A Study on Determinants of Short Term Return on Stocks in the S&P500

Abstract

Several research papers exist on whether stock market movement can be predicted or not. This paper uses data from 500 stocks in the Standards and Poor’s (S&P) 500 and checks the level of prediction of the selected variables and actual stock price movements, appreciation/depreciation against the predictor variables. The independent variables used are Price to Earnings (PE), Price to Sales (PS), Price to Earnings Growth (PEG), Price to Book (PB), Deviation from 200 day moving average (200D MVA), deviation from 50 day moving average (50D MVA), deviation from 52 week hi (52WH), deviation from 52 week low (52WL). Finally the yahoo finance 1 year target price (TGTP) is used as a predictor variable for the short term appreciation or depreciation of the stock price. Short term returns are assumed to be share price returns for the 10 month period between October 2014 and August 2015. The method used to calculate the return is the difference in stock price in August 2015 and October 2014, plus any dividends paid in the time period, the result divided by the stock price in October 2014.The calculated return is regressed against the variables PE, PEG, PB, PS, 200DMVA, 50DMVA, 52WH, 52WL, TGTP. Results indicate that none of PE, PB, PS, PB ratios is a good predictor of stock price changes (R-squared is less than .02 in each case). 200DMVA and 50DMVA are better predictors of stock price changes (R-squared is .10) and finally the stock returns are regressed against TGTP and the R-squared is close to .90, Thus, one can conclude safely that there is good correlation between 1 year target price stated in the yahoo finance portal and the actual stock price 10 months from that date. Given this level of predictive power, it may be tempting for the investor to create a portfolio of stocks where the target price exceeds the last trade price as on decision date. In order to test such portfolios, 25 random portfolios of 30 stocks were obtained from the data set of 500 stocks and the returns calculated for each of the random portfolio of stocks. If indeed the mean returns of the random portfolios exceeded the investment in an equal weighted portfolio of the 500 stocks for the same time period in question then target price is useful as a predictor. The results however indicate that the target price is not a predictor although the R-squared of the regression between target price and actual price is high and may tempt investors to use that parameter for equity purchases in the short term. Further research should explore on the disconnect between the high R-squared obtained in the regression and the inability for the investor to create a portfolio which can consistently beat the market.