## Using the Scaling Analysis to Characterize Financial Markets

### BibTeX

@MISC{A_usingthe,

author = {T. Di Matteo A and T. Aste B and M. M. Dacorogna C},

title = {Using the Scaling Analysis to Characterize Financial Markets},

year = {}

}

### OpenURL

### Abstract

The scaling properties encompass in a simple analysis many of the volatility characteristics of financial markets. That is why we use them to probe the different degree of markets development. We empirically study the scaling properties of daily Foreign Exchange rates, Stock Market indices and fixed income instruments by using the generalized Hurst approach. The robustness of the results is tested by both Monte-Carlo studies and a computation of the scaling in the frequency-domain. We show that the scaling exponents are associated with characteristics of the specific markets and can be used to differentiate markets in their stage of development.

### Citations

1897 |
An Introduction to Probability Theory and its Applications
- Feller
- 1971
(Show Context)
Citation Context ...to 2001 are reported in Table 7 for Foreign Exchange rates, Stock Market indices and Treasury rates. These values differ from the spectral density exponent expected for a pure Brownian motion (β = 2) =-=[62]-=-. However, we have shown in Section 4 that this method is biased and we have indeed found power spectra exponents around 1.8 for random walks using three different random numbers generators. It must b... |

992 |
Human Behavior and the Principle of Least Effort : An Introduction to Human Ecology
- Zipf
- 1949
(Show Context)
Citation Context ...affine analysis, detrended fluctuation analysis (DFA) [35,36,37,38,39,40], periodogram regression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method =-=[44]-=-, the moving-average analysis technique [45], the Average Wavelet Coefficient Method [46,47,48,49], and -in this respect- connection to multi-fractal/multi-affine analysis (the q order height-height c... |

583 | The variation of certain speculative prices - Mandelbrot - 1963 |

577 | The Behavior of Stock Market Prices
- Fama
(Show Context)
Citation Context ... characteristic of financial markets [7]. Generalizations of the classical Brownian motion were made by Mandelbrot and followers involving either fractional Brownian motions [9,17,18], or Lévy motion =-=[8,19,20,21,22,23]-=-. Closely related additive scaling models have also been developed [6,12,24]: Brownian, fractional Brownian, Lévy processes. The above approaches generally involve additive monofractal processes and a... |

504 |
Fractional Brownian motions, fractional noises and applications
- Mandelbrot, Ness
- 1968
(Show Context)
Citation Context ...ty clustering is a general characteristic of financial markets [7]. Generalizations of the classical Brownian motion were made by Mandelbrot and followers involving either fractional Brownian motions =-=[9,17,18]-=-, or Lévy motion [8,19,20,21,22,23]. Closely related additive scaling models have also been developed [6,12,24]: Brownian, fractional Brownian, Lévy processes. The above approaches generally involve a... |

325 |
The Estimation and Application of Long Memory Time Series Models
- Geweke, Porter-Hudak
- 1983
(Show Context)
Citation Context ... the rescaled range statistical analysis 4R/S [33] and modified R/S analysis [34], multiaffine analysis, detrended fluctuation analysis (DFA) [35,36,37,38,39,40], periodogram regression (GPH method) =-=[41]-=-, the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coefficient Method [46,47,48,49], and -in thi... |

286 | An Introduction to Econophysics - Mantegna, Stanley - 2000 |

230 |
Options, Futures and Other Derivatives
- Hull
- 1999
(Show Context)
Citation Context ...gna@converium.com (M. M. Dacorogna). Preprint submitted to Elsevier Science 2 February 20081 Introduction The scaling concept is increasingly applied outside the traditional physical sciences domain =-=[1,2,3,4,5,6,7]-=-. In the recent years, its application to financial markets, initiated by Mandelbrot in the 1960 [8,9], has largely increased also in consequence of the abundance of available data [1]. Two types of s... |

