Descriptive Statistics
Table 3 presents descriptive statistics of the data employed in the study stratified by company status. We portray two subsamples based on the trading status – active and delisted firms.^{11} The descriptive statistics include the number of observations, means, medians, minimum, maximum, standard deviations, skewness, and kurtosis for each subsample. It should be noted that because of the binary or dummy variables that have been used for some factors, the mean for these variables should be interpreted as the percentage of companies that satisfy a given criterion. The binary variables employed in this study include CM_NEXC, CM_DUAL, BACK, BIG5, IPO_9900, and VCBACKED.
Table 3. Descriptive Statistics of the Data  BD_SIZE  BD_INDP  CM_NEXC  CM_DUAL  TOP20  OF_PRICE  OF_SIZE  OF_AGE  RETAIN  BACK  BIG5  NUM_RISK  ROA  CUR  TAT  DET  C_SIZE  IPO_9900  VC_BACKED 


Survivor IPOs (N = 93) 
Mean  5.19  53.41  .64  .86  65.98  .89  32.95  5.80  62.16  .74  .53  12.72  −.29  7.17  .87  .43  7.27  .40  .11 
Median  5.00  60.00  1.00  1.00  70.00  .50  8.00  3.05  70.00  1.00  1.00  12.00  −.06  2.00  .61  .31  7.23  .00  .00 
Min  3.00  .00  .00  .00  14.40  .20  1.50  .00  .00  .00  .00  .00  −6.10  .02  .00  .00  5.62  .00  .00 
Max  10.00  83.00  1.00  1.00  94.14  4.60  421.09  38.46  96.34  1.00  1.00  31.00  .58  331.52  4.82  4.20  9.42  1.00  1.00 
Std dev.  1.32  19.59  .48  .35  18.67  .85  74.00  7.16  23.67  .44  .50  5.32  .75  20.68  .98  .53  .77  .49  .31 
Skewness  .61  −.68  −.60  −2.02  −.86  2.45  3.79  1.96  −1.14  −1.10  −.13  .80  −3.96  9.19  1.83  4.42  .50  .43  2.50 
Kurtosis  .95  −.10  −1.64  2.10  .04  7.21  14.43  4.74  .65  −.80  −1.99  2.02  21.56  116.27  3.62  25.57  .39  −1.82  4.28 
Nonsurvivor IPOs (N = 32) 
Mean  5.13  61.96  .70  .85  76.77  .93  135.10  6.24  70.48  .90  .70  14.25  −.35  7.05  .95  .50  7.41  .36  .31 
Median  5.00  67.00  1.00  1.00  78.41  1.00  12.00  4.51  74.34  1.00  1.00  13.00  −.01  1.81  .62  .34  7.35  .00  .00 
Min  3.00  .00  .00  .00  19.99  .20  1.00  .01  .00  .00  .00  7.00  −6.10  .02  .00  .00  5.61  .00  .00 
Max  9.00  89.00  1.00  1.00  98.28  2.00  6652.73  18.83  99.52  1.00  1.00  25.00  .58  567.03  4.82  4.20  9.42  1.00  1.00 
Std dev.  1.13  20.08  .46  .36  14.52  .50  873.75  5.50  20.06  .30  .46  3.91  1.17  43.39  .99  .60  .73  .48  .46 
Skewness  .85  −.89  −.86  −1.96  −.65  .29  7.41  .59  −1.02  −2.70  −.89  .91  −4.26  12.79  1.68  3.55  .16  .60  .83 
Kurtosis  1.75  .25  −1.28  1.84  .36  −.66  53.55  −.95  1.27  5.36  −1.22  .82  18.39  165.95  3.42  16.43  .33  −1.66  −1.33 
MannWhitney Utest  .09  2.59  .11  .22  7.21**  3.69†  .63  .26  .94  2.83†  2.25  1.99  1.09  .21  .58  1.46  3.33†  .12  5.53** 
pvalue  .77  .11  .74  .64  .01  .05  .43  .61  .33  .09  .13  .16  .30  .65  .45  .23  .07  .73  .02 
In order to prevent the influence of observations with extreme values, observations are truncated at the specified thresholds. All observations with covariate values higher than the 99th percentile of each covariate are set to that value. In the same way, all covariate values lower than the first percentile of each covariate are truncated. This procedure is similar to the one employed by Shumway (2001). The MannWhitney Utest, a nonparametric test, is employed to test for significant differences between the group means. Variables with significant differences in their group means will be expected to add information to a regression analysis. The variables TOP20 (U = 7.21, p < .05) and VCBACKED (U = 5.53, p < .05) display significant differences across the subsamples.
