I'll go into a little overkill. Let's consider a circular piston of radius z, mounted in an infinite rigid baffle. The noise radiated by this oscillating piston can be modeled in terms of numerous monopoles radiating together. Each monopole is radiating from a rigid, fully-reflecting plane, not from free space. Therefore, the sound pressure due to anyone of the baffled monopole is twice that of an equivalent monopole in free space[30].
<br />
p'(r,t) = \frac{ik\rho_0c}{2\pi r}Q_p e^{i(\omega t - kr)}
In this equation, Q represents the source strength of the monopole on the surface and is equal to U_p \delta S where U is the peak surface velocity of the monopole and \delta S is an elemental surface area. We can then integrate over the whole surface to get the resultant pressure fluctuation due to all the monopoles vibrating in phase.
<br />
p'(r,\theta,t) = \frac{ik\rho_0 c \pi z^2 U_p e^{i(\omega t - kr)}}{2\pi r}\left[\frac{2J_1(kz\sin\theta)}{kz\sin\theta}\right]
So, basically I'm just putting equations to words. The Bessel term on the right is a directivity term and puts in those nasty "lobes" that make certain seats at the opera house the "ideal" place to sit.
So, lower frequencies give us lower reduced frequencies, k. This gives us two variables to play with, U (the maximum velocity) and z (the piston, {e.g. speaker} diamter). In order to make the low frequency noise at a level that is equal to high frequency, one of these needs to go up. Increasing the velocity of the speaker can potentially cause damage, while increasing the piston diameter seems simple. In addition to this, because its exponential, we can get a greater effect simply by increasing the size.