XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 285
its significance with a similar coefficient. Previously non-
significant variables, Total Communications (p =0.039),
Total Assessment &Measurement (p =0.039), and Total
Incentives (p =0.068), achieved marginal significance.
Other variables, including “total_adoptofstandards,” “total_
donationfundings,” “total_modificationofprocedures,” etc.,
remained statistically insignificant in both models. Notably,
Total Asset Modifications’ significance vanished (p =0.504),
indicating no conclusive impact on ROA. The year 2021
emerged as significant (p =0.017), suggesting a difference
in ROA between 2020 and 2021 (Table 7).
Fixed Effects Regression with Multi-Way Clustering of
Errors
Robustness checks employing linear and instrumental-
variable regressions with various fixed effects levels were
conducted. The new approach facilitated multi-way error
clustering to address heteroskedasticity and autocorrela-
tion. This model yielded a substantially higher R-squared
(0.7541) compared to previous models (Table 8), indi-
cating a larger proportion of variance in ROA explained.
Control variables “size” (p =0.018) and “energy_efficiency”
(p =0.006, negative) maintained significance, with the lat-
ter confirming a negative association between energy effi-
ciency and ROA. “Total_associations” (p =0.029, negative)
remained significant, suggesting a negative relationship
with ROA. Notably, “Total_assetmodifications” achieved
significance (p =0.049), indicating a positive association
with ROA. Other variables, including “total_adoptof-
standards_rules,” “total_communications,” etc., remained
nonsignificant.
Hypotheses Testing and Analysis-Tobin’s Q
Building on the insights from the ROA analysis, a similar
staged approach was applied to examine the relationship
between sustainability behaviors and Tobin’s Q.
Hypothesis 1b—Association between specific
sustainability behaviours and financial performance
(Tobin’s Q):
• Null Hypothesis (H0): There is no statistically
significant association between specific sustainabil-
ity behaviours and higher financial performance
(Tobin’s Q).
• Alternative Hypothesis (H1): There is a statisti-
cally significant association between specific sustain-
ability behaviours and higher financial performance
(Tobin’s Q).
Tobin Q EE LAGE
SIZE
COMMODITY i.YEAR
Sustainability Behavior
f,t f,t f,t
f,t f,t
f f,t
i.f
3 4
5
8 11
a b b
b b6LEVARAGE
b b
+/bi
=++
++
++
Table 8. Regression results with multi-way clustering of errors, fixed effects
return_on_asset Coefficient std. err. t Pt [95% conf. interval] Sig
fncl_lvrg –4.414 2.323 –1.9 0.071 –9.232 0.404 *
size 7.696 3.020 2.55 0.018 1.433 13.959 **
energy_efficiency –7.911 2.571 –3.08 0.006 –13.243 –2.579 ***
total_adoptofstandards_rules 4.935 3.062 1.61 0.121 –1.414 11.285
total_associations –1.672 0.714 –2.34 0.029 –3.154 –0.191 **
total_communications 0.843 0.546 1.54 0.137 –0.289 1.974
total_donationfundings 0.149 0.616 0.24 0.811 –1.129 1.427
total_assetmodifications 0.953 0.459 2.08 0.049 0.002 1.904 **
total_incentives –3.156 1.713 –1.84 0.079 –6.708 0.397 *
total_modificationofprocedures 0.275 0.464 0.59 0.559 –0.686 1.237
total_newproducts –1.449 1.021 –1.42 0.17 –3.566 0.667
total_orgstructuring 1.159 1.279 0.91 0.375 –1.494 3.811
total_rdinvestments –0.337 0.892 –0.38 0.71 –2.187 1.514
total_training –0.169 0.810 –0.21 0.836 –1.850 1.511
total_volunteerism 3.326 4.104 0.81 0.426 –5.186 11.837
total_assess_measurement –0.677 0.408 –1.66 0.111 –1.523 0.169
_cons –61.753 30.615 –2.02 0.056 –125.245 1.739
R-squared 0.7541 Number of obs 90.000
F-test 66.97 Prob F 0
Number of clusters (comp_id) 23 Root MSE 5.277
***p.01, **p.05, *p.1
its significance with a similar coefficient. Previously non-
significant variables, Total Communications (p =0.039),
Total Assessment &Measurement (p =0.039), and Total
Incentives (p =0.068), achieved marginal significance.