226 | The jackknife and the bootstrap for general stationary observations
- Künsch
- 1989
(Show Context)
Citation Context ...ed Hurst exponents H(1) and H(2) in the whole time period (shown in Table 2) and in time periods of 250 days. Moreover, we tested the numerical robustness of our results by using the Jackknife method =-=[60]-=- which consists in taking out randomly 1/10 of the sample and iterates the procedure 10 times (every time taking out data which were not taken out in the previous runs). We observe (see Fig. 6) that t... |

216 | Long-term storage capacity of reservoirs - Hurst - 1951 |

172 |
An Introduction To High-Frequency Finance
- Dacorogna, Müller, et al.
- 2001
(Show Context)
Citation Context ...gna@converium.com (M. M. Dacorogna). Preprint submitted to Elsevier Science 2 February 20081 Introduction The scaling concept is increasingly applied outside the traditional physical sciences domain =-=[1,2,3,4,5,6,7]-=-. In the recent years, its application to financial markets, initiated by Mandelbrot in the 1960 [8,9], has largely increased also in consequence of the abundance of available data [1]. Two types of s... |

141 | Long-term memory in stock market prices, Econometrica 59 - Lo - 1991 |

137 | Fractals and Scaling in Finance - Mandelbrot - 1997 |

128 |
Fractals
- Feder
- 1998
(Show Context)
Citation Context ...-affine function X(t) has fluctuations on different time scales that can be rescaled so that the original signal X(t) is statistically equivalent to its rescaled version λ −H X(λt) for any positive λ =-=[54]-=-, i.e. X(t) ∼ λ −H X(λt). Brownian motion is self-affine by nature.) In the second case, when H(q) depends on q, the process is commonly called multi-scaling (or multi-fractal) [54,55] and different e... |

115 |
Maximum Likelihood Estimation of Stationary Univariate FractionallyIntegrated Time-Series Models
- Sowell
- 1992
(Show Context)
Citation Context ...modified R/S analysis [34], multiaffine analysis, detrended fluctuation analysis (DFA) [35,36,37,38,39,40], periodogram regression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood =-=[42,43]-=-, the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coefficient Method [46,47,48,49], and -in this respect- connection to multi-fractal/multi-affine analysi... |

87 |
Statistical Study of Foreign Exchange Rates, Empirical Evidence of a Price Change Scaling Law, and Intraday Analysis
- Muller, Dacorogna, et al.
- 1990
(Show Context)
Citation Context ...gna@converium.com (M. M. Dacorogna). Preprint submitted to Elsevier Science 2 February 20081 Introduction The scaling concept is increasingly applied outside the traditional physical sciences domain =-=[1,2,3,4,5,6,7]-=-. In the recent years, its application to financial markets, initiated by Mandelbrot in the 1960 [8,9], has largely increased also in consequence of the abundance of available data [1]. Two types of s... |

70 |
On the spectrum of Fractional Brownian Motions
- Flandrin
(Show Context)
Citation Context ...indeed associated with the scaling of the absolute spread in the increments. The exponent at q = 2, is associated with the scaling of the autocorrelation function and is related to the power spectrum =-=[56]-=-. A special case is associated with the value ∗(τ) scales of q = q∗ at which q∗H(q ∗ ) = 1. At this value of q, the moment Kq linearly in τ [9]. Since qH(q) is in general a monotonic growing function ... |

70 |
Long-term Storage Capacity of Reservoirs,” Trans
- Hurst
- 1951
(Show Context)
Citation Context ...d 1990-1996. 3 The theoretical framework The scaling properties in time series have been studied in the literature by means of several techniques such as the rescaled range statistical analysis 4R/S =-=[33]-=- and modified R/S analysis [34], multiaffine analysis, detrended fluctuation analysis (DFA) [35,36,37,38,39,40], periodogram regression (GPH method) [41], the ARFIMA estimation by exact maximum likeli... |

67 |
Mandelbrot and the stable Paretian hypothesis
- Fama
- 1963
(Show Context)
Citation Context ... characteristic of financial markets [7]. Generalizations of the classical Brownian motion were made by Mandelbrot and followers involving either fractional Brownian motions [9,17,18], or Lévy motion =-=[8,19,20,21,22,23]-=-. Closely related additive scaling models have also been developed [6,12,24]: Brownian, fractional Brownian, Lévy processes. The above approaches generally involve additive monofractal processes and a... |

