28  Raw Vermont

Your data will be a 1% sample of the people of Vermont. There are two special features of this mini-project:

Click the csv icon to download that data…but don’t try to use it without reading the rest of this chapter!

download Vermont 2023 1% sample

28.1 Background

The Bureau of the Census collects data on Americans—not just every ten years, but all the time as part of the American Community Survey.

In addition to summaries, they publish “microdata,” that is, data about individuals. In CODAP, you can get access to this data though the Microdata plugin, which you can find in the Plugins menu under Getting Data.

That plugin gives you nicely-formatted data that appears as a CODAP table.

In this project, however, we take one step backwards, and start with data that’s in the form in which we received it from IPUMS, a service at the University of Minnesota.

You’ll get a CSV that has not been decoded. What does that look like? Like this:

Ewww. But you can see what’s going on: the first line (which is four lines long) contains the names of the attributes—the column headings, separated by commas. Then each line has values, separated by commas. You can drag that csv file into CODAP (or a spreadsheet, or a browser), and it will become a table:

Better, but what do those numbers mean? You can see a plausible income value (INCTOT) over on the right, but SEX is full of ones and twos. Which one is female??

You’ll find that out in the documentation from IPUMS.

Tip

IPUMS is a free service, but you have to sign up.

Especially if you are a teacher, that’s easy. And it means that you can get a file just like this one for wherever—and whenever—you want.

The details of how you get exactly that can be tricky, but with some persistence, you can figure out how to make your request. Pro tip: you don’t want your whole State. But you might want your PUMA (Public Use Microdata Area). And that’s a column you can filter by.

28.2 Getting documentation on variables

Suppose we want to know what MARST is and what the codes mean.

Here is a link to a useful page on the IPUMS site. Go there and do this:

  • Under Select Harmonized Variables and A–Z, choose M (for MARST). You should see this (and more).

Aha! MARST is marital status!

Also, note the Type column, with values of P and H. This indicates whether the variable is a “person” or “household” variable. That distinction will become useful shortly.

  • Click the underlined MARST in the list. You will go to a page about Marital Status.
  • Click the CODES tab to see which numbers mean what:

Now you know, for example, that 5 means Widowed. Follow the same procedure to find codes for any of the columns.

Then, if you like, you should substutite suitable strings for the numbers in the dataset. That way, when you make a graph, you’ll see “widowed” instead of “5”.

This is a “calculating” data move, even though it requires no calculation: you’re changing the values for individual cases.

One really great way to do this is to group the data by MARST (or whatever)—by dragging left—and then to edit the values once in that leftmost table. Then drag it back to the right. We described this process in detail in the chapter on calculating and recoding.

28.3 Household variables

Your data includes data about households as well as data about individuals. What does that mean?

Each row in the file is a person. But people who live in the same household are grouped together. This may be clearer with an example. Here is that same CODAP table with a single household selected:

First: how do we know they’re in one household? There are several ways, but the one you can see in this illustration is in the column called PERNUM (person number). See how every houehold restarts with one?

So who is in this household? We looked up SEX and found that 2 is female. So we have what looks like a 39-year-old…mother? With three kids, aged 16, 15, and 7. The 15-year-old is a daughter. And they are all unmarried (MARST = 6).

Looking farther to the left, it looks as if their place has 5 bedrooms. Pretty deluxe! But wait: If we look up BEDROOMS in the documentation, we see that the code of 05 indicates four bedrooms. Tricky. Treacherous.

In any case, you can imagine that dealing with households by looking at PERNUM is difficult and time-consuming. So here’s what we recommend:

  • Scroll leftwards in the table to find the SERIAL column. This is the serial number for the household.
  • Drag SERIAL left, making new groups.

Since there is a fresh value of SERIAL for each household, this action separates the dataset into households.

Many other variables are “household” variables as well1; you can therefore drag them leftwards to join SERIAL in what you can rename as the “households” collection:

The red arrow shows where I have edited the name of the collection. Also, notice that when I selected the household on the left, all five of its people got selected on the right.

