football27 wrote: ↑Sat Oct 07, 2023 10:32 pm
Mountain Doc wrote: ↑Sat Oct 07, 2023 8:12 pm
football27 wrote: ↑Sat Oct 07, 2023 6:40 pmCan't someone just hold an S&P 500 ETF / Index Fund for long term and nothing else, wouldn't that outperform Boglehead 3 Fund portfolio in upcoming 30 years?
Can you articulate why you think the returns from 1957-2023 will be similar to the returns from 2023-2053?
I believe so because
I am just basing this off long past sample size, I am not sure what future returns will be like as in US, there's rumors of a potential recession, higher interest rates, and high inflation, and I am not sure how that will impact returns in the long term. I'm hoping there's still a strong return next 30 or so years.
Ah. There is an important problem in your statement "I am just basing this off long past sample size."
I would really recommend that you read The Misbehavior of Markets: A Fractal View of Financial Turbulence, by Benoit Mandelbrot and Richard Hudson.
Both in our intuitive thinking, and, if we have taken statistics courses, our statistical thinking, we are strongly influenced by the "normal distribution." The normal distribution is what we get when a natural phenomenon is structured as a sum of many independent random variables. Many things in the real world really do behave this way, or close enough, that the normal distribution really shows up. We can measure the degree of fluctuation as the "standard deviation," often represented by the Greek letter sigma, σ, which is a kind of average--the root-mean-square, to be precise--of the departures from the average value. In a normal distribution, we know, for example, that if we take many samples, 95% of them will like within ±2σ of the mean, and 99.7%, or virtually all, within ±3σ.
We also know that as we take more and more samples, the sample average "settles down" nicely in a regular way. In fact, the standard deviation of the average of N samples will be 1/√N of the standard deviation of the variable.
Well, financial data doesn't behave in that tame, manageable way. Not even close. And our judgements of the future based on the past are much less reliable than we intuitively expect.
In theory, there exists a statistical distribution, also found in the real world, called the Cauchy distribution, which also looks like a bell-shaped curve when plotted, but has "fat tails"--as you go farther and farther out, the probability looks as if it is declining just like a normal distribution, but it doesn't decline as fast. Extreme values are more likely to happen.
Fortunately, financial data is not this bad. Because the Cauchy distribution is a nightmare. Nothing about it settles down, no matter how many samples you take. It looks as if the average would be the peak, yet samples do not settle down and approximate the average, because no matter how many samples you take, sooner or later you get a wildly extreme sample, so extreme that it basically outweighs everything else you've averaged in.
Mandelbrot showed that financial data is somewhere in between normal and Cauchy. It's not as bad as Cauchy, but it's not as tame or well-behaved as (say) the distribution of human heights or weights.
An illustration of that was given during the collapse of the Long-Term Capital Management (LTCM) hedge fund in 1998. A principal of the firm said that the event that caused the collapse was a "ten-sigma event." If the financial variables had been following a normal distribution, a ten-sigma event would have occurred once in 100,000,000,000,000,000,000,000,000 years. Ten-sigma events are not common in finance, but they are more like once-in-a-decade events.
One big problem is that we have no reason at all to think that the stock market stays the same. Do you honestly believe that the stock market of the early 1900s, when there was no SEC... when "bulls" and "bears" meant operators who were making the market go up and down... and there was relatively little participation in the stock market by ordinary middle-class individuals... is quantitatively the same, and has the same averages and standard deviations and so on, as it does today?
So you have an insoluble problem. It is obvious that financial data has runs or periods or trends or "markets" (as in bull market or bear market) that persist, with seeming stability, for, say, five to fifteen years, and then change. So a period of 100 years isn't really 100 independent samples. It's more like, say, six or eight "markets." You don't really have 100 data points you have six or eight. So even a hundred years isn't a big sample. And yet it is long enough to raise serious doubts about whether the market has "stayed the same" over that whole period, or whether it might have "changed." By the time your data period gets to be long enough to get the sampling error down, it is so long that you doubt the hom*ogeneity of the sample... and, let's be clear, the quality of the data, too.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
