Trump was narrowly defeated in Iowa, where he was the favorite just days before the election. Cruz won (some say “stole”) and Rubio surged into a breathing-down-your-neck third place. The actual vote was quite different than the polls predicted. What happened?

Iowa polls versus resultsThis is an important question because if the polls were wrong about Trump winning Iowa, could they be wrong about Trump winning in New Hampshire, and perhaps even the GOP presidential nomination?

One part of the explanation is late-deciders, those voters who don’t make a decision until the week before the vote, or even the day of the vote. Oh, if a researcher calls and asks who they favor, they will give an answer. But some are not committed to that answer all the way through to the voting booth. They change their minds.

Surveys find anywhere from 25% to 45% of voters are still undecided a week before they vote. In 2012 entrance/exit polls (NY Times, CNN), researchers asked people as they were going into the vote (or leaving) “when did you make your decision?”. On average, 32% of voters made their decision THAT WEEK.

Slide3

Yikes! So there are a lot of people who answer polls months before they vote, but don’t stick with their choice on election day. They don’t decide until the last minute.

In Iowa, we see that nearly ONE HALF of voters didn’t make their choice until the week before the election. And the later they decided, the less likely they were to vote for Trump. Instead, they tended to favor Rubio (28%) and Cruz (22%). Of those who decided in the last week, only 15% selected Trump.

Slide4

That makes sense, doesn’t it? Many studies found the people who supported Trump were also the most convicted. Trump recently bragged he could shoot someone and not lose his supporters. One CNN reporter quipped “If Donald Trump punched a baby in the face, Trump’s supporters would say that baby had it coming.” But that also means if you weren’t swept up in Trumpomania three months ago, you aren’t likely to be swept up on voting day.

We saw the same thing in 2012. Ten days before the Iowa caucus, polls had Rick Santorum at just 8%. But he surprised with 25% of the vote! That’s, in part, because 39% of voters were undecided until the last minute. And among those undecided voters, 35% chose Santorum.

Slide5

The undecided voter will play a bigger part in the Republican race than in the Democratic race. We see that the polls were very accurate for Bernie and Hillary. But we also see that only 20% of voters were undecided heading into the Democratic caucuses. Again, this makes perfect sense. There are only two candidates, so it’s easier to differentiate their personalities and policies. And they are relatively divisive personalities. You either love or hate Bernie’s socialist ideas. And you either love or hate Hillary’s brash style. Nearly 60% had made up their minds several months ago!

Slide6

But it’s different for the Republicans. There are still 8 strong candidates duking it out, and it’s not so easy to distinguish their differences. You are likely to stay undecided longer and vacillate up to voting day. But as candidates drop out and it’s easier to distinguish the remaining candidates, voters will rally behind Cruz and Rubio – not Trump.

A Feb 4, 2016 survey by Public Policy Polling found that in a 3-candidate or 4-candidate race (Trump, Cruz, Rubio, Bush – strangely, they didn’t include Carson) voters rally behind Marco Rubio and vault him ahead of Trump.

Slide7And that’s one problem with the polls over the past several months: asking someone to choose among 14 candidates. That’s not how the primaries will play out in the coming months. Candidates will drop and we’ll be asked to choose from three or four candidates, not 14. Of the 8 remaining candidates, we’ll see more drop out after disappointing results in New Hampshire — probably Fiorina and Christie.

Those who remain will face a budget crisis – it’s expensive to keep campaigning beyond New Hampshire. Christie and Kasich each have less than $5 million in the bank (compared to $50 million for Cruz and $25 million for Rubio). Ben Carson recently slashed half his staff because of dwindling funds. Donors don’t like to keep spending money on a long shot. By mid-March we’ll be down to three or four candidates.

What does this mean? It means the January polls are not going to be good predictors of the Republican primary results. There will be a lot of people deciding on the last day. And those people not already committed to Trump won’t be easily swayed on the day of voting.

As the field winnows down, voters will migrate to the remaining candidates – likely Rubio and Cruz – and not the Donald.

New Hampshire will be a critical battleground. Trump has a commanding lead in the polls (29% vs 12% for both Rubio and Cruz). But in 2012, 46% of voters decided the week before the primary elections and all indications are that it will be similar this year.

