Statistics are bullskulls because if you asked the questions the right way, analyze the data the right way, and arrange the numbers the right way, you can make them say whatever you want them to say.

Example:

I ask you this, "Do you think the United States government spends too much or not enough money on illegal immigrants?" Most would answer, "Too much."

If I asked the same question with different wording, "Would you deny a child of an illegal immigrant healthcare or education?" Most would answer, "No."

The same person just gave me two conflicting answers to the same question. So, I get some preconceived notion as to how I want the poll to look and ask the questions accordingly.

After that, I can compile the data to manipulate the statistics even more. Unless you came up with the data and compiled it yourself, you really can't know whether or not to believe the statistics. And 86% of people agree with me, according to a poll I just made up for this joke.