ISBN 978-93-5156-328-0 International Conference of Advance Research and Innovation (ICARI-2015) 414 ICARI Comparative Analysis of Dividend Forecasting Methods R. Srivastava a , Chandramauli Gupta a , Himanshu Gupta a , Naunidh Singh a , Nand Kumar b a Department of Applied Mathematics, Delhi Technological University, New Delhi, India b Department of Humanities, Delhi Technological University, New Delhi, India Abstract Dividend Forecasting is a technique using which the future cash flows of a dividend paying stock can be found. Dividend is that part of the profits which the company distributes amongst its investors. Dividend Forecasting is an emerging field in stock market as it allows the shareholders to make wise decisions in buying and selling the stock and also predicts the performance of the company in the near future. With algorithm trading gaining a lot of popularity these days, technology has already started to govern the most complex financial markets of the world. And that is why a variety of techniques have been used to forecast the dividends. This paper presents a comparative study of the various approaches for dividend forecasting. In this paper, experiment with various techniques of forecasting dividend yields on secondary data of Infosys was done. The statistical packages of SPSS and Microsoft Excel were used for the analysis. The results of the study reveal that HP Filter with parameter α = 3200 gave the maximum accuracy of 80.30% with other methods giving relatively closer but lesser accuracies for the chosen dataset. 1. Introduction A dividend is a payment made by a corporation to its shareholders, usually as a distribution of profits. When a corporation earns a profit or surplus, it can either re-invest it in the business (called retained earnings), or it can distribute it to shareholders. A corporation may retain a portion of its earnings and pay the remainder as a dividend. Distribution to shareholders can be in cash (usually a deposit into a bank account) or, if the corporation has a dividend reinvestment plan, the amount can be paid by the issue of further shares or share repurchase [1]. A dividend is allocated as a fixed amount per share, with shareholders receiving a dividend in proportion to their shareholding. For the joint stock company, paying dividends is not an expense; rather, it is the division of after tax profits among shareholders. Retained earnings (profits that have not been distributed as dividends) are shown in the shareholders' equity section on the company's balance sheet - the same as its issued share capital. Public companies usually pay dividends on a fixed schedule, but may declare a dividend at any time, sometimes called a special dividend to distinguish it from the fixed schedule dividends. Cooperatives, on the other hand, allocate dividends according to members' activity, so their dividends are often considered to be a pre-tax expense. Dividend Forecasting is therefore influenced a lot by the financial condition of the company. In order to forecast dividend yields, a deep insight into the balance sheets is required and an in-depth knowledge of the forecasting techniques fitting different kinds of datasets is indispensable. The methods used for the study are explained below: 1.1 Principal Component Analysis (PCA) [2] PCA is a statistical method used for dimensionality reduction primarily. PCA is mathematically defined as an Corresponding Author, E-mail address: himanshuguptan@gmail.com All rights reserved: http://www.ijari.org orthogonal linear transformation that transforms the data to a new coordinate system such that the greatest variance by some projection of the data comes to lie on the first coordinate (called the first principal component), the second greatest variance on the second coordinate, and so on. 1.2 Factor Analysis Factor analysis is a statistical procedure used to identify a small number of factors that can be used to represent relationships among set of interrelated variables. To apply factor analysis, a correlation matrix of the quarterly balance sheets of Infosys from 2001-2014 was obtained for all 28 factors and varimax rotation was applied on the matrix so obtained to make the factors interpretable. The scores were then computed for every factor. NOTE: The following two tests were conducted to observe the feasibility of Factor Analysis on our dataset. 1.2.1 Bartlett’s test of Sphericity It is used to test the hypothesis that the correlation matrix is an identity matrix. 1.2.2 Kaiser Mayer Olkin Test It is a measure of sampling adequacy and is used to compare the magnitudes of the observed correlation coefficients in relation to the magnitudes of the partial correlation coefficients. Large KMO values are good because correlation between pairs of variables (potential factors) can be explained by other variables. If the KMO is below 0.5, factor analysis is not preferred. 1.3 Linear Regression Simple linear regression is a statistical method for approximating a straight line through a set of n data points. One assumes the presence of a straight line that describes the data, y = α + βx, where the intercept α and slope β are calculated using the least squares estimator, Article Info Article history: Received 8 January 2015 Received in revised form 15 January 2015 Accepted 22 January 2015 Available online 31 January 2015 Keywords Dividend, Principal Component Analysis, Factor Analysis, Regression, Moving Averages, HP Filter, etc