Enhancing security value by ownership restrictions: evidence from a natural experiment.

We present new evidence from a natural experiment to show circumstances in which ownership restrictions can enhance value. Our evidence is based on multiple restricted bond issues by an emerging market issuer at 150 basis points lower than comparable bonds, resulting in a billion dollars saving. This is intriguing: how can an emerging market issuer with junk bond ratings obtain such low yields? We argue ownership restrictions enhance value since they enable an issuer to precommit to renegotiate efficiently with a favored clientele in the potential default states, thereby circumventing deadweight costs of prolonged negotiations, particularly when the restricted clientele also values the underlying collateral higher than other investors. Ownership restrictions can also result in a transfer of value from holders of unrestricted bonds to holders of restricted bonds because of implicit seniority of the latter. We empirically test and find support for both value enhancement and value transfer and show robustness to several alternative explanations. Our evidence suggests' that firms can benefit from designing securities with ownership restrictions, by offering new securities exclusively to investors who value them the most.

**********

The literature on security design has typically given relatively little attention to the role of ownership restrictions in designing securities. (1) Interestingly, few articles in this area suggest restricting ownership in any fashion. The general view is that ownership restrictions are unnecessary or even detrimental. They are unnecessary because investors can self-select which securities they wish to hold (therefore, there is no need for a firm to restrict securities ex ante). They can be detrimental because ownership restrictions can affect demand, leading to a negative effect on a security's price because of constriction of demand. (2)

We hypothesize and present evidence that ownership restrictions can be beneficial to an issuer in obtaining higher prices (lower yields) for its bond offering to a targeted clientele. We present evidence from a natural experiment: multiple events of capital raising totalling $9.7 billion by an emerging market issuer (namely, India's largest bank, the State Bank of India) exclusively from Indians living abroad. We provide evidence that these bonds were priced substantially higher than comparable emerging market debt leading to a difference of about 150 basis points in yields which translates into large bottom line savings of over a billion dollars. We confirm this difference using a variety of methods including matching procedures based on Near Neighbor, Gaussian, and Epanechnikov kernel based estimators (see Heckman, Ichimura, and Todd, 1997, 1998), as well as regression based yield estimators. This is an intriguing issue because it raises the question, how can an issuer with a junk bond rating (India's credit rating at that time) obtain such low yields (high prices)? And are there lessons inherent in this security design that can be replicated elsewhere?

There are a number of reasons why a niche clientele may value a security more. First, there may be a perceived difference in credit rating by the niche clientele. A second reason is that a niche clientele may value the underlying collateral in the potential default states more than other investors. (3) For example, India may default if it does not have enough foreign exchange reserves, and it is plausible that they would pay in the local currency (Indian Rupees), which are not freely convertible into US dollars. Who values collateral in Indian Rupees more? Naturally, the Indians living abroad because they face a lower transaction cost in using the local currency (e.g., for purchase of real estate, jewelry, family support payments etc.--see Appendix A for details). Thus the value in the potential default states, and hence the expected value, ex ante is higher for this niche clientele. Finally, renegotiation costs are lower when a firm deals with a homogenous clientele in the default states. Gilson, John, and Lang (1990) provide related empirical evidence from the bankruptcy literature, and show that firms with more layers of creditors are less likely to restructure privately, out of court, an alternative that is less expensive than a bankruptcy court. Furthermore, these renegotiation costs are even lower if the niche clientele has reasons to value the underlying collateral more as suggested above.

The above mentioned reasons explain to some extent why a niche clientele has a higher expected valuation and is willing to pay more than other investor clienteles for the securities offered by the issuer. However, none of the above reasons by themselves explain why ownership restrictions are necessary. If the securities are offered freely, rational investors would take the above factors into account, resulting in the niche clientele holding the securities, and rendering the ownership restrictions unnecessary.

We present two alternative rationales for why ownership restrictions can be potentially valuable. The first one is that ownership restrictions allow for value enhancement through credible precommitments. To understand this rationale consider that the Indians living abroad clientele is an important one for India. They remit substantial foreign exchange, and can directly or indirectly vote in the political process. By restricting the security to a clientele that an issuer cares about, the issuer is able to precommit to treat this clientele more favorably than other investors in an event of default. That is, ownership restrictions ensure that other investors do not free ride (see Grossman and Hart, 1980) on the favored clientele. Absence of ownership restrictions can lead to multiple classes of investors (due to potential free-riding) and dilute an issuer's incentives to renegotiate efficiently in the potential default states. (4) Thus, ownership restrictions help an issuing firm to overcome this free-rider problem and make an effective precommitment to renegotiate efficiently in an event of default. More generally, where an investor clientele has a broader relationship with the firm, one that has dealings that extend beyond its current investment in the issuer's securities (e.g., the investor is a dependable provider of past and future capital and values the collateral more), the issuing firm is able to credibly precommit to an efficient ex-post renegotiation by issuing securities exclusively to that investor clientele. This rationale is somewhat similar to loans made by the mafia, which are invariably repaid as the cost of non-repayment can be very high (e.g., loss of a leg). By taking loans from a clientele that is critical for continued success of a business, such as the mafia (or the Indians living abroad in our context--See Section III.C for details), there is an implicit ex-post commitment to repay the debt which translates into higher prices ex ante.

Thus, we argue that ownership restrictions enhance value since they enable an issuer to precommit to renegotiate efficiently in the potential default states (as discussed above, by eliminating incentives to free-ride by other investors), thereby circumventing the deadweight costs of prolonged negotiations, particularly when a security is restricted to a homogenous clientele that values the underlying collateral higher than other investors.

A second rationale is that ownership restrictions can result in a transfer of value from other securities. The reason why this occurs is that ownership restrictions influence the priority structure of claims (and indirectly affect the value of collateral), effectively making the restricted securities implicitly senior to the unrestricted securities. This can result in a transfer of value from unrestricted securities to restricted securities.

We empirically test to see if the transfer of value explains the yield differential. Specifically, we test for an implicit transfer of wealth from the existing bond holders to the new bond holders by conducting an event study of the effect of the issuance of the restricted bonds on prices of the unrestricted bonds. One needs to be careful here because such an effect might also come from a restricted debt capacity of the issuer, and of the sovereign nation, so that any new issue of debt securities leads to a decline in the pricing of existing debt securities. Hence, we also check if there is a differential effect between the issuance of new restricted bonds versus new unrestricted bonds on the prices of existing unrestricted bonds. We find a significantly higher price decline for the former case, i.e., for the issuance of new restricted bonds. This is consistent with a transfer of wealth to the holders of restricted bonds from the holders of unrestricted bonds of other Indian firms.

However, we find that some, but not all, of the yield differential is explained by value transfer associated with the implicit higher seniority of the restricted bonds. This suggests that ownership restrictions also enhance value potentially through lower renegotiation costs (since they circumvent the deadweight costs of prolonged negotiations). We also examine whether our results can be rationalized by alternative explanations, such as a higher perceived credit rating, different measures of credit rating, (5) market segmentation, (6) commissions, and taxes. We find that our results are robust to controlling for these alternative explanations.

In summary, we find evidence in support of both the enhancement of value and the transfer of value (from holders of unrestricted securities to restricted securities). Our evidence suggests that firms can benefit from designing securities with ownership restrictions, where securities are offered exclusively to niche clienteles who value them the most. A natural extension of the evidence contained in this experiment is that issuers from other emerging market countries such as China, Israel, and Korea could bypass traditional underwriting routes to take advantage of their large expatriate populations by devising securities that get them higher prices (and lower yields). More generally, firms with niche clienteles where the investor clienteles have a broader relationship with the firm (such as, also being customers or suppliers) can benefit from issuing securities with ownership restrictions. Overall, our analysis shows the wider ramifications of ownership restrictions than suggested in the literature, as an integral part of security design, in lowering (rather than raising) the cost of capital for the issuing firm.

Our article contributes to the literature on foreign ownership restrictions by showing that ownership restrictions can sometimes be value enhancing, whereas the extant literature largely shows that securities with ownership restrictions trade at a substantially lower price than comparable securities without restrictions.

The remainder of the article is organized as follows. Section I outlines our data and sample selection. Section II describes the measurement of yield differentials, along with some robustness checks. In Section III, we present an economic rationale for why some investors are willing to pay more than others for a security and whether the issuing firm can benefit from restricting the offering to them. Section IV empirically tests an implication of restricting the ownership of securities, namely whether there is a transfer of wealth from holders of unrestricted securities to holders of restricted securities. Section V concludes.

I. Data and Sample Selection

To examine the role of ownership restrictions in the security design problem, we conducted an indepth analysis of a natural experiment: multiple events of capital raising by an emerging market company with ownership restrictions, namely $4.2 billion of Resurgent India Bonds and $5.5 billion of India Millennium Bonds offered by India's largest bank, State Bank of India, exclusively to Indians living abroad at approximately 150 basis points below comparable benchmarks.

We collected necessary data to confirm whether there is a significant yield difference between these bonds and comparable benchmarks, and additional data to test for sources of this yield differential. We obtained the data for our empirical analysis from several sources, such as the Securities Data Company's Global New Issues Database, the International Finance Corporation Emerging Markets Factbook, Datastream for historical price and yield data of bonds and bond indices, Moody's website for credit ratings, Dow Jones News Service for assessing credit outlook, Federal Reserve Bank of Chicago's web site for current and historical interest rates, and the websites of the Reserve Bank of India and the State Bank of India for information on Resurgent India Bonds and India Millennium Bonds.

Our sample selection procedure was as follows: First, we extracted the fixed-rate US dollar denominated non-convertible debt issues of foreign issuers from emerging markets (listed in the International Finance Corporation Emerging Markets Factbook) from the Securities Data Company's Global New Issues database that have a maturity of at least one year at the time of offering of the restricted bonds (i.e., Resurgent India Bonds or India Millennium Bonds). Second, we obtained price and yield information of these comparable sovereign and corporate bonds from Datastream. Third, we obtained Moody's credit rating (7) for Resurgent India Bonds, the India Millennium Bonds, and for the comparable bonds at the time of offering of the Resurgent India Bonds and the India Millennium Bonds. In addition, we incorporated additional information about the credit outlook on the issuer and whether the issuer's credit rating was under review for a potential upgrade or downgrade by searching Dow Jones News Services during the 12 months prior to the Resurgent India Bond or Millennium India Bond issue dates. Finally, to obtain yield spreads, we subtracted the yield on the Treasury of comparable maturity obtained from the Federal Reserve Bank of Chicago's web site.

A. Variables

We first test if there is a yield differential between the restricted securities (i.e., Resurgent India Bonds, and India Millennium Bonds) and the unrestricted securities (i.e., comparable bonds). Our main variable of interest is therefore YIELD SPREAD, defined as the ex ante yield of a debt security minus the ex ante yield of a US Treasury security of comparable maturity, measured in basis points. The independent variables for the regression results in Section II.C are as follows:

CREDIT RATING: Stands for a set of comprehensive credit rating dummy variables (B3, B2, B1, Ba3, Ba2, Ba1, Baa3, Baa2, Baa1, A3, A2, Aa3) based on Moody's comprehensive credit rating. (8) For example, Ba1 is a dummy variable which is one if Moody's comprehensive credit rating for the issue is Ba1. The dummy variable is zero otherwise.

MATURITY: The maturity of a debt issue measured in years at the time of offering of the Resurgent India Bonds or the India Millennium Bonds.

EXCHANGE: A dummy variable that takes a value of one ifa debt issue is traded on an exchange, and zero otherwise.

SOVEREIGN: A dummy variable that takes a value of one if a debt issue is a sovereign debt issue, and zero otherwise.

RESTRICTED: A dummy variable that takes a value of one for the Resurgent Indian Bond issue and for the Indian Millennium Bond issue, and zero otherwise.

