1: \begin{abstract}

2: We study, both analytically and numerically, an ARCH-like, multiscale model of volatility, which 

3: assumes that the volatility is governed by the observed past price changes over different time scales. 

4: With a power-law distribution of time horizons, we obtain a model that captures most stylized facts of 

5: financial time 

6: series: Student-like distribution of returns with a power-law tail, long-memory of the volatility, 

7: slow convergence of the distribution

8: of returns towards the Gaussian distribution, multifractality and anomalous volatility relaxation 

9: after shocks. 

10: At variance with recent multifractal models that are strictly time reversal invariant, 

11: the model also reproduces the 

12: time asymmetry of financial time series: past large scale volatility influence 

13: future small scale volatility. In order to quantitatively reproduce all empirical observations, 

14: the parameters must be chosen such that the 

15: model is close to an instability, meaning that (a) the feedback effect 

16: is important and substantially increases the volatility, and (b) that the model 

17: is intrinsically difficult to calibrate

18: because of the very long range nature of the correlations. 

19: By imposing consistency of the model predictions 

20: with a large set of different empirical observations, 

21: a reasonable range of the parameters value can be determined. 

22: The model can easily be generalized to account for jumps, skewness and multiasset correlations. 

23: \end{abstract}

24: