1. Environmental investment and firm performance: A panel VAR approach Tommy Lundgren, Shanshan Zhang, Wenchao Zhou Centre for Environmental and Resource Economics Umeå University and Swedish University of Agricultural Sciences

2. The study: summary Analyze the interactions between three dimensions of firm performance – productivity, energy efficiency, and environmental performance – and the role of environmental investment. Environmental investments are efforts to reduce environmental impact, which may also affect firm competitiveness, in terms of changes in productivity, and spur more (or less) efficient use of energy. DEA techniques used to derive Malmquist firm performance indexes Panel vector auto-regression (VAR) methodology is used to investigate the dynamic and causal relationship between the three dimensions of firm performance and environmental investment. Main results show that – Energy efficiency and environmental performance are integrated. – Energy efficiency and productivity positively reinforce each other. Hence, increasing energy efficiency, as advocated in many of today’s energy policies, would capture multiple benefits. – Improved environmental performance and environmental investments constrains next period productivity, a result that would be in contrast with 1) the so called Porter hypothesis, and 2) so called strategic corporate social (environmental) responsibility (CSR). – However, environmental investment, if channeled via improved energy efficiency, may affect productivity positively, signifying the role of investing in energy efficiency as a cost saving policy.

3. Background Corporate environmental impacts have received increasing attention in the last decades. Alongside with the increasing societal environmental concerns, firms have also experienced increasing pressure from governmental environmental policy. Whether environmental policy can improve firms’ competitiveness has been, and still remain, a debate since the Porter hypothesis was introduced in 1995. The Porter argument claims win-win solution of more stringent regulation. Firms may also go beyond compliance, take a proactive role in environmental protection (self- regulation), so called corporate social responsibility (CSR). CSR can be considered as firms’ strategic management aiming to meet societal expectations and minimize negative environmental impacts without compromising competitiveness. In the last few decades, studies on whether CSR can contribute to firms’ performance are flourishing, but evidence is not clear-cut. In economics, there is a debate on the validity of the methods used in these empirical studies. In any case, whether performance is driven by regulation or CSR, understanding the relationships between firms’ environmental investments and the actual economic and environmental and energy performances is crucial when evaluating the impacts of environmental management in general.

4. Background We ground firm performance measurements in microeconomic production theory. Follow the advice of Paul and Siegel (2006) who are critical to the vast amount of studies using subjective CSR scores and financial performance rather than economic performance indexes. We use Malmquist type of indexes of productivity, environmental performance, and energy efficiency, to define and measure firm performance. In a second stage regression analysis we study the relationships between the firm performances and environmental investment. To this end, we utilize a panel vector auto-regression (VAR) methodology: an econometric model that can examine the causal and dynamic relationships between the variables of interest, and can handle the inherent endogeneity problem present in our empirical application Thus we address properly the dynamic dimension, which is in line with for example Ambec et al. (2013) who argue that there is a lack of dynamic and causal concerns when assessing these relationships.

5. Contribution The present study adds to the existing literature in four respects. 1.Firm performance is assessed simultanously in three dimensions on firm level – productivity (and its components), energy efficiency, and environmental performance. 2.The indexes are consistent in the sense that all are estimated using the Malmquist index approach, which is soundly grounded in production theory. 3.In exploring the relationships, we integrate the four variables of interest (including environmental investments) into a system of multiple, cross- sectional time series, and as such, our model allows for estimating the causal effects between all four variables, without requiring to, a priori, explicitly specify the causal directions. 4.We use a representative sample of firm-level, industry-wide panel data consisting of 14 Swedish industry sectors, and thus our findings to a large extent are representative of the population of industrial firms as a whole, both in terms of environmental investments and firm performances.

6. The performance indexes Färe et al. (1989) define the Malmquist productivity index as a measure of productivity change of a decision making unit between two periods. We define Malmquist type of indexes for productivity, energy efficiency, and environmental performance. In constructing the indexes we use a combination of distance functions, which measures the distance of an output/input observation from itself to the best practice technology frontier. This means that in all aspects of performance, a firm is compared to its own “best-in-class” peers. The most efficient/productive firms make up the frontier which sets the benchmark.

