Polling Sensitivity to the DOW
September 19, 2008 by MTM
In stock market forums like Finance Manila's Trader's Lounge, emotions run high depending on how the market moves. One driver of sentiment is the performance of the US markets the previous night, specifically the DOW. Although I've done some study of the correlation of the DOW with the P6 and found that there is a positive correlation between the two in general:
(see more of this study here)
However, questions were posed on the forum of whether traders were really that sensitive to the DOW. Furthermore, was there a bias to traders in their sensitivity, or were people only looking for an UP DOW to satisfy their own innate bullishness.
I ran a small poll to address this in Finance Manila, and my questions were:
- If the DOW is up last night:
I am bullish today.
I am bearish today.
I am indifferent today. - If the DOW is down last night:
I am bullish today.
I am bearish today.
I am indifferent today.
Members were only allowed to select any two answers (one for DOW UP and DOWN respectively).
Here were the results:
To test for bias, we merge the Bullish and Bearish votes into simply "Affected"–and check the indifference level for each of the individual DOW UP and DOWN scenarios.
Here is the revised table:
We can then convert the results into percentage totals as follows:
This allows us to graph the results for better appreciation:
From an eyeball analysis, neither the DOW UP or DOW DOWN scenarios appear much different from the total sample. The sensitivity of respondents is about 2 out of 3, and conversely, the level of indifference remains constant throughout.
We can test the differences between the scenarios in 3 ways. First, a Chi-square test of significance shows that the results are significant to a .65 sensitivity which is not very significant (i.e. 2 out of 3 odds of being due to randomness).
Secondly, we can run a Population Stability Index using the logistic regression formula:
PSI = ln( actual % / benchmark % ) x ( actual % – benchmark % ) which results in the following:
For PSI to signify an instability or difference between two populations, a level of .25 or greater should be registered. In this case, the 3 populations (DOW UP and DOWN, and TOTAL) are practically identical.
Finally we can track Population Shifts between the 3 columns:
This shows a very minimal (5% or less) shift between the categories and against the total. To test bullishness or bearishness, we revert to the original categories, and keep I'M BULLISH and I'M BEARISH separate from each other. We then compute for percentages as follows:
Which allows us the graphical representation:
An eyeball analysis shows that for the total sample, there is practically an equal percentage of BULLISH, BEARISH and INDIFFERENT populations, but for the individual scenarios, these percentages do vary significantly.
A Chi-square signficance test shows the results to be significant to the 0.00000000769 level which is very significant or a 1 in 129 million odds of being due to randomness. Even if we halve the result that's still a 1 in 64 million odds.
The Population Stability Index using logistic regression (see above for formula) yields the following results:
A PSI of .69 and .64 for DOW UP and DOWN against the Total population shows significant instability or difference. A PSI of 2.67 for DOW UP vs. DOWN is even more signficant. Also noteworthy is the instability of the OPPOSITE vote population relative to the scenario i.e. Bullish votes when DOW is DOWN and vice-versa.
Tracking Population Shift:
This effectively shows that the bullish and bearish populations invert depending on the DOW result.
A final comparison was made with a similar poll done a year ago where respondents were asked whether or not they monitored the DOW more often lately. If we make a qualified assumption that those who said NO is equivalent to indifference, we can make a comparison with this latest result:
And display the percentages graphically for better appreciation:
This implies that while the sensitivity to the DOW remains high from last year, that sensitivity has lessened lately with more people claiming indifference to the DOW.
Inference and Application
Although I posed these same questions back on the forum page awaiting some reactions from FMers, I already have some ideas of why the results ended up the way they did.
Firstly, the level of correlation of the DOW to the P6 does translate or go hand-in-hand with the sensitivity of traders to the DOW. In fact the reason the correlation is positive is probably because of the sensitivity factor.
Secondly, why the level of sensitivity would drop or indifference would rise has probably something to do with the events that have happened since the first poll. The original poll was done in August 2007–when the markets plunged during the disclosure of subprime losses amongst US and British banks. Since then markets have been eroding in value up to September 2008 which is the worst performing month so far. This implies that people have either given up on their positions or liquidated them in the past year.
However, one thing that is interesting is that despite the rise in indifference, this has not in anyway diminished the correlation of the DOW and PSEi–and according to the previous study, the correlation has in fact strengthened over recent months.
whoaaaaaaaaaah!