It doesn't matter whether you connect in series or parallel; that only affects drive impedance.
You won't get 3dB per driver, but at most 3dB per doubling. Four drivers would give 6dB (two lots of doubling). This assumes, of course, that adding a driver affects efficiency at all - I don't know as I am not a speaker expert.
Note that you have to compare like with like: sound power out for electrical power in.
There are two ways of looking at this, power or voltage.
The voltage method gives 6dB per doubling (in parallel) so 4 drivers give +12dB for the same input voltage. This will draw 4x the power (6dB) so the power method gives only +6dB for the same input power. Amplifiers are voltage sources to a large degree, and speakers are "voltage sensitive", so in my opinion, the voltage method is correct. Think of it this way: If you take a single driver and measure, then add three more in parallel without changing amplifier settings, the SPL will go up 12dB and the power will go up by 6dB.
There is an additional factor that is not as easy to calculate, and that is how the drivers couple across frequencies. The coupling is best at low frequencies and lessens slightly at higher frequencies. You also get frequency dependent directivity effects at higher frequencies when dealing with an array.
Another thing: the effects of output impedance might start to become significant at really low impedances. IF you have a damping factor of 10, the output impedance is 0.8 ohms and the impedance of the 2 ohm speaker might be around 1.4 ohms above and below resonance, and perhaps 20 at resonance, so the amount of power "lost" in the output impedance becomes significant, and varies with frequency.