Down goes the Dow

Why are you avoiding the question? Did Portia plagiarize, or not?

wondering wrote:

Why are you avoiding the question? Did Portia plagiarize, or not?

We are avoiding your trolling nonsense. We know who you are.

Ok, fine. I assume your plagiarizing accusation was also trolling. Funny.

So a bit more on the "randomness" of the markets. Here are ten random market charts, each generated by a random shuffling of the actual daily changes in S&P 500 since 1950, with the actual chart shown in red:

[IMG]

http://i67.tinypic.com/dmxyxt.jpg[/IMG

]

Couple of things to notice:

- they all end around the same place. That's because I've used the same set of % changes. If you shuffle the order, their total effect is nearly the same. If I instead used an extreme value distribution that closely matched the real data, and the high and low outliers were different, the end point would be different

- each curve has long bear and bull markets, generated entirely randomly

The curves are a little hard to visualize with index values on a natural (linear) scale, and make a bit better sense on a log chart:

[IMG]

http://i66.tinypic.com/2wgagew.jpg[/IMG

]

Again, the red curve is actual. You can see that each random curve has generally similar character as the real curve. There are some important differences:

- you can't see it at this scale, but the random curves have more volatility on a daily scale

- the random curves are less constrained. They don't tend to stray much higher than the real curve (relative to an even trajectory, best fit of the real curve), but many of them go much lower than actual.

As noted earlier, if I loosened the constraints on the random curves a bit, using simulated (random) data rather than randomly reshuffled real data, we would probably see the curves wander much higher and lower; they would almost certainly not end at the current index value.

That's still not the stuff that piqued my interest...

Here is the S&P 500 since 1950 with a smooth best fit trend:

[IMG]

http://i66.tinypic.com/1z3v7ls.jpg[/IMG

]

You can draw bounds around the variation of actual index values from the best fit trend:

[IMG]

http://i68.tinypic.com/23ubdpt.jpg[/IMG

]

Here I've used 98th and 2nd percentiles (i.e. 2% of days lie above the upper line, and 2% of days lie below the lower line. If we believe this long term trend to be meaningful, and that it's upper and lower bounds constrain future index values, then we get the following forward-looking bounds (switching now from the log scale back to the more intuitive natural scale), using the absolute (highest and lowest) bounds instead of 98% and 2%:

[IMG]

http://i64.tinypic.com/s1n66g.jpg[/IMG

]

If we believe the past market behaviour has predictive power looking forward, then the lowest possible index value over the next year is about 1340, or about 3250 by 2030, and the highest possible values over the next year and by 2030 are 5850 and 14300. Median estimates for the end of 2018 and 2030 are 2935 and 7160.

Now I'll admit that even though the past pattern appears fairly compelling, I'm not necessarily convinced the future is bound by the past. Recall, if we miss only a few of the biggest or smallest daily drops, the whole chart changes. So either god (or something) is working to balance the big and small outliers to preserve an overall trend, or else, ell I guess or else she isn't... :-)

If instead we look at the forward-looking change for any given day, looking ahead to different time intervals (i.e. anchoring the projection on the current value, rather than constraining it by the overall trend), we get:

[IMG]

http://i64.tinypic.com/eg1pog.jpg[/IMG

]

This gives us a somewhat lower possible low, at about 1250, 2-years out.

So, while it seems unlikely we should consider past market behavior as a good indicator of future market behaviour, if we decide to, we can look at about 1250 to 1350 as a worst case bottom. However, it could persist for many years. Statistically speaking, of course. But the median projection, whether by considering the current index value, or its position within the historical trend, is for about 7-8% annual growth per year (on average), with equal probability of being better than that, or worse than that.

Albert Edwards has a lower bound:

https://www.zerohedge.com/news/2018-04-08/albert-edwards-next-recession-sp-will-drop-below-666

Ghost of Igloi wrote:Albert Edwards has a lower bound:

I understand. It is my belief that the fundamental distortions on the upside have been so enormous ($21 Trillion in global QE from central banks) that the lower bound of your range is quite likely, while the upper bound very unlikely.

Ghost of Igloi wrote:

I understand. It is my belief that the fundamental distortions on the upside have been so enormous ($21 Trillion in global QE from central banks) that the lower bound of your range is quite likely, while the upper bound very unlikely.

But then you also thought it unlikely that we’d reach our current record highs.

