The parameter carrying most of the weight here has a dull name and a long lever arm. Buyers, on average, pay less than a dollar today for a dollar's worth of future fuel savings [8]. The disagreement is over what that gap means. Read it as inattention and the buyer is leaving money on the table, so a standard that puts the efficient car on the lot leaves him better off. Read it as hidden costs and the same standard hands him cheaper fuel while taking away weight, performance or comfort he was also paying for [9]. Same car and same fuel savings, opposite sign in the welfare column. EPA took the first reading in 2024 and the second in 2026 [10][11], and the authors put the difference at hundreds of billions of dollars [12].
Their corrected split, drawn from a systematic review of the peer-reviewed literature, is 55 cents of inattention, 23 cents of genuine attribute trade-offs and 22 cents that buyers see and fully value [13]. So the missing portion is 78 cents, and inattention accounts for 55 of those 78, about 71 percent [17]. The 2024 analysis effectively used 100 percent; the 2026 analysis used zero. Neither matches the review, but one is far closer, which is what Arthur van Benthem of Wharton means in saying the 2024 analysis better reflects the academic evidence and is more internally consistent [14].
Internal consistency is the part that survives whatever you believe about consumer rationality. To price the technology, the 2026 analysis holds weight, power and comfort fixed and asks what fuel economy costs. To explain why buyers undervalue savings, it assumes those same attributes do move and that buyers absorb the loss [16]. Both propositions cannot be doing work in one document.
Two other inputs sit on the same ledger: avoided air pollution valued at zero, which the authors call obviously incorrect [7], and a gas price below what drivers pay at the pump [6]. Note also the shape of the published uncertainty. The rollback's benefit range runs from $600 billion to $790 billion, a spread of $190 billion [3][18], which is narrower than the swing the authors attribute to a single behavioral parameter [12].
What this analysis does not do is produce a rival headline number. It repairs parameters and reviews evidence rather than measuring new buyers [13], and the 23-cent trade-off share rests on recent EV studies [13], a narrower base than the parameter's leverage deserves. It is also an average across buyers, not a diagnosis of any one of them. Kenneth Gillingham's framing is the honest floor here: the swings between administrations are large enough that, if the arithmetic is being done correctly, one or both analyses must be wrong [15]. On this evidence I would read the $600 billion to $790 billion as a statement about the parameter set rather than about drivers.
