Monday, August 31, 2015

Best Bang?

An article in Columns, UGA's faculty and staff weekly newspaper, notes that the university is #3 in "best bang for the buck" in the southeast.

A few points:
  • Read "best bang for the buck" the wrong way and you'll get in trouble. So just stop.
  • While UGA is 3rd in the southeast, the article fails to note it's 52nd in the U.S., down from 32nd a year ago. Ouch.
  • UGA is actually tied for 1st in the southeast. Just listed third. Six schools have an overall score of "12." So that 52nd above nationally? We're in a tie for that, based on points.
The 2015 national rankings are here.

Ahead of us in the southeast are East Carolina University and North Carolina State University. For the life of me I can't figure out the tie-breakers. It's not alphabetic, not any of the obvious metrics listed across the page. Anyone see it? If so, lemme know.






Thursday, August 27, 2015

Dog Owners

I missed doing this for National Dog Day or whatever the hell it was called, but here's a quick look at how dog ownership correlates with other factors.

As compared to non-owners, dog owners tend to be:
  • more politically conservative
  • more likely Republican
  • more likely to believe in a literal interpretation of the Bible
  • younger
  • less educated
  • watch less TV news
  • read fewer newspapers
This is based on 2012 national survey data provides by ANES. When I have time, I'm gonna do a more complete cat vs dog owners because I have both questions available in a nice, fat data set with lots of other variables.



Excessive Drinking

Athens-Clarke County, home of UGA. Party School. Downtown bars. Vomiting students. Underage possession. You know the story. So you'd expect Clarke County to rank high in the state in terms of percent who meet the "excessive drinking" threshold.

No, not really.

According to the 2015 Georgia health data, the statewide average is 13 percent. The "winners" in the county excessive drinking contest are (envelope please) ...

1. Dawson (18%)
2. Fulton (16%)
3. (tie) Muscogee, Forsyth (15%)
5. (tie) Cobb, Chatham, DeKalb, and Richmond (13%)

So where's Athens-Clarke, home of UGA? Would you believe 31st place? At 9 percent "excessive drinking?" Wow. I'm not sure that word, excessive, means what they think it means.

Methodology note -- not every county is measured on excessive drinking. Perhaps people there were too drunk to ask other drunk people how much they drink. So the rankings cover just 58 of 159 counties.


Media Use and Science Knowledge

While skimming the latest issue of Journalism and Mass Communication Quarterly (stop laughing, I publish academic work there too), I came across this study:

Science News Consumption Patterns
and Their Implications for Public
Understanding of Science

First, the title is missing a colon. All academic articles, to be truly academic, must include a colon. There's even a name for this -- titular colonicity. Serious flaw.

Second, it's an interesting academic study. The gist of it is this, that those who rely on online-only sources are more likely to have a better understanding of science, and this includes statistical controls for such factors as age, education and whether a respondent was a science major (a neat control). For you nerds out there, I draw from Table 2, which looks at predictors of science knowledge. These are unstandardized regression coefficients after lots of other controls.

Primarily television -- .04 (non-significant)
Primarily newspapers -- .02 (ns)
Equal online and media -- .02 (ns)
Primarily online -- .08*** (sig at .001 level)


Wednesday, August 26, 2015

More Fun with Frosh Names

I wrote about this the other day, my analysis since 2008 of UGA freshman first names. Read it and enjoy. Today, let's look at the 2015 freshman class, it's first names, and do a little quick-and-dirty comparison.

Okay, so here's the top 10 of the most recent freshman UGA class.

1. (tie) Emily and John
3. Hannah
4. William
5. (tie) Anna and Caroline
7. Sarah
8. Mary
9. Madison
10. Matthew

For fun, I picked a likely birth year for freshman (1996) and looked at the Social Security's listing of popular baby names for that year, looking only from Georgia, to see how the UGA first names compare. Below, the Top 10 for that year, in Georgia.


