The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Series of Occasional Papers in Draft Form Prepared by Members of the Research Department for Review and Comment. 76-4 Financial Disclosure and Market Evaluations of Bank Debt Securities Chayim Herzig-Marx Federal Reserve Bank of Chicago h± + ♦ ±♦ * * * * * * * ♦[♦♦♦♦♦♦♦♦♦I ♦♦I Research Paper No. 76-4 FINANCIAL DISCLOSURE AND MARKET EVALUATIONS OF BANK DEBT SECURITIES By Chayim Herzig-Marx Department Federal The views and do not Reserve material to expressed stimulate of herein of is of discussion, the Research Bank are or the of Chicago solely represent Chicago contained permission Reserve necessarily Bank of those the v i e w s of of the authors the F e d e r a l Federal Reserve System. a preliminary nature, is and authors. is not to b e The circulated quoted without FINANCIAL DISCLOSURE AND MARKET Among EVALUATIONS all imprudence cacy than the of checks banks, giving great recurrent separated fied disclosure determine for the can achieve and its have SECURITIES* been is no publicity to their one devised of to greater the effi condition. . A actual Introduction ranks with capital Indeed, empirically, greater costs that in banking. theoretically, pressure policy issues BANK DEBT there I. Financial OF bank or in disclosure benefits desired of adequacy the two the public requires fuller as one cannot of be entirely mind. careful disclosure, so the Intensi research that to public goals. Federal bank regulators have long expressed their conviction that free markets are unable to evaluate the condition of commercial banks. Markets are said to be relatively naive in matters of bank operations and accounting: It ' should the has supervisory different however, — at been determine suggested the agencies judgment a much least for of more that adequacy of should their the not own. majority 1 play of to the ma r k e t and enforce approach market the of capital, presume This knowledgeable the v a s t free a b a n k ’s presupposes, than we n a t i o n ’s that a have 14,000 today banks.^ 2 The case is often stated even more strongly on the issue of market evaluation of leverage: In most industries, as the debt equity ratio increases, the cost of debt normally increases, reflecting creditor’s [sic] demands for higher risk premiums. This market discipline does not seem as effective in banking. ...I do not think it can be effectively argued that the mar ket itself can be relied upon to police the rate of bank asset expansion financed through leveraging. 3 In somewhat caricatured form, the regulatory argument is this: a) mar kets lack information to make "good" decisions about bank condition; b) even if markets had adequate information, the decisions they made would not be "good"; c) therefore: In a sense, the bank regulatory agencies are exercising for most banks the judgment as to capital adequacy which a perfectly informed market might be able to exercise.^ Notwithstanding such regulatory pronouncements, it is an empirical question whether or not markets are able to evaluate the condition of commercial banks and whether or not their evaluations are f,goodn or ade quate. The empirical question can be decomposed into three pieces: (1) do markets make consistent and systematic use of the information currently available to them? with regulatory judgments? (2) do market evaluations accord closely (3) if market evaluations depart systema tically from regulatory judgments, whose judgments are superior? This paper addresses part 1 of the empirical question. An economic point of view is adopted, which is to say that markets are seen as ef ficient information processing institutions. An empirical finding that markets are able to incorporate available information systematically and consistently into price-quantity decisions will be taken as sufficient 3 evidence to warrant further investigation into the relationship between market and regulatory assessments of bank condition. A finding to the contrary, however, cannot be taken as presumptive evidence that market judgment is inadequate, since regulatory agencies have the power to retard the development of well-functioning markets by restricting the flow of raw data. To this extent the case for disclosure is stacked in favor of greater dissemination of information. Our attention in this paper, however, will be confined to the issue whether or not markets use the information currently available. The philosophical bias in favor of greater disclosure will be balanced by an empirical bias in favor of the proposition that the mar ket cannot use even that information which it already has. In particu lar, "information currently available" is defined to be the Reports of Income and Condition. As an approximation to the amount of information the market has, these two documents certainly represent at best the greatest lower limit: they are the greatest lower limit only if they contain no misleading information. For the type of securities studied in this paper, additional information frequently is available and forth coming from banks. The type of security examined in this paper is a newly issued capital note or debenture. In principle there is no reason why seasoned capital notes could not be used. Indeed, using a single cross-section of pre viously issued capital notes has the advantage of avoiding some econome tric problems of pooling several cross-sections. In practice, however, very few bank debt issues are traded with sufficient activity to generate an adequate sample for study and one which reflects a range of bank sizes. 