Paul, Huckabee and Santorum have already dropped out, thinning the field – where will their supporters go?  Independent voters in New Hampshire can choose to vote in the Democratic or the Republican contest, further clouding where the votes will land. Strong showings by Cruz and Rubio in Iowa are likely to sway undecided GOP voters to put their votes “where they will count”. In fact the latest poll shows Rubio is up to 19% (Trump is still at 29%), illustrating the effect of the late deciders.

Trump will still be a formidable candidate throughout these primaries. But I expect undecided voters will put their weight behind Rubio and Cruz, especially as the field thins, and Trump’s results will be lower than the polls predict in most states. Come back later to see if I’m right.

And if you like my graphs, look for my new book “Storytelling with Graphs”, which shares all my secrets for analyzing data, choosing and designing graphs, and even using storytelling principles to make graphs more engaging. It should be released early this year.

Storytelling with Graphs cover

About the author: Bruce Gabrielle is author of Speaking PowerPoint: the New Language of Business, showing a 12-step method for creating clearer and more persuasive PowerPoint slides for boardroom presentations. Subscribe to this blog or join my LinkedIn group to get new posts sent to your inbox.


I just learned about #MakeoverMonday, an informal challenge and fun weekly event run by Andy Kriebel and Andy Cotgreave, where people do a graph makeover. This week, they asked, how can we improve this graph (raw data here)? As an educational opportunity, here’s how I approach it.

Hotel Revenue vs Travel Agents

1.Crossing lines

First, we have to be careful about letting lines cross on a dual-axis line graph. Research finds it’s more difficult to understand line graphs when the lines cross over.

The other problem is that when lines cross, it sends an unintended message that something has shifted. Something that used to be larger is now smaller. That’s not what we’re saying, and yet there is that subtle message out there. And in fact, there is no reason the lines have to cross. Think about it. They are only crossing because of the arbitrary choice of scales. We can change the crossover point by changing the scale.

 

Slide3

So, generally avoid having lines that criss-cross like this. Instead, change the scale so the lines never criss-cross, or use a combination bar/line graph, or use two different graphs.

Slide4

2. Graph title

Whatever point you are trying to make, write it out as the graph title. Right now, the graph title tells you what data is being measured, but not what conclusions you should draw. When you leave the point unstated, people may not get it, or may get the wrong idea. So spell it out clearly.

Slide5

3. Cause and Effect

I’m going to use two different graphs. Which one comes first? If you are making a cause-effect statement, as we are, put the cause first and the effect second.

Slide6

4. Graph choice

Now, I could use a line graph because nothing says TREND like a line graph. But I’m actually going to use a bar chart instead. Because I want you to get a sense of something DWINDLING down to nothing so I want there to be substance in the graph.

Slide75. Color

With bar graphs, use a soft color for the bars. When you use a dark color, it looks like zebra stripes and creates a moire effect, which is uncomfortable on the eyes. I’m also going to use a positive color for the growth of the online hotel booking revenues and a more negative color for the decline of the travel agents population.

Slide8

6. Pictures

No matter how hard you try, graphs will lack a lot of emotional impact. That’s because they are abstract, and appealing to people emotionally requires something more concrete. Pictures is the missing ingredient.

In particular, pictures of people give you the most emotional impact. And whenever I create a graph I want to ask myself “Are we talking about people right now?” Because if we’re talking about people, and the impact on people, I want to add pictures of those people.

In this case, we’re talking about the decline of travel agents (people!). We’re also talking about rising online hotel bookings. But hotel bookings don’t rise by themselves. They rise because someone is doing something (more people!). So I’m going to reflect that through pictures.

Slide9

7. Contrast

One last thing. Storytelling is about contrast. Good versus evil. Life versus death. Loneliness versus love. So if there’s contrast in the story we’re telling, we want to emphasize that. We’ve done that through the use of color (positive green versus negative red) and choice of pictures (happy older couple versus unhappy younger male). I’m going to punctuate that point with arrows (up versus down).

One blog visitor, Sheila B Robinson, pointed out in the comments section another reason for the arrows: some audience members are color-blind and cannot see certain shades of green and red. The arrows are another cue.

