Lockups can be used as a signal to reduce information asymmetry or as a commitment solution to mitigate moral hazard problems. Previous studies have provided different results regarding the relative importance of these two arguments in explaining IPO lockup length. Due to the unique characteristics of REITs, we may expect the determinants of lockup length for REIT IPOs to differ from those for industrial IPOs. We use the following Tobit model to examine the determinants that are specific to the lockup length of REITs IPOs:
LOCKUP = α + β1TOP6_AUDITOR + β2UNDERWRITER_RANK + β3LNPROCEEDS + β4MSE + β5SELF_MGT + β6RETAIL + β7RESIDENTIAL + (1) β8OFFICE_INDUSTRIAL + β9LODGING_RESORT + β10BULL,
where LOCKUP is the natural logarithm of 1 plus lockup length, which could be zero or positive.
TOP6_AUDITOR is a dummy variable to proxy for the quality of the issuing firm’s auditor, which is equal to 1 if the auditing firm is one of the top six auditing firms, and zero otherwise.1 UNDERWRITER_RANK is the Carter–Manaster reputation rank, which measures the reputation
1 These auditing firms are Arthur Andersen, Coopers & Lybrand, Deloitte & Touche, Ernst & Young, KPMG, and Price Waterhouse.
of a lead underwriter. Based on the Carter–Manaster ranking, we classify the reputations of underwriters into 10 categories ranging from zero to 9, with 9 being the highest reputation rank.
LNPROCEEDS is the offer size of a REIT IPO, which is the natural logarithm of the product of the number of IPO shares offered and the offer price. MSE is the mean square error of the market model regression, which proxies for idiosyncratic risk. We regress the daily REIT returns against the daily CRSP value-weighted index returns over the 60-day period after the IPO. SELF_MGT is a dummy variable, which equals 1 if the REIT is self-managed, and zero otherwise. RETAIL, RESIDENTIAL, OFFICE_INDUSTRIAL and LODGING_RESORT are dummy variables, which equal 1 for the corresponding property type, and zero otherwise. They are used as control variables to capture the property effect. BULL is a dummy variable, which equals 1 if the CRSP value-weighted index return is greater than the T-bill return over the 12–month period preceding the IPO, and zero otherwise.
Previous studies have found relations between lockup length and firm value, auditor quality and underwriter reputation, and venture capital backing. For example, by developing a signaling model, Courteau (1995) shows that the use of lockup agreement can signal the value of a firm. Brau, Lambson, and McQueen (2005) argue that auditor quality or underwriter reputation can provide certification for offerings to investors and improve transparency. Consequently, firms with high-quality auditors or reputable underwriters are more transparent and may rely less on lockups for the purpose of signaling. Conversely, opaque firms are more likely to impose longer lockups. Arthurs, Busenitz, Hoskisson, and Johnson (2009) show that a longer lockup period serves as a substitute for venture backing and that underwriter reputation serves as a signal to reduce the uncertainty embedded in industrial IPOs. Therefore the signaling hypothesis predicts a negative relation between auditor quality and underwriter rank and the lockup length.
A lockup agreement may also serve as a commitment device for alleviating moral hazard problems. Brav and Gompers (2003) find that, holding the quality of the firm constant, a firm with more severe moral hazard problems tend toward longer lockups whereas firms with other forms of certification suffice with shorter lockups. Consequently, if the commitment hypothesis holds, we expect an inverse relation between underwriter reputation/auditor quality and lockup length.
Firm size has been widely used in the finance literature as a proxy for both information asymmetry and moral hazard. In general, large firms are more transparent due to higher levels of analyst and media coverage and thus do not require long lockups to provide positive signals to investors. Thus, according to the signaling hypothesis, the LNPROCEEDS variable, which serves as a proxy for firm size, is expected to be negatively related to lockup length.
Previous studies show that larger firms are less likely to be disciplined by the market for corporate control due to a higher percentage of atomistic shareholders. In addition, due to the complexity of large firms, monitoring management may be more difficult. In testing their moral hazard theory of managerial compensation, Gayle and Miller (2009) show that the divergence of interests between managers and shareholders increases with firm size. If, in fact, moral hazard increases with firm size, firms with a large offer size are likely to have a longer lockup period to mitigate moral hazard problems. Thus, a positive relation between LNPROCEEDS and lockup length is consistent with the prediction of the commitment device hypothesis.
MSE is used as a proxy for idiosyncratic risk. A high level of idiosyncratic risk is associated with the high cost of being undiversified; therefore, pre-IPO shareholders are not likely to prefer a long lockup period. Consequently, short lockup periods signal that insiders are unwilling to hold risky assets for the long term. In addition, Brau et al. (2005) argue that risky
firms set up shorter lockups to avoid sending out costly, false signals. From this perspective, if the signaling hypothesis holds, we expect an inverse relation between MSE and lockup length.
SELF_MGT is included to control for the impact of management style. As previously discussed, lockup length is closely related to the management style of REITs. To analyze this relation, we must understand the possible impact of management style on the nature of the moral hazard problem. At first glance, externally managed REITs are expected to have a more severe moral hazard problem and higher agency costs than their self-managed counterparts. Prior evidence seems to support this claim. For example, Capozza, and Seguin (1988, 2000), using different performance measures, find that externally managed REITs underperform self-managed REITs. Such results imply that the operation of externally managed REITs is more complicated and costly. Thus, we expect that the decision making of self-managed REITs is more straightforward and transparent, resulting in a negative relation between SELF_MGT and lockup length.
