Friday, July 29, 2016

About That UGA Climate Survey

Last year UGA did a "climate" survey. Not climate as in weather, as in warming, but climate as in diversity and that sort of thing. I wrote about it a few days ago, based on this Athens Banner-Herald story. Now I can provide you the link to the actual study, which the ABH story did not do for some reason, and discuss it in some small detail.

First, here is the link to the study page. You can read the full report (all 447 pages of it), a summary, or look at specific parts. Enjoy.

Back to my points. First, I requested from UGA a copy of the report and it provided me with the link, which is terrific. I also asked for a copy of the actual data. That request was denied because the data is part of ongoing research. Here's the open records line quoted to me. My discussion follows, but I boldface the part I find of interest.
This specific exemption applicable to those records is O.C.G.A. § 50-18-72 (a)(36). That exemption applies to: Any data, records, or information developed, collected, or received by or on behalf of faculty, staff, employees, or students of an institution of higher education or any public or private entity supporting or participating in the activities of an institution of higher education in the conduct of, or as a result of, study or research on medical, scientific, technical, scholarly, or artistic issues, whether sponsored by the institution alone or in conjunction with a governmental body or private entity, until such information is published, patented, otherwise publicly disseminated, or released to an agency whereupon the request must be made to the agency. This paragraph shall apply to, but shall not be limited to, information provided by participants in research, research notes and data, discoveries, research projects, methodologies, protocols, and creative works.
By providing me a link to a published study and summary, isn't that published? My read of the open records law says, by publishing the full study, the data is now available. I've asked for a clarification on exactly that point, but I'm hardly in the mood to fight it out in court or anything like that. The stakes are too low.

Back to the report itself. I don't care so much for the results except how the methodology skewed them. See my earlier post linked above, but basically this is a self-selected survey, meaning anyone could participate, and it turns out that of the potential respondents among students, staff, faculty, and administrators, one-third of the same are admins and staff. That's so off as to make the results questionable.

So why not a professional job with a random sample? Here's the answer:
The goal of the survey was to hear as many voices as possible. Rankin and Associates Consulting recommended not using random sampling because that methodology may inadvertently exclude populations where numbers are very small (e.g., Native American faculty). The survey was open to all faculty, staff and students to include the broadest array of perspectives possible.
First, the goal of research is to provide a useful, a generalizable answer. This fails. That said, they've done the same thing at other places, such as this Dartmouth study. According to Rankin & Associates, the consulting firm, for response rates less than 30 percent "caution is recommended when generalizing the results to the entire constituent group." The UGA study fell below that threshold, plus I'm unfamiliar with any magic assigned to the 30 percent response rate. It may be true, I've just never heard it before.

Second, oversampling certain populations is accepted practice, if done via random sampling.

When you have a SLOP (a self-selected survey), you get far more responses from people either very pissed about something or those who need the results to look good (i.e., admins and their staff). In other words, the results are deeply biased, deeply flawed.

On another day I want to break down the respondents. It's surprising. For example, 224 of the respondents work in the vice president of student affairs office, 84 for the VP of instruction, 44 of the respondents work in the provost's office, and so on ... including 19 from the office of the UGA president. Make of all this what you will.



Thursday, July 28, 2016

Campaign Finance

No one likes political advertising. Well, TV stations like it. And media buyers. But most of us could live without it or, at least, a lot less of it.

So how about this survey question:
Currently, groups not working with a political candidate may spend unlimited amounts of money on advertisements during a political campaign. Do you favor, oppose, or neither favor nor oppose placing limits on this kind of spending?
About a third of 1,200 respondents in a national survey said they favored this "a great deal." Nearly a third neither favored nor opposed it. A measly 9.4 percent opposed this "a great deal" and another 4.3 percent opposed it "moderately. In other words, the results are strongly skewed in one direction, that of limiting the money spend on ads during a political campaign. Click on the graphic below for a better look.


