Tuesday, March 10, 2009

Dow Jones Rally - Relative vs. Absolute Change

AP News Story: Stocks Shoot Higher on Citigroup Profit News


Today the Dow Jones Industrial Average (DJIA) shot up over 300 points on good news from Citigroup. They are actually profitable over January and February, which was unexpected.


Fascinating as this is, what's more important to look at than the 300 points, called the absolute change, is the percentage change of the Dow Jones Industrial Average: 4%. In general, it is more important to seek this number, called the relative change.

In today's news, neither number seems that earth shattering, but early in our current recession woes, the media bewailed the fact that the DJIA had its largest ever one day drop. But what they didn't tell you was that the largest ever one day drop was the absolute change, not the relative change.

Let me give you an example of what I mean. Say we are watching two children playing tetherball, and one of them is clearly more coordinated than the other. Should the mother of the clumsier child be worried that her darling is growing up to be a personification of slapstick comedy? Probably not.

In this example, let's say the more coordinated child is older. We'd expect there to be a difference in agility based on age. But how much older? I ask the two mothers and find out there is a one year difference in age. Does that mean anything?

One year is the absolute change in their ages. Older child minus younger child is one year. But how important that one year is depends on how old they actually are. If our children are three and two (I don't know what mother is introducing tetherball at these tender ages, but it's my story, go with it), then one year makes a huge difference. A three year old has a lot of hand-eye coordination that a two year old lacks.

On the other hand, if they are twenty and twenty-one, there is really hardly any difference to be expected between them, contrary to liquor law cutoffs. So knowing their difference in age is one year doesn't tell us much.

Let's instead calculate the relative change in age. Start with our preschoolers.

Relative change = (older - younger)/younger = (3 - 2)/2 = 1/2 = 0.5 = 50%
A three year old is fifty percent older than a two year old. That's a huge change. What about our high school graduates?

Relative change = (older - younger)/younger = (21 - 20)/20 = 1/20 = 0.05 = 5%
Our twenty-one year old tetherball champ is only five percent older than our twenty year old. Far less significant.

So next time someone cries in anguish, "The Dow just lost a thousand points", well that would always be bad. But put it in perspective. If it was hovering around 30,000, (a 3.3% loss,) the loss would be bothersome, but not necessarily panic worthy. If on the other hand the Dow started about 4,000 and lost 1,000, (a 25% loss,) well, the banks better have plenty of tellers available.

Until the next fuzzy: Enjoy your numbers.

Friday, February 27, 2009

Revised GDP - Annualized Percentages

New York Times article: G.D.P. Revision Suggests a Long, Steep Downfall


Today the commerce department released figures stating that in the last quarter of 2008, the Gross Domestic Product in the United States fell by an annualized rate of 6.2%. This is big news because none too long ago, the government estimate the change in the GDP as a fall of 3.8%. The corrected rate of change is much worse.

But what exactly does this rate of change mean? What happened to the GDP between October 1, 2008 and December 31, 2008? Did it really change by 6.2%? No. That is the annualized rate of change. And what exactly does an annualized rate of change mean?

Rates of change are often given in an annualized form so they have meaning relative to one another. For example, someone else might quote the rate of change for January 2009. How do you compare a rate of change over one month to a different rate of change over three months? Things might seem better if someone told you January's rate of change was only a fall of 0.5%, but it wouldn't be. To truly compare those rates of change, we want all of them to apply to the same scale.

So here's what we do. We pretend: say everything were going to continue for the rest of the year exactly like it did for the last quarter of 2008? Then, if that were to happen, the GDP would fall by 6.2% Of course, not every quarter is going to be exactly the same as the last quarter of 2008. After four quarters, we may look back at an actual annual change that is worse than a 6.2% decline. Or, if we're optimistic, maybe it will be better.

Let's find the actual rate of change. We could approximate it pretty easily. There are four quarters in a year. So the rate of change over a quarter is approximately 6.2%/4 = 1.55%. And that's reasonably close to the true value, one might say close enough for government work. (Though I suspect the Commerce Department might disagree.)

Let's call the annualized rate of change a. We're saying that
a = -0.062
. (Remember 6.2% means 6.2 per hundred or 6.2/100 = 0.062. And since it is a decline, we have a negative rate of change.) So the new GDP at the end of our pretend year
Gyear = Gstart * (1 + a)
. Well, 1 + a = 1 + (-0.062) = 1 - 0.062 = 0.938


It doesn't matter what either of those G values are. We would care if we wanted to know the actual GDP figures, but we're only looking at the rate of change.


Now, the actual rate of change for the quarter, let's call it q. This is what we want to find.

Gone quarter = Gstart * (1 + q)


But we are pretending that this exact same rate of change happens four quarters in a row. So we follow this same pattern. Multiply each quarterly GDP by the change and get the new quarterly GDP.

G two quarters = Gone quarter * (1 + q) = Gstart * (1 + q) * (1 + q) = Gstart * (1 + q)2
Gthree quarters = Gtwo quarters * (1 + q) = Gstart * (1 + q)2 * (1 + q) = Gstart * (1 + q)3
Gyear = Gthree quarters * (1 + q) = Gstart * (1 + q)3 * (1 + q) = Gstart * (1 + q)4


Wow! That seemed like a lot of work to find out how we'd calculate the end of the year GDP based on each quarter. But of course, according to what we pretended (an economist would rather say assumed than pretended, but they really mean pretended), these two values should be the same:
Gyear = Gstart * (1 + q)4
Gyear = Gstart * (1 + a)


So:
(1 + q)4 = (1 + a)
(1 + q)4 = 0.938


I wish I could make the next part pretty, but there's no getting around taking fourth roots. I'm also going to round our answer to three decimal places.


1 + q = (0.938)1/4 = 0.984

q = 0.984 - 1 = -0.016


That's pretty darn close to our easy approximation. If we'd gone out more digits from the decimal place we would have found
q = -0.0159
. Even closer. So the approximation wasn't at all bad.

The final point being that over a three month period the Gross Domestic Product fell just over one and a half percent. That's not good. But it didn't fall 6.2% in that amount of time. If it had, that would be catastrophic.

Until the next fuzzy: Enjoy your numbers!