B. Discussion of Variables

The economic rationale for using these variables is as follows: credit rating of a debt issue is clearly important and one would expect lower-credit rated issues to have higher yield spreads. Maturity is another variable potentially affecting yield spreads. In particular, if the probability of default increases with debt maturity (see Flannery, 1986) then we should expect to see higher yield spreads with longer maturity issues. A debt issue listed on an exchange is more liquid and is associated with more public information, and one would expect an exchange-listed debt issue to have a lower yield spread. Sovereign issuers, due to their ability to tax local corporations and individuals, need to be distinguished from corporate issues who do not have such authority. Finally, the size and sign of the coefficient of the variable RESTRICTED indicates whether or not the offering yield spread of the Resurgent India Bonds and the India Millennium Bonds is lower than that of comparable benchmark bonds.

II. Do Ownership Restrictions Depict a Lower Yield?

In this section, we analyze whether ownership restrictions can result in a lower yield for the issuing firm. Specifically, we examine whether the offering yield on the Resurgent India Bonds and the India Millennium Bonds is significantly lower than that of comparable bonds using a variety of matching procedures and regression specifications. We also examine how robust our finding of any yield differential is to alternative explanations such as a higher perceived credit rating, different measures of credit rating, market segmentation, taxes, and commissions. We focus our analysis of the offering yields in US dollar terms since the proceeds from these bond issues were predominantly in US dollars (e.g., 95% in case of Resurgent India Bonds).

See Table I for salient features of the Resurgent India Bonds and India Millennium Bonds. Both these bonds carry a semi-annual coupon and mature in five years. Two key features of these bonds are their yields, touted in the Indian financial press as being highly attractive and that they are offered exclusively to Indians living abroad, also known as "Non-resident Indians". (9) Both these bonds are illiquid and are not traded on an exchange. However, transfer of ownership from one Non-resident Indian to another Non-resident Indian is possible through endorsement and physical delivery of the bonds. In this regard, the Resurgent India Bonds and India Millennium Bonds resemble standard bank deposits, albeit of a longer-maturity. Finally, these bonds are marketed on a best efforts basis through the branch networks of State Bank of India and of the collecting banks (who receive a commission based on the amount they collect) rather than sold to an underwriting syndicate of banks.

A. Univariate Results

Panel A of Table II presents the average yield spread for comparable debt issues, segmented based on the Moody's comprehensive credit rating on the day prior to the offering date of Resurgent India Bonds (i.e., August 4, 1998), and India Millennium Bonds (i.e., October 20, 2000). The average yield spread is the yield on a debt issue minus the yield on a US Treasury of comparable maturity, expressed in basis points. As can be seen, the average yield spread is 440 basis points for comparable debt issues with a Ba3 Moody's credit rating (same as that of State Bank of India's credit rating). However, the offering yield on Resurgent India Bonds and India Millennium Bonds was 231 and 281 basis points, respectively, which averages to 256 basis points, significantly lower than the estimated yield based on comparable debt issues of 440 basis points. This translates to an average yield saving of 184 (=440-256) basis points.

Panel B of Table II provides more detailed summary statistics of the comparable bonds, such as how many comparable bonds are from each country, average yield spread, average credit rating, average maturity, and the fraction of these bonds that are exchange-listed, and sovereign-grade. It is clear that the India-only sub sample is too small which limits us from making meaningful inferences on yield differentials using that sub sample. However, the summary statistics do provide an initial-cut on what to expect in subsequent results using regression methods and matching procedures. For example, Argentinean bonds with an average maturity of 4.91 years (relative to 5-year maturity of the restricted bonds), and an average numeric credit rating of 3.55 (relative to the numeric value of 4 associated with the Ba3 credit rating of State Bank of India) have an average yield spread of 391 basis points as compared to the 256 basis points of the restricted bonds. This results in an average yield saving of 135 (=391-256) basis points. The political economy of Argentina and India have some similarities, such as a democratic rule of law, GDP growth in high single-digits, high levels of government corruption, etc. Additionally, if we consider Turkey which has a similar credit rating as that of the restricted bonds, but a slightly higher maturity, the average yield savings is 324 (=580-256) basis points. Both choices of a comparable country for the benchmark bonds reveal a substantial yield savings associated with the restricted bonds.

Clearly, the yield savings (e.g., of 184 basis points based on Panel A of Table II) is only indicative since we must also control for the issue characteristics, such as the maturity of the issue, and whether the debt issue is traded on an exchange or not.

We proceed to control for the issue characteristics using two different approaches: 1) using a variety of matching procedures, such as the Near Neighbor, Gaussian, and Epanechnikov estimators (see Heckman, Ichimura, and Todd, 1997, 1998 for details), and 2) multivariate linear regressions.

B. Yields Computed by Propensity Scores and Matching Procedures

The formal econometric methods of matching are developed in Rosenbaum and Rubin (1983), Heckman and Robb (1986), and Heckman, Ichimura, and Todd (1998). Below, we provide a summary of their results.

We consider the case where a bond can belong to one of two groups, numbered 1 and 0. Let D=1 denote the treatment, which in this case is a restricted bond offering, such as the Resurgent India Bonds and India Millennium Bonds, and let D=0 represent the control, which is if the bond is unrestricted. In principle, the ith bond has both a price [Y.sub.1i] that would result if it had received treatment and another price [Y.sub.0i] that would result if it did not receive the treatment. The effect of interest is a mean effect of the difference between [Y.sub.1] and [Y.sub.0]. However, since we only observe [Y.sub.1] for our sample of restricted bonds, we have a missing data problem that cannot be solved at the level of the individual, so we reformulate the problem at the population level. We focus on the mean effect of the difference between restricted bonds and unrestricted bonds with characteristics X:

(P - 1) E([Y.sub.1] - [Y.sub.0]|D = 1,X).

The mean E([Y.sub.1]|D = 1,X) can be identified from the data on restricted bonds. However, assumptions must be invoked to identify the unobservable counterfactual mean, E([Y.sub.0]|D = 1,X) The observable outcome of self-selected unrestricted bonds E([Y.sub.0]|D = 0,X) can be used to approximated E([Y.sub.0]|D = 1,X). The selection bias that arises from this approximation is:

B(X) = E([Y.sub.0]|D = 1,X) - E([Y.sub.0]|D = 0,X).

The method of matching solves the evaluation problem. Following Heckman and Robb (1986), we assume that all relevant differences between restricted and unrestricted bonds are captured by their observable characteristics X. Let:

(A - 1) ([Y.sub.0], [Y.sub.1]) [perpendicular to] D|X.

denoted the statistical independence of ([Y.sub.0], [Y.sub.1]) and D conditional on X. Rosenbaum and Rubin (1983) establish that when (A-1) and:

(A - 2) 0 < P(X)<1.

(which are referred to as the strong ignorability conditions) are satisfied, then ([Y.sub.0], [Y.sub.1]) [perpendicular to] D|P(X), where P(X) = Pr(D = 1|X). While it is often difficult to match on high dimension X, this result allows us to match based on the one-dimensional P(X) alone. P(X), known as the propensity score, can be estimated using probit or logit models.

Heckman et al. (1998) extend this result by showing that the strong ignorability conditions are overly restrictive for the estimation of (P-1). Instead, a weaker mean independence condition is all that is required:

(A-3) E([Y.sub.]0|D = 1, P(X)) = E([Y.sub.0]|D = 0, P(X)).

We use the Near Neighbor, Gaussian and Epanechnikov estimators discussed in Heckman, Ichimura, and Todd (1997, 1998). These estimators construct matches for each restricted bond by using weighted averages of the outcomes of multiple observations of the corresponding unrestricted bonds. Let [Y.sub.1i] be the yield spread of a restricted bond, and let [Y.sub.0j] be the yield spread of a corresponding unrestricted bond, and let [[bar.Y].sub.0j] represent the weighted average of yield spread of the unrestricted bonds.

The Near Neighbor method chooses for each restricted bond, the n non-restricted bonds with the closest propensity score. We specify n=10 and n=25 to obtain two separate estimates using the Near Neighbor method. We also use two different kernels to compute [[bar.Y].sub.0j]. Specifically, if the weights from a typical symmetric, non-negative, unimodal kernel K(.) are used, then the kernel places higher weight on bonds close in terms of P(X) and lower or zero weight on more distant observations. The Gaussian kernel is characterized by K(u) [varies] exp(-u/2). The Epanechnikov kernel only uses unrestricted bonds with a propensity score P([X.sub.0j]) that falls within the fixed bandwidth h of P([X.sub.1i]). When [absolute value of P([X.sub.1i]) - P([X.sub.0j])] < h and [absolute value of u] < 1, then the Epanechnikov kernel is K(u) [varies] (1 - [u.sup.2]). Otherwise the kernel has a value of zero. For each i, we compute [Y.sub.1i] - [[bar.Y].sub.0j]. Formally, for each [Y.sub.1i], we match a corresponding [[bar.Y].sub.0j] where

[[bar.Y].sub.0j] = [[summation].sub.j]K(P([X.sub.1i]) - P([X.sub.0j])/h) [Y.sub.0j] / [[summation].sub.j]K (P([X.sub.1i]) - P([X.sub.0j])/h).

and h is a fixed bandwidth. To compute P(X), for each set of a single restricted bond and its corresponding unrestricted bonds, we run a logit or probit model:

RESTRICTED = [[beta].sub.0] + [[beta].sub.cr] CREDIT RATING + [[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3] SOVEREIGN + error.

where RESTRICTED takes a value of one for the restricted bonds, and zero otherwise, CREDIT RATING is the numerical counterpart of the comprehensive credit rating of the bond as shown in Appendix B (e.g., Aaa = 16, Aa1 = 15, Aa2 = 14), MATURITY is the maturity of the bonds in years, and EXCHANGE takes a value of one if the bond is listed on an exchange, and zero otherwise. (10)

The results using the Near Neighbor, Gaussian, and Epanechnikov estimators for the pooled sample of the Resurgent Indian Bond issue, the India Millennium Bond issues, and the comparable benchmark bonds for both issues are summarized in Column 1 of Table III. The yield differential ranges between 145.27-223.15 basis points, which averages to 188.92 basis points (based on an equal-weighting of the estimates in Column 1 of Table III).

Next, we separately examine the estimated yield difference for the Resurgent India Bonds and the India Millennium Bonds using the Near Neighbor, Gaussian and Epanechnikov estimators. The results are summarized in Columns 2 and 3 of Table III. The estimated yield differential for the Resurgent India Bonds ranges from 115.40-201.20 basis points, which averages to 156.48 basis points (based on an equal-weighting of the estimates in Column 2 of Table III). (11) The estimated average yield differential for the India Millennium Bonds (based on an equal-weighting of the estimates in Column 3 of Table III) is higher at 247.95 basis points (range 145.33-334.13 basis points).

Overall, our empirical results in Table III based on the matching procedures described above indicate a yield differential in the range of 145.27-334.13 basis points, which averages to 197.78 basis points (based on an equal-weighting of all the estimates in Table III). We next examine estimates of yield difference using a variety of regression specifications, and compare them with estimates from the matching procedures.

