PF Investing Club: The Stock Market & Compounding Interest

[vela]

Amazon and Google are obviously giants, but they now have extremely steep prices. Would they still be a buy if you are looking at holding them for 20-30 years? Or is it better to play completely safe and throw everything into Vanguard 500? So hold tech giants for 30 years or Van500?

It doesn't have to be one or the other. You could put most of your money into index funds and use the rest to invest in individual stocks.

Index funds are the easy way to invest. Index funds are diversified, so they're relatively safe, but you're not going to get the dramatic growth you can from finding stocks that end up taking off. If you don't want to devote considerable time managing your investments, they're the way to go. You might want to look at a fund that mirrors the broader market though.

If you want to try your hand at picking individual stocks, I'd recommend reading books on investing in stocks and investing in general before you start trading. I liked Peter Lynch's books. Lynch ran Fidelity's Magellan Fund with great success over a pretty long run. In his books, he goes over the factors he looked at when evaluating different types of stocks. I also liked Burton Malkiel's A Random Walk Down Wall Street. He explains why technical analysis is stupid, and you'll learn the idea behind index funds. Vanguard's website also has a lot of good info on investing.

 

[Vanadium 50]

Many banks are offering 'offset' accounts these days

In the US? Washington Mutual had something like this right before they went under.

. Index funds are diversified, so they're relatively safe, but you're not going to get the dramatic growth you can from finding stocks that end up taking off.

But you're also avoiding the dramatic losses you can get from finding stocks that end up tanking.

 

[kyphysics]

Actually, your mutual-fund company (if you invest in their funds directly with them) or your brokerage (for stocks and ETFs) has to keep track of cost basis for you. This is a fairly recent thing, since about 2010 or so. At least for mutual-fund companies, not sure about brokerages. Every year they send you a form 1099-B that reports the details of the gains on your sales, for your tax returns. TurboTax and other tax software can import these online so you don't have to type in the numbers yourself. I still like to do it by hand with a spreadsheet, to check their figures and because I feel more comfortable when I understand what's going on.

However, all my ongoing monthly contributions have been to my tax-deferred 403b plan at work, where this doesn't come into play.

I set up my taxable brokerage account with inherited money, and make trades in it only to rebalance, once a year or so. I don't reinvest dividends in that account. I'm retired now, so I collect them and use them to support part of my spending. If I do start reinvesting dividends there at some point, I'll let them accumulate and reinvest them at rebalalancing time. I keep things simple there: Vanguard Total Stock Market Index ETF (VTI), Vanguard Total International Stock Index ETF (VXUS) and Vanguard Total Bond Market Index ETF (BND).

Appreciate the response. Gotta be honest, jtbell. Too complicated for me!

Applause that you actually take the time to understand it for yourself. I think that's a great and proper way of learning. Feynman would be proud!

I probably would do the same, but since I'm not at that stage of investing yet, I don't bother to learn the tax related stuff yet. At absolute worst, I figure I could ask a professional to do it for me.

 

[jtbell]

The basic principle is simple: when you sell something for a profit, you owe tax on the profit. That means keeping track of how much you paid for it (the cost basis) so you can subtract it from the gross proceeds of the sale. Applying this principle is messy when you have a portfolio with many chunks of stock (tax lots) bought at different prices over a period of many years.

It's sort of like physics. Newton's Laws are simple, but applying them to a real-world situation with lots of moving parts can be messy.

 

[Vanadium 50]

At absolute worst, I figure I could ask a professional to do it for me.

I think you mean "hire a professional to do it for me..

 

[kyphysics]

The 3 to 9 months is guideline, personally I ran much lower savings when I was younger. In my 50's I had about months 15, then boom the company offers a severance program and I retired.

I feel like in today's volatile and slow-growth economy that having that 6 months of living expenses saved up could be wise.

One of the worst things you can do with an IRA is to put money in and then raid it for emergency cash. It's not so much the fees (I think there might not even be any on the principal?), but the lost ability to get compounding on the investment. It'd feel like a major waste of energy and time to put money in and have to take it out before you get the full effect of time and compounding. I'd want to put money in, leave it, and never have to touch it for 20 years.

A six month stash of living expenses seems like it'd cover most cases of lost jobs, random budget breaking expenses, and emergencies that could crop up. Some people might even say 9-12 months. All subjective and reasonable.

I like a bare minimum of six months. Seems like a good baseline. After that, I'd feel more relaxed and I could put money into an investment account and likely not have to raid it for cash.

 

[Stephen Tashi]

I think the majority of trades on the major stock exchanges nowadays are done by computer algorithms. It's an interesting question whether human "day traders" have suffered because of this - but that doesn't bear on the conventional wisdom for human investors, which is to do "long term" investing.

However, it's worth considering whether the historical statistics for returns on stocks will continue to be valid if computers and artificial intelligence tackle long term investing.

Any discussion of that possibility will mainly be a debate on whether adequate technology will exist for it in our life time. However, (to me) the fundamental question is one of economic theory. Do returns on stocks depend on some sort of inefficiency in how people set current prices on stocks? - or are returns on stocks mainly due to differences in tastes - personal preferences for spending now or investing for later? ( There is the notion that both sides win when a trade is executed. The seller prefers something tangible now. The buy is willing to wait.)

