More often than not data will contain no outliers. Something measured or recorded incorrectly can lead to an outlier. If you can find an explanation for a value being an outlier you would exclude it from the analysis. For example, if you were looking at annual factory production and found an unusually low value for one year you would take a closer look at where the data came from. If, say, there had been a fire that year it would be reasonable to exclude the value from the analysis. You should not exclude values if there is no explanation for them being unusual.