Is it really hard to motivate those who are not highly educated? Empirical evidence from a semi-parametric analysis

ABSTRACT

In this study, we estimate the principal-agent model of moral hazard using a semi-parametric estimation method. We estimate the principal-agent model of moral hazard using longitudinal data on firms and managerial compensation collected from the database of Beijing Juyuan Ruisi Data Technology Corporation Limited, called RESSET. The estimated primitives in the contracting model show that most managers with high school and master educations have significantly higher costs of moral hazard than managers with university and doctor educations. Managers with high school and master educations are harder to motivate because they lack the knowledge to fulfill the delegated tasks.

KEYWORDS:

• Moral hazard
• compensation
• conflits
• education
• semi-parametric
• china

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

¹ The number means thirty-seven thousand six hundred and ninety-seven.

² The number means two thousand six hundred and seventy-one

³ We assume that extant public disclosure laws prevent managers from anonymously trading the assets of their own firms

⁴ $100\mathrm{\%}$ shirking means being employed as a manager at the firm but pursuing interests totally different from those of the shareholders

⁵ Working diligently means being employed as a manager at the firm and working entirely in the shareholders’ interests

⁶ The participation constraint is met with equality when $j=1$

⁷ The participation constraint is met with equality when $j=2$. And incentive compatibility constraint is binding, too.

⁸ The same assumption is used in Gayle and Miller (2015)

⁹ In the compensation contracts under monitoring, managers are risk adverse $\rho >0$. The detailed satisfaction of condition $\frac{b_{t+1}}{b_t-1}>0$ can refer to Margiotta and Miller (2000).

¹⁰ The identification equation for risk parameter $\rho$ holds for its positive value. The identification equation for risk parameter $\rho$ does not hold for any of its negative values. Hence, $\rho$, identified from the fourth equation in Theorem 3.1, is positive.

¹¹ The detailed satisfaction of this condition can refer to Margiotta and Miller (2000)

¹² This condition has been used in Gayle and Miller (2009) to identify parameters in the truncated normal distribution of abnormal returns.

¹³ The definitions of variable $W_{ti}^{\ast }$, constants $c_2$, $c_1$, and $C$ can refer to Fan, Hu, and Truong (1994)

¹⁴ Details about the five steps of the ``naive” two-step approach can be found in Ackerberg et al. (2014)

¹⁵ It will be helpful to take log of sales and total assets in estimation. However, in the estimation process, we use the six indicators of firm characteristics only to divide firms into three different kinds, which are small firms with low return, medium firms with medium return and large firms with high return. And seven different moment functions are estimated conditional on three different kinds of firms. Therefore, we do not need to take log of these two variables. Furthermore, their correlations help us characterize firms into different sizes.

¹⁶ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as both one-sided hypothesis tests can imply the significant differences.

¹⁷ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as one-sided hypothesis test can imply the significant differences.

¹⁸ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as one-sided hypothesis test can imply the significant differences.

¹⁹ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare their differences. The differences are significant as long as one-sided hypothesis test can imply the significant differences.

²⁰ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare their differences. The differences are significant as long as both one-sided hypothesis tests can imply the significant differences.

²¹ Our bond price series is based on Treasury Bills with maturities of $1$, $3$, $5$, $7$, $10$, and $30$ years from the China Central Depository and Clearing Co., Ltd. Data on minimum wage are collected from the Ministry of Human Resources and Social Security of the People’s Republic of China

²² Hypothesis tests with significance on both sides of $0.05$ and $0.95$ are carried out based on 2,000 bootstrap samples to compare their differences.

²³ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as both one-sided hypothesis tests can imply the significant differences.

²⁴ Firms with different sizes and returns are selected based on firm characteristics.

²⁵ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as both one-sided hypothesis tests can imply the significant differences.

²⁶ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the difference. The difference is significant as long as both one-sided hypothesis tests can imply the significant difference.

²⁷ These positions’ numbers in the RESSET database are $10$, $16$, $23$, $42$. $44$. $45$, $46$, $101$.

²⁸ We did not characterize positions of operators as non-CEO position. Because there are only three positions of operators in the database, the numbers of which are $12$, $24$, $120$, can be characterized as non-CEO position.

²⁹ These positions’ numbers in the RESSET database are $15$, $17$, $19$, $25$, $27$. $29$. $30$, $35$, $40$.

³⁰ These positions’ numbers in the RESSET database are $11$, $18$, $20$, $22$, $28$, $34$, $41$.

³¹ These positions’ numbers in the RESSET database are $13$, $26$, $31$, $38$, $39$, $43$, $125$, $210$, $213$, $500$, $999$.

³² The costs of moral hazard of managers in CEO position are presented in the left figure of the first row. The costs of moral hazard of managers in COS position are presented in the right figure of the first row. The costs of moral hazard of managers in DCOS position are presented in the left figure of the second row. The costs of moral hazard of managers in lower position are presented in the right figure of the second row.

³³ Hypothesis tests with significance on both sides of $0.025$ and $0.975$ are carried out to compare the differences. The differences are significant as long as both one-sided hypothesis tests can imply the significant differences.

Additional information

Funding

This research is funded by the excellent talents program, with award ID as 40447357360001, sponsored by Beijing Municipal Commission of Education.