 

[russ_watters]

However, it's worth considering whether the historical statistics for returns on stocks will continue to be valid if computers and artificial intelligence tackle long term investing.

I don't see why they wouldn't. Fundamentally, the value of a stock is tied to the health of the company. I don't see how automation of trading could change that.

Do returns on stocks depend on some sort of inefficiency in how people set current prices on stocks? - or are returns on stocks mainly due to differences in tastes - personal preferences for spending now or investing for later?

I don't understand the questions... or maybe you're missing the third option - the actual answer - to the questions? Your first question seems to imply that stocks grow in value over time because they are priced to low now. No: they grow over time because they - the company they represent - become more valuable over time.

 

[Stephen Tashi]

I don't see why they wouldn't. Fundamentally, the value of a stock is tied to the health of the company. I don't see how automation of trading could change that.

Aren't historical returns on stocks dependent on the degree of accuracy (as far as predictive accuracy) that went into setting the day-to-day price of stocks over that history? If an investor (human or computer) was statistically more accurate than other investors at predicting future stock prices, it seems to me that his historical returns would be higher than other investors. The historical returns on stocks are based on historical prices that were negotiated among a mix of investors with diverse skills at prediction.

Automation of trading, need not not change the current or future health of a company, but it might (eventually) be able to make more reliable estimates of current and future health of a company and thus be more skilled at setting a fair current price for a stock than non-algorithmic traders.

If computer algorithms become more skilled at setting fair prices for stocks than negotiations among a mix non-algorithm investors then these algorithms will dominate the market. They will be buying and selling versus other algorithms with similar skills. The daily market price of a stock will be dominated by what the algorithms cause to happen.

I agree that a market dominated by algorithms can still show positive returns from a buy-and-hold strategy. I'm just asking whether estimating the size of these returns from historical data is a good guide to what would happen under such different circumstances.

 

[russ_watters]

Aren't historical returns on stocks dependent on the degree of accuracy (as far as predictive accuracy) that went into setting the day-to-day price of stocks over that history?

That's a confusing statement. The return is based on selecting/purchasing a mix of stocks and the return of those stocks. Selecting the stock mix doesn't "settle" (set?) the price. I'm not sure what you are getting at with that word. Perhaps more importantly, the whole point of the buy-and-hold strategy is that the short-term fluctuations average-out over the long term.

If an investor (human or computer) was statistically more accurate than other investors at predicting future stock prices, it seems to me that his historical returns would be higher than other investors.

Yes.

The historical returns on stocks are based on historical prices that were negotiated among a mix of investors with diverse skills at prediction.

Yes.

Automation of trading, need not not change the current or future health of a company, but it might (eventually) be able to make more reliable estimates of current and future health of a company and thus be more skilled at setting a fair current price for a stock than non-algorithmic traders.

Yes.

If computer algorithms become more skilled at setting fair prices for stocks than negotiations among a mix non-algorithm investors then these algorithms will dominate the market. They will be buying and selling versus other algorithms with similar skills. The daily market price of a stock will be dominated by what the algorithms cause to happen.

Ok. So what?

I agree that a market dominated by algorithms can still show positive returns from a buy-and-hold strategy. I'm just asking whether estimating the size of these returns from historical data is a good guide to what would happen under such different circumstances.

If an algorithm is ever created that is able to beat the market, it will be the first, and congratulations to its creator. But that doesn't have anything to do with the long-term investing we are discussing. It won't change the long-term prospects of normal mutual funds. If anything, it would reduce the short term fluctuations by adding a buffer against unwarranted speculation.

It almost sounds like you are intending to imply - but not coming out and saying - that if automated stock trading ever becomes successful it will reduce the long-term gains of things like index funds. Is that what you are suggesting? Can you explain why?

 

[Stephen Tashi]

That's a confusing statement. The return is based on selecting/purchasing a mix of stocks and the return of those stocks. Selecting the stock mix doesn't "settle" (set?) the price.

Calculating return from historical data requires assuming you bought a stock (or a mix of stocks) at some historically "current" prices. Return calculations are not independent of your purchase prices. If purchase prices become based on more accurate forecasts then their relation to future prices may change.

For example, investments such as bonds have a higher degree of predictability that stocks. Their historical returns are (I think) lower than returns on stocks because their purchase prices are computed free of much of the unpredicatability associated with stocks. If the future of stocks were to become more predictable, they might become more like bond investments and give lower percentage returns.

If an algorithm is ever created that is able to beat the market, it will be the first, and congratulations to its creator.

As I said, one objection to the scenario of computers taking over long term trading is that it is impossible. But it's worth pointing out that the computer take-over of short term trading is not due some new kind of intelligence. Computer trading algorithms make the same decisions that common sense human traders make, but computers do things faster. Short term computer trading algorithms do beat the market.

But that doesn't have anything to do with the long-term investing we are discussing.

Anything that causes a major change in how current prices for stocks are derived will have an effect on long term investing. Long term investors must buy at some current price.

 

[russ_watters]

Calculating return from historical data requires assuming you bought a stock (or a mix of stocks) at some historically "current" prices. Return calculations are not independent of your purchase prices.

