Research shows typing typically is faster.
Reach for the mouse: .4s
Move the mouse to target: 1.1s
Click: .2s per (ie. a double click is .4s)
Move hand back to keyboard: .4s
Keeping the hand on the mouse saves 1s per operation, which isn't common for data entry apps.
Moderate typists do about .2s per keystroke (control keys are counted as separate keystrokes). Fast typists do about .08s per keystroke. So if the choice is between typing "emacs" and clicking on the emacs icon, the comparison will be:
Mouse: .4 + 1.1 + .2 + .4 == 2.1s\nKeyboard (moderate): .2 * 5 == 1.0s\nKeyboard (fast): .08 * 5 == 0.4s
These are numbers from research done for the GOMS method of UI analysis, which has been extensively studied (and uses timings a bit more accurate than a stopwatch).
Note that I have also not included the congnitive timings; typically you put a modifier where the user has to stop and think about what to do next (like "which icon should I click"). These timings are subjective and can vary widely based on familiarity with the UI. Typical cognitive timings are on the order of 1.35s.
Bottom line: I'm wary of anyone's "stopwatch numbers" that purport to prove that a mouse is faster than the keyboard, because such a result goes against a lot of verified research. Stopwatch numbers also have situational bias: "OK, I'm going to time you, so click on that icon as fast as you can! Ready, set, go..."