63 |
Spectrum analysis--A modern perspective
- Kay, Marple
- 1981
(Show Context)
Citation Context ...2 Spectral analysis In order to empirically investigate the statistical properties of the time series in the frequency domain we perform a spectral analysis computing the power spectral density (PSD) =-=[61]-=- by using the periodogram approach, that is currently one of the most popular and computationally efficient PSD estimator. This is a sensitive way to estimate the limits of the scaling regime of the d... |

51 |
Scaling behavior in the dynamics of an economic index, Nature 376
- Mantegna, Stanley
- 1995
(Show Context)
Citation Context ...roenendijk et al. [10]. In this study, we are interested in the first type of analysis. Until now, most of the work has concentrated in studies of particular markets: Foreign Exchange [1,7,11], Stock =-=[12]-=- or Fixed Income [13]. These studies showed that empirical scaling laws hold in all these markets and for a large range of frequencies: from few minutes to few months. In a recent book [7], the hypoth... |

43 |
Effect of trends on detrended fluctuation analysis
- Hu
- 2001
(Show Context)
Citation Context ... in the literature by means of several techniques such as the rescaled range statistical analysis 4R/S [33] and modified R/S analysis [34], multiaffine analysis, detrended fluctuation analysis (DFA) =-=[35,36,37,38,39,40]-=-, periodogram regression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coe... |

40 |
Interest-Rate Option Models
- Rebonato
- 1996
(Show Context)
Citation Context |

38 |
Scaling in Economics: a Reader’s Guide
- Brock
- 1999
(Show Context)
Citation Context ...also reproduce. The purpose of this paper is to point out how a relatively simple statistics gives us indications on the market characteristics, very much along the lines of the review paper by Brock =-=[31]-=-. In Section 3, we recall the theoretical framework and we introduce the generalized Hurst exponents analysis. In Section 4 we describe the methodology utilized to empirically analyze the data. In Sec... |

37 |
An Introduction to Econophysics (Cambridge
- Mantegna, Stanley
- 1999
(Show Context)
Citation Context |

32 |
Wavelet Methods for Time Series Analysis Cambridge
- Percival, Walden
- 2000
(Show Context)
Citation Context ...gression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coefficient Method =-=[46,47,48,49]-=-, and -in this respect- connection to multi-fractal/multi-affine analysis (the q order height-height correlation) have been made [50,51,52]. The rescaled range statistical analysis (R/S analysis) was ... |

30 | Consistent High-Precision Volatility from High-Frequency Data,” Economic Notes
- Corsi, Zumbach, et al.
- 2001
(Show Context)
Citation Context ...llent paper by Groenendijk et al. [10]. In this study, we are interested in the first type of analysis. Until now, most of the work has concentrated in studies of particular markets: Foreign Exchange =-=[1,7,11]-=-, Stock [12] or Fixed Income [13]. These studies showed that empirical scaling laws hold in all these markets and for a large range of frequencies: from few minutes to few months. In a recent book [7]... |

28 |
Stochastic volatility as a simple generator of apparent financial power laws and long memory. Quantitative Finance
- LeBaron
- 2001
(Show Context)
Citation Context ...processes. The above approaches generally involve additive monofractal processes and analyses; but, in contrast, several scaling systems appear to be more complex. Recently, a controversy has erupted =-=[25,26,27,28]-=- to know if the processes that describe financial data are truly scaling or simply an artifact of the data. Moreover, several publications propose new scaling models or empirical analyses that better ... |

25 |
The variation of some other speculative price
- Mandelbrot
- 1967
(Show Context)
Citation Context ... characteristic of financial markets [7]. Generalizations of the classical Brownian motion were made by Mandelbrot and followers involving either fractional Brownian motions [9,17,18], or Lévy motion =-=[8,19,20,21,22,23]-=-. Closely related additive scaling models have also been developed [6,12,24]: Brownian, fractional Brownian, Lévy processes. The above approaches generally involve additive monofractal processes and a... |