C. Yields Computed by Regression Analysis 

We additionally check for yield differentials by means of a regression equation, an approach often taken in the finance literature. For this purpose we run the following regression (see Section I.A for the definition of the variables):

YIELD SPREAD = [[beta].sub.0] + [[beta].sub.cr] CREDIT RATING + [[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3] SOVEREIGN + [[beta].sub.4] RESTRICTED + error. (1)

Column 1 of Table IV tabulates the regression results for the pooled sample of the Resurgent India Bond issue, India Millennium Bond issue, and the comparable benchmark bonds for both issues. The estimated yield spread for the Resurgent India Bond issue and the India Millennium Bond issue is lower by 170.90 basis points (based on the coefficient estimate of RESTRICTED) than that of comparable benchmark bonds after we control for the comprehensive credit rating, maturity, exchange listing, and whether it is a sovereign bond, and this difference is statistically significant at the 1% level. (12)

We next run the regression in Equation (1) separately for the subsample of the Resurgent India bond issue and the comparable benchmark bonds, and for the subsample of the India Millennium Bond issue and the comparable benchmark bonds. The regression results are tabulated in Columns 2 and 3 of Table IV. The estimated yield differential is 112.82 basis points for the Resurgent India Bonds and 316.14 basis points for the India Millennium Bonds, which averages to 214.48 basis points (based on an equal-weighting of the estimated yield differentials of the Resurgent India Bonds and that of the India Millennium Bonds in Columns 2 and 3 of Table IV).

Overall, the multivariate results in Table 1V indicate an estimated yield differential in the range of 112.82-316.14 basis points, which averages to 214.48 basis points (based on an equal weighting of the estimated yield differentials in Table IV).

In summary, based on our estimates from the univariate results, matching procedures and the multivariate results, the average yield differential is in the range of 184.00-214.48 basis points. To be conservative, we use 150 basis points as our estimated average yield differential for the remaining part of this article. The 150 basis points yield savings translates to a substantial bottomline savings of $1.08 billion. (13)

D. Alternative Explanations

We examine whether the yield differential can be explained by other factors, such as potential inaccuracies of credit ratings, market segmentation, taxes, and commissions.

1. Potential Inaccuracies of Credit Ratings

We analyze the robustness of our results to potential inaccuracies of credit ratings in two different ways. First, we consider two other alternative measures of the bond issuer's credit rating, namely the International Country Risk Guide (ICRG) composite credit rating, and the Institutional Investor (II) country credit rating for the bond issuer. (14) A few differences between these alternative credit rating measures that we use here and the Moody's credit rating is that the alternative measures are numeric (a higher number implies a better credit rating) rather than a letter rating (as in the case of Moody's), and are available as a time series (monthly for ICRG, and bi-annual for II ratings, rather than at specific points of time when Moody's changes a bond's rating). We replace our credit rating dummies in the regression in Column 1 of Table IV with the ICRG composite credit rating for the Resurgent India Bonds, the India Millennium Bonds and the comparable bonds at the time of issuance of the Resurgent India Bonds and India Millennium Bonds. The results are reported in Panel A of Table V. The RESTRICTED variable continues to be negative (with a similar economic magnitude as in Column 1 of Table IV) and is statistically significant at the 1% level. The results are qualitatively similar when we replace the credit rating dummy variables in the regression in Column 1 of Table IV with the II country credit rating for the Resurgent India Bonds, the India Millennium Bonds and the comparable bonds at the time of issuance of the Resurgent India Bonds and India Millennium Bonds (see Panel B of Table V).

Second, if we reestimate the regression in Column 1 of Table IV using Moody's comprehensive credit rating of Ba1 (prior to June 19, 1998) (15) for the RIB issue, the results are qualitatively similar (the RESTRICTED variable has a coefficient of -202.57 basis points and is statistically significant at the 1% level). Moreover, even if we reestimate the regression in Column 1 of Table IV simply using the explicit credit rating (ignoring the credit watch information) of Baa3, the RESTRICTED variable has a coefficient of -95.15 basis points, significant at the 1% level. (16)

The evidence suggests that even after taking into potential inaccuracies of credit ratings as detailed above, the restricted bonds still yield about 100 basis points lower than comparable benchmarks.

2. Market Segmentation

One could argue that the yield differential we document in this article is driven by market segmentation. We analyze the market segmentation explanation in three ways. First, even if bonds are denominated in hard currencies, the inflation rate of the home country may play a role, as local investors in a high inflation economy may aggressively seek investments in hard currencies. In such a situation, a segmentation or demand driven explanation might explain the yield differential, and not ownership restrictions. (17) To test this view, we include the lagged annual inflation rate at the time of the restricted bond issue obtained from the World Economic Indicators database in our regression in Column 1 of Table IV. The results, reported in Column 1 of Table VI are qualitatively similar to those in Column 1 of Table IV.

Second, despite the fact that ownership restrictions typically increase market segmentation of the local capital markets, the market wide segmentation in India may be still less restrictive relative to other nations, e.g., due to India's size. Consequently, the broader notion of economy wide segmentation across nations might still predict that India's bonds yield less than similar bonds in other nations, and the segmentation that arises from ownership restrictions might not be sufficiently large. Given that the India-only sample is too small (see Table II), we include national fixed effects in the regression in Column 1 of Table IV. The results, presented in Column 2 of Table VI are qualitatively unchanged.

Finally, we examine if there is any direct evidence of an increase in market wide segmentation (i.e., a decrease in the market integration of India's capital markets with the rest of the world) in India after the issuance of the restricted bonds. Specifically, we do not see any increase in market segmentation for India during 1998-2000 as evidenced in the graphs on time-varying market integration of India in Bekaert and Harvey (1995), and in Bekaert, Harvey, and Lundblad (2003).

Our evidence shows that even after taking into consideration differences in economy wide segmentation across nations, the restricted bonds still yield about 120 basis points lower than comparable benchmarks.

3. Taxes

One could argue that the motivation for issuing Resurgent India Bonds or India Millennium Bonds is to provide implicit tax benefits (18) to the transferees, e.g., family members of the Nonresident Indians, which will be reflected in the lower yields for Resurgent India Bonds and India Millennium Bonds as compared to benchmark securities. For instance, Resurgent India Bonds may be gifted to a domestic resident, such as a family member in India without any gift tax in India. (19) However, the gift tax was repealed on July 21, 1998 for all Indians (not just Non resident Indians), a few months prior to the issue date of the Resurgent India Bonds. Clearly, this cannot explain the yield difference.

Given that Resurgent India Bonds and the India Millennium Bonds are exempt from Indian taxes for original holders (see Table I), one might argue that the effective yield, on an equivalent pre-tax basis is much higher, perhaps closer to the yield on a comparable benchmark. However, the comparable bonds, which mostly trade in the euro bond market (e.g., London and Luxembourg) enjoy a similar tax benefit due to tax anonymity and absence of withholding taxes. In addition, given that approximately 50% of the Resurgent India Bond inflows and the India Millennium Bond inflows came from the Middle East with zero taxes further negates this hypothesis.

4. Commissions

A collecting bank, to gain a larger share of the Resurgent India Bond issue or the India Millennium Bonds issue may choose to keep only a small part of the commission (e.g., 0.25% to 0.50%) and pass on the rest of the commission to Non-resident Indian investors as an investment incentive. (20) Naturally, this increases the effective yield on the Resurgent India Bonds and the India Millennium Bonds since the price paid is the nominal amount minus the commission kickback. In computing the effective yield, we assume the most aggressive kickback strategy, namely to part with the entire commission (capped at 1.5% in case of Resurgent India Bonds). It turns out that the effective yield, taking into account the commission kickback of 1.5% up front for the Resurgent India Bonds is 8.12% (21), approximately 37 basis points higher than the coupon of 7.75%. However, that still leaves it approximately 113 basis points (= 150-3 7) lower than yields on comparable debt issues. Similar analysis can be conducted for the India Millennium Bonds, and results are qualitatively unchanged. In other words, the commission kickbacks account for only a small part of the difference in yields between Resurgent India Bonds or India Millennium Bonds and their comparable benchmark securities.

In summary, the empirical results suggest that the offering yield on the Resurgent India Bonds and India Millennium Bonds is significantly lower than that of comparable securities, and that this finding is robust to alternative explanations, such as a higher perceived credit rating, different measures of credit rating, market segmentation, taxes, and commissions.

III. Economic Rationale for Ownership Restrictions

In this section, we present the intuition behind why an investor clientele might be willing to pay more for a security than other investors and examine whether there are any benefits to the issuing firm in restricting the offering to the investor clientele.

A. Reasons for a Higher Valuation by an Investor Clientele

One reason for a higher valuation by an investor clientele relative to that of other investors is that it has a broader relationship with the issuer and values the collateral (i.e., what an investor gets in exchange for extinguishing his or her claim) in the potential default states higher than other investors.

In the context of the Resurgent India Bonds and India Millennium Bonds, the investor clientele (Non-resident Indians) face a much lower transaction cost in utilizing the collateral in the local currency (Indian Rupees, a currency that is not freely convertible) than foreign investors as they have a economic need for utilizing rupees. (22) For instance, Non-resident Indians have been known to provide financial assistance for their parents and other family members living in India, buy residential and business properties, operate businesses in India and invest in securities of local companies, and these transactions are facilitated directly or indirectly through the banking network of the State Bank of India (see Appendix A).

An example of a similar difference in the value of the collateral is between a car dealership, such as the BMW or Lexus and a bank. The car dealership is in a position to offer a lower interest rate than a bank on a car lease because it has a ready market for its cars after the lease period in the form of the "certified pre-owned" program. That is, the car dealership incurs a significantly lower transaction cost on utilizing the collateral (i.e., the leased car that is returned to the dealership on expiry of the lease).

B. Benefits in Restricting Securities Offering to an Investor Clientele

A related issue is that even if an investor clientele is willing to pay more for a security than other investors, why would the issuing firm restrict the offering of the securities to the investor clientele? Investors who value the security should self select in buying the security making such restrictions unnecessary.

We present a number of economic reasons why ownership restrictions are potentially beneficial. Generally, these fall under two broad categories: real savings or enhancement of value, and transfer of value (from holders of other securities).

First, ownership restrictions can result in an enhancement of value. The issuing firm incurs lower renegotiation costs in the event of a default when it renegotiates with only one homogenous class of investors (the investor clientele) rather than with multiple classes of investors. Gilson et al. (1990) provide empirical evidence that is consistent with this viewpoint in a study of 169 financially distressed firms during 1978-1987. They show that firms with more layers of creditors are less likely to restructure out of court and save a costly bankruptcy procedure. Second, the ex ante restriction can also serve as a precommitment device to facilitate an efficient ex-post renegotiation in the potential default states with the investor clientele because of the broader relationship that the investor clientele has with the issuer (i.e., dealings that extend beyond its current investment in the issuer's securities). This rationale is somewhat similar to loans made by the mafia, which are invariably repaid as the cost of non-repayment can be very high (e.g., loss of a leg). By taking loans from a clientele that is critical for continued success of a business, such as the mafia (or the Non-resident Indians in our context--See Section III.C below), there is an implicit ex-post commitment to repay the debt which translates into higher prices ex ante. Absent ownership restrictions other investors may wish to free-ride (see Grossman and Hart, 1980) on the investor clientele's ability to obtain a favorable negotiation in the default states. This in turn can lead to multiple classes of investors and dilutes the issuer's incentives to renegotiate efficiently in the potential default states. Ownership restrictions help firms overcome this free-rider problem and make an effective precommittment to renegotiate efficiently in an event of default.

Second, ownership restrictions can result in a transfer of value from unrestricted securities to restricted securities. The reason why this occurs is that ownership restrictions can influence the priority structure of claims (and indirectly affect the value of collateral), as explained above (e.g., in the context of loans from a clientele that is critical for continued success of a business), effectively making the restricted securities implicitly senior to unrestricted securities (see Section IV for some evidence on such wealth transfers). Both these effects (value enhancement and value transfer) imply that the investor clientele bids a higher price (accepts a lower yield) when the offering of securities is restricted to them.

