Between politics and the pandemic, it has been difficult to determine what to believe these days. A recent data analysis from Nate Silver’s FiveThirtyEight site has former Vice President Joe Biden winning this election in a walk:
Joe Biden’s chances of winning the Electoral College rose to 76.7%, according to the latest run of poll aggregator FiveThirtyEight’s election forecasting model, from 76.6% on Sept. 24. He is predicted to win 352 of 538 electoral votes.
A piece in one of our local papers ran this headline, which is problematic on several levels: “With an abundance of polls favoring Joe Biden, is Wisconsin still up for grabs?”
- Question headlines are essentially worthless in terms of informing people. I’ve seen so many of them lately, I keep thinking they’re an unreported symptom of COVID-19.
- Well, polls don’t vote. People do. Thus, what the polls favor and what people are willing to do in the voting booth are two different things. It’s the same reason that nobody ever admitted to watching “The Dukes of Hazzard” and yet it was in the top 10 every week when the ratings came out.
- Polls, including Nate Silver’s, got a lot of stuff wrong back in 2016, something we’ll get back to later…
These are the data folks talking, and I’m a huge fan of data. However, I have some observations that seem to run contrary to this sense that the former VP is going to be capping off a cheesehead victory on election night.
First, everywhere I go (and I mean everywhere I drive), I see dozens and dozens of Trump-Pence signs. Not only are they simple lawn signs, but they’re these giant monstrosities of cardboard and vinyl. This is in the larger cities (not counting Madison, which is both our liberal bastion of the state and our second-largest city) and along the rural roads I drive to get between home and school. Just as polls don’t vote, signs don’t vote and I’m sure I could make some crack about “size doesn’t matter” when it comes to these signs, I can’t remember seeing this many of these things for ANY presidential candidate, including Obama in ’08.
Second, Wisconsin is always a wiggler. If you look at our statewide elections, we tend to go either/or. If you look at our legislature, it looks like we’re as red as Indiana, based on seats. Trump won the state in 2016 over Hillary Clinton by 22, 748 votes. That’s less than either Jill Stein pulled in, and she placed fifth. Put another way, that gap is smaller than the average attendance of a Milwaukee Brewers home game last year. When I lived in Indiana, I knew my vote didn’t matter either way. If I voted Republican, I was throwing votes on the overwhelming stack. If I voted Democrat, it was the Charge of the Light Brigade at best. Wisconsin? I matter and not just in a metaphoric sense.
Third, the noise. I stand in a lot of lines for groceries, to pay for gas, to pick up take-out and to wait for my kid to get out of school. I listen to other people’s conversations and I’ve not heard once anyone saying that the plague, the economy, the riots, the racism or anything else is in anyway related to anything the president is doing or not doing. Instead, I see the MAGA hats, I hear “Boy, we’d be in REAL trouble if we didn’t have Trump” comments and more. Again, each of these people has one vote and I’m not living in the San Francisco of the Midwest out here, but it just… sounds… like the data about “the forest” is somehow overlooking “the trees.”
The same thing is true with the coronavirus pandemic. According to news articles, Wisconsin is racking up cases like we’re playing COVID-19 Pokemon. (Gotta catch ’em all!) Our cases are spiking out of control, universities elsewhere in the state are going on lock downs and our flagship campus has even cancelled spring break as a preventative measure for next term.
Here at UWO, I lost four of six kids in one section in my blogging class in less than 24 hours. I’ve been teaching classes of two students (instead of 15) due to kids dropping like flies to this or being forced to quarantine. My colleagues in the department are also reporting similar concerns, with tiny classes, sick students and more.
However, according to our COVID-19 dashboard, we’re in great shape:
The data points reported here suggest that we’re seeing fewer cases each day. The positive percentage is dropping precipitously and our aggregate numbers are declining as well. Even as I’m seeing more cases (as is the state, the county and other UW campuses), UWO is defying the odds and beating the virus, this data says.
On the day in which I had four students test positive, we only had 14 positive tests on campus. Is it possible that I know roughly 30% of all the positive cases on that one day? Sure. Is it possible that as our department (and at least one of my friends’ departments) are seeing kids disappearing like they’re in the opening scene of “The Leftovers?” Absolutely. Is there a chance that what UWO is reporting and what I’m seeing are not being measured the same way? Could be.
Do I doubt the data? Um. Yeah.
And that bothers me quite a bit as someone who understands data and knows that “what my neighbor Dan says” isn’t the best representation of reality.
