Hughes, et. al.: The Structure of IFs

Hughes, Barry, Anwar Hossain and Mohammod T. Irfan. (2004). "The Structure of IFs". unpublished IFs working paper on the Frederick S. Pardee Center for International Futures website: Denver, CO. http://www.ifs.du.edu/reports.htm.

This document begins by identifying a large number of global trends that are unfolding. It makes the claim that the International Futures forecasting platform will offer assistance in a variety of ways in light of these trends. IFs will be able to identify tensions in political, economic and environmental risk (3). It will also explore long-term trends. IFs explores the world through examining dynamic systems.

The authors then claim that three goals of forecasting are important in light of the above global trends and possibilities afforded by the IFs model. We should use these tools and trends to create and clarify global priorities. Also, we need to explore the world as it changes in a variety of directions through scenario analysis. Finally, examining our changing world through the perspective of an agent-class based approach allows us to look at points of contingency.

Four assumptions underlie the development of IFs: issues affecting human development must be explored globally and locally; goals of humans are increasingly being iterated; understandings related to human systems is growing rapidly; and the domain of human choice is broadening.

There are then four ways in that the IFs project was taken from an abstract understanding of global system interactions to a forma computer model. These involved the highlighting of different parameters necessary for formal forecasting, the selection of global systems and sub-models, the identification of both theoretical and philosophical foundations for modeling and the embrace of a technical approach to modeling. These four parts are further identified as this paper progresses.

M1-M4 represent, “…four design parameter decisions” (5). I1-I4 explores issues of interface vis-à-vis such a model.

M1: the geographic representation of IFs at the time of this writing was 164 countries with the plan to move to 182. Current versions do forecast 182 countries and are currently in the process of breaking those countries down into provinces.

The model uses a pre-processor to prepare the raw data for the base year forecast. The pre-processor structure allows the re-running of the base year to happen with relative ease. Also, the structure allows for a relatively easy addition of new countries into the model.

M2: The issue areas that are represented in IFs are demographic, economic, energy, food, environmental and socio-political. There has since been the addition of an education model and a health model.

M3: Time Horizon: the model runs to 2100.

M4: The model uses extensive sets of data bases in order to dynamically link sub-models. Simple extrapolation is adequate for some short-term forecasts, but dynamic linkage is required for mid or long-term forecasts.

I1: History of IFs availability.

I2: Usability.

I3: Interventions and Scenario Development: This has developed substantially to allow users to create and save their own global scenarios.

I4: Transparency: While IFs tries to be as simple as possible, any long-term model must be sufficiently complex to represent the dynamic linkages between different variables. Therefore, transparency is important. This is accomplished where possible through help systems, etc.

There is a discussion of the philosophy of the structure of IFs. Important aspects of this discussion involve the ability to represent both stocks and flows within a system. This is accomplished, in part, thorough an input-output matrix, as in the SAM. Additionally, the structure must be able to represent non-linear, or dynamic changes within the model. This is accomplished by imposing a disequilibrium causing lever that the model enacts and an equilibrium seeking structure. Therefore, equilibrium is never quarantined to one year, but rather always exists and is sought somewhere in the future. The ability of the model to accomplish this relies on stocks and flows: if trade balance, for example, is not achieved in one year, stocks can be held and this affects future year trade interactions. Also, agents and classes are represented through households, firms and governments. Some of this behavior is clearly market based, but other behavior is not (i.e., governmental).

Relative to other modeling platforms, the authors claim that IFs represents an “eclectic” approach (13). This is because IFs is neither fully an econometric, systems dynamics or optimization model. They authors describe the model as, “…structure based…agent-class driven, dynamic modeling” (13). While it pays close attention to stocks and flows, it isn’t a system dynamics model because it represents some systems embedded in the structure of other systems. Also, the model is not a micro-agent model either because it represents classes. Finally, it isn’t fully econometric but it does rely heavily on data.

The authors also posit that there should be an additional question posed of modelers that tends to be left out: how easy is it to intervene in your model? The IFs model is designed to allow users to quickly and easily make parameter or variable changes within the model.

Demographics:

The demographic sub-model of IFs relies on stocks and flows being represented by age/sex cohorts similar to the ones used the UN or US census. These are then used to determine births within a year. However, the TFR must be determined by some metric derived from other areas of the model. Simply regressing different data sets allows patterns to emerge. In Table 3.1, the authors present a few different estimations of TFR using GDP per capita at purchasing power parity, total education, female education and contraceptive use. While they find a higher adjusted R-squared with one set of variables, they opt for a second estimation set that gives less priority to the education component, as this is not fully robustly forecasted in the model at the time of this writing.

Also, the authors point out that the relationship between TFR and GDP changes over time. The curve essentially shifts “down” showing a relationship where levels of TFR are associated with lower and lower levels of GDP. This is incorporated into the model by using a time dependent factor that decays over time due to saturation effects.

Also, longevity is determined by the model. It is found that GDP highly correlates to this, though there are exogenous factors that may speed the increase in life expectancy, like technology, etc. These are forecasted in the model by once again adding a time lag factor.