7. The indexes: formal representation Productivity Energy efficiency Environmental performance To estimate each distance function in these performance measures we use Data Envelopement Analysis (DEA) or Activity Analysis. This is a linear programming technique that ”envelops” the data and defines the frontier. Index > 1 progress Index < 1 regress

9. The pVAR model Panel VAR methodology is used to investigate the relationships between firm performances and the impact of environmental investment. This approach allows that all the variables in a system of cross-sectional (multiple) time series can affect each other, and allows for the unobserved individual heterogeneity associated with panel data.

10. The pVAR estimation System GMM method is used to estimate the coefficients. Since these Malmquist indexes are calculated deterministically by solving the distance functions and they are not observed data generated from some stochastic process. According to Simar and Wilson (1999), these calculated indexes are biased and need to be corrected. We therefore adopt the Simar and Wilson (1999) bootstrap procedure to bias-correct our estimated Malmquist measurements and then use the corrected values in the second-stage panel VAR estimation.

11. Data The empirical study is performed by using a sample of firm-level, industry-wide data from Swedish industry. The unbalanced panel data set covers the years 2002 - 2008, and includes 14 sectors: pulp and paper, basic iron and steel, chemicals, mining, wood products, stone and mineral, food, motor vehicles, machinery, rubber and plastic, electro, fabricated metal products, textile, and printing. We select observations that are available for at least two consecutive years. Output is calculated using firm’s final sales divided by its corresponding sector- level producer price index. Non-energy inputs are capital stock and labor. Energy inputs are fossil fuel (coal, oil and gaseous fuel), non-fossil fuel (wood fuel and district heating) and electricity. Undesirable output is carbon dioxide (CO 2 ) emission.

12. Results: the indexes PeriodsMPMEMEP 2001/20021.0070.9890.994 2002/20030.9870.9811.004 2003/20041.0111.0001.003 2004/20051.0081.0000.993 2005/20061.0511.0030.997 2006/20071.0240.9960.998 2007/20080.9841.0130.999 Average1.0100.9970.998 Aggregated mean value of Malmquist indexes of all sectors, 2002-2008

13. The indexes and environmental investments: simple correlations MPMEMEPEI MP1 ME0.378 (0.000)1 MEP0.145 (0.000)0.345 (0.000)1 EI0.041 (0.120)0.032 (0.023)0.010 (0.731)1 Pearson-type correlation test between Malmquist indexes and environmental investment.

14. The pVAR results MP t ME t MEP t EI t MP t-1 0.027 0.092*** 0.010 0.028 ME t-1 0.415**-0.077 0.069** 0.274*** MEP t-1 -2.054*** 0.557*** 0.293*** 0.086 EI t-1 -0.057** 0.012** 0.001 0.058 No. of observations: 1966 Estimation results of the panel VAR(1) model Lags 1 to 4 are used as instruments in our GMM estimation.

15. Results I Previous period productivity generates positive impacts on all current firm performances and the environmental investment, but only significantly on energy efficiency. In line with Boyd and Pang (2000), who find that in US flat glass industry, a high productivity is associated with high energy efficiency. Previous period energy efficiency significantly impact current productivity, environmental performance and environmental investment. Increased energy efficiency would conserve more resources, thereby contributing to the growth of productivity and give more room for environmental performance. As the conservation of energy resources save costs it creates opportunity to increase the environmental investment further.

16. Results II Previous period environmental performance significantly impact current energy efficiency and itself. Also, previous environmental performance progress impede current productivity. This result contradicts Porter. The rationale is that there is a trade-off between environmental performance and productivity. This In line with, e.g., Gray and Shadbegian, 1993; Wally and Whitehead, 1994; Boyd and McClelland, 1999; Brännlund and Lundgren, 2010. Previous period environmental investments significantly impact on the current energy efficiency gain, while the impact on productivity growth is significantly negative. Expanding environmental investment would induce energy efficiency improvement, but it may occupy resources and thus generate barrier on spurring productivity growth further.

17. Conclusions Energy efficiency and environmental performance are integrated. Energy efficiency and productivity positively reinforce each other. Hence, increasing energy efficiency, as advocated in many of today’s energy policies, would capture multiple benefits. Improved environmental performance and environmental investments constrains next period productivity, a result that would be in contrast with 1) the so called Porter hypothesis, and 2) so called strategic corporate social (environmental) responsibility (CSR). However, environmental investment, if channeled via improved energy efficiency, may affect productivity positively in future periods, signifying the role of investing in energy efficiency as a cost saving policy.

18. Policy implications