From Stone Street Advisors' take on Paulson's testimony to the Financial Crisis Inquiry Commission:

Their research approach was pretty straight-forward: Focus on subprime, where they were amazed at how low quality the underwriting was, and how low the credit characteristics were on the loans. They found the average FICO was around 630, and over half of the loans were for cash-out refi?s, which were based on appraised, not sales prices (so ?value? could be manipulated). For many of these loans, LTV was very, very high, 80, 90, 100% with many of them concentrated in California (no surprise there). Close to have of the mortgages they looked at were of the stated-income, no-doc variety.

Those who did report incomes had D/I ratios of > 40% before taxes and insurance. 80% of them were ARMs, so-called 2/28?s with teaser rates around 6-7% for those first 2 years, but after they reset, the rates were L+ 600bps which at the point would have doubled the interest rate on these loans, and Paulson & Co thought there was very little ? if any ? chance borrowers would be able to afford the higher payments.

Once the rates reset, the only thing these borrowers could do would be to sell, refinance, or default. These were people spending > 40% of their gross income on their mortgages already, once the rate jumped up after the teaser period, they expected that many borrowers would simply default, and the price of the RMBS into which these loans were securitized would fall drastically, while the price of the protection (CDS, etc) Paulson bought on them would skyrocket.
...

Paulson also made a distinction missed by many if not most: Everyone was looking at nominal home price appreciation, but real appreciation numbers were much different. Going back 25+ years using real growth rates, they found that prices had never appreciated nearly as quickly as they had from 2000-2005, and that this trend was unlikely to continue for much longer, i.e. there would be a correction and then mean reversion. Their thought was that once this correction came about, because of the poor mortgage quality and questionable assumptions/structures in mortgage securities, losses would be much worse than estimated.

Does research help you build a better model of reality? Paulson certainly thinks so.