It Walks, It Talks: How Google Consumer Surveys may one day replace Public Opinon Polling as we know it
Public opinion pollsters must decide whether DIY Google Consumer Surveys (#gsurveys) represent an opportunity or a threat. Google seeks to disintermediate public opinion pollsters by telling those who want to ask questions that they can use Google Consumer Surveys to access directly those who want to give answers, bypassing the middle-man research companies. Do-It-Yourself #gsurveys appear to be positioned primarily at businesses who want to learn something about their customers, and/or their market.
But: consider these #gsurveys now in rotation, which seek to measure preference for Barack Obama vs Mitt Romney.
In example #1, on the left, any breathing human is asked whether he/she prefers Obama or Romney. In theory, this question has all kinds of limitations, including …
- You can answer the question if you live outside of the United States.
- You can answer the question if you are younger than 18.
- You can answer the question if you live inside the United States but are not registered to vote.
- You can answer the question if you live inside the United States, are registered to vote, but are not likely to vote in the November election.
- You are not offered a “undecided” choice.
But in practice, there is no telling how the answers to this ultimately simple #gsurveys question construction might turn out, and maybe, against all odds, the question will have some real-time value, or perhaps some predictive value.
A different approach appears in example #2 (above right), where a screening question is used to “get rid” of a) kids, b) non-registered voters and c) the 95% of the world’s population who live outside of the USA. From there, eligible voters are offered the exact same choice, using the same wording and same punctuation, which indicates to us that both of these #gsurveys were fielded by the same outfit, as a parallel test, to compare results from test A to results from test B.
What about Ron Paul, you ask? That, it turns out, is covered in parallel-test example #3, below:
The inclusion of “I am not a registered voter” in the above example #3 turns this leg of the test into something that is neither fish nor fowl, in our opinion, but we are fascinated to see the apparent rigor with which this experiment is being pursued.
In a separate post, we have wondered about how many of the clicks on Google Consumer Surveys are “junk clicks,” and unless Michael Bloomberg knows something that we don’t know, we think parallel test example #4, below, is designed to test for junk clicks. By that we mean: we are not aware of anyone seriously considering a 4th-party run for president. Pat Buchanan has a book to sell, Al Sharpton has a TV show and Ralph Nader’s moment has passed. Yet, we are offered “other” in the example #4 below:
Unfortunately, the parallel test jumps the shark, in our opinion, in example #5 below, which the Google-Matic answer-choice Randomizer has turned from meaningful into mush.
Had the parallel-test #5 been completed rewritten to ask:
In the 2012 presidential election, are you…?
- Leaning toward Mitt Romney
- Locked-in on Mitt Romney
- Leaning toward Barack Obama
- Locked-in on Barack Obama
… data from the #gsurveys question could have been salvaged. As written and presented, parallel-test #5 is a (however well intentioned) random-wreck.
And just when you think the parallel-test could not become any more over-thought or overwrought, comes up for consideration example #6:
Stepping way back:
Public opinion polling was …
- once conducted door-to-door using canvassers, who walked the blocks;
- then was conducted using U.S. mail;
- then was conducted using live operators on the telephone;
- then was conducted using recorded-voice.
Some public opinion polling has migrated from telephone to the internet. But none of the internet solutions available today are as disruptive as the Google Consumer Surveys (#gsurveys) solution.
When you see parallel tests such as these being vetted using Google Consumer Surveys, do you think we are getting our first glimpse at the future of public opinion polling? Are we looking at the replacement for The Gallup Poll? Or, are we looking at some Frankenstein that escaped from some a Mountain View laboratory?