By 2006, S&P was making its own study of such loans' performance. It singled out 639,981 loans made in 2002 to see if its benign assumptions had held up. They hadn't. Loans with piggybacks were 43% more likely to default than other loans, S&P found. In April 2006, S&P said it would raise by July the amount of collateral underwriters must include in many new mortgage portfolios. For instance, S&P could require that mortgage pools have extra loans in them, since it now expected a larger number to go bad. Still, S&P didn't lower its ratings on existing securities, saying it had to further monitor the performance of loans backing them. It thus helped the market for these loans hold up through the end of 2006.
The Wall Street Journal Aug 2007