Author Topic: thoughts on the past performance disclosure and data mining  (Read 2255 times)

farangster

  • 5 O'Clock Shadow
  • *
  • Posts: 24
thoughts on the past performance disclosure and data mining
« on: March 12, 2016, 02:20:09 PM »
past performance may not be indicative of future results.......Bogle shows how today's morningstar five star funds are tomorrow's losers.  But expanding on those thoughts about past performance, we are all basing our investment choices on prior returns when you really think about it.  We invest in 60% stocks/40% bonds because historically stocks have had better performance than bonds.  We invest 50% or more in the U.S. because it has had higher historical returns than foreign investments......i doubt we would all be investing so heavily in the U.S. if foreign stocks had better long term historical returns than the U.S. even if the fees for investing in foreign stocks was a little higher. 

Let's look at everyone's FAVORITE personal finance guru, Dave Ramsey. His broken record speech is to invest in growth stock mutual funds with loooooong track records.  I do not like his advice at all for many reasons, but can we really knock him for shelling out this advice based on past performance?  His long term definition is something like 30 years.....can we really say that is horrible invest advice when we are all basing are investing decisions on historical returns? (just a longer time horizon).  Furthermore, if we use a longer time horizon, say 1000 years...we have seen the rise and fall of many hegemonic powers- Rome, Spain, U.K, etc. during that time.  Many top academics, i.e. Noam Chomsky, state that we are a hegemonic power in decline and obviously no super power has been able to sustain that status indefinitely.  therefore is it really wise to invest the bulk of our assets in a declining economic power based on it's superior stock market past performance during its rise to power over the last 100 years? 

Also, consider the traditional market-cap weighted index funds.  When Google went public it was only a tiny fraction of an total stock market index but today it is a major part of that index in comparison.  Companies that have grown to the largest size are weighted the heaviest in a market-cap index, therefore it is naturally placing heavier emphasis on business's with the strongest past performance.  Logically then, a traditional market-cap weighted index is a big bet that past performance will be indicative of future returns. 

I believe we are all investing in marketable assets based on past performance and should not be so quick to admonish other's that are data mining and investing in such a way to get higher returns..........all of us investing in stocks or bonds are data mining and have expectations based on those assets past performance.

GrowingTheGreen

  • Bristles
  • ***
  • Posts: 356
    • Growing The Green
Re: thoughts on the past performance disclosure and data mining
« Reply #1 on: March 12, 2016, 06:29:57 PM »
It's kind of hard to invest based on future returns :)

What do you mean by data mining in this context?

bobechs

  • Handlebar Stache
  • *****
  • Posts: 1068
Re: thoughts on the past performance disclosure and data mining
« Reply #2 on: March 12, 2016, 09:04:24 PM »
And how is one to find an investment market with no past, in order to avoid having one's purity of motive uncorrupted by knowledge of former events?

MustacheAndaHalf

  • Handlebar Stache
  • *****
  • Posts: 1771
Re: thoughts on the past performance disclosure and data mining
« Reply #3 on: March 13, 2016, 12:59:19 AM »
If you open up the Wall Street Journal, you'll see the "best" and "worst" funds in various categories.  And the time frame is year-to-date.  You could very well be looking at a major publication suggesting that 2 months of past performance is significant.  So let's start with that example, and say a few months don't predict future performance.

Once you're making predictions with 30 years of data, the data has much more perspective built into it.  Thirty year performance has multiple corrections and multiple business cycles.  You see times when international beat U.S. stocks, and how a mixture might improve on either one in isolation.

