So now that I’ve pointed out who were the pollsters in question, you don’t care about that anymore? Ha ha ha. Next time do a little research before opening your trap, otherwise it makes you look like an idiot.

But to answer your question, yes they do teach statistics in high school (did you not go to high school in the US?); however, I was fortunate enough to learn statistics in college from one of the leading authorities in the field: professor Bob Rosenthal.

To start off, the term “target population” is only used when you’re talking about biased samples which are typically used when you want to target a specific population and there’s no possible way to define what their scores are in a random manner (for instance, sampling women who enjoy high-end clothing, thus sending out questionnaires to Vogue subscribers). As the name suggests, the probability of bias is quite high especially if the questions incite feelings of apathy, hubris, or disgust, as is the case with immigration. Therefore, this type of sampling is not used by the most trusted political pollsters.

In fact, political pollsters have been extra cautious about bias sampling ever since that very famous incident in the 30s where Reader Digest, I believe, was supposed to predict the results of the election. They sent out literally millions of questionnaires and concluded that Republicans would win in a landslide. As you’re probably aware, if you know your history, Franklin Roosevelt was actually the one who killed at the polls. (you can google it if you want)

When you want to produce a truly representative sample there are three things to keep in mind:

1) Explain what the scores are in the population (in other words, define it).
2)Identify every member of the population (although not necessarily sample it).
3)Choose scores so that every sample has the same probability of getting picked.

So when you deal with inferential statistics (as opposed to descriptive statistics) there are fortunately specific techniques to reduce the uncertainty of the results. For instance, research sampling is supposed to yield a statistic similar to the population parameter; therefore political pollsters use carefully selected nonrandom samples - which have reached accurate conclusions in the past - to account for the uncertainty that exists in all statistics. Otherwise the results are not published.

Another technique they use to reduce the uncertainty in research sampling is to use sampling distributions. In a nutshell, sampling distributions is always the distribution of a particular statistic. In other words, there’s a sampling distribution of the mean, a sampling distribution of the range, a sampling distribution of the variance, etc etc.

Think of it this way: there are many samples all coming from the same population (the US). The same statistic is calculated for each sample. Then all of the statistics are scattered into a frequency distribution and graphed as a frequency polygon. Next, the standard deviation and the mean are calculated. Depending on what the frequency polygon looks like (let’s assume it’s a normal curve, or bell curve, if you will) you can go ahead and calculate Z scores which allow you to figure out the probability of a particular statistical value.

But guess what? These highly skilled pollsters working for Pew et al already know about confidence intervals. And therefore, it is highly unlikely that EACH AND EVERY SINGLE ONE of them reached the same conclusion in error, particularly when the collective questioning is so thorough. This is why people familiar with statistics knew who was going to win the presidential race days before the actual election. Those who weren’t familiar kept fretting over the fact that Fox News was touting a story that McCain pollsters had him winning in Pennsylvania and New Hampshire.

So again my advise to you is to do some research first, before spouting a factually erroneous opinion.