Logo Logo
Help
Contact
Switch language to German
The calibration of economic growth. an application to carbon emission scenarios and to the DICE model
The calibration of economic growth. an application to carbon emission scenarios and to the DICE model
In chapter 1, I construct carbon dioxide emission scenarios for Europe until 2100. The three most important ways in which this chapter contributes to the literature are, first, that economic growth is driven by endogenous investments into research and development. Second, the model is formally calibrated, using data that span a period (1850-2008) longer than the projection period (2008-2100). Third, this work provides statistically valid confidence intervals of economic growth, which translate into a measure of uncertainty regarding future carbon emissions. Forecasts are made on the regional level for Europe as a whole, Western Europe, Eastern Europe and the former USSR. The calibration of economic growth yields higher future annual growth rates in Western Europe than in Eastern Europe and the former USSR. In order to offset this higher economic growth (or equivalently in order to keep future carbon emissions until 2100 at the same level as today), Western Europe would have to reduce its energy and emission intensities by roughly 1% annually, while in Eastern Europe and former USSR countries approximately 0.5% of annual reductions would be sufficient. Because of past economic turmoil the estimated uncertainty that is tied to future economic growth is larger in Eastern Europe and former USSR countries than in Western Europe. Chapter 2: In the Integrated Assessment literature there has been a growing need for sound calibration techniques of growth models which go beyond the medium run and aid as a forecasting device of macroeconomic trends. To model the future costs of climate change, we need to know more about carbon emissions in the long run, and they will crucially depend on future economic growth and technological advancements. Because the majority of Integrated Assessment Models is deterministic, in this chapter I develop a robust calibration technique for deterministic models of long run growth. The aim is to present a standardized approach towards calibration, which is straight forward and easy to adapt. I suggest a Bayesian inversion technique to elicit the distributions of all parameters of calibration and to project the confidence intervals of future income and consumption shares. Since in this chapter I propose a technique to calibrate deterministic models of economic growth in the long run, I set my self apart from the Bayesian calibration literature of stochastic macro-economic models (see for instance Fernández-Villaverde and Rubio-Ramírez (2007) and Fernández-Villaverde (2010)). To integrate over Bayes' law I use a Markov chain Monte Carlo (MCMC) algorithm. The likelihood function is derived from a stochastic process, which describes the residuals between the observed and the simulated data. Since the majority of Integrated Assessment models is based on a Ramsey type growth model, I demonstrate this procedure by calibrating a standard Ramsey model of exogenous growth as well as an endogenous growth model by Aghion and Howitt (1999). The resulting growth trajectories until 2050 from both models are similar with an average growth rate of the median projection of 2.3% in the Ramsey model and 2.2% in the Aghion & Howitt model. All parameters of calibration are well identified and have clear-cut distributions. Therefore, I conclude that this approach is highly flexible for the calibration of the growth component in Integrated Assessment Models and that it can be adopted for a wide range of different growth models. Chapter 3 analyzes the negative impact of climate change on economic growth caused by a reduction of the return on general R&D and consequently of in-vestments into the same. The framework is based on an Integrated Assessment Model, the DICE model by Nordhaus (2008). The DICE model builds on Ramsey type growth where environmental damages cause a negative level effect on GDP. In a version of this model, I substitute the growth component by endogenous Schumpeterian type growth and calibrate it to the original. In the socially `Optimal Scenario', the social planner is able to mitigate climate change in two ways. First, he can invest in the reduction of carbon emissions and, second, he can shift his spending away from the carbon-emitting capital stock. In addition, in the endogenous Schumpeterian growth setting, the return on investment in R&D declines due to environmental damages and thus investments into the general R&D sector are reduced. Since endogenous investments into R&D are what drives economic growth in a Schumpeterian model, global warming has a lasting and negative impact on GDP growth through this channel. It can be understood as an additional effect which adds to the channel of directed technical change as described in Aacemoglu et al. (2012). In both model versions, the reallocation of resources reduces future total output. However, the negative effect in the endogenous growth setting is stronger, since investments into R&D are allowed to go down. Comparing the Ramsey and the Schumpeterian version of the DICE model, this long-lasting, negative growth effect is even stronger in a `Constrained Optimum Scenario', where households cannot actively mitigate to reduce climate damages. On the contrary, in a `Business as Usual Scenario', where the climate externality is not internalized and the private return on investment is not affected by climate change, there are no negative growth effects due to the reallocation of resources. In this scenario, however, higher growth rates cause more climate damages, which eventually overcompensate an initially higher rate of economic growth. Chapter 4: In the Integrated Assessment literature, economic growth is a major determinant of projected carbon emissions and climate damages. Nevertheless, its importance is often overlooked. While most macroeconomic models of growth are run for a couple of decades at best, in an environmental context these same growth models are often solved for centuries. This increases the dependency of all growth projections on their underlying model assumptions. In this chapter, I carefully recalibrate the growth component of the DICE-2016R model as described in chapter 3 using the Bayesian calibration approach developed in chapter 2. One major advantage of this Bayesian calibration technique is that it quantifies the uncertainty that is tied to economic growth, given the model assumptions and the observed hostorical data. In the DICE model, the link between economic growth and carbon emissions is particularly intense, since GDP translates directly into carbon emissions at an exogenously given proportion, which shrinks over time as fossil energy is gradually substituted by clean energies. The expected mean temperature increase compared to pre-industrial levels, in the `Optimal Scenario' in the re-calibrated version of the DICE model, amounts to 3.8°C. Interestingly, the results show that even though the expected variation of future gross income is very high, the expected variation from the mean temperature increase by 2100 is relatively low, with only 4% within the 90% confidence interval. This is because in the `Optimal Scenario' the mitigation of carbon emissions amounts to almost 100% in 2100. In the `Constrained Optimum Scenario', where households have no instrument of direct mitigation, the mean temperature increase of the atmosphere compared to pre-industrial levels amounts to 4.7°C, with an expected variation of 11% in the 90% confidence interval in 2100. Thus, the implementation of effective climate change policies aimed at reducing carbon emissions does not only lower the level of the future temperature increase significantly, but also the uncertainty over the magnitude of future climate damages.
Climate Change, Bayesian Calibration, Endogenous Growth
Ciesielski, Anna
2018
English
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Ciesielski, Anna (2018): The calibration of economic growth: an application to carbon emission scenarios and to the DICE model. Dissertation, LMU München: Faculty of Economics
[img]
Preview
PDF
Ciesielski_Anna.pdf

1MB

Abstract

In chapter 1, I construct carbon dioxide emission scenarios for Europe until 2100. The three most important ways in which this chapter contributes to the literature are, first, that economic growth is driven by endogenous investments into research and development. Second, the model is formally calibrated, using data that span a period (1850-2008) longer than the projection period (2008-2100). Third, this work provides statistically valid confidence intervals of economic growth, which translate into a measure of uncertainty regarding future carbon emissions. Forecasts are made on the regional level for Europe as a whole, Western Europe, Eastern Europe and the former USSR. The calibration of economic growth yields higher future annual growth rates in Western Europe than in Eastern Europe and the former USSR. In order to offset this higher economic growth (or equivalently in order to keep future carbon emissions until 2100 at the same level as today), Western Europe would have to reduce its energy and emission intensities by roughly 1% annually, while in Eastern Europe and former USSR countries approximately 0.5% of annual reductions would be sufficient. Because of past economic turmoil the estimated uncertainty that is tied to future economic growth is larger in Eastern Europe and former USSR countries than in Western Europe. Chapter 2: In the Integrated Assessment literature there has been a growing need for sound calibration techniques of growth models which go beyond the medium run and aid as a forecasting device of macroeconomic trends. To model the future costs of climate change, we need to know more about carbon emissions in the long run, and they will crucially depend on future economic growth and technological advancements. Because the majority of Integrated Assessment Models is deterministic, in this chapter I develop a robust calibration technique for deterministic models of long run growth. The aim is to present a standardized approach towards calibration, which is straight forward and easy to adapt. I suggest a Bayesian inversion technique to elicit the distributions of all parameters of calibration and to project the confidence intervals of future income and consumption shares. Since in this chapter I propose a technique to calibrate deterministic models of economic growth in the long run, I set my self apart from the