According to Table 3, the median number of directors for both survivors and nonsurvivors is five, which is consistent with Lamberto and Rath (2008). They find that the majority of IPO companies have fewer than six directors on the board which is the minimum number of directors recommended by the ASX for good governance. The mean percentages of nonexecutive directors on the board were 53.41 and 61.96 for active and nonsurvivor companies, respectively. This figure implies that the majority of directors on the new economy Australian IPO company boards are independent directors. In addition, 64 and 70 percent of active and nonsurvivor new economy IPO companies, respectively, have a nonexecutive chairperson, and 86 and 85 percent of these companies have the positions of CEO and chairperson held by different persons. These results suggest that the majority of new economy Australian IPO companies have boards which can be considered independent. Furthermore, the mean percentages of the top 20 shareholders for active and nonsurvivor companies are 65.98 and 76.77 percent, respectively.
In terms of the offering characteristics, the median offering price is A$0.50 for the survivors and A$1.00 for the nonsurvivors. The median offer sizes are A$8 and A$12 million and the medians of offering age are 3.04 and 4.51 years for the survivor and nonsurvivor companies, respectively. These results suggest that the new economy Australian IPO companies are relatively young and small, consistent with the results reported by Lamberto and Rath (2008).
Additionally, 74 and 90 percent of the offerings by survivor and nonsurvivor companies are underwritten, while 53 and 70 percent of the offerings by active and nonsurvivor companies have an auditor from one of the Big 5 accounting firms. The median number of risk factors identified in the prospectus is 13 and 14 for active and nonsurvivor companies, respectively. The means of retained ownership by preIPO owners are 62.16 and 70.48 percent for active and nonsurvivor IPO companies, respectively, which implies that the control of new economy IPO companies was retained by the original owners. It is also interesting to note that 40 and 36 percent of active and nonsurvivor IPOs companies are listed during the 1999 to April 2000 period.
The profitability ratios, which show the ability of the company to generate profit, are negative for both groups. The means of ROA for active and nonsurvivor companies are −.29 and −.35, respectively. This result suggests that nonsurvivor IPOs companies have lower earnings than active companies. But the difference is not statistically significant. The mean of the liquidity ratio, CUR, of nonsurvivor companies is higher than that of the active firm subsample. The mean of debt ratio, DET, indicates that the nonsurvivor companies have higher leverage than that of active companies. For the activity ratio, TAT, the mean of nonsurvivor companies is higher than that of the survivors. However, the MannWhitney Utest indicates that there is no significant difference in means of these ratios between active and nonsurvivor new economy IPO companies.
The mean SIZE of active and nonsurvivor companies is 7.27 and 7.41, respectively. The MannWhitney Utest shows that, on average, the size of active and nonsurvivor new economy IPO companies in our sample are marginally statistically significantly different (U = 3.33, p < .10). Finally, the survivor and nonsurvivor samples differ significantly with respect to the percentage of firms backed by venture capitalists. Only 11 percent of survivors are backed by venture capitalists while 31 percent of the nonsurvivors have VCbacking.
The Pearson correlation coefficients across the variables are shown in Table 4. The results suggest weak relationships across the variables. We do not find any large and significant coefficients that indicate serious problems of multicollinearity.