4 The hypothesis of the paper, simply stated, is that the rate of return required by the market (the yield to maturity) systematically incorpo rates information concerning the issuing bank’s condition and earnings prospects. Only securities issued directly by banks are included in this study. Issues of holding companies, even though the proceeds were channeled directly to the bank subsidiary, are excluded for two basic reasons. The financial structures of holding companies are considerably more complex even than those of banks, which makes comparisons between banks and holding companies quite difficult. There is, in addition, some evidence that financial markets are still in the process of learning how to eval uate holding companies and their financial conditions.^ Despite the exclusion of direct issues by holding companies, it is impossible to avoid such complications entirely since many banks are holding company affiliates. To the extent possible, however, such a compartmentalization will be attempted. II. The Model It is assumed that investors maximize utility functions which are increasing in rate of return and decreasing in risk. When a security is offered to an investor for purchase, the amount of the issue, the coupon rate, and the term to maturity are data. termines the yield to maturity. A bid price then de In assessing the risk-return charac teristics of any security, the investor will compare (at least intuitively) that issue with a security free of credit risk which matures on the same day. The risky security must then offer sufficient increased return to compensate for the increased risk over the life of the security. 5 In pricing security issues, the market will consider both charac teristics of the debt instrument itself and characteristics of the issuer. The following characteristics of the instrument were considered relevant to the market pricing mechanism (their presence or absence will be indicated by dummy variables): convertibility; callability; subordination to other debt; provision for payment in instalments; pro vision for sinking fund; private placement; issuer a holding company affiliate; restrictions on dividend payout; restrictions on issuance of other debt.^ A multitude of variables describing the condition and prospects of a bank can be constructed from the Income and Condition Reports. Fortunately, literature does exist to guide the empiricist in devising his variables. Proper use of analytical concepts and available data requires some considered thought, however. Two points are especially important, namely, the implications of focusing on investors in debt (as opposed to equity) securities and the manner of accounting for the new security in the bank’s financial condition. The proper definition of income hinges on the type of security under consideration. Payments of interest have a claim on gross reve nues prior to payments of taxes or dividends. Thus, income before taxes is the relevant earnings measure for purposes of this paper. By the same logic, one should not adjust pre-tax earnings to reflect taxexempt earnings on a "fully taxable basis." Tax-exempt earnings bene fit holders of equity securities because a larger proportion of pre-tax earnings flow through to after-tax earnings. Since returns to debt holders come out of pre-tax income, tax-exempt status is of no benefit. 6 This is not to imply that tax-exempt status of earnings is of no con sequence to debt holders; for, other things equal, holders of debt should prefer a given dollar volume of taxable securities to the same dollar volume of tax-exempt securities, since the taxable securities will gen erally give rise to a larger flow of pre-tax earnings. No clear answer can be given to the question of how to account for a (proposed) new issue of debt in evaluating a bank’s financial con dition. Given the nature of the decision model outlined above, all financial information was gathered for the year prior to the year in which the security was issued. One then has two more or less limiting views of the bank, one pessimistic, the other optimistic. According to the pessimistic view, the new debt issue raises the degree of lever age and financial risk. Therefore, in calculating, e.g., a debt/equity ratio, one should include the new issue in the numerator. The optimistic view holds that the bank is an ongoing enterprise and can be expected to restore the previous degree of leveraging. According to this view, one would exclude the new issue from the debt/equity ratio on the grounds that to include it would misrepresent the long-term financial structure of the bank. The truth is likely to be somewhere in between. The only fair way to conduct the analysis is to try both methods. The inclusion o or exclusion of the new debt applies also to interest coverage ratios. With these considerations out of the way, we can present the finan cial variables used in this study. They fall into two basic categories, variables constructed from the balance sheet alone, and variables using information from the income statement. 7 Balance sheet variables correspond, for the most part, to ratios well known in banking literature. Given the great problems inherent in such rudimentary summary ratios as capital to deposits or capital to assets, some attempt to improve upon these was made. (1) Ratios to total assets. As a general measure of the riskiness of the bank's portfolio, a ratio of (credit) riskless to total assets was constructed. Riskless assets are defined as cash and due from banks, Treasury securities, and U.S. agency securities. As a measure of the ability of capital to absorb losses in the asset portfolio, the ratio of capital available to absorb losses to assets was taken. less capital stock. Capital available to absorb losses is equity capital Q This exclusion was made because in many, if not most, jurisdictions impairment of capital stock requires the immediate liquidation or forced merger of the institution. Total borrowings, total borrowings less time and savings deposits, and debt capital were expressed as percentages of total assets in the attempt to measure financial leverage of assets. (2) Ratios to total loans. The loan portfolio displays a wide