Table of Contents
- 2.Economic Modelling
- 3.Ecological Modelling
- 4.Integrated models
Suggested citation: Rennings, K. (2013). Modelling. Economic modelling. LIAISE Toolbox. Retrieved date, from http://beta.liaise-toolbox.eu/ia-methods/modelling.
Modeling in the context of IA describes numerical and quantitative techniques for analyzing the impacts of policy proposals simulated with computer software. They represent a form of quantitative in-depth analysis of the most significant impacts of a policy initiative. The main objective is to provide simplified versions of complex real-world phenomena in order to infer statements on the system under consideration. Models are used to simulate the likely impacts of selected exogenous variables (e.g. policy changes or tariff reductions) under various circumstances on different markets (commodity markets, financial markets, labor markets, resource use, emissions, etc.) or systems as endogenous variables explained by equations within the model.
Each model is designed for a specific purpose or area of analysis (e.g. policies under the Common Agricultural Policy) and is based on a set of assumptions that is suitable for this purpose. These assumptions need to be made transparent for users in every modeling exercise since they have relevant effects on modeling results and are also relevant for interpretation of these results. They are usually embodied in large-scale computer algorithms, requiring data sets tailored to the particular problem area.
Combination with other methods
In the following sections different areas of modelling are described including the various sub-types of economic modelling as well as approaches in ecological and integrated modelling.
Lead Editor of this section is Klaus Rennings
Definition & Objectives
Economic-environmental modeling comprises a lot of different models, for example input-output-models, microeconometric models and computable general equilibrium (CGE) models. The following text will give a brief overview of the features, strengths and weaknesses of each model (based on descriptions from Cambridge Econometrics (Pollitt, Hector and Chewpreecha, Unnada, 2009, Revision of the IA TOOLS Model Inventory - Final Report, Cambridge Econometrics, Annex 2).
The models described on this section are:
- Input output models
- Computable general equilibrium (CGE) models
- Econometric models
- Single sector models
- Microsimulation models
- Integrated models
At the end of the page we will summarise the main aspects of the most important models in a table.
Input-output and Multiplier analysis
An alternative approach to large-scale economic modelling is input-output (IO) and multiplier analysis. This is based around economic input-output tables, which indicate the values of purchases between economic sectors in a particular year. Thus they can be used to build a picture of supply chains. In addition, Supply and Use tables (SUTs) link the components of final demand (consumption, investment, etc) to economic sectors, allowing a picture of the whole economy to be built up.
Input-output tables are usually available at the national level, although this is based on data availability rather than on theoretical grounds. Some countries produce regional IO tables and national tables may be aggregated to obtain a European IO table.
The main advantage of using input-output analysis over a fully-specified modelling approach is its flexibility. While the sectoral disaggregation in most economic models is fixed, input-output analysis can be applied to any sector that is defined in the input-output table. In fact, many economic models with a sectoral dimension include input-output tables in their parameters, although sometimes in a more aggregate form.
The second main advantage that input-output analysis has over a modelling approach is its relative simplicity, which means results are easier to interpret, and relatively few resources are required.
The main disadvantages of input-output analysis are that the assumptions required to simplify the analysis may not always hold. The main assumptions are: Supply chains are fixed, with no economies of scale.
Shares of inputs met by domestic production and imports do not change
Input-output analysis is described as “static”, meaning that it does not take into account changes over time. For example, it is assumed that technologies and prices do not change; therefore input-output analysis is usually only considered appropriate for the year for which the input-output tables are available. Unfortunately these are often several years in the past.
A common and easy way to interpret the output of input-output analysis is the economic multiplier. This tells the user what the impact will be of an increase in economic activity in a particular sector and can easily be derived from the input-output table.
For example, if a national government spends an additional €1m on health care, there will be benefits for both the health industry (direct impacts) and its suppliers (indirect impacts), such as equipment manufacturers and pharmaceuticals companies. If these increases total €2m in additional output then the multiplier, defined as the sum of the direct and indirect impacts over the direct impacts, is 2.
The most common types of multipliers are:
Type I economic multiplier: This is based on the goods and services that industries purchase from each other, as defined in the input-output table.
Type II economic multiplier: This is similar to the Type I multiplier but also includes the effects of employment and household incomes (e.g. if the health care sector employs 100 extra people, they have incomes to spend on household goods). A Type II multiplier is always higher than the equivalent Type I multiplier, because it allows household spending (part of final demand) to increase, while a Type I multiplier assumes this is fixed.
