Docstoc

Is there an Environmental Kuznets Curve at the Firm Level

Document Sample
Is there an Environmental Kuznets Curve at the Firm Level Powered By Docstoc
					Is there an Environmental Kuznets Curve at the Firm Level?

                        Evidence from the Forest Industry


                        Jesse Yamaguchi, University of Victoria REPA


Abstract

This study examines the relationship between corporate environmental performance (CEP) and corporate

financial/economic performance (CFP) at the firm level with specific application to the North American forest

product industry. An unbalanced panel of firm level observations is constructed using data from

PricewaterhouseCoopers’ (PWC) Global Forest, Paper and Packaging Industry Survey, the EPA’s Toxic Release

Inventory (TRI), and Environment Canada’s National Pollutant Release Inventory (NPRI). With formaldehyde

and methanol as dependant variables, we find evidence of a firm level Environmental Kuznets Curve (EKC) in

static models. In dynamic models, the EKC loses significances. However, there is weak evidence of a negative

linear relationship for formaldehyde.

                                                                                                                1
                          Motivation

1. “win-win” hypothesis

 Porter and van der Linde (1995), “Toward a new conception of the

 environment-competitiveness relationship”



 ‘end of the pipe’ vs. ‘innovation offsets’ ?



2. Environmental Kuznets Curve

 Macro-phenomenon, Grossman and Krueger (1991)

 - Environmental quality is a normal good

 - Technological and composition effects overtake scale effect.
                                                                    2
              Macro EKC and Firm Level EKC

Pollution per capita

Pollution per product




                                Country i’s GDP per capita

                                Firm i’s Profitability




      Viewed over cross-section, over time, or both.
                                                             3
                         Research Questions



1. By modeling the relationship of pollution as a quadratic function of

  profitability, will the resulting parameter estimates produce an

  inverted U-shaped, firm-level EKC?



2. If so, will the firm-level EKC be robust to realistic modifications,

  such as capturing the dynamic effects by the inclusion of lagged

  dependant variables?


                                                                          4

                  Support for ‘win-win’ hypothesis?
                                Methods



1. Sample selection: PricewaterhouseCoopers’ Global Forest, Paper

and Packaging Industry Survey. Years 1997 to 2006.

21 U.S. and Canadian companies, observed from 1996 to 2005

(unbalanced)



2. Dependent variables: formaldehyde (H2CO) and methanol

(CH3OH) from EPA’s TRI and Environment Canada’s NPRI online

registries at facility level.

                                                                    5
                         Choice of Pollutants?

   1.Availability

   2.Human health and Environmental Damage

   3.Listed as Hazardous Air Pollutants in U.S. EPA’s Clean Air

     Act.

-Milota (2006) measures methanol and formaldehyde emissions from

drying process of wood from different species of trees.

  “Title III requires Maximum Achievable Control Technology (MACT)

standards to be established for hazardous air pollutants and requires that

control equipment be placed on some point sources, such as certain types of

dryers and presses.” (Milota, 2006, p. 79)                                    6
                     Explanatory Variables



1.Return on Capital Employed:

ROCE = (Net Income/(Average Total Assest – Average Current

Liabilities))×100%



2.# Standard Industrial Code facilities:

SIC 24 = Lumber and wood products, SIC 26 = Pulp and paper




                                                             7
                               Summary Statistics
                     Mean    Median Maximum Minimum Std Dev. Observations
Formaldehyde
(Kgs.)       126,144.3 28,215.49       795,374.4    53.5239 193,545.3   112

Formaldehyde/
sales
(Kgs./$U.S.
million)           39.6918   15.2415    276.935    0.041363 54.35043    112

Methanol
(Kgs.)            2,205,638 967,819.1 15,732,335 59,012.37 3,378,795    127

Methanol/sales
(Kgs./$U.S.
million)       508.2755 402.0923       2,249.261   61.05604 397.7971    127

ROCE (%)          5.400962      4.55        23.7       -8.5 4.844639    208

Sales
($U.S. million)    5326.75      2035      28180        471   6,804.53   208

SIC 24            9.314286         2         62          0 14.56108     210

SIC 26            8.657143         5         34          0      9.551   210
                                                                         8
                                      Correlations
                          Formaldehyde           Methanol
                             per unit of          per unit                        SIC SIC
             Formaldehyde         Sales Methanol  of Sales ROCE           Sales    24 26
Formaldehyde            1
Formaldehyde
per unit of
Sales                   0.344*            1
Methanol                0.558*        -0.133        1
Methanol per
unit of Sales               -0.094   -0.231*   0.203*        1
ROCE                        -0.088   0.217*    -0.180*   -0.006      1
Sales                   0.686*        -0.093   0.902*    -0.059   0.007      1
SIC 24                  0.847*        0.077    0.572*    -0.046 -0.172* 0.644*      1
SIC 26                  0.711*        -0.155   0.805*    0.010 -0.193* 0.780* 0.481*        1
* Significant at 5% level




