![Sample our Environment & Agriculture journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5934/TOC-Subject-Sample-2021-Environment-Agriculture.jpg"})
Abstract
This study provides new evidence on the determinants of environmental expenditure and investment. In particular, it investigates how environmental expenditure and investment of Swedish industrial firms responded to climate policies, such as the European Union's Emission Trading System (EU ETS) and the Swedish CO2 tax, directed to mitigate air pollution. Overall, an important conclusion of this analysis is that climate policies, both on the national and international levels, were highly relevant motivations for firm environmental expenditure. However, the findings do not support the expectations that the EU ETS and the Swedish CO2 tax encouraged investment in air pollution abatement.
Notes
1. Ellerman, Convery, and Perthuis Citation(2010) is probably the most authoritative reference on the EU ETS. It contains a detailed discussion of the design of the ETS and provides a comprehensive analysis of Phase I.
2. See Zhang and Wei Citation(2010) and Venmans Citation(2012) for the overviews of research on the EU ETS.
3. Another study related to ours, in the sense that it uses the Swedish environmental expenditure and investment data, is by Broberg, Marklund, and Samakovlis Citation(2013). They analyse the effect of expenditure and investment (proxies for policy pressure) on firm level technical efficiency to test the so called Porter hypothesis of which they find no support.
4. This sector suffers from bad quality data on energy use for many years.
5. This decoupling of economic performance and pollution is also corroborated by the results for Swedish industrial firms during 1990–2004 in Brännlund, Lundgren, and Marklund Citation(Forthcoming).
6. See Hammar and Åkerfeldt Citation(2011) for more details on CO2 taxation in Sweden.
7. See Table A1 in the Appendix for more detailed definitions of environmental investment and expenditure.
8. One might argue that this is not a perfect instrument; however, after trying various other types of instruments, we find that the selection of the instrument does not affect the main results in any significant way. As a robustness test, we also run the models without these exclusion variables. Again, the main results are not affected in any significant way. These additional robustness results are available from the authors upon request.
9. We use all energy sources, including primary fossil fuel consumption, converted to a MWh equivalent.
10. Our Törnqvist price index is calculated as the geometric average of energy input prices relative to the base period prices that are weighted by the arithmetic average of the particular energy product value shares for the analysed time period. We estimated firm-specific energy input prices by dividing the energy specific cost of each firm by the corresponding energy input quantities used by the firm.
11. For the models focusing on the EU ETS effects, we perform an additional robustness test, where we control for the dynamics of the CO2 tax by including total energy consumed (in MWh). This variable is closely related to total CO2 tax expenses. The main results are not affected in any significant way. A full set of results are available from the authors upon request.
12. In 2005, due to the administrative burden on enterprises, Statistics Sweden decided to raise the cut off to enterprises with 50 employees or more instead of the usual 20 employees. To make comparisons with earlier years possible, estimations were made for the size group 20–49 employees.
13. More detailed information on the sampling frame, its variation over time, the representativeness, and the actual questionnaires that were distributed to the firms in the sample can be found on the SCB website: http://www.scb.se/Pages/Product____12799.aspx?Produktkod=MI1302&displaypublications=true.
14. To the best of our knowledge, this merger of different datasets has not been done before.
15. Some variables are available for the shorter period.
16. The first-stage probit models are run separately for each year. Then, the inverse Mills ratios are constructed and used in the second stage. Thus, each fixed-effects model has at least eight corresponding probit models. These results can be provided by the authors upon request. Note that the results from the probit models for each individual year are somewhat different from each other and from the results of the pooled probit model. These differences further motivate the case for our chosen panel model approach. However, we refrain from explaining these yearly differences due to the complexity issue that involves a much deeper analysis of yearly investment decision making subtleties what is not our main purpose of this article.
17. Baltagi and Pinnoi Citation(1995) explain that this may happen as fixed-effects models estimate more short-run effects and pooled OLS models produce the expected sign, suggesting the long-run impact. Thus, it is important to consider both estimation methods.
18. See Table A3 in the Appendix for the first-stage results.
19. Note that only a small subsample is available with firms with information on CO2 tax expenses. The differences of coefficients between and are largely due to the different samples, not due to the inclusion of the additional CO2 tax variable. The fixed-effects model results are in line with the ones in .