Boy’s Name
Girl’s Name
1
Christopher
Ashley
2
Joshua
Hannah
3
William
Taylor
4
Michael
Jessica
5
Austin
Emily
6
James
Sarah
7
Jacob
Alexis
8
Brandon
Brittany
9
John
Madison
10
Matthew
Kayla


So at UGA, John is the most popular boy's name, but in the state of Georgia in 1996 (a rough guess as to the birth year), it was 9th. The next most popular boy's name for UGA is William. It's third on the Georgia 1996 list. Matthew is 10th at UGA, 10th in the state. So there's a decent correlation here. Among girl's names, the top UGA one has always been either Emily or Sarah. Emily is 5th in Georgia that year, Sarah 6th.

So let's flip the data. Christopher is the top boy's name in 1996. Where is it in the 2015 freshman class? Would you believe 32nd place? And Joshua, Georgia's #2? At UGA, 49th. With the girls names, it's kinda the same. In Georgia Ashley was tops in 1996, but at UGA it's only 24th. Hannah is a better fit. At UGA it's 4th, in Georgia that year it's 2nd.

The Special Case of Madison

Madison has never made UGA's top 10 until this year. Indeed, in 2008 it was in 235th place at UGA. But look at 1996. It's 9th in Georgia. That explains, I suppose, its move up. Madison was 22nd in 1995 in Georgia. So it's growing popularity helped it slip into the UGA top 10.

Yeah, I do this with data all day.

Tuesday, August 25, 2015

Frosh First Names

Every academic year I ask UGA for the first names of the latest freshman class. There's a lot you can do with these data, but let's keep it simple and ask what's the most popular first name, and has that changed over time?

First, the Top 10 of Fall 2015:

1. (tie) Emily and John
3. Hannah
4. William
5. (tie) Anna and Caroline
7. Sarah
8. Mary
9. Madison
10. Matthew

Keep in mind these ranks represent small actual numbers. Of the 5,035 first names provided to me, 61 students are named either Emily or John. Third place, Hannah, has 56 students sharing that name. Here are a few interesting factoids:

  • I've asked for these data since 2008 and never has a male name done better than second place (also John, back in 2011).
  • Since 2008, either Sarah or Emily have held first place. 
  • Sarah fell this year, but it also fell in 2010. We're talking small numbers here, so there's some random walk involved in the data.
  • Madison has never made the Top 10. In 2014 it was in 15th place. In 2008, it was tied for ... wait for it ... 235th place? Wow.
Below is a quick-and-dirty image that shows the Top 10 across years.



I highlighted Emily, Sarah, and John to show how much they've dominated over the years. When I have more time, I can crank the data some more. For example, it's fun to compare what proportion of unique names (one person only with it) we see over time. My hypothesis has always been that as UGA improves in diversity, we should see more one-name-only students, but so far the data has not supported that hypothesis. Damn those data.

Finally, I leave you with a word cloud of the 2015 freshman names because all bright visual people hate word clouds, and I'm not a visual guy -- so there.

 

Thursday, August 20, 2015

Survey: Exactly How Many?

I often criticize national, regional and local news media on how they handle public opinion survey stories. Today it's The Red & Black's turn.

In an otherwise fine story about how the law has changed concerning underage drinking, this shows up in the fifteenth graf of the online version:
According to an anonymous straw poll conducted  by The Red & Black with almost 150 200 underage university students, the new law will not alter student habits.*
OK. I can quibble about using an N of 150 without stating its unscientific -- which is kinda the journalistic equivalent of why bother using it in the first place. Let's set that aside for the moment. Later in the story there's this from the fifteenth graph of the print version:
According to an anonymous straw poll conducted by The Red & Black with almost 150 underage university students, the new law will not alter student habits.
Weird. Different N for different stories, but wait. This kinda makes sense. The print version no doubt went to bed far sooner than the online version, so the N of 150 for print versus an N of 200 for online kinda makes sense, if you continued to collect surveys from your "sample" (note my intentional typographic sneer of the word sample). We know, from the story, that:
Over 90 percent of the non-freshmen students surveyed said they either illegally drank before the change and plan to continue, or they did not drink before and do not have plans to start now. Roughly two-thirds of non-freshmen respondents said they planned to drink underage, while less than half of the incoming freshman respondents answered the same way.
What's bothersome to me is there are no details on how this "sample" was collected. It's a straw poll, but what the hell does that mean? Did you stand at Tate Plaza and grab people? Survey folks at your favorite watering hole? Snag people hiking up Baxter Hill? Troll the dining halls? When you report a poll, even a bad poll like this, readers deserve just a little more information on how it was done. Spend a few words if you're going to report on a poll, just like you'd spend a few words describing a source quoted in a story.

