I remember, many many years ago, being surprised to learn that the credit rating of a mortgage borrower was as important to underwriters as it was; the loan, after all, is secured by an asset worth 25% more than the loan, and it seemed to me that the focus would be on verifying the value of the asset, conditional on which it wouldn't really matter all that much if the borrower defaulted. The problem is, even in what I'm going to date myself by calling "normal times", seizing collateral and selling it is a real nuisance, and is not what banks specialize in; they welcome the backstop, sure, but not least because it strengthens the borrower's interest in repaying the loan. The bank really just wants a borrower who's going to repay the debt.
Matt Levine recently commented that the very essence of collateralized lending is lending against illiquid assets — viz. assets that would be annoying to repossess and sell; for liquid assets, the owner can raise funds by selling the asset. This comment is more surprising to me than it is wrong, which is not to say that it isn't at least frequently somewhat wrong, starting with the housing market, where the whole point of your garden-variety mortgage is that you want to own the asset, and not that you have some other need for funds for which you might prefer to temporarily liquidate it. What he had in mind is the banking industry; both banks with abstract assets and other large companies with physical capital frequently use somewhat illiquid assets as collateral where, in an idealized world of perfect markets and divisible assets they might sell off a third of a factory and buy it back when they no longer needed the cash. Some of the models I've played with emphasize the extent to which an asset's value increases because it can be sold — because of its market liquidity — but notwithstanding some literature on the extent to which assets gain value because they can be used as collateral, I maybe haven't appreciated enough the extent to which the ability to sell an object and the ability to use it as collateral can substitute for each other.
 This is before it wasn't, before it now is again.
 Again, back when people put down 20%, at least for the mortgage loans under discussion here.
 In fact, the most frequent use of "collateralized borrowing" in finance actually tends to use very liquid collateral, but are not in fact legally "collateralized borrowing" at all; they are the sale of an asset and simultaneous agreement to buy it back shortly thereafter.
 The clear example is Kiyotaki and Moore (1997): "Credit Cycles," Journal of Political Economy, 105(2): 211--48, though Geanakoplos has a whole oeuvre that hovers closely around this, and Gary Gorton is probably worth mentioning, too.