Ronald Burt is grateful to the University of Chicago Booth School of Business for financial support during the work reported here. Sonja Opper is grateful to the Marianne and Marcus Wallenberg Foundation for support of this work. We are grateful to the Jan Wallanders and Tom Hedelius Foundation for the grant to Sonja Opper that funded the fieldwork in China providing the data analyzed here, and to Na Zou and Ke Zeng for assistance coding event descriptions into kinds. This manuscript, supplement materials, and the network instrument in English are available online (and, for variation on the theme, the questionnaire with event name generators used in the 1998 GSB Alumnae Survey, GSBAS1.pdf): http://faculty.chicagobooth.edu/ronald.burt/research.
 Family are not often cited by the entrepreneurs (8.31% of cited contacts are nuclear or extended family), but when cited, family members are often cited multiple times, and for significant events. A regression model with respondent fixed effects predicting number of events in which a contact is cited from a dummy variable distinguishing family contacts returns a 1.15 coefficient and 20.53 test statistic. A logit model with respondent fixed effects predicting from the family variable which contact was valued in founding the business returns a 2.39 coefficient and 20.59 test statistic.
 The Chinese word ‘信任’ in the trust question in Table 1 is a term as ambiguous in Chinese as ‘trust’ is in English. Interviewers were trained to guide respondent queries about what we mean by trust in the following way: ‘Consider the extent to which you trust each of the listed people. For example, suppose one of the people asked for your help. The help is not extreme, but it is substantial. It is a level of help you cannot offer to many people. To what extent would you trust each person to give you all the information you need to decide on the help? For example, if the person were asking for a loan, would they fully inform you about the risks of them being able to repay the loan? If the person was asking you give a job to one of their relatives, would they fully inform you about their relative's poor work attitude or weak abilities, or other qualities that would make you prefer not to hire the relative’? In this situation, it is valuable to know for our network analysis that trust scores vary primarily within rather than between networks. Trust variance across relationships is 60% network differences within respondents, 10% individual differences between respondents, and 30% random error (Burt, Bian, & Opper, 2017).
 Survey network items, like most survey questions, are affected by respondent mood when interviewed. Bailey and Marsden (1999) show that preceding the General Social Survey network questions with questions about the respondent's family predispose people to think about family issues when naming network contacts. Smith, Menon, and Thompson (2012) show that people exposed to material about losing one's job before a network interview, are more likely to report a network of densely interconnected contacts.
 This observation is based on a cluster analysis of time profiles. We created a time profile for each respondent defined by the years in which events occurred. For example, the time profile for the Figure 1 entrepreneur is 2, 3, 7, 10, 13, corresponding to the years in which the entrepreneur's five significant events occur. The squared Euclidean distance between two profiles is small to the extent that events in each profile occur in the same years after founding. Cluster analysis of profiles for the 675 entrepreneurs who reported five events, using the Ward minimum-variance method in Stata (see supplement Figure S1) reveals three distinct clusters: a cluster of profiles that occur within the first decade of business, a cluster of profiles that occur within the first 15 years of business, and a cluster of profiles that occur within the first 22 years of business. Within each cluster, events are about evenly distributed over time (mean year in which each event occurs for each cluster is in the inset box to the right of the cluster dendogram in Figure S1). The first cluster is young businesses (8.84 years old on average), the second is older businesses (13.95 years old on average), and the third is still older businesses (22.10 years old on average).
 Business age is held constant by measuring events as a proportion of business age. For example, an event .5 in proportional time occurred half way between founding and the 2012 survey. Cluster analysis of proportional time profiles (same method as in the previous footnote and reported in supplement Figure S2) also reveals three clusters. Profiles for each cluster are reported in the inset box to the left of the cluster dendogram in Figure S2. Events are distributed about evenly over time, differing in the first event: The first cluster spreads over the whole life of a business. The second cluster begins with the first event late (about a third of the way into the business’ life). The third cluster begins at about the same time as the first cluster, but with a larger gap between the first and second events. We tested for trust and success association with time to first event. A control for time to first event adds nothing to the prediction of trust in Table S1 (−1.68 t-test for years to first event, 0.37 t-test for proportional time to first event), nothing to the prediction of success in Table 6 (respective t-tests of 1.00 and 0.41), and nothing to the prediction of success in Table 7 from networks limited to current, founding, and Event 1 contacts (respective t-tests of 1.28 and 0.63). Therefore, we focus in the text on event order, rather than physical or proportional time.