It's no surprise that there's a strong partisan division on this question and, hell, just about everything these days. Republicans were more likely to oppose it, Democrats were more likely to support it. Even so, let's look at those Republican responses. Among self-identified GOPers, most were more likely to fall in the "neither favor nor oppose" camp (37.1 percent). Only 12.1 percent of Republicans opposed it "a great deal." More fell the other way, with 16.6 percent of Republicans favoring it "a great deal" and another 14.3 percent favoring it "moderately." Let that sink in. Yes, the numbers are nowhere near the Democratic response to this question (for example, 41.2 percent of Dems favor it "a great deal"), but still it's kinda surprising.

I got curious about responses to this campaign finance reform question, so I dug deeper in the data. Warning, statistics ahead. I constructed a model that allows all the various factors to statistically control for one another and see which ones matter, which ones don't. I even created a variable based on whether respondents live in battleground states from 2012 and who may have been exposed to far more political ads than the rest of us.

Let's skip to the meat of the analysis. In my model were the usual socio-demographic variables, like age and income. It also included the state you lived (battleground or not), whether you voted in 2012 or expect to vote in 2016, whether you give money yourself to political campaigns, your political knowledge (based on three questions), how much you follow politics, and of course your party identification and ideology. The winners? Party ID and ideology, by far. For you nerds out there, the 13-variable model included few statistically significant results once Party ID and ideology stole all the variance. Party ID (high is Republican, beta = -.12, p<.001), and Ideology (high is conservative, beta = -.33, p<.001) explained most of the model. A few others emerge. Race (whites more likely to favor limits, beta = .39, p<.01), and an interest in politics (beta = .24, p<.01).

I thought for sure if you lived in a battleground state (Florida, Virginia, etc.) you'd be sick of ads and more likely to support limits. Nope. The regression coefficient is non-significant, even before we enter party ID and ideology. There's simply nuthin there. I was a bit surprised at the lack of a relationship with having voted, or expecting to vote. No idea why as to my surprise, but I kinda guessed voters may be a little more forgiving of political ads than those who don't vote and might prefer the whole election thing would simply go away.

Methodological Details

The "battleground" states based on political advertising spending in 2012 were: Colorado, Florida, Iowa, Nevada, New Hampshire, North Carolina, Ohio, Virginia, and Wisconsin. Spending in these states far outstripped that of others.

The survey and data are available here. Do your own analysis.

Final Model Results (beta weight in parentheses, * if statistically significant)

Age (-.03)
Race (.09*)
Sex (.05)
Income (.00)
Education (.05)
State (.02)
Voted 12 (-.03)
Will Vote 16 (.04)
Give $ (.00)
Party ID (-.13*)
Ideology (-.20*)
Political Knowledge (-.00)
Political Interest (.11*)













Wednesday, July 27, 2016

Will UGA Change Admission Policy?

Ever since a U.S. Supreme Court decision upholding the limited use of race in the University of Texas admission policy, I wondered if UGA may change its system.

Apparently not. Though it's hard to say. Maybe. Or not. This story just posted in my local paper suggests no changes are gonna happen. Here's the important, or actually unimportant, graph:
“The university’s current admission decision practices and procedures enroll an academically strong and diverse class and is working well as indicated by our first-year retention rate of 95 percent,” said Jan Gleason, UGA’s executive director for strategic marketing.
God save us from friggin flacks who say things without saying anything. Lemme guess, Jan Gleason. You wrote an email response. Lemme guess, ABH, you let her get away with an email response. Our "strategic marketing" person has managed to strategically say ... nothing. In fairness, maybe she said we have no plans to change anything and the ABH managed to leave that vital tidbit out of the story. All we get is it's "working well." If you read the story, you'll find no other support for the hed, which says:
UGA won't change admissions policies to allow race as a factor
I can only assume there's other stuff, not in the story, that supports this hed. It'd be nice to have it in the story.




Tuesday, July 26, 2016

107.77.235.84 Hates UGA

Meant to get around to this sooner but a talk I gave today reminded me of it. If you look at the "history" of edits on UGA's Wikipedia page you'll find weird edits by someone from IP 107.77.235.84. Here are a couple in which the UGA entry was changed to odd stuff, just brief lines.