Agreed.

If purchase prices become based on more accurate forecasts then their relation to future prices may change.

For individual cases or in the short term, but not in the aggregate. On the whole, the value of the stock market is what it is.

For example, investments such as bonds have a higher degree of predictability that stocks. Their historical returns are (I think) lower than returns on stocks because their purchase prices are computed free of much of the unpredicatability associated with stocks.

Correct: bond returns are guaranteed.

If the future of stocks were to become more predictable, they might become more like bond investments and give lower percentage returns.

Not correct. You should never forget that in the aggregate the value of stocks is the value of the companies the stocks represent. The only way for the returns to be substantially different over the long term is for the value of the companies to be substantially "wrong" over the long term. "Unpredictable" means some are wrong high and some are wrong low over the short term, but over the long term the returns become more stable -- the unpredictability averages out.

For example, if a lot are wrong high, that's called a "bubble". Bubbles don't last forever though; they burst, bringing the market on the average (the aggregate) back to a reasonable price.

As I said, one objection to the scenario of computers taking over long term trading is that it is impossible.

I don't know who would say that/what it means because it is quite possible and is being done. Heck, to some degree computers are trading for me: buying stocks automatically for me every payday!

But it's worth pointing out that the computer take-over of short term trading is not due some new kind of intelligence. Computer trading algorithms make the same decisions that common sense human traders make, but computers do things faster. Short term computer trading algorithms do beat the market.

On the whole, no they don't. If someone had a winning formula, everyone would know it and everyone would invest with that company.

Anything that causes a major change in how current prices for stocks are derived will have an effect on long term investing.

Again: not on the whole, it won't. In order for the long term value to be affected, either the price now or the price then has to be substantially "wrong". And on the whole, the value of the stock market isn't "wrong", it is what it is.

Edit:
Let's be a bit more specific:

When someone says the current stock market is "wrong" or "right" (over or under valued), what they are talking about is the p/e ratio. Here's the historical p/e ratio:
http://www.multpl.com/

The historical average is between 15 and 20. It might vary naturally due to economic growth or recession, but overall it should always return to that range. If the computers have the effect of making the market more accurately rated, what they will do is flatten-out that graph, softening the boom and bust cycle. But they will not change the range, because in the aggregate that's what the range is "supposed to" be.

 

[Stephen Tashi]

On the whole, the value of the stock market is what it is.

That's a tautology, of course. Are you saying there is a "true" value of a collection of stocks that doesn't depend on their current prices?

Not correct. You should never forget that in the aggregate the value of stocks is the value of the companies the stocks represent.

You seem to be making a distinction between "value" and "price". I agree that such a distinction is theoretically possible, but how do you intend to quantify "value"?

"Unpredictable" means some are wrong high and some are wrong low over the short term, but over the long term the returns become more stable -- the unpredictability averages out.

Computing long term statistics based on historical data shows post-dictability, not pre-dicitability. What I'm asking about hypothesizes a change in the nature of the mix of "some are wrong high and some are wrong low".

On the whole, no they don't. If someone had a winning formula, everyone would know it and everyone would invest with that company.

On the whole they do "beat the market" - they trick the market. High frequency trading algorithms don't try to pick companies. They deal in stock prices. There are many different stock exchanges. Beside the major exchanges, there are also the "dark pool" exchanges. As a simple case, if an offer to sell 1000 shares of company X at 14 comes up on one exchange and an offer to by 950 shares of company X at 15 1/4 comes up on another exchange, a computer program can quickly detect the existence of this pair of offers in the stream of real time data and simultaneously buy 1000 shares of X at 14 on one exchange and sell 950 shares of X on the other at 15 1/4. (There are probably much more complicated examples of such "sure thing" situations that involve trading more than one stock and including puts , calls, futures etc.)

A human trader can, in principle, try to find the same sorts of deals, but he can't execute the algorithm (including the search for deals) as quickly as a computer.

When someone says the current stock market is "wrong" or "right" (over or under valued), what they are talking about is the p/e ratio.

That's not true of people who seek "growth stocks" The terms "over valued" and "under valued" have no standard definition.

If the computers have the effect of making the market more accurately rated, what they will do is flatten-out that graph, softening the boom and bust cycle.

I agree, that's plausible.

But they will not change the range, because in the aggregate that's what the range is "supposed to" be.

Do you mean the range of p/e's ? I think your argument that is that the financials of companies (e.g. revenue, expenses, liabilities, etc) would not be affected if predictions of the future stock prices became more accurate. That's a complicated question because a company's ability to borrow often depends on the outlook for the price of its stock. Also the "p" in p/e is based on stock price.

 

[russ_watters]

That's a tautology, of course.

Yes.

Are you saying there is a "true" value of a collection of stocks that doesn't depend on their current prices?

You seem to be making a distinction between "value" and "price". I agree that such a distinction is theoretically possible, but how do you intend to quantify "value"?

No, it seems that you are suggesting these things! I'm saying the value of the stock market (a collection of stocks) is what it is.

Computing long term statistics based on historical data shows post-dictability, not pre-dicitability. What I'm asking about hypothesizes a change in the nature of the mix of "some are wrong high and some are wrong low".

