Please keep in mind that the salary is the purported average from the respondents of the survey. If the average is represented by the "arithmetic mean", then a high salary can sway the results and if only a few responses came from DC with a non-representative high salary for a small subset of the DC respondents, the average from DC would be skewed. The best representation for the salaries would probably have been the "median". The median would have reflected the midpoint salary for all the responses; both the regional and aggregate levels.