Understanding the double-level Influence of Guanxi 2 on Construction Innovation in China ： the Mediating 3 Role of Interpersonal Knowledge Sharing and the 4 Cross-level Moderating Role of Inter-organizational 5 Relationships 6

: Guanxi, a Chinese term that defines social networks of power and benefits, can be divided 20 into inter-personal and inter-organizational relationships, and guanxi significantly influences 21 construction innovation in China. Many studies have examined the relationship between guanxi 22 and construction innovation at the project or organizational level. However, few of these studies 23 explained how guanxi could affect an individual’s innovative behaviour from a double-level 24 perspective. This paper builds on social capital theory and social exchange theory to examine 25 guanxi’s role in motivating innovative behaviour in a China-specific construction context. It 26 investigates the main effects of inter-personal relationships on innovative behaviour, the mediating 27 effects of knowledge sharing, and the cross-level moderating effects of inter-organizational 28 relationships. These elements were tested using a survey that received 178 responses from 35 29 different organizations. The results were analysed using Hierarchical Linear Modelling (HLM) and 30 revealed that inter-personal relationships have positive influences on innovative behaviour, thus 31 highlighting the partial mediating effects of knowledge sharing. In addition, the analyses showed 32 that inter-organizational relationships augment inter-personal relationships and knowledge sharing 33 on innovative behaviour by cross-level interaction. The research findings enhance an understanding 34 of guanxi and innovative behaviour in China-specific construction project settings, as well as 35 verifying the significance of guanxi in stimulating innovative behaviour. 36

setting up new production functions [24,25]. This theory of innovation has attracted much attention 103 from scholars and institutions, which has contributed to refining the definition of innovation. For 104 instance, Damanpour (1992)

116
Social capital refers to all resources embedded in social network relationships [33], which implies 117 that social actors engaging in such relationships can obtain access to resources to further their own 118 interests [34]. The social capital theory emphasizes the exchange of non-financial resources, 119 establishment of common resources [35], and that the exchange partners have a responsibility to 120 mutually contribute valuable resources that may be helpful [36]. Thus, by utilizing social capital, 121 actors (e.g., individuals, organizations, and commercial entities) can gain indispensable external 122 resources that promote innovation and enhance performance. Guanxi, a China-specific concept that 123 dominates business activities throughout the country [16], has been closely related to the western 124 culture concept of social capital; consequently, guanxi has attracted the attention of scholars in 125 management and business fields. Some of them have found that guanxi has produced significant 126 effects on technological innovation [37], and innovation performance [38].

127
Social exchange theory postulates that all social behaviours result from an exchange process [39] 128 , and an important assumption of the theory is that the behaviours are based on reciprocal exchanges 129 [40,41]. In essence, social exchange theory is one of the most influential conceptual paradigms 130 applied to understanding workplace behaviour [42], exchange rules and norms that shape social 131 behaviours, and resource exchanges [43]. Furthermore, social exchange tends to generate emotions 132 related to individual obligation, gratitude and trust [39], which may influence personal innovation 133 behaviour. Knowledge sharing, one of a specific pattern in social exchange, also has an impact on 134 innovation and has been investigated by several scholars [44][45][46]. Innovation practices in construction 135 projects tend to rely heavily on an individual's knowledge, skill and experience. Meanwhile, 136 knowledge sharing activities, as important ways to improve personal knowledge, can be 137 simultaneously seen as necessary for innovation in the construction process.

140
Guanxi is viewed as an intimate and common relationship amongst individuals or organizations via 141 high-quality social activities and reciprocal interest exchanges [47]. Inter-personal relationships are a 142 complex notion and comprises emotions and feelings toward others [2]. The family tie is a fundamental pattern in inter-personal relationships, and the scope of this tie can be extended to other social groups, such as kin, friends, and acquaintances [48]. Thus, people can develop inter-personal relationships within families, friends, classmates, colleagues and so on. Good interpersonal 146 relationships mean that there is at least a kind of guanxi within families, friends, classmates or 147 colleagues. Meanwhile, inter-personal relationships are also widely recognized as assets at a utilized, inter-personal relationships can cut cross organizational boundaries to get resources by 150 providing an alternative, informal and efficient network. For instance, Chen et al., (2015) affirmed 151 that Chinese entrepreneurs could gain information and resources via their guanxi networks, thereby 152 influencing a firm's success [50].