I'm aware. And I'm saying that the way that would manifest does not change the long-term outlook either forward or backwards in history.

On the whole they do "beat the market" - they trick the market.

High frequency trading algorithms don't try to pick companies. They deal in stock prices. There are many different stock exchanges. Beside the major exchanges, there are also the "dark pool" exchanges. As a simple case, if an offer to sell 1000 shares of company X at 14 comes up on one exchange and an offer to by 950 shares of company X at 15 1/4 comes up on another exchange, a computer program can quickly detect the existence of this pair of offers in the stream of real time data and simultaneously buy 1000 shares of X at 14 on one exchange and sell 950 shares of X on the other at 15 1/4. (There are probably much more complicated examples of such "sure thing" situations that involve trading more than one stock and including puts , calls, futures etc.)

Please show me an example of a fund that does this. I want to invest in it!

Anyway, this suggested approach a) doesn't have anything to do with what we are discussing (that stocks are over-valued and computers will figure that out) and b) represents either a technical flaw or illegal activity. I believe I've read about companies getting in trouble for purposely delaying stock price reporting in order to create and exploit such a discrepancy.

That's not true of people who seek "growth stocks" The terms "over valued" and "under valued" have no standard definition.

Yes, it is/does:
http://www.investopedia.com/terms/o/overvalued.asp

I agree, that's plausible.

Ok...then you seem to be acknowledging what you've been suggesting is wrong? Are we done?

Do you mean the range of p/e's ? I think your argument that is that the financials of companies (e.g. revenue, expenses, liabilities, etc) would not be affected if predictions of the future stock prices became more accurate. That's a complicated question because a company's ability to borrow often depends on the outlook for the price of its stock. Also the "p" in p/e is based on stock price.

No, I have not made that argument. You have been suggesting that the value of companies might, in the aggregate be very wrong and "fixing" that would change the future stock price outlook and make market returns drop. This would have to manifest as the "normal" p/e ratio for stocks going down dramatically and continuously and I'm saying that isn't possible.

Here's what your suggestion would look like:

Currently, say the historical average p/e ratio is 15 and say the computers figure out that that's wrong and it should have always been 12. The market will then drop by 20%, once, and then return to its long-term average growth rate after making that correction.

On the other hand, if what you suggest were true - that the long term growth rate could be substantially and permanently reduced - the "right" p/e ratio would have to continuously drop (12 next year, 11 the year after that, etc...), which is a contradiction in your suggested mechanism and can't happen.

 

[Borg]

I have been investing for about 30 years. I originally invested only in index funds offered through my 401K plans. These were easy to deal with, didn't require a lot of thought or maintainence but, didn't have the greatest returns. As my ability to invest increased (i.e. no debt), my wife and I started investing in individual stocks. We have made some bad decisions and some very fortunate ones.

One thing to note about the 'fortunate' stocks. Several of our high gainers went up drastically because they were purchased by other companies. Two recent ones were turned into cash when the sale finalized and we were forced to pay capital gains that year. If the gain is significant enough, this could impact your federal taxes such that you could end up underpaying by more than 10% at the end of the year. If that happens, the IRS will impose a penality.

I currently use an investment strategy designed by a friend with a masters in Economics that was turned into a program by myself. Basically, we score stocks according to many factors such as P/E ratio, historical debt, cash, etc. Blue chip and aristocratic stocks get a small bump in their score as well. The score of a stock determines its investment value which, in our case, means how many shares we would be willing to purchase. We then examine the stock manually. Finally, we don't purchase a stock if it is above its current 6 year average. This is so that we can pick up good stocks that are probably undervalued. While this limits our ability to purchase stocks that never fall below their average, it does tend to limit short term losses. I have not had any serious losers since I have been following the advice of this program.

 

[MarneMath]

Due to my job, I'm not allowed to give much advice regarding this. I will reiterate my position that for most people, picking their own stocks is absurd and for individuals I've seen who have done so their profile typically carries ungodly amount of risk. Are there people who do it right, sure. It isn't common. There are many theories to picking stocks, and the most popular on this forum seems to be the more traditional value based approach. There's nothing wrong with that. I'll say though, that focusing on solely PE, and ignoring the 4 other fundamentals to markets is foolish.

However, if you're goal is simply to beat your saving's account, that's relatively trivial task. Investing to live off your gains is a completely different ball game.

 

[Vanadium 50]

Stephen, it sounds like you are describing arbitrage - the simultaneous buying and selling to take advantage of a differing price. You don't need to go to multiple exchanges to see this: it commonly appears during mergers: if ABC merges with XYZ such that an owner of XYZ gets two shares of ABC, and ABC is trading at 10, you'd expect XYZ to be worth 20. But it might be worth 19.99 or 20.01.

What you will find is that the ratio of the prices of the two stocks is not 2.00 exactly, but moves around in a narrow band. The width of that band is determined by transaction costs and the return you would get on a "safe" investment. If the merger is in a month, to make this worth doing, you have to make more money with all this buying and selling - after transaction costs - than you would by simply buying a 28-day treasury.

What you are describing is a more efficient arbitrage. That tends to reduce transaction costs, and that tends to reduce the width of this band. It doesn't do anything to the underlying value, which is driven by future returns.

