Investment "science"

I recently bought the book Investment Science by David Luenberger. I chose it because it was one of the best-sellers on wilmott.com, and it appeared to have the right mix of mathematics and economics. I am not unhappy with my purchase, it's a nice little book. I have read it once at an accelerating pace, and now I am going through for a second, more detailed reading.

The first thing that comes to my mind is the somewhat impertinent question (borrowed from a critic of computer science), just which part of the scientific method does "investment science" use?

The book is organized around the idea of cash-flow analysis. The first part of the book is about deterministic cash flow streams, interest rates and valuations. The second and third parts introduce random cash flow streams of increasing complexity. The fourth part is about futures, options and related problems. Put this way, the theory of investment becomes easy and reasonable. The basic kind of deterministic cash flow stream is the bond. Because bonds are standarized there is a well-established theory of them. The presentation would have made more sense to me on a first reading if the price/yield curve of a bond were explicitly related to the concept of internal rate of return introduced earlier for general cash flow streams.

The second "flaw" I found in the book came when the author discussed the "term structure" of interest rates. Basically, there is a different market interest rate for money loaned for different lengths of time. Usually, the interest rate increases monotonically with the loan time, although sometimes this is not so. Right now, looking at Bloomberg.com one finds that the interest rate for UK government bonds peaks for maturities of about 8 years, and decrases for longer maturities. This is not as expected, usually, and the bonds issued by all other countries monitored by Bloomberg show monotonic (sometimes very steep) yield curves.

Anyway, the book gives three different "explanations" of the term structure of interest. The first is "market segmentation": the theory that investors fall in different classes according to the maturity of the bonds they prefer to buy, and that so there need be no relation between the interests paid by bonds of different maturities. Since the yields are not observed to fluctuate wildly with maturity date, this theory does not seem very compelling.

The second theory has to do with risk. A longer maturity involves a larger risk that the issuer of the bond will default, and so should be associated to a higher yield. The problem is that this is essentially a probabilistic situation, but the book says nothing about needing to wait for later for a fuller analysis of this idea. Now, what does this theory imply about the market expectations for the British economy given the Bloomberg data?

And, now that I mention "market expectations", this is the third theory about the term structure of interest presented, and apparently the one preferred by the author. The argument goes as follows. A given term structure has implicit information about the market expectations for the term structure in the future. For instance, money invested today for 2 years at the 2-year spot rate should produce the same yied as if it were invested at the 1-year spot rate and then reinvested at next year's 1-year spot rate. This allows one to compute the forward rate for a 1-year investment made 1 year from now. The idea is that, if this forward rate did not match the market expectations, it would be possible for investors to make money by playing with the difference between the two (a no-arbitrage argument). The problem with this is that arbitrage arguments involve zero-risk profit, and loaning money at interest always involves risk. Moreover, if the spot rates increase with the maturity time, as they usually do, the expectation dynamics predict that interest rates will grow in the future, someties considerably. The troubling thing is that the book admits that this is not the observed behaviour, but that "expectation dinamics are very logical" and so are preferred to the other explanations.

Why not simply admit that without an analysis of default risk one cannot explain the term structure and move on to random cash flows? Why insist on using expectation dynamics all over the place and proving invariance theorems about them if that is not the way the market behaves? I find it hard to believe that people will actually invest money according to expectation dynamics... that sounds like a losing strategy.