Oh, and explain the discrepancy in your online version, saying it has updated numbers.

* I screwed up earlier, repeated the 150 twice. Online has 200, print 150. What I get for a rushed blog just before class.




Tuesday, August 18, 2015

Public Records Request ... for public records requests

Every year I ask UGA (where I teach journalism) for the same set of data and public records. I use these to show my students the kinds of stories that can be done. Today, let's talk about one of my favorites: a public records requests for ... public records requests. In other words, I wanna know what other people have asked for.

What? Why do this?
  1. Keep an eye on what the competition is asking for
  2. Fish for story ideas
  3. Because it's just fun to say
I have last academic year's requests. Two-thirds of them are for, sigh, sports. Most are for contracts, expenses, correspondence, free tickets, etc. (just kidding on that last one). Mostly it's about coaching salaries.

Setting aside sports, any good story ideas? Maybe. I plucked out a few:
  • Investigative files from the feds about a complaint filed against Dr. Richard Supilta. This one got covered by the R&B.
  • Someone wanting all studies mentioned by the UGA prez in his argument for merit raises for faculty, etc. 
  • All partnership agreements between UGA and a bunch of companies, like Apple and Google.
  • Records about an incident in at the Ramsey Center in which a student "was injured when she jumped from the 10-meter diving platform into the diving well below."
  • Oooh. A census of all dogs and cats used for research.
  • Personnel file for a (former?) employee.
  • Payments to a certain company associated with a guy named in the request.
  • Copies of all "employment agreements" between the present and former president of UGA.
Yeah, I didn't name names above. If you want to find them, make your own open records request for open records requests. But many years I've seen damn good stories. If UGA were my beat, I'd ask for this every month.

Monday, August 10, 2015

New Trump Poll

So this poll came out and, like others, shows Trump holding onto his lead after the GOP presidential circus debate.

Here's the lede:
Donald Trump continued to defy the laws of political gravity on Monday as a Reuters/Ipsos poll found the real estate mogul holding onto a wide lead among Republicans in the U.S. presidential race despite an acerbic debate and a feud with a female television anchor that have bolstered charges of sexism.
OK, wow. Trump is Teflon.  He has, according to this poll, 24 percent. In second is Jeb Bush, at 12 percent.

This is a Reuters/Ipsos poll and I have to point out something at the very bottom. Here it is:
The online poll of 278 self-identified Republicans has a credibility interval of 6.7 percentage points.
Two things should jump out at you, the educated reader of polls. First, "online poll." Second, only 278 people participated, meaning a margin of error (credibility interval? what the hell is that?) of almost 7 percentage points. That 7 percentage points means Trump's numbers could be as low as 17 percent, and Bush's could be as high as 19 percent. In other words, it's a statistical tie. I should point out that given other recent polls, Trump's lead is real, but the story should point out the statistical tie.

Saturday, August 8, 2015

Why Even Bother?

I love polls. I teach a graduate seminar every year in public opinion. But how some polls are presented or reported makes me nuts. Here's one set of crosstabs of an August 6 Landmark/RosettaStone poll of 600 Georgia Republican voters. The news angle is obvious -- over a third favor Donald Trump, while Jeb Bush and Scott Walker are the only others in double figures. But look across to the race breakdown. Fifty-seven percent of blacks favor Trump. You'd think, holy cow. That's amazing. Look harder. Apparently this survey could only find seven -- yes, seven -- black Republicans. So 57.1 percent of 7 means that four (yes, 4) black Republicans favored Trump.