153
Furthermore, many scholars have stressed that inter-personal relationships are a key variable 154 for innovation. During an analysis of a firm's innovation, Arribas et al., (2013)

248
The greater the trust amongst government agencies, owners, designers, construction units, suppliers

301
The questionnaire, developed from the literature review and initial interviews, was separated 302 into two parts. The first part consisted of respondents' personal information (i.e. the gender,

372
One sample item was "We believe the information that this partner provides us"; the commitment 373 between inter-organizations was measured via four items validated by Gu et al [85]. Another sample 374 item was "We are committed to this partner". The responses with higher scores indicated that inter-375 organizational relationships were significant.

377
Analysis of the results obtained from the questionnaire survey was undertaken in three phases 378 as follows:

379
Firstly, in the preliminary analyses of these data, reliability test, exploratory factor analysis 380 (EFA) for each measurement item were conducted to estimate whether these variables' structures 381 were in accord with anticipated results. The software SPSS22.0 was utilized to carry out the reliability 382 test and EFA, thereby allowing for a discussion of the results for these measurement items.

383
Secondly, confirmatory factor analysis was conducted to validate the distinctiveness of these 384 variables, including inter-personal relationships, knowledge sharing, inter-organizational 385 relationships and innovative behaviour. The software Amos17.0 was utilized to examine the 386 measuring models of these latent variables, and the results suggested that the hypothesized model 387 with four latent variables excellently fitted our data when compared to other models.

388
Thirdly, given the multilevel characteristic of our data, it is necessary to discriminate the 389 variance at the individual and organizational levels in examination of the hypotheses. Thus, 390 hierarchical liner modelling (HLM) was applied to test our research hypotheses, including the main 391 effect, as well as the mediating and moderating effects, through using the software HLM version 6.08.
Subsequently, innovative behaviour and knowledge sharing were respectively regarded as 393 dependent variables. The main effect was examined with inter-personal relationships, and 394 knowledge sharing was seen as the mediator in the mediating effect testing, inter-organization 395 relationship as the moderator in moderating effect testing. All the results were estimated according 396 to the significance of the coefficients and R-square. the scales that satisfy the requirements of reliability, validity, and to test the hypotheses.

400
Testing for criterion-related validity convergent validity

443
innovative behaviour as an outcome in Table 4).

444
The third step was to examine the moderating effects of this study at the cross-level, and 445 following are key formulas for level 1 and level 2 models for innovative behaviour: (1) Table 4. Hypothesis 1 postulated that inter-personal relationships have a positive 461 influence on innovative behaviour in construction projects, which was the main effect of this study.

462
The results indicated that inter-personal relationships significantly related to innovative behaviour 463 (β2 = 0.412; see Model 1).

464
Hypothesis 2 proposed that a significant relationship existed between inter-personal

509
Firstly, we found that inter-personal relationships had significant positive influences on 510 innovative behaviour, which is consistent with social capital theory that stresses that guanxi as a type 511 of social capital can effectively stimulate innovative behaviour [51], and that these influences were 512 more significant in more innovative construction projects. In addition, this research provided 513 evidence that knowledge acted as partial mediator between inter-personal relationships and

519
Secondly, after testifying that inter-personal relationships and knowledge sharing were 520 respectively associated with innovative behaviour, we found there were some individual differences 521 in these antecedents of innovative behaviour, which originated from diverse organizations, as shown 522 by the R square of Model 4 being higher than that of Model 1 in Table 4

542
This study is also the first cross-level empirical test of inter-organizational relationships 543 moderating the direct and indirect influence of inter-personal relationships on individuals' 544 innovation behaviour. Prior studies on links between guanxi and innovation have focused on 545 performances at the firm's level [23], but this research tried to bridge the gap by utilizing multilevel 546 analyses to consider simultaneously individual-level and organizational-level variables.

547
Finally, this research differentiates itself from the prior studies because the social capital and 548 exchange theories were applied to examine the links between the inter-personal relationships and 549 individuals' innovative behaviour.     b. I feel that I have learned from each other by sharing information or ideas.
c. I am willing to share information or ideas with the other member of project-based organization.
d. In the project, i am willing to exchange and combine ideas to find solutions to problems.
Inter-organizational relationship Trust from senior managers, being able to represent the organizatsions： a. We believe the information that this partner provides us.
b. We trust this partner keeps our best interests in mind c. This partner keeps promises it makes to our firm d. This partner is trustworthy e. We find it necessary to be cautious with this partner Commitment from senior managers,being able to represent the organizatsions:f. We expect relationship to continue for a long time g. We are committed to this partner.
h. We expect relationship to strengthen over time.
i. Considerable effort and investment in innovation activity.