Can we use the stock market itself as a useful prediction market in any way? For example can we get useful information about how long Moore’s law type growth in microprocessors will likely continue based on how much the market values certain companies? Or are there too many auxiliary factors, so that reverse engineering anything interesting from price information is hopeless?
Yes, look at the ratio of a stock’s price to it’s current earnings to get a good prediction of how the company’s earnings will change over the next decade.
I don’t think I understand how that is consistent with EMH. You mean that P:E predicts high variance but the mean is still priced in correctly so one can’t make a profit just by looking at P:E? A high P:E meaning something like the market saying ‘this company’s earnings will either go way up or way down but I don’t know which’?
Imagine that the stock prices of companies A and B were equal. Last year Company A had low earnings per share while company B had high earnings per share. EMH implies that the market expects that in the future Company A’s earnings will increase relative to Company B’s earnings.
EMH implies that stock prices, not profits, follow random walks.
Say both of our businesses made $1 million this year. Everyone expects your profits to increase but mine to decline. We have the same number of shares of stock outstanding. Your company’s stock price will be a lot higher than mine, giving you a much greater price/earnings ratio reflecting the market’s expectations of increases in your future profits.
Do you want a citation for that P/E ratios reflect market expectations about future value (and thus earnings), or do you want a citation for the claim that the market is good at predicting what earnings will be in the future?
The claim that the market is good at predicting future earnings.
It probably is, but economics does not yet have the empirical grounding to give me high confidence in its theories (the way I would be for fields like physics or chemistry; I still think economic theory has a strongly positive correlation with reality).
The claim that the market is good at predicting future earnings.
This is conceptually very easy to test; get historical stock price and earnings data, compute P/E ratios at each snapshot, compute earnings growth across snapshots, and then look at the relationship between the two. Vanguard ran the numbers here (page 7), and two ways of calculating the P/E ratio were the strongest two factors. (As one would expect, 1-year returns were very difficult to predict at all, and they were mostly useful for 10-year returns.)
If I’ve understood that document correctly, they aren’t saying anything about predicting performance of individual companies, they’re looking at some sort of averaged P/E ratio and relating it to future performance of the stock market as a whole.
One way for that to work would be for individual stocks’ P/E ratio to be indicative of their future performance, but there are others; e.g., maybe overall economic conditions influence both investors’ attitudes and future performance. In the latter case, P/E ratio might be less useful (or outright useless) for comparing companies at a single time.
Agreed. I went with the first Google search result that was at all close to the question at hand; I recommend anyone more interested in the subject collect the data and run the numbers themselves.
In particular, one might want to compare residual P/E ratios—that company’s P/E minus the S&P 500′s P/E or the total stock market P/E—to future earnings growth in order to try to remove some of the time-dependent effects and specifically judge the market’s ability to guess the earnings growth of individual companies.
One could, if they knew historical industrial groupings, judge the market’s ability to price the overall market, industries, and individual companies. It seems like we would expect the first to be better than the second, which is itself better than the third. This is both for the raw statistical reason that the sample being averaged over is smaller as we go down, making the range of reasonable numbers larger, and the financial reason that forces on larger scales may be more visible or predictable than forces on smaller scales.
No, companies are too complicated. To address your particular example, Intel has three advantages over its competitors. One is that it pushes the boundary of Moore’s law, the one point you are interested in. But the other two are that it is probably the best in the world at arranging transistors and that it has close to a monopoly on the 386 architecture, and thus inertia. Even if Moore’s law did break down, it is not clear how long it would take for the competitors to catch up.
Can we use the stock market itself as a useful prediction market in any way? For example can we get useful information about how long Moore’s law type growth in microprocessors will likely continue based on how much the market values certain companies? Or are there too many auxiliary factors, so that reverse engineering anything interesting from price information is hopeless?
Yes, look at the ratio of a stock’s price to it’s current earnings to get a good prediction of how the company’s earnings will change over the next decade.
I don’t think I understand how that is consistent with EMH. You mean that P:E predicts high variance but the mean is still priced in correctly so one can’t make a profit just by looking at P:E? A high P:E meaning something like the market saying ‘this company’s earnings will either go way up or way down but I don’t know which’?
Imagine that the stock prices of companies A and B were equal. Last year Company A had low earnings per share while company B had high earnings per share. EMH implies that the market expects that in the future Company A’s earnings will increase relative to Company B’s earnings.
EMH implies that stock prices, not profits, follow random walks.
Say both of our businesses made $1 million this year. Everyone expects your profits to increase but mine to decline. We have the same number of shares of stock outstanding. Your company’s stock price will be a lot higher than mine, giving you a much greater price/earnings ratio reflecting the market’s expectations of increases in your future profits.
Really? Given the massive bubble in tech stocks and the US stock market in general?
It’s very hard to know if you are in a bubble.
Citation needed.
Do you want a citation for that P/E ratios reflect market expectations about future value (and thus earnings), or do you want a citation for the claim that the market is good at predicting what earnings will be in the future?
The claim that the market is good at predicting future earnings.
It probably is, but economics does not yet have the empirical grounding to give me high confidence in its theories (the way I would be for fields like physics or chemistry; I still think economic theory has a strongly positive correlation with reality).
This is conceptually very easy to test; get historical stock price and earnings data, compute P/E ratios at each snapshot, compute earnings growth across snapshots, and then look at the relationship between the two. Vanguard ran the numbers here (page 7), and two ways of calculating the P/E ratio were the strongest two factors. (As one would expect, 1-year returns were very difficult to predict at all, and they were mostly useful for 10-year returns.)
If I’ve understood that document correctly, they aren’t saying anything about predicting performance of individual companies, they’re looking at some sort of averaged P/E ratio and relating it to future performance of the stock market as a whole.
One way for that to work would be for individual stocks’ P/E ratio to be indicative of their future performance, but there are others; e.g., maybe overall economic conditions influence both investors’ attitudes and future performance. In the latter case, P/E ratio might be less useful (or outright useless) for comparing companies at a single time.
Agreed. I went with the first Google search result that was at all close to the question at hand; I recommend anyone more interested in the subject collect the data and run the numbers themselves.
In particular, one might want to compare residual P/E ratios—that company’s P/E minus the S&P 500′s P/E or the total stock market P/E—to future earnings growth in order to try to remove some of the time-dependent effects and specifically judge the market’s ability to guess the earnings growth of individual companies.
One could, if they knew historical industrial groupings, judge the market’s ability to price the overall market, industries, and individual companies. It seems like we would expect the first to be better than the second, which is itself better than the third. This is both for the raw statistical reason that the sample being averaged over is smaller as we go down, making the range of reasonable numbers larger, and the financial reason that forces on larger scales may be more visible or predictable than forces on smaller scales.
That’s very useful info. Thanks for the link.
Here you go:
Yes, of course, though it depends on what you consider “useful”.
No, companies are too complicated. To address your particular example, Intel has three advantages over its competitors. One is that it pushes the boundary of Moore’s law, the one point you are interested in. But the other two are that it is probably the best in the world at arranging transistors and that it has close to a monopoly on the 386 architecture, and thus inertia. Even if Moore’s law did break down, it is not clear how long it would take for the competitors to catch up.