I'll say though, that focusing on solely PE, and ignoring the 4 other fundamentals to markets is foolish.

I think one thing which hasn't been addressed as much as it should is that a low P/E may mean that a stock is cheap. Or it may mean that it's on a slide into oblivion.

 

[Stephen Tashi]

Please show me an example of a fund that does this. I want to invest in it!

I don't know of an pure plays on high freqency trading open to public investment. You can invest in large banks, like Goldman-Sachs, which have high frequency trading departments.

Anyway, this suggested approach a) doesn't have anything to do with what we are discussing (that stocks are over-valued and computers will figure that out) and b) represents either a technical flaw or illegal activity.

It does seem an "unfair" activity, but it is not illegal. In fact, the secrecy of the computer codes companies use is protected by laws (https://en.wikipedia.org/wiki/Sergey_Aleynikov). The example only has to do with your remark that high frequency trading doesn't beat the market.

I believe I've read about companies getting in trouble for purposely delaying stock price reporting in order to create and exploit such a discrepancy.

The new IEX stock exchange delays quotes in order to thwart high frequency trading. Real time quotes are what high frequency traders want.

Yes, it is/does:
http://www.investopedia.com/terms/o/overvalued.asp

That says the p/e ratio is the "most popular" method of valuing stocks - not the only method.

Ok...then you seem to be acknowledging what you've been suggesting is wrong?

On the contrary, it would be an example of what I'm suggesting.

No, I have not made that argument. You have been suggesting that the value of companies might, in the aggregate be very wrong and "fixing" that would change the future stock price outlook and make market returns drop. This would have to manifest as the "normal" p/e ratio for stocks going down dramatically and continuously and I'm saying that isn't possible.

Why do you say the "normal" p/e would go down? And why continuously?

Here's what your suggestion would look like:

Currently, say the historical average p/e ratio is 15 and say the computers figure out that that's wrong and it should have always been 12. The market will then drop by 20%, once, and then return to its long-term average growth rate after making that correction.

I don't know whose long term average growth rate you're talking about - the market's or the stock's. Any wiggly curve can be be assign a "long term average" rate of change by computing an average. An over any historical period, the curve "returns" to its average rate because that's how the average rate is computed.

On the other hand, if what you suggest were true - that the long term growth rate could be substantially and permanently reduced - the "right" p/e ratio would have to continuously drop (12 next year, 11 the year after that, etc...), which is a contradiction in your suggested mechanism and can't happen.

If a future stock price could be more accurately predicted, it does not follow that its current market price would drop or that that its current market value would rise. The current market value of an investment with a known future value depends on what people (or computer programs) are willing to tie-up in the investment for a period of time in order to get a larger amount back at a later time.

 

[Stephen Tashi]

Stephen, it sounds like you are describing arbitrage

Yes, but the topic came up as a digression from the topic of the OP [ meaning my OP in the thread] , which wasn't about short term trading.

That tends to reduce transaction costs, and that tends to reduce the width of this band.

I don't know why it would tend to reduce transaction costs - and whose costs are we talking about ? The exchange's? The traders? - and that's yet another digression.

It doesn't do anything to the underlying value, which is driven by future returns.

"Underlying value" is ambiguous concept, but I certainly agree that the value-as-measured-by-current-price-of-a-stock set by an investor does depend on his predictions for that same measure of value in the future.

I think one thing which hasn't been addressed as much as it should is that a low P/E may mean that a stock is cheap. Or it may mean that it's on a slide into oblivion.

And promising start-ups have infinite p/e's !

 

[MarneMath]

I'm not sure what the point of the discussion between Russ and Stephen is really about. Comparing HFT to long term investments is like comparing collecting quarters for the laundry machine vs saving for a mortgage. Yeah your saving money for a task but the time frame and goals are completely different. HFT typically deal in low capital short positions profiting on cents and crowding of markets. HFT have many strategies most if not all strategies have little concern for market value but rather liquidity and over-crowding.

If the discussion is more about how HFT may effect long term value. That's not such an open and shut question. There exist papers regarding how the algorithms may help long term investors, but there are also reasonable papers on how they may negatively effect long term investors. Heck there's even papers on how it has no effect. Point is that it isn't simple to say one way or another.

 

[Stephen Tashi]

I'm not sure what the point of the discussion between Russ and Stephen is really about.

I'm not either!

Comparing HFT to long term investments is like comparing collecting quarters for the laundry machine vs saving for a mortgage.

The point of mentioning HFT was to only raise the additional possibility that (future) computer programs might also dominate long term investing. HFT illustrates the fact that current computer programs do have the ability to deal with stock transactions and are effective over the short time spans they seek to handle.

If the discussion is more about how HFT may effect long term value. That's not such an open and shut question. There exist papers regarding how the algorithms may help long term investors, but there are also reasonable papers on how they may negatively effect long term investors. Heck there's even papers on how it has no effect. Point is that it isn't simple to say one way or another.

I agree, it isn't simple!

 

[MarneMath]

The point of mentioning HFT was to only raise the additional possibility that (future) computer programs might also dominate long term investing. HFT illustrates the fact that current computer programs do have the ability to deal with stock transactions and are effective over the short time spans they seek to handle.

