volatility around the time of the events, which are defined by spikes in our index. Also,
average volatility in the post-event window is larger than average volatility in the pre-
event window. Indeed, this can also be confirmed by inspecting the statistics reported at
the bottom of the window. The second, smaller plot displays the density functions of the
pre-event samples and post-event samples; thus, the fact that the pale curve (the post-
event density function) is shifted to the right vis-a`-vis the dark curve (the pre-event
density function) means that there has been an upward shift in volatility, on the average,
as a result of the events.
In Figure 3.3 the impact of these events seems clear, but for other indices the impact
may be less visually apparent, and thus it is important to measure the impact using
rigorous statistical techniques, which we propose in Sections 3.5.3–3.5.5. Nevertheless,
it is instructive to consider two more examples.
Figure 3.4 shows an event study for our Agriculture index. This time, the currency
pair is AUD/USD and return is the variable of interest. Upon visual inspection, there
seems to be no significant change as a result of the events. The density plots to the right
confirm this as well, as do the statistical tests described below. This is not surprising,
however, since it is not clear that the presence of agriculture news would tend to drive
exchange rates in one particular direction. Nonetheless, there is an impact which can
again be seen by examining exchange rate volatility (this event study is shown in
Figure 3.5).
Figure 3.5 shows an increase in volatility after surges in agriculture news, although it
is a more modest effect than the example of macroeconomic news in Figure 3.3. Even so,
84 Quantifying news: Alternative metrics
Figure 3.3.Screenshot of event analysis tool GUI, coded in MATLAB.