Employment multiplier: This is based on the same principals as the economic multipliers but measures impacts in numbers of jobs, rather than monetary terms.
Analysis over time
As published IO tables are fixed over time they are often used for analysis of the past. Attempts have been made to allow IO tables to change dynamically over time, although this is usually within a fully-specified modelling framework. Examples include adjusting coefficients over time to reflect technological change, changes in relative prices, changes in investment patterns and links to physical models where IO coefficients are adjusted to match patterns in demand for energy and raw materials. This is more complicated than basic IO analysis and is an area of ongoing research.
Links to international trade
The input-output tables published by Eurostat (and most other statistical agencies) separate demands met by imports and those met domestically. This provides links to international trade which can be expanded on using bilateral trade data (telling the user where imports come from). Combining these data sets can give a basic assessment of trade impacts but has more recently been applied to non-economic analysis, for example estimating the scale of “indirect” carbon emissions caused by shifting production overseas and embodied materials in imported goods.
Relationships with fully-specified economic models
The basic IO/multiplier structure discussed above may be expanded upon by making the components of final demand endogenously determined. The Type II multiplier is in fact the first stage of this by allowing household spending to change; the other components of final demand, price and labour market equations may be added sequentially until the model is “fully-specified” (usually defined as all the economic relationships in the national accounts being covered). Most of the models in the IA Tools inventory that include a sectoral disaggregation make use of IO tables to link the sectors together.
Extensions to include non-economic analysis
Recently, basic input-output analysis has been extended to cover non-economic topics, in particular energy use and demand for raw materials. These follow the same basic ideas as economic analysis but use physical units rather than monetary values. The aim of this analysis is to determine which sectors contribute the most to energy use (and implied atmospheric emissions) and material consumption, when both the direct and indirect effects are taken into account.
Example of a study using input-output analysis:
Exiopol (DG Research):
This is an FP6 project looking at creating “environmentally-extended” input-output tables to include external environmental costs in the traditional input-output framework.
Computable General Equilibrium (CGE) models
CGE models calculate a vector of prices such that all the markets of the economy are in equilibrium, implying that resources are allocated efficiently. They are based on economic theory and theoretical coherence (i.e. the Walrasian representation of the economy). Behavioural relationships in the models are calibrated with mathematical methods, typically using a single base year of data. The GTAP database provides a common data source for many CGE models.
CGE models can be static (comparing the situation at one or more dates) or dynamic, showing developments from one period to another. They can be used for analysing the long-term effects of general economic policies like public finance, taxation and social policy.
The strength of CGE models is their internal consistency; they allow for comparative analysis of policy scenarios by ensuring that in all scenarios the economic system remains in general equilibrium. CGE models integrate micro-economic mechanisms and institutional features into a consistent macro-economic framework and include clear feedback mechanisms between sectors. All behavioural equations (demand and supply) are derived from micro-economic principles (i.e. utility-maximising individuals, profit-maximising enterprises).
Since CGE models are calibrated to a single base year data set, their data requirement is limited even if disaggregation is high. Ready-made global databases exist so that, unlike other modelling approaches, the models do not require customised data sets. This means it is often possible to build a customised CGE model for a particular purpose and this allows greater flexibility for the evaluation of distributional effects across countries, economic sectors and other parts of the economy.
The main weakness of CGE models is their theoretical construction and reliance on assumptions. Care must be taken when interpreting results as model assumptions can heavily influence outcomes. For example, as CGE models assume that all “best available technologies” are already in use and resources are being used efficiently, the costs of reducing CO2 emissions may appear higher than results from other modelling approaches.
The assumption of equilibrium means that, unlike econometric models, CGE models are not generally considered suitable for short-term analysis or forecasting because, although they have internal consistency (through theoretical relationships) they often lack external consistency (with real world data and behaviour).
For analysis of a particular sector, it is unlikely that CGE models will be able to provide the same level of detail as single-sector models. They do, however, provide detailed links between sectors.
Example for CGE Models
We will use in the following a CGE model application for illustration by using the paper “The EU Decarbonisation Roadmap 2050 – What way to walk?” by Hübler and Löschel (Hübler, Michael und Andreas Löschel (2013), The EU Decarbonisation Roadmap 2050: What Way to Walk?, Energy Policy 55, 190-207).
This study assesses the sectoral effects and the overall costs of decarbonisation given certain policy scenarios.
Process & Method
Hübler and Löschel use a detailed computable general equilibrium (CGE) analysis called PACE. It is a multi-sector, multi-region CGE-model of global energy production, consumption, trade and use. This study extends the PACE model from 2020 to 2050.