range of asset liquidities and risks. Four percentages of the loan portfolio were calculated in the attempt to measure liquidity and risk better: a) secured loans, where secured loans are loans secured by farmland, mortgage loans on one- to four-family dwellings and on multi-family residential properties, loans secured by nonfarm nonresidential properties, auto loans, loans on mobile homes, and instalment loans to repair and modernize residential property; 8 b) financial loans, which are loans to financial institutions and loans for purchasing or carrying securities; c) commercial and industrial loans; and d) personal loans, which are credit card and related loans, instal ment loans to purchase retail consumer goods other than mobile homes, instalment loans other than those already mentioned under a), and single payment loans for household or personal expenditures. The ratios of loan loss reserves and actual loan losses to loans were calculated as measures of risk in the loan portfolio. The ratio of "core deposits11 to loans measures the proportion of loans which are sup ported by relatively stable liabilities.**"® (3) Ratios to total securities. The ratio of municipal securities to total securities was discussed above, in connection with tax-exempt income. The ratio of reserves on securities to total securities was calculated as a rough measure of the degree of risk inherent in the securities portfolio. (4) Ratios to equity capital Two measures of leverage in the capital account were calculated, debt capital to equity capital and total debt to equity capital, where total debt includes total deposits, federal funds purchased, other indeb tedness, capital notes previously outstanding, and preferred stock. The new security issue was included in these two ratios, in accordance with the discussion above. Several earnings coverage ratios were constructed. These include fixed charges to gross income, fixed charges to expenses, and times-chargesearned (based on net income). Each of these three was calculated using both net and gross occupancy expense, including debt service on the newly issued security, but excluding interest on deposits. 9 The ratio of securities income to total income was calculated to in dicate stability of the income stream. Three other income ratios were calculated, based upon the writings of well-known analysts. Net operating earnings are expressed as a pro portion of total deposits. ^ The "margin of safety11 is the ratio of dividends plus retained earnings to gross income from operations before deduction of expenses or charges. "On an assumption that costs and non operating income are completely rigid, the margin of safety would repre sent the maximum proportion by which total sales might shrink and fixed charges still be earned in full."^ Since this measure is based upon the ratio of after-tax income to gross earnings, it would seem to be overly conservative: if gross revenues fall to a level that covers only fixed charges and expenses, no taxes need be paid. Therefore, a pre tax margin of safety was al^o calculated. Finally, the rate of return on stockholder equity (book value) was calculated, on the grounds that investors in bank debt securities are only well protected when the bank can adjust its capital structure by issuing new equity.^ III. Findings of Previous Research Four pieces of research are of interest for this study, in that they have presented results of considerable interest for the present topic. The seminal article on determinants of risk premia was written by Lawrence Fisher.^ He selected a sample of outstanding corporate bonds in five different years, calculated risk premia as the difference in yield to maturity between the corporate issue and a government issue maturing at the same time, and regressed the risk premium on four var 10 iables. The independent variables he used were earnings variability, period of solvency, equity/debt ratio, and size of the issue; all were usually significant. The model was specified in logarithmic form, on the grounds that one expected strong interaction among the independent variables. It is not clear from the article, however, whether or not he experimented with straight linear forms. Peter E. Sloane was interested in yield curves and the determinants of yield differentials at various points along the yield curve.^ The yield curves he constructed, for Treasury securities and for corporates rated Aaa or Baa, were based on averages of the yields for individual securities. Two findings of this research are important for present purposes. First, a linear regression model was used with success in explaining yield differentials. Second, Sloane also found that the yield differential should be expressed as an absolute, not percentage, difference. Richard H. Pettway has recently carried out a study quite similar to this one, testing a model of the determinants of risk premia for a sample of capital notes issued by commercial banks and holding companies.^ The question of interest in that study was whether or not the yield to maturity on capital note issues is influenced by the capital structure of the firm and in particular by measures of capital adequacy. The two capital adequacy ratios used were equity capital to assets and borrowings (including deposits but excluding debt capital) to total capital. The most important finding for present purposes is that yield to maturity was not related to the capital adequacy variables. Within the context of fi nancial disclosure, the conclusion from Pettway’s article is that the 11 market does ing caital to In not make adequacy another on whether structure salient ingly time of significant nificant. to assess the By in the price, from this not this company financial is is In 1970 1970 and leverage greater markets securities, and Jacobs 177 relat sensitive their in interest to financial equities. the m a r k e