wilmott.com

Could you please give me some help on the following? How can one express the return of the Dow Jones Industrial Average index as a function of the returns of its individual components? It is not simply the sum of the products of stock weights times returns as I have already checked.

Did you account for stock splits? The weights depend only on price, so a stock split would reduce the weight of a stock by half.

I was drinking with the bloke who does a lot of inddex calcs for Reuters last night. He used words like bollocks and senseless about the Dow. Apparently it isn't weighted in any rational way.

Which weight are you using? In the composition file on Dow Jones Indexes they give you tomorrows weights.The correct weights are calculated as yesterdays close price for a specific stock divided the sum of yesterdays close prices. In the composition file they use todays close prices.

The DJIA is weighted in perfectly rational way. It's a Price Weighted Index Hence every time a price changes, the weights change. Most Engines are too stupid to handle that sort of index, but it's a perfectly normal way of weighting something. The good thing is that crap stocks will reduce their weightings automatically, so the index will tend to perform better. it's a sort of gamma on stocks versus the index... Regards.

They adjust for spilts though don't they? They double yesterday's weight in any stock that splits today.

QuoteThey adjust for spilts though don't they? They double yesterday's weight in any stock that splits today. There is often some confusion concerning the weights. In the Dow Jones Average indices each stockprice is always equally weighted, or price-weighted as FDAXHunter said. These two clasifications are equivalent. (1) The "equally-weighted" comes from the fact that, if you are trying to replicate the index, you should alway buy an equal number of each stock. (2) The "price-weighted" comes from the formula I gave in my last reply (that you weigh the return by yesterdays price to aggregate these figures to the index return). However, this formula is a direct result of the fact that the index is equally weighted, and vice versa.To answer your question Beans, the weights you will need to use to replicate the index will remain the same (as they are always equally weighted). However, if we by "weights" mean the the weight-vector used to aggregate the individual stocks returns to the retur of the index, then, as (a) the price of the splitting stock will decrease, and (b) this is not caught in the calculations of the weights on the day before the split, this price adjustment has to be made to the stockprice to be able to calculate these weights. Two things:1) The weight of the splitting stock will decrease, not increase.2) I think this is more complex way of thinking about the weights. There is no benifit of defining the weights in this way as it doesn't make the results better. It's just a mathematical result of the fact that the index is equally weighted.A long answer to a short question. That means it was a good question!

right now i remember. they change the divisor to accomodate for splits. there is still one of each stock, so the "weight" of the splittting stock relative to the other 29 goes down but they change the divisor, keeping the index at at the same level (so the index doesn't crash everytime a big component gets split)

exactly