In the context of the Resurgent India Bonds and India Millennium Bonds, the restriction serves as a precommitment device to facilitate a efficient renegotiation for the Non-resident Indians. Given State Bank of India's pivotal role in India's economy (see Appendix A for details), it may be considered "too big to fail" and consequently, one would expect that the Government of India will play a major role in a potential renegotiation. (23) In such a situation, the restriction ensures that the Government of India has the right incentives to renegotiate efficiently with the Non-resident Indians for the fear of losing potentially valuable stable and recurring foreign exchange inflows, jobs in businesses owned by the Non-resident Indians and electoral votes. Such a precommitment to an efficient ex-post renegotiation makes the restricted bonds (i.e., Resurgent India Bonds and India Millennium Bonds) implicitly senior to unrestricted bonds (i.e., foreign currency denominated bonds issued by Indian companies), and potentially result in a wealth transfer from holders of unrestricted bonds to restricted securities. We present some evidence on such wealth transfers in the next section. Overall, the lowering of renegotiation costs and the commitment to renegotiate favorably in the potential default states, ex-post, translates into higher prices (lower yields) ex ante.

C. Are Indians Living Abroad Critical to the Success of India's Economy?

Indians living abroad, also known as the Non-resident Indians play a critical role in ensuring the success of India's economy. According to the Economist Intelligence Unit Country Report for India as of the 4th quarter of 1999 (EIU, 1999), "Non-resident Indians (NRIs) have been the main stay of the Indian balance of payments for the past 25 years. (24) Annual inflows of remittances and investments have helped India through two oil shocks, and shielded the economy from the weaknesses of merchandise trade account." In a recent country report as of March 2005 (EIU, 2005), "Figures released by the Ministry of Finance (of India) show that, of India's external debt of US$113.6bn (in billions) at end-September 2004, US$30.6bn was owed to non-resident Indians (NRIs) ..." making them India's largest creditor, ahead of multilateral debt of US$30.1 billion, the second largest creditor.

Kapur (2004a) suggests that the Non-resident Indians have a significant influence on public policy in India. For example, in 2000 the government of India constituted a group comprising select non-resident Indians to formulate a global strategy for India. This group advised the government on information-technology (IT) policies, telecom infrastructure, guidelines on venture capital funds, and issues related to IT education. On a general note, meetings with the Non-resident Indian community has become socially obligatory for politicians from India, with the Non-resident Indians from the United States being most important in this respect. Some of these meetings resulted in immediate policy changes, such as the introduction of automatic clearing scheme for foreign direct investment proposals, under which a proposal is deemed to be cleared if it did not receive a negative response from the Reserve Bank of India within a short period of time (as opposed to waiting to hear of a positive response), and allowing importation of many capital goods without the requirement of a license from the government. In a related study, Kapur (2004b) documents that the probability that a household in India has members abroad sharply increases with socioeconomic classification, defined in terms of income level and professional background, and that the households in the highest socioeconomic classification--nearly a quarter of households in India--have immediate or extended family members abroad. The extent of influence that the overseas members (i.e., Non-resident Indians) have on the domestic counterparts has in recent times been further amplified by the ease of travel and communication.

Finally, Walton-Roberts (2004) argues that globalization processes have encouraged a renewed national interest with Non-resident Indians as a force to assist India to engage with the global economy. Also, see Appendix A for additional details.

In essence, the Indians living abroad are extremely critical for the success of India's economy, as evidenced in their role in influencing India's balance of payments situation (e.g., in helping maintain adequate foreign exchange reserves and current account surpluses), as India's largest creditor, and in engaging India with the global economy.

IV. Empirical Evidence on Implicit Seniority and Wealth Transfers

In the previous section we hypothesized two reasons for lower yields on restricted securities: enhancement of value and transfer of value. Value enhancement can arise by designing a security to circumvent the deadweight costs of prolonged negotiations, particularly when a security is restricted to a homogenous clientele that values the underlying collateral higher than other investors in the default states. Transfer of value can occur from existing security holders to the new security holders.

We next test for the value transfer effects by employing an event study to determine whether there is a price decline in other foreign currency denominated bonds issued by other Indian companies upon the announcement of the restricted bonds. Specifically, we test whether the implicit seniority, as evidenced by a price decline in other foreign currency denominated securities issued by Indian companies fully explains the difference in yields between the Non-resident Bonds and the comparable benchmarks documented in Section II. This takes a more general view that the restricted bonds may be implicitly senior to the foreign currency denominated unrestricted bonds of all Indian companies (and not just the State bank of India) as one may expect that the Government of India plays an important role in a renegotiation in the potential default states since the State Bank of India may be considered "too big to fail".

A. Evidence of Implicit Seniority

The required data for the event study is obtained from Datastream. Our sample for the event study consists of 32 foreign currency denominated debt securities (e.g., US dollar denominated bonds, non-US dollar foreign currency bonds, fixed-rate non-convertible bonds, and fixed-rate convertible bonds etc.) issued by Indian companies listed on an exchange for which daily price data (to compute the daily returns) was available on Datastream. We use market adjusted returns, where the market index is the J.P. Morgan Emerging Markets Index (JPMEMBI), also obtained from Datastream. Following Brown and Warner (1985), we use the following event study methodology:

(2) [A.sub.i,t] = [R.sub.i,t] - [R.sub.m,t],

where the event date, i.e., day 0 is defined as announcement dates of June 2, 1998 for Resurgent India Bonds and October 9, 2000 for India Millennium Bonds. The announcement dates were obtained from Dow Jones Newswires, being the first date of a credible mention of these bond issues during time period that is one year prior to the issue date of these bonds. [R.sub.i,t] is the observed arithmetic return for security i at date t, [R.sub.m,t] is the observed arithmetic return for the market index at date t, and [A.sub.i,t] is the excess return for security i at day t.

The market-adjusted average abnormal return on the event day is -1.40%, statistically significant at the 1% level (See Column 1 of Table VII). We next examine whether the negative price reaction is pervasive across both the restricted bonds.

Columns 2 and 3 of Table VII presents the announcement effect segmented by the type of the Non-resident Indian bond, namely for the Resurgent India Bonds and India Millennium Bonds separately. The market-adjusted average abnormal return on the event day is -1.59% and -0.99% for the announcement of the Resurgent India Bonds and India Millennium Bonds respectively, each of which is statistically significant at the 5% level. In other words, there is a similar negative price reaction for each of these restricted bonds, albeit the magnitudes of these price reactions are different.

Overall, the evidence of wealth transfers from unrestricted bonds to restricted bonds presented here is consistent with an implicit seniority of the restricted bonds vis-a-vis the unrestricted bonds. However, one needs to be careful here because such an effect might also come from restricted debt capacity of the issuer or of the sovereign nation, so that any new issue of debt securities leads to a decline in the pricing of existing debt securities. We turn our attention to this issue next.

B. Comparison Against Other Debt Issuance Effects

We presented evidence of a wealth transfer from unrestricted bonds to restricted bonds in Section IV.A. We examine the robustness of this result by investigating whether it is driven by the debt capacity constraint of the issuer or more generally of the sovereign nation (India). Specifically, if the overall debt capacity of the issuer and of the sovereign nation is constant (e.g., in the short-term), any new debt issue (whether restricted or unrestricted) would have negative implications for existing debt. Consequently, if we do not see a difference in the price reaction on existing unrestricted bonds upon announcement of issuance of new restricted bonds versus issuance of new unrestricted bonds, then we cannot attribute a wealth transfer due to ownership restrictions.

If we assume a constant debt capacity for India, at least in the short-term, one could argue that there is likely a negative price reaction (similar to the one documented above) for all foreign currency denominated bond issues by Indian issuers. If so, one could infer evidence consistent with implicit seniority only if the price decline documented in Section IV.A is higher than what one would expect for any foreign currency (unrestricted) bond issue. We conduct an event study (similar to the one in Section IV.A) to measure the price reaction associated with announcement of foreign currency bonds of Indian issuers on outstanding foreign currency bonds of other Indian issuers. As before, the announcement dates correspond to the first credible mention about the debt issue in Dow Jones Newswires during a time period that is one year prior to the issue date. The market-adjusted abnormal return, is significantly less (-0.33% versus -1.40%), suggesting that there is a transfer of wealth from the existing unrestricted bonds to new restricted bonds in excess of the effect associated with a new unrestricted bond issue.

Consequently, we view that our evidence of a significantly higher price decline for the issuance of new restricted bonds vis-a-vis new unrestricted bonds on existing unrestricted bonds, namely a transfer of wealth to restricted bonds from holders of unrestricted bonds, is consistent with implicit seniority of the restricted bonds.

C. Yield Differentials and Implicit Seniority

We next examine whether the implicit seniority, as evidenced by a price decline in other foreign currency denominated securities issued by Indian companies fully explains the difference in yields between the Non-resident Bonds and comparable benchmarks.

To estimate the value loss from implicit seniority of the Non-resident Indian Bonds, we multiply the average price decline on the event day (e.g., in Column 1 of Table VII) with the amount of foreign currency denominated securities issued by Indian companies that are outstanding on the event date. We find that the value loss on the event day is $127.07 million (corresponding to Column 1 of Table VII), significantly lower than the present value of interest savings of $722.07 million pertaining to the yield differential. (25) Note that this is gross value loss based on the negative announcement effect of -1.40% that we document in Section IV.A. If we were to compute the net value loss (netting off the -0.33% negative announcement effect that we document in Section IV.B for new unrestricted bonds) this is lower at $97.12 million.

D. Robustness Test

The above evidence does not control for the size of each new bond issue, as well as the influence of the limited number of bond issues on the standard errors for the announcement returns associated with them. In other words, if the results are driven by finite issuance (i.e., constant debt capacity for India), larger new bond issues (restricted or unrestricted) will have larger negative impact on bond yields, and it is plausible that the inferior (i.e., more adverse) announcement returns associated with restricted new bond issues arises simply because they are larger than unrestricted issues. To test for this, we regress the abnormal announcement return of the unrestricted bonds in response to the announcement of a new bond issue on: a) the size of the each new bond issue, b) the size of the unrestricted bond whose abnormal bond price reactions is being measured, c) an indicator variable RESTRICTED for whether the new bond issue is restricted or not, and d) fixed effects for the new bond issues. The results, reported in Table VIII are stronger once we control for these additional effects the indicator variable RESTRICTED is negative and statistically significant at the 1% level, and the explanatory power of this regression is relatively high. In other words, after we control for the size of the new bond issue, and include fixed effects for each new bond issue, there is strong evidence of a transfer of value from the holders of other unrestricted bonds to the holders of restricted bonds, consistent with an implicit seniority of the restricted bonds vis-a-vis the unrestricted bonds. Nevertheless, the value loss from the implicit seniority of the restricted bonds is still substantially lower than the present value of interest savings from the yield differential of the restricted bonds.

Based on the above, we conclude that the implicit seniority of the Non-resident Indian bonds alone accounts for some but not all the difference in yields between the Non-resident Indian Bonds and comparable benchmarks. (26) This suggests that in addition to the transfer of value, there is potentially value enhancement from some of the factors we discussed such as real savings from lower renegotiation costs.

V. Conclusions

In this article, we examine the role of ownership restrictions in raising capital from niche clienteles. We argue that ownership restrictions must be viewed broadly in the context of security design since they have direct implications for the renegotiation costs and the value of collateral. We show that restricting ownership of securities can be beneficial in lowering (rather than raising) the cost of capital of an issuing firm.

We argue that ownership restrictions may enhance value as well as transfer value from holders of other securities. Ownership restrictions can enhance value by circumventing the deadweight costs of prolonged negotiations, particularly when a security is restricted to a homogenous clientele that values the underlying collateral higher than other investors. Restricting the ownership to a homogenous class of investors lowers the renegotiation costs and serves as a precommitment to ensuring an efficient ex-post renegotiation in the potential default states, resulting in a lower ex ante offering yield (and a higher offer price). This also results in an implicit seniority of holders of these restricted bonds vis-a-vis holders of unrestricted bonds. We empirically test and find support for both value enhancement and value transfer from ownership restrictions.