Table 4. Pearson Correlation CoefficientsVariable  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20 


1. SUR_TIME  1.00  .03  −.05  −.05  −.01  −.27  .03  −.10  .15  −.18  −.02  .03  −.31  .04  −.10  .01  .03  .16  −.07  −.15 
2. BD_SIZE   1.00  .10  .04  .16  .03  .44  .26  −.10  −.06  −.08  .18  .05  .06  −.02  −.03  .00  .52  −.15  .21 
3. BD_INDP    1.00  .37  .24  .12  −.01  .09  .00  .01  .12  .16  .14  −.09  .02  −.03  .07  −.09  .05  .22 
4. CM_NEXC     1.00  .34  −.14  .01  .01  .01  −.00  .03  −.12  .08  −.09  .01  −.02  −.05  −.10  .06  .04 
5. CM_DUAL      1.00  −.03  .07  .04  −.15  −.04  .12  .02  .12  −.03  −.10  .07  .06  .07  .10  .03 
6. TOP20       1.00  .11  .04  .16  .37  .14  −.04  .13  .04  −.06  .10  .09  .07  −.20  .11 
7. OF_PRICE        1.00  .18  −.04  −.03  −.17  .09  .03  .15  −.09  .06  .08  .54  −.02  .05 
8. OF_SIZE         1.00  −.01  −.20  −.15  .06  .01  .04  −.02  .05  .09  .24  −.04  .13 
9. OF_AGE          1.00  .09  .15  −.02  −.16  .13  −.11  .11  .02  .04  .05  −.02 
10. RETAIN           1.00  .16  .00  .23  −.08  −.11  .01  .06  −.10  −.10  .03 
11. BACK            1.00  .02  −.12  .02  −.05  .17  .08  −.04  .04  .02 
12. BIG5             1.00  .08  .01  −.02  −.12  .05  .13  .01  .18 
13. NUM_RISK              1.00  −.05  −.02  .001  .04  −.00  .08  .15 
14. ROA               1.00  .04  −.02  −.48  .47  −.00  −.01 
15. CUR                1.00  −.15  −.16  −.08  −.05  −.04 
16. TAT                 1.00  .40  .02  .0  −.06 
17. DET                  1.00  −.15  .06  −.02 
18. C_SIZE                   1.00  −.08  .10 
19. IPO_9900                    1.00  −.08 
20. VC_BACKED                     1.00 
Cox Proportional Hazards Model Estimation Results
We employ the Cox proportional hazards model to investigate the influence of corporate governance variables on the survival likelihood of new economy IPO companies. In addition to corporate governance variables, we also include offering characteristics, financial ratios, and companyspecific variables. The estimation results are presented in Table 5.^{12}
Table 5. Estimation Results of Multivariate Cox Proportional Hazards Model of the Entire SampleCovariate  Coefficient  Standard error  χ^{2} Statistic  pvalue  Hazard ratio 


BD_SIZE  2.58†  1.53  2.86  .09  13.19 
BD_SIZESQ  −.25†  .14  3.27  .07  .78 
BD_INDP  −.11*  .04  6.05  .01  .90 
BD_INDPSQ  .00*  .00  5.90  .02  1.00 
CM_DUAL  .47  .87  .29  .59  1.60 
CM_NEXC  .07  .50  .02  .88  1.08 
TOP20  −.06  .08  .66  .42  .94 
TOP20SQ  .00  .00  1.64  .20  1.00 
OF_SIZE  .00†  .00  3.80  .05  1.00 
BACK  1.21†  .68  3.17  .07  3.36 
VC_BACKED  .91†  .50  3.36  .07  2.49 
DET  .67†  .35  3.76  .05  1.96 
C_SIZE  .68†  .36  3.61  .06  1.97 
We present the coefficients, estimated standard error of this estimate, Wald chisquare tests along with the relative Pvalue for testing the null hypothesis that the coefficient of each covariate is equal to zero. Finally, the hazard ratio is presented in the last column. The hazard ratio is obtained by computing e^{β}, where β is the coefficient in the proportional hazards model. A hazard ratio equal to one indicates that the covariate has no effect on survival. If the hazard ratio is greater (less) than one, then this indicates a more rapid (slower) hazard timing. We report only coefficients and test statistics of significant control variables in order to conserve space. Our estimations include all available variables. We categorize survival on the basis of whether the firm continues to trade in the Australian exchange as of December 31, 2007. All delisted firms – regardless of the reason for delisting – are treated as failed firms.
Since we do not have a theory regarding a functional form of the relationship between board independence and probability of survival, we include quadratic and higher order terms. We find that board independence exhibits a negative coefficient, indicating that independence has a beneficial effect on firm survival (β = −.11, p < .05). The quadratic term is statistically significant at the 5 percent level (β = .001, p < .05). We also tried higher order terms. These are not statistically significant. Therefore in order to conserve degrees of freedom, we stop with the quadratic term.
