What most people miss on Y2K is that given the constraints at the time most of those programs were written, the cost of avoiding the problem would have been roughly triple (in real terms, discounted dollars, etc.) as ultimately fixing it was.
Yes, economists went back and actually measured, estimated, etc. Also, most of the code from the era did not, in fact, make it to Y2K, and would have incurred the costs without receiving a benefit.
Although the resources would be negligible today, we are talking about the extra labor to punch two more digits on each and every 72 column card, the limited number of digits on those punch cards, memory for which it was a break thought when the *rental* price dropped down to a buck a byte a month, the extra processing time to handle three or four digits, etc.
It also didn't help that so many (a strong majority?) coded in the year ending in 00 exception to leap years without including the divisible by 400 exception to the exception . . . other attempts to accommodate 2000 might have had similar blunders.