19 |
Scaling properties of foreign exchange volatility, Physica A 289
- Gencay, Selcuk, et al.
- 2001
(Show Context)
Citation Context ...gression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coefficient Method =-=[46,47,48,49]-=-, and -in this respect- connection to multi-fractal/multi-affine analysis (the q order height-height correlation) have been made [50,51,52]. The rescaled range statistical analysis (R/S analysis) was ... |

18 | Long-term storage of reservoirs: an experimental study - Hurst - 1951 |

17 |
A critical look at Lo's modified R/S statistic
- Teverovsky, Taqqu, et al.
- 1999
(Show Context)
Citation Context ...t the lack of robustness of the original Hurst R/S approach in the presence of short memory, heterocedasticity, multiple scale behaviors has been largely discussed in the literature (see for instance =-=[34,63,64]-=-) and therefore several alternative approaches have been proposed. Also the fact that the 13range relies on maxima and minima makes the method error prone because any outlier present in the data woul... |

14 | Apparent multifractality in financial time series
- Bouchaud, Potters, et al.
(Show Context)
Citation Context ...t describe financial data are truly scaling or simply an artifact of the data. Moreover, several publications propose new scaling models or empirical analyses that better describe empirical evidences =-=[25,26,27,28,29]-=-. It should be however noted that - as underlined by Stanley et al. [30] - in statistical physics, when a large number of microscopic elements interact without characteristic scale, universal macrosco... |

14 |
Multi-affine analysis of typical currency exchange rates
- Vandewalle, Ausloos
- 1998
(Show Context)
Citation Context ...iation from Brownian, fractional Brownian, Lévy, and fractional Lévy models, as already seen in FX rates [1]. (Other cases showing marked deviations from Brownian motion have been discussed elsewhere =-=[50,52,57,58,59]-=-.) 3.2 Scaling spectral density and Hurst exponent For financial time series, as well as for many other stochastic processes, the spectral density S(f) is empirically found to scale with the frequency... |

13 | Low q-moment multifractal analysis of gold price, DowJones industrial average and BGL-USD exchange rate, Eur - Ivanova, Ausloos - 1999 |

13 |
Scaling and universality in economics: empirical results and theoretical interpretation
- Stanley, Plerou
- 2001
(Show Context)
Citation Context ...processes. The above approaches generally involve additive monofractal processes and analyses; but, in contrast, several scaling systems appear to be more complex. Recently, a controversy has erupted =-=[25,26,27,28]-=- to know if the processes that describe financial data are truly scaling or simply an artifact of the data. Moreover, several publications propose new scaling models or empirical analyses that better ... |

12 | Theorie des Risques Financiers - Bouchaud, Potters - 1997 |

10 |
Determination of the Hurst exponent by use of wavelet transforms
- Simonsen, Hansen, et al.
- 1998
(Show Context)
Citation Context ...gression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coefficient Method =-=[46,47,48,49]-=-, and -in this respect- connection to multi-fractal/multi-affine analysis (the q order height-height correlation) have been made [50,51,52]. The rescaled range statistical analysis (R/S analysis) was ... |

9 |
Statistical physics in foreign exchange currency and stock markets
- Ausloos
- 1996
(Show Context)
Citation Context ... in the literature by means of several techniques such as the rescaled range statistical analysis 4R/S [33] and modified R/S analysis [34], multiaffine analysis, detrended fluctuation analysis (DFA) =-=[35,36,37,38,39,40]-=-, periodogram regression (GPH method) [41], the ARFIMA estimation by exact maximum likelihood [42,43], the (m, k)- Zipf method [44], the moving-average analysis technique [45], the Average Wavelet Coe... |

9 |
Multifractality of self-affine fractals
- Barabasi, Viscek
- 1991
(Show Context)
Citation Context ...st should therefore be associated with the scaling behavior of statistically significant properties of the signal. To this purpose we analyze the q-order moments of the distribution of the increments =-=[9,53]-=- which is a good characterization of the statistical evolution of a stochastic variable X(t). It is given by the: Kq(τ) = 〈|X(t + τ) − X(t)|q 〉 〈|X(t)| q 〉 , (1) where the time-interval τ can vary bet... |

