Dear stata users,
I am working on a replication study for an assignment, but am stuck.
One of the tables reports coefficients from estimating the first-stage regression of the 2SLS model. the description is as follows:
The determinants of CEO decision horizon. This table reports coefficients from estimating the first-stage regression of the 2SLS model. The first step involves a
regression model wherein decision horizon (DH) is estimated. The predicted value of DH is used in the second-stage models in other tables. DH = decision
horizon, SIZE = log of total assets, LEV = long-term debt scaled by total assets. EBIT = earnings before interest and taxes, CAPX = capital expenditure, R&D =
research and development expenditure, ADV=advertising expenditure, SALES=Gross sales, TOBINQ=Tobin's q, ISIZE=the percentage of other firms that are
larger than the firm in the same industry, ICOMP = the percentage of other CEOs who are paid more than the CEO in the same industry, ECOMP = the ratio of
equity compensation to total compensation. *, **, and *** denote significance at the 10%, 5%, and 1% levels.
I have trouble generating ICOMP and ISIZE and running the right regression when I have found these variables.
If there is someone able to help, it is really appreciated.
kind regards,
Stef
Additional info:
To reduce the potential problem of endogeneity in the multi-factor regression models, we use a two-stage least squares (2SLS)
model that involves the estimation of two regression models, one for CEO decision horizon and one for firm performance. The 2SLS
procedure requires that the first-stage equation contain at least one instrumental variable that is unrelated to the error term in the
second-stage model. Here, we use industry characteristics as instruments. Our full 2SLS model is structured as follows.
DH = f ðfirm performance; firm characteristics; CEO compensation package; and instrumentalsÞ; ð3Þ
where firm performance (TOBINQ) is computed as [market value of common equity+preferred stock liquidating value+longterm
debt−(short-term assets−short-term liabilities)] / (total assets), following Chung and Pruitt (1994). Firm characteristics
are: the log of total assets (SIZE), the ratio of long-term debt to total assets (LEV), profitability (EBIT/SALES), capital expenditure
(CAPX/SALES), and intangibles' intensity that is measured as the sum of R&D expenditure (R&D) and advertising expenditure
(ADV) scaled by SALES. The first-stage model also accounts for the fact CEOs approaching retirement are likely to receive different
pay packages than those recently put into office (see Anderson et al., 2006).23 For example, to mitigate the horizon problem,
associated with CEOs nearing retirement, firms are likely to increase incentive-based compensation.24 To account for that, we
include a CEO's equity compensation ratio (ECOMP), which is computed as the value of unexercised stock options (OPTION)
divided by the value of total compensation (TCOMP). Gibbons and Murphy (1992) argue that since career concerns are weaker
when executives near retirement, incentive contracts should be the strongest for these workers. We therefore expect to find a
negative relationship between DH and ECOMP. In addition, we include industry characteristics because the job market
environment in the industry in which a CEO competes can be related to the CEO's decision horizon. Based on this expectation, we
control for two variables: the percentage of other firms that are larger than the firm in the same industry (ISIZE) and the
percentage of other CEOs who are paid more than the CEO in the same industry (ICOMP). Both ISIZE and ICOMP serve as proxies for
the existence of better employment opportunities in the industry, which are expected to have an impact on CEO decision horizon.
Specifically, we anticipate that in the presence of good employment opportunities in the industry, a CEO will be more inclined to
adopt a long-term horizon in order to improve external employment opportunities in the future. Finally, several indicator
variables that account for year, industry and exchange listing effects are also included. The industry indicators are based on the
Fama-French 12 industry classification.
I am working on a replication study for an assignment, but am stuck.
One of the tables reports coefficients from estimating the first-stage regression of the 2SLS model. the description is as follows:
The determinants of CEO decision horizon. This table reports coefficients from estimating the first-stage regression of the 2SLS model. The first step involves a
regression model wherein decision horizon (DH) is estimated. The predicted value of DH is used in the second-stage models in other tables. DH = decision
horizon, SIZE = log of total assets, LEV = long-term debt scaled by total assets. EBIT = earnings before interest and taxes, CAPX = capital expenditure, R&D =
research and development expenditure, ADV=advertising expenditure, SALES=Gross sales, TOBINQ=Tobin's q, ISIZE=the percentage of other firms that are
larger than the firm in the same industry, ICOMP = the percentage of other CEOs who are paid more than the CEO in the same industry, ECOMP = the ratio of
equity compensation to total compensation. *, **, and *** denote significance at the 10%, 5%, and 1% levels.
I have trouble generating ICOMP and ISIZE and running the right regression when I have found these variables.
If there is someone able to help, it is really appreciated.
kind regards,
Stef
Additional info:
To reduce the potential problem of endogeneity in the multi-factor regression models, we use a two-stage least squares (2SLS)
model that involves the estimation of two regression models, one for CEO decision horizon and one for firm performance. The 2SLS
procedure requires that the first-stage equation contain at least one instrumental variable that is unrelated to the error term in the
second-stage model. Here, we use industry characteristics as instruments. Our full 2SLS model is structured as follows.
DH = f ðfirm performance; firm characteristics; CEO compensation package; and instrumentalsÞ; ð3Þ
where firm performance (TOBINQ) is computed as [market value of common equity+preferred stock liquidating value+longterm
debt−(short-term assets−short-term liabilities)] / (total assets), following Chung and Pruitt (1994). Firm characteristics
are: the log of total assets (SIZE), the ratio of long-term debt to total assets (LEV), profitability (EBIT/SALES), capital expenditure
(CAPX/SALES), and intangibles' intensity that is measured as the sum of R&D expenditure (R&D) and advertising expenditure
(ADV) scaled by SALES. The first-stage model also accounts for the fact CEOs approaching retirement are likely to receive different
pay packages than those recently put into office (see Anderson et al., 2006).23 For example, to mitigate the horizon problem,
associated with CEOs nearing retirement, firms are likely to increase incentive-based compensation.24 To account for that, we
include a CEO's equity compensation ratio (ECOMP), which is computed as the value of unexercised stock options (OPTION)
divided by the value of total compensation (TCOMP). Gibbons and Murphy (1992) argue that since career concerns are weaker
when executives near retirement, incentive contracts should be the strongest for these workers. We therefore expect to find a
negative relationship between DH and ECOMP. In addition, we include industry characteristics because the job market
environment in the industry in which a CEO competes can be related to the CEO's decision horizon. Based on this expectation, we
control for two variables: the percentage of other firms that are larger than the firm in the same industry (ISIZE) and the
percentage of other CEOs who are paid more than the CEO in the same industry (ICOMP). Both ISIZE and ICOMP serve as proxies for
the existence of better employment opportunities in the industry, which are expected to have an impact on CEO decision horizon.
Specifically, we anticipate that in the presence of good employment opportunities in the industry, a CEO will be more inclined to
adopt a long-term horizon in order to improve external employment opportunities in the future. Finally, several indicator
variables that account for year, industry and exchange listing effects are also included. The industry indicators are based on the
Fama-French 12 industry classification.
Comment