Many other emerging market firms can also benefit from designing securities with ownership restrictions where the new securities are offered only to investors who value them the most. For example, firms from emerging markets with large expatriate populations, such as China, Israel and Korea could bypass traditional underwriting route and devise securities that could get them higher prices (and lower yields). Similarly, for many other firms, e.g., hi-tech firms, where customers are also investors, and hence have a broader relationship with the firm, there can be benefits to ownership restrictions. Overall, our analysis shows the wider ramifications of ownership restrictions than suggested in the literature, as an integral part of security design, in lowering (rather than raising) the issuer's cost of capital.

Appendix A. Brief Overview of India, State Bank of India, and the Non-Resident Indians

This appendix outlines a brief overview of India, State Bank of India's history, its role in India's economy, explains the characteristics of the Non-Resident Indians, and their role in India's economy.

A. Brief Overview of India

India is the largest democratic nation in the world and the second largest country in terms of population (next only to China) with approximately one billion people. India gained its independence from the British in 1947. For a large part of the post-independence period, India has been governed by the Congress party. Principal among India's leaders were Jawahar Lal Nehru, his daughter, Indira Gandhi, and her son, Rajiv Gandhi. Economic reforms since 1991 has led to a strong economic growth of the Indian economy. The result of these reforms has been moderate rates of inflation, an increase in the foreign exchange reserves, a modest balance of payments deficit, higher investment flows and a higher growth in trade.

Most of India is in the villages and the urban population accounts for only a fourth of the total population. More than a third of the population still lives below the poverty line and nearly half of the population is illiterate. However, India has a burgeoning middle class whose population equals that of the European community and is the target market of the manufacturers of many international consumer brands. While the fundamentals of India's economy have been moving in the right direction, the economy is still hampered by inadequate infrastructure, high interest rates, and a large fiscal deficit.

B. State Bank of India's history and Its Role in India's Economy

The origins of the State Bank of India dates to 1806 when the Bank of Calcutta (later called the Bank of Bengal) was established. In 1921, the Bank of Bengal and two other banks (Bank of Madras and Bank of Bombay) were amalgamated to form the Imperial Bank of India. In 1955, the controlling interests of the Imperial Bank of India were acquired by the Reserve Bank of India (the equivalent of the Federal Reserve Bank in the United States), and the State Bank of India was created by an act of the Parliament to succeed the Imperial Bank of India. The Reserve Bank of India, which has supervisory control over the banking system in India, is also the single largest shareholder of State Bank of India (with 59.73% share holding). Consequently, some market participants tend to view State Bank of India as quasi-sovereign risk.

State Bank of India is the largest bank in India in terms of profits, assets, deposits, branches and employees. As of March 31, 1998, State Bank of India possessed total assets worth USS 45,487 million and total deposits worth USS 33,188 million. With a network of 8,895 branches in India and 52 foreign offices in 34 countries, State Bank of India commands about one-fifth of the total deposits and loans in the country. State Bank of India's shares and bonds are listed for trading on all the major Indian stock exchanges and State Bank of India has one of the largest market capitalizations of all companies traded on such exchanges.

State Bank of India plays a key role in India's economy through its product set that covers consumer lending, working capital financing, infrastructure lending, merchant banking, government securities dealing, and priority lending (e.g., agriculture and small business sectors of the economy). In addition, due to State Bank of India's extensive branch network, almost every villager in India has an account with State Bank of India. In other words, approximately three-fourths of India's population which lives in the villages relies extensively on State Bank of India for their banking services.

State Bank of India is an active player in the Non-resident Indian market through its short to medium term foreign currency deposits (which range from three months to three years) and accounts, denominated either in foreign currency or in Indian Rupees. Long considered the flagship of Indian banking, State Bank of India is also the privileged financial intermediary of the Indian government in raising sovereign debt.

C. Role of Non-Resident Indians in India's Economy

There are estimated to be approximately 9.6 million Non-resident Indians in the world. Non-resident Indians have certain distinguishing characteristics. First, many of the Nonresident Indians provide financial assistance to their parents, siblings and relatives living in India. Second, Non-resident Indians intend to return to India in the future, e.g., if there is a crisis in the foreign country, such as the invasion of Kuwait by Iraq in early 1990s, or after their children go to college, or after retirement. Finally, driven by their experiences abroad, Non-resident Indians desire to make fundamental changes in India's economy. To meet the above mentioned objectives, Non-resident Indians have been known to send money to their parents, siblings and relatives living in India, buy property in India, own businesses in India, own stocks and bonds in local companies in India, and contribute significant amounts to charitable organizations in India.

Non-resident Indians play a critical role in India's economy in several ways. First, as a provider of recurring and stable foreign currency cash inflows. Appendix D presents the outstanding balances under various Non-resident Indian deposits and accounts as of March 31,2001. Interestingly, a non-trivial amount of the traditional Non-resident Indian net capital flows have been in the form of non-repatriable rupee accounts during the five years preceding the Resurgent India Bond issue in 1998. This suggests that Non-resident Indians, unlike foreign investors (who are unlikely to hold rupees given the limited convertibility of the Indian rupee into a foreign currency), have an economic need to utilize rupees for the reasons mentioned earlier, such as providing financial assistance to their parents, siblings and relatives living in India, buying property in India consistent with their future plans to return to India, owning businesses in India, and contributing significant amounts to charitable organizations in India. Second, by providing employment through owning businesses in India. For example, several Indian entrepreneurs from silicon valley have set up software enterprises in cities, such as Bangalore and Hyderabad. Finally, through family, relatives and affiliations with political organizations control a non-trivial proportion of electoral votes.

Appendix B. Description of Moody's Credit Ratings

This appendix presents a brief description of Moody's credit ratings,
and the corresponding numeric rating created by the authors.

Moody's Credit
Rating            Numeric Rating            Brief Description

Investment Grade--High Creditworthiness

Aaa                   16            Gilt edge, prime, maximum safety
Aa3-Aa1               13-15         Very high grade, high quality
A3-A1                 10-12         Upper medium grade
Baa3-Baa1             7-9           Lower medium grade

Distinctly Speculative--Low Creditworthiness

Ba3-Ba1               4-6           Low grade, speculative
B3-B1                 1-3           Highly Speculative

Predominantly Speculative--Substantial Risk or in Default

Caa                   0             Substantial risk, in poor standing
Ca                    0             May be in default, extremely
                                      speculative
C                     0             Even more speculative than those
                                      above

Source: Wilson, R.S. and Fabozzi, F.J., 1990, The New Corporate Bond
Market, Chicago, IL, Probus Publishing Company. Numeric rating is
created by the authors.

Appendix C. List of Matched Bonds

This appendix provides the list of matched bonds based on the Near
Neighbor (N=10) matching method for the Resurgent India Bonds as
reported in Column 2 of Table III with the propensity score estimated
based on a probit model. We report the nation of the issuer, yield
spread (in basis points), numeric credit rating (based on the numeric
scale in Appendix B), maturity (in years), whether exchange listed (1
if exchange listed and 0 otherwise), and whether sovereign-grade (1 if
sovereign grade and 0 otherwise). The average yield savings is 201.20
basis points reported in Column 2 of Table III which equals the average
yield spread of 432.20 basis points (see below) minus the 231 basis
points yield spread of Resurgent India Bonds.

Issuer's                                        Maturity in
Nation         Yield Spread    Credit Rating       Years

Argentina         425.00           4.00            5.00
Argentina         281.00           4.00            5.00
Argentina         253.00           4.00            8.00
Argentina         236.00           4.00            5.00
Mexico            256.00           5.00            6.00
South Korea       643.00           5.00            5.00
South Korea       563.00           5.00            4.00
South Korea       585.00           5.00            2.00
South Korea       537.00           5.00            2.00
South Korea       543.00           5.00            3.00

Average           432.20           4.60            4.50

Issuer's       Exchange
Nation          Listed     Sovereign

Argentina        0.00        1.00
Argentina        0.00        0.00
Argentina        0.00        0.00
Argentina        0.00        0.00
Mexico           0.00        1.00
South Korea      0.00        0.00
South Korea      0.00        0.00
South Korea      0.00        0.00
South Korea      0.00        0.00
South Korea      0.00        0.00

Average          0.00        0.20

Appendix D. Outstanding Balances under Various Non-Resident
Indian Deposit Schemes during 1996-01

This appendix presents the outstanding balances in millions of US
dollars as at the end of March under various Non-Resident Indian
deposit schemes during 1996-01, namely, Foreign Currency Non-Resident
(FCNR) deposits, Non-Resident External (NRE) Rupee Accounts, and
Non-Resident Non-Repatriable (NRNR) Rupee Deposits. The FCNR deposits
are held in foreign currency and are also repaid in foreign currency.
The NRE accounts are held in Indian rupees, converted at the current
spot exchange rate and are fully repatriable at the future spot
exchange rate (at the time of withdrawal). The NRNR accounts are held
in Indian rupees, converted at the current spot exchange rate and are
not repatriable in foreign currency, i.e., repayment is only in Indian
rupees.

Category                        4/95-3/96    4196-3/97    4/97-3/98

Foreign Currency
Non-Resident (FCNR) Deposits       9975         9802         8468
Non-Resident External
(NRE) Rupee Accounts               3916         4983         5637
Non-Resident Non-Repatriable
(NRNR) Rupee Deposits              3542         5604         6262

Total Amount                      17433        20389        20367

Category                        4/98-3/99    4/99-3/00    4/00-3/01

Foreign Currency
Non-Resident (FCNR) Deposits       7835         8172         9076
Non-Resident External
(NRE) Rupee Accounts               6045         6758         7147
Non-Resident Non-Repatriable
(NRNR) Rupee Deposits              6618         6754         6849

Total Amount                      20498        21684        23072

Source: Reserve Bank of India (RBI) Bulletin, Trade and Balance of
Payment, 2002.

We thank the seminar participants at the American Finance Association (AFA) annual meeting in Washington, D.C., the Georgia Tech/Fortis International Finance Conference in Atlanta, the Global Corporate Governance Forum's regional workshop in Hyderabad, India, the Financial Management Association (FMA) annual meeting in Denver, and at Vanderbilt University for helpful comments. We thank an anonymous referee, the Editors (Lemma Senbet and Alex Triantis), Geert Bekaert, Peter DeMarzo, Vikram Jaipuria (Citibank Non-resident Indian services), Roger Huang, Ken Singleton, and particularly Jeremy Stein .for helpful comments. We also thank Nilesh Jain of Lexicon Finance Limited, India for providing useful data and Alex Franta for his research assistance. A number of students provided data and related inputs that helped improve this article.

References

Allen, F. and D. Gale, 1988, "Optimal Security Design," Review of Financial Studies 1,229-263.

Bailey, W., Chung, EY., and J. Kang, 1999, "Foreign Ownership Restrictions and Equity Price Premiums: What Drives the Demand for Cross-Border Investments?" Journal of Financial and Quantitative Analysis 34, 489-511.

Berger, A.N. and G.F. Udell, 1990, "Collateral, Loan Quality and Bank Risk," Journal of Monetary Economics 25, 21-42.

Bekaert, G. and C.R. Harvey, 1995, "Time-Varying World Market Integration," Journal of Finance 50, 403-444.

Bekaert, G. and C.R. Harvey, 2003, "Emerging Markets Finance," Journal of Empirical Finance 10, 3-55.

Bekaert, G., Harvey, C.R., and C.T. Lundblad, 2003, "Equity Market Liberalization in Emerging Markets," Journal of Financial Research 26, 275-299.

Bensanko, D. and A. Thakor, 1987, "Competitive Equilibrium in the Credit Market under Asymmetric Information," Journal of Economic Theory 42, 167-182.

