environment in a country annually starting with 2004.

Allen, Qian, and Qian (2005) discuss several dimensions of China’s legal and regulatory

environment and compare to countries that are included in the La Porta et al (1998) indexes.

For one of the La Porta et al (1998) series that we use, anti-director rights, Allen, Qian, and

Qian (2005) estimate an index value of 3 for China while Pistor, Raiser, and Gelfer (2000)

estimate an index value of 3.45 for Russia. We substitute these values into any tests than

Include the anti-director rights index. 23 Some of the more detailed measures of the workings

of the legal system may also not fully describe the environment in a less developed country.

For example, Djankov et al (2003) reports values for the time it takes to collect on a bad check

or evict a tenant for non-payment of rent, and their tables include China. However, checks are

not as widely used in China as in other economies and evictions do not often involve formal

legal proceedings. Measures of the ease of raising capital and adjudicating legal and regulatory

problems are interpreted as if maximizing shareholder wealth is the primary goal of

corporations. Finally, the environment perceived by investors and the measures tabulated by

these sources can depend on whether the investor is a local or a foreigner. The apparent success

of China’s economy in the absence of a high quality legal and regulatory environment raises

concerns about the research efficacy of any such measures (Allen, Qian, and Zhang, 2011).

23

Alternatively, we can consider China as the base case reflected in our regression intercepts.

23

Thus, there are limitations to the usefulness of these indicators, particularly as they apply to one

of our most interesting and important home countries, China.

3.3 Methodology

To test our hypotheses from Section 3.1, we begin with several test specifications inspired

by Doidge, Karolyi, and Stulz (2004). We regress Tobin’s q on firm and home country

characteristics and an OFC dummy variable:

qit = α0 + α1Dofc,it + β’xit + δ’cit + εit

(8)

Dofc,it is the OFC dummy equal to one if firm i is incorporated in an OFC in year t. xit is a vector

of company characteristics, and cit is a vector of country characteristics. The observations are

firm-years for two types of firms. Sample firms are those located in a non OFC country but

incorporated in an OFC country, such as a US firm incorporated in the Cayman Islands. In

addition, our OFC sample also includes a small number of firms located in an OFC country but

incorporated in a different OFC country. Control firms are both located and incorporated in non

OFC countries. Firms which are both located and incorporated in the same OFC are excluded.

This excludes the relatively large number of firms that are both located and incorporated in

Hong Kong, Ireland, Singapore, and Switzerland.

24

issue. The classic “treatment effects” problem for (8) is that an apparent relationship between

individual firm Tobin’s q and incorporation in an OFC can induced by the decision about where

to incorporate, rather than from OFC incorporation itself. The effect is akin to that of an

omitted variable in (8) if the decision to incorporate, Dofc,it, is a function of firm characteristics,

xit, and country characteristics, cit, and the error terms in (8) and the selection equation (which

explains the unobserved value driving Dofc,it) are correlated. Therefore, we estimate a two-stage

procedure following Doidge, Karolyi, and Stulz (2004) and Bailey, Karolyi, and Salva (2006).

The first stage is a Probit regression to identify factors that predict whether a firm chooses to

incorporate in an OFC. The dependent variable is the OFC dummy. The independent

variables are country and firm characteristics. The second stage regresses firm-specific Tobin’s

q on the OFC dummy variable and country and firm characteristics as in (8), plus the inverse

Mills ratio from the first stage.

As an alternative to the Heckit procedure, we can think of (8) as suffering from an

endogenous variable, Dofc,it, among its explanatory variables. Therefore, we also estimate a

two-stage least-squares (2SLS) version of (8).

Another single-stage variation on (8) adds slope dummy terms equal to the explanatory

variables times the OFC dummy. This decomposes the OFC valuation discount or premium

measured by the slope, α1, on the OFC dummy in (8).

25

disclosure quality measures on other firm characteristics, home country characteristics, and

OFC dummy variable in a specification similar to (8):

yit = α0 + α1Dofc,it + β’xit + δ’cit + εit

(9)

yit represents a firm specific governance or disclosure characteristic. If governance and

disclosure measures are just cross-section or do not vary significantly over time, then this

regression must be purely cross sectional. Else it can be estimated as a pooled regression.

There may be some firm specific governance measures (such as the concentration of ownership

from Datastream) that vary over time. Variations on (9) include the two-stage Heckit, two stage

least squares, and the slope dummies specification described above for (8).

In addition, we find a few dozen firms which experienced an OFC reincorporation event,

that is, either moving incorporation to or from an OFC. We conduct the event study following

Bailey, Karolyi, and Salva (2006). Regressions similar to (8) and (9) as described above also

include the OFC incorporation dummy to distinguish firm-years when the firm was

incorporated in an OFC from other firm-years.

26

performance, and relate this to offering, firm, and incorporation country characteristics.24 For

a first look at IPOs for this draft of the paper, we focus on firms from China only, given the

interest and controversy surrounding them.

4. Empirical results and discussion

4.1 An overview of the data

Table 1 tabulates country of incorporation, country of address, and listing exchange for

the firms in the OFC sample. Panel A includes an observation for each pair of firm and listing

exchange (2397 total). This means that a firm with two listings on two different exchanges

counts twice. On the other hand, Panel B includes an observation for each distinct firm only

(1372 total). Comparison of the two panels suggests how many of the OFC-incorporated

firms are listed in more than one market.

Across Panels A and B of Table 1, it is evident that country of incorporation is heavily

concentrated in two jurisdictions, Bermuda and the Cayman Islands, which account for over

75% of the sample firms. The British Virgin Islands is a distant third, with less than 5% of the

OFC sample firms. Among the home countries for these OFC incorporated firms, Hong Kong

firms account for more than half of the OFC sample, Chinese firms about a fifth, and UK and

USA firms each less than 10% of other OFC sample firms. Most Hong Kong address firms are

24

See, for example, Welch (1991) for a cross sectional regression approach.

27