As I see it, the main cause was systemic, namely the securitization of mortgages (e.g., CDOs). This means mortgages being repackaged by lenders and resold to other investors who wanted the income from the monthly payments. This was first done in 1970 by the Government National Mortgage Association (GNMA or Ginnie Mae), but securitization didn’t really start taking off until after the dot com bubble burst in 2000.
One big side-effect of securitization was that the parties supplying funds used to lend out as mortgages were no longer those that were actually directly interacting with the mortgage borrowers.
The graph below from Keys et al (2007) shows the number of loans given out by FICO score (a measure of a borrower’s creditworthiness). The vertical line shows the FICO score of 620, and loans to borrowers with a higher score than this were easier to resell (the less likely it is for a mortgage to be defaulted upon, the more attractive it is to a CDO investor). As you can see, there are significantly more loans granted right above that 620 threshold, and the gap increased over time.
This next graph from the same paper shows loan delinquencies in 2002 by FICO score. If FICO scores accurately reflected borrower risk as they are supposed to, there should be a continuous downward slope. Instead, borrowers right at 620 and above were more likely to default than borrowers with a slightly worse FICO score.
Other years show similar effects, albeit somewhat less dramatic than this. This is clear evidence that lenders were inflating the FICO scores for borrowers applying for loans that were intended be securitized and resold. Somehow, I feel like I’ve heard this story before.
As the authors put it:
[W]e find that there are more than twice as many loans securitized above the credit threshold at 620+ vs. below the threshold at 620−… [T]his result confirms that it is easier to securitize loans above the FICO threshold.
Strikingly, we find that while 620+ loans should be of slightly better credit quality than those at 620−, low documentation loans that are originated above the credit threshold tend to default… at a rate 20% higher than the mean default rate… As the only difference between the loans around the threshold is the increased ease of securitization, the greater default probability of loans above the credit threshold must be due to a reduction in screening by lenders.
Furthermore, Mian and Sufi (2008) found large increases in mortgage loans and decreases in denial rates for zip codes where there was a sharp increase in securitization even though those same zip codes “experienced significantly negative relative income and employment growth” in the same time period.
Before securitization, lenders had a strong incentive to care about default risk. Afterwards, they were passing this risk along to investors who were mostly large financial firms. You might think those institutions would have cared just as much about default risk, but there were a couple of reasons why they didn’t.
First of all, since most of the investors in securitized mortgages were financial institutions, this meant that the investment decision was made by an employee who was betting company money, not his/her own. That being shielded from risk encourages too much risk is a well-known problem known as moral hazard (and notice that this also applies to the originators being shielded from default risk). In the worst case, the employees risked losing their job, which seems significant until you compare it to the massive amounts of money they were investing for their employers (or to the tremendous bonuses they stood to gain if their bets paid off).
Secondly, these financial institutions play a key role in our economy in translating savings to investment, and their myriad interconnections with other financial firms made them too important for the government to let them fail. So, the management of the firm largely sanctioned having their trading desks place huge bets because, again, the potential payoff was huge, and the worst case was probably just getting a bailout from the government. Note that preventing both of these problems is the rationale behind the Volcker Rule that basically bans firms from playing the market.
The investors also were just not in as good a position to screen borrowers. They had to rely on whatever information the lender passed along, and some of the research I linked above showed that they passed bad information to make the bundles more attractive (e.g., inflating the borrowers FICO score). This is a form of asymmetrical information, specifically “the lemons problem” where the sellers know more about the deficiencies of a product being sold (e.g., a “lemon” in the used-car market) than the buyers.
These institutional investors also saw CDOs as low-risk bets because they were “insuring” them with credit-default swaps, but they underestimated the danger that the insurer itself might go out of business (counterparty risk). Plus the ratings agencies gave these securitized mortgage bundles much higher ratings than they should have due to lack of competition and conflicts of interest.
There are a lot more actors involved, like the ratings agencies and the Fed, but this was the core cause. Here are my recommendations for further reading. For a short read, I recommend this short primer by Comiskey and Madhogarhia. It was relatively early, before a lot of the more compelling empirical research regarding securitization came out (the first two of which I cited above), but it covers a lot of the bases. It also has good arguments against the talking point that the crisis was caused by the Community Reinvestment Act (CRA). Investopedia’s take also looks very promising, although I haven’t read all of it yet.
The one book I most highly recommend is All the Devils are Here by McLean and Nocera. I found it to be very comprehensive and also nonpartisan, and very readable as well. The main shortcoming is that neither author is an economist and so they gloss over the role played by the Federal Reserve in engaging in overly loose monetary policy. The first half of Nouriel Roubini’s Crisis Economics does a good job filling that in (the second half calls talks about the reforms he proposes, and I disagree with much of it). Michael Lewis’s The Big Short isn’t bad, but much more narrow in scope. Also, Mian & Sufi have also written a book, House of Debt, but I haven’t yet read it.
Anyway, here’s an easy way to tell if you’re reading a partisan explanation. Conservative partisans invariably call it entirely a government mistake. Liberal partisans try to blame it entirely on the free market. The truth, of course, lies somewhere in between. Both parties championed home ownership as the American Dream, and while the very first securitized mortgage was created by Ginnie Mae, it was the private markets that ran with it and made securitization obscenely huge.
Slightly edited from my original answer at Quora.