I concur. Betterment basically uses algorithms to buy ETF and automates the risk pool allocations. It isn't a leap to think algorithms can take in earning statements, and market conditions as inputs and make stock selections as well as your common day trader.

 

[Vanadium 50]

Algorithms are nothing new. When Daddy Warbucks would say "Buy it at 8 and sell it at 10!" that's algorithmic trading. A less fictional example is a stop-loss order. An example that people don't usually think of as algorithmic is an index fund: if ABC drops off the index and XYZ replaces it, the fund has to sell ABC and buy XYZ.

 

[Stephen Tashi]

Algorithms are nothing new. When Daddy Warbucks would say "Buy it at 8 and sell it at 10!" that's algorithmic trading. A less fictional example is a stop-loss order. An example that people don't usually think of as algorithmic is an index fund: if ABC drops off the index and XYZ replaces it, the fund has to sell ABC and buy XYZ.

That's correct. What's new is the speed at which algorithms can be executed and the amount of data they can take as inputs. For example, "disciplined" investors who take a long term approach may apply the same series of calculations to the financial statements of each company they analyze. How long does it take a human being to do such work? A human being can probably be competitive with a computer if the scope of the investigation covers only the amount of material that human would examine. But if we extend the scope of the investigation beyond that limit by using algorithms designed by humans on the theme of "These are things I would do if I only had the time and resources" then computer programs have the advantage. (For example, it is claimed that some HFT algorithms can consider breaking news stores - i.e. they analyze text data. So is it a stretch to think that a program trying to do long-term investing could analyze articles written about a company ?)

Human long term investors aren't supposed to "buy and hold" blindly. They are suppose to re-examine their investments periodically to detect fundamental changes. How often do they have time to do this?

I suppose it's possible that increasing the amount of data used to make a prediction about a stock will not increase the reliability of the prediction. That would keep the world of long term investing safe from the invasion of computers.

 

[NTL2009]

...

Human long term investors aren't supposed to "buy and hold" blindly. They are suppose to re-examine their investments periodically to detect fundamental changes. How often do they have time to do this? ...

I think people way overestimate the value of "examine an investment for fundamental changes".

Let's say a company has experienced a "fundamental change" - their market has shrunk, or they are no longer competitive in that market, etc. OK, but that information is available and is already reflected in its stock price. There is nothing you can do about that.

If you sell, the real question is will what you buy do better than the one you just sold from that point forward.

So what you are concerned with is, how will this stock do in the future (compared to the market)? That is not known. They may find a way out of their mess, or re-invent themselves, or get bought out. But we don't have a functioning Crystal Ball, so we can't make a decision based on that. But the market will consider those possibilities, and assign some sort of value to it, and can be right or wrong in aggregate. But none of that really helps us, does it?

I've read some studies that say there is value in buying stocks that are "out of favor", companies that people expect to do badly. Their price has been beaten down, and human nature tends towards "group think", so too many people sell, driving the price lower than fundamentals might indicate. In aggregate, these stocks will exceed their artificially low expectations, and do better than market averages.

If that study holds, the results are counter to what an individual would do on the above advice. They would sell at a low price to dump a stock with poor fundamentals. That might be the worst approach. Or using the reverse Crystal Ball quote from Will Rogers again, "If it don't go up, don't buy it!".

This is why I'm a fan of buying broad based index funds and ignoring them. Works just fine.

 

[russ_watters]

That's correct. What's new is the speed at which algorithms can be executed and the amount of data they can take as inputs.

I'll get back to the other stuff later (I'd still like to know what, if anything, you were really getting at), but I'll put in my agreement with @NTL2009:

Succinctly: It is a myth that managed funds do better than index funds. (this should be added to the financial knowledge thread). They don't. In fact, almost all of them do worse.

http://www.marketwatch.com/story/wh...-funds-beat-the-sp-than-we-thought-2017-04-24

The performance of actively managed funds is so bad that just by luck, you would expect half of them to beat the S&P and 5% basically means they are all bad. So you can essentially say that all managed funds are a bad idea because the only way for them to beat the S&P is by luck.

Applying this fact to the current discussion gives us:

For example, "disciplined" investors who take a long term approach may apply the same series of calculations to the financial statements of each company they analyze. How long does it take a human being to do such work? A human being can probably be competitive with a computer if the scope of the investigation covers only the amount of material that human would examine.

Per the above: no investor, disciplined or otherwise, professional or amateur, human or computer should be doing any work to analyze individual companies when it comes to setting up long-term retirement investment funds.

But if we extend the scope of the investigation beyond that limit by using algorithms designed by humans on the theme of "These are things I would do if I only had the time and resources" then computer programs have the advantage.

Certainly: but anything times zero is still zero. Since the above stats show us conclusively that managed funds don't beat the market, we flat-out shouldn't be using them -- human or computer managed.

(For example, it is claimed that some HFT algorithms can consider breaking news stores - i.e. they analyze text data. So is it a stretch to think that a program trying to do long-term investing could analyze articles written about a company ?)

