Environmental investment and firm performance: A network approach

Transcription

1 Environmental investment and firm performance: A network approach Moriah Bostian 1 Rolf Färe 2 Shawna Grosskopf 2 Tommy Lundgren 3 1 Department of Economics, Lewis & Clark College and University of Turku 2 Department of Economics, Oregon State University 3 Centre for Environmental and Resource Economics, Umeå University June 17, 2015

2 Introduction Reductions to energy use and emissions often require investment in new production methods. Environmental investments may spur innovations that increase productivity, through design improvements and spillover effects (Clarke et al., 2006). We introduce firm-level investments, both environmental and production oriented, into a network production model to better account for their role in: productivity, efficiency change and technology change energy efficiency environmental performance We apply this framework to firm-level data on Swedish manufacturing.

3 Introduction Existing environmental network studies include: Murty et al. (2012) develop a bi-production technology, connecting conventional production to a pollution-generating technology. Färe et al. (2013) break the technology into a production stage and an abatement stage. Hampf (2014) incorporates a materials balance condition into a similar two-stage network. This study connects the production technologies for good and bad outputs through intertemporal investment decisions.

4 Environmental Investments and Expenditures Environmental investments: - Pollution treatment, or end-or-pipe techniques (e.g., air filters and scrubbers) - Pollution prevention processes (e.g., fuel switching/saving equipment and re-circulation of process gases) Environmental expenditures: - Operating costs of existing environmental equipment, internal monitoring, personnel training, and remediation costs. (Juraite et al., 2014)

5 Modeling Framework We model the production technology,t, for time period t as T t = { (x t, i t 1, y t, u t ) : x t and i t 1 can produce y t and u t}, x t = (x1 t,..., x N t ) denotes current period resource inputs i t 1 is total investment from the previous period y t = (y1 t,..., y M t ) is current period production output u t = (u1 t,..., ut J ) is current period pollution emissions. We further decompose the technology: i t = pi t + ei t, production and environmental investments x t = ex t + ox t, energy inputs and other inputs

6 Modeling Framework Figure : The Environmental Investment Technology Network We estimate the network technology nonparametrically, using Data Envelopment Analysis (DEA) methods (Charnes et al., 1978).

7 DEA Network Technology Model T t ={y t, u t, x t, pi t 1, ei t 1 } : K ym t zk t y km t, m = 1,..., M, K uj t = z t k ut kj, j = 1,..., J, K xn t zk t x kn t, n = 1,..., N, K pi t 1 K ei t 1 z t k z t k z t k pi t 1 k, ei t q k, 0, k = 1,..., K.

8 Network Technology Indexes Our performance indexes are based on quantity indexes, constructed from distance functions (Shephard, 1970), for production output, energy use and pollution emissions. Beginning with production output, y, in time period τ, we let D τ y (x τ, i τ 1, y τ, u τ ) = inf {θ : (x τ, i τ 1, y τ θ, uτ ) T τ } θ measures the feasible proportional increase in current production output, given current inputs, previous investment, and the production technology. 0 θ 1 and θ = 1 at the frontier.

9 Production Output Distance Estimation (D t y (x t, i t 1, y t, u t )) 1 = max θ K θyk t m zk t y km t, m = 1,..., M, u t k j = K z t k ut kj, j = 1,..., J, K xk t n zk t x kn t, n = 1,..., N, ei t 1 k pi t 1 k z t k K K z t k z t k ei t 1 k, pi t 1 k, 0, k = 1,..., K.

10 Malmquist Productivity Index This gives rise to the widely applied Malmquist (1953) productivity index, adapted to our network setting, M y (τ, τ + 1) = [( ) ( D τ y (x τ+1, i τ, y τ+1, u τ+1 ) D τ+1 D τ y (x τ, i τ 1, y τ, u τ ) where efficiency change is measured as y (x τ+1, i τ, y τ+1, u τ+1 ) D τ+1 y (x τ, i τ 1, y τ, u τ ) EC y (τ, τ + 1) = Dτ+1 y (x τ+1, i τ, y τ+1, u τ+1 ) Dy τ (x τ, i τ 1, y τ, u τ, ) and technology change is TC y (τ, τ + 1) = [( D τ y (x τ+1, i τ, y τ+1, u τ+1 ) D τ+1 y (x τ+1, i τ, y τ+1, u τ+1 ) ) ( D τ y (x τ, i τ 1, y τ, u τ ) )] 1 2, )] 1 2 Dy τ+1. (x τ, i τ 1, y τ, u τ )

11 Energy Efficiency Index (EEI) To measure energy efficiency, we consider feasible contractions of energy use, given other inputs, investment and outputs using D τ ex(ex τ, ox τ, i τ 1, y τ, u τ ) = sup{λ : ( ex τ λ, ox τ, i τ 1, y τ, u τ ) T τ }, λ 1 and λ = 1 at the frontier.

