That might be true for a single pairwise preference, but not for filling out the whole ballot. Producing a full preference ordering for more than a few items is a lot more work.

Empirical studies have shown that data collection via rating scales usually takes only a third of the time required for a complete ranking of the value items (Munson / McIntyre 1979: 49).

subjects take an average of one third as long to complete the scaling procedure as the ranking procedure. This significant saving in respondent time and effort apparently is achieved at little or no cost in measurement accuracy (Mcintyre and Ryans, 1977).

Of course it also destroys any information about strength of preference, which makes it less likely that a voting method elects the most-representative candidate, as shown in the original article this thread branched from.