In early 2000, I met a guy called Ray Safaie.  Turned out, he had been extremely successful trading his own personal funds in the stock market and had amassed a small fortunte he managed from his San Francisco apartment.  He had educated himself with books he purchased at Amazon.com.

When I discovered the extent of his success, I asked him for some book recommendations so I could educate myself too.  I purchased the three books he recommended the next day and read all three within one week of their arrival.  I then went back to Amazon to do my own research and selected 32 books covering various areas of the stock market and technical analysis.

One of those 32 books was "The Roaring 2000s" by Harry Dent.  The author's hypothesis revolved around the Baby Boom generation and the established fact that the average American reaches their peak spending year at age 47.  As such, you could simply add 47 years to the basic birth rate statistics and get a fairly good estimate of consumer spending and, by association, stock market performance.

Harry Dent's hypothesis is overly simplistic to yield meaningful results but the underlying concept caught my attention and I began building a model to map everyone in the population into a dynamic population platform, allowing me to project age-specific income, spending and savings data to calculate aggregate figures.  The first version of my model became extremely complicated because I introduced too many variables including inflation, wage growth and a dynamic mortality curve, clouding the correlation I was looking for.

When the model reached 26 megs and the computations were stalling my computer, I abandoned the effort for almost a year.  Then, after continually thinking of the underlying concept, I started again with fewer variables.  I used constant 2000 dollars and included only historical wage growth (not future growth).  I also restricted the dynamic mortality curve to historical figures, assuming the current mortality curve would remain stable in the future.

I understand these assumptions are naive but the resulting simplicity made it much easier to isolate the correlation I was looking for.  Again, the curve itself is a function of changes over time, not dollars in one single year, so the use of simplifying assumptions detracts nothing from the model's findings.

All the input statistics were obtained free of charge from the US Census Bureau website, a tremendous resource.  Anyone could get the same data I used and build an identical model.  If you would like to do so, please contact me.  I would be happy to help.  If you have questions about my findings, contact me.  I spent hundreds of hours on this thing and am clearly enthusiastic about the project.