It may be more natural or intuitive to look at the two "predictions" with index value on a log scale:

[IMG]

http://i65.tinypic.com/2qaljjr.jpg[/IMG

]

The two parallel lines (light blue on top and dark green below) represent projections based on the overall trend and absolute bounds of past index values. The funkier curves represent projections from the current index value, independent of the overall trend, based on the complete set of forward changes from every other day in 70 years. Depending on how you prefer your poison (shaken versus stirred) you might like one projection over the other. Or if you're an idiot like me, they may both leave you stupefied, with no better basis for a decision than before... :-)

J. Hardy wrote:

Ghost of Igloi wrote:

I understand. It is my belief that the fundamental distortions on the upside have been so enormous ($21 Trillion in global QE from central banks) that the lower bound of your range is quite likely, while the upper bound very unlikely.

But then you also thought it unlikely that we’d reach our current record highs.

Perhaps, the view was/is it doesn’t matter, since for most all gains going back to 1998 will be wipped out.

Ghost of Igloi wrote:

J. Hardy wrote:

But then you also thought it unlikely that we’d reach our current record highs.

Perhaps, the view was/is it doesn’t matter, since for most all gains going back to 1998 will be wipped out.

Or it doesn’t matter because any losses will eventually be recovered and more.

J. Hardy wrote:

Ghost of Igloi wrote:

Perhaps, the view was/is it doesn’t matter, since for most all gains going back to 1998 will be wipped out.

Or it doesn’t matter because any losses will eventually be recovered and more.

Yes, in 10-12 years.

Igy, you keep saying "will." In the words of Inigo Montoya, I'm not sure you know what that means... :-)

One last burst from me. There's a lot of discussion about the Shiller CAPE, a measure of 10-year trailing PE. Because it's been high at the time of other market crashes, I guess we're supposed to loo at high values as an indication of an impending crash. Here's his index, showing January values from 1950 onward:

[IMG]

http://i67.tinypic.com/anyq8.jpg[/IMG

]

Like I did with the "death cross," I wondered whether this metric has any forward-prediction power. I also looked at whether it was a useful hind-casting tool, assuming it ought to be better at that, since it incorporates trailing data. Here are plots for 6-month, 2-year and 10-year leading (prior) change, with best fits and R^2 values:

[IMG]

http://i66.tinypic.com/xf9h8g.jpg[/IMG

]

There is some (weak) correlation with leading 10-year change, and essentially no useful link with prior 2-year and 6-month change.

Here's what the plots look like if we look forward from given values of Shiller's CAPE:

[IMG]

http://i67.tinypic.com/24wbig4.jpg[/IMG

]

While the best fits look appealing, the R^2 values and the obvious scatter tell us there is only a very weak link between Shiller CAPE and forward change in the index for any meaningful period.

This got me curious, so I dug a little deeper. First, I looked back at the market's actual behaviour and best fit:

[IMG]

http://i67.tinypic.com/zv78ud.jpg[/IMG

]

Then I generated a ratio of actual values to best fit (black curve below):

[IMG]

http://i68.tinypic.com/29dfk1e.jpg[/IMG

]

This is effectively normalizing the chart to its long term trend. Where the black curve is above the straight line, the actual chart (blue curve) lies above its long term trend.

I can plot Shiller's CAPE along with this ratio:

[IMG]

http://i66.tinypic.com/2edy5ol.jpg[/IMG

]

You can see the CAPE value has similar overall pattern, but rises over time. If I normalize the Shiller CAPE to its long term trend, creating an "Idiot Edjusted Shiller CAPE" (or I-E-S-CAPE), I get this:

[IMG]

http://i66.tinypic.com/2edy5ol.jpg[/IMG

]

Here, we can see the I-E-S-CAPE follows the normalized market very well. From this, I deduce two things:

- Shiller's CAPE has a long term upward trend, suggesting to me that a value of 28 today is equivalent to a value of12 in 1950

- normalized for its long-term trend, the I-E-S-CAPE does a decent job of following the market

- like Shiller's unmodified CAPE, the I-E-S-CAPE has relatively weak hind-casting power, and very weak forecasting power.

In sum, I think the Shiller CAPE does a pretty good job of telling us something about what the market is doing today, if we read it right, but has little to bo ability to tell us very much about what the markets will look like in years to come.

OK, done, finito!

Damn, messed up... here is the correct "I-E-S-CAPE" plot (I think):

[IMG]

http://i68.tinypic.com/2rylyrp.jpg[/IMG

]

nice. And all we have to do is look at the thing...it says we should have been out/wary of stocks since the mid 90s. With one short exception I think. Meanwhile the market has soared. It's not a predictor.

Of course that assumption is made at record high valuations and ignores almost all fundamental metrics. That is true of idiot’s charting as well. One thing is for sure the character of the market is quite a bit different from last year.

Your conclusion mimics what has been said since the the inception of the CAPE. That is, it is not a good predictive tool.

I enjoy reading your analysis.