However, self-managed REITs may have longer lockups than externally managed REITs for several reasons. First, as previous studies show, self-managed REITs outperform externally managed REITs. That is, pre-IPO shareholders of self-managed REITs are more willing to be long-term investors and accept a longer lockup period to demonstrate their commitment than pre-IPO shareholders of externally managed REITs.
Second, managers of self-managed REITs and externally managed REITs are compensated on the basis of different performance measures. External managers are compensated on the basis of property level income and asset size, and self-managed REITs are compensated according to shareholder value. Most important, external managers are less likely to invest in REITs under their own management, meaning that lockups do not serve as an
effective bonding solution to the moral hazard problem. Because external managers have no (or limited) investments in the companies they manage, lockup agreements are less relevant to them.
In contrast, internal managers are compensated according to performance measures that are closely related to the share values of their companies, making a lockup agreement a more effective tool in controlling the moral hazard problem for self-managed REITs.
Third, other mechanisms such as financial leverage are used to control the moral hazard problem for externally managed REITs. The extant evidence indicates that externally managed REITs are more leveraged than self-managed REITs. For example, Capozza and Seguin (2000) report that the debt ratio for self-managed (externally-managed) REITs is approximately 33%
(44%). Jensen (1986) suggests that leverage can reduce the agency cost of free cash flows.
Accordingly, the moral hazard problem of externally managed REITs is under control to some extent, and externally managed REITs are not necessarily more associated with longer lockup periods than self-managed REITs.
As previously discussed, REITs with different property types have different characteristics (see Table 3, Panel E), and these differences are likely to affect the lockup decision. For example, Gyourko and Nelling (1996) show that the systematic risk of equity REITs varies by property type. They find that REITs with a large proportion of their assets invested in retail properties have higher betas. To control for the impact of property type, we include dummy variables for the four major types of REITs (retail, residential, office/industrial, and lodging/resort) in the Tobit analysis.
Market conditions may also affect lockup length because lower levels of asymmetric information may prevail during a market upturn. Therefore, a shorter (longer) lockup period is expected during a bullish (bearish) market. In a bull market, investors may be too optimistic to
be concerned about the moral hazard problem and thus are willing to accept a shorter lockup period.
Empirical results
Table 4 presents the results of Tobit regression. Model 1 includes the key variables used to test the signaling hypothesis and the commitment device hypothesis, with MSE controlling for idiosyncratic risk. Model 2 incorporates dummy variables for property types, and Model 3 controls for market conditions by including the variable BULL.
[TABLE 4 ABOUT HERE]
In Model 1, auditor quality is not significantly associated with lockup length, but the coefficient of UNDERWRITER_RANK is negative and significant at the 5% level. We find that REITs that hire more reputable underwriters tend to have shorter lockup periods. This result is consistent with the signaling argument in that reputable underwriters can use lockups as a substitute to reduce the degree of investor uncertainty. It is also consistent with the prediction of the commitment hypothesis as set forth by Brav and Gompers (2003).
The coefficients of offer size (LNPROCEEDS) and management structure (SELF_MGT) are significantly positive at the 1% and 5% levels, respectively. Contrary to the prediction of the signaling hypothesis, firms with larger offer size tend to use a longer lockup period. This evidence suggests that large firms may suffer from a more severe moral hazard problem than small firms and thus rely on a longer lockup period to mitigate this problem. Note that our result does not imply that large firms are more informational asymmetric, it just indicates that the
signaling hypothesis cannot explain the negative relation between firm size and lockup length.
The significant and positive association between SELF_MGT and lockup length is consistent with the notion that lockups are more relevant to self-managed REITs than to externally managed REITs. Our findings imply that a lockup is a more effective tool in reducing the moral hazard problem for self-managed REITs, thus supporting the prediction of the commitment device hypothesis.
The result in Model 2 shows that the inclusion of property type does not change our major findings in Model 1, except that the coefficient of UNDERWRITER_RANK is no longer significant at the conventional level. Both the coefficients of LNPROCEEDS and SELF_MGT remain positive and significant. Out of the four dummy variables used to control for property type, only LODGING_RESORT is significantly and positively related to lockup length. The result indicates that among all REITs—regardless of whether they have lockups—managers of lodging/resort REITs tend to set longer lockup periods to show a stronger level of commitment.
However, as shown in Table 3, Panel E, across all lockup groups, lodging/resort REITs have a much shorter average lockup period than REITs that focus on any other property types.
For Model 3, we find no significant relation between market conditions and lockup length.
This result implies that market conditions do not prompt a change in either the level of asymmetric information or the moral hazard problem. The coefficients of LNPROCEEDS and SELF_MGT are again significantly positive. Similar to Brau, Lambson, and McQueen (2005), we find no significant relation between idiosyncratic risk and lockup length in any of the three models. We also find that the coefficients of auditor quality and underwriter reputation remain insignificant.
In summary, the results for TOP6_AUDITOR and UNDERWRITER_RANK, as reported
in Table 4, provide no strong support to the prediction of the signaling hypothesis. While all the coefficients of TOP6_AUDITOR are insignificant, the coefficient of UNDERWRITER_RANK is significant with the expected sign in only one model. In addition, the positive relation between offer size and lockup length is not consistent with the signaling hypothesis. Our results are, however, consistent with the commitment device hypothesis in that the moral hazard problem is more severe for self-managed and larger REITs, which use longer lockups to solve this problem.