Another blog visitor, Andy Cotgreave who helps run #MakeoverMonday, also posted a helpful comment that the arrows I originally used are too large (see original here). I agree. I like adding annotations to graphs, like arrows, because they quickly summarize data. But the rule is to make them only as large as they need to be, and no more. And I agree the original arrows were like throwing one too many persons into the lifeboat, so here’s my revised final with smaller arrows. (I’ve also used one of my favorite techniques for titles: using ellipses to join two different graph titles into a coherent story.)

CrossingLines Final

Storytelling with graphs is not about accurately visualizing the data. It’s about determining the story in the graphs and then intelligently using design principles to bring that story into our minds and straight through to our hearts. If you want all my techniques for creating clearer, more engaging and more persuasive graphs, look for my new book “Storytelling with Graphs” which I’m hoping to release this year.

Storytelling with Graphs cover

About the author: Bruce Gabrielle is author of Speaking PowerPoint: the New Language of Business, showing a 12-step method for creating clearer and more persuasive PowerPoint slides for boardroom presentations. Subscribe to this blog or join my LinkedIn group to get new posts sent to your inbox.


Could Donald Trump ever be president of the United States? Recent polls have him the #1 candidate for the Republican party at 22%, 12 points ahead of his closest rival Jeb Bush. Pretty scary.

The answer is an unqualified no, he’ll never be president. Here’s why.

When the party chooses a presidential candidate, they want someone with the best chance of winning the election. Hard-core Republicans will vote for any reasonable Republican candidate. And hard-core Democrats will vote against them. So to start, they want a candidate who is popular in the party.

The more important battle is over independent voters, who make up 43% of all voters. But it’s not favorability they are concerned about, because independents tend to be moderate, easily swayed by good arguments that appeal to their self-interests. No, it’s their level of unfavorability. Which candidates do independent voters hate? Which ones will they oppose and never ever under any circumstances vote for?

If we look at a history of U.S. presidential elections, we see that the candidates most likely selected to represent their parties are rated high on FAVORABLE by their own party’s voters, and low on UNFAVORABLE by independent voters.

2008

In 2008, both the Republicans and Democrats were looking for presidential candidates. Hillary Clinton had the most support from Republican voters. But that didn’t earn her the nomination because she also had the least support among independent voters. And both parties chose the candidate favorable to their own party, and low on unfavorable with independent voters: Obama and McCain. (source)

2012

In April 2011, there were 16 presidential candidates. Trump was rated favorable by 58% of Republicans. But he was rated unfavorable by 57% of independents – not electable. Mike Huckabee had the most support from Republican voters, but Mitt Romney had less opposition from independents and was chosen the Republican presidential candidate. (source)

2016

And how about this year? Well, the numbers may change as we learn more about the candidates. But as of August 2015, Trump is once again in that quadrant with high Republican support but high opposition among independent voters. In fact, Jeb Bush is in that same quadrant! Even though Trump and Bush are the front-runners right now to get the Republican presidential nomination, it’s very likely that neither one has enough support to win an election.

So who are the candidates with high Republican support and little objection from independents? They are Ted Cruz, Scott Walker and Marco Rubio. Rand Paul is in the hunt as well. Carly Fiorina could also be a contender if she could get more Republican support; independents have little objection to her. (source)

And what about the Democratic party? Well, Hillary Clinton is considered the front-runner. But 60% of independent are opposed to her so she may be unelectable. The one with the right balance of support from Democrats and lack of opposition from independent voters is Bernie Sanders. (source) But he’s not a strong candidate, just outside of that magic quadrant. In fact, the Dems don’t seem to have any candidates with the right mix of high party support and low independent voter opposition.

Things are likely to change before the election. But at this point, my prediction is Bernie Sanders will be selected for the Democrats although he’s not super-popular among Dems. And notorious Tea Party trouble-maker Ted Cruz may actually be popular enough to get the Republican nod. But Trump has virtually no chance of representing the GOP. Check back in 2016 to see if I was right.

About the author: Bruce Gabrielle is author of Speaking PowerPoint: the New Language of Business, showing a 12-step method for creating clearer and more persuasive PowerPoint slides for boardroom presentations. Subscribe to this blog or join my LinkedIn group to get new posts sent to your inbox.

 


Chris Borland made headlines last week when he announced, after one year in the NFL, he was retiring because the risk of concussions was too great. It was especially pertinent given March 2105 is Brain Injury Awareness Month. The NFL responded that “football has never been safer“.

