A few corrections & clarifications to Mr. Scott Mueller's write-up are in order.

First, the effect of the arm length on overall rate is not linear. It's partly an [arm length]^2 function (as lever arm length affects the torsion in the lateral center section) and partly an
[arm length]^3 function as regards bending of the arm itself.

Next, there are a few terms besides the arm bending that have been overlooked. Mostly, they are small enough to be neglected, though bending in the central torsional section does indeed occur (particularly with relatively thin front bars; in such instances this effect can reduce bar stiffness by 5% or more).

Once you start playing around with the length of the central torsional section, the other terms become more significant, although that effect is likely dwarfed by the change in control arm attachment (and the associated reduction in motion ratio).

This would make a rather neat problem for engineering students to derive a solution for by working through standard beam formulas.

Norm