We study the distribution of fluctuations of the S&P 500 index over a time
scale Delta t by analyzing three distinct databases. Database (i) contains
approximately 1200000 records, sampled at 1-min intervals, for the IS-year
period 1984-1996, database (ii) contains 8686 daily records for the 35-year
period 1962-1996, and database (iii) contains 852 monthly records for the
71-year period 1926-1996. We compute the probability distributions of retur
ns over a time scale Delta t, where Delta t varies approximately over a fac
tor of 10(4)-from 1 min up to more than one month. We find that the distrib
utions for Delta t less than or equal to 4 d (1560 min) are consistent with
a power-law asymptotic behavior, characterized by an exponent alpha approx
imate to 3, well outside the stable Levy regime 0 < alpha < 2. To test the
robustness of the S&P result, we perform a parallel analysis on two other f
inancial market indices. Database (iv) contains 3560 daily records of the N
IKKEI index for the 18-year period 1984-1997, and database (v) contains 464
9 daily records of the Hang-Seng index for the Is-year period 1980-1997. We
find estimates of alpha consistent with those describing the distribution
of S&P 500 daily returns. One possible reason for the scaling of these dist
ributions is the long persistence of the autocorrelation function of the vo
latility. For time scales longer than (Delta t)(x) approximate to 4 d, our
results are consistent with a slow convergence to Gaussian behavior. [S1063
-651X(99)11211-X].