How to Know What They Think: Win with a Robo Poll

Ever watch that Discovery show “Gold Rush“?  It reminds me of  conducting a campaign poll.

I know this sounds silly.  And frankly I am a little reluctant to admit that I actually enjoy the show.

The exciting thing about the show is that people win (they strike gold) or they loose (they don’t strike gold).  It is a lot like running for political office.

But something that has caught my eye on the show is that the real successful miners pre-drill their mining sites.

That is, the smart miners test for gold before they start spending tens of thousands of dollars on mining.

Political campaigns are very similar.

Winning campaigns do polling.

Winning campaigns test to see where the voters are on candidates and issues before they spend hundreds and thousands on voter contact.

Like the gold miners , you want to know where the gold is before they spend money mining.

A campaign needs to know where the votes are before spending money on voter outreach.

It is vital to your campaign’s success to know if voters even care about your message or the platform you are running on.

There is a couple ways to do this.

By far, the best type of polling is probably with live calls, where an interviewer actually talks directly to a voter and asks a variety of questions.

Live calls can get expensive.  A good live poll can run well over fifteen thousand dollars or more depending on the number of questions you ask.

Many campaigns have started using online polling.

But online polling presents a host of challenges.

Because not everyone has access to the internet, it is difficult to target enough emails and capture enough responses in a targeted district.

With the biggest users of the Internet being those between the ages of 18-29 but making up only 13 percent of voters, any online poll is demographically skewed to non-voters.

This brings us to IVR or Automated or most commonly known simply as robo polling or a robo-poll.

Robo polling is the use of an  automated telephone surveys in which participants are directed by recorded instructions to press numbers on their telephone keypads in response to polling questions.

Robo polling has many critics for sure.

With robo polling you don’t really know who you are talking to. People get bored and drop off the call.

And you can’t call cell phones so respondents are going to most often be older voters.

Additionally, the FCC recently passed a rule allowing phone companies to make available to users a technology that allows for the blocking of all robo calls.

Use of the technology will be voluntary and the the polling industry is trying to win exemptions, but who knows where this ultimately ends.

The FCC ruling has the potential of shutting down robo polling.

With polling it is too expensive to call everyone you want to test as a representative sample of the overall universe.

This is where a generally understanding of margin of error is important.

Margin of error provides us a likelihood that a sample result is similar to what would be found if a whole population was polled.

To determine a margin of error simple visit this link and it will be calculated for you.

If you are really old school and like to do things your self there are statistical formulas that may be employed.

For more about margin of error visit Wikipedia,

A magrin of error of around 5 (95 percent accuracy)  is generally considered solid.

Straight out of Compton

Not calling cell phones will reduce your sampling universe and potentially increase your error rate.

It is illegal to make automated calls to cell phones.  This can be a problem with robo polling.

Even though we generally find about a 20 to 30 percent participation rate, only calling land line phones makes sampling smaller universes challenging.

Recently we conducted a robo poll in the southern California city of Compton.

We started with a universe of 5,496 voters.  Keep in mind this was total voters not a targeted likely voter universe.

After withdrawing all cell phones, we called a universe of 2,253 households and had 469 voters participate in the poll.  This gave us a margin of error of  4.03 – not bad.

But the problem is that they were not likely voters.

Had we gone with a likely voter universe, our initial universe would have been around 1,100 people.

After taking out cell phones we may have only had about 300 numbers to call.

With a sampling universe of 300 voters we would have needed to obtain a voter participation rate of over 50 percent.

Additionally,  Compton is unusual.  There are actually a lot of older voters in Compton, in fact most voters are well over 50 years or age.

In Compton there are a lot of homeowners and many tend to still have land lines.

With all the advantage in Compton we still could only poll the total universe of registered voters. It is just too small a universe.

When polling in more transient communities, where you have a lot of renters,  and a younger population, it can  be very hard to get that type of participation or even find many land lines (remember you can’t call cell phones).

If you are trying to sample a very specific group of voters such as likely voters, men, women, by party or a particular ethnic group, you can run in to real problems reaching an adequate sampling size.

Here is where robo polling seems to work well.

A large voter universe makes robo polling easier to pull off.

Last year I was the lead consultant on a Los Angeles Community College Board Race.

With 2,478,158 voters, the Los Angeles Community College District is huge.

However, these elections are held in odd years when turnout tends to be low. In 2015 only 9 percent – less then 250,000 voters cast a ballot.

The majority of voters was over 65 years of age or older, were homeowners, and there were lots of landlines.

This is prime real estate for a robo poll.

We developed a likely universe of 37,212 voters with land lines.  These were very high propensity voters.

We also had to be concerned with our budget so we narrowed the sampling size down as far as reasonable.

We asked 9 questions.

My client was an incumbent board member who was being challenged by a well funded opponent who came into the race with the ballot title of “Educator”.

“Educator” is a great ballot title as opposed to “Community College Board Member”.

For sure there has been problems with community colleges in California and Los Angeles County. Classes have been cut, facilities are old, and tuition keeps going up.

Though my guy had done a good job and was President of the board, we were a bit concerned.

So we did a poll.

Our initial question: ” Are you voting?” had an affirmative response of 78 percent.  This was followed by” “Who would you vote for?”.

This second question put my guy behind by about one point.  Not bad but not what an incumbent wants to see.

The following questions were issue questions regarding every thing my guy had worked on and would continue to work on if re-elected.

We polled: Bringing back classes — make colleges safer — keep community college affordable — free college —  would you vote for a bond — my candidate’s led effort to establish a 10 percent budget reserve.

Voters were also asked if they cared that he was President of the Board – they did not.

However, all the issues he was good on were highly valued by the voters.

We cross tabbed (used a contingency table) the results and were able to determine very specifically what issues mattered to what voters (Contingency table defined here).

Our strategy was to pound these issues home as past accomplishments and as future goals to the voters who cared about them.

It cost a few dollars in mail and other outreach efforts but we got the job done and won with 62.2 percent of the vote.

We earned 120,322 votes compared to our opponents 71,926, or 37.4 percent of the vote.

Had we not done polling and cross tabulations, we would not have been sure what to talk about with the voters.

Without polling and cross tabs we might have been talking about the wrong thing to the wrong people.

Who knows what the election result would have been?

With our polling, just like the miners on Discovery’s “Gold Rush,” we  hit gold.

Polling is essential to building and targeting your message to voters

Traditional polling is great but can be expensive.  Not all campaigns can afford it.

A robo poll can cost a lot less.

Robo polls can run $1,500, to $3,000, sometimes more,  depending on the number of questions you ask.

Weather or not you use a live poll vs a robo poll largely depends on your budget.

But if you are spending $20,000 on mail and it goes to the wrong people with the wrong message, it is kind of like throwing money away.

Spend an extra $10,000, or at least $2,000 on a poll, and be like a gold miner, and “pre-drill” that electorate so you have some idea where the gold is and win more votes.