TMT heterogeneity refers to heterogeneity not only in the demographic diversity, such as races, age, tenure, education, but also in the cognition and values among TMT members [5] . Julian et al. [6] found that TMT heterogeneity affected corporate openness and supplied diverse information for team decision which contributed to improve decision quality and performance, when they researched joint ventures in Thailand. Yang, L. and Wang [7] pointed that TMT heterogeneity in age, gender and work experience lead enterprises to position the strategies that help to improve performance. However, Huang Y. and Yang N. D. found that TMT average age and age heterogeneity had negative relationship with corporate performance and equity centralization made negative moderating effect on the relationship, when they did empirical research on 153 companies listed on Shanghai exchange. Fan M. and Jiao L. B. [8] also found that TMT age heterogeneity and tenure heterogeneity were negatively related with corporate performance since they took the data in 2010 from Shanghai and Shenzhen GEM (Growth Enterprise Market) exchanges. Velinov et al. [9] even found no relationship between TMT age, tenure heterogeneity and corporate performance, as they studied on 37 IPO listed enterprises during 2008 to 2012. Since present researches show difference and even contradiction in the relationship between TMT heterogeneity and corporate performance, we propose that it is limited to say TMT heterogeneity has linear relationship with corporate performance, and the non-linear relationship and effects of leadership structure have been neglected.

The strategy in performance-declining enterprises can not be adapted to the business environment they confront, so TMT need to adjust strategy to reverse the non-performing situation. TMT heterogeneity reflects TMT members’ competence in communication and cooperation as well as capability of collection and filtering to market information and data, which affect strategic quality. If TMT heterogeneity is low, members have similar experience and values, so they have similar manners and strategic decision to keep less conflict and better communication but lack insights and diverse viewpoints in a whole [10] . And the team is inclined to keep the original strategy and hard to motivate innovation, which is easy to miss the opportunity to reverse the non-per- forming situation. If TMT heterogeneity reaches a temperate level, members share intelligence, expertise, capital and so on, which keep decision innovation and equality to improve corporate performance. If TMT heterogeneity goes beyond the limit, motivation to innovate disappear while emotional conflicts increase and communication becomes less, leading to low efficiency and equality in decision [11] . Since China is in a transitional period with social structure and institutional environment increasingly changing, lack of coordination and regulation will hinder communication and make decision failure, and lead to performance decline ultimately. Thus, this study proposes the first hypothesis:

Hypothesis 1: with the new CEO succession in performance-declining corporate, TMT heterogeneity (age/ tenure/education) has inverted U-shaped relationship with corporate performance.

As corporate governance mechanism, leadership structure means that whether chair of the board and CEO position are taken upon one person or different persons. It reflects the balance between independence and supervisorship of the board, and innovation freedom of TMT [12] , which affects TMT innovation behavior and strategic decision that influences corporate performance ultimately. Heretofore there rarely has been common conclusion about the effect of leadership structure. Principal-agent theory proposes that the chair-CEO duality will weaken supervision and control of the board and increase governance risk. Meanwhile, such a combination promotes the interaction and information communication between TMT and the board, and prevents conflict [13] . Management autonomy theory and decision theory hold that corporate performance has much closer relationship with TMT characteristics under chair-CEO duality. On the other hand, the separation of chair and CEO will constrain TMT’s innovation and autonomy, and weaken TMT’s effects on corporate performance, even though the separation guarantees supervision and control of the board and avoid TMT opportunistic behavior.

As to the performance-declining enterprises with chair-CEO duality, TMT heterogeneity’s effects on corporate performance could be more significant. If TMT heterogeneity is low, chair-CEO duality will make the board out of control of CEO who might take opportunities to pay more attention to individual benefit but less to long-term corporate development, which lead to lower performance. If TMT heterogeneity reach to a temperate level, there will be a balance between TMT innovation freedom and board independence, as chair-CEO duality gives CEO more power to accelerate decision speed that adjusts performance-declining enterprises to environment. Meanwhile, chair-CEO duality improves coordination between the board and management to obtain external resource and make full use of social capital like social network and political status in the board, so that strategic decision can be implemented smoothly to improve performance. If TMT heterogeneity is high, chair- CEO duality will intensify TMT conflicts as lack of supervision mechanism, which make performance worse.

