I am having trouble understanding how the Y! dividend adjustment option is applied. I have found a case where a stock split and dividend fall on the same date and am struggling to work out how, even after setting split and dividend adjustments, the share price falls approx 97% overnight.

Please refer to KSU on 7/13/2000.
With split and dividend adj set on the Y! provider. Close price on the 7/12/2000 was 96.70. Opened at 5.38 on 7/13/2000. There was an dividend of 90.80. 2:1 split.
I have been playing around with the Wealth Lab formula used to calculate dividend adjustment factor (1-(dividend/prevClose)) and it looks to me that, when a split and dividend fall on the same date, the wrong variable is being passed to the formula.

It looks to me when div adjustment is the only option selected it is being calculated using the split adjusted price which is then applied to the raw data. ie.
1-(90.8/187.5) = 0.515 applied to raw close of 93.75 = 48.35.

Selecting the div and split adj option results in 96.7 (48.35 * 2).

If the div adjustment was calculated from the raw data then we would get:
1-(90.8/93.75) = 0.0315 applied to raw close = 2.95 adjusting for split = 5.9 which seems more realistic to the next open of 5.38.

As I said, with some playing around it looks like the dividend and split falling on the same date has created a calculation error. Some clarification would be appreciated.

Thanks

Even though I think you're looking at it correctly, the stock split was 1:2, not 2:1 (shares were reduced, not added).

If you look at the chart without any adjustments, the close on 7/12/2000 is 93.75. So, applying only the 1:2 split, that price would be 93.75 * 2 = 187.5. Subsequently applying the dividend adjustment (1 - (90.8/187.5) = 0.5157), you get 0.5157 * 187.5 = 96.7. And that's the value displayed on the chart with splits and dividends applied, and, it's the Y! adjusted close value as well, Check it out.

Sorry, you are right about about the 1:2 split. The calculation makes sense (thanks)... just not so sure about the data??

Cone if the dividend adjustment was applied before the stock split is factored in the result would be a resonable value.
1-(90.8/93.75) = 0.0315 * 93.75 = 2.95
Adj for split = 2.95 * 2 = 5.9

I see where you're going with this, but right or wrong, a dividend adjustment of that magnitude could create a lot of havoc with backtesting prior to the dividend in Version 5 since limit order stock trades are rounded to 2 decimals.

If someone else doesn't know which method is correct (and why) we should research it further - and find out why Y! back-adjusts this way too. Of course, there are even more ways to back-adjust, as gbeltrame described here.

Just thinking about what's actually happening with this dividend and reverse split, the company is returning 97% of its value to shareholders and retiring half the shares. When you look backwards in a chart the company was "worth more" than it is worth in the present day. The value of the remaining shares reflect that "loss" in value, though I understand that from a shareholder perspective, it doesn't represent a loss - though they're forced to realize gains in the distribution on their tax form that year.

I appreciate that the rounding error will affect backtesting, but at the same time a 97% fall also can create a lot of havoc - especially since this is reflected in the backtest as a loss (not representing what actually happened (as you described)).

Even though Y! adjust the same way, I would argue that applying the dividend adjustment first would be the correct method, as this (best) reflects the real world scenario. I would also argue that, if you were to use the other method (split then div adj) the split factor should also be applied to the dividend. This would give the same result. As I mentioned before, this may just be an oversight in the calculation sequence as a split and div adj falling on the same is relatively rare event.

Please create a Support Ticket for this. For the intended purpose of adjusting the data, I'd have to say that you're right. Either way, percentage changes in price and volume are preserved, but for adjustments of this magnitude, you could run into precision problems when backtesting stocks with limit-order scripts.

It just bothers me that we're now showing the same adjustment as Y! does, and this is a Y! Data Provider. But if they're doing it wrong, it doesn't make sense for us to do it wrong too.

Even if we decide to change it, it's not going to happen overnight. In the meantime, you could write your own code to adjust price and volume for dividends and splits. Note that the Y! adapter always stores the raw price data and performs the adjustments on the fly. You can do the same by turning off the dividend and split adjustment, access the "Dividend (Yahoo! Finance)" and "Split (Yahoo! Finance)" fundamental items, and apply them to the chart data before doing anything else.

Thanks for your help on this Cone. I have created a support ticket. I will look into how many cases there are where a split and dividend fall on the same day. If not many, a work around may just be to apply a split manually by a factor. Otherwise I will write my on adjustment code.