Summing up, we observe a nonlinearity in the relationship between board independence and probability of survival. It appears that the benefits of board independence increase at a decreasing rate. Our evidence suggests that there exists an optimal level of board independence – somewhere in the middle, neither too little nor too much. It appears that insiders and outsiders play complementary roles in preventing firm failure. We find strong support for Hypothesis 1.
The leadership structure variables such as CM_NEXC and CM_DUAL do not significantly alter the IPO firms' chance of survival. Our empirical results do not support Hypotheses 2 and 3. Our results imply that CEO duality, which is deemed to be important for complex firms, neither increases nor reduces the IPO firms' likelihood of survival.
In order to check the robustness of our results, especially in the light of nonsignificance of CEO duality and nonexecutive chairperson, we used several alternate specifications. Arguably, large firms are more complex than small firms and may have different advising requirements. To explore this possibility, we interacted our governance variables with firm size (C_SIZE). We find that none of these additional variables are significant. Therefore, in the interests of brevity, we do not report these results.
Another possibility is that CEO duality and the presence of a nonexecutive chairperson may capture entrenchment effects. When the same person is both the CEO and chairperson, it increases the likelihood that the managers will resist a takeover attempt. Thus the observed coefficient captures the effects of entrenchment in addition to performance that we are essentially interested in. In our robustness checks, reported in the next section, we explicitly deal with this possibility.
Our results indicate that board size has a positive estimated coefficient (β = 2.58, p < .10). The square of board size has a negative coefficient (β = −.25, p < .10). Board size variables are marginally significant. Since there is no theory regarding the exact functional form of the relationship between board size and survival likelihood, we tried several higher order terms. These are not statistically significant. Therefore in the interests of conserving degrees of freedom, we stop with the quadratic term. It appears that there is a nonlinear effect of board size on survival likelihood. Our empirical results suggest that a small size board, and to a lesser extent large size board, have longer survival times compared to a mediumsized board. Our result is similar to that of Dowell et al. (2007). Thus we find weak support for Hypothesis 4.
The relationship between board size and survival is graphically portrayed in Figure 1.^{13} In Panel A, we map the survival function for firms with small, medium and large size boards. We characterize a board with fewer than four members as small, those with between four and six as medium, and those with more than six members as large. The graph shows the probability of survival over time (since listing). The graph clearly shows that firms with medium size boards have the lowest chance of survival at any given time. It is seen that firms with small and large board sizes are more likely to survive compared to firms with moderatesized boards. We reach a similar conclusion when we lump small and large boards into one category and medium into the other group. This is portrayed in Panel B. We, therefore, conclude that Hypothesis 4 is supported by data. We find support for the “stuck in the middle” form of the hypothesis and reject the “best of both worlds” version.
Among the control variables, offer size, underwriter backing, venture capital backing, debt equity ratio, and company size are statistically significant. The estimated hazard ratio for the variable VC_BACKED is 2.490 which indicates that the probability of financial distress for venture capitalbacked IPO companies increases by about 149 percent compared to nonventure capitalbacked IPO companies (β = .91, p < .10). Our finding of VCbacking being associated with higher failure likelihood could potentially be explained by a form of grandstanding that is studied in Gompers (1996). An alternate explanation is that the shortterm focus displayed by venture capitalists during the IPO period is deleterious for longterm survival. Similarly, the estimated hazard ratio for the BACK variable is 3.36, signifying that firms backed by underwriters are 3.36 times as likely to fail compared to firms which are not underwriter backed (β = 1.21, p < .10). This counterintuitive result may be explained by the possibility that risky firms seek underwriter backing. However, both these variables only display marginal significance.
Considering financial ratios, DET is the only financial ratio which is statistically significant in explaining the survival of IPO firms. The parameter estimates are positive for DET, which means that the IPO companies with low debt ratio are less likely to fail (β = .67, p < .05). The estimated hazard ratio for DET is 1.96 which indicates that for every unit increase in debt ratio, the risk of failing increases by 96.3 percent. C_SIZE is marginally significant (β = .68, p < .10), with a hazard ratio of 1.97. The positive sign of C_SIZE means that the larger the size of IPO companies, the higher the likelihood of companies entering into financial distress. Our results are consistent with prior research (Lamberto & Rath, 2008) but inconsistent with our expectation as outlined in the previous section. A possible explanation for this finding is that large firms are more complex than small firms and as such are more prone to failure, other things being equal.