In putting the final touches on the second edition of the reporting book last week, I found myself reviewing a comment from one of the professional journalists who contributed to the text. Jaimi Dowdell is an award-winning reporter, former IRE educator and Spotlight Fellow at the Boston Globe. A long-time data journalist, Dowdell offered this observation about observations:
“Get away from anecdotes. Verify information,” Dowdell said. “Stop relying on what people say. If the police chief says crime has decreased, get the data and prove it yourself. If the mayor says your community took a financial hit because of lost tax revenue, get the revenue data to see if that is true. By verifying facts and not relying on anecdotes we can inform issues in the community.”
Jaimi is totally right. If you want the perfect example of how “size doesn’t matter” in an election when it comes to the number of signs or plots of land, check out Illinois. The Chicagoland area (Chicago and its immediate suburbs) contains 9.5 million of the state’s 12.67 million people. It’s basically a Democrat stronghold, so no matter how many Trump signs you would see driving from Edwardsville through Springfield, up past Bloomington-Normal, and around to DeKalb, that state isn’t going red. Interview every farmer you meet in Wayne County who would vote for a potato if it were on the Republican ticket and you’re STILL not seeing reality as the data will bear out.
Jaimi’s also right when I stop and think about the coronavirus testing. The numbers I’m looking at are a day behind the tests (Monday’s testing is released Tuesday etc.) and they’re ONLY for the ones that are done on campus. We have various other places where students go for testing, based on their insurance, their workplace connections and so forth. My wife has been to the Sunnyview Expo center for testing so often, she practically has her own parking place over there. Those folks don’t go into the mix as positive tests in these figures.
As we have discovered in many corners of life, reality usually sits somewhere between what the data tell us and what we can see with our own eyes. It’s not so much that one is better than the other in all cases, but more about what we need to keep in mind as journalists who are working on stories in which various sides want us to see certain things in specific ways. For example, the data told us that it wasn’t going to rain while we were out at the Adams County Flea Market. Cloudy, but no rain. We could see some dark clouds building in west and we heard a crack of thunder. (One optimistic vendor yelled, “That’s over there! We’re fine over here!”)
Suddenly, the skies opened up as people scrambled to cover and protect their wares, all the while muttering variations on, “They said it wasn’t supposed to rain today!”
So how is it that we know which is the best way to look at ongoing and developing situations that matter to our audiences? Here are a few thoughts and suggestions:
Understand the type of data you are examining: Jaimi’s point about going to get data to examine the veracity of an anecdote works well in her example because of the type of data she wants. In this case, it’s data that is already collected for a specific purpose about an event that already occurred. That stuff will almost always be rock-solid, presupposing the people who collected it didn’t screw up or aren’t hiding something.
When you’re talking about things like polling data, you’re trying to predict something that hasn’t happened yet based on what people tell you they plan to do. That data is more variable and less reliable than hard numbers based on finished outcomes.
The difference between these two would be akin to looking at the attendance of a county fair versus a poll regarding people’s intentions to go to the county fair. The first one, you have actual numbers that already have been collected based on actions people took: attending the fair. The second one, you have people telling you what they plan to do, but those plans can change. Something better might come along. They might get sick. Or, and this is likely, it might rain that day and nobody wants to be stuck at a county fair ground during a rainstorm. (Trust me on that one…)
Understand the poll and how it works: If you’re playing with polling data, a lot of things factor into it. In fact, the National Council on Public Polls has a list of 20 questions journalist should ask about polls and their data. These questions cover everything from how questions were asked, when they were asked and who asked them.
A lot of what went wrong (if you read the post-game analysis above on Nate Silver) comes down to some glitches in the answers to the 20 questions. It also doesn’t help when the polls have been consistently RIGHT for years and years, leading people to a false sense of security about these things, even as the ground is shifting under their feet.
I’ll be honest, I didn’t see Trump winning in 2016 for the same reasons a lot of people didn’t see it. We looked at the polls and we looked at our own sense of reality: Rich people with limited experience in politics (Ross Perot), people who had been accused of… let’s call it “extra-marital sex stuff” (Gary Hart, John Edwards), people with racism issues (George Wallace) and “political outsiders” (Basically every third-party candidate out there) got smoked in presidential elections. Al Gore had to apologize for “being rude” when he sighed during a presidential debate and George H.W. Bush got grief for looking at his watch during the 1992 run.
Every time I saw Trump on TV in the news in the 15 months leading up to the election, it was either a poll saying he had no shot or him checking off one of the “is this the death knell for his campaign?” issues listed above (and more). What I didn’t do, and should have done, was start looking around me and noticing the signs, the rallies and the groundswell and think, “Maybe this guy does have a shot…”
Understand the biases in your own observations: Observation is based on individuals looking at things through their own eyes and making judgments on what they see. The problem with that is people often overvalue their own sense of reality when they decide what is or isn’t happening.