9 | Some portable very-long-period random number generators - Marsaglia, Zaman - 1994 |

9 |
Can statistical physics contribute to the science of economics?", Fractals
- Amaral, Buldyrev, et al.
- 1996
(Show Context)
Citation Context ...ver, several publications propose new scaling models or empirical analyses that better describe empirical evidences [25,26,27,28,29]. It should be however noted that - as underlined by Stanley et al. =-=[30]-=- - in statistical physics, when a large number of microscopic elements interact without characteristic scale, universal macroscopic scaling laws may be obtained independently of the microscopic detail... |

9 |
Fractals and Scaling in Finance; Springer-Verlag
- Mandelbrot
- 1997
(Show Context)
Citation Context ...caling concept is increasingly applied outside the traditional physical sciences domain [1,2,3,4,5,6,7]. In the recent years, its application to financial markets, initiated by Mandelbrot in the 1960 =-=[8,9]-=-, has largely increased also in consequence of the abundance of available data [1]. Two types of scaling behaviors are studied in the finance literature: (1) The behavior of some forms of volatility m... |

7 | Fractal geometry of financial time series
- Evertsz
- 1995
(Show Context)
Citation Context ...otion were made by Mandelbrot and followers involving either fractional Brownian motions [9,17,18], or Lévy motion [8,19,20,21,22,23]. Closely related additive scaling models have also been developed =-=[6,12,24]-=-: Brownian, fractional Brownian, Lévy processes. The above approaches generally involve additive monofractal processes and analyses; but, in contrast, several scaling systems appear to be more complex... |

7 |
How does the Eurodollars interest rate behave
- Matteo, Aste
- 2001
(Show Context)
Citation Context ...y have maturities of 3 and 6 months and 1 year. • ER: The Eurodollar interbank interest rates (Table 4) are bid rates with different maturity dates and they are daily data in the time period 19901996 =-=[32]-=-. As an example, the behaviors of FX rates and the SM rates for Japan (JPY/USD and Nikkei 225) and Thailand (THB/USD and Bangkok SET) as a function of time t are shown in Fig. 1 in the time period 199... |

6 | A hybrid joint moment ratio test for financial time series - Groenendijk, Lucas, et al. - 1998 |

6 |
Characterization of long-range correlations in complex distributions and profiles, Phys
- Mehrabi, Rassamdana, et al.
- 1997
(Show Context)
Citation Context |

6 |
Sparseness and Roughness of Foreign Exchange Rates
- Vandewalle, Ausloos
- 1998
(Show Context)
Citation Context ...is technique [45], the Average Wavelet Coefficient Method [46,47,48,49], and -in this respect- connection to multi-fractal/multi-affine analysis (the q order height-height correlation) have been made =-=[50,51,52]-=-. The rescaled range statistical analysis (R/S analysis) was first introduced by Harold Edwin Hurst [33] to describe the long-term dependence of water levels in rivers and reservoirs. It provides a se... |

6 |
Hurst analysis of electricity price dynamics. Physica A 283, 462-468. 50 Strozzi [et al.], Application of non-linear time series analysis techniques to the nordic spot electricity market data
- Weron, Przybyłowicz
- 2000
(Show Context)
Citation Context ...t the lack of robustness of the original Hurst R/S approach in the presence of short memory, heterocedasticity, multiple scale behaviors has been largely discussed in the literature (see for instance =-=[34,63,64]-=-) and therefore several alternative approaches have been proposed. Also the fact that the 13range relies on maxima and minima makes the method error prone because any outlier present in the data woul... |

5 |
Turbulence in financial markets: the surprising explanatory power of simple cascade models, Quantitative Finance 1
- Lux
- 2001
(Show Context)
Citation Context ...processes. The above approaches generally involve additive monofractal processes and analyses; but, in contrast, several scaling systems appear to be more complex. Recently, a controversy has erupted =-=[25,26,27,28]-=- to know if the processes that describe financial data are truly scaling or simply an artifact of the data. Moreover, several publications propose new scaling models or empirical analyses that better ... |