Republicans think it's wasteful and wrong?
First you need to understand how the survey works.
Why is the Census Bureau doing something other than the normal census?
The first US Census in 1790 was very simple. It counted the number of males older than 16, which helped gauge the country’s military potential, as well as the number of free people and slaves. It surveyed the head of each household, and by law the census schedules had to be posted in the "two most public places" within each jurisdiction.
Here's what a completed census looked like:
But in 1790, the numbers were small enough that counting everyone wasn't an impossible task. Not anymore.
These days, the Census Bureau counts Americans every 10 years using a 10-question survey. Frankly, this is the boring part for data geeks, but the Constitution requires it because it helps the country allot federal representatives. (There are a host of problems with how this works, but that's not for this story.) This survey is known as the "decennial census."
Over time, the census started going beyond its 10-question set, adding more and more questions. For example, it started asking Americans about their education and vocation. And the survey started getting quite long: In 2000, about one in six households had to answer 52 questions.
These "longform" surveys existed until the 2010 census, when everything was simplified back to 10 questions. The other questions, like when people first came to America or the languages they speak at home, were asked in the American Community Survey.
What can the American Community Survey tell us about America?
It helps us get a macro-level view.
Imagine that you're in a helicopter, looking down at America. In this scenario, you have super vision, so the people look like 320 million little dots. The decennial census helps you see basic traits, like how old those dots are and what race they identify as.
Imagine yourself in this helicopter, with super vision, looking down on every person in America.
But what if you wanted to know how long each person's commute was? That's not on the decennial census. In fact, a lot of very important questions aren't on the census but were moved to the ACS. But ACS doesn't count everyone. Instead, it surveys 3.5 million people each year — and it tries to estimate other things (like commute length) based on that small population.
That said, if you're only talking to 3.5 million people each year, that means only a tiny number of people in your neighborhood answer the questionnaire annually. So there's not enough of a sample size to accurately calculate the average commute time of people in your neighborhood.
So the ACS uses a little trick: It looks at five years' worth of data to get a better idea of the average commute time for people in your neighborhood.
That's the data that was released today. It's the most recent and accurate data set that helps you get a bird's-eye view of rural areas and small ethnic and racial populations. It helps unseen populations be seen.
That said, five-year estimates aren't up to date, by definition. For example, the ACS asks whether you have health insurance — but if you look at five-year estimates that span before and after the Affordable Care Act, it's not useful.
In addition, you can't compare last year's five-year estimates with this year's five-year estimates. The Census Bureau has a full explanation here.
If you're not sure which data set to use, the bureau has a nice chart here.
Why does the census undercount the underprivileged?
An analysis of the 1980 census found that African Americans were severely undercounted.
The 2010 census missed 1.5 million minorities.
The ACS is not a comprehensive survey, and it has never fined or prosecuted anyone for not responding, so there's no such thing as "missing" people. But it can underrepresent people. This is largely because minorities and other underprivileged people fall into "hard-to-count" groups — people who are renters are those who are more transient. As Congress cuts funding for the ACS, it makes it harder to get proper samples of sparsely populated rural communities and smaller racial and ethnic communities.
Why are Republicans trying to get rid of the ACS?
Congressional Republicans have tried to kill the ACS, arguing that it's both a waste of money and an intrusion on Americans' privacy.
"This is a program that intrudes on people’s lives, just like the Environmental Protection Agency or the bank regulators," Daniel Webster, a Republican representative from Florida who sponsored a bill to kill ACS in 2012, told the New York Times.
The ACS is still around, but the House has cut the Census Bureau's budget and voted to make the ACS voluntary. Poor response rates are a notorious problem with voluntary studies. That said, studies show voluntary surveys tend to underrepresent the wealthy, educated, and mobile populations.
Are there fascinating moments from census history?
You bet there are!
The 1880 census took almost a decade to complete — and that was with the help of a rudimentary counting machine built by Charles Seaton. By the time the 1880 census was counted, it was time to begin the next count. Since hand counting was no longer feasible, the 1890 census required a new technological innovation.
Herman Hollerith, a former Census Bureau employee, created the first electric tabulating machine, which required the questionnaires to be transferred to a punch card and run through the machine.
Another key moment for the census was when Congress special-ordered one in 1930. It was right after the market crash in 1929. As the United States sank into the Great Depression, statisticians believed the 1930 unemployment number was unreliable — and probably too low. Congress agreed, and wanted more up-to-date numbers, so it ordered a special unemployment census in 1931 — and another one in 1937. We don't have this problem today, though. The ongoing ACS questionnaire means we don't need to order a new census to find out information between decades.
Okay, I'm sold. How do I explore this data?
There are three ways, spanning from easy to hard, that I like to explore census data:
1. Census Reporter (easy): This site holds your hand through the entire process. It's the most user-friendly, but it's also the least flexible.
2. American FactFinder (medium): This site is the way the Census Bureau expects most people to browse its data. You can select what data set you want (one- or five-year estimates) and browse dozens of different geographic breakdowns. This is where the new five-year estimates are available.
3. IPUMS (hard): If you need a specific data set going back to 1850, this is your tool. It's probably unlike any data site you've ever used. The learning curve is quite steep, but try these YouTube tutorials if you really want to go down this rabbit hole: