In celebration of the new year, I will be dedicating this blog post to policy analysis given 2020 targets. My aim is to analyze Norwegian carbon emissions and understand the impact (if any) of the carbon taxes or emission trading schemes in place.
2020 is a target year for many policies. For example, the EU’s climate and energy targets for 2020 were: 20% cut in greenhouse gas emissions (from 1990 levels), 20% of EU energy from renewables, and 20% improvement in energy efficiency. Whether or not the EU has attained its targets is out of the scope of this blog post, but it will certainly be an interesting research question.
Norway had similar climate targets for 2020:
“Norway is committed to an ambitious climate policy and has a target of 30 % GHG emissions reduction by 2020 relative to 1990 levels.”
As part of Norway’s ambitious climate policy, a carbon tax was created in 1991. The aim of this tax was to ensure market prices reflect the social cost of greenhouse gas emissions into the atmosphere, and incentivize the reduction of these emissions. However, the carbon tax varied across sectors, going up to 500 kroner per ton of CO2-equivalent for the oil and gas sector. The picture below illustrates this problem: back in 2006, sectors like oil and gas (to the right) paid the highest carbon tax while sectors of aluminium, cement or fisheries paid basically zero carbon tax. Since 2006, the carbon tax has increased to today’s maximum level of 500 kroners (roughly 50 US dollars).
Source: SSB website (https://www.ssb.no/natur-og-miljo/artikler-og-publikasjoner/for-komplisert-klimapolitikk)
So, even though the carbon tax can be considered a “treatment”, its effect might be unequal across sectors.
Bruvoll and Larsen (2004) find that the Norwegian carbon tax only contributed to a reduction in CO2 emissions of 2%. The authors applied a general equilibrium simulation, while I am using historical data to analyze the impact of the carbon tax.
The second policy instrument that affects carbon emissions is the emissions trading scheme (ETS). In 2005, the European Union (EU) promoted its own emissions trading scheme. Norwegian companies also participate in the EU ETS by buying allowances at their market price.
More than 80% of the Norwegian greenhouse gas emissions are covered by either carbon taxes or the EU ETS. Almost all sectors of the Norwegian economy have to pay carbon taxes. Oil production, power generation, heating and air travel sectors are also covered by the EU ETS (Emissions trading Scheme). However, the agricultural sector is not covered by neither carbon taxes nor the EU ETS.
The differences across sectors means that I may consider some sectors at some specific moment in time as the treatment group and some sectors as the control group. For example the agricultural sector seems not to be covered by any policy instrument, so its emissions are not expected to exhibit any impact due to a carbon tax or the ETS. A note however, is that the database I built for this study is not very reliable, since within one sector, some firms are covered by a scheme, and other might not. I am also not confident about the unbiased in the measurement of the dummy variables. However, data availability forces me to consider sectors instead of firms.
My goal is to analyze the joint impact of the carbon tax and the EU ETS in Norwegian carbon emissions. To do so, I use the data from an Excel data file that I created. It contains selected variables extracted from SSB (Statistisk sentralbyrå). The variables are:
- Sektor and Sektor_ID. Sektor provides a description of which sector of the economy the observation pertains to, and Sektor_ID is a unique identifier for each sector.
- Greenhouse gas emissions for Norway, which are originally taken from here: https://www.ssb.no/statbank/list/klimagassn/
- Value Added by Industry, which can be found here: https://www.ssb.no/en/nasjonalregnskap-og-konjunkturer/tables/nr-tables
- Population in Norway, which can be found here: https://www.ssb.no/en/statbank/table/06913/
- Two dummy variables: Carbon_Tax and EU_ETS. They take the value 1 if the sector is covered by either a carbon tax or the EU ETS, and 0 if not covered.
In order to interpret the coefficients as elasticities, I will consider the natural logarithm of GHG emissions as my dependent variable.
data$ln_GHG <- log(data$`GHG Emissions`)
I can now run a simple OLS regression to see the impact of the carbon tax and the emission trading scheme:
model1 <- lm(ln_GHG ~ Carbon_Tax + EU_ETS , data=data) summary(model1)
Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 8.39585 0.17235 48.713 <2e-16 *** Carbon_Tax 0.05071 0.19517 0.260 0.7953 EU_ETS 0.33299 0.17893 1.861 0.0642 .
According to the estimated coefficients, instead of having a negative impact on GHG emissions, the EU ETS has a positive effect on GHG emissions. This means being covered by the EU ETS actually increases GHG emissions. However, this effect is only statistically significant at the 10% level.
However, there is likely omitted variable bias, since there are other variables that explain GHG emissions and that might be correlated with either policy instruments. For example, the IPAT framework establishes that the environmental impact, such as GHG emissions, are a function of population, added value and technology. Thus, it is important to account for these variables in their logarithmic form, especially if they are correlated with the dummy for carbon tax or the EU ETS.
The second consideration I want to make is that there might be sector-specific effects. These aspects generally do not vary over time, hence they are called fixed effects. I account for these by including a dummy for each sector.
The third consideration is to include is the GHG emissions from the previous year. The amount a sector pollutes can be highly correlated with the amount they have polluted in the previous year, and it is likely to be the strongest predictor. To address potential autocorrelation, it is important to add lagged GHG emissions when explaining current GHG emissions.
Therefore, I take fixed effects, lagged GHG emissions and the natural logarithm of population and value added into account. I run the following regression:
model2 <- lm(ln_GHG ~ factor(Sektor_ID) + Carbon_Tax + EU_ETS + ln_Pop + ln_VA + ln_GHG_1, data=data) summary(model2)
The output looks like this:
Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 1.22417 3.26985 0.374 0.70855 factor(Sektor_ID)3 -0.03032 0.03106 -0.976 0.33027 factor(Sektor_ID)4 -0.12730 0.09424 -1.351 0.17841 factor(Sektor_ID)5 -0.16798 0.08318 -2.019 0.04489 * factor(Sektor_ID)6 -0.04460 0.03403 -1.310 0.19169 factor(Sektor_ID)7 -0.04634 0.03405 -1.361 0.17522 factor(Sektor_ID)8 -0.37607 0.13439 -2.798 0.00568 ** Carbon_Tax -0.31759 0.10607 -2.994 0.00313 ** EU_ETS -0.04115 0.02774 -1.484 0.13962 ln_Pop -0.03134 0.23331 -0.134 0.89328 ln_VA 0.02411 0.03618 0.667 0.50591 ln_GHG_1 0.92609 0.02886 32.088 < 2e-16 ***
Interestingly, the carbon tax seems to have a negative impact on Norwegian GHG emissions. If a sector is covered by a carbon tax, GHG emissions decrease on average by 32%. This effect is statistically significant, while the EU ETS coefficient is not. Some sector fixed effects are significant, such as the agricultural (Sector_ID=8) or households (Sector_ID=5).
This is a rather simple exercise that treats the panel nature of the data rather simply. The effect of the Norwegian carbon tax seems to be a lot stronger than that estimated in Bruvoll and Larsen (2004), which leads me to believe there might be some bias in the results.
Bruvoll, A., & Larsen, B. M. (2004). Greenhouse gas emissions in Norway: do carbon taxes work?. Energy policy, 32(4), 493-505.