Abstract. In stochastic finance, one traditionally considers the return as a competitive measure of an asset, i.e., the profit generated by that asset after some fixed time span ∆t, say one week or one year. This measures how well (or how bad) the asset performs over that given period of time. It has been established that the distribution of returns exhibits "fat tails" indicating that large returns occur more frequently than what is expected from standard Gaussian stochastic processes [1,2,3]. Instead of estimating this "fat tail" distribution of returns, we propose here an alternative approach, which is outlined by addressing the following question: What is the smallest time interval needed for an asset to cross a fixed return level of say 10%? For a particular asset, we refer to this time as the investment horizon and the corresponding distribution as the investment horizon distribution. This latter distribution complements that of returns and provides new and possibly crucial information for portfolio design and risk-management, as well as for pricing of more exotic options. By considering historical financial data, exemplified by the Dow Jones Industrial Average, we obtain a novel set of probability distributions for the investment horizons which can be used to estimate the optimal investment horizon for a stock or a future contract.
PACS.Financial data have been recorded for a long time as they represent an invaluable source of information for statistical investigations of financial markets. In the early days of stochastic finance it was argued that the distribution of returns (see definition below) of an asset should follow a normal (Gaussian) distribution [4,5]. However, by analysing large, and often high-frequency, financial data sets, it has been established that these distributions on short time scales -typically less then a month, or so -can posses so-called "fat-tails", i.e. distributions that show strong deviations from that of a Gaussian [1,2,3] with higher probabilities for large events. This is similar to the distributions found for turbulence in air and fluids which have led to comparisons between the statistics of financial markets and that of turbulent fluids [5,6,7,8]. In turbulence, one obtains stretched exponential distributions which find their analogy in finance when considering higher order correlations of the asset price [9,10].In order to get a deeper understanding of the fluctuation of financial markets it is important to supplement this established information of fluctuations in the returns with alternative measures. In the present paper, we therefore ask the following "inverse" question: "What is the typical time span needed to generate a fluctuation or a movement a Email address: ingves@nordita.dk b Email address: mhjensen@nbi.dk c Email address: johansen@nbi.dk Correspondence to: Ingve Simonsen (in the price) of a given size" [11,12,13,14,15,16]. Given a fixed log-return barrier, ρ, of a stock or an index as well as a fixed investment date, the corresponding time span is estimate...