Economics and statistics have a relatively strange relationship. They are perfect together, yet rarely do they get along. It comes down to size, sample size that is. Statistics begs for a large enough sample of data that is as uniform as possible. The average number of bighorn sheep in the state of Colorado over the last decade and calculating the probability you’ll see at least one on your summer vacation to the Rockies is right up statistic’s alley. Economics does not lend itself to such clean-cut plentiful data. Well, most economic data anyway.

The unemployment rate, which has been produced much in the same way since the 1930s, offers the best chance for the marriage of economics and statistics to last. Going back to 1948 — the dawning of the post-war economy — we have a sample size with over 700 monthly units that are as close to uniform as economics gets. The distribution of the monthly percentage change, as shown below, bears this out in a bell-curve like format.

For roughly 50 years, the average monthly change in the unemployment rate was 0.0 percent with a standard deviation of 0.2 percent. For all you statistics geeks, over this period you could’ve said with 99 percent confidence that the monthly change in the unemployment rate would always fall in-between -0.4 to 0.4 percent. But that is not the case anymore.

You’ll notice that since 2000, the distribution appears to have undergone a slight shift to the right. The mean over the last 13 years has remained at 0.0 percent and the standard deviation is still basically 0.2 percent(just a hair less at 1.7 percent actually). However, the tails have filled in quite dramatically in a very short amount of time. For instance where it took 51 years to get 5 months of a 0.5 percent increase in monthly unemployment (0.8 percent of the time), it has only take 156 months since the turn of the century to get 3 occurrences of a 0.5 percent monthly unemployment increase (1.9 percent of the time).

In order to balance this statistical anomaly, yet typical mass-layoff business dynamic, the unemployment rate tends to fall at a slower monthly pact more often. For example, while looking at the 1948-1999 data while it is true that more instances of outlier exists on the plus side (11+5+14 = 30 versus 17 +4 +3 = 24) but is justified by the fact that the rate fell by 0.1 percent by almost 25 percent more than it increased by 0.1 percent (126 versus 101).

Therefore, it isn’t really a surprise that the unemployment rate fell by 0.1 percent last month although the headlines would make you think it was. Rather it is the numbers working themselves out to fit that bell-curve distribution. And yes, we are more than due for a few large drops in the unemployment rate going forward, which I am sure will be another “surprise” when it happens.