Case in point: the COVID numbers at UWO. I’m seeing students dropping like flies all around me. I’m seeing colleagues have the same situations. I’m seeing a line a block long outside Albee Hall for testing every day. My sense of observation tells me we’re about six days away from the zombie apocalypse.
However, the numbers are telling me something else, which is that our overall cases are going down. This doesn’t jibe with what I’m seeing, which gives me some serious cognitive dissonance.
The bias I’m dealing with is that I’m working with only one group of students (journalism students) and so are most of my colleagues. These folks tend to work and play together in class and through student organizations (student media etc.). Thus, if one of them gets sick, it’s probable more of them will as well. (Our student newsrooms used to be like petri dishes for illness. One kid coughed and suddenly we’re down to three staffers.)
In addition, I understand that I have a bias against simply trusting what I’m told. In journalism, we say “If your mother says she loves you, go check it out.” Also, I’ve been burned a few times, so I find myself constantly in that “non-denominational skeptic” role, trusting almost nothing until I can feel as sure as possible about something.
Where you are looking (rural vs. urban is always a thing in election data) and when you are looking (how many “I’m with Bernie” bumper stickers are still floating around now as opposed to eight months ago) can have a huge impact on your own observations.
Learn to look behind both data and your observations: One of the things the NCPP discusses in its list of 20 poll questions is the issue of who’s doing the poll and what it’s trying to show. This kind of question can be easily applied to both data and observations as well.
When I worked with student media, we used to study the data we got each year as part of the Clery Report. This report, which is required due to the Jeanne Clery Act, provides data on crime and safety issues on each college campus that participates in federal financial aid programs. The idea behind the act and the report is that people have a right to know how safe (or unsafe) their campus environment truly is.
When we would look at the data in some years on some campuses, we were a bit disturbed. It wasn’t because our campus was so violent or scary, but rather that we knew we had reported on incidents involving students that should have made the numbers look scarier. In other words, we might have reported on six or seven assaults in a given year, but the data told us that zero had occurred.
What we started to realize was that the institutions were able to set some parameters and report some data in ways that excluded some places around campus where bad stuff happened. (On one particular campus, Frat Row was basically turned into a Mad Max movie every weekend, but fortunately for the bean counters, it was just far enough away to not be considered “on campus” if they didn’t want it to be.) It wasn’t that they were lying, but rather reporting the data in a way that was most advantageous to the position they wanted to demonstrate. Knowing this meant we had to look beyond that data before writing a story that told our readers, “This place is totally safe and you can be secure in everything you do!”
The same thing is true of observations: Looking beyond what you’re seeing can help a great deal. It’s not always “citizens” who are putting up large signs for candidates or putting up billboards that profess a position on something. In many cases, larger organizations do the heavy lifting for them. My grandmother lived on the main drag of her city for decades, and politicians were always asking if they could put a sign up in her yard. Unless she REALLY hated someone, she was pretty OK with it. That didn’t mean she would necessarily vote for that guy or gal, let alone make a strong case for him or her. However, the optics were that she was a proud supporter, saying it loud and proud.
In driving between small towns a few weeks back, we ended up essentially following a truck that had giant signs for one candidate. The truck stopped every so often and pounded one of these things in the ground (I’m assuming this was done with owners’ permission or whatever; I’m not trying to get anyone in trouble here.) before moving on. Maybe the other candidate would be rolling through later to get some signs up. Maybe not. Maybe everyone in that area loved that candidate. Maybe not.
Maybe it was just two dudes in a truck with a bunch of signs trying to make us think that the entire stretch of our state was in favor of this one candidate. Who knows?
The point is, going beyond what you can see with a quick glance or an assumed position of accuracy will do you a lot of good when examining data and observational points.
It ain’t over ’til it’s over: Yogi Berra always had a way of making sense of things by not making sense at all. (I worked for about five minutes trying to weave his line “You can observe a lot by watching” into my spiel on observations. I failed.) This favorite Yogi-ism really does hit the nail on the head when it comes to data and observation.
You need to keep reporting a story until it is complete. Assuming you have the story when the pieces are “almost” in place will set you up for failure. Figuring that what you saw last week is probably the same as what you would see this week, is likely to get you into trouble.
Regardless of what you can see, hear and sense at any point in time, stick with the story until it’s over. It ain’t over until then.