The problem here is power. Look at how it works if you talk about voltage instead of power:
Q: How is the voltage distributed when two speakers are hooked up in parallel?
A: Both speakers get the same voltage.
Historically, engineers cared about power because we're interested in the thermodynamic efficiency of systems. Also, power is a crude way to compare systems that have different impedances. Today, power is a nice thing to put on a spec sheet, but a poor quantity to use when trying to understand what's going on with amps and speakers. Most basic testing is done with voltage measurements.
Amps and speakers are voltage devices. What's this mean? An ideal amp is designed to reproduce the input voltage at a higher amplitude. Its output voltage is proportional to its input voltage, over a wide range of output current levels. Mainstream speakers ideally produce roughly constant SPL within their passband when fed with signals of equal voltage amplitude. Frequency response and sensitivity curves assume a constant voltage input signal, not constant current or constant power.
Thus it's the nature of speakers -- as voltage devices -- that makes it useful to design amplifiers as voltage devices as well.
Now, hooking up those two speakers: The one with lower impedance (at a given frequency) will receive more current, and thus draw more power at that frequency.