Learning portal - Modelling approaches

• Existence of well-tailored regional or macro models aligned with the needs of an ex post evaluation.
• Updated information and data reflecting an actual situation during the implementation of a given programme.
• High quantitative skills of the evaluators.

The technique can be applied to assess the effect of CAP support on the evolution of the values of the impact indicators listed in the following table.

• Step 1 – Construct the model with the appropriate data: In the case of the recursive-dynamic CGE, the most demanding data needs are associated with model construction. The basis for a regional/rural CGE model is a mechanically constructed social accounting matrix (SAM). As information necessary for the SAM construction requires regional employment and accounting data at the sectoral level, it is advised that the CGE model is built at the level of relevant rural NUTS 3 regions, as defined by the Eurostat urban-rural typology.
  • sectoral employment data is needed to downscale an available national input-output table (which should ideally be for a year close to the start of an implementation period) for the programme area level;
  • data needed for filling the inter-institutional and factor-institution flows can be (usually) obtained from regional accounts, household income and expenditure surveys;
  • the latter should also be used in the (very frequent) case where households are disaggregated into different types (e.g. income levels).
• Step 2 – Calibrate the model: The calibration of the dynamic CGE model requires the specification of a wide range of production, trade and household consumption elasticities. When the analysis is at the regional level, such elasticities are often based on reviews of the relevant literature. The same holds for the definition of exogenous parameters, which are often available at the national level, and hence, significant fine-tuning is often needed to downscale them. Deep knowledge (and often expert opinion) is used to specify study-area-specific closure rules on factor markets, the government budget, the regional current account, and the investment and savings account. In the case of measure-specific financial flows, the needed information is annual expenditure (which has to be converted into model baseline prices) and data on the sectoral targeting of flows for each measure.
• Step 3 – Control model dynamics with appropriate adjustments: To control model dynamics, a number of exogenous ‘between period’ adjustments on variables can be imposed in the recursive-dynamic version of the model, such as productivity growth, government spending, population and labour supply. These adjustments should be imposed through the use of real data for the implementation period and projections for the period for which real observations do not exist. Capital adjustment for each sector between periods is typically endogenous, with investment by commodity in the solution of the model in ‘period t-1’, used to update capital stocks before the model solution in ‘period t’. Assuming that the commodity composition of capital stock is identical across activities, the allocation of new capital across activities uses a partial adjustment mechanism with those activities where returns are higher than average, obtaining a higher-than-average share of the available capital.
• Step 4 – Estimate the impact indicators with appropriate additional data: CGE model outputs include programme-specific annual impacts on employment, household income and GDP. Therefore, to estimate the aforementioned impact indicators, the following additional data should be obtained:
  • Rural employment – study-area-specific changes in population aged 15-20 and over, since the start of a programming period, net of non-programme effects. These estimates can be generated through the application of a qualitative method.
  • Poverty rate and rural GDP – study-area-specific changes in total population, since the start of the programming period, net of non-programme effects. These estimates can be generated through the application of a qualitative method.
  • Rural GDP – PPS conversion rates which are available from Eurostat.

To sum up, with the exception of programme implementation data, which should normally be available from programme implementing authorities, data availability for the construction and calibration of the model is a rather case-specific issue. Research experience has shown that data availability varies considerably amongst case studies, leading (in the case of restrictions) to second-best choices in designing a model with a more aggregate sectoral structure.

• Structural models are essential for estimating unobservable economic or behavioural factors in programme evaluation.
• They allow for comprehensive policy simulations and forecasting effects in various economic scenarios.
• Capable of covering a wide range of complexities, these models are versatile in application.
• Their usage, especially in forecasting and evaluating programmes, demands high-level quantitative skills and detailed data.

• European Evaluation Network for Rural Development for the CAP (2014)
  Capturing the success of your RDP: Guidelines for the ex post evaluation of 2007-2013 RDPs