Talk:Random walk hypothesis

Clarification required in introductory paragraph
The introduction should specify that the so-called "hypothesis" is in fact not really a hypothesis, but a modelling assumption.

Everyone in finance knows that the data does not follow a random walk. Random walk models, however, simplify mathematical analyses, and may furthermore yield useful insights.

Compare with the so-called "Laws" of Thermodynamics (which are actually probabilistic outcomes), or the "Theory" of Relativity, which is well-supported by data, for precedents in which the conventional language misrepresents the role played by the model. —Preceding unsigned comment added by 211.28.54.169 (talk) 00:43, 20 April 2009 (UTC)


 * I am looking forward to how you want to use data analysis to prove that something is or is not random. — Preceding unsigned comment added by 119.228.239.82 (talk) 13:36, 10 April 2017 (UTC)

Repetitive sections
Please note that the sections "A non-random walk hypothesis" and "Random walk hypothesis vs. market trends" essentially repeat each other, and should be merged. Askari Mark (Talk) 05:04, 1 February 2007 (UTC)

Some comments and questions
1. How has it been shown that the financial markets are a random walk? Was there a study done that showed no correlation between past and future price movement?

2. I'd like some clarification of the hypothesis: does it refer only to the unpredictability of a single subsequent price from the immediately preceding one? What about price X time steps into the future?

3. A comment on "Testing the Hypothesis:" The first example (coin toss) is surely not a test of the hypothesis. It is only a test of the "chartist," it says nothing about financial markets.

Mwtzz (talk) 20:49, 8 January 2008 (UTC)

4. The test for the chartist cannot prove that price is random. Someone who believes stock prices are random should testify that stock prices do not randomly move is wrong. Yao Zhou, Sep 16, 2010 (IIT) —Preceding unsigned comment added by 24.148.75.11 (talk) 03:25, 17 September 2010 (UTC)

Some notes
The NBA example is truly unneccessary to include. Clearly whether stock markets can be modelled as a random walk has nothing to do with whether basket ball shots can be so modelled, and the example hardly illuminates anything.

Also it might be interesting to note that volatility tends to cluster. (See eg the ARCH model.) This is not compatible with a random walk but does not mean that prices are predictable. —Preceding unsigned comment added by 65.95.27.133 (talk) 17:01, 7 November 2008 (UTC)

Removing it now. Bazuz (talk) 19:17, 2 April 2017 (UTC)

Robert Hall's random walk hypothesis
There's another hypothesis with the same name proposed by the Robert Hall. It concerns the consumption function. Should we have a new page for that or just include it here? __earth (Talk) 05:28, 22 February 2009 (UTC)

=
It appears that finance economists aren't interested in contributing for free to a public good. This article is of low quality. There is in principle no direct test of whether a set of data is generated randomly. One can only test whether the data fits given nonrandom patterns, one at a time; or whether some group of people (chartists, for example) have demonstrated success making predictions. Makiel's book discusses such tests, not just his classroom experiment.

Argument against Weber
Weber's discovery that well perfoming companies in first 5 years of a 10 year observation period will underperform later. This can be supported by Daniel Kahneman's explanation in Thinking, Fast and Slow on regression to the mean. The explanation is that many people do not realize that a distribution of outcomes will regress to the mean. For example: an air force pilot will not fly equally well all the time, but will have a distribution about his mean skill, sometimes better, sometimes worse. However, the commanding officer will react to the immediate observation that the pilot is flying poorly and reprimand him, and observe an "improvement" the next flight. Similarly, a pilot flying better than usual may be praised, but then will perform worse the next flight. Both of these are due simply to regression to the mean, but the officer believes it is the scolding that improved behaviour, and praise that worsens it, so will dispense more scoldings and withhold praise. Analogously, companies performing better now will perform worse later, and vice versa, due simply to regresion to the mean performance of the company. 128.174.229.198 (talk) 02:45, 28 November 2012 (UTC)


 * The Reification of Regression to the Mean is a logical error. "Regression to the Mean" is not a thing.  It does not act.  It cannot cause, nor explain anything.  Regression to the mean is simply an observation about many statistical compilations of past events.  Some distributions will continue to show it.  Some won't.
 * David Lloyd-Jones (talk) 12:52, 23 April 2017 (UTC)

Testing the hypothesis
"Testing the hypothesis" section has nothing to do with thesting the RWH and is just an anectode. It should rather contain a summarry of test's performed pre-EMH (Cowles ; Kendall ), a short description of Bachelier-Osborne's and Mandelbrot's definition of the Random Walk, Fama's findings. Further, we've got a number of studies reporting predictability of stock returns: Rozeff & Kinney ; French & Roll and variance ratio test (Lo & McKinley ) mentioned in the subsequent section, that should be a part of this one (since the alternative hypothesis to RW is obviously non-random walk). And after Brock et al. we have a real boom of studies providing evidence that simple chartist rule are working. Zirr (talk) 21:14, 3 July 2019 (UTC)