The validity of an experiment - and likewise a poll or survey - depends both on its structure, and on the analysis of its results.  We all know that numbers and statistics can be manipulated, even if they were obtained in a scientifically robust manner - which these political tools usually are not, in any case.  Or perhaps the numbers are analysed properly, but the structure of the poll/survey is not robust: Maybe the sample population used was not adequately large, or not proportionally representative of the whole voter population, or unplanned variables influenced the performance of the one being tested, and so on. 

There are many polls and surveys (for brevity I will use the word ‘polls' hereafter to mean both) that are deliberately skewed for propaganda purposes, psychological warfare, if you will:  Boost up your voters that might not want to back a losing cause, and discourage those who would vote for your opponent by making them feel they are wasting their vote.  The discrepancy between polls and real-life results are so large, so frequently, that one wonders why reputable media would even bother publishing them, since they reflect political agendas far more than they do objective reality.

So, have you guessed the answer?  Social engineering.  Polls are used politically not to reflect public opinion, but to form it.  Who is doing/commissioning the poll, regarding both bias (intended or not) and professional expertise, is important.  Often politicians want the genuine data for themselves and employ the professional means to obtain it, but prefer to publicise material more favourable to their campaign. The law of the jungle reigns in the realm of these political polls.  According to experts, however, even those polls done with objective and honest intent are not always scientifically robust enough to be reliable. 

To give the simplest example, the polls usually use landlines.  This means that already the sectors of the population least likely to have landlines such as young adults, soldiers, and hotel workers, are excluded from the poll.  Some of the polls make calls during the morning hours.  This means that more housewives or affluent people who don't need to work will be caught in the ‘net' than people who are out at work during those hours.  The calls are in Hebrew - which means that non-Hebrew speakers will not wish or be able to participate. 

The pollsters are highly unlikely to have studied the demographics of Eilat and made sure their sample populations reflected all the subpopulations proportionately.  If a conscientious pollster did make sure that his samples include exactly the right percent of, say, Russians or Ethiopians, soldiers or pensioners, affluent or poor, and so on, would he take the added step of making sure he made as many calls to ‘extra' ones as necessary to preserve the proportion against those who don't answer their phones or participate?  

Then there's what happens during the questioning.  If the call starts out "This is so-and-so from Candidate X's campaign", there are those who will say that "of course" they're voting for Candidate X, just to get off the line and avoid the hassle of being campaigned at.  Then you get the ‘box' questions, where the formulation of none of the questions is what you'd actually choose, but you must select the closest to what fits even though it's not accurate.  There are also the ploys of what is included in a question, how it is phrased, what order names are listed in, even tone of voice, and various other little psychological tricks that can subconsciously influence the respondent. 

Also look out for contradictions, such as a poll claiming that Candidate Y is in only 3^rd or even 4^th place, while at the same time using a question pitting only Y (not the others) against Candidate X in a runoff election to ask which of the two you would then choose. 

Another tactic is diversions, distracting your attention to tangential issues such as why Candidate Y doesn't refute Candidate X's polls or other propaganda.  (Implication being "because he can't, they're true.")  But maybe Y prefers to spend his energies building on his own merits rather than wasting them trying to disprove a constant stream of attacks and misrepresentations from his opposition.  Or maybe he doesn't want to play the game by the opposition's rules, or lower himself to their tactics, or maybe he feels his case is strong enough without it, or a myriad of other reasons.

Again, the law of the jungle at play:  Anyone can say almost anything about someone else, and the work of disproving it or going through the courts for slander/libel is costly and time-consuming.  In the case of the upcoming elections, the verdict would come too late to help.  In the case of rumours and innuendo, it's almost impossible to disprove anyway, like the classic Israeli saying ‘prove you don't have a sister.'  It's easy to prove you do have one, but how do you prove that you don't?  Or one I heard from America:  "Do you beat your dog?"  (Well, of course the guy will say ‘no' whether he beats his dog or not).  Rumours are like silt stirred up to muddy the waters, obscuring the truth while being too disperse and fluid to catch up in one's hands and dispose of so that clarity is restored.

['Gossip' by Norman Rockwell] 

The final stage is analysis of the results.  The analysis usually does not include those who didn't answer the phone, those who did but declined to participate in the poll, and often those who answered ‘undecided', ‘don't know', or the like.  The poll might force an answer, along the lines of an insistent "but if you had to vote today, which candidate would you choose?", giving an unreliable result.  Or worse, if 25% - 30% of those who picked up the phone refuse to answer questions give  ‘undecided' or similar answers, pollsters usually omit them from the results.  This is a mistake, because it implies that their results would have been divided in the exact same proportions as those who did answer, which is almost never true.  Different kinds of statistical tests can be run, different sets of figures included in the testing, and more.

The bottom line is, as with all the other kinds of propaganda, don't be fooled.  Check sources, check professionalism, check objectivity, check the ‘materials and methods' of the poll as well as how the data was analysed - in short, be an aware voter and use your intelligence to outfox the foxes and arrive at an informed decision.  It is our lives - our incomes, our homes and neighbourhoods, our children's schools, our elderly's conditions, our earning and spending power, our irreplaceable natural resources, and more - that depends on how we vote.