Christopher L. Smith, at the Fed, makes the best stab at this question to date:
Since the beginning of the recent recession, the employment-population ratio for high-school age youth (16-17 years old) has fallen by nearly a third, to its lowest level ever. However, this recession has exacerbated a longer-run downward trend that actually began in the 1990s and accelerated in the early 2000s. There is little research regarding why teen employment has fallen. Some earlier work emphasized labor supply explanations related to schooling and education, such as an increased emphasis on college preparation (Aaronson, Park, and Sullivan 2006), while others have argued that adult immigrants have crowded out teens, at least in part because adult immigrants and native teens tend to be employed in similar occupations (Sum, Garrington, and Khatiwada 2006, Camarota and Jensenius 2010, Smith 2012). This paper presents updated trends in teen employment and participation across multiple demographic characteristics, and argues that, in addition to immigration, occupational polarization in the U.S. adult labor market has resulted in increased competition for jobs that teens traditionally hold. Testing various supply and demand explanations for the decline since the mid-1980s, I find that demand factors can explain at least half of the decline unexplained by the business cycle, and that supply factors can explain much of the remaining decline.
This paper is now the place where all others should start. I thank several loyal MR readers for the pointer.