Boot, W.A., Milbourn, T.T., and A. Schmeits, 2005, "Credit Ratings as Coordination Mechanisms," Review of Financial Studies (Forthcoming).

Boot, W.A. and A.V. Thakor, 1993, "Security Design," Journal of Finance 48, 1349-1378.

Boot, W.A., Thakor, A.V., and G.F. Udell, 1991, "Secured Lending and Default Risk: Equilibrium Analysis, Policy Implications, and Empirical Results," Economic Journal I 01,458-472.

Brown, S.J. and J.B. Warner, 1985, "Using Daily Stock Returns: The Case of Event Studies," Journal of Financial Economics 14, 3-31.

Chan, Y. and G. Kanatas, 1985, "Asymmetric Valuation and the Role of Collateral in Loan Agreements," Journal of Money, Credit and Banking 17, 85-95.

Chan, Y. and A.V. Thakor, 1987, "Collateral and Competitive Equilibria with Moral Hazard and Private Information," Journal of Finance 42, 345-364.

Dallas, G., 1997, "The Role of Ratings in the Public and Private Debt Markets," Standard & Poor's Document.

Domowitz, I., Glen, J., and A. Madhavan, 1997, "Market Segmentation and Stock Prices: Evidence from an Emerging Market," Journal of Finance 52, 1059-1085.

Ederington, L.H., Yawitz, J.B., and B.E. Roberts, 1987, "The Information Content of Bond Ratings," Journal of Financial Research 10, 211-226.

Erb, C.B., Harvey, C.R., and T.E. Viskanta, 1996, "The Influence of Political, Economic, and Financial Risk of Expected Fixed-income Returns," Journal of Fixed Income 6, 7-30.

Erb, C.B., Harvey, C.R. and T.E. Viskanta, 2000, "Understanding Emerging Market Bonds," Emerging Markets Quarterly 4, 7-24.

EIU, 2005, "Country Report--India as of March 2005," The Economist Intelligence Unit, London, UK.

EIU, 1999, "Country Report--India as of December 1999," The Economist Intelligence Unit, London, UK.

Flannery, M.J., 1986, "Asymmetric Information and Risky Debt Maturity Choice," Journal of Finance 41, 19-38.

Gande, A. and D.P. Parsley, 2005, "News Spillovers in the Sovereign Debt Market," Journal of Financial Economics 75, 691-734.

Gilson, S.C., John, K., and L. Lang, 1990, "Troubled Debt Restructurings: An Empirical Study of Private Reorganization of Firms in Default," Journal of Financial Economics 27, 315-353.

Grossman, S. and O. Hart, 1980, "Takeover Bids, the Free-rider Problem, and the Theory of the Corporation," Bell Journal of Economics 11,42-64.

Heckman, J.H., Ichimura, H., and P.E. Todd, 1997, "Matching as an Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Programme," Review of Economic Studies 64, 605-654.

Heckman, J.H., Ichimura, H., and P.E. Todd, 1998, "Matching as an Econometric Evaluation Estimator," Review of Economic Studies 65, 261-294.

Heckman, J. and R. Robb, 1986, "Alternative Method for Solving the Problem of Selection Bias in Evaluating the Impact of Treatments on Outcomes," in H. Wainer, Ed., Drawing Inferences from Self-Selected Samples, New York, NY, Springer-Verlag.

John, K., Lynch, A.W., and M. Puri, 2003, "Credit Ratings, Collateral, and Loan Characteristics: Implications for Yield," Journal of Business 76, 371-409.

Kapur, D., 2004a, "Ideas and Economic Reforms in India: The Role of International Migration and the Indian Diaspora," India Review 3, 364-384.

Kapur, D., 2004b, "Diasporas and Development: The Impact of Migration from India to India," Harvard University Mimeo.

Mayers, D. and C.W. Smith, 1987, "Death and Taxes: The Market for Flower Bonds," Journal of Finance 42, 685-698.

Rosenbaum, P. and D. Rubin, 1983, "The Central Role of the Propensity Score in Observational Studies for Causal Effects," Biometrika 70, 41-55.

Stulz, R.M. and W. Wasserfallen, 1995, "Foreign Equity Investment Restrictions, Capital Flight, and Shareholder Wealth Maximization: Theory and Evidence," Review of Financial Studies 8, 1019-1057.

Walton-Roberts, M., 2004, "Globalization, National Autonomy and Non-resident Indians," Contemporary South Asia 13, 53-69.

White, H., 1980, "A Heteroskedasticity-consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity," Econometriea 48, 817-838.

(1) For example, Boot and Thakor (1993) show that the revenue maximizing strategy for a firm selling securities in an asymmetric information environment is to split the claims on the cash flow from the asset into an information-sensitive security that promotes informed trading and a second claim that is less sensitive to information. Allen and Gale (1988) incorporate transaction costs of issuing securities to investors facing short-sales constraints. They examine the security design problem from the viewpoint of providing optimal risk-sharing among investors, and show that the issuing firm's income stream should be split so that in every state all payoffs are allocated to the security held by the group that values it most.

(2) This view is consistent with the empirical evidence on foreign equity ownership restrictions, namely, economically significant stock price premia documented for unrestricted shares (i.e., those open for investment to all investors) relative to restricted shares (i.e., those open for investment only to Mexican investors, as in the case of Domowitz, Glen, and Madhavan, 1997) of the same company. Also, see Stulz-Wasserfallen (1995) for similar evidence from Switzerland, and Bailey, Chung, and Kang (1999) for similar evidence for a larger sample of 11 countries with foreign ownership restrictions.

(3) There is a large literature on collateral. For example, Berger and Udell (1990) document that collateral plays an important role in more than two-thirds of commercial and industrial loans in the United States. Boot, Thakor, and Udell (1991) endogenize the use of outside collateral in a model with moral hazard and adverse selection. John, Lynch, and Puri (2003) study how collateral affects bond yields, and Bensanko and Thakor (1987), Chan and Kanatas (1985), and Chan and Thakor (1987) elaborate on the role of collateral in the presence of asymmetric information. Our article differs from these articles in that we consider the effect of differential valuation of collateral by different investor clienteles on security design.

(4) A non-niche clientele investor may hold a security for reasons such as portfolio diversification, higher perceived credit quality etc.

(5) Specifically, we use country credit ratings of Institutional Investor Service, International Country Risk Guide, and Moody's Investor Services in our analysis. Our results are qualitatively similar (see Section II.D.1 for details). For an extensive list of articles that examine the relationship between emerging market spreads and credit ratings, see http://www.duke.edu/~charvey/research.htm, for example, Erb, Harvey, and Viskanta (1996, 2000). Also see Bekaert and Harvey (2003) for an excellent survey of emerging market finance.

(6) We see no significant increase in market segmentation for India during the sample period. See Bekaert and Harvey (1995) and Bekaert, Harvey, and Lundblad (2003) for details.

(7) See Ederington, Yawitz, and Roberts (1987) for the details of the bond rating process and Dallas (1997) for the role of bond ratings in debt markets.

(8) See Appendix B for a brief description of Moody's credit ratings. Quite frequently, rating agencies release information about the credit outlook of the issuer, especially the possibility of a potential upgrade or downgrade, in addition to the letter rating (e.g., Aaa, A3 etc.). Markets typically factor in such information in determining the price and yield of a debt issue. We construct a comprehensive credit rating measure (see Gande and Parsley (2005) for details) to capture such information as follows: if a rating agency places an issuer under review for a potential upgrade (downgrade) or if the associated outlook for an issuer is positive (negative), we increase (decrease) the credit rating by a notch. For example, while Colombia's sovereign credit rating at the time the Resurgent India Bonds were issued (August 1998) was Baa3, Moody's had a negative outlook on Colombia for a possible downward revision since March 1998. Consequently, the comprehensive credit rating for Colombian sovereign debt is Ba1 instead of Baa3. However, our results are qualitatively unchanged when we use Moody's explicit credit rating (not reported in the article).

(9) An article in the Financial Express, a prominent Indian financial newspaper (September 2, 1998, Banking section) quotes a J. P. Morgan's report: "... the confidence expressed by expatriate Indians is encouraging and displays the fact that the cost at which funds have been raised imply a perceived sovereign rating three or four notches higher than the current levels. For comparison sake, China's 2003 yankee bond issue rated A3 by Moody's trades at a spread of 280 basis points over the 10-year US Treasury implying a dollar yield of 7.90% and the Resurgent India Bond issue is not even sovereign risk."

(10) The results are qualitatively similar when we replace the numerical counterpart of the credit rating with a set of credit rating dummy variables (as in Table IV), and additionally replace MATURITY with its logarithmic term, LN(MATURITY) to control for potential non-linearities.

(11) For illustration purposes, as suggested by a referee, we report in Appendix C the list of matched bonds for the Resurgent India Bond issue using the Near Neighbor (n=10) matching method where propensity scores are estimated through a probit model. The list of matched bonds for the India Millennium Bond issue (not reported here) is qualitatively similar.

(12) The results are qualitatively unchanged when we replace maturity with its logarithmic term, LN(MATURITY) to control for potential non-linearities in Column 1 of Table IV.

(13) The estimated yield savings from the 150 basis points yield differential based on semi-annual compounding is 4,200[[1+(7.75%+1.50%)/2].sup.10] - 4,200[[1+(7.75%/2)].sup.10] = $458 million for Resurgent India Bonds, and 5,500[[1+(8.5%+1.50%)/2].sup.10] - 5,500[1+(8.5%/2)].sup.10] = $620 million for India Millennium Bonds. Together, this represents a saving of $1.08 billion.

(14) The relationship between emerging market spreads and credit ratings (such as the ICRG and II ratings) is well-established--see http://www.duke.edu/~charvey/research.htm. We thank Cam Harvey for providing us the Institutional Investor country credit ratings data. The source for the International Country Risk Guide (ICRG) composite credit ratings is the Political Risk Services Group (http://www.prsgroup.com/icrg/icrg.html).

(15) On June 19, 1998 Moody's lowered India's sovereign credit rating to Ba2 and State Bank of India's long-term foreign currency rating to Ba3 on concern over escalating tensions with its neighbor, Pakistan. The previous rating activity was on January 9, 1998 when Moody's placed India's sovereign credit rating and State Bank of India's long-term foreign currency rating of Baa3 under review for a possible downgrade. Hence, as described in footnote 8, the explicit credit rating is Baa3 and the comprehensive credit rating is Bal prior to June 19, 1998.

(16) In fact, Boot, Milbourn, Schmeits (2005) argue that credit rating agencies provide credible monitoring services through the credit watch procedures. That is, the credit watch procedure allows for an implicit contract between the firm and the credit rating agency where the former implicitly promises to undertake specific actions ("recovery effort") to mitigate the possible deterioration of its credit standing.

(17) We thank a referee for drawing our attention to the inflation issue and its link to market segmentation.

(18) Mayers and Smith (1987) document an instance of certain US government securities, known as flower bonds that trade at a substantial premium (i.e., lower yield) than comparable non-flower bonds. These flower bonds, issued prior to 1966, could be redeemed at par plus accrued interest for the purpose of paying estate taxes, if held at the time of death. Such a flower bond, which may be viewed as a straight bond plus an insurance policy, will consequently trade at a lower yield reflecting the embedded insurance policy.

(19) Since domestic residents are not allowed to hold foreign exchange as per the exchange control regulations of the Reserve Bank of India, a domestic resident who receives the Resurgent India Bonds or the India Millennium Bonds as a gift gets an equivalent amount in Indian Rupees converted at the prevailing spot exchange rate.

(20) Such behavior of passing on a part of the commission to investors as an investment incentive is not unique to these bonds. Rather, it is a fairly common practice in some other markets, such as the firm-commitment underwritten debt issues in the euro markets.

