If sending them somewhere to be measured is too costly, you can so your own matching. It won't be as close as a match made in a proper facility, and you won't end up with an acurate response graph for each mic, but all it takes is your time, careful methodology and gear you probably already have.
The simplest type of matching compares the mic output levels for a given input. Play a 1Khz sine though a speaker, record that output though each mic one at a time- be very careful to position each mic in exacly the same location and orientation when under test and also careful to not change the playback or recording levels. A good check to make sure you have things arranged close enough is to re-run the test group a second time, checking to make sure that the results for each individual mic are the same. Then analyze the recordings and check the RMS levels for each file. Pair up the mics with output levels that are closest to each other.
Matching for spectum is probably more important than matching for level though, because level differences are easy to adjust during or after recording. To do so simply, set up the same test situation as above and record some pink noise instead of a sine signal (you might run both these tests one after the other to keep from having to move mics and set things up twice). Listen to the recordings and choose mic pairs from those where the recorded pink noise sounds closest in timbre. In this case you are making a subjective listening comparison between mics.
You can make your timbre matching more accurate by running the resulting pink noise recordings though an audio analyzer plugin which averages the output of the file over the lenght of the file and displays an average reponse curve. Save each curve (or even an image of the curve) and compare the responses. You can do it the old fashioned paper way and print them out, line them up and hold them to the light like John mentions if you like. The resulting curves will not be very accurate representations of the actual response of the microphone like as you would get from a professional response measurement using calibrated test equipment, but that's not important here. What is important isn't the shape of the curve from each mic, but the differences between the curves of the mics. Pair the mics with curves that match most closely.