Note(s): This table presents 2SLS and system-GMM estimates for the effect of OK on investment inefficiency. 2SLS estimates are presented first in columns 1–4. The first stage 2SLS regression outputs are reported in column (1) for the first OC measure (OKTA) and column (3) for the second measure (OKTC). The second stage results are shown in columns 2 (OKTA) and 4 (OKTC). System-GMM estimates results are shown in columns 5 and 6 for OKTA and OKTC, respectively. For the system-GMM, the Sargan test statistics is the test statistics of overidentifying restrictions for the validity of instruments, and AR(2) is a test for autocorrelation indifferences. Additionally, for the system-GMM regressions, we add the second lag of the dependent variable as a control variable to control for the persistence of investment inefficiency. For the 2SLS regressions, we follow extant literature (Li et al., 2018; Hasan and Cheung, 2018; Carlin et al., 2012) and employ industry-level growth uncertainty as an instrumental variable. For both 2SLS and system-GMM models, OKTA is the ratio of OK to total assets, while OKTC is the ratio of OK to total capital. INV represents investment inefficiency or the deviation from the optimal investment and is computed as the residuals from the investment model. The control variables include Size, ROA, Slack, Tangibility, Firm age, Leverage, Tobin’s q, and Loss. All remaining variables are defined according to  Appendix 1. Figures in parentheses are standard errors that are clustered at the firm level for heteroscedasticity. (^***) (^**) (^*) significance at the (1%) (5%) (10%) levels

Source(s): Authors’ own work