It is well-known that partial equilibrium (PE) and computable general equilibrium (CGE) models have structural differences both in terms of the data and the behavioural elements (i.e., explicit or implicit elasticities), which can generate divergent results, whilst previous precedents in the literature even show that CGE and PE can generate contradictory findings for the same scenario. Although this is well recognized within the modelling community, in the policy arena it can often be hard to reconcile the findings of both models when presenting a consistent story line for a given policy reform.
In the past, previous work commissioned by the Joint Research Centre (JRC), Seville, on behalf of DG Agri, forged a ‘soft’ model linkage (Helming et al., 2010; Nowicki et al, 2006, 2009), such that both models generate a mutually consistent storyline. Typically, a soft linkage is driven by a more ad hoc assessment of the overall results (i.e., are the models broadly telling the same story?), whilst one plays to the strengths of each model to serve as a source of input to the other. For example, the CGE model, with an explicit or endogenous treatment of factor markets, world trade and macro aggregates, could conceivably be used within a PE model. Similarly, the sectoral detail and econometric foundation in supply response which serves some PE models well could be employed to assess and improve the veracity of the CGE model results.
Under the auspices of project 154208-2014-A08-NL, entitled, “Scenar2030, parameters and model chain preparation”, the Economic of Agriculture unit of the JRC requested a further look at this issue to better understand the merits of different model linkage options. More specifically, as part of technical specification for task 5 (‘preparation of model chain’), two forms of model linkage, broadly labelled as ‘soft’ and ‘hard’ linkage are considered.
The advantage of the soft approach is that it is relatively straightforward to implement in terms of the necessary modelling modifications. On the other hand, the ‘soft’ approach adopted in the Scenar2020 project through linkage of variables was, as noted above, implemented more on an ad hoc basis, rather than following a systematic framework. Thus, subject to the prejudices of the model scenario (i.e., the scenario design, the type of shocks etc.), the use of variable linkage could conceivably vary considerably. This, in turn, has led to the alternative choice of a ‘hard’ linkage which seeks to forge a union between the structural or behavioural elements of the model (see, for example, Britz and Hertel, 2011; Pelikan et al., 2015). Whilst this approach is intuitively appealing because it follows a very specific methodological approach, it requires considerably more modelling expertise to implement, whilst the potential robustness of the two models being linked is, at the current time, far from certain. A fuller exposition of the hard linkage approach is given in section four below with some reflections of its potential suitability for advanced policy analysis using the MAGNET model.
For the purposes of the current (tentative experiments), in section two, a ‘test bed’ study is described, which considers a more systematic class of ‘soft’ model linkage between two well-known and respected models from the iMAP platform, namely, the Common Agricultural Policy Regionalised Impact (CAPRI) PE model and the Modular Applied GeNeral Equilibrium Tool (MAGNET) CGE model. In CAPRI, a standard CAP baseline is run, whilst in the MAGNET model, two specific experiments are implemented. The first runs a standard CAP baseline in the MAGNET model, whilst the second implements the same baseline shocks with the inclusion of model predictions of output taken from CAPRI. The aim of the exercise is simply to ascertain the extent to which the MAGNET model results (section three) diverge between the two experiments and assess the degree of compromise required in MAGNET to accommodate said changes.
Clearly, if considerable divergences are found, and one considers that the CAPRI sectoral output results are superior, then this could potentially warrant the need for a more extensive research effort to provide a systematic, theoretically consistent and scientifically rigorous approach to model linkage for future policy impact assessments.