Skip navigation

Pardee Logo International Futures at the Pardee Center

International Futures Help System

IP Equations: Overview of Threat Formulation

The threat that one country poses to others is a key concept in IFs. Unlike most variables in IFs, it is dyadic (actor country to target country). It is also different from most IFs variables in that it is a concept that has a probabilistic element in its implications for forward linkages. In fact, it is possible to think about threat as being the probability of military challenge or war, and that is the conceptualization in IFs. The database on Militarized Interstate Disputes (MIDS) was used in both conceptualization and initialization of threat in IFs.

Because of its importance, a substantial sub-project, sponsored by the Strategic Assessments Group (SAG) of the CIA devoted time to specifying the drivers of threat and the formulation for creating forecasts based on those drivers. Although none of the participants in that subproject bear ultimate responsibility for the treatment of threat in IFs, the model owes a substantial debt to the sponsors and participants of that sub-project.

Three key distinctions are important to understanding the overall threat formulation and its use in forecasting:

  1. Using history to initialize threat levels versus using predictive formulations. The argument for using data to initialize dyadic threat levels is obvious: data tell us about historic relationships between countries like India and Pakistan, often carrying information that is not available in a predictive formulation calculated across many dyads and not picking up the historic path elements of a particular dyad. Yet the argument for not relying too heavily on such dyadic data in forecasting is also obvious: the U.S.-Russian relationship has fundamentally changed since the collapse of communism and the break-up of the Soviet Union, so that a forecast based heavily on historic data would now be questionable. The IFs formulation provides forecasts that are rooted in data, but it allows the user to relax the ties to historic data over time.
  2. The complicated contribution of constant terms, switches, and variables. The single best predictor of conflict among countries historically may well be their physical proximity, with contiguous or geographically touching countries being much more conflict prone. But because contiguity is a constant, it is near useless in determining how the threat of overt conflict will change in the future. Somewhat similarly, territorial disputes are a near constant over time, but can be switched on or off. Quite differently, power levels and commitment to democracy fluctuate substantially over time. The different types of variables enter differently into the formulation.
  3. The contribution of power-based drivers and other drivers. For purposes of clarity of conceptualization and presentation of it, there is value in distinguishing between drivers of threat that have their roots primarily in state power and those, like democracy level, that do not.

Taking into account these important distinctions, the IFs formulation of threat has three key components. The first is a constant base term rooted in data and/or predictive theory. When it is rooted in predictive theory, the term draws on the constant and switch inputs to threat such as contiguity and territorial dispute. When it is rooted in data it represents recent history for the dyad as computed by the MIDs data. The second term is a delta or variable term rooted in power variables and the third term is a delta term rooted in non-power variables. The model user can use a parameter (wpthrconv) to determine whether the ultimate threat calculation (THREAT) should remain tied to the empirical initial condition (THREATIDATA), as modified by the delta terms, or should converge over time to a fully predicted threat formulation (THREATIPRED), again modified by delta terms.


It is useful also to be able to see a summary measure of the average world threat level (WTHREAT) over time. The sum of all threat terms is normalized by the product of the number of regions (NR) times the number – 1 (there are no non-zero and meaningful self-threat terms).


Detail on the component terms of this general formulation is available: