In general, you don't - and it's rather meaningless to do so.
Where this most often comes up is in earnings. Take the following
examples:
Company A Company B Company C
Last Year -$0.01 $0.00 $0.01
This year $0.01 $0.02 $0.03
The way to calculate %change is usually given as
=(this year - last year)/(last year) * 100%
or
=(this year/last year - 1) * 100%
For each company, earnings increased by $0.02. but their %changes
are
Company A = (.01-(-.01))/(-.01)*100% = -200%
Company B = (.02-0)/0 * 100% = Undefined
Company C = (.03-.01)/.01 * 100% = +200%
The changes range from -200% to 200%, wildly fluctuating where the
base period is zero. Yet all had about the same increase in
"success" - a small profit or loss to a slight profit.
Also, consider Company D which went from +0.01 last year to -0.01
this year. The %Change is (-0.01-0.01)/0.01*100% = -200%, or exactly
the same percent change as Company A. Silly.
People for some reason still want to see %change on earnings - the
WSJ, however, only publishes %Change when both this period and last
period were profits. If either period had a loss (or zero, IIRC),
they list only P or L (corresponding to profit or loss for the most
recent period)
It's a bit different when values can only be positive (or negative)
- then %change gives a better indication of relative change.
