The metrics used by a word
processor are *not* the autohinted metrics. It would be completely
crazy to do so since you get a dependency on the font size and
screen/printer resolution. Autohinted metrics are good for screen
menus and the like only. The same is true for the bytecode
interpreter, BTW.
In case the unhinted metrics for Times New Roman and Liberation Serif
are identical, they will stay identically on paper, regardless whether
you use the autohinter or the bytecode interpreter for rendering
glyphs on screen. In case you get different results, something is
severely broken in your word processing application.