Forecast Development & Documentation

ARC used an integrated system of three sophisticated models to develop the forecasts that support the Atlanta Region’s Plan. The REMI model (details below) produces a regional forecast, which is then disaggregated by a small-area model called PECAS, which then feeds distributed socioeconomic data into an activity-based travel model (ABM).

This general modeling flow is illustrated in the diagram below, with more details by model set, provided in the sections that follow.


The REMI model (Regional Economic Models, Inc.) is a very widely used regional economic policy analysis model. The model is used by government agencies on the national, state, and local level, as well as by private consulting firms, utilities, and universities. REMI is a structural economic forecasting and policy analysis model. It integrates input output, computable general equilibrium, econometric, and economic geography methodologies. The model is dynamic, with forecasts and simulations generated on an annual basis and behavioral responses to wage, price, and other economic factors.

ARC evaluated and implemented REMI from 2007 to 2008. Policy Insight was the specific model used from 2008-2010. TranSight (TS) has been the working model since 2010 and was used for the previous forecast series (Series 13); TS enables direct use of travel model output and is built on a Policy Insight base. In the spring of 2015, ARC used the REMI model to generate an updated set of regional control forecasts. A Technical Advisory Committee (TAC) of local economists reviewed the forecasts and a general agreement on a reasonable forecast was reached in April. We will post more details soon about that process. ARC staff then began allocating the regional controls to smaller geography, with PECAS.

For more information check out the Download Center to the right of the page for links to Policy Insight and TranSight documentation.

REMI Resources

PECAS for Small Area Forecasting (Land Use Allocation)

ARC reviewed state-of-the art land use models, to allocate the forecast population and employment totals to small areas, between 2007 and 2008 and selected PECAS (Production Exchange Consumption Allocation System). PECAS’ main purpose is to simulate the future location of activities (industries, households and government), and the development of space by developers, for both forecasting and policy analysis. It has been used in the conformity process for the first time in 2015.

The ARC PECAS model includes the two standard PECAS modules: the Activity Allocation module (AA) and the Space Development module (SD). AA follows an aggregate approach and represents how and why industries, government and households choose to locate in different zones or locations in the region. SD follows a microsimulation approach and simulates development at the parcel level, taking into account developers’ profit-motivated behavior as well as land and market characteristics. These two modules interact with each other, and both also interact with the Atlanta transport model by providing it with land use data. The transport model, in turn, provides an indication of travel conditions for use in AA.

PECAS output was not used as a “black box,” as an extensive outreach process consulted with local jurisdictions on the forecasts over spring and summer of 2015. We met with planners and economic developers to discuss key areas (and the extent of growth in those areas) for small-area allocation. More details on that process will be posted soon.

See the links at right for much more information about PECAS. The first document discusses the theory behind PECAS. The second details PECAS development in Atlanta from 2008 through 2014. The third documents the most recent model refinements and application.

PECAS Resources

Activity-Based Model (ABM) for Travel Modeling

The Activity-Based (ABM) model is an evolution from the previous 4-step travel model, and has been fully implemented over the past few years. The model addresses both household-level and person-level travel choices including intra-household interactions between household members, as compared to other more simplified Activity-Based Models where all travel choices are modeled at the person level, independently of choices made by other household members within the household. The Activity-Based Model reflects and responds to detailed demographic information (from REMI and from PECAS), including household structure, aging, income changes, and other key attributes.

Visit our Activity-Based Modeling page for more information.