I'll start by saying, I've only done one canvass, though I've passed up opportunities to do several. (I have had reasonable reasons, but I feel I could have done more so far) This is as much for me as it is for everyone else.
I'll also state that I do network and regulatory stuff at a competitive phone company (Not an incumbent like AT&T and Verizon). I write from the viewpoint of someone with an intimate understanding of telecommunications network infrastructure, and a professional daily working understanding of the telecommunications regulatory environment at the state and federal level, though I am not a lawyer.
I also wrote and run a website that provides telephone numbering and network information for free that is used and relied on by hundreds of phone companies every day.
However, I am only loosely familiar with polling, and know enough statistics to predict whether I'll run out of cash in the bank before I get my next paycheck or not. So take what I say with a grain of salt with respect to that.
And this is my first diary, so if I'm doing it wrong, I apologize.
I hear an oft repeated refrain when it comes to political polling, and that's that it ignores cell phone users, and favors old people who "still have their landlines". (I'm 27 and still have mine, but I may be an anomaly as I need something that works no matter what, as if my network is having problems, I still need to make phone calls.)
I think this claim has merit, but there's something else that's being missed.
I'll have to go into a bit of detail how telephone numbering and directory data work. As this is the only real visibility that a pollster can realistically have into dialing random telephone numbers with any reasonable degree of success, it's important to understand.
Firstly, before all the competitive carriers like the one I work for came around, telephone numbers were more or less assigned by fiat, and coordinated through a branch of the Bell System called Bellcore. (This was later spun off into two responsibilities - the NANPA for assignment and recording, and Telcordia which is what bellcore became after divestiture and subsequent spinoffs and mergers. They now are the defacto registrar for routing information) The government did little to decide what number blocks go where, because frankly, they didn't care. In the old days of telecom (I'm talking about the mid 1980s here, but the telephone network started in the late 1800s), telephone numbers were assigned to an exchange. I'll use real world examples here:
248-350 - Southfield, MI
248-591 - Royal Oak, MI
616-842 - Grand Haven, MI
As telephone numbers are 10 digits long, each of these provided 10,000 numbers that could be provided to subscribers. I'll refer to them as "Ten Thousand Blocks" (or 10k blocks) going forward. Well, when us raggedy upstarts in the competitive industry came along, and the PCS and Wireless carriers came along, we needed phone numbers to assign to new customers. Since it's hard to share a block of numbers, and at the time, technological means didn't exist to do proper number portability, we needed a way to get our own numbers. So we got our own 10k blocks too!
(In the interest of avoiding any possible reprisal, I'll be specifically avoiding numbering resources belonging to my employer, so if you see something here, it's NOT where I work. :) )
So now you have:
248-350 - Southfield, MI (AT&T Midwest)
248-440 - Southfield, MI (Long Distance of Michigan)
248-915 - Southfield, MI (Verizon Wireless)
248-591 - Royal Oak, MI (AT&T Midwest)
248-298 - Royal Oak, MI (Long Distance of Michigan)
248-217 - Royal Oak, MI (Cingular Wireless)
616-248 - Grand Haven, MI (AT&T Midwest)
616-223 - Grand Haven, MI (Lucre)
616-638 - Grand Haven, MI (Nextel Communications)
Obviously, this doesn't scale well when you have thousands of entrants. This is why you had so many area code splits and changes back in the late '90s. Everyone was starting a phone company, and they all needed numbers. Each 10k block can only be assigned to one region and carrier for technical and billing reasons, so everyone needed one, everywhere. (we've since as an industry came up with ways to split up 10k blocks into smaller "thousands" blocks, but that has little to no effect on what I'm saying here, so I'll leave that part at that)
Of course, the majority of customers were then, and are now, still customers of the incumbent carrier. So if you're going to spend efforts dialing randomly, you're going to call the incumbent carrier's numbers, and ignore the competitors. They often port over the incumbent numbers anyway, and most of the newly issued numbers from the competitors have typically been modem banks and business lines, neither of which are good for pollsters.
There's a few problems with this though - you skip wireless customers, possibly on purpose, because there until recently been considerable expense in getting a wireless call, and nobody wanted some law to come into place banning the practice, and possibly more importantly for the pollster, you don't REALLY know where the person is! I know a number of people who have a cell number, move across the country, and keep the number because everyone has it and they don't want to change it. So if you call a bunch of Detroit cell phones, how do you know for sure you're reaching Detroit residents? There's another problem though, and I'll get into that here - directory information.
So everyone looks at me here for a moment, and says "huh? What do phone directories have to do with it?". Well, pollsters getting information directly from the phone companies about who has a number and even whether it's active or not would run afoul of federal privacy laws. So the only real source of information you can get on someone (unless you're Carly Fiorina, anyway) is to get directory data. But isn't something missing from the phone directory?
Of course - unlisted numbers.