 We measure the tendency for two characteristics to appear in the same relations with a Jaccard coefficient, which is the number of relations in which the two characteristics occur together, divided by the total number of relations in which either occurs. The two-dimension solution in Figure 4 fits the data well. The first dimension is defined by the eigenvector associated with an eigenvalue of 5.63. The second dimension corresponds to an eigenvalue of 3.23, and the third to a 1.64 eigenvalue. The first two dimensions together describe 79% of the association variance, and are drawn in Figure 4 in proportion to their eigenvalues.
 We focus on ‘no role’ contacts being none of the seven familiar sources of contacts listed in Table 5 because that we know for certain. With less clarity, we know that many of them are or were co-workers. Table 1 lists ‘colleague’ as one of the roles a contact could play, which the respondent's worksheet defined as ‘you and the person have been employed in the same organization’. Of the 3,645 ‘no role’ contacts in Table 5 and Figure 5, most are ‘colleagues’ (79%). However, we failed in the questionnaire to distinguish between colleagues in the current organization versus former employers. The ambiguity should be removed in future data collection. We put a warning about this point on the downloadable network questionnaire in the acknowledgement note.
 We went one step further to see whether the Figure 6 association between trust and closure is different in family firms versus other firms. We use the common definition of family firms: owner-operated firms in which the respondent's spouse or children are employees. By this criterion, 254 of the 700 businesses are family firms. Respondents for family firms are almost twice as likely to turn to family in founding the business: 44% of family firms cite family as founding contacts, versus 24% of other firms (versus 31% for the whole sample, see the Figure 3 left-most graph). Regardless, Figure 6 coefficient estimates for family versus other firms are given below, holding constant contact frequency, years known, respondent fixed effects, and whether a contact was family to the respondent (t-tests in parentheses, N is number of relations across which estimates are computed). The three coefficients measuring association between trust, closure, and founding as a guanxi event are similar for both kinds of firms.
 Distinctions between kinds of events require subjective judgments, so the irrelevance of such distinctions in Figure 8 made us concerned about the reliability of the distinctions in Table 3. The coding was reviewed by the author fluent in Chinese, and seemed sensible, but as a further check we had a second research associate working in Beijing code all 4,163 events into the Table 3 categories to compare with the coding we had. Reliability is high on average. The two coders agreed whether an event was a gain or a loss on 98% of the events, and agreed in their assignment of 74% of events to the Table 3 categories. The coders disagreed most clearly on customer events versus collaboration events (categories 3 and 7 in Table 3). Entrepreneurs often collaborated with others to produce a new product or secure a customer contract. Most of the disagreements between the coders were one coding an event as a customer issue while the other coded the event as a collaboration issue. If customer and collaboration issues are combined, the two coders agreed in their assignment of 84% of events. Given no statistical difference between customer and collaboration issues in Figure 8, we are confident in our conclusion in the text: all substantive kinds of significant events have the potential to generate guanxi ties.
 Our conclusion is robust to years known. In networks around Western managers, time distinguishes relations that can be discussed as guanxi-like ties in that trust is independent of structural embedding and high for colleagues with whom respondents have worked for multiple years. The time required to establish a guanxi-like tie in an organization can be determined by replicating trust correlations with network closure for contacts within intervals of time known. For example, among bankers and analysts, the trust-closure association is strong for colleagues known for a year or two, then the correlation drops to zero, and average trust increases, for colleagues known more than two years (Burt & Burzynski, 2017: Figure 4). In other words, two years is the time required to establish a guanxi-like tie for the bankers and analysts. We checked for change in the trust-closure association across the years for which a Chinese entrepreneur had known a contact. The dashed-line strong, positive trust-closure associations for nonevent contacts in Figures 7 and 8, and the solid-line negligible trust-closure associations for event contacts, are consistent across the years for which an entrepreneur has known the contact (see Figure S3 in the supplement materials). Years known adds nothing to our guanxi distinction between event and nonevent contacts.