(21) The commission kickback of 1.5% is equivalent to a Non-resident Indian paying 98.5% of par value for the bond instead of 100% of par value. The annualized yield that equates 7.75% coupon paid semi-annually in exchange for 98.5% of the par value is 8.12%. In other words, y = 8.12% in the following equation:

98.5 = 3.875/(1+y/2) + 3.875/[(1+y/2).sup.2] + ... + 3.875/[(1+y/2).sup.9] + 103.875/[(1+y/2).sup.10].

(22) In an analogous fashion, an investor clientele may have a higher perceived credit rating (i.e., lower probability of default). See Section II.D.1 for details.

(23) In the US, Continental Illinois (deposits of $29 billion and assets of $40 billion, as reported in the Federal Reserve Bank of Chicago's website), which has a slightly smaller asset base than State Bank of India, and certainly much smaller relative to the US economy than State Bank of India is to the Indian economy, was bailed out in 1984 as it was considered "too big to fail." When placed in this context, State Bank of India, the largest bank in India, and of greater relative importance to its national economy, is likely to be considered "too big to fail" by the Government of India.

(24) The Economist Intelligence Unit, which is part of the Economist business group (that includes the well-known Economist magazine), is a leading provider of country, industry and management analysis for the past 60 years. Their extensive international reach and unfettered independence makes them one of the most trusted and valuable resource for international companies, financial institutions, universities and government agencies. See http://db.eiu.com for additional details.

(25) Using Footnote 13, the present value of interest savings from Resurgent India Bonds is: 458/[(1+(7.75%/2).sup.10]=$313.16 million. Similarly, the present value of interest savings from India Millennium Bonds is: 620/[(1+8.5%/2).sup.10]=$408.91, which totals to $722.07 million.

(26) Our conclusion holds even if we use a lower yield savings to account for alternative explanations in Section II.D. For example, the present value of yield savings (estimated from footnotes 13 and 25 using 100 basis points) is significantly higher than the $127.07 million corresponding to Column 1 of Table VII.

Amar Gande and Manju Puri *

* Amar Gande is an Assistant Professor of Finance at Vanderbilt University in Nashville TN. Manju Puri is an Associate Professor of Finance at Duke University and NBER, in Durham, NC.

Table I. Salient features of Resurgent India Bonds (RIBs) and India
Millennium Bonds (IMBs)

This table presents the salient features of Resurgent India Bonds
(RIBs) and India Millennium Bonds (IMBs).

Issue Opened On           August 5, 1998 (RIBs); October 21, 2000
                            (IMBs)
Maturity                  Five years from the date of issue
Coupon rate               7.75% ($), 8.00% ([pounds sterling]), 6.25%
                          (DM) on an annualized basis for RIBs and
                          8.50% ($), 7.85% ([pounds sterling]) and
                          6.85% (Euro). Investors can receive interest
                          either half-yearly or on a cumulative basis
                          at maturity.
Repatriability            The principal amount as well as the interest
                          earned on Resurgent India Bonds and India
                          Millennium Bonds is fully repatriable in the
                          currency of denomination for all
                          Non-resident holders. Premature encashment is
                          permitted without any penalty after six
                          months from the date of issue on a non-
                          repatriable basis in Indian Rupees.
Loans against Bonds       The Resurgent India Bonds and India
                          Millennium Bonds can be presented as
                          collateral for bank loans up to 90% of the
                          invested amount, either in Indian Rupees or
                          in foreign currency. The interest rate on
                          such loans is at the discretion of individual
                          banks. However, loans in foreign currency
                          require approval of the Reserve Bank of
                          India.
Holding/Transferability   The Resurgent India Bonds and India
                          Millennium Bonds can be held jointly by
                          Non-Resident Indians with resident Indians on
                          a "Former or Survivor Basis." The Resurgent
                          India Bonds and India Millennium Bonds are
                          freely transferable between Non-resident
                          Indians and/or Overseas Corporate Bodies
                          (owned directly or indirectly to the extent
                          of at least 60% by Non-resident Indians).
Tax Benefits              The interest earned on Resurgent India Bonds
                          and India Millennium Bonds is not subject
                          to any withholding tax or income-tax in
                          India. The Resurgent India Bonds and India
                          Millennium Bonds are free from wealth-tax
                          liability in India and would not attract any
                          gift-tax in India either in the hands of the
                          donors or the donees. All of these tax
                          benefits will continue to apply in the hands
                          of the transferees at least in the case of
                          the first transfer.
Issue Closed On           August 24, 1998 (RIBs); November 20, 2000
                          (IMBs).

Source: Dow Jones Newswires.

Table II. Average Yield Spread of Comparable Issues

This table presents the average yield spread (in basis points) of
comparable fixed-rate US dollar denominated issues from emerging
markets. Yield spread is measured as the ex ante yield of a debt
security minus the ex ante yield of a US Treasury security of
comparable maturity. Panel A presents the average yield spread in
basis points classified by the credit rating. Panel B presents the
average yield spread, average credit rating (based on the numeric
scale in Appendix B), average maturity (in years), fraction of issues
that are exchange listed, and fraction of issues that are
sovereign-grade (these variables are used in regressions in
subsequent tables) classified by country.

Panel A. By Credit Rating

Credit Rating   # Issues    Yield Spread

B3                  2           771
B2                  5           618
B1                 52           458
Ba3                21           440
Ba2                46           490
Ba1                29           547
Baa3               46           333
Baa2               28           231
Baa1               10           384
A3                 26           198
A2                  3           171
Aa3                 1           199

Total             269           400

Panel B. By Nation

                                 Average       Average
                                  Yield         Credit
Nation            # Issues        Spread        Rating

Argentina            22            391           3.55
Brazil               24            456           3.00
Chile                 3            350           9.00
China                17            234          10.00
Colombia             13            613           5.77
El Salvador           1            314           7.00
Hungary              12            137           9.17
India                 4            324           5.00
Indonesia             2            771           1.00
Israel                5            134          10.60
Lebanon              10            337           3.00
Malaysia              3            250           8.67
Mexico               40            373           6.10
Panama                4            418           9.00
Philippines          17            633           6.00
Poland                2            378           8.00
South Africa          4            346           7.00
South Korea          49            377           6.57
Taiwan                1            199          13.00
Thailand              4            379           6.75
Turkey               12            580           3.75
Uruguay              10            260           7.00
Venezuela            10            629           2.50

Total               269            400           5.94

                  Average        Fraction
                Maturity in      Exchange      Fraction
Nation             Years          Listed      Sovereign

Argentina           4.91           0.36          0.41
Brazil              5.75           0.71          0.29
Chile               7.33           0.67          0.33
China               6.94           0.35          0.88
Colombia           10.54           0.15          1.00
El Salvador         6.00           1.00          1.00
Hungary             5.58           0.50          0.17
India               6.75           1.00          0.00
Indonesia           7.00           0.00          1.00
Israel              5.00           0.40          1.00
Lebanon             4.80           1.00          1.00
Malaysia            5.00           0.33          0.67
Mexico              6.80           0.50          0.38
Panama             16.50           0.00          1.00
Philippines        10.24           0.18          0.41
Poland              5.50           0.50          0.50
South Africa       11.50           0.00          1.00
South Korea         4.47           0.27          0.08
Taiwan              9.00           0.00          0.00
Thailand            7.00           0.50          0.75
Turkey              7.92           0.58          0.92
Uruguay            11.00           0.50          0.90
Venezuela          11.40           0.20          0.40

Total               6.95           0.42          0.48

Table III. Near Neighbor, Gaussian, and Epanechnikov Estimates of Yield
Difference between Non-Resident Indian Bonds and Matched Bonds

This table presents the estimated yield difference between Non-resident
Indian bonds and matched bonds using the NEAR NEIGHBOR, GAUSSIAN, and
EPANECHNIKOV estimators developed in Heckman, Ichimura, and Todd (1997,
1998). To implement these estimators, we estimate propensity scores for
the restricted bonds, namely the Non-resident Indian bonds, such as the
Resurgent India Bonds, and India Millennium Bonds, and for the
non-restricted bonds through a logit model (top panel) and a probit
model (bottom panel):

RESTRICTED = [[beta].sub.0] + [[beta].sub.cr] CREDIT RATING +
[[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3]
SOVEREIGN + error.

where RESTRICTED takes a value of one for Non-resident Indian bonds
(and zero otherwise). CREDIT RATING is the Moody's credit rating of a
bond. Each rating is given a numerical counterpart as shown in Appendix
2, e.g., Aaa = 16, Aa1 = 15, Aa2 = 14. MATURITY is the maturity of the
bonds in years. EXCHANGE takes a value of one if the bond is listed on
an exchange, and zero otherwise. SOVEREIGN stands for whether the debt
issue is a sovereign bond. The estimators are defined as follows: NEAR
NEIGHBOR chooses for each restricted bond, the n non-restricted bonds
with closest propensity scores. We specify n=10 AND n=25. GAUSSIAN and
EPANECHNIKOV use a weighted average of non-restricted bonds, with more
weight given to a non-restricted bond with propensity score that is
closer to a restricted bond's propensity score. GAUSSIAN uses all
non-restricted bonds, while for EPANECHNIKOV, we specify a propensity
score bandwidth (h=0.01) that limits the sample of non-restricted
bonds. For each estimator, we have a sample of matched pairs as
described above, for which we compute the difference between the yields
of restricted and matched bonds, and report the average yield
difference in basis points. Column 1 reports the results for all
restricted bonds and the other two columns report the results for
Resurgent India Bonds and India Millennium Bonds separately. The
standard error and the t-statistics are estimated using bootstrapping.

                            (1)                         (2)

                       All Restricted             Resurgent India
                           Bonds                        Bonds

                 Yield Diff.                 Yield Diff.
Estimator           (bps)        T ratio        (bps)        T ratio

Logit Model
NEAR NEIGHBOR
  (n=10)           -223.15      -2.80 (a)      -183.00      -2.22 (b)
NEAR NEIGHBOR
  (n=25)           -202.56      -5.15 (a)      -115.40      -2.06 (c)
GAUSSIAN           -145.27      -5.28 (a)      -138.83      -2.94 (a)
EPANECHNIKOV       -199.24      -2.12 (b)      -172.30      -1.19
Probit Model
NEAR NEIGHBOR
  (n=10)           -173.45      -2.20 (b)      -201.20      -2.56 (b)
NEAR NEIGHBOR
  (n=25)           -204.44      -5.37 (a)      -127.32      -2.29 (b)
GAUSSIAN           -145.32      -5.31 (a)      -139.10      -2.89 (a)
EPANECHNIKOV       -217.91      -2.13 (b)      -174.70      -1.59

                            (3)

                     India Millennium
                           Bonds

                 Yield Diff.
Estimator           (bps)        T ratio

Logit Model
NEAR NEIGHBOR
  (n=10)           -297.75      -3.74 (a)
NEAR NEIGHBOR
  (n=25)           -210.48      -3.05 (a)
GAUSSIAN           -145.33      -2.88 (a)
EPANECHNIKOV       -310.89      -1.76 (c)
Probit Model
NEAR NEIGHBOR
  (n=10)           -325.55      -3.74 (a)
NEAR NEIGHBOR
  (n=25)           -213.90      -3.21 (a)
GAUSSIAN           -145.54      -2.58 (a)
EPANECHNIKOV       -334.13      -1.61

(a), (b), and (c) stand for significance at the 1%, 5%, and 10% levels
using a two-tailed test.

Table IV. Multivariate Regressions of Yield Spread

This table gives the OLS estimates of the following equation:

YIELD SPREAD = [[beta].sub.0] + [[beta].sub.cr] CREDIT RATING +
[[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3]
SOVEREIGN + [[beta].sub.4] RESTRICTED + error.