t became increas 1973, the and 1971, variables in 1973 period was moderately it w a s highly a negative in than 1973 apparently particular of explaining exerted are The to leverage in studied Their significant financial was that are that from equation, and equities. statistically influence study Boyd, in p r i c i n g drawn. increasing information issues. study consolidated financial markets leverage regression which holding of of company companies sample was 1972 note securities financial Furthermore, conclusion to v a l u e result leverage were price. share not use Beighley, bank holding or to over which article, holding empirical measuring on bank sensitive share of consistent in p r i c i n g recent valuation models focused any in still influence 1972. still sig The learning learning leverage. To gather the results of these four studies together, Fisher’s work indicates that a logarithmic specification is appropriate, while Sloane and Pettway used linear specifications. All three expressed the risk premium as the absolute difference in yield to maturity between the corporate issue and a Treasury issue maturing at the same time. Finally, Pettway found that differences in capitalization exerted no influence on risk premium for bank and holding company issues, while Beighley, Boyd was and going Jacobs found through a that, learning for holding process in companies evaluating at least, financial the m a r k e t leverage. 12 Since P e t t w a y ’s s a m p l e learning later phenomenon years may covers is 1971 occurring; reveal to and significant IV. Sample 1974, thus be that restricting effects and Data it m a y of the the same sample to capitalization. Construction The sample to test the model described above was assembled by com paring end of year call report data for two consecutive years, which allows one easily to determine which banks issued capital was done was notes during 18 the ye a r as elapsed. of D e c e m b e r amounts for of year In p r a c t i c e w h a t 31, of capital T-l. and All year T-l all year T than in y e a r during year T. years were relied from publicly Income from of the and it impossible and 1968 1969 represented to or in not at order larger from T amount as the and of having for initial available: a listing, reporting non-zero comparable for year notes of T but not capital 1970 if any for in notes through issue, and listing outstanding issued sample amount as m u c h a obtain years M o o d y ’s B a n k gather of 1973. the coupon rate, Finance Manuals information as possible sources. sheet data were and are duplicate after. obtained Condition substantial the States listing repeated issue. to Income data the identified deleted were Because 1969 year of a United in y e a r in procedure was price in Reports in T-l were balance port, available showing available issuance. is appearing items upon the banks were to m a t u r i t y , in outstanding This data banks notes banks Observations following all to not for For sample. for revisions directly 1968 these or for all sample the year prior were made in comparable. earlier reasons, the 1970 the to banks the year income re Furthermore, increased is the detail earliest 13 The buted The above by year of two m a j o r price that of the the formal procedure data day issuance the conclusions formulated the few information in A variables the iables. components infinite variables tions. which The and first has second to b o t h the tinued, there as fact that all all at issue need 28 15; of banks, 1972, the sample be the distri 11; 1973, are and m a t u r i t y not linear second of 1. the (note same as the and used, the a may of first is all that combinations third as m a n y the of of from var criterion of the The independent linear combina combination with are orthogonal component are principal a of Naturally, orthogonality. linear principal set independent being components. original components that the variables"). possible that of useful. the combination among analysis, on principal nature be of criterion linear principal the certain combinations the model regression "independent select empirical While regression analysis combinations a maximum strong results. method the to The no combinations component linear use of being component. and being to maximum variance among among first is the linear second principal principal imum variance the of actual this linear in component maximum variance the are size of using multiple type needed possible criteria principle are are 1971, available, of of of this referred criteria maximum variance first discussion (hereafter Two sample regardless from the dates freedom prompted brief 1; to m a r k e t expectation available specific among of sample security). small drawn total 1970, precise the a limiting issue went on the in follows: the extremely Principal some date degrees components. and the can be was as constraints issue, exact Given issue resulted has pair-wise The max orthogonal process components to as is con there 14 are independent useful when degrees from for of the variables. one faces freedom. fact that overcoming for Their all of for components That 80 nature, principal of m u l t i c o l l i n e a r i t y value insufficent upwards their problems of m a x i m u m variance. count By overcoming are of is, first the percent of freedom the few or insufficient orthogonal. arises from components total are multicollinearity pair-wise degrees components arises Their the value criterion typically ac variance of the independent components as regressors variables. Two place aim problems of is the to pendent variable those are dependent That is, one may well nent of a dependent of words, the set of twelve will yet for important are the of it independent regressors important does the example, the variables. not imply regressors. be the into it of Thus, knowledge with the case that components, compo to knowing