Sigh. Just collapse racial breakdowns when they're so meaningless, though I admit there's a certain entertainment factor that a survey can only find so few black Republicans.

Here's a story on the poll itself.

What's Next in GOP Debates?

OK, the Fox debate was great fun. So what's next?
  • September 16, on CNN (an "A" and "B" debate)
  • October 28 on CNBC, nothing yet on who is invited
  • November (no date given) on Fox Business Network (honest, I forgot it still exists)
  • December 15 on CNN
  • January (no date) 2016, on Fox News Channel
  • February 6, 2016, on ABC
  • February 13, 2016, on CBS
  • February 26 on NBC and Telemundo (Trump irony alert)
  • March (no date) on Fox News
  • March 10 on CNN
And, of course, the primaries themselves will do much to winnow the Group of 18, or however many there are.

Friday, August 7, 2015

Carson Won?

There's at least one poll out that says, believe it or not, Ben Carson won last night's GOP presidential debate. No. Really.

Now this is a poll conducted by Gravis Marketing, called in one story the worst poll in America. It's a robo-poll conducted right after the debate of registered Republican voters. It's key to note it's not of likely primary voters. Still, let's get to the results. See the screen grab below.


Carson won, followed by Trump, within the 3 percent margin of error. After that comes Rubio (who did have a good debate) and Bush (who was mediocre) and Huckabee (who was kooky). Kasich and the rest are all kinda close.

That Paul lost the debate, that's actually my sense as well. That Trump finishes second as both having "won" and "lost" the debate fits nicely with his high negatives. If you didn't like him before the debate, you sure as hell didn't like him after.

Again, don't make much of this. It's a robo poll, which can result in a biased sample, and it's of registered, not likely, voters. Still, it is kinda interesting as I don't know of any educated analysts who would say Carson won. For much of the debate, he looked frightened by the flood lights. He did well later, so perhaps his strong ending helped.

Who won? No one. But Rubio and Kasich helped themselves the most.

Who won?

Who "won" last night's GOP debate? I'm still waiting on a legitimate poll, so instead let's look at some crap polls done by various media organizations. SLOPs, these are called, self-selected polls. In other words, complete bullshit.

According to Fayetteville Observer, it was (1) Trump, (2) Carson, and (3) Rubio. Really? Carson? Other than one good line, he bumbled his way through the night and most of the time fit that old cliche of a deer caught in the headlights. The Roanoke Rapids paper says, well, exactly the same, because they're using the same site as the place above. Now the sites give us the percentages, but not the raw votes. See their "results" below, at least so far (as of 12:58 p.m. today).


As readers of my blog know, I hate SLOPs, and I firmly believe that while they can be entertaining and are okay on silly topics, on serious matters no journalism organization should use them. At least no legitimate journalism organization. Yes, I'm looking at you, papers above. Stop hurting America (written in honor of Jon Stewart).




Thursday, August 6, 2015

Trump and Georgia

A new Georgia poll has Trump way ahead of his competitors for the Republican nomination. Forget the news stories about the poll. Instead, look at the actual results (pdf).

For you nerds out there, first the methodology. It's a survey of 569 likely Republican primary voters, with a 4.1 percent margin of error. The data were weighted by age, race, and gender to correct for the survey not matching the 2012 primary voter stats. IMPORTANT: this was a robo-poll of landline only phones. Remember landlines? Well, when people couldn't be reached by landline, they were reached by mobile. Not sure exactly how that works, but 83 percent of all completed surveys were by landline. Skewed sample.