                                                                                        9
                        Model Specifications

 Static Panel:



 Log(Pit/Sit ) = β1ROCEit + β2ROCE2it + δ′x it + μi + єit



 for i = 1,…, N firms and t = 1996,…, 2005

   - єit i.i.d. with mean zero, constant variance

   - μi capture firm specific effects.



Estimation methods: OLS, Random Effects, Fixed Effects (LSDV)
                                                            10
                              Results
Dependent Variable: log(Formaldehyde/Sales)
               Estimation methoda)
                                   Random
Variable           Pooled OLS        Effects            Fixed Effects
ROCE                   0.086       0.108**                  0.108**
                       (0.072)     (0.036)                  (0.035)
ROCE-
squared                -0.002      -0.009**                  -0.009**
                       (0.005)     (0.002)                   (0.002)
SIC 24                 0.054**     0.007                     -0.013
                       (0.013)     (0.011)                   (0.014)
SIC 26                -0.062**     -0.020                    0.049
                       (0.022)     (0.029)                   (0.089)
Intercept              2.419**     2.636**                   2.547**
                       (0.266)     (0.364)                   (0.489)
Observations           112         112                       112
R-squared              0.163       0.015                     0.009
Wald test (μi = 0),
χ2                                                           464.36**
Breusch-Pagan
test, χ2                           152.92**
                 2
Hausman test, χ                                              8.820*
Turning Point
ROCEc)                 26.300      6.276**                   6.038**
Notes
a) Standard errors are in parenthesis. Overall R-squared presented for
all models.
* and ** indicates significance at the 10% and 5% level, respectively.
b) Significance of Turning Point ROCE found by testing the non-linear
restriction: b1 / -2b2 = 0.                                              11
                       Results Contd.
Dependent Variable: log(Methanol/Sales)
               Estimation
               methoda)
                                  Random
Variable          Pooled OLS        Effects            Fixed Effects
ROCE                  -0.030      0.036**                  0.045**
                      (0.034)     (0.017)                  (0.017)
ROCE-
squared               0.001       -0.002*                   -0.002**
                      (0.002)     (0.001)                   (0.001)
SIC 24                0.005       0.000                     -0.006
                      (0.007)     (0.006)                   (0.007)
SIC 26                0.003       -0.021                    -0.066*
                      (0.010)     (0.015)                   (0.023)
Intercept             5.984**     6.003**                   6.375**
                      (0.161)     (0.236)                   (0.200)
Observations          127         127                       127
R-squared             0.024       0.016                     0.014
Wald test (μi = 0),
χ2                                                          643.32**
Breusch-Pagan
test, χ2                          282.31**
                 2
Hausman test, χ                                             6.630
Turning Point
ROCEc)                NA          10.796**                  10.934**
Notes
a) Standard errors are in parenthesis. Overall R-squared presented for
all models.
* and ** indicates significance at the 10% and 5% level, respectively.
b) Significance of Turning Point ROCE found by testing the non-linear
restriction:                                                             12
b1 / -2b2 = 0.
           Dynamic Environmental Kuznets Curve



First order autoregressive function:

Log(Pit/Sit ) = β1ROCEit + β2ROCEit2 + δ′x it + ρLog(Pi,t-1/Si,t-1 ) +

                                                               μi + єit

With μi and єit as random errors, єit is serially uncorrelated and,
for stationarity, ρ < ⎟ 1 ⎜.

  - Estimation methods: OLS, Fixed Effects, Blundell Bond
  GMM, Bias corrected fixed effects (Bruno, 2005).

  - Evidence that methanol series is non-stationary, ρ ≥ ⎟ 1 ⎜. Not
    a formal test.

