Several authors have suggested that the use of the mean absolute percentage
error (MAPE) as a measure of forecast accuracy should be avoided because t
hey argue it treats forecast errors above the actual observation differentl
y from those below this value. To counter this, the use of a symmetric (or
modified) MAPE has been proposed. This paper shows that, in its treatment o
f negative and positive errors, the proposed modification is far from symme
tric, particularly where these errors have large absolute values. It also s
hows that, under some circumstances, a non-monotonic relationship can occur
between the symmetric MAPE and the absolute forecast errors. (C) 1999 Else
vier Science B.V. All rights reserved.