Population projections rely on a multitude of assumptions. Reearchesrs must decide they expect mortality, fertility, immigration, and other factors to change at the same rate over time or grapply with complex algorithms in an attempt to capture changing population dynamics. It is impossible to know how populations will change inthe future, but we aim to equip scholars with the tools they need to uncover the inhertent assumptions in their calculations.

The Assumption-Relative DEMographic Information System (ARDEMIS) is designed to visualize the impact of these decisions on projections. What if the mortality rate is applied to individuals instead of the pupulation at large? What if we calculate rates of change on a smaller population? What if we apply the fertility reate before the mortality reate? None of these questions have a right answer, necessarily, but each will change the outcome.

ARDEMIS does not tell us “what will be,” but asks, “what if?”

Explore ARDEMIS

ARDEMIS is a product of the Modeling Religious Change project. Our partners involved in the creation of this tool are the Center for Mind and Culture, the Virginia Modeling Analysis & Simulation Center, the Center for the Study of Global Christianity, the Center for Modeling Social Systems, and New York University,

ARDEMIS visualizes the impact of assumptions on religious demographic projections. Traditional methods used do not support sensitivity analyses or estimates of the effect of measurement assumptions on projections. Computer simulation is more flexible and can help estimate the impact of assumptions.

Option toggle descriptions

Agent Count: This is how many artificial agents are present in the simulation model at the earliest modeled time point. Traditional demographic projections apply fertility, mortality, and migration rates to aggregates of the population. Simulations apply the rates as probabilities to individual agents. Starting simulations with fewer agents can introduce rounding errors (agents can’t have half a baby!), so it is necessary to maximize the agent count for demographic simulations. Learn more about these options here. 

Model Step: This indicates whether the simulation model operates in one-year increments or five-year increments. Most traditional demographic projections operate in five-year steps, but simulations are often designed in one-year steps to improve their realism. Converting between time steps requires a careful re-designing of the simulation to avoid introducing error. Learn more about these options here. 

Layout: This specifies whether the simulation is more “top-down” or “bottom-up” in its design. Top-down simulation models are deterministic. This means the exact number of agents experiencing an event (e.g. giving birth, migrating, or dying) is calculated at each time step based on how many agents are at risk of experiencing the event at that moment. A bottom-up simulation model, however, introduces more chance into the world. For each event, individual agents have a given probability of experiencing it, and each model will have slightly more or fewer agents experience the event. This approach is typical in simulations, and you can see a cone of uncertainty appear when the option is selected. Learn more about these options here. 

Split/Intuitive: This indicates whether demographic events in the simulation model operate in an order that best mimics traditional demographic projection methods (called “Split Fertility”), or in a more intuitive order. The “Split Fertility” option sets the order of events as: round one fertility, mortality, aging, migration, round two fertility. The intuitive order is: fertility, mortality, migration, aging. Learn more about these options here.