The world economy gets described by several equations within the model. In PACE there are two classes of conditions characterizing a perfectly competitive equilibrium: zero-profit and market-clearing conditions. To validate these conditions the model assumes profit maximization in firms, utility maximization of consumers, constant returns to scale in production and homotheticity of consumer preferences.
Each region is described by a microeconomic consumer (chooses bundle of consumption goods which maximizes utility given preferences and budget) and producer (chooses bundle which maximizes his profit). The model includes all types of taxes, but excludes costs due to climate damages.
After a detailed description of the model used for their study Hübler and Löschel create seven different scenarios. For each scenario they analyze future outcomes.
At last the results of the scenarios are discussed and evaluated.
Combination with other methods
PACE model is implemented in MPSGE a subsystem of GAMS using PATH for solving a Mixed Complementary Problem.
Types of data needed
The authors took the data for the base year data from GTAP 7 database and future economic development data up to the year of 2050 from International Energy Outlook 2008/2010 (regional data for energy consumption and carbon emission, assumptions on GDP development, fossil fuel prices and other factors, data takes population growth and technical progress into account). Data for substitution elasticities were taken from the GTAP database as well and certain CES elasticities of substitution between production factors are based on “Estimation of Substitution Elasticities for CGE Models” by Okagawa and Ban.
Strengths & weaknesses
The Strengths of this model are:
1. Identifies detailed sectoral results which are determining for the implementation of emission reductions in on the industry level
2. Combines high sectoral resolution with bottom-up representation of energy sources
3. Extends time horizon from 2020 to 2050 and creates a new reference scenario that includes existing EU-Policies.
4. Focus on comparison of different policy designs implementing the same emission targets.
Econometric models are based on empirical relationships and are therefore developed using coherent (usually time-series) data sets. The parameters of the equations are estimated with formal econometric methods. The original econometric models were built for short-term macro-economic forecasting, but can also be applied for policy analysis.
The main strength of econometric models is the validation of their equations by statistical methods, meaning that results are determined by observed, rather than theoretical micro-economic, relationships.
Unlike CGE models, econometric models are generally suitable for short and medium-term forecasting and analysis.
The most commonly cited weakness of econometric models is the underlying assumption that relationships estimated using historical data may be used to predict future behaviour. In particular where there are large structural changes, historical relationships may break down. A particular branch of this criticism relating to policy analysis is known as the Lucas Critique, and states that behavioural relationships may change endogenously as a result of government policy, something which the fixed coefficients of an econometric model are usually unable to account for.
The second weakness of econometric models compared to the other model groups is the data requirements; for time-series models large data sets are required, and if these data are unreliable this will be reflected in the model parameters.
Finally, macro-econometric models are sometimes accused of lacking any micro-economic foundations (sometimes called micro-consistency). This accusation may not apply to all econometric models, some of which incorporate and estimate micro-economic behavioural relationships, but those that do not are effectively not grounded in any underlying principles of theoretical economic behaviour.
Single-sector models are built to analyse one specific policy area. The general structure of these models is determined by the particular features of the sector in question. Usually, in a manner similar to that used in CGE models, this sector is assumed to be in equilibrium, so these models are also referred to as partial-equilibrium models.
The strength of single-sector models is that they focus only on one policy area or economic sector and are thus able to offer a relatively high degree of disaggregation and a detailed representation of specific economic and institutional factors. This may also allow the application of more specialised modelling techniques, such as factors accounting for non-linear relationships.
Transport modelling provides a good example of how single-sector models are able to go into much more detail than other modelling approaches. The most detailed transport models include a geographical representation of Europe’s transport network, for each mode of transport. In contrast the more general frameworks may treat transport as a single sector.
Partial models are an appropriate tool if the focus of policy analysis is on a specific sector and if feedbacks to the economy or other policy areas can be ignored. However, if these feedbacks cannot be assumed to be zero, the analysis will need to be supplemented using an alternative approach, such as a CGE or econometric model.
For example, it would be reasonable to assess the impacts of building a new road on traffic volume using a single-sector transport model, if it could be assumed that economic growth was not affected. However, if more efficient transport links led to higher economic growth this could in turn increase traffic volume; a single-sector transport model could not be expected to include this effect.
Introduction and relative strengths
Microsimulation models are a specific type of single-sector model. They are usually based on micro-economic data and are typically used to assess the impacts of various policy changes on small units such as individuals, households or firms. Like other single-sector models they are able to incorporate detailed factors that are relevant to the area of analysis, such as income and expenditures, age, family status or profits. By using a representative sample, micro-level changes can be aggregated in order to estimate macro-level effects.