12 Energy Input Distance Estimation (D ex (ex t, ox t, i t 1, y t, )) 1 = min λ, s.t. K yk t m zk t y km t, m = 1,..., M, u t k j = K z t k ut kj, j = 1,..., J, K xk t n zk t x kn t, n = 1,..., N 1, K λexk t zk t ex k t, ei t 1 k pi t 1 K z t k ei t 1 k, K z t pi t 1,

13 Energy Efficiency Index (EEI) We measure energy efficiency change similarly, using the energy efficiency index (EEI), EEI (τ, τ + 1) = D ex (ex τ, ox τ, i τ 1, y τ, u τ ) D ex (ex τ+1, ox τ+1, i τ, y τ+1, u τ+1 ), EEI values greater than one indicate improvements to energy efficiency.

14 Environmental Performance Index (EPI) To measure environmental performance, we construct quantity indexes for both desirable and undesirable outputs. Our desirable quantity index is written as Q y (τ, τ + 1) = Dτ+1 y (x τ+1, i τ, y τ+1, u τ+1 ) Dy τ (x τ, i τ 1, y τ, u τ. )

15 Environmental Performance Index (EPI) For pollution emissions, we measure feasible contractions, given inputs, investment, and production output with D τ u (x τ, i τ 1, y τ, u τ ) = sup{δ : (x τ, i τ 1, y τ, uτ δ ) T τ }. δ 1 and δ = 1 at the frontier.

16 Pollution Output Distance Estimation (D u (x t, i t 1, y t, u t )) 1 = min δ s.t. K yk t m zk t y km t, m = 1,..., M, K xk t n zk t x kn t, n = 1,..., N, δu t k j = K ei t 1 k pi t 1 k z t k K K z t k ut kj z t k z t k, j = 1,..., J, ei t 1 k, pi t 1 k, 0, k = 1,..., K,

17 Environmental Performance Index (EPI) The corresponding pollution quantity index is Q u (τ, τ + 1) = Dτ+1 u (x τ+1, i τ, fy τ+1, u τ+1 ) Du τ (x τ, i τ 1, y τ, u τ. ) Note that Q u (τ, τ + 1) < 1 indicates improvement.

18 Environmental Performance Index (EPI) We then construct the EPI as the ratio of good to bad quantity indexes, following Färe et al. (2006; 2010) as, EPI (y τ, y τ+1, u τ, u τ+1 ) = Q y (τ, τ + 1) Q u (τ, τ + 1). An EPI value greater than 1 indicates a relative increase in goods to bads.

19 Application to Swedish Manufacturing Firms We apply our environmental investment network technology and associated performance index framework to an unbalanced panel of 66 firms from Sweden s Pulp and Paper manufacturing sector, for the years Our data includes at the plant level: Expenditures for standard production and energy inputs Expenditures for environmental management practices and environmental investments Emissions of CO 2, SO 2, and NO X

20 Application to Swedish Manufacturing Firms Capital and Output are measured in M 2008 SEK. Environmental Investments and Expenditures are measured in 1,000s 2008 SEK. Table : Summary Statistics, Variable Median Mean Min Max Capital , , Labor , Energy (MWH) 369, , , ,319, EnvExp 9, , , Env. Inv. 1,962 17, , Output 120, , , ,228, CO2 (T) 20, , , SO2 (T) NOX (T)

21 Index Results Table : Industry Geometric Means by Year Year Malm. EffCh. TechCh. EEI EPI Qy Qb The Malmquist means indicate a slight overall decline in productivity, driven more by a decrease in Technology. The EEI results indicate general improvements to energy efficiency; the EPI indicates an overall decline to environmental performance, due largely to the period.

22 Industry Trends Figure : Malmquist Industry Cumulative Geometric Means

23 Industry Trends Figure : Malmquist, EEI and EPI Cumulative Geometric Means

24 Index Results at the Plant Level Figure : EPI and Malmquist Results at the Plant Level

25 Index Results at the Plant Level Figure : EEI and Malmquist Results at the Plant Level

26 Index Results at the Plant Level Figure : EEI and EPI Results at the Plant Level

27 Conclusion We develop a network technology representation to measure energy efficiency, environmental performance and productivity change, accounting for intertemporal investment decisions. For firms in Sweden s pulp and paper sector, we find: General improvements to energy efficiency Overall decline in environmental performance Slight decline to overall productivity Positive relationship between energy efficiency, environmental performance, and productivity

28 Conclusion To our knowledge, this is the first study to include environmental investments in a network environmental technology framework. Future extensions include: Integrating the investment network with an abatement technology network. Incorporate dynamic optimization, to estimate optimal investment time paths, for both environmental and production objectives.