Summing up, the results of our study show that new economy IPO companies with smaller value of total assets, lower leverage and those that are not VCbacked are more likely to survive. Interestingly, the dictum that the majority of the board should be composed of independent directors proves to be useful in reducing firm failure likelihood. We would like to add that the benefit of board independence increases at a decreasing rate signifying that insiders and outsiders act in a complementary manner to enhance the effectiveness of the board. Another remarkable finding is that small boards and very large boards are associated with a lower chance of corporate failure compared to medium size boards. The commonly touted recommendations of conventional leadership structure wisdom do not help in mitigating the risk of corporate failure. These are: (a) the chairperson should be an independent director and (b) the roles of chairperson and chief executive officer should not be exercised by the same individual.
Robustness Checks
We conduct two sets of robustness checks. These are based on alternate methods of classifying survivors and nonsurvivors. Our method of classifying survivors and nonsurvivors on the basis of delisting is subject to criticism. One may argue that delisting may be due to poor performance or takeovers. Delisting due to takeover by another firm is not necessarily indicative of poor performance. It may be the optimal response to increase shareholder wealth in the wake of a lucrative bid from an acquiring firm.
First, we estimate the survival likelihood for two subsets of firms – those that were delisted due to financial distress and those that were delisted due to takeovers and acquisitions by applying the competing risks model. Second, we estimate the Cox proportional hazards model excluding firms with good performance which were taken over. In the first method, we exclude all firms which were delisted due to takeovers. The resulting sample of nonsurvivors thus represents a clean sample of firms that performed poorly prior to delisting. In the second method, we only exclude firms with good performance which were taken over. Since our basic motivation in this paper is to identify the board characteristics that impact a new economy firm's survival likelihood, firms delisted due to other extraneous reasons are best left out. One potential problem with our attempt to construct a “cleaner” sample of nonsurvivors is the loss of sample size and the consequent reduction in the power of our statistical tests.
We report the estimation results from applying the competing risks Cox proportional hazards model in Table 6. For these tests, the set of firms delisted due to financial distress is categorized as nonsurvivors in Panel A. In Panel B, nonsurvivors are those firms delisted due to takeovers. Our results incorporate potential nonlinearity in the governance and ownership structure variables.
Table 6. Competing Risks Model of the SubsamplesCovariate  Coefficient  Standard error  χ^{2} Statistic  pvalue  Hazard ratio 


Panel A: Subsample of delisted firms due to financial distress 
BD_SIZE  6.22†  3.52  3.12  .08  501.29 
BD_SIZESQ  −.61†  .35  3.10  .08  .54 
BD_INDP  −.17*  .07  6.51  .01  .84 
BD_INDPSQ  .00*  .00  5.84  .02  1.00 
CM_DUAL  −.87  1.01  .74  .39  .42 
CM_NEXC  −.13  .74  .03  .86  .87 
TOP20  .14  .22  .42  .52  1.15 
TOP20SQ  −.00  .00  .20  .66  1.00 
OF_AGE  −.11†  .06  3.64  .06  .90 
IPO_9900  2.25*  .81  7.71  .01  9.48 
VC_BACKED  1.44†  .80  3.22  .07  4.24 
C_SIZE  1.14*  .51  4.97  .03  3.12 
Panel B: Subsample of delisted firms due to takeovers and acquisitions 
BD_SIZE  1.51  2.00  .57  .45  4.512 
BD_SIZESQ  −.14  .18  .65  .42  .87 
BD_INDP  −.12†  .07  2.85  .09  .89 
BD_INDPSQ  .00†  .00  3.16  .08  1.00 
CM_DUAL  16.90  2154  .00  .99  21887627 
CM_NEXC  .58  .84  .48  .49  1.79 
TOP20  −.24*  .11  4.72  .03  .79 
TOP20SQ  .00*  .00  5.54  .02  1.00 
DET  .92†  .52  3.05  .08  2.50 
The results in Panel A indicate that the significance levels of the board independence (β = −.17, p < .05) and board size (β = 6.22, p < .10) are similar to the full sample results reported in Table 5. As in the whole sample, C_SIZE is statistically significant (β = 1.14, p < .05). Once again, we observe nonlinearity in the relation between board size (and board independence) on survival likelihood. As before, we find strong support for Hypothesis 1, weak support for Hypothesis 4 and no support for the other hypotheses. In addition, age of the company (β = −.11, p < .10), and VC_BACKED (β = 1.44, p < .10) are marginally significant. OF_AGE has a hazard ratio of .90 indicating that increasing the age of the firm by one year on the offer date reduced financial distress likelihood by 10.3 percent. The significance of the dummy variable IPO_9900 (β = 2.25, p < .05) indicates that if a firm went public during the years 1999 or 2000, the chances of delisting increased by 848 percent.
