Tuesday, August 19, 2008

Investor Sentiment that works

Ronald Domingues published a brief summary of an article drawn from his Masters of Economics entitled "Finance Theory and the Investor Sentiment". His remit:

By merely calculating and make investor sentiment survey indicators public, we are going against efficient market hypothesis (EMH) which considers that, in any time, asset prices reflect all available information. Taking into consideration that we cannot predict upcoming information, market practitioners should not expect future p rise or fall. Even if any person decides to make his/her expectations public, they should not be taken into consideration in forecasting asset price changes.

His G100 Sentiment survey looks to tackle the three main weaknesses of direct sentiment surveys:

  • Both time and discipline needed to fill out surveys
  • Results are obsolete when released
  • Potential conflicts of interest

    One of the key features of Ronald's sentiment survey is the need to retain high scores to remain part of the survey sample. A second pool of participants, a resource group as such, provide the promotion material to the survey group as the weakest of the survey group are relegated to the resource pool. The shuffle takes place at monthly intervals. The following table shows the comparison between direct sentiment measures (as provided by Ronald)

    Participants are asked whether markets are bullish (greater than 0.1%), neutral (+ to - 0.1%) or bearish (loss of 0.1% or more) for the S&P the next day. Emails of the questionnaire are sent as soon as the market closes and results of the survey are released prior to the market open. His full list of rules can be obtained here.

    Here is his most recent ranking (I'm ranked in 15th place based on performance over the past month). You can see the four set for the chopping block:

    And five due for promotion if they can hold their position come reconstitution time:

    But the meat in the pie is the performance of the sentiment indicator. Based on results since April 11th of this year the G100 had a 65.2% success rate in calling the S&P.

    The TickerSense Blogger Sentiment Poll is similar to the G100 in it provides a time dependent survey of S&P performance, but over 30-days instead of the next day. However, unlike the G100 there is no qualification or internal ranking of TickerSense participants. In addition, CXO Advisory found:

    ...analysis of Ticker Sense Blogger Sentiment Poll results suggests that aggregate blogger sentiment is largely non-predictive for future stock market direction.

    But perhaps all that is missing from the Ticker Sense Poll is a qualification filter. That gives me another idea for a blog article....

    Dr. Declan Fallon, Senior Market Technician, Zignals.com the free stock alerts, market alerts, and stock charts website