Πρόβλεψη της δεσμευμένης κατανομής της μεταβλητότητας των αποδόσεων του χρυσού
View/
Keywords
Χρυσός ; Realized volatility ; Quantile regressionAbstract
This study examines the behavior of realized volatility of gold price returns, with the aim
of analyzing how it behaves at different levels of the distribution. Gold price data for the
period 1833–2025 are used to calculate logarithmic returns and realized volatility, which
is defined as the square of returns. Initially realized volatility is examined and it is
observed that it presents asymmetry and extreme values, indicating that it does not follow
a normal distribution.
Then, quantile regression is applied to estimate the relationship between current realized
volatility and the realized volatility of the previous period at the quantiles τ = 0.10, 0.50,
and 0.90. The results show that the effect of past volatility is not the same at all levels.
Specifically at low and medium levels of volatility the effect of the previous period is
small, while at high levels the effect of past volatility on the corresponding quantile of
current realized volatility is much stronger. This indicates that during periods of increased
uncertainty volatility remains high and has a stronger influence on the upper quantile of
future volatility.
Overall the use of quantile regression provides a better understanding of volatility
behavior as it allows the analysis not only at the average level but also at low, medium,
and high levels of volatility.