The dependent variable YIELD SPREAD is the ex ante yield spread (in
basis points) of a new debt issue, i.e., ex ante offering yield of new
debt security minus the ex ante yield of a U.S. Treasury security of
comparable maturity. The independent variables are: CREDIT RATING is a
set of comprehensive credit rating dummy variables. For example, Ba1 is
a dummy variable which is one if Moody's comprehensive credit rating
for the issue is Bal. MATURITY is the maturity of a debt issue measured
in years. EXCHANGE takes the value one if a debt issue is traded on an
exchange. SOVEREIGN stands for whether the debt issue is a sovereign
bond. RESTRICTED takes a value of one for Non-resident Indian bonds,
such as the Resurgent India Bonds, and the India Millennium Bonds. All
dummy variables are zero otherwise. Column 1 reports the results for
all restricted bonds and the other two columns report the results for
Resurgent India Bonds and India Millennium Bonds separately. The T
ratios are adjusted for heteroskedasticity using White's (1980)
variance-covariance matrix.

                              (1)                       (2)

                        All Restricted bonds     Resurgent India Bonds

Variable               Coeff        T ratio      Coeff        T ratio

INTERCEPT               72.24       3.55 (a)     118.37       3.52 (a)
B3                     609.39      14.24 (a)     684.62      31.01 (a)
B2                     396.71       5.02 (a)
B1                     286.32       6.92 (a)     323.31       7.17 (a)
Ba3                    265.84       6.08 (a)     153.96       5.36 (a)
Ba2                    312.35       9.62 (a)     261.01       7.70 (a)
Ba1                    354.31       7.39 (a)     261.09       6.30 (a)
Baa3                   137.30       3.55 (a)      89.45       2.13 (b)
Baa2                    85.96       3.10 (a)     -16.69      -0.26
Baa1                   149.53       2.61 (a)     176.70       7.52 (a)
A3                      33.95       0.46 (a)
A2                      26.95       0.68 (a)
MATURITY                14.08       6.23 (a)       9.96       4.18 (a)
EXCHANGE                32.32       1.29           4.03       0.14
SOVEREIGN               -9.72      -0.30         -63.67      -2.21 (b)
RESTRICTED            -170.90      -5.52 (a)    -112.82      -4.25 (a)
Observations                   273                       117
Adjusted [R.sup.2]           0.3468                    0.4676

                              (3)

                       India Millenium Bonds

Variable               Coeff        T ratio

INTERCEPT               61.71       2.18 (b)
B3                     589.71      11.01 (a)
B2                     397.83       4.88 (a)
B1                     288.05       4.67 (a)
Ba3                    437.98       7.15 (a)
Ba2                    501.28       5.94 (a)
Ba1                    482.99       5.52 (a)
Baa3                   175.09       3.54 (a)
Baa2                   119.09       3.69 (a)
Baa1                   151.90       2.21 (b)
A3                      89.54       0.80
A2                      39.99       0.68
MATURITY                15.25       4.85 (a)
EXCHANGE                40.94       1.14
SOVEREIGN              -20.95      -0.43
RESTRICTED            -316.14      -6.52 (a)
Observations                   156
Adjusted [R.sup.2]           0.4175

(a), (b), and (c) stand for significance at the 1%, 5%, and 10% levels
using a two-tailed test.

Table V. Multivariate Regressions of Yield Spread (using Alternative
Measures of Credit Rating)

This table gives the OLS estimates of the following equation:

YIELD SPREAD = [[beta].sub.0] + [[beta].sub.cr] CREDITRATING +
[[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3]
SOVEREIGN + [[beta].sub.4] RESTRICTED + error.

The dependent variable YIELD SPREAD is the ex ante yield spread (in
basis points) of a new debt issue, i.e., ex ante offering yield of new
debt security minus the ex ante yield of a US Treasury security of
comparable maturity. The independent variables are: CREDIT RATING
stands for the International Country Risk Guide (ICRG) composite credit
rating in Panel A, and the Institutional Investor (II) country credit
rating in Panel B for the bond issuer. MATURITY is the maturity of a
debt issue measured in years. EXCHANGE takes the value one if a debt
issue is traded on an exchange. SOVEREIGN stands for whether the debt
issue is a sovereign bond. RESTRICTED takes a value of one for
Non-resident Indian bonds, such as the Resurgent India Bonds, and the
India Millennium Bonds. All dummy variables are zero otherwise. The T
ratios are adjusted for heteroskedasticity using White's (1980)
variance-covariance matrix.

         Panel A. ICRG Composite Credit Rating

Variable               Coeff                T ratio

INTERCEPT             1435.54               9.64 (a)
CREDIT RATING          -15.91              -8.00 (a)
MATURITY                15.25               5.91 (a)
EXCHANGE                 5.73               0.23
SOVEREIGN              -89.95              -3.40 (a)
RESTRICTED            -204.19              -5.87 (a)
Observations                       273
Adjusted [R.sup.2]               0.2975

        Panel B. Institutional Investor Rating

INTERCEPT              592.26               8.10 (a)
CREDIT RATING           -5.24              -4.29 (a)
MATURITY                14.83               5.51 (a)
EXCHANGE                10.46               0.39
SOVEREIGN              -62.28              -2.23 (b)
RESTRICTED            -122.37              -3.94 (a)
Observations                       273
Adjusted [R.sup.2]               0.1813

(a), (b), and (c) stand for significance at the 1%, 5%, and 10% levels
using a two-tailed test.

Table VI. Multivariate Regressions of Yield Spread Controlling for
Inflation or Country Fixed Effects

This table gives the OLS estimates of the following equation:

YIELD SPREAD = [[beta].sub.0] + [[beta].sub.cr] CREDITRATING +
[[beta].sub.1] MATURITY + [[beta].sub.2] EXCHANGE + [[beta].sub.3]
SOVEREIGN + [[beta].sub.4] RESTRICTED + [[beta].sub.ctrl] CTRL + error.

The dependent variable YIELD SPREAD is the ex ante yield spread (in
basis points) of a new debt issue, i.e., ex ante offering yield of new
debt security minus the ex ante yield of a US Treasury security of
comparable maturity. The independent variables are: CREDIT RATING is a
set of comprehensive credit rating dummy variables. For example, Ba1 is
a dummy variable which is one if Moody's comprehensive credit rating
for the issue is Ba1. MATURITY is the maturity of a debt issue measured
in years. EXCHANGE takes the value one if a debt issue is traded on an
exchange. SOVEREIGN stands for whether the debt issue is a sovereign
bond. RESTRICTED takes a value of one for Non-resident Indian bonds,
such as the Resurgent India Bonds, and the India Millennium Bonds. CTRL
stands for INFLATION in Column 1 (where we control for inflation) and
for COUNTRY in Column 2 (where we control for country fixed effects),
where INFLATION is the lagged annual inflation rate at the time of the
restricted bond issue obtained from the World Economic Indicators
database and COUNTRY represents the set of country indicator variables
reflecting the country fixed effects included in the regression. All
dummy variables are zero otherwise. The T ratios are adjusted for
heteroskedasticity using White's (1980) variance-covariance matrix.

                              (1)                       (2)

                            Inflation            Country Fixed Effects

Variable               Coeff        T ratio      Coeff        T ratio

INTERCEPT               74.59       3.65 (a)      78.75       3.92 (a)
B3                     596.95      12.49 (a)     627.01      12.01 (a)
B2                     371.23       4.54 (a)     303.58       2.30 (b)
B1                     273.53       6.40 (a)     298.26       3.99 (a)
Ba3                    236.04       5.81 (a)     196.13       5.06 (a)
Ba2                    299.36       8.86 (a)     296.68       8.26 (a)
Ba1                    349.28       7.25 (a)     267.01       5.07 (a)
Baa3                   125.18       3.20 (a)     179.22       2.88 (a)
Baa2                    81.09       2.83 (a)     115.78       3.94 (a)
Baa1                   149.87       2.65 (a)     214.08       1.17
A3                      32.30       0.44         189.02       4.42 (a)
A2                      24.22       0.61         290.63       6.37 (a)
MATURITY                13.80       6.07 (a)      13.36       5.98 (a)
EXCHANGE                30.78       1.23          53.64       1.95 (c)
SOVEREIGN              -13.55      -0.42         -28.79      -0.67
RESTRICTED            -155.55      -5.01 (a)    -121.57      -3.50 (a)
INFLATION                1.18       1.72 (c)
Country Fixed
  Effects                                                Yes
Observations                   273                       273
Adjusted [R.sup.2]           0.3494                    0.3951

(a), (b), and (c) stand for significance at the 1%, 5%, and 10% levels
using a two-tailed test.

Table VII. Average Abnormal Return Around the Announcement of
Non-Resident Indian Bonds

This table presents the average abnormal return on foreign currency
bonds of Indian issuers surrounding the announcement of Resurgent
India Bonds (announced on June 2, 1998), and India Millennium Bonds
(announced on September 21, 2000). Abnormal returns are based on the
market adjustment method, where the market index is the J.P. Morgan
Emerging Markets Index (JPMEMBI). The results are based on a 100 day
estimation time period, from day -102 to -3 relative to the
announcement date. Column 1 reports the results for all restricted
bonds and the other two columns report the results for Resurgent India
Bonds and India Millennium Bonds separately. The T ratios are computed
using the methodology of Brown and Warner (1985) that considers both
time series and cross-sectional dependence in returns.

                  (1)                        (2)

          All Restricted Bonds      Resurgent India Bonds

         Abnormal                   Abnormal
Date    Return (%)     T ratio     Return (%)    T ratio

-2         0.00        0.08           0.11         0.15
-1         0.64        1.25           0.64         0.93
 0        -1.40       -2.74 (a)      -1.59        -2.32
 1         0.13        0.25           0.05         0.07
 2        -0.29       -0.57          -0.10        -0.15

                  (3)

         India Millennium Bonds

         Abnormal
Date    Return (%)     T ratio

-2        -0.24       -0.47
-1         0.64        1.30
 0        -0.99       -1.98 (b)
 1         0.32        0.64
 2        -0.73       -1.46

(a), (b), and (c) stand for significance at the 1%, 5%, and 10% levels
using a two-tailed test.

Table VIII. Multivariate Regressions of Announcement Effects Associated
with New Bond Issues

This table gives the OLS estimates of the following equation:

ABNORMAL RETURN = [[beta].sub.0] + [[beta].sub.nbi] NEW BOND ISSUE
+ [[beta].sub.1] LN(NEW BOND SIZE) + [beta]LN(BOND SIZE) [[beta].sub.3]
RESTRICTED + error.

The dependent variable ABNORMAL RETURN is the day 0 abnormal return of
an unrestricted bond in response to the announcement of a new bond
issue, measured as a percentage of the bond price on the previous day.
The independent variables are: NEW BOND ISSUE is a set of new bond
issue fixed effects (i.e., indicator variables). LN(NEW BOND SIZE) is
the natural log of the offering size of a new bond issue, measured in
$ millions. LN(BOND SIZE) is the natural log of the size of the
unrestricted bond (in $ millions) whose abnormal return is being
measured. RESTRICTED takes a value of one if a new bond issue is a
restricted bond, such as the Resurgent India Bonds or the India
Millennium Bonds, and zero otherwise. The coefficient estimates and T
ratios for the new bond issue fixed effects are not reported although
they are included in the regressions. The T ratios are adjusted for
heteroskedasticity using White's (1980) variance-covariance matrix.

Variable                        Coeff       T ratio

INTERCEPT                       -2.16      -2.27 (b)
LN(NEW BOND SIZE)                0.41       2.63 (a)
LN(BOND SIZE)                   -0.01      -0.07
RESTRICTED                      -2.85      -4.60 (a)
New bond issue fixed effects           Yes
Observations                           275
Adjusted [R.sup.2]                   0.4339