the of in information possible to calculate of w h i c h the other of translate coefficients which the same way. principal total de correlation the v a r i a n c e regression difficult the in the correlation with the is in sixth significant of the of m a x i m u m largest while extremely not 4 percent most components is correlations components proportion Second, regression model typically only The small variables. variables, for a criterion rank encompass a very coefficients the display may contain and in partial It w i l l that, variables. independent find First, whose will independent regression iables variable variable, independent errors variables of m a x i m u m v a r i a n c e the principal variables. largest. with set in u s i n g independent locate criterion arise of the standard components independent are var 15 Table for of Independent Regression Variables on P r i n c i p a l Components Variable name Variable description 3 ISSUE size of 5 TERM term to m a t u r i t y 10 CONVERT dummy variable for convertibility 11 CALL dummy variable for callability 14 SINKING dummy variable for sinking fund 15 PRIVATE dummy variable for private placement 16 HC dummy variable for h o l d i n g 17 D IV RES dummy variable for dividend 18 OTH RES dummy variable for restrictions 25 %ASSET2 capital available 29 %L0AN1 secured loans 30 %L0AN2 financial loans 34 %L0AN6 loss rate on 35 %L0AN7 core deposits 40 %INC1 securites 42 FIXED2 fixed 50 RETURN rate 51 MARGIN3 margin issue from absorb issue date provisions company affiliation restrictions on losses other over debt assets loans over loans loans over income return of to over charges of in yea r s loans over (gross on safety total income occupancy) equity before taxes over income 16 V. was The entire input into set the p r incipal of the as second input. Those gression also in runs same they a the of the absolute note and yield collinearity results reported. smaller set of 18 listed the difference to m a t u r i t y in the in order of a Treasury III then among a step variables run few were was table. used Re eigenvalues variable yield were this preceeding dependent between on from independent components correlation with as extreme to not entering section Regressions empirical are in routine. are 18 v a r i ables made Due described and (risk premium, to m a t u r i t y security on the maturing at the time). Results When Table eigenvalue tion) . was step, were order expressed capital variables, Consequently a 19 Results variables components components. independent As independent a principal on unstable. of Empirical (note limited Judging regression the index stipulation or equal being analysis the of to components, which total variance the risk for regards sign both the and in can total One can original size, regression the is generated for variables, that the by just are that having the that not the 77 are many of equa eigenvalues a total first the quite of program. six percent significant several of in by equation computer over variables signs regression resulted conclude independent and into This note entered the components account are entering equation. also Issue components the easily independent premium. coefficients table one only entered. into variance when entered that be entered F-values, of one the N e w components components cipal plaining of by the Include results of than by to number six* c o m p o n e n t s An is D e f i n e d 1 shows The greater Debt prin of in the ex calculated unstable independent as varia- DEPENDENT VARIABLE 54 PREMIUM TOTAL SUM OF SQUARES DEGREES OF FREEDOM MEAN SQUARE 4720.2031 27 . 174.82233 CORRELATION BETWEEN PRINCIPAL COMPONti.TS AND DEPENDENT VARIABLE -0*07557 -0.16761 -0.02644 -0.00971 -0.08061 -0.10066 REGRESSION COEFFICIENTS OF PRINCIPAL COMPONENTS CONSTANT COMPONENTS (MEAN OF Y) 3.52589 -0.45425 -1.28837 -0.24213 -1.30958 -0.10152 -0.88206 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS INDEX OF RESIDUAL F-VALUES COMPONENTS SUM OF REGRESSION COMPONENT ENTERINGl SQUARES TO ENTER MODEL 1 2 3 4 5 6 4693.24609 4560.64062 4557.33984 4556.89453 4526.21875 4478.39453 0.15 0.44 0.29 0.21 0.19 0.19 0.15 0.73 0.02 0.00 0.15 0.22 R2 0.0057 0.0338 0.0345 0.0346 0.0411 0.0512 CONSTANT 4.6292 5.3065 4.7435 5.1355 3.2925 3.0722 VARIABLES 10 CONVERT 11 CALL 3 ISSUE 5 TERM 14 SINKING 15 PRIVATE -0.2534 -0.0057 0.0028 0.3238 -0.1772 -0.0402 -0.0428 -0.9732 -0.5771 -0.0080 -1.0021 -0.6343 -0.8979 -0.0091 -0.6019 -0.7989 -0.0287 -0.3898 -0.7379 -0.0089 -0.6804 -0.0277 -0.7187 -0.4783 -0.5084 -2.2283 -0.0115 -1.6765 -0.0192 -1.2901 -3.6980 -0.0287 -2.0244 -0.8676 -0.0245 -0.2832 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL i COMPONENTS (CONTINUED) VARIABLES 16 HC )7 DIV RES 18 OTH RES 25 %ASSET2 29 %LOAN-l 34 %L0AN6 30 9SL0AN2 -0.28758 0.25749 -0.11770 -0.07020 0.00957 -0.01204 0.48564 -0.26394 0.39139 -0.05367 -3.92914 -0.75040 -0.40827 0.02868 0.03014 -0.21285 -0.05807 -0.45613 -0.74101 0.49780 -4.15610 -0.18657 -0.78583 0.43361 -0.48375 -0.05938 -4.25293 0.03072 -0.33886 0.03404 -0.06215 0.55943 1.24830 -0.36817 -1.98292 0.54884 1.61589 -2.25996 -3.04109 0.21935 0.03389 -0.05830 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 50 RETURN 51 MARGIN3 -0.02419 -0.02040 0.01975 0.15323 0.19333 0.02285 0.18430 0.01600 0.23370 -0.00687 0.05595 -0.05918 Table 1 35 %LOAN7 0.00554 0.00809 0.00990 0.00964 0.00609 0.02103 40 %INC1 0.02074 0.02551 0.01277 0.01176 0.04233 0.00620 42 FIXED2 -0.03444 -0.011P4 -0.01393 -0.00851 -0.00830 0.04178 17 bles are counter regression (for strictions on absorb fixed charges entered 2 in variable over again of by stipulation the greater three or Analysis ten ponent achieves by is the the F again least squares variables ponents are Great regression entered for when the variable 42, are dependent equation was The limited correlations resulted in a total of equation. regression model explanatory nearly