What it reflects is something that we mentioned earlier, which was us improving our prepayment controls as we went along, because with these programs one didn't know at the time how long they were going to last or what subsequent programs would be there. As they stayed on for a while, we developed additional tools to be able to have better prepayment control.
We learned. As Jean-François said, we were flying the plane as we were fixing it. This table is actually a reflection of that. As you see, at the very beginning there was more going out that wasn't getting caught by prepayment controls. As time went by, they were becoming more and more effective. As you said, 17.9% went down to 0.5% by the time we got to the lockdown benefit. That's really what was driving that.
There was an element, too, of additional data that came in that allowed us to do our work better. I think we talked earlier about our getting the income tax filings from 2019. It really is just a reflection of our gaining experience with the programs and putting in place some better controls to have less money go out the door before we were comfortable with it.