Consumption and production-based CO₂ pricing policies: macroeconomic trade-offs and carbon leakage

Abstract

This paper applies a DYNK (Dynamic New Keynesian) model to compare the traditional environmental tax reform for greenhouse gas (GHG) emissions with a taxation scheme that taxes GHG emissions embodied in consumption within the framework of a unilateral policy of the EU-27. The embodied emissions of different commodities are taxed independently of their origin. The GHG tax rates applied are identical and new revenues are in both cases recycled via lower social security contributions of employers. The results show the macroeconomic results, driven by the different impact of the taxation schemes on price competitiveness of EU-27 firms. These differences drive the leakage and show negative leakage in the case of taxing embodied GHG emissions. Both taxation schemes are also regressive for household incomes emphasizing the importance of the choice of revenue recycling. In terms of emission reduction, we find the taxation of emissions embodied in consumption less effective.

KEYWORDS:

• GHG tax
• input-output modelling
• carbon leakage
• environmental fiscal devaluation
• income distribution

Acknowledgments

We highly appreciate the valuable input of Dr. Mathias Kirchner and Dr. Ina Meyer. Comments by three anonymous reviewers contributed substantially in providing a more consistent and transparent paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Status of the European Union in 2014, i.e. without Croatia and with the United Kingdom.

2 And further work to derive the social cost of CO₂ emission by Tol (2009), Anthoff et al. (2011) or Sunstein (2013).

3 For details on the model see Kratena (2017), Kratena et al. (2017), and Kratena and Streicher (2009), Kirchner et al. (2019) and Sommer and Kratena (2017).

4 Data on emissions stems from the environmental tables of the world input output data base www.wiod.org and covers CO₂, NO₂ and CH₄ emissions.

5 i.e. the same €/ton CO_2.

6 This is known as the Double Dividend effect (Goulder, 1995).

7 Even though to assume that labor tax reductions can be adjusted to be equal to CO₂-Tax revenues instantly in the same year is unrealistic, a real implementation would oscillate around these ‘instant’ rates calculated in our model.

8 These industries represent the Statistical Classification of Economic Activities in the European Community, commonly referred to as NACE (for the French term ‘nomenclature statistique des activités économiques dans la Communauté européenne’) and in our case is aggregated to 59 sectors.

9 These consumption categories represent the COICOP classification (Classification of Individual Consumption according to Purpose) that comprises – in the three-digit structure – 47 classes.

10 For a detailed documentation on assumptions and theory of the macroeconomic FIDELIO model family, of which DYNK is a part of, see Kratena et al. (2013, 2017). FIDELIO also comprises the global ADAGIO model (Kratena and Streicher, 2014).

11 The first stage refers to those durables for which purchases the households obligate themselves. Appliances might be purchased on credit are not included there in this model version for the sake of simplicity. Nevertheless, appliances are energy relevant durables. We model the appliances stock via the accumulated consumption of appliances, C_app, which in turn is explained in a log linear specification like total transport demand (see section ‘extended model description’ in the Appendix).

12 Almost ideal demand system (AIDS) developed by Deaton and Muellbauer (1980).

13 Please note that all variables are listed in the online Appendix and monetary values are in current prices except – in some equations – when explicitly deflated by a price (e.g. q/p).

14 EUROSTAT, Data tables on Final consumption expenditures of households by consumption purpose (COICOP three digit).

15 WIOD (www.wiod.org) Environmental Accounts of the 2013 version.

16 IEA database (www.iea.org) on international fuel prices.

17 ODYSSEE database (www.odyssee-mure.eu).

18 GAINS project (http://www.iiasa.ac.at/web/home/research/researchPrograms/air/GAINS.html).

19 The estimated coefficients in the AIDS model can be transferred to point own-price elasticities that depend the respective expenditure share. Since the five household groups show different shares in their consumption, the resulting elasticities are different among households.

20 See Table 2 for the eight groups.

21 NACE (Nomenclature statistique des activités économiques dans la Communauté européenne) is the statistical classification of economic activities in the European Community.

22 These energy prices π_h represent monetary values per physical energy content (€/TJ). They are the weighted sum of the subshare of energy types (S_he) of which the energy commodity consists of and their respective energy type prices (π_e), i.e. ${\mathit{\pi }}_{\text{h}}={\mathit{\pi }}_{\text{e}}{\mathbf{\text{S}}}_{\text{he}}$.

23 Direct GHG emissions of households consist only of CO₂. Nevertheless, for consistency reasons we stick with the term ‘GHG’.

24 A disaggregation of both, electricity and district heat was not possible due to data restrictions.

25 We are indebted to Iñaki Arto for providing us with the aggregate results of his work for the EU-27.

26 This translog model comprises substitutions of the fuel mix in the sectoral inputs, see Section 2.2.

27 The statistical classification of products by activity, abbreviated as CPA, is the classification of products (goods as well as services) at the level of the European Union (EU).

28 These prices represent the prices of the 5 fuel aggregates (k), which are weighted averages of 26 underlying energy carrier (fuels and electricity) prices.

29 This carbon price is in the range of previous and recent scenarios of the EU Roadmap (EU Commission, 2011, p. 210).

30 This rate is the unemployment rate the economy is set to tend to in the long run.

31 World market prices are exogenous in the model, i.e. we implicitly assume that the measures in the EU have no effect on these prices. Hence, increasing domestic prices lead to a relative increase in comparison to the world market prices.

32 The disposable income of each household quintile is deflated by the specific consumer price index of each group’s consumption basket.

33 By keeping model parameters constant, we implicitly assume that the distribution of new employees amongst household quintiles is equal to the distribution of the working population in the base year (here 2010). Hence, by implementing more-detailed modeling, via agent-based models, the income and consumption divergence between households could be less.

34 36% of income stems from wage in the 1st quintile compared to 70% in the 5th – based on SILC (‘Community Statistics on Income and Living Conditions’) statistics for the base year (2010) in the EU-27.

35 That is, 17% of the domestic reduction is offset by emissions increases abroad. Given by dividing the change of leakage (compared to baseline domestic emissions) by total domestic emissions. Based on Table 4 3.1/(−17.4) = 0.17 in Table 4.

36 The employment effects of all sectors is noted in the Appendix.

37 Results from adding the change of domestic GHG emission (−17.8) to leakage (+3.1) in Table 4.

38 Results from adding the change of domestic GHG emission (−5.2) to leakage (−1.9) in Table 7.

39 World Bank and Ecofys (2017)

40 See for example Murray and Rivers (2015) and Yamazaki (2017).

Additional information

Funding

The research presented was mainly developed in the project “WWWFOREUROPE” which was funded by European Commission's Seventh Framework Programme FP7/2007-2013 under grant agreement no. 290647.