High Dynamic Range (HDR) Imaging is all about dynamic range. That is a way of describing the difference between the darkest and brightest parts of a scene we wish to photograph. Each stop is double the light of the previous one. A scene with, for example, 6 stops dynamic range has highlights that are 64 times as bright as the shadows.
Any JPEG, regardless of how it was produced, is limited to about 6 stops of dynamic range. Each color channel uses 8-bits, but about two of those bits are taken up by what is known as the noise floor. Most current DSLRs and other advanced digital cameras record either 12 or 14 bit RAW files with 10 to 12 stops of dynamic range at their base ISO sensitivity. What HDR (High Dynamic Range) Imaging attempts to do is squeeze as much dynamic range as possible into the 8-bits for each color channel in the JPEG standard.
As we mentioned above, most advanced digital cameras can record 10-12 stops of dynamic range in a single exposure. To get the same amount of information in a series of JPEG files you would need one exposure three stops darker and one exposure three stops brighter than a 'correct' exposure. Since there would be very little overlap between the darkest parts of the bright exposure and the brightest parts of the dark exposure we would also need to add a third exposure in the middle between the other two. This is where the -3, 0, +3 series of bracketed exposures to use in HDR processing comes from. So in most cases a properly exposed single RAW file can contain as much information as a -3, 0, +3 series of JPEGs.
But some scenes contain even more than the 10-12 stops of dynamic range a single RAW file can record. If we shoot a series of RAW files we can expand the total amount of information from the darkest to the brightest parts of the scene even more. By shooting a -3, 0, +3 series of RAW files, we can add an additional 6 stops of dynamic range for a total of 16-18 stops! That means that the brightest exposure records highlights that are 262,144 times as bright as the shadows in the darkest exposure. The more intermediate steps of exposure we have, the better the software we use to combine the images can do so smoothly.