### 1. Introduction

### 2. Literature Review and Hypothesis Development

### 3. Data and Research Design

### 3.1 Data

### 3.2 Variable Construction

#### 3.2.1 Firm’s GHG Emission

*Scope 1 GHG Emission and Scope 2 GHG Emission*, respectively.

#### 3.2.2 Institutional Ownership

#### 3.2.3 Control Variables

*Firm Growth*) in regressions. Firms with higher growth rates are likely to emit more GHG because they manufacture more products. Moreover, we control for intangible assets as measured by selling and general administration expenses (Eisfeldt and Papanikolaou, 2013). Investment in intangible assets, such as information and organizational capital, is crucial to increase firms’ survival and growth (e.g., Lev et al., 2009; Li et al., 2018; Francis et al., 2021, Kim et al., 2021). Therefore, firms with more intangible assets are less likely to generate GHS emissions because they are better able to focus on CER activities to enhance their long-term value. The intangible assets of a firm are measured as selling, general, and administrative expenses divided by total assets (denoted as

*Intangibility*).

*Cash*) is computed as cash and short-term investments divided by total assets, whereas leverage (denoted as

*Leverage*) is calculated as the sum of short- and long-term debt divided by total assets. In addition, larger firms are more likely to effectively manage their GHG emissions due to economies of scale. Thus, we control for firm size (denoted as

*Firm*

*Size*), which is measured as the natural log of the total assets plus 1. To minimize the outlier effect, the top and bottom 1% of all continuous variables are winsorized. All variable definitions are provided in the Appendix.

#### 3.2.4 Summary Statistics and Correlations

*median*) Scope 1 GHG emission and Scope 2 GHG emission are 2,493,776 (

*112,310.7*) and 647,234.6 (

*102,088.5*), respectively. These GHG emission variables have mean values that are greater than their median values, indicating that they are right-skewed. In regression analyses, we therefore use their logarithmically transformed values to account for the positive skewness.

##### <Table 1>

### 4. Empirical Findings

### 4.1 Univariate Tests

*the other group*with institutional ownership above and

*below*the sample firms’ median value, respectively. Subsequently, we test whether significant differences exist among these two subsamples.

*median*) of

*Scope 1 GHG Emission*(GHG emissions / total sales) is 137.207 (

*22.638*). In contrast, this mean (

*median*) value of this variable for those with lower institutional ownership is 243.754 (

*27.637*). The mean and median differences among these two groups are statistically significant at the 1% level. In addition, the same analyses performed for

*Scope 2 GHG Emission*yield the identical results. Overall, our results show that firms with a higher proportion of institutional ownership tend to generate fewer GHG emissions than those with a lower proportion of institutional ownership.

##### <Table 2>

### 4.2 Baseline Regression Analysis

*GHG Emission*is the GHG emission rate of firm

_{i,t}*i*in year

*t*. Institutional Ownership

_{i,t-1}is the percentage ownership of firm

*i*held by institutional investors in year

*t-1*. X is a set of control variables, including

*Firm Growth, Intangibility, Cash, Leverage, and Firm Size*. Detailed definitions of the variables are provided in the Appendix. The regressions include year and industry fixed effects, which are defined based on two-digit SIC codes, to control for the specific effects of year and industry. All independent variables are 1-year lagged to mitigate concerns about reverse causality. Standard errors are robust to heteroscedasticity and clustered at the firm level.

*Scope 1 GHG Emission*, which measures the direct GHG emissions generated by a firm’s operations, as our main dependent variable. In addition, we employ

*Scope 2 GHG Emission*, which measures the firm’s indirect GHG emissions, as our complemental variable. Results on the relationship between institutional ownership and a firm’s GHG emissions are presented in <Table 3>.2) The dependent variables in Columns (1) and (2) are

*Scope 1 GHG Emission*and

*Scope 2 GHG Emission*, respectively. In both columns, the coefficients on institutional ownership (i.e., -0.519 and -0.295) are significantly negative at the 1% level , thus indicating that there is a negative association between institutional ownership and a firm’s direct and indirect GHG Emissions.3) Overall, the results suggest that institutional investors induce firms to decrease their GHG emissions, leading to an increase in CER activities.

##### <Table 3>

### 4.3 Endogeneity Issue

#### 4.3.1 Two-stage Least Squares Regressions

*Illiquidity*), which is equal to one if a firm’s stock illiquidity is above the industry median; otherwise, it is zero. Institutional investors prefer more liquid stocks because illiquid stocks may convey to the market a bad signal about the firms’ performance, facing demand pressure (Amihud et al., 2006).

*Institutional Ownership*is the percentage ownership of firm

_{i,t}*i*held by institutional investors in year

*t*.

*Illiquidity*is equal to one if stock illiquidity for firm

_{i,t-1}*i*in year

*t-1*is above the industry median value; otherwise, it is zero.

*GHG Emission*is the GHG emission rate of firm

_{i,t}*i*in year

*t*. ∑

*C*

_{i,t-1}is a set of control variables for firm

*i*in year

*t*. ∑

*I*and ∑

_{i}*ηI*are industry dummies, while ∑

_{i}*Y*and ∑

_{t}*θY*are year dummies.

_{t}*Instrumented Institutional Ownership*is the fitted value obtained from the first stage regression.

*Illiquidity*is negatively associated with

*Institutional Ownership*at the 1% level. The F-statistics for the weak instrumental test at the first-stage regression is 28.038, which is greater than the acceptance value of 10 suggested by Stock and Yogo (2005). This result shows that the null hypothesis of weak instrumental variables is rejected by the test.

##### <Table 4>

*Scope 1 GHG Emission*as a dependent variable. The coefficient on

*Instrumented Institutional Ownership*is significantly negative at the 1% level. Further, we conduct the 2SLS regressions, including

*Scope 2 GHG Emission*. The coefficient on

*Instrumented Institutional Ownership*in column (4) is significantly negative. The results suggest that institutional investors tend to induce firms to reduce their GHG emissions, confirming that our baseline findings are unlikely to be driven by endogeneity issues.

#### 4.3.2 Change Regressions

*t-1*to

*t*are denoted as Δ

*Scope 1 GHG Emission*(or Δ

*Scope 2 GHG Emission*). Moreover, changes in institutional ownership from year

*t-2*to

*t-1*are denoted as Δ

*Institutional Ownership*. <Table 5>’s Columns (1) and (2) show that the coefficients for Δ

*Institutional Ownership*are significantly negative. These findings indicate that an increase in institutional ownership will likely induce firms to reduce their GHG emissions.

##### <Table 5>

#### 4.3.3 Regressions with Firm Fixed Effects

*Scope 1 GHG Emission*and

*Scope 2 GHG Emission*as dependent variables in columns (1) and (2), respectively. We find that the negative association between institutional ownership and a firm’s direct GHG emissions is statistically significant. Meanwhile, the relationship between institutional ownership and the firm’s indirect GHG emissions is insignificant. These results suggest that earlier findings on the firm’s direct GHG emissions still persist after addressing an omitted variable issue.