Basically yes, but it's suitability depends very much on the intended frequency range and measurement accuracy. Furthermore, a differential port has more network parameters than a single ended one, if you also consider common mode properties. If you have an ideal balun (actually it doesn't exist of course), common mode parameters would be ignored, with a real balun, it's limited symmetry and unknown common mode parameters of the differential port can cause large errors.
Another solution is to treat the differential port as two single ended ports and calculate all parameters from a two-port measurement. It works well in principle, but the accuracy may be possibly worse than with a balun measurement.