How economic models work

An operative economic model consists of a set of mathematical rules that govern the responses of economic actors to changing prices. The models used for ILUC modelling are often described as general or partial equilibrium models. Equilibrium modelling involves constructing a baseline that is in economic equilibrium (supply matches demand and prices are in balance so that no industry is making losses or excess profits). Some exogenous change¹³, sometimes referred to as a ‘shock’, is then made to one or more numerical values in the model (for example increasing biofuel demand) and the model is allowed to find a new balance, generally through a computationally intense iterative process. Figure 3 provides a much-simplified schematic representation of the sort of adjustments that occur as an equilibrium model looks for a new equilibrium following a shock.

13 Exogenous means originating outside of the model structure, i.e. a change made directly by the modeller. This is contrasted with endogenous changes, which emerge from the numerical rules of the model.

Economic models for assessing ILUC

Figure 3 Schematic representation of the cascade of adjustments when finding a new model equilibrium

Depending on the level of regional disaggregation present in a model, a shock could be applied at the national level (as in modelling of corn ethanol mandates in the U.S.) or regional level (for instance the EU). The location of a shock can be important because land use change dynamics can differ between regions. In the case of soybean production, for example, while soy is grown in both North and South America, most soy-associated deforestation occurs in South America. We might therefore expect that there will be differences in land use change emissions outcomes for the creation of new soy demand in the United States versus Argentina, even if the markets are somewhat connected.

While the basic premise of adjusting to find new equilibria in response to shocks is the same for all equilibrium models, the mechanisms the models use to decide how to reach a new equilibrium differ considerably. The most fundamental difference is between ‘partial’ and

‘general’ equilibrium models. Partial equilibrium models include only a few sectors of the economy but do so in considerable detail, whereas general models include the whole economy but do so in less detail.

Partial equilibrium models tend to work in terms of physical quantities such as tonnes of material produced and area of land used, whereas general equilibrium models tend to work in more financial terms, for example considering land in terms of rents rather than physical areas. Prices allow physical units to be converted to and from financial units, but it is noteworthy that using financial flows as a functional unit in general equilibrium modelling can lead to counter-intuitive outcomes – for example, if explicit constraints aren’t introduced in general equilibrium models, then quantities like mass or area may not be preserved.

Partial and general equilibrium modelling are not mutually exclusive philosophies, and it is possible in principle to use the results of one approach to inform the other. For example, partial equilibrium modelling tools that have more detailed characterisations of agricultural decision making may be used to calibrate the agricultural sector responses in a general equilibrium model. Similarly, results relating to the broader economy from a

Demand for

general equilibrium model (for example relating to the fossil fuel rebound) could be used to apply adjustments to agricultural results from partial equilibrium modelling. One exercise coupling general and partial modelling frameworks is presented by Britz & Hertel (2011), which couples a partial-general modelling framework using the CAPRI (Common Agricultural Policy Regionalised Impact) model for the agricultural sector in Europe and the GTAP (Global Trade Analysis Project) model for the rest of the global economy. In this work, the EU agricultural market behaviours predicted by CAPRI are mathematically integrated into the broader GTAP framework. This is shown to result in significant differences in the location of predicted production and land use changes, and allows more detailed results to be reported within the EU.

While Britz & Hertel (2011) firmly endorse the idea of coupling models in this way, the practice has not been widely adopted in ILUC analysis. This may be explained in large part by the challenges associated with coordinating work between different modellers and modelling teams. The economic modelling tools used for ILUC modelling were originally developed to assess other questions, generally relating to analysis of agricultural policy and of the impacts of changing trade rules, and continue to be used for these purposes.

Despite its importance in biofuel policy, ILUC analysis is not the only or even main concern of the communities that develop the various economic models, and the resources available for ILUC modelling are limited. In the EU, ILUC modelling has been undertaken under contract to the European Commission, and bids must be competitive if they are to succeed. The overheads associated with coupling different modelling frameworks together are significant, and therefore unless contracting authorities make it an explicit requirement, modellers are likely to prioritise the development of a single modelling system over the integration of separate systems.

The details and mathematics of economic model structure are necessarily complex, and therefore many authors have tried to simplify the explanation of ILUC modelling by focusing on the most important underlying factors that dictate the amount of land use change predicted and the associated GHG emissions. Woltjer et al. (2017) presents a decomposition approach that splits ILUC modelling into six basic steps:

1. Identify the gross land demand, i.e. the amount of land necessary to grow the amount of feedstock required for biofuels at typical yields;

2. Area requirement is reduced if co-products are returned to the market;

3. Area requirement is reduced by reductions in demand for agricultural commodities for other uses (primarily food, feed, pharmaceutical);

4. Area requirement is reduced if yields of crop production are increased;

5. Area requirement is affected (up or down) by relocation of crop production to areas with different yield potential;

6. GHG emissions are assessed based on the land use changes predicted.¹⁴

In practice, economic models address all of these steps at once, and the net changes the system are an aggregate across thousands of productivity shifts, demand changes and production changes across the modelled system. Identifying broad themes is useful

14 We note that the decomposition as presented in (Woltjer et al., 2017) relates to land use area only, we have added to the list the step of conversion of area changes to emissions.

Economic models for assessing ILUC

because it provides a more narrative way to engage with the model outcomes, and to compare modelled behaviours with observed behaviours. For example, when considering the second step (impact of co-products) we can look at evidence from livestock feed markets and from analysis of the properties of different feeds, and then ask whether the outcomes in a given model seem consistent with what our understanding of feed markets leads us to expect (this is discussed further in section 3.4.2).