Originally, these cost plenty of money so people didn't get them. But there's something interesting about directory listings for competitive carriers - you do unlisted numbers by "dropping the listing" - literally removing the data from the directory database. This costs nothing to do because there are a number of reasons it can be done, all of them legitimate (disconnection, reassignment of the number, or if the number is part of a multiline hunt group and you have a listing for the main phone number, or of course, if you want the number unlisted AND not available to directory assistance. You can pay a bit to have it visible to directory assistance and not appear in the phone book, though). Anyway, so since this is free to the competitive carrier (and thus, it's easy to pass it on), and many people found that they got a lot of calls from solicitors, and everything else, by not being listed in the directory, there is little impetus to add people to the directory when signing up new customers in your own number blocks. You can sell it to people as a free service, and for those who are privacy conscious, that's a savings of $5-$10 bucks right there, making it a no brainer to switch (or sign up for new service) even if the rest of the deal was a breakeven.
How do you most efficiently target your autodialers to target numbers that are connected, working, and have people answering them, and avoid calling a bunch of fax lines, elevators, payphones and other undesirable numbers? You look in this giant listing of active telephone numbers that lands on your doorstep for free every year! Conveniently, it even has names attached, so you can match it up with the qualified voter files from the state you're working in, and call ONLY registered voters. Neat!
Except we have a problem now. Cell phones are cheap, and everyone seems to have one nowadays. If they have a home phone, they rarely use it, and keep it around for emergencies, or the babysitter, or their directv or tivo requires it, or some similar thing. Or they have this nifty thing called VoIP that everyone loves a lot. I have several voip lines myself (I build phone networks, remember, so it's not like I pay for them or anything, and I need them to test our networks if there's a problem after hours). The thing is, the incumbent carriers who run the directories were, and still are, slow on the uptake with Voice over IP. At least for a while, and even still to this day, many people opt to shut off their old lines and get new numbers with the voip service - not only because the voip carriers often do a poor job coordinating number porting, but because we're into instant gratification, and porting takes at least a week even if you do it right. These new numbers, are, you guessed it, taken from competitive carrier 10k blocks, and directory listings are never entered. (to enter directory listings takes more manpower at the VoIP carrier, so many don't even offer the option, and consumers generally don't mind).
So directory listing data is becoming less relevant because many people are switching to competitive carriers or cell phones and opting to remove their data from the directories (in the case of cell phones, directory listings are removed under an industry agreement that keeps cell numbers unlisted as a whole).
So it would seem that the sane idea would be to just call numbers randomly, right? But we run again into the issue of dialing hundreds and maybe thousands of disconnected numbers, lots of new numbers being assigned out of competitive carrier and cell phone blocks, and lots of incumbent carrier numbers being disconnected or ported to other carriers.
But what are the alternatives these days?
To recap:
* Privacy law precludes phone companies providing a list of active numbers
* Numbers can be assigned out of pretty much an option of tons of number blocks, and it's impossible, without a connection to the live industry database, to know for sure if you're calling a landline number that's been ported to a cell phone for example, and realtime access to that database is expensive, and there's no way to buy snapshots, again, due to federal privacy laws regarding retention of "customer proprietary network information"
* Directory data is still important to determine residential versus business - wasting your time randomly calling switchboards at businesses gets you nowhere too.
* Even if you look at disappearances from the directory, it's hard to figure out why that might be that they went away, especially as economic problems come to a head and people shut off their phones because they're running out of places to trim the budget. Did they turn it off (or get shut off due to nonpayment)? Switch it to VoIP or a cell phone? Who knows! With number portability (and intermodal number portability, the ability to port your landline over to a cell, and vice versa), you have no insight into what's causing that listing to go away anymore.
There's an even bigger problem too - most new technologies and services come with free caller ID. There's a cost to providing the caller id with name service, but it's fractions of a cent per call, so many places bundle it for free, and caller ID with number only costs absolutely nothing so every cell phone has it. So you get a lot more people who have caller ID. As the economy declines, these people who have caller ID, and have given their number to companies (companies selling customer data being a good way to get numbers without having to use directory data), are getting calls from collections companies, and nonprofits asking for donations. Another number comes up unknown, private, or with some number they don't recognise. Do they answer it? Heck no.
Unfortunately, the same people being inundated with calls and having caller ID are probably more opinionated than most about political issues, as they're actively suffering. We'll definitely see them at the polls unless their home gets foreclosed on, of course but they're unlikely to answer calls from political pollsters, or even calls from political campaigns.
Even people who aren't in bad financial straits are likely to screen their calls, either through caller ID, an anwering machine, or both, making them more likely to ignore political calls, calls asking for donations, and pollsters. They're busy handling their familial obligations after a long day at their job (or possibly jobs) and the last thing they want to do is talk to a stranger about politics when there's dinner that needs to be put on the table, clothes that need washing, and all those good season premiers won't watch themselves, after all.
So we went from being able to take our feet off the streets to be able to bring our message and polls to the people using technology, to people using further advances in technology to avoid us. The only way to respond now is to get back on the streets and do this the old fashioned way. Sure, it's easy to ignore the door, but it could be someone important! In our case it is, but not the kind of important they had in mind when they came running. :)
If you need further evidence of this, just look at the difference in accuracy between phone polling, and exit polls. It's more than just people having made up their minds solidly before going, although that's obviously part of it, and it's not just that these are the people that actually voted, although that is also a contributing factor. It's that you're getting a good sample population of real people, who aren't trying to avoid you using technology.
If I can get away from building phone networks long enough to spend some time on it, I'll see you in the neighborhoods!