to the m o d e l ) . simple This re components the entering having 18, equity). principal regression entered. this of the on squares available loans; correlation with the into total least variable capital return components significant coefficients signs 0.21, quite entered for m a n y counter regression. In addition, dissimilar by from reveals variable. significant that This at com one the com 5 percent run previous the component component number encompass only 14 2. was sign and entered of component the size one are into Furthermore, percent of w h i c h the the original one would independent expect coefficients in of an the regression model vari ordinary independent in which com eigenvalues. in and of to w h a t differences principal variables, be ordinary test. have are only of 25, over 50, components 0.15 an R-square Calculated ables to being their into only loans an provision; variable results of from call variable index of v a r i a n c e ponent level equal components and entered that 11, secured order the components than 29, regression decreasing expect issuances; income; shows number variable debt variable (note one would example, other losses; Table to w h a t of coefficients to b e total 2 by expected. the m o d e l components in both 10, variance itself between makes 18, Only this one regression and of the up nearly runs, 2 altogether independent 12 percent. DEPENDENT VARIABLE 54 PREMIUM TOTAL SUM OF SQUARES DEGREES OF FREEDOM MEAN SQUARE 4720.2031 27 . 174.02233 CORRELATION BETWEEN PRINCIPAL COMPONENTS AND DEPENDENT VARIABLE -0.07557 -0.16761 -0.02644 -0.00971 -0.08061 0.08994 -0.04758 -0.11152 0.13155 0.07536 REGRESSION COEFFICIENTS OF PRINCIPAL COMPONENTS CONSTANT COMPONENTS (MEAN OF Y) 3.52509 -1.28837 -0.45425 -0.24213 1.87164 -1.32035 -3.38431 -0.1015? 4.30386 -0.88206 4.31991 0.10066 0.12767 0.09198 0.07081 -0.02286 -0.17592 -1.30958 -8.30844 1.34292 -5.43838 -0.34904 -30.07352 -0.05407 0.84428 -0.45476 -7.61399 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS INDEX OF RESIDUAL F-VAL'UES COMPONENTS SUM OF REGRESSION COMPONENT ENTERING SQUARES TO ENTER MODEL 10 18 2 3744.01880 3597.93359 3465.33130 6.78 3.90 2.90 6.78 1.02 0.92 R2 0.2068 0.2378 0.2659 CONSTANT 25.8788 2.3326 3.0099 VARIABLES 14 SINKING 3 ISSUE 5 TERM 10 CONVERT 11 CALL -2.1076 -0.2027 0.0351 0.0360 4.1402 -1.9884 0.3230 12.1873 28.2172 30.1972 11.7875 0.3207 29.2353 -2.0340 27.2590 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 29 *L0AN1 16 HC 17 DIV RES 18 OTH RES 25 %ASSET2 30 %L0AN2 34 %L0AN6 3.97911 -1.64179 -0.06529 -0.00377 5.25114 7.06732 -2.27115 -7.78109 -0.92918 -1.01619 -2.01528 16.51372 44.28165 -0.08232 -8.78890 -1.26725 -1.99163 17.02280 -0.06321 -1.05783 39.86685 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 50 RETURN 51 MARGIN3 -0.68219 -0.23237 0.47359 -0.80282 0.64721 -0.75888 Table 2 35 %L0AN7 0.06120 -0.00648 -0.00394 40 *INC1 -0.04459 0.41394 0.41872 PRIVATE 12.5524 14.2896 13.5698 FIXED2 -1.01440 0.37123 0.39442 18 The pal problem components principal which nal is encompass the An tion On attempt in of were thus principal components for the all squares. 10, Results have mean zero therefore in sets of these hand, greater than or very low correlation with the one data the variance set encompasses (about and and variance. regressions regressions one), origi depen signifi very little of analysis interpreta The all run for independent upon the using presented in ori standardized Eigenvectors original Based are to is of variance. were the components, for m e a n component. the regression problem principal unit equal six 2 percent). multiple difficult first of component which variable princi the total indicate which each one on the calculating standardized from the straight-forward Before three On regressing of dependent avoiding heavily weighted component hand, original of of eigenvalues percent a results study. exhibit utilize variables are more 77 the above. variables this other the hopes discussed ginal over with to the (those w i t h the variance made, in variables, correlated total was evident just variable. cantly interpreting components independent dent of the variables eigenvector ordinary the next least three tables. The independent variables having principal component able 16, -0.4414; 15 and to expectations, charges 16 have to significant the three in variable pay normal 15, that smaller regression values values The risk statistical sign banks in the are: -0.3116). signs. indicating the largest (eigenvector plausible income by variables with premia. criteria. Table 3 are eigenvector variable The of of 42, higher None of 42 the of is ratios the 0.5428; coefficients variable of those of tenth vari variables counter fixed coefficients is R E G R E S S I O N T I T L E .......................................................................................R U N 1 D A T A : I N C L U D E S D E P E N D E N T V A R I A B L E ......................................... ........................................ 54 PREMIUM T O L E R A N C E . .............................................. ..... ..............................................0 . 0 1 0 0 A L L DATA C O N S I D E R 0 AS A S I N G L E GROUP M U LTIP LE M U LTIP LE R R-SOUARE A N A LYSIS OF STD. ERROR OF EST. ISSU E 1 3 .2 3 09 VARIANCE REG RESSIO N RESID U A L VAR I ARI_ E IN TER C EP T PRIV A TE HC FIX ED 2 0 .3 3 1 5 0 .1 0 4 9 NEW S U ^ Op - S Q U A R E S 5 ]3 .8 5 9 4 ? 0 1 .3 4 3 C O EFFIC IEN T 15 16 42 10.767 -2.8 90 8.2 7 3 -l.U ? l STO . ERROR DF 3 24 MEAN S Q U A R E 172.953 175.056 S T :J . REG COEFF 1 3 .6 6 9 5 .9 4 1 0 .6 9 7 -0.041 0 .2 9 8 -0.3 15 Table 3 F T - 0 ..211 1.. 3 9 3 -1 ..465 P (2 RA TIO 0 .9 8 b TA IL 0 .8 3 5 0 .1 7 6 0 .1 5 6 P (TA IL) 0 .41511 REG RESSIO N T I T L E . . . . ................................................................R l l N l D A T A : I N C L U D E S D E P E N D E N T V A R I A B L E . ............................................................... BA P R E M I u M T O L E R A N C E ......................................................... ..... ........................................ 