Microsimulation models are tools used for policy recommendations: During the last ten years they have been widely used particularly in empirical tax policy analysis in several European and OECD countries. Typical applications of tax-benefit models are, for example, the calculation of the distributional effects of different tax-benefit policy scenarios (i.e. the calculation of the tax payable, identification of individuals who would benefit or be negatively affected by a specific policy, etc.).
Microsimulation models have two main weaknesses: The first is the lack of feedback to other parts of the economy (similar to other single-sector models). The second is that the data are often drawn from surveys so there is the possibility of sampling error influencing the final model outcomes.
Table of suitability of approaches to policy questions
Single Sector Models
Micro Simulation Models
Input- Output Models
- able to produce very disaggregated results as models require only one (base) year of data
- provide detailed information on the policy impact of a certain variable
- allow for consistent comparative analysis by ensuring that the economic system remains in general equilibrium
- relationships econometrically estimated from historical past
- based on detailed time-series data
- rely less on economic assumptions
- capture the process of dynamic adjustment and structural changes
- focus on one sector and therefore provided high degree of disaggregation within the sector
- very disaggregated - usually focus on individuals, households or firms
- can be used to draw conclusions that apply to a higher level of aggregation
- flexible: can be applied to various topics
- flexible: analysis can be applied to any sector that is defined in the IO table
- relatively simple: results are easier to interpret, and relatively few resources are required
- transparent to use in policy decision making
- lack of historical validation
- heavily rely on economic assumptions
- cannot be used for forecasting
- cannot cope well if there are major structural changes
- require detailed time-series data
- not suitable for short-term analysis
- relatively inflexible – changes to the models require large resources
- unable to capture the feedbacks between other sectors and the sector in question
- unable to capture the effects in other markets
- require very disaggregated detailed data
- data sampling error e.g. data may not include information on all actors of interest
- not all resource flows may be represented in the model
- simple treatment of the model
- rely heavily on assumptions
- can only be used for static analysis as the model doesn’t take into account changes over time
- macro-economic relationships are consistent with micro economic theory
- system is always in long-run equilibrium
- historical behaviour - relationships hold in the future
- feedbacks between the rest of the economy are ignored
- factors outside the sector are exogenous
- samples are good representatives of more aggregate level
- assumptions based on the nature of IO tables e.g. supply chains are fixed, no economies of scales, shares of inputs met by domestic production and imports do not change
- capturing one-off and long-term effects from shocks
- capturing short/medium/long-term effects from shocks
- detailed analysis of a specific sector
- analysis at detailed micro level e.g. tax effect on income distribution
- multiplier analysis
- analysis of supply chains, how industries are related
Ecosystem models as mathematical computer-based models synthesize information about different components of an ecosystem. They are designed to predict and quantify changes in dynamic ecological systems in response to external influences.
In the context of an impact assessment they can
- Increase the understanding of a policy’s or policy options’ impacts on an ecosystem;
- Represent the trade-offs related to changes in an ecosystem which are relevant to different stakeholders;
- Support monitoring and evaluation of a policy’s effects on the region represented in the model.
- Visualise the different potential outcomes of a planned policy on an ecological system.
Here we define integrated models as ones that combine other relevant models together. Usually these models are the results of research projects that consist of a series of modules. The integrated model links these modules together, providing a tool that can be applied to assess impacts in several policy areas simultaneously.
For example, a project concerning the impacts of pollution may have one module concentrating on the impact on human health and another module concentrating on economic impacts. The integrated model for this project links both modules together and provides users with a tool that allows them to directly analyse the impact of pollution on both health and on the economy within a single framework.
Another example of an integrated model is one that links similar types of models in each country to provide analysis of an aggregated region.
The integrated model provides an analytical tool that maintains the consistency of its sub-models while allowing the intrinsic diversity that is specific to the sub-models. The integrated model has the benefit that it is able to include feedbacks that would be missing from running separate analyses.
The models provide relatively detailed analysis and allow users to study wider impact horizons. This is because the models draw on different areas of specialised expertise and combine them together.
Despite its strengths, an integrated model requires a great deal of resources to construct. This is evident from the limited number of integrated models. The difficulties lie in both theoretical approaches, with models that may be based on different assumptions, and the practicalities of linking different sets of computer code, model classifications, etc.
Each of the sub-models is also subject to its own strengths and weaknesses.