I think it is entirely possible that HFT trading, news analysis and psychology may combine to beat human trends in investing. The problem is that since HFT already dominates trading, the ability of humans to change market values with impulse-buys/sells is already been reduced.

And in any case, as someone pointed out above, this has nothing to do with long-term investing. They are fundamentally different games that do not impact each other. Indeed, day-trading, unlike investing is zero-sum. It is theoretically required to have both winners and losers, whereas for investing it is theoretically possible to have only winners.

Human long term investors aren't supposed to "buy and hold" blindly.

Yes they are. That's what "buy and hold" means. Since nobody can beat the market in the long term except by luck, nobody should be trying.

The only significant change people should be making is adjusting to life change and aging:
-Sell a house, buy stock with the the proceeds.
-Get older, reduce your risk level.

I suppose it's possible that increasing the amount of data used to make a prediction about a stock will not increase the reliability of the prediction. That would keep the world of long term investing safe from the invasion of computers.

This goes back to our earlier conversation: you still have not outlined exactly what that would look like/how it would manifest. I'm contending that because day-trading/HFT and investing are fundamentally different games, they do not impact each other in any meaningful way. I think you are operaing on a misconception about how the markets work, but I haven't figured out what yet -- perhaps the zero sum game misconception?

 

[NTL2009]

...
Per the above: no investor, disciplined or otherwise, professional or amateur, human or computer should be doing any work to analyze individual companies when it comes to setting up long-term retirement investment funds. ...

Excellent points about the professional money managers who simply do not beat the market consistently. I just want to add that this appears so counter-intuitive to most people, that I feel it is worth some expansion:

We live in a world where we have learned that we can generally expect a professional, such as a plumber, carpenter or brain surgeon, to do a better job than an untrained amateur. They have the education, training, experience, and tools for the job. So people expect the same from financial 'professionals' in terms of stock picking.

But as we've shown, stock picking is about future prices. And no amount of education, training, experience, or tools can help predict the future. And whether you accept this explanation or not is actually irrelevant. The facts are as russ_watters stated - most money managers do not consistently beat the indexes. What makes you think you can?

I put more faith in the idea that a broad group of companies in general will continue to create value, and therefore enrich me with my investments in them, than I have faith in the idea that I can pick the 'winners' and 'losers' among them. I'm retired, so being wrong could be very painful. I'll stick to the game that has an evidence based (this is a physics forum after all) advantage for me.

 

[Stephen Tashi]

competitive in that market, etc. OK, but that information is available and is already reflected in its stock price.
.

It's an interesting assumption that the consensus reached by "the market" implicitly accounts for all the relevant information. Why is that assumption sound? For example an investor with a million dollars to invest in a stock can have more affect on its price than an investor with half that amount. Is the investor with more money twice as wise?

I agree with the argument that by the time the average human investor had realized there is a change in the fundamentals of a stock then the price of the stock will probably already have changed. However, speedy investors or speedy computer programs might realize things quicker than the consensus of the market.

I recall a situation in a book by Feynman where he told of a controversy about the content of science textbook. I think he reviewed it for a school board. His criticisms of the book were met by the objection that the book was given a high average rating in a survey of a large number of engineers and scientists. Feynman said he didn't claim his opinion was better than that of the smartest engineers and scientists, but he thought it was better than the average opinion.

 

[NTL2009]

It's an interesting assumption that the consensus reached by "the market" implicitly accounts for all the relevant information. Why is that assumption sound? ...

Did I say it was "sound"? No. I only said "it was". So for you to have a different assessment than "the market" would indicate you know something the others don't. I don't think that is likely (and it may be illegal) on a consistent basis. But I do think it is possible from time to time. I do think "the market" can over/under-react to news, or the view of the average person may be different than some people more familiar with the industry (but the "pros" doing the big buying know specialists in the field).

But as I said earlier, the reasons barely matter. If this was do-able consistently, we'd see a LOT of managed funds consistently outperforming the market, there is huge motivation for this - but we don't. If you are convinced you can do this, start a mutual fund.

Could computers identify some of this, either anomalies or just be faster to react to news? Possibly, but some of that still is predicting the future. But they will be in competition with other computers, and things will average out over time anyhow. If company xyz hits $100, it matters little if I bought it for $20 ten years ago, or $19.99 ten years ago. And most of these computer algorithms are down to scrapping a penny or less here and there. I don't think it will fundamentally change things. I could be wrong, but if it is fundamentals that drive the market in the long run, I just don't see how some computers getting in a little early can have any large effect.

 

[Stephen Tashi]

The performance of actively managed funds is so bad that just by luck, you would expect half of them to beat the S&P and 5% basically means they are all bad. So you can essentially say that all managed funds are a bad idea because the only way for them to beat the S&P is by luck.

That's like saying if 95% of high school students can't understand theoretical physics, we should give up expecting any of them to understand it.

I agree with the advice that index funds are good choice if the decision is to pick an index fund at random versus pick a managed fund at random.

Per the above: no investor, disciplined or otherwise, professional or amateur, human or computer should be doing any work to analyze individual companies when it comes to setting up long-term retirement investment funds.

(!) I wonder what salaries pension fund managers make.

Since the above stats show us conclusively that managed funds don't beat the market, we flat-out shouldn't be using them -- human or computer managed.