0 . 0 1 0 0 A L L DAT A C O N S I D E R E D AS A S I N G L E GROUP M U LTIP LE M U LTIP LE R R-SQUARE A N A LYSIS OF 0.4539 0 .2061 e.RROR OF ISSU E 1 3 .3 3 87 E.ST. VARIANCE SUN 3E REG RESSIO N RESID UA L C O EFFIC IEN T v a r ia b le IN TE R C E P T TERM PRIV A TE HC OTH R E S %LOAN7 FIX ED 2 STL). NE* s 15 16 18 35 4? 4 .6 6 3 -0.3 70 2 .6 8 3 1 0 .6 7 9 0 .9 7 B 0.141 -0.7 95 SQUARE' S V 7 2 . t>4 8 3 7 47 .65 9 : STO . ERROR OF 6 21 F MEAN S Q U A R E 162.108 1 7 8.455 S T U . REG COFFF 0 .3 0 2 15.0 2 9 7 .5 0 0 6 .5 1 5 0 .1 1 9 0 .7 6 3 -0.2 50 0 .0 3 8 0 .3 8 4 0 .0 3 6 0 .3 0 6 -0 .2 4 5 Table 4 T -1.2 23 0.1 7 9 1 .4 2 4 0 .1 5 0 1 .1 6 9 -1 .0 3 5 P (2 RA TIO 0.9 0 8 TA IL 0 .2 3 5 0.8 6 0 0 .1 6 9 0 .8 8 2 0 .2 4 8 0 .3 1 3 P (TA IL) 0 .5 0 7 8 2 R E G R E S S I O N T I T L F .................................................... ..................................R J N 1 D A T A : I N C L U D E S D E P E N D E N T V A R I A B L E ................................................................................. 54 P R E M U M T O L E R A N C E . . . ..................................................................... 0 .0100 A L L DAT A C O N S I D E R E D AS A S I N G L E GROUP M U LTIP LE M U LTIP LE R R-SQUARE A N A LYSIS OF 0 .3 5 1 7 0 .1 4 5 7 STD . SUM OF OF 5 22 OF EST ISSU E 13 .5 3 90 VARIANCE SQUARES 687.536 4032.671 REGRESSIO N RESID UA L V A RIA BLE IN TE R C E P T TERM SIN K IN G PRIV A TE HC FIX ED 2 ERROR NEW C O EFFIC IEN T 5 14 15 16 42 16.276 - l.i • 2 7 0 -0 .3 42 -0.741 i . d 29 -1.1 34 STD . ERROR MEAN S T D . REG COEFF -0.1 83 -0 .0 1 3 -0.011 0 .3 0 7 -0.351 0.6 8 3 1 2 .875 14 .6 8 7 6 .5 1 2 0 .7 3 0 Table 5 SQUARE 13 7.5 07 183.303 T -0.3 96 -0 .0 2 7 -0.0 50 1 .3 1 0 -1•560 p F RATI.O 0 .7 5 0 (2 TA IL) 0.6 9 6 0.9 7 9 0 .4 6 0 0.2 0 4 0 .1 3 3 P (TA IL) 0 .5 9 4 78 19 In the Table 4, eigenvector variables for regression model. variable holding 42 company company banks sistent with that The becomes significance. 5), the corresponding tenth sign even less the holding pay is higher findings of companies of from the eighteenth principal from the tenth seen two to b e When Debt Tables Table than total 6 shows that or equal to variance Five equal however, not is Defined of the one, to that R-square the 0.21 have achieved analysis debt to and for that their holding debt, (see is con footnote themselves. important variables important private the placements equations. regression The equation. Issue of Tables just variables, the statistical three most the N e w for positive Jacobs for to into variable regression all in do 1 through with eigenvalues over 77 not percent correlate 5. greater of the signi variable. is component when the components correlations ten and values entered implying float analysis again independent The to the six principal component due repeat dependent to three at are approach two m o s t Exclude accounting the 0.15. to to variable the contribute 10 the components The do the largest dummy over-leverage dummy in to Boyd, component The The sign, premia to six switches one positive 5 adds unstable through 6 ficantly with or highly new variables Results Table component. only risk tend regression is the Beighley, The 15 significant. coefficient's must the component of v a r i a b l e affiliation Its principal to of variance again 10 was with the alone defined risk in only is to premium Table greater 7 indicates, significant 0.16, component. considerably include than the new less issue. than DEPENDENT VARIABLE 5<+ PREMIUM TOTAL SUM OF SQUARES DEGREES OF FREEDOM MEAN SQUARE 4720.2031 27 . 174.82233 CORRELATION BETWEEN PRINCIPAL COMPONENTS AND DEPENDENT VARIABLE -0.08388 -0.16913 -0.03658 -0.01871 -0.06601 -0.04080 REGRESSION COEFFICIENTS OF PRINCIPAL COMPONENTS CONSTANT COMPONENTS (MEAN OF Y) 3.52589 -0.50949 -1.30267 -0.33371 -0.52537 -0.19735 -0.72147 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS INDEX OF RESIDUAL F-VALUES COMPONENTS SUM OF REGRESSION COMPONENT ENTERING SQUARES TO ENTER! «2 MODEL 1 2 3 4 5 6 4686.98823 4551.96875 4545.65234 4543.99609 4523.42576 4515.56641 0.18 0.46 0.31 0.22 0.19 0.16 0.18 0.74 0.03 0.01 0.10 0.04 0.0071) 0.035b 0.0370 0.0373 0.0417 0.0434 CONSTANT 4.8432 5 • 974 4.7368 5.4945 3.5527 3.1750 VARIABLES 5 TERM 10 CONVERT 11 CALL 3 ISSUE 14 SINKING 15 PRIVATE -0.0064 0.3576 -0.2043 *•0.0481 -0.2762 0.0031 -0.6349 -0.5771 •1.0254 -1.0043 -0.0435 -0.0081 -0.7494 -0.3108 -0.0093 -0.8931 -0.5862 -0.0243 -0.0089 -0.4934 -0.6242 -0.4937 -0.0222 -0.8440 -1.9418 -1.0819 -0.4714 -0.0120 -0.0151 -1.5197 -2.5011 -0.0168 -0.0203 -1.6212 -0.8673 -0.3943 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 30 %L0AN2 25 *>ASSET2 29 ^LOANl 34 %L0AN6 17 DIV RES 13 OTH RES 16 HC -0.13037 -0.08597 0.28165 0.48674 0.01106 -0.32472 -0.01*02 -0.05*91 0.40769 0.02929 -0.76541 -0.27236 -0.41763 -3.99553 -0.20074 0.55645 -0.48291 -0.75758 0.03106 -0.06063 -4.31083 -0.53476 -0.84650 0.45025 -0.063*5 -4.46607 -0.12986 0.03220 -0.43629 -0.41825 -0.06532 -2.57164 0.44734 1.15615 0.03497 -0.19294 -0.06567 -1.12016 0.03547 0.41815 1.24930 -3.02721 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 50 RETURN 51 MARG1N3 -0.02159 -0.02734 0.15857 0.02095 0.02611 0.21434 0.19967 0.01296 0.24501 -0.00227 -0.02158 0.18136 Table 6 35 %LOAN7 0.00637 0.00838 0.01082 0.01016 0.00778 0.01316 40 %INC1 0.02321 0.02637 0.00879 0.00699 0.03139 0.01373 *2 FIXED2 -0.03595 -0.02*57 -0.03*26 -0.02551 -0.01510 0.02610 DEPENDENT VARIABLE TOTAL SUM OF SQUARES DEGREES OF FREEDOM 54 PREMIUM MEAN square 4720.2031 27 • 174.62233 CORRELATION BETWEEN PHINCIpAL COMPONENTS AND DEPENDENT VARIABLE -0.08388 -3.16913 -0.03658 -0.01871 -0.06601 -0.00241 0.06001 0.10814 0.1016? 