GPRNK index

In our paper, Corporate Investment under North Korea Threat, we construct a monthly index of Geopolitical Risk from North Korea (GPRNK index) using articles published in 18 Korean newspaper and broadcast outlets.

First, we identify articles that contain "North Korea" in the following media outlets: Chosun Ilbo, Dong-a Ilbo, Joong-ang Ilbo, Kyunghyang Shinmun, Kookmin Ilbo, Munhwa Ilbo, Seoul Shinmun, Segye Ilbo, Hankyoreh, Hankook Ilbo, Maeil Business Newspaper, Money Today, Seoul Gyeongje, Hankook Gyeongju, Herald Economy, KBS, MBC and SBS.

Second, within the set of articles selected by the first step, we flag articles that mention four types of geopolitical risk as follows:

1.     Military Tensions: Articles that contain {nuclear or missile or military or war} and {threat or tension or provocation}, excluding articles that contain {peace}.

2. Sanctions: Articles that contain {sanction or pressure} and {refute or dissent or criticize}.

3. Talks & Agreements: Articles that contain {talks or dialogue} and {resume or agreement or negotiation}, excluding articles that contain {fail or break or boycott}.

4. Economic Cooperation: {"economic cooperation" or its abbreviations} and {progress or expectation}, excluding articles that contain {concern}.

Third, we compute the frequency of news articles (N_(j,it)) for each risk type j=1,2,3,4, outlet i, and month t. Then the frequency of negative geopolitical risk articles is computed as N_(neg,it)= N_(1,it)+N_(2,it) (Military Tensions + Sanctions) and the frequency of positive geopolitical risk articles as N_(pos,it)= N_(3,it)+N_(4,it) (Talks & Agreements + Economic Cooperation).

Fourth, we consider the relative net frequency of negative articles as X_it= (N_(neg,it)-N_(pos,it) )/N_(All,it), where N_(All,it) is the total number of articles published in outlet i and month t. We transform this net frequency count to have positive values.

Finally, we standardize this value by dividing it through by its time-series standard deviation from January 1995 to December 2016. We average this standardized outlet-level index across outlets by month to obtain their geopolitical risk index, which we multiplicatively rescale to a mean value of 100 from January 1995 to December 2016.