Bayesian calibration literature of stochastic macro-economic models (see for instance Fernández-Villaverde and Rubio-Ramírez (2007) and Fernández-Villaverde (2010)). To integrate over Bayes' law I use a Markov chain Monte Carlo (MCMC) algorithm. The likelihood function is derived from a stochastic process, which describes the residuals between the observed and the simulated data. Since the majority of Integrated Assessment models is based on a Ramsey type growth model, I demonstrate this procedure by calibrating a standard Ramsey model of exogenous growth as well as an endogenous growth model by Aghion and Howitt (1999). The resulting growth trajectories until 2050 from both models are similar with an average growth rate of the median projection of 2.3% in the Ramsey model and 2.2% in the Aghion & Howitt model. All parameters of calibration are well identified and have clear-cut distributions. Therefore, I conclude that this approach is highly flexible for the calibration of the growth component in Integrated Assessment Models and that it can be adopted for a wide range of different growth models. Chapter 3 analyzes the negative impact of climate change on economic growth caused by a reduction of the return on general R&D and consequently of in-vestments into the same. The framework is based on an Integrated Assessment Model, the DICE model by Nordhaus (2008). The DICE model builds on Ramsey type growth where environmental damages cause a negative level effect on GDP. In a version of this model, I substitute the growth component by endogenous Schumpeterian type growth and calibrate it to the original. In the socially `Optimal Scenario', the social planner is able to mitigate climate change in two ways. First, he can invest in the reduction of carbon emissions and, second, he can shift his spending away from the carbon-emitting capital stock. In addition, in the endogenous Schumpeterian growth setting, the return on investment in R&D declines due to environmental damages and thus investments into the general R&D sector are reduced. Since endogenous investments into R&D are what drives economic growth in a Schumpeterian model, global warming has a lasting and negative impact on GDP growth through this channel. It can be understood as an additional effect which adds to the channel of directed technical change as described in Aacemoglu et al. (2012). In both model versions, the reallocation of resources reduces future total output. However, the negative effect in the endogenous growth setting is stronger, since investments into R&D are allowed to go down. Comparing the Ramsey and the Schumpeterian version of the DICE model, this long-lasting, negative growth effect is even stronger in a `Constrained Optimum Scenario', where households cannot actively mitigate to reduce climate damages. On the contrary, in a `Business as Usual Scenario', where the climate externality is not internalized and the private return on investment is not affected by climate change, there are no negative growth effects due to the reallocation of resources. In this scenario, however, higher growth rates cause more climate damages, which eventually overcompensate an initially higher rate of economic growth. Chapter 4: In the Integrated Assessment literature, economic growth is a major determinant of projected carbon emissions and climate damages. Nevertheless, its importance is often overlooked. While most macroeconomic models of growth are run for a couple of decades at best, in an environmental context these same growth models are often solved for centuries. This increases the dependency of all growth projections on their underlying model assumptions. In this chapter, I carefully recalibrate the growth component of the DICE-2016R model as described in chapter 3 using the Bayesian calibration approach developed in chapter 2. One major advantage of this Bayesian calibration technique is that it quantifies the uncertainty that is tied to economic growth, given the model assumptions and the observed hostorical data. In the DICE model, the link between economic growth and carbon emissions is particularly intense, since GDP translates directly into carbon emissions at an exogenously given proportion, which shrinks over time as fossil energy is gradually substituted by clean energies. The expected mean temperature increase compared to pre-industrial levels, in the `Optimal Scenario' in the re-calibrated version of the DICE model, amounts to 3.8°C. Interestingly, the results show that even though the expected variation of future gross income is very high, the expected variation from the mean temperature increase by 2100 is relatively low, with only 4% within the 90% confidence interval. This is because in the `Optimal Scenario' the mitigation of carbon emissions amounts to almost 100% in 2100. In the `Constrained Optimum Scenario', where households have no instrument of direct mitigation, the mean temperature increase of the atmosphere compared to pre-industrial levels amounts to 4.7°C, with an expected variation of 11% in the 90% confidence interval in 2100. Thus, the implementation of effective climate change policies aimed at reducing carbon emissions does not only lower the level of the future temperature increase significantly, but also the uncertainty over the magnitude of future climate damages.