North Korea is as unpredictable a regime as a political scientist might imagine. Its decision-making is secret. Its leader is a posterchild for the madman theory: the idea that simulating insanity may be a useful tool of intimidation.

Yet, a new paper by economists Andy Kim, Hyoung Goo Kang, and Jong Kyu Lee suggests that North Korea’s aggressive military and diplomatic actions do follow some predictable, though surprising, patterns.

The authors analyze 287 of the country’s military and diplomatic interactions between 1999 and 2012. They assign different levels of hostility to these actions, with missile tests as the most aggressive category. They then examine whether a range of different variables can predict the more aggressive acts. Two findings stand out.

Read British newspapers

The first standout finding comes from their coding of print journalism. The authors collected more than 70,000 newspaper reports about North Korea. For each article they used automated text analysis to code how negative the tone of the article was. They then examined whether highly negative reports were correlated with an aggressive North Korean action one or two days later.

It turns out that the tone of newspaper reporting did help the model predict North Korean aggression, improving its predictive value by as much as 47 percent.

But here’s what’s really surprising: South Korean and U.S. coverage didn’t correlate with North Korean action. Only British news stories did. This doesn’t imply that the articles’ attitudes prompted those actions — only that we can systematically observe those negative British articles coming before the aggressive actions.

The authors suggest that this is so because British journalists may have more information about what’s happening in North Korea. The United Kingdom has an embassy in North Korea, unlike either South Korea or the United States. That’s consistent with another finding by Andy Kim and Hosung Jung that British hedge funds show the most significant abnormal short-selling of the South Korean stock market (KOSPI) index two to three trading days before North Korean military actions.

Another possibility (not suggested by the authors) is that British news stories reveal stronger attitudes, period – whether positive or negative — than do either South Korean or U.S. reporting, and are therefore easier to code using automated techniques.

Take note of Dear Leader birthdays

The second key finding is that North Korea is more likely to undertake aggressive acts within five days of a past or present “Dear Leader’s” birthday. That wasn’t true for organizational anniversaries, such as the establishment of the North Korea government (Sept. 9), the Army (April 25), or the Labour Party (Oct. 10).

A note of warning: Kim Jong Il’s birthday was Feb. 16, so we’re in the danger zone.

That’s a reminder — in case we had forgotten — that North Korea is a personal dictatorship, with an agenda focused much more on entertaining the “Dear Leader” than upholding the country’s institutions.

So how does this matter?

North Korea’s recent satellite launch, which apparently flew over the Superbowl site last Sunday, illustrated once again that the country will continue to take actions that are threatening to South Korea and the international community. We don’t know if the South Korean and U.S. governments have predictive models based on classified information that perform better than this one.

Whether or not they do, the rest of us — citizens, businesses, and other non-governmental actors — also have an interest in knowing when to expect aggression. This analysis shows that even these highly secretive and seemingly random actions follow somewhat predictable and publicly observable patterns.