Imagine a friend asks you to guess how tall a series of Miss America contestants are as they walk by on stage. Your friend also tells you that the average height for a contestant last year was 5’9”. If you forgot your glasses that day and were sitting so far away that you really couldn’t see them, a good strategy would be to guess they are all 5’9”. The volatility of your answers would be 0 (meaning each answer you give wouldn’t deviate from your average answer at all) and the error term would be random (meaning you’re not making a predictable mistake every time). If you were very good at guessing heights and didn’t forget your glasses, the best outcome you could hope for is to guess every height correctly. In this scenario, the volatility of your answers would match the volatility of the sample pool, and your error term would be 0. Chances are that most people would fall somewhere in between, not completely blind but not 100% accurate. If a reasonable person saw someone walk by who was actually 6’6”, they would think “this is unusual, 6’6” is very big, maybe I should temper what I believe I see with my knowledge of the sample mean… 6’3” sounds more likely”. This person’s estimates should actually have lower volatility than the sample pool and an error term commensurate with skill. If however, the guesser is a narcissist who believes himself to be much better at guessing than he actually is, he wont temper his estimate with the sample mean. This guesser would probably have a guess volatility similar to the sample mean, but a non-zero error term. If this narcissist was also manic, and got excited about unusually tall or short contestants, he might exaggerate his guesses and end up with a guess volatility greater than the sample pool volatility and a non-zero error term.
Using similar logic, Shiller demonstrates in a number of different ways how the level of observed volatility in the stock market indicates that the efficient market hypothesis is at best flawed. A stock should be worth the present value of future dividends. Given that we don’t know exactly what future dividends might look like, we have to make a guess today about the present value. However, with decades of historic stock price and dividend data, Shiller looks at how the price of a stock at some point in history compared to the actual value of discounted dividends moving forward into the future. Not surprisingly, the guesses were not only wrong, but also far more volatile than you would expect from rational and efficient market participants.
I give this book 4/5 stars despite the fact that it appears to deserve full credit because a) I’m not smart enough to fully understand Shiller’s work, and b) the average book should earn around 3 stars. By recognizing my own lack of understanding, I admit that portions of Shiller’s analysis which appear to be both interesting and correct, could in fact be wrong or misleading and therefore less deserving of stars. My lack of understanding creates volatility in my rating ability. By averaging my belief that this book is worth 5 stars with the long-run average review of 3 stars, I should reduce the expected error and volatility in my book reviews.
The book is very interesting if you like this sort of thing, but it’s also a very challenging read. My background in finance is stronger than my background in math and statistics, so I usually had to read the book with wikipedia nearby to reference math terms. Even with wikipedia, the equations got too hairy for me to follow with less than 100% concentration. I found myself flipping back pages almost as often as I moved forward, and there were parts I had to just skip over. If you read this book be ready for a serious challenge. Caveat emptor.