THE ECONOMICS OF THE PRIVATE MARKET
Appendix H. Borrower Substitution between the Public and Private Markets
Empirical research on the relation between public and private bond markets has focused primarily on the ability of borrowers to switch between the two. The research in this regard is limited, but it has covered a considerable period of time. In one of the earliest studies, which covered 1951 to 1961, Cohan (1967) concluded that the two markets were segmented. This conclusion was based on a finding that yields, adjusted for issuance costs, were generally higher on private placements than on public offerings. Cohan attributed the lack of similar borrowing costs to the structure of the markets: The public bond market served primarily large, better-known corporations, and the private market served smaller, less-well-known corporations. The preference of the public market for better-known corporations prevented smaller companies from shifting out of the private market into the public market to take advantage of the lower borrowing costs.
Cohan's research did not examine the sensitivity of the demand for funds in the two markets to changes in yields. Wolf (1974) corrected this shortcoming by estimating demand equations over 1956-70 for two groups of firms: One included Fortune 500 companies, and the other included all other industrial corporations. 209
For each group, the dependent variable in a regression equation was the ratio of gross issuance in the private market to total gross issuance in both the private and the public markets; the explanatory variables included the spread between yields in the private and public markets, a measure of the supply of funds in the private market, and average issue size. With smaller borrowers having no access to the public market, Wolf hypothesized that only the Fortune 500 companies would be sensitive to the yield spread. The empirical results were consistent with this hypothesis; the coefficient on the yield spread was statistically significantly different from zero only in the equations for the Fortune 500 companies. Moreover, when these companies were further disaggregated, the estimated elasticity of demand with respect to the yield spread for the top 100 corporations was, in magnitude, nearly three times that for the top 250 corporations.
Although Wolf's approach was an improvement over Cohan's, the demand equations were misspecified because other relevant explanatory variables, such as issuance costs, credit quality, and characteristics of the instruments, were not included. In this regard, Blackwell and Kidwell's (1988) study improved on Wolf's by incorporating more institutional features of the markets into an analysis of a borrower's choice of the public or the private market. They assumed that, holding all else constant, a borrower chooses to issue in the least expensive market. The basic elements of the decision can be illustrated by assuming the borrower is contemplating issuing a one-year bond with one annual interest payment. The all-in cost per dollar borrowed, r, can be computed from the expression
S - K = (1 + c)S/(1 + r),
where S is the issue price, K is the fixed cost of issuance, and c is the coupon rate of interest. The bond is assumed to be issued at par; thus, the issue price and face value are equal. Variable costs can be ignored as they do not produce economies of scale in issue size. 210 From this expression, the all-in cost can be derived as
As issue size S increases, the all-in cost r falls, thereby producing the economies of scale.
Two conditions are necessary for a borrower to find that the private market is less expensive for small issues and more expensive for large issues. One is that the fixed issuance cost must be lower in the private market, and the other is that the coupon rate must be lower in the public market. Assuming both are satisfied, the break-even issue size S*, the point at which the all-in cost is the same in both markets, is
where the subscript pr indicates the private market and the subscript pu, the public market. For an issue less than S*, the borrower would choose the private market; for an issue greater than S*, the borrower would choose the public market.
If either of the two conditions is not satisfied, borrowing in only one market is always cheaper. Issuance costs are lower in the private market because of the absence of both SEC registration expenses and underwriting fees. However, the coupon rate condition is not necessarily satisfied. On the one hand, for some borrowers the difference in coupon rates in the two markets may be fully accounted for by an illiquidity premium that makes the coupon rate in the private market higher than that in the public market. In this case, issue size determines the choice of the market. On the other hand, information-problematic borrowers should find that the coupon rate is higher in the public market, because of the unwillingness of public-market investors to undertake the credit analysis and monitoring required to lend to these borrowers. Under this circumstance, these borrowers, in effect, have no choice but to issue in the private market.
Blackwell and Kidwell provide some weak empirical evidence that information problems are more important than issuance costs for small companies using the private market. To examine this issue, Blackwell and Kidwell first estimated a linear version of equation 1 for a sample of public bonds issued by large companies that had used only the public market. The explanatory variables in the regression estimating equation 1 included issue size, characteristics of the issuer, features of the securities, and market conditions. Based upon this regression, Blackwell and Kidwell then predicted the all-in cost in the public market for a sample of private bonds issued by small companies that had used only the private market. The predicted value was then compared with the actual all-in cost of the private placements to determine what, if any, cost saving was achieved in the private placement market.
The estimated cost saving from using the private placement market averaged 132 basis points. The average saving was 301 basis points for below-investment-grade companies but was -16 basis points for investment-grade companies. This finding is consistent with the cost saving reflecting the information-problematic nature of the companies issuing the private placements as below-investment-grade firms are more likely (though not certain) to be information problematic. Because such companies require the most due diligence and monitoring, they must pay the largest premium to borrow in the public market, where investors generally forgo the risk control inherent in due diligence and monitoring. Some market participants have indicated also that issuance costs tend to be higher for such companies, but they are unlikely to be as large as Blackwell and Kidwell's estimates.
Several problems temper Blackwell and Kidwell's results. One is that the issue size variable in the regression estimates was not significant, indicating an absence of economies of scale in issue size in the public market. Another problem is that the values of the explanatory variables for the private issues are generally well outside the ranges of the variables used to estimate the model, leading to extremely imprecise estimates of the cost savings. Finally, the negative estimate for the investment-grade issues, even though statistically insignificant, implies that nonprice factors not included in the regression model dominate price considerations for high-grade borrowers. Otherwise, it makes no sense for these companies to borrow in the private market when the public market is the cheaper alternative. The omission of relevant variables in the regression model could also lead to biased estimates of the cost saving in the private market.
In a second regression, Blackwell and Kidwell produce results that can be interpreted as implying that those companies with access to the public market utilize the private market primarily because of special borrowing needs and not because of cost considerations. This regression involved estimating equation 1 for a sample containing both private and public issues; the public issues were those described above, whereas the private issues were those of companies that had issued in both the public and private markets. To distinguish between the public and the private issues, the regression included dummy variables that allowed the intercept and the coefficients of several explanatory variables to shift with the market.
The dummy variables were generally insignificant, indicating that the all-in cost of issuing in the two markets was the same. This finding is inconsistent with the underlying model of borrower choice and certainly is at odds with market participants' description of the cost differentials between the two markets. The modeling strategy may well account for this result and, in fact, may lack the power to detect differences in issuance costs. One problem concerns those firms that had issued in both markets, the so-called switch hitters. They are large relative to companies issuing only private placements, and consequently their observations cluster around the break-even issue size S*. Thus the estimated equation provides unreliable estimates of issuance costs for small issues (which are generally by small firms), and precisely for such issues the differences in the all-in cost between the private and public markets should be most pronounced. An even more critical problem stemming from the presence of switch hitters in the sample is their tendency to use the private market for non-price-related reasons, such as speed or complexity of a particular issue, rather than for cost alone. Thus, their use of the private placement market may provide little information about relative issuance costs in the two markets and, indeed, could lead to a finding that cost minimization is irrelevant. That is, the negative results for issuance cost actually point to the importance of nonprice factors.
In general, Blackwell and Kidwell's work points to the need for a richer model of borrower choice. Choosing is not simply a matter of selecting the market with the lower all-in cost of borrowing. As our analysis in part 1 emphasizes, nonprice terms, such as disclosure requirements, covenants and their renegotiation, and speed, can matter and should be incorporated into models of borrowers' choice of markets.