                                                                      13
                          Results
Dependent Variable: log(Formaldehyde/Sales)
              Estimation
              method
                  Pooled        Fixed
                    OLS        Effects BB GMM          LSDVC
                                        Instrument Blundell
                                        Limit:       Bond
                                        Collapsed, Initial
Variable                                4 Lags       Values
ROCE                0.049        0.006         0.022       0.007
                  (0.055)      (0.037)       (0.037)     (0.057)
ROCE
squared            -0.002       -0.002        -0.002      -0.002
                  (0.004)      (0.003)       (0.003)     (0.004)
Lagged
Dependent
Variable         0.733**         0.118      0.524**     0.518**
                  (0.056)      (0.094)       (0.168)     (0.110)
SIC 24              0.004       -0.015         0.010      -0.012
                  (0.010)      (0.012)       (0.013)     (0.026)
SIC 26             -0.006        0.032        -0.016       0.059
                  (0.019)      (0.072)       (0.019)     (0.126)
Year             0.000**        -0.008      0.001**       -0.029
                  (0.000)      (0.022)       (0.000)     (0.042)
Observations           92           92            92          92
Firms                  16           16            16          16
Instruments                                        9
Standard errors are in parenthesis. Bootstrapped standard
errors in italic
* and ** indicates significance at the 10% and 5% level.           14
           Results, predetermined ROCE
Dependent Variable: log(Formaldehyde/Sales)
                 Estimation method
                 Blundell Bond Two-Step Estimator
                 Instrument limit:
                                        Collapsed, 3 Collapsed, 2
Variable                Collapsed              Lags         Lags
ROCE                          0.039            0.047       -0.001
                            (0.057)          (0.039)      (0.057)
ROCE-
squared                      -0.003           -0.005       -0.003
                            (0.004)          (0.004)      (0.003)
Lagged
Dependent
Variable                   0.639**          0.657**      0.497**
                            (0.158)          (0.116)      (0.125)
SIC 24                        0.002            0.005        0.005
                            (0.016)          (0.010)      (0.017)
SIC 26                       -0.001           -0.009        0.001
                            (0.033)          (0.022)      (0.039)
Year                       0.001**          0.001**      0.001**
                            (0.000)          (0.000)      (0.000)
Observations                     92               92           92
Instruments                      32               15           12
Firms                            16               16           16
Sargan Test,
χ2                         61.59**          40.49**      43.82**
          2
Hansen, χ                      8.75             6.19         3.48
Standard errors with Windmeijer’s (2005) correction are in
parenthesis.
* and ** indicates significance at the 10% and 5% level,            15
respectively.
         Results Contd. Linear function
Dependent Variable: log(Formaldehyde/Sales)
               Estimation method
               Blundell Bond Two-Step Estimator
               Instrument limit:
                             Collapsed, 3 Collapsed, 2
Variable       Collapsed     Lags          Lags
ROCE             -0.036       -0.033       -0.042**
                (0.038)      (0.030)        (0.016)


Lagged
Dependent
Variable        0.512**       0.588**        0.482**
                 (0.204)       (0.150)        (0.104)
SIC 24             0.006         0.007          0.000
                 (0.020)       (0.014)        (0.016)
SIC 26            -0.015        -0.013          0.009
                 (0.036)       (0.024)        (0.036)
Year            0.001**       0.001**        0.001**
                 (0.000)       (0.000)        (0.000)
Observations          92            92             92
Instruments           22            11              9
Firms                 16            16             16
Sargan Test,
χ2               28.02**        13.67**        11.56**
          2
Hansen, χ            11.9           9.44          1.63
Standard errors with Windmeijer’s (2005) correction are in
parenthesis.
* and ** indicates significance at the 10% and 5% level,     16

respectively.
                         Conclusions



1.Firm level EKC not robust to simple, realistic modification



2.Evidence, although very weak, of negative relationship

 between formaldehyde and ROCE.



3.Methanol pollution intensity really non stationary?




                                                                17
                       Recommendations



1.Increase width and length of data, especially on relevant

 pollutants



2.Structural equations based on theory




                                                              18
References


Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics,
   87, 11–143.


Bruno, G. (2005). XTLSDVC: Stata module to estimate bias corrected LSDV dynamic panel data models. Statistical Software
    Components S450101, Boston College Department of Economics, revised Sep 8, 2005.

Grossman, G. M., & Krueger, A. B. (1991). Environmental Impacts of a North American Free Trade Agreement. National Bureau of
   Economic Research Working Paper 3914, NBER, Cambridge MA.

Milota, M.R. (2006). Hazardous air pollutant emissions from lumber drying. Forest Products Journal, 56(7-8), 79-84.

Porter, M. E., & van der Linde, C. (1995). Toward a new conception of the environment-competitiveness relationship. The Journal of
    Economic Perspectives, 9(4), 97-118.

 Windmeijer, F. (2005). A finite sample correction for the variance of linear efficient two-step GMM estimators. Journal of
    Econometrics 126: 25–51.




                                                                                                                                     19