As to the performance-declining enterprises with separation of chair and CEO, TMT heterogeneity has less significant effects on corporate performance. Separation of chair and CEO takes more effective supervision and control on CEO and TMT, and enhances independence of the board to protect corporate from risk, which but also misses chance to develop easily. Therefore, no matter TMT heterogeneity is high or low, performance-de- clining enterprises with separation of chair and CEO are inclined to take conservative strategies to keep stability. Thus, this study proposes the second hypothesis:

Hypothesis 2: with the new CEO succession in performance-declining corporate, leadership structure moderates the relationship between TMT heterogeneity (age/tenure/education) and corporate performance. Namely, the higher leadership structure is, the more significantly TMT heterogeneity affects corporate performance in the inverted U shape.

This study takes China performance-declining enterprises as research subjects, and measures the subjects according to ‘absolutely declining performance in succeeding 3 years at least’ proposed by Ran M.. If ROAci (Corporate Return on Assets) in three years before CEO succession minus ROAIi (Industry Average Return on Assets) is negative value and the absolute value consistently increase in this three years, we define the corporate as performance-declining corporate. As to TMT definition, there is no common standard. CHO proposed that TMT includes vice-president and the above titles [14] . Sun H. F. et al. hold that TMT includes chair of the board, general manager, function directors and managers who participate decisions on the top. According to China listed corporation regulation, we select vice-president, deputy general manager, chief accountant, chief engineer and other top managers as TMT members.

The sample enterprises selected in this study are fit in following standards: 1) the corporates listed on Shanghai and Shenzhen exchanges with three years succeeding performance decline, and experiencing CEO succession during 2009 to 2011 but staying CEO unchanged in the next three years; 2) the corporates that succeed CEO because of promotion, death, crime, term expiration or demission should be eliminated; 3) the corporates where chair-CEO state has changed during 2009-2011 should be eliminated; 4) the corporates in finance industry, ST, PT and without complete data should be eliminated. As the statistics tested in this study go through the previous three years as well as the next three years, the actual samples used spread 2006 to 2014. Finally, we assure 130 sample enterprises.

Statistics used in this study are from CSMAR database, Gildata and Corporate Financial Statement, with part of information from Sina Finance net and SPSS20.0 software for data processing.

Variables classification and coding rules are shown in Table 1.

The results of variables descriptive statistics and correlation analysis are shown in Table 2. We can see that TMT age, tenure and education heterogeneity in the sample enterprises are high, as the values are 0.882, 0.629, 0.609. Besides, some variables have close relationships that have passed significance test. However, as influenced by other factors like sample capacity and control variables, all of the variables need to be a further analysis.

Considering interacting influence, we processed data in hierarchical regression models by joining control variables, independence variables, moderators and interaction items of independence variables and moderators step by step. In order to avoid multicollinearity problem, we made centralization process on independence variables

Note. N = 130; ^***p < 0.001, ^**p < 0.01, ^*p < 0.05, ⁺p < 0.1 two-tailed test.

and moderators, and computed interaction items that were then taken into regression equation. From Table 3 we can see deviation expansion coefficient of each variable in four models is below 5, which means there is no multicollinearity problem.

Model 1 is the regression model of control variables and dependence variables; model 2 is the regression model of control variables, independence variables and dependence variables; model 3 takes centralized quadratic terms of TMT heterogeneity into the regression equation, of which the results show a positive relationship between quadratic terms of TMT heterogeneity and corporate performance with the regression coefficient 0.142 (p < 0.1), reflecting a U-shaped relationship between TMT heterogeneity and corporate performance (Figure 1). Quadratic term coefficients of tenure heterogeneity and education heterogeneity are −0.252 (p < 0.05), −0.190 (p < 0.05), reflecting an inverted U-shaped relationship between tenure, education heterogeneity and corporate performance (Figure 2 & Figure 3). Thus, hypothesis 1 is partly supported.