 We also looked at the network association with success as a Western investor would want to experience it – profits. We measure profits by return on assets (net income divided by book value of assets, both for the last full year, 2011). When we predict return on assets from the variables in Table 6, plus a control for log assets, profits are significantly lower for entrepreneurs in relatively closed networks (−2.88 t-test for log network constraint), and average returns over the last three years are similarly lower in relatively closed networks (−2.63 t-test for log network constraint).
 A business is founded when formally registered as a private enterprise. However, many of the sample businesses had been in operation before they were registered. Some operated under a different legal form. Others started operations, and even signed their first contract, without formal registration. In its first year as a registered private enterprise, the median business had 20 full-time employees and sales of 1,500,000 yuan (about 180 thousand U.S. dollars at the turn of the century). Without the control for founding success, Burt and Burzynska (2017, Table 1) report a −0.440 regression coefficient for log network constraint with a .131 standard error. Table 6 shows that holding constant success-at-founding weakens the coefficient slightly, but shrinks the standard error more, resulting in a stronger test statistic for the network association with success (−3.64 here versus −3.36 in Burt & Burzynska).
 We also looked into an extension that turned out to be negligible. The Chinese national constitution was amended in 2004, increasing the status of private enterprise and institutional protection of private property (http://www.npc.gov.cn/englishnpc/Constitution/node_2825.htm). Suspecting the network association with success might be stronger for businesses founded after the amendment, we added level and slope adjustments to Table 6 for businesses registered after the amendment. Both adjustments are negligible. Success is negligibly lower for businesses founded after the amendment (−0.37 test statistic) and negligibly less associated with having a large, open network (0.26 test statistic).
 The order of events matters. When we predict success from networks composed of current contacts plus contacts cited for the most recent events – events four and five –the added contacts do not improve prediction. Entries for a new row in Table 7 would be a −0.130 coefficient, .095 standard error, and a negligible −1.36 t-test. In short, predicting success depends on including contacts helpful in early events.
 The first event is exceptional in terms of recovering the success association with success (Table 7) and the strength of relationship with contacts cited for the first event (Figure 3), so we tested for success variation across kinds of first events. Distinctions between kinds of events that are irrelevant in general (Table 8) could be consequential in the first event. They are not. Adding to Table 6 dummy variables distinguishing the eight kinds of post-founding events does not improve the success prediction (F(7,683) = 0.86, P ~ 0.54).
 Differences between the three categories are statistically significant. Predicting business success from a 1, 0, −1 contrast between the three rows in the Figure 10 table yields a 2.25 test statistic (P ~ 0.03). Logit regression yields a −4.62 test statistic (P < 0.001) predicting, from the same contrast, which entrepreneurs cite a family member as most valued contact in founding the business. The result in Figure 10 is not about family firms. It is about who entrepreneurs turn to at founding. In fact, the heads of family firms are more likely to turn to family at founding. Using the definition in footnote 8 of a family firm, about half of the entrepreneurs running a family firm turn to family at founding, versus a quarter of those running non-family firms (44% versus 24% respectively, 27.86 chi-square, P < 0.001). But family firms are more likely the solid dots in Figure 10 rather than the lower hollow dots who turned to family at founding (about half of the solid dots in Figure 10 are family firms versus a third of the hollow dots; 47% versus 32% respectively, 13.13 chi-square, P < 0.001). More, the negligible success association in the first row of Table 7 for current contacts is quite strong for family firms (−0.560 coefficient, 0.191 standard error, −2.93 test statistic, P ~ 0.003, for the slope adjustment for family firms when a family-firm dummy is added to the equation). We do not discuss this in the text because the slope adjustment for family firms is negligible when the founding and first event contacts are included in an entrepreneur's network (third row of Table 7, −1.80 test statistic for family-firm slope adjustment), and accordingly quite negligible when all event contacts are included (bottom row of the table, −0.57 test statistic). In short, current contacts in the networks around family-firm entrepreneurs better capture the diversity of the entrepreneur's contacts because family is a source of both current and event contacts, but the family-firm difference is negligible when founding and first-event contacts are included in the networks, and disappears when all event contacts are included.
Bailey, S., & Marsden, P. V.
1999. Interpretation and interview context: Examining the General Social Survey name generator using cognitive methods. Social Networks, 21
Batjargal, B., Hitt, M. A., Tsui, A. S., Arregle, J.-L., Webb, J. W., & Miller, T. L.
