# Three percentage indicators into one (or two measures) in regression moder

I am working with a data set resembling the extract below:

MeasuerA MeasureB MeasureC Cost
10       89       1        100
5        83       2        120
30       69       1        100
20       79       1        90
25       74       1        95
15       83       2        100
5        94       1        110
4        95       1        10
10       88       2        110
6        93       1        85
7        92       1        100


The nature of the data set is that the two measures (MeasureA and MeasureB) are reversly proportional with some "noise" being introduced by the MeasureC. I'm interesting in doing a regression analysis answering on the relationship of the measures to the Cost variable. Naturally, when introduced to the model the measures A and B will display a high degree of multicollinearity. In effect, I can run the model with only MeasureA or MeasureB; however, I would like to make statements about both of the measures as impacting (or not) the Cost variable. In particular, I would like to be able to state that: "This change in MeasureA and related change in MeasureB results in Cost being higher/lower" Presently, I don't have access to any other variables than those described in the data extract. My question is how can I work around the multicollinearity problem? Would it be sensible to attempt to transform the measures into one "index-like" indicator reflecting the proportionality of distribution between those two variables and then introducing the generated index variable to the model? I presume that I can always run the model with only one of the measures (A or B) but I won't be able to make the statment that change in MeasureA and corresponding change in MeasureB result in this change to cost as the existence of MeasureC suggests that MeasureB is not simply (100% - MeasureA) as the MeasureC impacts the distribution in effect: $$100\% = MeasureA + MeasureB + MeasureC$$