I was reading an article regarding Time Series Analysis and came across suggestion that it is not wise to use MEAN of salaries to forecast future salaries if TREND is present (Probably because salary is bound to grow in future).
But my assumption is MEAN too would increase as increase in salary. So it should be alright to use MEAN as forecaster.
The best example I have read on using the mean is for real estate. In my area if I look at the prices of the last 20 homes sold I might see 18 which sold for $150,000 with one that sold for 2.3 million and one which sold for $75,000. If I take the average I will get a larger number due to one large sale. If I use the mean I get a result which is closer to the true price of homes being sold, similar to Olympic scoring where the lowest and highest scores are discarded. If I apply this to a data set of salaries, I have to expect the CEO salary would have a similar impact. In short I think using the mean is a better method.