Forecasting the Demand for U.S. Ground Forces
A tool for comparing potential military outcomes over the next 20 years
RAND modeled thousands of scenarios exploring how the future demands for U.S. Army Forces could change depending on U.S. policy decisions and shifts in global order. In the tool below, you can explore these potential future scenarios by modifying baseline assumptions about U.S. policy and global order. You can then compare the results across military outcomes, such as the number of interstate wars, U.S. military interventions. and troops needed for military interventions. You can also compare the results across geographic regions.
This tool evaluates projected trends in demand for ground forces based on future U.S. policy, international conditions, and types of conflict and interventions
RAND research modeled hundreds of scenarios exploring how demands for U.S. ground forces could change in the future. Based on that modeling, this tool allows users to explore potential future scenarios by modifying assumptions about U.S. policy and key aspects of the international order. After defining a scenario, you can compare the results across different outcomes, such as the number of interstate wars, U.S. military interventions, and troops needed for military interventions. You can also compare the results across geographic regions.
This tool was completed in February 2019, followed by security review by the sponsor and the Office of the Chief of Public Affairs, with final sign-off in April 2022.
Worldwide
Interstate Wars
Field Definitions
U.S. Policy Assumptions
Our baseline assumptions for U.S. policy through 2040 are that military size and spending will slightly decline, existing U.S. force posture with remain in place, and existing alliances will remain in place. We modeled a number of other policy decisions and their projected outcomes based on the following definitions:
Military size and spending
- Baseline
- Slight declines in personnel (1%) and spending (4%) by 2040
- Increase to modernize
- Substantial increases in personnel (14%) and spending (57%) by 2040
- Reduce for restraint
- Substantial decreases in personnel (25%) and spending (20%) by 2040
Military force posture
- Baseline
- Existing U.S. force posture remains in place
- Prioritize defense of European NATO allies
- 11,000 additional troops in NATO Europe
- Prioritize defense of U.S. Asian allies
- 23,000 additional troops in U.S. Asian allies
- Combine prioritization of European and Asian allies
- 34,000 additional troops in NATO Europe, U.S. Asian allies
- Prioritize forces available for Mideast contingencies
- 5,000 additional troops in Kuwait, UAE
- Withdraw U.S. forces from overseas
- Remove all major concentrations of troops from U.S. overseas locations
Extent of U.S. alliances
- Baseline
- Existing U.S. alliances remain in place
- Continued expansion
- U.S. concludes formal defensive alliances with Georgia, Sweden, Finland, and India
- Sharp pullback
- U.S. withdraws from NATO, cancels defense treaty with the Philippines
International Order Assumptions
Our baseline assumptions for international order through 2040 are that there will be a modest increase in trade between states and a modest increase in democracies. We modeled a number of other options and their projected outcomes based on the following definitions:
Importance of International Trade
- Baseline
- Modest increases in trade between states of roughly 5%, with regional variations
- U.S. joins TPP, TTIP
- U.S. membership increases trade between U.S. and other member countries by 10 to 40%
- Fracturing of global trading system
- System dissolves into four competing regionally-based blocs, decreasing trade between most state by 10-20%
Prevalence of democracy
- Baseline
- Modest increase in number of democracies, with limited reversals
- Democratic reversals
- Several key countries revert to anocracy, including Brazil, Nigeria, and Poland
- Expanded democratic growth
- Increase in democratic regimes in sub-Saharan Africa, Southeast Asia. Limited democratic experiments in parts of Eurasia.
Methodology
The RAND tool for forecasting the demand for U.S. ground forces allows the user to estimate U.S. military requirements under a variety of assumptions. The model underlying these forecasts has four central components.
First, the model predicts levels of intrastate armed conflict in the current year. We fit a statistical model (a logistic regression model), using data from historical cases of intrastate conflict, to establish the strength of the association between intrastate conflict onset and its key determinants, including demographic factors, political institutions, and previous experience with conflict. The model uses these historical patterns to calculate the predicted probability that each state will experience intrastate conflict in the current year based on its projected characteristics. Each state is assigned either a ‘1' or ‘0' for ‘conflict' or ‘no conflict, where the probability of drawing a ‘1' is set to the predicted probability determined by the model. Next, the process is repeated for interstate war: we fit a model based on historical data to specify the relationship between interstate war and its key predictors, such as relative power, territorial disputes, and dyadic democracy, and we predict a probability of war for each pair of states. The model assigns either war onset or peace based on these predicted probabilities. The model similarly determines the likelihood of termination for all armed conflicts, providing a dataset of all ongoing armed conflicts, by state, in each year.
These forecasted armed conflicts provide opportunities for U.S. Army interventions. We consider three types of interventions. All states in the system may host a deterrent intervention if the risk of conflict outbreak is sufficient. States engaged in a current intrastate conflict or interstate war may be subject to an armed conflict intervention, where the U.S. will remain committed until the conflict ends. All U.S. armed conflict interventions might then transition to a stabilization mission when the conflict ceases, and any state that has experienced the cessation of armed conflict within the past 5 years may likewise host a post-conflict stabilization mission. In a process that parallels the forecast of armed conflict, the model predicts which states will experience an intervention in a given year, given factors such as U.S. alliance commitments, the economic status and strategic resources of the partner, and the potential threat facing the state, as well as which interventions will persist and which will conclude. The final component of the forecasting model provides estimates of troop size and type. We leverage historic U.S. military interventions to provide a categorization of U.S. interventions and their "typical" sizes and force mixes. We place projected future interventions into these categories based on their type and factors such as the level of threat or adversary strength, the size of the country, and its political relationship with the United States.
We repeat this process to produce forecasts of armed conflict and interventions for each year from 2017 to 2040, where each yearly forecast incorporates the predictions from the prior year. Since the forecasts are developed using predicted probabilities, the model's predictions about the onset and cessation of armed conflicts and interventions are partly probabilistic. That is, while states having higher predicted probabilities are more likely to draw a ‘1' than states with a lower predicted probability, they are not guaranteed to. This means the results can change each time the model is run. Additionally, since the components of the forecasting model are interconnected — for example, the model will only forecast the start of an armed conflict intervention if it has previously forecast the start of an armed conflict and U.S. deterrent interventions influence the propensity of future interstate wars in a region — slight changes in the model's early projections can ultimately lead to significantly diverging forecasts between different runs of the model.
To ensure the robustness of our forecasts, we iterate the entire process 500 times, and base our forecasts on both the average predictions across those iterations and their distribution. A single iteration of our model therefore involves the full simulation of each component for each year from 2017 to 2040, and each subsequent iteration then re-simulates the entire 2017 to 2040 period. We present the mean predicted trends in armed conflicts and U.S. Army interventions across these 500 iterations to demonstrate the most likely single outcome, as well as the 10th and 90th percentile projections to show the range of plausible forecasts.