Endogeneity

Table 21.6 presents regression models on the determinants of exit outcomes that are similar to those in Table 21.5 but with one major exception. In Table 21.6, we examine the effect of endogeneity of control rights on exits. We use two-step instrumental variable estimates. Step 1 of Models 3 and 4 accounts for the factors that affect the extent of VC control. Step 1 of Model 3 considers binomial logit estimates for majority board seats and the right to replace the CEO. Step 1 of Model 4 considers ordered logit estimates of the control rights index for the sum of dummy variables for drag-along, redemption, and antidilution rights. Step 2 of Models 3 and 4 estimates the multinomial logit model of exit outcomes. In Table 21.6 , we use the exogenous variables that are included in step 1 and excluded in step 2 of Models 3 and 4. These variables are the La Porta et al. (1998) indexes for creditor rights and antidirector (shareholder) rights. These instruments are intuitively related to contract terms, as confirmed by all of the correlations that are reported in Table 21.7 . We do not expect to find a relation between creditor rights and exit outcomes, unless the efficiency of bankruptcy law is related to write-offs. Table 21.7 indicates that creditor rights are not statistically related to any of the exit outcomes, which confirms the suitability of that instrument for the data. Antidirector rights may be positively related to the probability of an IPO (La Porta et al., 1998), but the data here indicate the opposite. The correlation between antidirector rights and IPO exits is 0.18. This negative correlation can only be explained by independent factors driving the IPO exits, and by little or no direct relation between IPOs and antidirector rights. In contrast, the antidirector rights index is significant and negatively related to the contract terms. This finding is consistent with the view that as shareholders, VCs use contracts as substitutes for the absence of strong legal protection (Lerner and Schoar, 2005). Creditor rights are positively and significantly related to contract terms, which is intuitive, because when decisions have be made in times of financial distress, VCs as shareholders want priority over and above creditors. Also, we use the industry market/book value at the time of first investment as an instrumental variable. This is a valid instrument because the second-step regressions use the industry market/book value at the time of exit. The results of the second step regressions are similar when we use the different instruments and different specifications for the first-step regressions and are available on request. In Table 21.6 , the step 1 evidence in Models 3 and 4 indicates that creditor rights are statistically and positively related to the extent of VC control and to the right to replace the CEO, but not statistically related to majority board seats. A one-point increase in the creditor rights index (on the scale of zero to four) increases the probability of the VC acquiring the right to replace the CEO by 31.6%. We note that although the creditor and antidirector rights are negatively correlated ( 0.42 for this sample), the statistical and economic significance of the regressions is not materially affected by the inclusion or exclusion of these variables. For instance, excluding antidirector rights in step 1 of Model 3 in Table 21.6 increases the economic significance of creditor rights from 31.6 to 34.3% for the right to replace the CEO without changing the statistical significance. A one-point increase in the creditor rights index also increases the probability of an extra control right (drag-along, redemption, or antidilution) by on average approximately 3%. The economic significance of a one-point increase in creditor rights is 4.9% to move from two to three, and 2.6% to move from three to four in step 1 of Model 4. In contrast, antidirector rights are more closely tied to majority board seats. An increase of one point in the index (on the scale of 0 to 6) reduces the probability of VC majority board seats by 12.6%. The step 1 regressions in Table 21.6 further indicate that VCs are more likely to have a majority on the board and to have greater control rights for those firms in industries with higher market/ book ratios (high-tech industries). VCs are less likely to take majority board seats, have the right to replace the CEO, and to have other control rights for earlier-stage investments. VCs are also less likely to have the right to replace the CEO and other control rights when the entrepreneur has a higher experience ranking. The step 2 regressions in Models 3 and 4 in Table 21.6 are consistent with those reported for Models 1 and 2 in Table 21.5 . Majority board seats lead to a 22.3% increase in the probability of an acquisition exit (Model 3), and each additional control right increases the probability of an acquisition by 14.5% (Model 4). The one result that differs from Models 1 and 2 is that the right to replace the CEO is statistically insignificant in Model 3. In general, the alternative specifications for the step 1 and 2 regressions in Table 21.6 invariably support at least one of the VC control right variables as being significantly related to acquisition exits. The instrumental variables regressions are not completely robust to alternative specifications, but nevertheless indicate support for the central proposition in the chapter that VC control influences the IPO versus acquisition choice, even after controlling for endogeneity.