Other variables, including “total_adoptofstandards,” “total_
donationfundings,” “total_modificationofprocedures,” etc.,
remained statistically insignificant in both models. Notably,
Total Asset Modifications’ significance vanished (p =0.504),
indicating no conclusive impact on ROA. The year 2021
emerged as significant (p =0.017), suggesting a difference
in ROA between 2020 and 2021 (Table 7).
Fixed Effects Regression with Multi-Way Clustering of
Errors
Robustness checks employing linear and instrumental-
variable regressions with various fixed effects levels were
conducted. The new approach facilitated multi-way error
clustering to address heteroskedasticity and autocorrela-
tion. This model yielded a substantially higher R-squared
(0.7541) compared to previous models (Table 8), indi-
cating a larger proportion of variance in ROA explained.
Control variables “size” (p =0.018) and “energy_efficiency”
(p =0.006, negative) maintained significance, with the lat-
ter confirming a negative association between energy effi-
ciency and ROA. “Total_associations” (p =0.029, negative)
remained significant, suggesting a negative relationship
with ROA. Notably, “Total_assetmodifications” achieved
significance (p =0.049), indicating a positive association
with ROA. Other variables, including “total_adoptof-
standards_rules,” “total_communications,” etc., remained
nonsignificant.
Hypotheses Testing and Analysis-Tobin’s Q
Building on the insights from the ROA analysis, a similar
staged approach was applied to examine the relationship
between sustainability behaviors and Tobin’s Q.
Hypothesis 1b—Association between specific
sustainability behaviours and financial performance
(Tobin’s Q):
• Null Hypothesis (H0): There is no statistically
significant association between specific sustainabil-
ity behaviours and higher financial performance
(Tobin’s Q).
• Alternative Hypothesis (H1): There is a statisti-
cally significant association between specific sustain-
ability behaviours and higher financial performance
(Tobin’s Q).
Tobin Q EE LAGE
SIZE
COMMODITY i.YEAR
Sustainability Behavior
f,t f,t f,t
f,t f,t
f f,t
i.f
3 4
5
8 11
a b b
b b6LEVARAGE
b b
+/bi
=++
++
++
Table 8. Regression results with multi-way clustering of errors, fixed effects
return_on_asset Coefficient std. err. t Pt [95% conf. interval] Sig
fncl_lvrg –4.414 2.323 –1.9 0.071 –9.232 0.404 *
size 7.696 3.020 2.55 0.018 1.433 13.959 **
energy_efficiency –7.911 2.571 –3.08 0.006 –13.243 –2.579 ***
total_adoptofstandards_rules 4.935 3.062 1.61 0.121 –1.414 11.285
total_associations –1.672 0.714 –2.34 0.029 –3.154 –0.191 **
total_communications 0.843 0.546 1.54 0.137 –0.289 1.974
total_donationfundings 0.149 0.616 0.24 0.811 –1.129 1.427
total_assetmodifications 0.953 0.459 2.08 0.049 0.002 1.904 **
total_incentives –3.156 1.713 –1.84 0.079 –6.708 0.397 *
total_modificationofprocedures 0.275 0.464 0.59 0.559 –0.686 1.237
total_newproducts –1.449 1.021 –1.42 0.17 –3.566 0.667
total_orgstructuring 1.159 1.279 0.91 0.375 –1.494 3.811
total_rdinvestments –0.337 0.892 –0.38 0.71 –2.187 1.514
total_training –0.169 0.810 –0.21 0.836 –1.850 1.511
total_volunteerism 3.326 4.104 0.81 0.426 –5.186 11.837
total_assess_measurement –0.677 0.408 –1.66 0.111 –1.523 0.169
_cons –61.753 30.615 –2.02 0.056 –125.245 1.739
R-squared 0.7541 Number of obs 90.000
F-test 66.97 Prob F 0
Number of clusters (comp_id) 23 Root MSE 5.277
***p.01, **p.05, *p.1