GTAP Resources: Resource Display
| GTAP Resource #6365 |
|---|
|
"Combining large-scale sensitivity/uncertainty analysis in Computable General Equilibrium analysis with Machine Learning to find optimal policy mixes" by Britz, Wolfgang, Jingwen Li and Linmei Shang Abstract Policy design nowadays needs to consider goals related to multiple sustainability dimensions simultaneously. The Tinbergen principle suggests addressing each issue individually with a targeted policy measure. But each single resulting command-and-control, incentive or market-based policy is likely to affect also other policy goals, questioning overall policy coherence (cf. May et al. 2006). CGE modelling can contribute here by quantifying market-mediated impacts of changing policies against the given benchmark, considering key policy indicators such as income and its distribution, Green House Gas Emissions or land cover changes. But each single experiment considers one specific policy mix, from a public choice decision space which is extremely large, once we consider a larger set of policy domains (general economic and social policy, different environmental fields etc.). The growing interaction of global value chains implies that regional policy choices also increasingly affect sustainability outcomes elsewhere, a viewpoint increasingly addressed in policy impact assessments as well. Finally, impacts of policy choice in each region also depend on the policy chosen in others. We combine large-scale sensitivity analysis, changing both policy instruments and key model parameters in different regions, focusing on key sustainability metrics, and fit a neural network to the results. The considered policy instruments are indirect tax rates changes relating to bio-economy sectors, while land supply and different substitution elasticities are subject simultaneously to sensitivity analysis. We find a very good fit to cases where only policy instruments are changed and still quite high once when also parameters are changed. We conclude from there that machine learning techniques are able to provide robust meta-models of CGEs and can be used to predict or even optimize over the response space of the CGE. |
| Resource Details (Export Citation) | GTAP Keywords | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
- Software and modeling tools - Renewable energy - Climate change policy - Land use |
| Attachments |
|---|
|
If you have trouble accessing any of the attachments below due to disability, please contact the authors listed above.
Public Access  Paper (1.1 MB) Replicated: 0 time(s)Restricted Access No documents have been attached. Special Instructions No instructions have been specified. |
| Comments (0 posted) |
|---|
|
You must log in before entering comments.
No comments have been posted. |
Last Modified: 9/15/2023 1:05:45 PM