In model 4 where leadership structure, interaction items of leadership structure and TMT heterogeneity, and interaction items of leadership structure and TMT heterogeneity quadratic terms were joined, F value is 6.678 (p < 0.001), and adjusted value is 0.563 that has increased, reflecting leadership structure’s significant moderating effect on relationship between TMT heterogeneity and corporate performance. The regression coefficients of interaction items of age heterogeneity quadratic term, tenure heterogeneity quadratic term and leadership structure are 0.705 (p < 0.001), −0.301 (p < 0.1), which means that the higher leadership structure is, the more significant TMT age heterogeneity’s U-shaped effect is (Figure 4), and the more significant TMT tenure heterogeneity’s inverted U-shaped effect is (Figure 5). Thus, hypothesis 2 is partly supported.

We find that TMT tenure, education heterogeneity have inverted U-shaped relationship with corporate performance. Namely, corporate performance reaches to the best when TMT tenure and education heterogeneity are moderate. If TMT tenure and education heterogeneity is below the moderate level, organization will lack innovation and miss a lot of opportunities. But once TMT tenure and education heterogeneity go beyond the level, team cohesion and satisfaction decrease with conflicts increase. It implicates that enterprises should set appropriate recruitment and selection standards as well as tenure length, and consciously allocate and adjust TMT members in principle of moderation, to make the most optimum combination for performance improvement.

Findings show a U-shaped relationship between TMT age heterogeneity and corporate performance which is coordinated with research by Richard and Shelor [15] but different with our hypothesis. The reason might be age heterogeneity’s special influence compared with tenure and education heterogeneity under China present institutional background and managerial situation. As coordination and cooperation are needed between functions in

Note: N = 130, ^***p < 0.001, ^**p < 0.01, ^*p < 0.05, ⁺p < 0.10, two-tailed test.

performance-declining enterprises, TMT with low age heterogeneity keeps unanimous cognitive ability to reach agreement that helps to improve performance. When age heterogeneity reaches to a moderate level, corporate performance goes down, which might be caused by: on one hand, as a dominant variable to classify oneself [16] , moderate age heterogeneity makes team polarized and influences corporate performance; on the other hand, heterogeneity advantages are restricted by institutional environment and collectivism. However, higher age heterogeneity contributes to obtain resources and support from social network, especially the older members do well in dealing with government and society. Meanwhile, the younger members are more innovative and adaptive to improve decision quality and corporate performance. It implicates that it’s better to build a team with low age heterogeneity to keep common cognition and accelerate decision speed, or a team with high age heterogeneity to motivate innovation to improve performance.

From Figure 4 & Figure 5 it can be seen that leadership structure makes significant moderating effect on the relationship between TMT age, tenure heterogeneity and corporate performance. Namely, the higher leadership structure is, the more significant TMT age heterogeneity’s U-shaped effect is (Figure 4), and the more significant TMT tenure heterogeneity’s inverted U-shaped effect is (Figure 5). Besides, relationship between leadership structure and corporate performance is determined by L-H fit (L-leadership structure; H-TMT heterogeneity). It implicates that management can adjust L-H fit to improve performance. In details, CEO or chair can be changed to influence interaction between CEO and the board; or enterprises keep leadership structure unchanged to avoid organizational fluctuation while adjusting TMT members to fit leadership structure and improve performance. As China corporate external governance mechanism is relatively void, enterprises should not only depend on governance effect of leadership structure, but also pay more attention to combination effect of internal governance mechanism, like management holding share, rational design of equity structure, and market-oriented personnel mechanism.