And that got me thinking. How BAD is the concussion problem in the NFL? Is the risk too great, as Borland says? Or is it relatively safe, as the NFL wants us to believe?

But first, let’s appreciate what it means to suffer a concussion.

 

The long-term complications from concussions can include chronic migraines,  fuzzy thinking, forgetfulness, seizures, early-onset Alzheimer’s, and depression, often leading to suicidal thoughts. One of the example is Andre Waters, a safety with the Philadelphia Eagles known for his punishing tackles, who shot himself at the age of 44, a few years after retiring from the NFL. Examiners studied his brain tissue and found it was like that of an 85 year-old dementia patient. Other studies have found that 76 of 79 brains of deceased NFL players also showed signs of the same degeneration. However, the NFL has always denied there is a proven connection between concussions and later brain degeneration.

Let’s start with the NFL’s position. They say that over the past three years, the number of concussions has been decreasing. And the data show they are right.

NFL Concussions are declining since 2012 But hold on. Whenever you look at timeline data, always be skeptical of the starting date. Because the story can change based on your starting date. If we go back to 2009, the first year PBS Concussion Watch started tracking official injury reports, we see that the concussion problem got progressively worse until 2012, and it is now abating. So, the NFL’s story is true, but only half of the story. (note: NFL reports different numbers than Concussion Watch).

NFL concussions 2009-2014

But are these numbers good? Is 111 concussions in the NFL too high? Or “safe”? There are 32 teams and 53 players per team, for a total of 1,696 players. If 111 get a concussion, that’s 6.5%. That seems pretty low, I guess.

But it turns out that the chances of getting a concussion are different based on your position. In 2014, only 2% of quarterbacks got a concussion. But 14% of cornerbacks. 2% seems pretty low. 14% seems kinda high to me.

In fact, if you look at the trends by position, you see the percentage has decreased for offensive players, especially running backs, tight ends and wide receivers. That’s in part because of the NFL’s new rules protecting ball carriers from vicious tackles, especially helmet-to-helmet. But they have stayed stubbornly high for defensive players, especially cornerbacks, safeties and linebackers.

NFL concussions, by position 2012-2014

This shows the percentage chance of getting a concussion in a single season. But what if you have a 10-year career? Clearly, the chances of getting a concussion sometime during your career will be higher the longer you play.

Chance of NFL concussion, by length of career

Most players don’t last 10 years. According to the NFL, the average NFL career is 6 years. So, assuming you are a rookie starting in the NFL, what is the chance you will get a concussion sometime during your 6-year career? The numbers are more sobering.

Chance of concussion during 6-year NFL career

Now we see the true extent of the NFL’s concussion problem. For the “speed positions” (the fastest players on the field), and especially those involved in the passing game, at least one-in-three will get a concussion during their career. If you are brave enough to be a cornerback or safety, one of every two players will get at least one concussion during their career.

So does the NFL have a concussion problem? That’s a normative question, based on what you think is “normal”.

But put the question this way: imagine there was a paint, and if you were exposed to it long enough there’s a 35% chance that the rest of your life will be complicated by depression, early-onset Alzheimer’s and suicidal thoughts. Would the government allow you to pain the walls of your workplace with this stuff?

Not likely. I’d say the NFL does have a concussion problem.

 

About the author: Bruce Gabrielle is author of Speaking PowerPoint: the New Language of Business, showing a 12-step method for creating clearer and more persuasive PowerPoint slides for boardroom presentations. Subscribe to this blog or join my LinkedIn group to get new posts sent to your inbox.


Facebook’s data visualization contest Viz Cup gives me mixed emotions — excitement about the growing interest in smart data visualization, and sadness about the poor data viz practices among the entrants.

Case in point is last year’s winner, Eric Rynerson, with this entry. Now, to be fair, Eric only had an hour to put this data viz together (he studied the data the night before) and there are a lot of good practices here. But as an educational opportunity for you, dear reader, I want to point out the weaknesses in this data viz and suggest a makeover.

