Last week our discussion took us as far as to a description of the normal or Bell curve.
See a depiction above. The numbers at the bottom of the graph refer to “standard deviations” from
the norm.
We were discussing specifically mileage errors on maps. One standard deviation from the mean (the area between -1 and +1 on our graph above) accounts for roughly 68 percent of all the maps in our hypothetical database. Two standard deviations from the mean (the area between -2 and +2) account for roughly 95 percent of the maps. Three standard deviations account for 99 percent.
We were discussing specifically mileage errors on maps. One standard deviation from the mean (the area between -1 and +1 on our graph above) accounts for roughly 68 percent of all the maps in our hypothetical database. Two standard deviations from the mean (the area between -2 and +2) account for roughly 95 percent of the maps. Three standard deviations account for 99 percent.
A normal curve for a phenomenon is
taken as evidence of randomness. If there were some reason why the mapmakers of
17th century England were inclined to make a particular error that
reason would show up as some non-normality in this chart. Perhaps the roads
between these two places were especially good by the standard of the day, and
the ease of travel created a general impression that the cities were closer
than was/is the case. That might lead to a lop-sided (or as statisticians say a
skewed) curve.
Or perhaps some famous troubadour of
the day had sung of the “191 cursed miles” between his own home and that of his
beloved. The song became quite popular and the right mileage number stuck in
people’s heads. This would make mistake
less likely, and it would give the resulting curve a pointier look or (again to
use the right jargon) a higher kurtosis than the above.
Now: it isn’t difficult to apply the
same general idea of probability to the movement of the price of a stock.
Consider every day in which XYZ’s stock has been listed on an exchange as a
distinct datum. On any given day, the price has either risen, or fallen, or
held its ground. As a first approximation, we might take a price that stays at
the end of the day where it was at the start at the mid-point of our curve. As a first hypothesis, we might expect small upward moves
and small downward moves to be about equally likely with each other, and
slightly less likely than a flat day. Huge moves in either direction would be
quite unlikely, some number of standard deviations away from the center.
If we create a chart for XYZ stock
on these principles and it turns out not to look like a Bell curve, then could
we fairly conclude that there is some unexpected joker in the deck? Is that how
we might make our case that the Morgan Stanley downgrade had an impact? Because
such reports create non-random distortions? Maybe the downgrade is acting
something like our imaginary troubadour or like those atypically good roads
that might once have distorted cartography.
Maybe. But we have to take account
of other complications.
For example, there is a long-run
upward trend in stock market prices, so that if we really tried to use no change as a mid-point for a
particular stock’s prices, we’d risk a much skewed curve. The long-run trend is
in some periods the result of inflation, in some periods the result of real
underlying economic growth, and in some periods includes elements of both.
Still, it exists, and some modification of our method is in order.
Suppose we compared a particular
stock’s performance to that of a broader market (thus controlling for factors
like price inflation and over-all economic growth). As a second approximation,
then, we might say that a stock price that at the end of the day moves just as
far, up or down, in percentage terms, as did the broader index on the same day (the Dow Jones,
the S&P, the Nikkei, whatever) is at the midpoint of the
curve. If we think that stock prices move randomly, we would expect days in
which the stock of XYZ did a little better than the broader market, or a little
worse, to be about equally likely with one another, and slightly less likely
than a day on which XYZ does just what the broader market does. Huge departures
from the broad market in either direction would be quite unlikely, some number of
standard deviations from the center.
Comments
Post a Comment