# Relative merits of Adjusted versus Closing prices for market predictions

## Basic question

I am familiar with the data returned from Yahoo. For indices and the like (e.g. ETFs) there are seven columns of data: Date, Open, High, Low, Close, Volume, Adjusted. We only need consider Close and Adjusted here. Here is a brief definition of Adusted-Prices, which states that they are used when examining historical prices (without any references).

I would like to know what the relative merits are for using the market closing prices for any given asset, versus the adjusted prices, which take events such as dividend-payouts into account. As the accepted answer in this thread states (in the personal finance SO), which of the two you should select depends on what you want to use it for. This is where I am unsure.

I feel like the adjusted prices would perhaps be more telling of the market sentiment towards and actual asset, as a dividend payout for example would likely improve investor confidence towards that stock. On the other hand, I am looking at predicting number based on other numbers, which are those actually displaying what was most recently paid for that asset, so it feels more concrete to use the closing prices.

Any references to literature on this topic would be greatly appreciated.

## More detailed example

I am using a wide variety of data in a model that predicts the return of the Dow Jones. This includes the returns of other indices as well as commoodities and currencies etc. Specifically for the indices that I am looking at, I had a look to see what the difference was between the Close and Adjusted prices. Here is the code (in R using quantmodto get the relevant data) and the results

library(quantmod)     # to get market data
library(data.table)   # my preferred way to manipulate the data

myDAX <- getSymbols(Symbols =  c("^GDAXI"), auto.assign = FALSE)
myDOW <- getSymbols(Symbols = c("^DJI"), auto.assign = FALSE)
myEEM <- getSymbols(Symbols = c("EEM"), auto.assign = FALSE)
myFTSE <- getSymbols(Symbols = c("^FTSE"), auto.assign = FALSE)
myGSPC <- getSymbols(Symbols = c("^GSPC"), auto.assign = FALSE)
myN225 <- getSymbols(Symbols = c("^N225"), auto.assign = FALSE)

## Create data tables
dax <- as.data.table(myDAX)
dow <- as.data.table(myDOW)
eem <- as.data.table(myEEM)
ftse <- as.data.table(myFTSE)
sp500 <- as.data.table(myGSPC)
n225 <- as.data.table(myN225)

## create new column 'delta' - the difference between adjusted and closing prices
dax[, delta := GDAXI.Adjusted - GDAXI.Close]
dow[, delta := DJI.Adjusted - DJI.Close]
eem[, delta := EEM.Adjusted - EEM.Close]
ftse[, delta := FTSE.Adjusted - FTSE.Close]
sp500[, delta := GSPC.Adjusted - GSPC.Close]
n225[, delta := N225.Adjusted - N225.Close]


Here we get a results printed out, saying how many of the values for delta were zero.

print(paste0(sum(dax$delta == 0) / nrow(dax) * 100, "% zero delta")) [1] "100% zero delta" print(paste0(sum(dow$delta == 0) / nrow(dow) * 100, "% zero delta"))
[1] "100 % zero delta"

print(paste0(sum(eem$delta == 0) / nrow(eem) * 100, "% zero delta")) [1] "1.66 % zero delta" print(paste0(sum(ftse$delta == 0) / nrow(ftse) * 100, "% zero delta"))
[1] "100 % zero delta"

print(paste0(sum(sp500$delta == 0) / nrow(sp500) * 100, "% zero delta")) [1] "100% zero delta" print(paste0(sum(n225$delta == 0) / nrow(n225) * 100, "% zero delta"))
[1] "100% zero delta"

print(paste0(sum(sse\$delta == 0) / nrow(sse) * 100, "% zero delta"))
[1] "100% zero delta"


We can see that only in the case of the ETF EEM did we get any differences in the prices, only 1.66% of the delta values being zero. Looking at the data for individual stocks (for example: Apple, I wouldn't expect hardly any delta values to be equal to zero!

A final question remains: Why are the prices of the indices above not adjusted at all, considering their constituents are adjusted?