OK, the survey details aside, a few results
  • Trump is way ahead at 30.4 percent. 
  • Bush holds second at 17.3 percent. No one else in double figures unless you round up, and then Carson has 10 percent.
  • Undecided is only 3.8 percent, but it bests several other candidates.
  • Trump sucks among 18-29 year olds, tied for 4th place with "someone else." Then again, how many 18-29 year olds have a landline phone? That age group made up only 6 percent of the sample. In other words, about 34 people. In even other words, meaningless.
  • So how about black Georgia Republicans? It's Trump, then Walker, then Carson (who is black). Then again, this is based on about six people. Should you even report such a breakdown? No, not at all. Never.
So what do we learn from all of this? Not a hell of a lot, since much of Trump's popularity seems to be a combination of voter anger, media exposure, and possibly drawing in folks who normally and rationally would be tuned out of the process more than a year ahead of an election.

Then again, no polling this early matters a lot, which is why many pollsters cautioned Fox and others to not rely on them to decide who gets in the main debate tonight versus being sent to the kids' table.



Do Debates Matter?

Tonight at the big kid table, 10 GOP prez candidates will line up behind podiums exactly 42 inches apart and debate. No one knows what will happen. Will Donald Trump go postal? Will Jeb Bush stumble? Will any of the others score?

My question is ... do we learn anything from such debates?

The research says yeah, some. And post-debate analyses can subtly, and sometimes not so subtly, influence who we think "won" and who we think "lost," which has other consequences. Post-debate analysis has changed people's minds. Not often, mind you, and perhaps most famously in 1976 when Gerald Ford "freed" Poland. People thought he did well, then the post-debate analysts told them otherwise, and a couple of days later opinions changed.

But, and this is important, nearly all the research on debates and post-debate analysis was focused on those final debates between the Republican and Democrat (and rarely, independent) nominees. This is a GOP debate in August, more than a year before the general election, at a time when -- still -- only the truly politically interested are paying attention. As such, don't expect a lot of movement in opinions and support. A point here, a point there, that's about it, short of some dramatic moment when someone forgets the name of the third branch of government they want to cut.

Why is Donald Trump doing so well? Because, in part, he's drawing in people who normally would not be paying all that much attention. He's the wild card, as everyone knows. If he successfully zings some candidates and remains unscathed, he wins. If his lack of policy knowledge is exposed, he loses (a little). If someone hammers him right back ("at long last sir, have you no sense of decency?") then he loses even more.

Of course, who won and lost will be spoon fed to us by the networks. That's their job.



Monday, August 3, 2015

This Week's Favorite Surveys

I love to scour the net for the latest and greatest of surveys. Here are a few that caught my eye.

Pig Farmers: Yes, a survey of pig farmers in the UK to tackle the question of resistance to worming.

Financial Literacy: Yes, this is always low, and in Australia among the young it's even worse, according to a survey.

Hispanics and Social Security: A survey finds Hispanics lack knowledge of their benefits. See a trend here? Surveys, often sponsored by special interests, always find a lack of knowledge about their particular area.

COPD: Such as this condition, according to a survey by, yup, COPD folks.

Loneliness: Yes, the campaign to combat loneliness has a survey about loneliness, especially among the aged.

Debate Drinking Game


As you no doubt know -- and have already stocked up on popcorn and beer for -- the GOP debate is Thursday night. We'll learn Tuesday which 10 of the 17 announced candidates earn a podium on the main stage and which  are voted off the political island and into debate oblivion.

I take this stuff seriously. So in that spirit, I ask -- what's a good drinking game for Thursday night? Some suggestions:
  • Every insult by or toward Donald Trump
  • Every time Hillary Clinton mentioned
  • Every time Trump touches his hair
  • Mentions of Obama
  • Mentions of ISIS/ISIL/ETC
  • Mentions of Obamacare
  • Every attack on Jeb Bush 
  • Every time Trump is ignored
You get the idea. I think the insults by or toward Trump will provide you with an adequate buzz, suitable at least for a Thursday night. If the debate were on a Friday night, we'd have to go to several of these to really get a buzz on.

Other suggestions welcome.