This time the mom is divorced, and has four sons. The oldest one seems to have a job! Mom has a GED; to decode that 64, I looked up EDUC and clicked a button for “detailed”—which is EDUCD.

Use “choosy” to reduce the number of columns

Your table may be littered with columns you don’t want to see any more. Mine sure was.

So I opened choosy from the Plugiuns menu under Data Moves and

  • Made all columns invisible by clicking on the slash-eye icon.
  • Made the ones I really wanted visible one by one.

Of course you can use choosy to un-hide any attributes you want back. Read all about choosy by clicking its info button after you launch it.

Now you can explore questions about household data or person data—or questions that mix the two. For example, you could claim that people in the cities make more money than people in the country. Income is attached to the person, but what we’ll call urban is attached to the household. So you can make a graph like the one in the margin—which suggests that the urban advantage may not be all that big.

The median income for urban folks is slightly higher.

28.4 Group Quarters (GQ)

You may notice that the household variables for the first 400 or so households look strange.

That’s connected to the GQ variable: “Group Quarters.” This is for people who don’t live in a traditional household, like an apartment or house. There is a wide variety of group quarters. They include:

  • Homeless shelters
  • Skilled nursing facilities
  • Military barracks
  • Prisons
  • College dorms

Be aware of this if you’re drawing conclusions about households. You might want to filter the group quarters out—or not.

28.5 PUMAs

No, not big cats. These are Public Use Microdata Areas. The idea is that each PUMA has at least 100,000 people in it. This makes it easier to make the data anonymous, but still have some geographical relevance.

Since Vermont is small (which is why we chose it for this example), there are only 4 PUMAs in the State.

Vermont PUMAs
number PUMA
00100 Northern Vermont—Grand Isle, Franklin, Lamoille, Orleans, Caledonia & Essex Counties
00200 Central Vermont—Addison, Washington & Orange Counties
00300 Chittenden County
00400 Southern Vermont—Rutland, Windsor, Bennington & Windham Counties

And since PUMA is a household variable, you can use it to, for example, compare income or housing size between the northern and southern parts of the State.

28.6 Wait…what am I supposed to do?

  • Download the data (link near the top of this page). It’s a csv file. Figure out where it is on your computer (probably in something like a Downloads folder).
  • Drag the file into an open, blank CODAP document.
  • When it asks, tell it you want to import all of the data.
  • Drag SERIAL to the left to group by household.
  • Recode several of the variables (start with SEX and MARST) using the instructions above.
  • Before you go too far, make at least a graph or two to establish that the data make sense.
  • Explore more data and graphs, thinking about how to incorporate household data.
  • Maybe drag additional variables (such as GQ) leftwards to join SERIAL.

During this process, you will encounter things that don’t seem to make sense. Persist! Get help! See if you can understand them!

  • Come up with a simple, straightforward claim or question that you can address using the data. (Desperate? Try “people with more money have bigger homes”)
  • Make a graphic that addresses the claim.

Then, eventually,

  • Be critical of that original claim, and do additional analysis to add nuance.

28.7 Heartfelt reflection and advice

These data give you a view into the lives of over 6000 Vermonters in 2023. For me, adding the household data makes the data much more three-dimensional.

So here’s a task your teacher might even assign:

Scroll somewhere in the middle of the file and find a household. And tell its story. Like in the last illustration, we have a divorced mom with four sons. She makes under $30000 a year, and has a rent of $560. What is her life like? Why didn’t she finish high school? When did she get her GED? What’s it like to be the oldest son, who is only 15 and made $1200 last year? Do the two little ones share a bedroom? What’s that like?

The point being: We make generalizations about the world and our society when we make graphs comparing income and education; or when we search out sources of injustice, exploring how persons of color don’t get the educational attainment that whites do; or when we study history, seeing how the percentage of women in the workforce has increased dramatically in the last century.

But in every one of our graphs—especially the ones made of dots—we should remember that they are made up of actual individual people, and every person has their own story.

Those individual stories do in fact contribute to trends, trends we measure and name, trends that suggest policies that might make people’s lives better.

But every dot is someone different. And one of the dots is you.

  1. In fact, all of the household variables start out to the left of the person variables in the csv file.↩︎