See Poynton. It's been discussed on the forums before in further detail, but just a quick excerpt:
As I explained in Rendering intent, on page 81, it is important for perceptual reasons to alter the tone scale of an image reproduced at a luminance substantially lower than that of the original scene, reproduced with limited contrast ratio, or viewed in a dim surround. The dim surround condition is characteristic of television viewing. In video, the alteration is accomplished at the camera by slightly undercompensating the actual power function of the CRT, to obtain an end-to-end power function whose exponent is about 1.25, as indicated in Equation 23.1 in the margin. This achieves end-to-end reproduction that is subjectively correct (though not mathematically linear).
YsubE ~ 0.5; YsubD ~ 2.5; YsubE * YsubD ~1.25
Rec. 709 encoding assumes that encoded R'G'B signals will be converted to tristimulus values at a CRT (or some other display device) with a 2.5 power function (YsubD ~ 2.5).
The product of the effective 0.5 exponent at the camera and the 2.5 exponent at the display produces an end-to-end power of about 1.25, suitable for typical television display environment...