Since every sample taken from a population will contain different values, the values taken by an estimator will have a distribution of their own. An estimator is unbiased if its average value is equal to the actual value of the parameter it is estimating.