Here’s Eric’s winning entry:

1. Bubble Chart

Let’s start with a critique of the bubble chart in the upper left. What I like about it:

  1. a. The graph title summarizes the main point, requiring less work for the reader
  2. b. It’s a good choice for comparing two groups. You can tell from the slope if there is a bias toward penalizing the Away team
  3. c. The graph is relatively subdued with light gridlines whispering in the background and muted colors for the bubbles

But here’s the things that don’t work

  1. a. First, the graph title is too small. It’s the same size font as the axes numbers. It should be more prominent
  2. b. The bubbles are different colors but I don’t know why. Is this so you can distinguish the tiny bubbles clustered in the lower left? Or is it to add variety to the graph? Generally, the reader will assume the colors have meaning. In this case, I’d avoid the randomness of the many colors and use red for the refs that hand out more red cards to Away teams, and gray for refs with no apparent bias
  3. c. And I’m not going to just use any red. I’m going to use the exact red on a penalty card because those who are familiar with that red have learned a subconscious reaction to that particular shade of red. And I want to generate that reaction
  4. d. The bubbles are also different sizes. Again, it isn’t clear if this is supposed to mean something or not. If it does, indicate that somewhere on the graph. If not, then keep them all the same size
  5. e. Eric has added a regression line to the chart. But a better option would be to add a 45-degree line, showing the expected position of refs who penalize Home and Away teams evenly
  6. f. You should also add text labels to each half of the chart, so readers clearly understand what it means to be above or below that line
  7. g. The annotation explains there is a similar but weaker bias for yellow cards. But the extensive text clutters the graph unnecessarily. Better to remove that annotation, or minimize the amount of text

 

So my makeover of the bubble chart would look like this

2. Column Charts

The column charts in the upper right are intended to appear as you click on each bubble. I like a couple of things about these charts

  1. a. First, I like that we’re seeing some trend data tracking how a referee’s bias may, or may not, change over the years. The basic format of a story is a beginning, middle and end and so a timeline graph gets us closer to a story
  2. b. Eric shows the absolute number of cards for Home and Away teams, but then he also explicitly calculates those differences and plots them on another chart
  3. c. The mirrored bar chart is generally a good choice for showing how two data series compare

What I don’t like

  1. a. The text block to the left of the graphs is supposed to invite the reader to click. But it’s too much text to draw attention. Better would be to use an image with a prominent call to click
  2. b. Almost always, time series data should go left to right. The reader intuitively understands this. A top to bottom timeline is less intuitive
  3. c. When comparing two data series, and one is clearly larger than the other, I prefer to use an overlapping column chart
  4. d. The legend at the bottom should be integrated into the graph, where it’s easier to read, and not set outside the graph
  5. e. The two charts should use a similar scale. Right now, the differences look enormous compared to the absolute values, just because that scale is stretched out
  6. f. I’m also not a fan of the monochromatic purple for the mirrored bar chart. The purple doesn’t complement the red and seems to be chosen at random. Better to select from our current red color palette and complement it with a strong but otherwise neutral black.

We end up with a graph like this:

 

3. Layout

Finally, we study the overall layout, including the title, pictures and flow through the piece. I like

  1. a. The attempt to add a (clip art) picture next to the title, increasing visual interest and suggesting quickly the topic of the data visualization
  2. b. The title summarizes the conclusions of the graphs, not leaving that work to the reader
  3. c. The use of faces is especially effective at drawing the reader into the data visualization and adding some human drama and emotion to what would otherwise be emotionally inert data

What I’d improve

  1. a. This clip art is weak, has no emotion and generally trivializes the importance of the piece. I’d choose an image with more emotion, like a referee forcefully holding up a red card
  2. b. The title is a bit vague. I’d write it more directly
  3. c. The face images below are good, but they are a bit remote from the column charts. In this case, the data in the column charts applies to only one referee — the one in the pictures below. We need to link this more clearly perhaps using a color band

Adding all these elements together gets us a final data visualization that look like this, and tells a better story too, don’t you think? (Hover over the image to compare with the original).

What do you think? Is the story clearer? What other changes would you make? Leave a comment in the comments section. And if you’re interested in knowing when my new book “Storytelling with Graphs” is available, just sign up for my LinkedIn group or subscribe to this blog.

About the author: Bruce Gabrielle is author of Speaking PowerPoint: the New Language of Business, showing a 12-step method for creating clearer and more persuasive PowerPoint slides for boardroom presentations. Subscribe to this blog or join my LinkedIn group to get new posts sent to your inbox.