"Pharos Research Group" first came to my attention on Wednesday via a very thin piece in The Hill regarding a new ND-Sen poll purporting to put Democrat Heidi Heitkamp up 49-48 over Rick Berg.
Curious to know more about this firm, I started Googling. In particular, I was intrigued by Pharos's claim (in The Hill's writeup) to have conducted prior ND-Sen polls before. Since we watch all polling like a hawk here at Daily Kos Elections, I was surprised to have never seen a single Pharos poll before—nor even heard of the firm. (The Hill claims that Pharos's principle, Steve Lachtman, is a "former Gallup pollster"—not exactly a ringing endorsement, these days.)
There were extremely few hits for "Pharos Research Group" (in quotation marks) on Google when I first started looking. (Though now there are more, thanks to me—more on that in a moment.) I found their website easily, and I also noticed they had a whole ton of essentially unreleased polls. So I took the liberty of re-uploading all of them to Scribd (which is why there are now more hits when you search their name). You can find them all in one pile here.
So, on top of ND, that's another 7 polls: FL, IN, MT, NE, OH, PA, and the city of San Diego. There was also a national poll I didn't bother with, making nine in total. The more I explored, though, the more red flags I began to notice:
- In their methodology statements, Pharos says they used live callers in every instance (and called land lines only, FWIW). That would mean we're talking about, very roughly, at least $100K worth of polling.
- In a number of the writeups, they refer, weirdly, to prior polls—often multiple prior polls—they claim not only to have taken but to have released. Example from NE:
In the Senate race, the race has tightened significantly against the previous three polls we have released. The race is no longer outside the margin of error. In four weeks, Kerrey has managed to chip away at Fischer’s support, and pick upundecideds at a fair rate.
Like I say, we've never seen a single Pharos poll before, and had never even heard of the firm. (And oh yeah, they somehow have Deb Fischer up just 2 points over Bob Kerrey. That's their craziest result. The rest are all what I'd characterize as leaning Dem—a synopsis is below the fold.)
- They report numbers to the hundredths of a percent. This kind of false accuracy is a hallmark of questionable pollsters.
- They went to the trouble of testing ND-Gov (and every other imaginable statewide race in all these states) but didn't bother with IN-Gov, for some odd reason.
- They mis-spelled Kim Gillan's name as "Gillen" in MT-AL.
- Their website "about page" is super-thin and doesn't mention a single name.
- They claim to have been in business for some 30 years, but I can't find a single payment to Pharos on CQ's Political MoneyLine.
And that's only what I've learned so far. There may be more reasons for skepticism that I haven't uncovered yet. So I'm curious to know what, if anything, anyone else knows about this firm, and what other red flags you might have spotted in their poll releases or data.
Polling Summary:
FL-Pres: 47-47 Obama (D)
FL-Sen: 55-44 Nelson (D)
IN-Pres: 51-38 Romney (R)
IN-Sen: 46-46 Mourdock (R)
MT-Pres: 47-41 Romney (R)
MT-Sen: 48-46 Tester (D)
MT-Gov: 47-45 Bullock (D)
MT-AL: 47-41 Daines (R)
NE-Pres: 51-41 Romney (R)
NE-Sen: 48-46 Fischer (R)
ND-Pres: 49-39 Romney (R)
ND-Sen: 49-48 Heitkamp (D)
ND-Gov: 63-32 Dalrymple (R)
ND-AL: 54-42 Cramer (R)
OH-Pres: 50-45 Obama (D)
OH-Sen: 52-41 Brown (D)
PA-Pres: 49-46 Obama (D)
PA-Sen: 52-42 Casey (D)
San Diego Pres: 47-41 Obama (D)
San Diego Mayor: 47-41 Filner (D)
6:32 PM PT: Dreaminonempty points out that Pharos's web domain was created only on Oct. 1 of this year, according to a WHOIS search.
7:14 PM PT: I've received an interesting and informative email from Pharos's Steve Leuchtman (not "Lachtman," as The Hill had it). He's given me permission to post it in full:
David:
To answer the questions you raise in order:
1) Cost is SIGNIFICANTLY lower than that. Say what you want, we use low cost call centers, cost of polls is nowhere near 100k.
2) We have been doing prior polls, and releasing to media. Doesn't get any attention, not one freaking play until this most recent ND Senate poll.
3) Odd numbers in N (I can explain that if you like). Report to hundredth because that's what we came up with.
4) I live in ND. I don't live in IND. The two races are about as competitive.
5) Right. Thanks for the correction.
6) Website is place holder while we complete the more thorough one. We're primarily in commercial market research, surveys for hospitals, clinics, etc.
7) I worked at Gallup for a number of years in the 1990s (as you say, right now, not exactly an endorsement). Michael, my partner, has worked in local politics off and on since the 1980s in CA.
On to the substantive points.
1) I'm skeptical of the NE Senate numbers too. I suspect the poll is an outlier. On the other hand, there hasn't been a ton of polling done on the race. It is certainly possible it has been tightening. That said the respondents we got were a little tilted to Dem/Independents, and a little heavy on the 2nd CD, which means it tilts towards Kerrey.
2) We are relatively new to the political polling business as a business entity, and the basic response to our report releases has been essentially... "you aren't on the aggregators, you aren't serious, we won't report it."
In short, your response seems to be the same as theirs, I haven't heard of them, they must not be real. Honestly, I understand that. And honestly, if you want to feel that way, you are welcome to. The numbers are well inside the trendlines (with the exception of the NE Senate, a number that as I acknowledged, I'm skeptical of as well.) The voter models are solid, let's compare our results with the election results on 11/6.
If you have any questions, I would be happy to answer them for you.
Thu Oct 25, 2012 at 8:29 AM PT: Leuchtman has provided me with copies of his firm's prior round of polls taken in mid-October. You can download them all here (warning: ZIP file) if you want to take a look.