2013. Institutional polycentrism, entrepreneur's social networks, and new venture growth. Academy of Management Journal, 56
1997. Bringing strong ties back in: Indirect ties, network bridges, and job searches in China. American Sociological Review, 62
Bian, Y. Forthcoming. The prevalence and increasing significance of guanxi. China Quarterly, In Press.
Bian, Y., & Li, L.
2012. The Chinese General Social Survey (2003–8). Chinese Sociological Review, 45: 70–97.
Borgatti, S. P.
2002. NetDraw. Boston, MA: Analytic Technologies.
Burt, R. S.
1992. Structural holes. Cambridge, MA: Harvard University Press.
Burt, R. S.
2005. Brokerage and closure. Oxford: Oxford University Press.
Burt, R. S.
2010. Neighbor networks. Oxford: Oxford University Press.
Burt, R. S., & Burzynska, K.
2017. Chinese entrepreneurs, social networks, and guanxi
. Management and Organization Review, 13
Burt, R. S., Bian, Y., & Opper, S.
2017. More or less guanxi: Trust is 60% network context, 10% individual difference. Working paper, University of Chicago.
Burt, R. S., Kilduff, M., & Tasselli, S.
2013. Social network analysis: Foundations and frontiers on network advantage. Annual Review of Psychology, 64: 537–547.
Chen, C. C., Chen, X.-P., & Huang, S.
2013. Chinese guanxi: An integrative review and new directions for future research. Management and Organization Review, 9
Guo, C., & Miller, J. K.
Guanxi dynamics and entrepreneurial firm creation and development in China. Management and Organization Review, 6
1991. Trusting persons, trusting institutions. In Zeckhauser, R. J. (Ed.), Strategy and choice: 185–209. Cambridge, MA: MIT Press.
1990. The peasant family and rural development in the Yangzi Delta, 1350–1988. Stanford, CA: Stanford University Press.
IMD. 2005. IMD World Competitiveness Yearbook 2005. Lausanne: IMD.
1994. The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology, 100
Li, H., Meng, L., Wang, Q., & Zhou, L.-A.
2008. Political connections, financing and firm performance: Evidence from Chinese private firms. Journal of Development Economics, 87
Luo, J.-D., Cheng, M.-Y., & Zhang, T.
Guanxi circle and organizational citizenship behavior: Context of a Chinese workplace. Asia Pacific Journal of Management, 33
Luo, Y., Huang, Y., & Wang, S. L.
Guanxi and organizational performance: A meta-analysis. Management and Organization Review, 8
Marsden, P. V.
2011. Survey methods for network data. In Scott, J. P. & Carrington, P. J. (Eds.), The Sage handbook of social network analysis: 370–388. Thousand Oaks, CA: Sage.
Nee, V., & Opper, S.
2012. Capitalism from below: Markets and institutional change in China. Cambridge, MA: Harvard University Press.
Nee, V., & Opper, S.
2010. Political capital in a market economy. Social Forces, 88
Peng, M. W., & Luo, Y.
2000. Managerial ties and firm performance in a transition economy: The nature of a micro-macro link. Academy of Management Journal, 43
1998. The content of the General Social Survey discussion networks: An exploration of General Social Survey discussion name generator in a Chinese context. Social Networks, 20
Ruef, M., & Grigoryeva, A.
2017. Jim Crow, ethnic economies, and status attainment: Occupational mobility among U.S. Blacks, 1880–1940. Working Paper. Department of Sociology, Duke University.
Smith, E. B., Menon, T., & Thompson, L.
2012. Status differences in the cognitive activation of social networks. Organization Science, 23
Stinchombe, A. L.
1965. Social structure and organizations. In March, J. G. (Ed.), Handbook of organizations: 142–194. Chicago, IL: Rand McNally.
1995. Managing legitimacy: Strategic and institutional approaches. Academy of Management Review, 20
Xiao, Z., & Tsui, A. S.
2007. When brokers may not work: The cultural contingency of social capital in Chinese high-tech firms. Administrative Science Quarterly, 52
Xin, K., & Pearce, J. L.
Guanxi: Connections as substitutes for formal institutional support. Academy of Management Journal, 39