I have taken a good deal of statistics courses and know more about statistics than most people who don't have a degree in statistics itself. Statistics is a very valuable tool for us to understand large dynamics that would otherwise be unknown to us, and is one of the prime sources of many of the scientific advances in recent times.
This being said, most of the statistics that get quoted in the news are complete and total crap. They are lies dressed up to look authorative, plain and simple. In order to actually get good statistics several things are necessary. The data gathering needs to be very unbiased, the data analyzers need to be unbiased, the analyzers need to have a very strong grounding in statistical methods, and the analyzers need to actually want to generate unbiased results. If any of this is missing then the statistics is probably worthless.
Take the average political political poll. These are almost always completely worthless at best, if not actively counter-productive. They are usually generated from appalingly small sample sets, on the order of hundreds of people, when any decent political poll would need tens of thounsands of respondants at least in order to be effective. The people conducting the polls generally have an extreme bias towards one of the outcomes, often directly tied to one of the candidates or positions. This makes them less than eager to find out that the public wants what they don't, and often willing to purposefully perform bad statistics: better to achieve their goals than be bothered with mere honesty. Even if they want to be unbiased they usually have a degree in political science or something similar and at best only had one course in statistics at college, which wasn't even calculus-based. Without a graduate-level knowledge of statistical methods most statistical studies involving human beings is out of the question. The result is that Party A's poll says their candidate is going to win by a landslide, and Party B's poll says exactly the opposite.
With that in mind, there are actually well-done statistical results out there, even dealing with people and their opinions. They are generated by people with Ph.D's in statistics, usually working for the Census Bureau or similar instutions, and are generally very boring and completely ignored by most of the rest of the world. But if the Census says that there is a net increase in the percentage of the population living in poverty in State C, then I will generally believe them. If doctors say that smoking cigarattes causes cancer even though they can't actually demonstrate the exact mechanism yet then it probably does.