A longer time frame has another implication - how often will you change your portfolio?  If you follow the Wall Street Journal's lead, each month can bring a different set of year to date winners.  You see that the year-to-date performance differs from 1 year and 5 year performance.  You might even wonder why these great funds didn't exist 5 years ago, which is often the case.  Compare that to a portfolio based on 30 years of data, and implicitly you're only changing when the 30 year performance changes.

dmn

  • Stubble
  • **
  • Posts: 101
  • Location: Switzerland
Re: thoughts on the past performance disclosure and data mining
« Reply #4 on: March 13, 2016, 05:46:55 AM »
We invest in 60% stocks/40% bonds because historically stocks have had better performance than bonds. [... ]I believe we are all investing in marketable assets based on past performance and should not be so quick to admonish other's that are data mining and investing in such a way to get higher returns..........all of us investing in stocks or bonds are data mining and have expectations based on those assets past performance.

If it were just historical returns, I would not risk so much of my money in stocks. However, there are also very good economic theory reasons why stocks should overperform. For example, you can check some numbers to see that stocks can be expected to outperform. Buying stocks at a P/E ratio of 20, you have an estimated 5% yield from reinvested or distributed profits (though of course with much higher uncertainty because profits fluctuate wildly). Buying bonds at a 0% real interest rate, you know your return will be zero (but this is guaranteed at least).

With active funds, theory tells you that they will on average underperform due to fees.

The historical data serves as a consistency check for these theoretical arguments. So looking back at international historical data, you generally see that stocks outperform bonds and those outperform cash over long time periods in agreement with theoretical arguments. You also see that the overall performance of active funds is statistically consistent with overall stock performance minus fees, with the variation being just random year on year (some funds will have outperformed in the past just by chance, but one would not expect that they continue to outperform going forward).

So I disagree that one has to base investment decisions on historical returns. I invest in stocks because there are sound fundamental reasons for why they should perform well. If stocks reached P/E ratios of 100, I would not invest in stocks because then there is no longer a fundamental reason for why they overperform. I do not invest in bonds at the moment despite their historic performance, since with inflaton-adjusted yields below zero, their long-term return is now negative.

Seppia

  • Pencil Stache
  • ****
  • Posts: 616
  • Age: 39
  • Location: NYC
Re: thoughts on the past performance disclosure and data mining
« Reply #5 on: March 13, 2016, 05:21:22 PM »
It boggles my mind how someone would actually go out and buy a bond with negative interest rate.
"Hey you! Part with your money, give it to me, so that in three years I will give you back LESS".

When "stuffing money under the mattress" has better yield than an investment, one has to be very motivated to stick to its AA to keep on buying said investment.

dmn

  • Stubble
  • **
  • Posts: 101
  • Location: Switzerland
Re: thoughts on the past performance disclosure and data mining
« Reply #6 on: March 13, 2016, 05:31:48 PM »
When "stuffing money under the mattress" has better yield than an investment, one has to be very motivated to stick to its AA to keep on buying said investment.

I understand that "cash under the mattress" has a negative nominal yield because it can be lost in fires or stolen by burglars.

In Europe, banks have started to look into this option due to negative ECB interest rates. I have read estimates that if one includes storage and insurance costs, "stuffing cash in bank safes" has a nominal yield of around -1%.

Seppia

  • Pencil Stache
  • ****
  • Posts: 616
  • Age: 39
  • Location: NYC
thoughts on the past performance disclosure and data mining
« Reply #7 on: March 13, 2016, 05:37:21 PM »
Is this a serious remark?
I deem it unlikely that my online bank account that I pay exactly zero for will go out in a fire.
I'm willing to give you 100-1 odds whenever you please.

dmn

  • Stubble
  • **
  • Posts: 101
  • Location: Switzerland
Re: thoughts on the past performance disclosure and data mining
« Reply #8 on: March 13, 2016, 06:20:51 PM »
Is this a serious remark?

Well, yes, it is serious: see this news article.

European banks offer zero-interest accounts for private customers because they expect positive revenues from other parts of the customer relationship (e.g. eventually offering a mortgage to that customer).

For large institutional clients negative interest rates are becoming a reality, and they have to accept it because they cannot accept a 1-in-100 chance that their uninsured cash literally goes out in a fire.