I like working with Yahoo! data because I can set Yahoo! Data>Perform Split Adjustment to keep the prices constant without having to completely reload the data (and Perform Dividend Adjustment to do the same thing in the case of extraordinary dividends).

However, I would prefer that the adjustment NOT adjust the volume. In my experience, the volume does not change in proportion to the split/dividend. For example, Citigroup (C) had a reverse split of 1:10 on 5/9/11. Obviously, without performing the Split Adjustment, it looks like there was a 1000% increase in the price of the stock that day. Performing the Split Adjustment eliminates that, as it is supposed to.

The problem is that there was no significant change in the average volume before and after the split. Performing the Split Adjustment inflates the volume prior to the split by a factor of 10, making it look like there was a huge drop in volume after the split. Obviously, the reverse is true for a 10:1 split.

The practical problem is that if you have a strategy which utilizes price and volume, you have the unpleasant choice of either having the price giving incorrect signals or the volume giving incorrect signals.

QUOTE:
The problem is that there was no significant change in the average volume before and after the split.

Are you sure? The day before split the volume dropped 5 times. And the week after the split it increased 1,5-2 times compared to the week before. Maybe it was just a side effect of that reverse split because by comparing it with some other reverse splits, I see no problem at all. Could you point me at more examples illustrating the issue?

Rather than debating which is better, I'll just "roll my own". I assume the following is essentially what happens when Perform Dividend Adjustment and Perform Split Adjustment are checked, in which case, I can just insert a streamlined version of this (without the debug) into my Strategies.

I would appreciate any comments regarding whether I am making the adjustments correctly or suggestions for improving the code.

CODE:
Please log in to see this code.

It's likely an isolated case that you were observing. So-called "Splits" in Yahoo! can be misleading and many times represent a dilution during M&A activity, in which case the price usually requires another adjustment that Y! does not provide. These are cases when Y! data are unusable for backtesting (read: cliffs) and a "roll-your-on" solution (or simply a special adjustment for a particular stock) may be required.

I want to have the best back-testing possible. Therefore, I wrote the following Strategy, to determine whether it is better to adjust or not adjust the Yahoo! volume for stock splits.

Essentially what it is supposed to do is look at the 30 day average volume before and after the split. My assumption is that the "best" decision is the one that comes closest to having the average volume before and after the split be the same. This should result in the least amount of volume change after a split.

The key output is the last two numbers. The very last number is simply the ratio of the total of the previous volumes to the total of the subsequent volumes. The comparison is the total of the previous volumes (adjusted for the average split factor) to the total of the subsequent volumes. Please let me know if there is something wrong with my methodology or code.

I ran this on the Russell 1000 for All Data (which is truly all data available from Yahoo). Of the more than 2000 splits, adjusting the volume resulted in a ratio of previous to subsequent volume of 0.57. The unadjusted ratio was 1.15. Running it against the Russell 2000 yielded similar results -- 0.76 adjusted and 1.11 unadjusted.

This tells me that not adjusting the Yahoo! volume is more likely to give me the proper volume for back-testing. Obviously, the results may be different for specific stocks and specific splits, but over a large number of symbols and splits, it makes more sense to me not to adjust the Yahoo! volume.

I welcome opinions on this, especially Eugene and Cone.

CODE:
Please log in to see this code.

I'm closing this tread because after looking at specific symbols, I see that Yahoo! sometimes adjusts the pre-split volume and sometimes they don't. I think Cone is correct that if you're going to work with Yahoo! data, there is no way to solve the volume issue unless you do it symbol by symbol.

This is more fully explained in my request for on update on the best data provider.