VII. SCC Implications of Learning and Adaptation Flashcards
Welfare cost of risk in Learning vs no-learning
In the learning case, disasters are extra costly because beliefs jump to a high point (overreaction). Investment decreases even more than would be optimal, making risk even more costly when there is learning.
Policy implication of learning
As a result of learning and over-reaction the planner spends even more into mitigation when there is learning, especially at low levels of risk
Why does SCC go up over time?
In no-learning
- As the world grows, there is more carbon, larger carbon stock. Carbon stock depreciates slowly. Therefore, over time, there is more damage.
With learning, the story changes.
Optimal Mix of Fossil Fuel and Capital Taxation
In this economy, SCC along with the capital tax (adaptation) will give you together a good public policy that leads to higher social welfare.
If you didn’t have adaptations, SCC would be very high because only way to stop disasters is cutting fossil fuel use. But if you do have adaptations, that’s the substitute to mitigation.
What is the half life for a pulse of carbon?
3%
How does introducing learning into DICE change the model?
Previously, temperature had a deterministic impact on production. Now, we have shocks and all things are probabilistic. Different results based on shocks we draw.
By introducing shocks, we have a new mechanism of investing to protect/mitigate capital stock that goes into the investment block.
Due to the new existence of risk and uncertainty, I form beliefs about shocks through learning. Given my estimates about how many shocks I will face, I will invest more or less.
SCC in a No-Learning Environment
Explain the concave, increase shape of the curve.
Curve includes mitigation, extreme events, damages not in the models of first half of course. Just no learning.
Carbon stock is large emitting at the start and then we emit less and less until it plateaus and we hopefully stop adding more damages.
No-learning so you know the probabilities.
SCC with Learning
What do the lines mean in the quantiles graph?
Running 100 simulations. Given probabilistic nature, simulations of the same model have different disasters at different times.
Jumps are because of the occurrence of disasters
10% is the simulations with lowest SCC.
99% means very bad state because if SCC is higher then that means welfare cost of that scenario is higher.
- It’s flatness suggests disasters happening very very often and you learn at a constant rate.
SCC with Learning
Explain the u-shaped SCC curve
Very high SCC at the start and then it decreases
In the first few years, a period of high uncertainty, every shock happening is supremely costly for welfare because lots of over-reaction.
- Volatile early beliefs cause SCC to sky rocket.
- Having so few datapoints causes huge swings in beliefs.
As time goes on you have resolution of uncertainty.
- As stock of carbon increases, disasters are more frequent. More data points. Your assessment of probability becomes closer to the true one and the baseline of the no-learning.
- Hence, go back to normal shape after 10 years. Become less afraid of extremes and resolution of uncertainty is welfare improving.
Why should the SCC be higher today than in the future in a model with learning?
SCC should be higher today than in the future because we have way more uncertainty today than in the future. Over-reactions are costly for welfare.
Is the capital tax trend always rising?
Yes
Evaluate the use of SCC vs Adaptation in a learning world.
Carbon taxes distort production since if you can’t have carbon you cant have output
Adaptation such as building sea walls is less distortive since I take some capital and deploy on defensive.
How do the dynamics of beliefs reinforce our understanding of a u-shaped SCC curve with learning?
At the start, lots of volatility which tends to be the over-reactions.
These events make the extreme scenarios more likely in my head.
Fits the story of over-reactions and higher SCC initially.