Are there yet many computer managed funds? Do we have extensive statistics about their performance?

This goes back to our earlier conversation: you still have not outlined exactly what that would look like/how it would manifest.

I merely asked a question. I don't know how computers will impact the long term behavior of the stock market.

I'm contending that because day-trading/HFT and investing are fundamentally different games, they do not impact each other in any meaningful way.

That's wasn't the focus of my question. HFT came up only as an example that computers are already a factor in the markets.

I think you are operaing on a misconception about how the markets work, but I haven't figured out what yet -- perhaps the zero sum game misconception?

At least I know some facts about HFT and the various stock exchanges.

 

[NTL2009]

That's like saying if 95% of high school students can't understand theoretical physics, we should give up expecting any of them to understand it ...

No, it's not like that at all.

The funds in the study have professional managers, not just picked at random.

It's more like saying if 95% of scientists cannot replicate the published results of an experiment, then maybe that result should be questioned?

I agree with the advice that index funds are good choice if the decision is to pick an index fund at random versus pick a managed fund at random.

No again. Look some more into the studies of actively manged funds. Very few of those 5% that beat the market in a five year period, are able to repeat the performance in the next 5 year period (which is what matters to us).

So how are you going to pick a 'better than average fund'? And if I do a modicum of research on an index fund, I have a very good chance that it will perform in a very tight band around its benchmark, trailing a bit by the small fees/expenses most of these have, and a bit up/down depending on how closely they can replicate the benchmark. It won't be a random pick, it will be an educated one based on information, not a Crystal Ball.

 

[Stephen Tashi]

Did I say it was "sound"? No. I only said "it was". So for you to have a different assessment than "the market" would indicate you know something the others don't. I don't think that is likely (and it may be illegal) on a consistent basis. But I do think it is possible from time to time.

I agree if we are talking about "you know" meaning me, since I am a human being. There is difference between my knowing something other people don't know because I have illegal inside information versus my knowing something because I have taken the trouble to ferret out important facts from a mass of publicly available data that others have not had time to examine. A significant significant advantage that computers have is being able to apply search procedures to large volumes of data. I think some investor's who make the current market already use sophisticated computer programs in the typical person-to-machine interface of mouse and keyboard - computer summarizes data, human analyzes the summary. I'm curious what will happen if many firms begin taking the human out of the loop.

(It interesting to me that many people who don't accept consensus opinions about politics, nuclear power, global warming etc, are willing to defer to the consensus opinion of "the market" about stock prices. I haven't followed your (NTL2009) posts on diverse issues, so I don't know if you are such a maverick. I myself am not such a maverick that I'm determined to ignore consensus opinions on all subjects, but when it comes to "the market" I'll be a little skeptical.

 

[russ_watters]

Succinctly: It is a myth that managed funds do better than index funds. (this should be added to the financial knowledge thread). They don't. In fact, almost all of them do worse.

http://www.marketwatch.com/story/wh...-funds-beat-the-sp-than-we-thought-2017-04-24

The performance of actively managed funds is so bad that just by luck, you would expect half of them to beat the S&P and 5% basically means they are all bad.

My point and my source, but I'm feeling like this shouldn't be possible. By definition of the word "average", half should do better and half should do worse (minor quibble about median vs mean). The fees charged should skew to somewhat less than half doing better, but 5% sounds impossibly low.

I wonder if when they factored-out the selection/success bias in managed funds, they ignored the exact same bias in indexes?

Whatever: even if the true answer is closer to 50% it doesn't change the point: managed funds are on the whole like going to a casino: the house always wins and you lose, even when you win.

 

[Stephen Tashi]

No again. Look some more into the studies of actively manged funds. Very few of those 5% that beat the market in a five year period, are able to repeat the performance in the next 5 year period (which is what matters to us).

We can look at statistics on active-human-managed investment funds versus human-managed index funds and perhaps computer-managed-index funds. As far as I know there are yet no long term active-computer-managed investment funds (is "Betterment" one of them?) , so I don't know what statistics we'd use to compare their performance.

Isn't (or wasn't) the conventional wisdom that statistics prove that short term (human) traders have poor performance? HFT added a new wrinkle to that situation.

 

[russ_watters]

It's an interesting assumption that the consensus reached by "the market" implicitly accounts for all the relevant information. Why is that assumption sound?

To say what @NTL2009 said in a different way: it isn't necessarily "sound", it is circular. The value is what it is because it takes certain information into account and in doing so decides/makes that information the relevant information.

For example an investor with a million dollars to invest in a stock can have more affect on its price than an investor with half that amount. Is the investor with more money twice as wise?

No, but he can change what information is relevant/dominant in affecting the price. I think you think this fact is important/relevant/meaninful/dangerous, but it isn't. It's all in there already.

I agree with the argument that by the time the average human investor had realized there is a change in the fundamentals of a stock then the price of the stock will probably already have changed. However, speedy investors or speedy computer programs might realize things quicker than the consensus of the market.

So I think we largely agree that short term trading is a gambling game where some can win and others lose. But I think you are still under the mistaken impression that this is the same game long-term investors are playing and can affect that game. It isn't/can't.