Using Historical Financial Data

Henry K. Hebeler



“He uses statistics as a drunken man uses lamp-posts . . . for support rather than illumination.”

-Andrew Lang (1844-1912)


History is a record of the past.  It is not a record of the future.  The financial world uses lots of historical data including returns, inflation, gross domestic product, price to earnings, etc.  People expect the future to be similar to the past.  Attaching a statistical measure (mean, standard deviation, etc.) describes the past variations but does not necessarily reflect what the statistics will be for the remainder of your personal future.


It’s very common for financial analysts and journalists (who parrot the analysts) to say that you have a certain percent chance that your money will last your lifetime if you withdraw a certain percent next year and increase that amount by inflation in each subsequent year.  They should know better.  What they should say is that if they were living in the past, they would have succeeded a certain percentage of the time.  But this doesn’t sound as good and casts some doubt on the results.


It should cast some doubt!  One famous planner with all kinds of degrees working for a prestigious company said that in order to get an answer you could really count on from statistical retirement programs, you should run about a million simulations per data point.  After having done that, he said he learned that the best retirement portfolio was one with 100% stocks.  If you withdrew 7.5% of the balance in the first year, there would be only a 10.4% chance of failure for a male and 15.6% for a female.  (Note that he was confident enough to give the results down to 0.1%.) This paper appeared in several professional magazines, and was lauded by the financial press. Three years later, anyone who had followed that advice would have a 100% chance of failure!  He withdrew the paper several years after publication in light of mounting evidence that these withdrawal rates were leading people into the poor house—fast!


To make matters worse, some financial sites pretend to offer the data representing your particular choice of funds.  Of course none of these funds existed for very long, so the statistics are faked.  They assume a mean and standard deviation for each fund and may estimate an adjustment for the fund manager’s assumed style and costs of the fund.  When I went to school, we called this “dry labbing,” meaning that the results came from thin air, not from actual lab tests.


To further add to the historical uncertainty, the data completely ignores all of the securities that drop out of the data base because of bankruptcies or, in the case of funds, because their performance was so bad.  Every year mutual firms will start several funds in the hope that one will turn out well, so they can tout it.  The poor performing ones disappear in a couple of years.  For example, a fund that would represent the Dow Jones Industrials invests in the same 30 major companies as the Dow Jones Industrial index.  Only General Electric is left from the original firms.  The rest have underperformed and were replaced by “more representative” firms.  The same is true of all index funds.  In other words, the indexes represent the better performing securities.


The picture gets really messy when you start trying to account for inflation.  You should start to feel queasy when the statistical program asks you to input a constant inflation rate.  Inflation was different for each security return data point.  Some programs help put this in better perspective by using “real returns,” that is, returns less inflation.  (More precisely, returns less inflation divided by the quantity 1 + inflation.)


Still, how meaningful will historical data be  in a future where returns may go down when inflation goes up over an extended period?  The feds are faced with this quandary right now as they try to stimulate the economy by reducing interest.  But inflation is rising rapidly due to many factors including what is happening to import prices (particularly petroleum), diversion of grains to energy, and the declining value of the dollar compared to foreign currencies.  To fight inflation, they’d like to increase inflation rates to slow the economy, but they dare not do it.


Now, I don’t want to imply that you can’t learn anything from security statistics.  At the very least, you should get the sense that things are uncertain.  But you should never be left to believe that there is any accuracy in your projection.  I personally believe that the most useful results from simulations are to help you compare alternative strategies.  That’s why the comprehensive programs on have two programs working simultaneously:  One represents one set of assumptions and the other another set of assumptions.  You get a side-by-side comparison of the results.


The other thing I believe, is that the most truly representative statistics of the past are ones that use the actual time histories of inflation and returns of the past, not a jumbled-up set of numbers where the results from the Great Depression or the booming nineties are diced and cut and then mixed in a statistical salad bowl so that a data point from 1975 is back-to-back with a data point from 1995 as if the order of history had no influence on the result.  I believe that economic performance is very dependent on both what happened in the recent past as well as ever changing taxes, debt, global economics, wars, etc.  Statistical programs are oblivious to these things.


For these reasons, I chose to represent the actual time histories in the planning programs on  That way you can see what would have happened if you retired in, say, the year 1962.  There is no pretense that this represents what will happen in your future.  It’s simply a description of what would have happened to your investments taking into account the actual year-by-year histories of returns and inflation the way they actually occurred, not all mixed in a statistical salad bowl with pretence of foretelling the future with a certain degree of precision.


In my view, it’s a very bad idea to assume that what happened to a security in the past five, ten or even twenty years will represent what will happen in the next five, ten or twenty years.  Yet, that’s a very common thing for financial analysts.  That’s true for any period, but especially now when national savings are so low, debts so high, energy problems already on us, large numbers of people going from the workforce into retirement, and international changes taking place to say nothing of what are likely to be major tax law changes as the Congress tries to cope with Social Security, Medicare, immigration, and redistribution of wealth issues.


So use statistics with great caution when making projections.  Overly optimistic retirement projections get you to spend too much too early.  Early overspending depletes savings that must then be spread more thinly over the later years of you life.