By design, frontal New Car Assessment Program (NCAP) tests focus on a narrow portion of the spectrum of field crash events. A simple, high level parsing of towaway crashes from NHTSA's National Automotive Sampling System - Crashworthiness Data System (NASS-CDS) files shows that only a small fraction of occupants (but a somewhat larger portion of their harm as measured by ISS) find themselves in crash circumstances remotely similar to NCAP crash conditions. Looking only at seat location, area of damage, direction of force, distribution of damage, and estimated delta-V filters significantly restricts the relevance of NCAP even before critical factors like belt use and vehicle crash partner are considered. Given the limited scope of frontal NCAP, it should not be surprising that it has limited usefulness in discriminating among various vehicles' overall performance in the field.
Abstract