This paper has applied spectral analysis to reveal possible cyclical components in the Indian stock returns series. For power spectrum analysis, the study used daily data of four broad-based market indices viz., BSE Sensitive Index, BSE National Index, NSE S&P CNX 500 Index and NSE S&P CNX Nifty Index during the period from January 1991 to December 2001. The univariate spectra (power spectrum) for the four returns series suggest that there are no significant cyclical patterns present in four stock price indices. Apart from that, the study has considered three major developed stock markets indices for crossspectrum analysis, viz., Dow Jones Industrial Average (DJIA) of the US, FTSE 100 index of the UK, and Nikkei 225 of Japan, spanning over the period from January 2000 through December 2001. The crossspectrum analysis suggests that there are no similar long-term development features between India and developed stock markets studied here.

The rich stock of theoretical and empirical literature in the area of stock price behavior reveals that stock price is not merely a number as we observe; rather, it is a manifestation of all available information in the market at a certain point of time. For practitioners, studying pattern in stock price behavior is quite important in the sense that it is to be used to predict future movements in prices, which will guide the trading rule. The policy makers, on the other hand, view stock prices as indicators of policy changes on the monetary and fiscal front, i.e., any change in the economy gets reflected in stock prices and hence studying stock price behavior assumes immense importance from the policy point of view. Any kind of extraordinary or random movement in prices exhibiting gross deviation from that implied by economic fundamentals raises concern both for practitioners in the market place as well as policy makers. Therefore, understanding and analyzing stock price behavior have been of direct interest to academics in general and model builders in particular.

Over the past few decades, there has been a dramatic reversal in opinion concerning the degree of information contained in the data, leading to a new approach towards analysis of economic and financial time series. There are two major approaches to analyze the behavior of a time series, viz., time domain and frequency domain. In fact, the analysis of time series commenced in the frequency domain. The idea of decomposing a time series into different components with different frequencies was technically termed as spectral analysis of time series, which developed as a distinct and parallel discipline in contrast with the analysis of time series in time domain.