You get the idea. If you look at the UGA history, you'll see the page has been "protected" a few times due to "persistent vandalism." The IP is via AT&T mobility out of Atlanta, but there's no way to know exactly where or who it is, that's just the carrier. This IP has made the two changes above, but one other time "blanked" the UGA page. In all cases it was caught, replaced, and locked for a time. It's open now.

Indeed, earlier strange edits to other pages ended with this warning:


Yes, stuff like this fascinates me. I suspect this happens to other schools as well, I just haven't taken the time to see.






Hillary Clinton, Lucifer, and Ohio

PPP loves throwing odd questions in its surveys. In an Ohio survey this week it asked:
Do you think Hillary Clinton has ties to Lucifer, or not? 
  • Clinton has ties to Lucifer 19%
  • Clinton does not have ties to Lucifer 63%
  • Not sure 18%
So nearly 1-out-of-5 have magically sensed her ties to Lucifer, and 18 percent just aren't sure. But if we view just Republicans (see below), a third of them are certain about her Lucifer-ness. Among Independents, 1-out-of-5 are convinced she's in league with the Prince of Darkness.


Democrats
Republicans
Independents
Clinton has
ties to
Lucifer

  6%

33%

20%
Clinton does
not have ties
to Lucifer

86%

43%

56%
Not sure
  8%
24%
24%

This is a good time to point out my favorite all-time quote about defining Public Opinion is from a play entitled, yes, Prince Lucifer.
Public opinion is no more than this,
what people think other people think.


Hillary Clinton, Lucifer, and Ohio

PPP loves throwing odd questions in its surveys. In an Ohio survey this week it asked:
Do you think Hillary Clinton has ties to Lucifer, or not? 
  • Clinton has ties to Lucifer 19%
  • Clinton does not have ties to Lucifer 63%
  • Not sure 18%
So nearly 1-out-of-5 have magically sensed her ties to Lucifer, and 18 percent just aren't sure. But if we view just Republicans (see below), a third of them are certain about her Lucifer-ness. Among Independents, 1-out-of-5 are convinced she's in league with the Prince of Darkness.


Democrats
Republicans
Independents
Clinton has
ties to
Lucifer

  6%

33%

20%
Clinton does
not have ties
to Lucifer

86%

43%

56%
Not sure
  8%
24%
24%

This is a good time to point out my favorite all-time quote about defining Public Opinion is from a play entitled, yes, Prince Lucifer.
Public opinion is no more than this,
what people think other people think.


Support for Campaign Finance Reform

Here's a survey question for you:
Currently, groups not working with a political candidate may spend unlimited amounts of money on advertisements during a political campaign. Do you favor, oppose, or neither favor nor oppose placing limits on this kind of spending?
This is from the 2016 ANES pilot study, a survey designed to test new questions for possible inclusion in the traditional ANES surveys conducted before and after each election. In our case, it allows us to see not only what people think about money in political campaigns (other surveys ask this kind of question too) but also to examine it in relation to other factors, from the obvious (party identification, ideology) to the less obvious.

First, let's look at the breakdown. There's a graphic below, but as you can see, in a national survey of 1,200 respondents, more favor than oppose limits. The key here may be use of the word "advertisements," as no one likes them. A lot of folks, obviously, go to the middle and easy answer, a result we often see in surveys. Clearly, though, the data are weighted toward the limiting of campaign spending on advertisements.


As you'd expect, the more strongly you describe yourself as Republican or conservative, the more you oppose such limits. Also the more interest you have in the news, the less you like the idea. The more educated you are, the more you favor limits. Also interesting is political knowledge is unrelated to an opinion about limits, as is age. Whites favored limits more than blacks, women slightly more then men.

A real test, of course, would be to set these factors up to compete with one another to see which ones truly predict opinions about campaign finance. My very quick and really dirty regression analysis, in which all the factors statistically control for one another, says party identification and ideology trump most other factors. Even so, a little statistical room is left for education and interest in news. The other factors drop out.

Like so many other issues, this one appears to be largely partisan.