bostonbubble.com

[GenXer]

[Kaidran]

Technically you did say "PM" and emailing on a pm request seems like bad forum etiquette to me.

Anyway, I guess I wrote everything I wanted to ask in the post above anyway.

[GenXer]

Short answer is that mean reversion doesn't exist (and as a result, prices are 'maintained' by other means). The long answer would be explaining why that would be the case. Otherwise, that's the answer

Also, another part of the answer is that you have to be careful about drawing inferences from averages. Average of 1M and 500k is 750k, if that tells you anything (i.e. an average for a 'phantom' house that doesn't exist except as a statistic - there may be a bunch of 1M houses and a bunch of 500k houses). Median is similar. A 'median' house also doesn't exist.

Also, just because you know a median salary doesn't tell you what the 'effective' salary is - some people can survive on 20k a year, and some need the entire 100k. Lots of 'high order of magnitude' aspects to this that are not trivial. One has to always be careful about the assumptions - otherwise the results may not make much sense.

[Kaidran]

That does make sense, though is a little depressing.

[CL]

GenXer - saying mean reversion does not exist is quite a sweeping statement and I don't fully agree, so would like to hear your long answer.

And while mean/average can be "phantom", median by definition has to be one of the data points of the data sample/population. So a median house does exist. How much inference/implication you can draw from it is, of course, up to (endless) debate.

[admin]

[GenXer]

True, a median house does physically exists, my mistake.

I will reverse that on you. Show me one paper that proves mean reversion exists, for example, in the stock market. I have several references to the contrary, all by reputable researchers.

Think about this. Take a simple coin tossing game. Does mean reversion occur? No, not by a long shot. If you run 1000 scenarios, your expected 'return' would be 0, true. But for any one sample run, your return could be anything, again, bounded by the Gaussian sqrt(t).

This is everything you ever need to know about mean reversion:

http://homepage.mac.com/j.norstad/finance/rtm-and-forecasting.html

Bottom line: nobody has ever seen it enough to make a profit on it. Don't assume something that may not be there, especially when you are betting money on it.

[GenXer]

[CL]

Apologies to everyone for getting side-track

GenXer - if you believe the stock market is 100% efficient, then yes mean reversion will not exist since the market will not deviate from instrinic value, and thus the future return is random. But if you believe market is 100% efficient, then market bubble cannot exist neither. We all believe bubble exists, I assume.

One evidence of mean reversion exist in stock market is the performance of value indicators (Price to Earnings, Price to Book, etc). Value indicators essentially is betting on market deviate on their intrinsic value, and revert to mean/normal. Long term, it has been proved to add value. It won't add value 100% of the time, but high enough hit rate for it to be profitable. You can look up Fama French research papers in mid-90s for reference.

[GenXer]