Finally, migration is forecasted. Large migrations throughout history have not at all been sustainable. Therefore, there is an upper limit placed on migration patterns so that the “maximum inflow or outflow of population is reduced over about 20 years to one percent of a country’s population” (21). Also, because it is typically largely populated countries that experience migration, the flow is capped if there is an overall decline in one year of more than 0.5%.

Economics:

The best approach to understanding the robust IFs economic sub-model is taken in three steps: first, understand how goods and services are produced. Secondly, broaden your attention to understand how the larger goods and services market interacts with consumption and exchange. Thirdly, the social accounting system must be explored to understand how the stocks and flows in the above systems are represented.

Production of Goods and Services:

The model uses a modified Cobb-Douglas production function. The Cobb-Douglas production function states that total output of an economy can be characterized by the interaction of labor multiplied by an exponent representing elasticity of labor time capital multiplied by an exponent representing elasticity of capital. The “elasticity of capital” and the “elasticity of labor” measures how quickly output will respond to a change in either labor our capital.

This production function is modified with an endogenously created multi-factor production variable. The authors note a variety of others who have determined that the Cobb-Douglas function is applicable, but limited. Some have pointed out that up to 50% of production increase comes from technological increase.

There were two ways that a MFP could be introduced into the model, one simple, the other complex. The more complex, endogenously derived model was selected, as it provided the opportunity for a more thoroughly responsive model to policy interventions. The derivation of this value involves five categories:

The convergence base: This is the base rate of MFP growth. It involves a convergence principle whereby developing countries converge on the technological superiority of a hegemon, assuming that the developing countries have passed some threshold of development. The base rate of the technological leader is a value pulled out of the air by the developers, but that can be changed to represent technological waves of advance and stagnation.

Knowledge creation and diffusion: Government spending on R&D.

Human capital quality: This addition to MFP involves spending on both health and education relative to GDP. The estimation is that a 1.5% increase in government spending on education equates to a 0.3% increase in economic growth.

Social capital quality: This tracks the effect that economic freedom can have on GDP growth. This was accomplished by looking at economic freedom and GDP cross sectionally to determine a possible relationship.

Physical capital quality: This tracks the relationship between energy supply availability and economic growth. If world energy price increases substantially, then much capital stock depreciates. IFs computes this price change by looking at the previous year’s price.

The Goods and Services Market:

This is the demand side of the model. It has embedded inside of it the supply, or production side of the model: the production function. This aspect of the model is equilibrium chasing, which means that it is never fully in equilibrium, as an economy would never fully be in equilibrium. It is able to do this through sectoral and country based inventories. These allow aspects of the economy that are not in equilibrium to rest year to year, as well as providing signals which can move both the supply and demand side towards equilibrium. This is accomplished through the setting of desired levels of stocks.

Total consumption is directly linked to income, which is based on labor earnings, returns on capital and transfer payments. These features are all addressed in the SAM discussion. Additionally, household consumption and savings is responsive to an interest rate. Consumption also takes place by sector, as households with higher levels of income tend to consume more services.

The size of the government’s share of involvement in the economy also continues to grow with time. This interaction is capped over time. IFs also does not represent trade bilaterally, but rather as a pooled feature.

IFs is equilibrium seeking in its economic sub-model through maintain a balance between supply and demand. This is accomplished through prices and stocks. Prices are mediated by elasticities and control supply and demand domestically. Stocks drive investment by destination. This equilibrium seeking behavior is controlled by a PID. “A PID-driven adjustment process responds proportionately to the integral of the error (the stock of discrepancy) and to the derivative of the error (the change in stock term)” (30).

The production function is the driver of both supply and, in many ways, demand. It accomplished this by driving consumption (through incomes). Consumption can change the dynamics of the goods and services market, but only at the margin.

The Social Accounting Matrix:

The SAM is embedded in the Goods and Services Market. It “…tracks and dynamically represents the financial stocks (assets and liabilities) and flows associated with key agent-classes” (31-2). This feature is an accounting tool for looking at how different sectoral and inter-state financial flows create abundance or dearth of certain stocks. For example, if a population is aging and a country doesn’t spend enough resources on pensions, possible adverse effects can be seen through the SAMs.

The internal SAM is a tool for looking at the distribution of stocks and flows among different sectors of the economy, households, firms and the government. The SAM is, once again, equilibrium chasing. It also does not represent a large lever that allows for policy interventions within the model, but is rather a tool for tracking the distribution of stocks, flows, and therefore production, a feature that can have effects on the model more broadly (see the MFP in the Production Function).

There is also an inter-state SAM which tracks the relationship between FDI and different countries as well as international debt. Once again, the SAM looks at assets and liabilities and uses signals from stocks to chase equilibrium over time. This is also accomplished by states acting as either agents of providing FDI or demanding FDI. These levels are then used to produce flows of capital gains to both asset country and liability country.

Energy:

Stopped here to focus on the Economic sub-model in other literature.