I tried plotting the difference 1999-2013 by percentile, first in absolute terms (£ difference):
Which support that the wealthiest part of the population (upper middle class) benefited the most from the past 15 years.
However if you do the same in relative terms:
it is the lowest part of the income distribution who benefited the most, seeing their wages doubling. The very much upper end too. But not the middle class.
Depending on how you look at it, it can be used as an argument both ways. Which is why I hate when people reduce something as complex as a distribution into one number. It's too easy to lie with statistics.