It's time for a reality check ...

Most surveys of public opinion are crap
There are two key elements to surveys – the way you derive a representative sample, and the precise way you frame and ask the question.

However, anyone who has taken a survey research course knows that while there are statistically accurate ways to draw a representative sample, how you ask a question largely determines the answer you’ll get. 

So by stating your question in a certain way, you can obtain almost any answer you want, even within a statistically accurate representative sample. 

If you’re asked if you believe the rich should pay more in taxes than you do, most of us would probably say yes.  If you’re asked whether the tax rate on the rich should be significantly higher than the tax rate most Americans pay, yes votes would drop substantially. 

But if you’re asked if the rich should be penalized with higher taxes for being more successful, very few yes votes would be recorded. 

As another example, if you were asked if people currently not paying any Federal income tax should be compelled now to pay their fair share, most would agree.

However, if you clarified exactly who is not paying Federal income tax – the poor, as well as a large percentage of ordinary taxpayers and businesses through legitimate tax credits and deductions they’ve earned – the response would be quite different. 

So whenever you see survey results that make no sense, see the question.  Chances are it’s heavily weighted toward a desired outcome.  And it’s probably crap.  


Look at the sampling, too.  Drill down to the actual percentages of one group versus another.  Age, education, household income, marital status, etc.  A lot of "independent" pollsters manipulate sampling to get a certain result. So liberals over-sample demographics they think will tend to agree with them in their polls and conservatives do the same.  Almost nobody draws a true representative sample of the population, if they have a predetermined result they want.  


It's a rigged game.  And produces more crap results.  

Also beware of the results from any self-selecting audiences – such as online polls – because:  1) the people who volunteer to take surveys are not usually representative of the norm; and 2) the technology or medium used to acquire and interview respondents may substantially distort the results. 

The greatest historical example of the latter is the Dewey/Truman election: pollsters only surveyed people with telephones, which was not truly representative of the general voting public and especially not true of Truman supporters.  While they predicted a solid win for Dewey based on this bad sampling, Truman actually won.  


Telephone surveys today are probably less representative of public opinion than ever before for a couple of reasons.  


Do you really answer those calls?  Most normal people don't want to be bothered.  So who actually answers those calls and spends 10 minutes or more on the phone with some canvasser?  Do you think they are representative of the general public? They're probably the same people who look forward to getting called for jury duty, look forward to a chat with a telemarketer at dinner time, and have 23 cats and bundles of old newspapers around the house.  


Face it:  they are not you. 


And with the number of people who no longer have land lines and now only have smartphones that limit -- by law -- telemarketer access, the "representative" nature of samples drawn from conventional land-line telephone canvassing is getting weaker every day.  In time, you might as well survey people who heat their homes with coal.   


So beware.  There are lies, damned lies, and public opinion polls, to paraphrase Mark Twain.