In Panel B, we examine the subset of firms that delisted due to takeovers and acquisitions. Board independence (β = −.12, p < .10) and leverage (β = .92, p < .10) have marginally significant influence on the likelihood of survival. TOP20 is significant with a negative coefficient (β = −.24, p < .05). Both BD_INDP and TOP20 reduce the likelihood of delisting while leverage exacerbates the odds. We note that the significance levels of board independence have dropped compared to the full sample and financial distress subsamples. Our results from Table 5 and Panel A of Table 6 indicate that more independent boards are associated with a higher likelihood of firm survival and that the benefits of independence increase at a decreasing rate. Arguably, board independence is also associated with a higher probability of takeovers and acquisitions. Thus board independence is less of a distinguishing factor explaining new economy IPO firms' survival when delisting due to takeovers and acquisitions is considered as nonsurvival. Furthermore, board size is no longer significant. Our earlier results indicate that small boards, due to their speed of response, and large boards, due to their greater advising capability, are associated with a higher likelihood of firm survival. The exact same features should also be associated with the increased likelihood of being taken over. Thus board size ceases to be a distinguishing feature that explains survival likelihood when nonsurvival is characterized by delisting due to takeovers and acquisitions.
Our results indicate that ownership concentration is associated with higher corporate longevity or lower probability of being taken over. Our finding with respect to TOP20 is consistent with agency cost theory but inconsistent with the findings of Woo et al. (1995). This is because shareholders with significant holdings are more likely to have an influence on management's decisions and they will expend more monitoring costs as their stake in the firm increases (Jensen & Meckling, 1976). Higher ownership concentration is also likely to deter takeovers and acquisitions. Thus the observed significance of TOP20 is due to both effects – better monitoring and lower chance of takeovers. TOP20 is not significant in Panel A of Table 6 and in Table 5. We interpret this finding to imply that the takeover effect is much more significant than the monitoring effect. We find that the squared term (TOP20SQ) is significant (β = .002, p < .05), indicating a nonlinear relationship between ownership concentration and survival likelihood. Perhaps, this signifies the entrenchment effect if TOP20 shareholders include controlling shareholders.
We conduct further robustness checks using the Cox proportional hazards model excluding firms which had good performance and were taken over. Good performance is signified by nonnegative earnings during the two years preceding takeovers. Ostensibly, the acquired firms were taken over not because of distress and are therefore not classified as nonsurvivors. We confirm the significance of board size, the square of board size, board independence, the square of board independence, VC_BACKED, and DET. Overall, our results indicate that leadership structure does not significantly affect survival likelihood of new economy IPO firms. These results are not reported for the sake of brevity.
We also considered the possibility of simultaneity affecting our estimated results. Board size and other included control variables, such as firm size (C_SIZE), offer size (OF_SIZE), and offer price (OF_PRICE), could potentially be affected by simultaneity (Lehn et al., 2009). We considered two approaches to account for potential simultaneity. First, we reestimated our results without C_SIZE, OF_SIZE, and OF_PRICE. We obtained qualitatively similar results. Second, we used an instrumental variables approach. We regressed board size on C_SIZE, OF_SIZE, and OF_PRICE. The residual from this estimation was then added to our original model instead of C_SIZE, OF_SIZE, and OF_PRICE. The estimated results did not indicate simultaneity. These results are not reported for the sake of brevity.