0.04949 -0.04080 0.13593 -0.16282 -0.02868 0.17623 -0.18964 -0.12603 0.40200 REGRESSION COEFFICIENTS OF PRINCIPAL COMPONENTS CONSTANT COMPONENTS (MEAN OF Y) -1.30267 -0.33371 3.52589 -0.50949 1.67175 3.27062 -1.73112 -0.52537 8.54583 -2.35074 -2.13871 2.58475 -30.60114 -1.99250 6.50381 -0.19735 3.44721 -0.73147 2.69648 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSIUN ON PRINCIPAL COMPONENTS F-VALUES INDEX OF PESIDUAL COMPONENTS SUM OF k EGRESSI ON COMPONtNT ENTERING SQUARE S mo del TO ENTtR 10 18 8 2 7 3957.41235 3787.65381 3641.06665 3506.04765 3380.91526 5.01 3.08 2.37 1.99 1.74 5.01 1.12 0.9 1 0.89 0.81 R2 U.1616 0.1976 0.2 286 0.2572 0.2837 CONSTANT 20.8629 -3.6392 8.0570 8.7112 9.1152 VARIABLES 10 CONVERT 11 CALL 3 ISSUE 5 TERM 14 SINKING 15 PRIVATE 12.5138 4.5806 -0.0368 1.3226 -0.2235 -2.9816 13.4784 16.3464 29.8527 -2.0853 0.2692 29.0793 -2.0997 29.1277 29.7607 8.1592 0.3279 13.3751 28.7834 0.3262 -2.1463 13.0023 7.4310 28.1352 27.7071 11.8487 29.5388 9.1748 -2.1216 0.3531 COEFFICIENTS OF VARIABLES 0bT0INfc.U FROM k EGRESSION ON PRINCIPAL COMPONENTS (CONTINUED) VARIABLES 29 *L0AN1 34 *L0AN6 30 %L0AN2 16 OTH PES 25 '■*ASSET2 17 DIV RES 16 HC 2.06674 -0.04326 -0 •045QS 4.32093 -0.89552 3.22770 7.36550 0.57295 -0.98582 30.94792 -2.07389 16.55031 -0.01240 -1.57960 -0.95292 -0.48637 31.96846 -0.0195b 16.32765 -2.27672 -1.84767 27.48618 -3.32379 16.86571 -0.81803 -0.99382 -0.00135 -1.79531 25.15753 -0.95511 -0.32966 -0.00797 16.11189 -2.12791 -6.16217 COEFFICIENTS OF VARIABLES OBTAINED FROM REGRESSION ON PRINCIPAL components (CONTINUED) VARIABLES 50 RETURN 51 MARGIN3 -0.66024 -0.23300 -1.09429 0.87217 0.64184 - 1.20282 -1.15453 0.82200 0.72097 -1.13643 Table 7 35 *>L0AN7 0.03932 -0.04250 -0.02969 -0.02768 -0.00990 40 %INC1 -0.11966 0.21557 0.05067 0.05383 0.16789 42 FIXED2 -0.46841 0.72987 0.63583 0.64721 0.30082 R E G R E S S I O N T I T L E . . . .......................................................... . •RON2 D A T A : E X C L U D E S D E P E N D E N T V A R I A B L E ................................................................................. 54 PREMIUM TOLERANCE ..................................................... 0.0100 A L L DATA C O N S I D E R E D AS A S I N G L E GROUP M U LTIP LE M U LTIP LE R R-SQIJARE A N A LYSIS OF SU-1 V A RIA BLE ERROR OF EST. ISSU E 13.6 7 53 VARIANCE OF REG RESSIO N RESID U A L IN TE R C E P T TERM PRIVATE" HC STU. 0 .2 2 lb 0.0491 ME* C O EFFIC IEN T 5 15 16 A. 7 6 6 -0.205 - 3.991 •4 8 1 SQUARES 231.892 4 *8 8 .3 1 2 s TO. ERROR OF 3 24 MEAN 8 T D » Rt.G COEFF 0.2 9 7 14.169 6 .5 6 8 -0 .1 3 8 -0 .0 5 7 0 .1 6 1 Table 8 SQUARE 7 7 .2 9 / 187.013 F T - 0 , >6 3 9 - 0 1. 2 6 2 Oi, & 0 2 P (2 RA TIO 0 .4 1 3 TA IL 0 .4 9 7 0 .7 6 1 0.4 3 0 P (TA IL) 0 .7 4 4 94 R E G R E S S I O N T I T L E ...................................................................................... RIJN2 D A T A ; E X C L U D E S D E P E N D E N T V A R I A b L F ................................................................................. 54 PREMIUM TOLERANCE ............................. ..... ...............................................................0 . 0 1 0 0 A L L DAT A C O N S I D E R E D AS a S I n G L E GROUP M U LTIP LE M U LTIP LE R R-SQUARE A N A LYSIS OF 0 .4 3 3 6 0 . 1 FRO ERROR OF EST. ISSU E 13 .5 0 97 VARIANCE SU* REGRESSION RESID U A L V A R IA B LE IN TERC EPT TERM CALL PRIV A TE HC OT H R E S FIX ED ? S Ti). MEtf OK S Q U A R E S 387.448 3832.759 C O EFFIC IEN T 5 11 IB 16 IP 42 13.610 -0.336 3.3 2 9 -1.0 80 8.6 9 2 2 .0 7 4 -1 .2 89 STD. ERROR OF 6 21 « EA N S T D . REG COEFF -0 .2 2 7 0 .1 1 3 -0 .0 1 5 0 .3 0 6 6 .8 5 5 15 .0 8 9 6.0 3 7 6 .0 9 0 .7 3 2 0 .2 *8 0 .0 7 7 -0.3 78 Table 9 SQUARE 147.908 182.612 T -1.1 01 0 .5 5 9 -0.0 72 1 .0 0 8 0 .3 2 2 -1.7 60 F RA TIO 0.8 1 0 P (2 TA IL) 0 .2 3 3 0.58.2 0 .9 4 4 0 .3 2 6 0 .7 5 1 (J.093 P (TA IL) 0 .5 7 3 5 2 R E G R E S S I O N T I T L E ...................................................................................... R U N 2 D A T A ! E X C L U D E S 54 PREMIUM D E P E N D E N T V A R I A B L E . . . ................................................................ TOLERANCE . . . . . . 0.0100 A L L DAT A C O N S I D E R E D AS A S I N G L E GROUP M U LTIP LE M U LTIP LE R R-SQUARE A N A LYSIS OF ERROR OF IS S U E 13.9 5 45 EST. VARIANCE SUM REG RESSIO N R ESID U A L V A RIA BLE IN TERC EPT TERM SIN K IN G PRIV A TE HC STD . 0.2 2 6 2 0.0511 NEW .C O EFFIC IEN T OF S Q U A R E S 241.433 4478.773 STO . ERROR MEAN OF 4 23 SQUARE 6 0 .3 5 8 194.729 S T D . RtG COEFF PA TIO 0 .3 1 0 R (2 TA IL) 2.261 5 14 15 16 -0.0 67 -2 .9 1 3 -3 .0 5 8 5 .0 8 2 0 .6 9 1 13.161 15 .0 6 0 6 .3 1 4 ■0.046 •0.108 •0.044 0.1 8 3 Table 10 ■0.098 • 0.221 •0.203 0 .8 0 5 0 .9 2 3 0 .8 2 7 0 .8 4 1 0.4 2 9 P (TA IL) 0 .8 6 8 33 20 A comparison of Tables 2 and 7 indicates that empirical results for this model may differ significantly according to whether one defines debt to include or exclude the new issue. Examining the two tables that transform the regression coefficients of the tenth principal component back into regression coefficients for the original independent variables (that is, looking at "coefficients of variables obtained from regression on principal components," the first line only), one notes that two variables, issue size and dividend restrictions, change sign and that several variables’ coefficients experience large changes in size— up to two orders of magnitude for variable 10. Although the sample is too small to generate stable results, this finding indicates that strikingly different empirical results emerge depending upon whether one tests an optimistic or pessimistic model of bank capital structure. Given the substantial difference in R-squares between Tables 2 and 7, on the basis of this limited test one could conclude that investors subscribe rather more to the pessimistic view. Tables 8, 9, and 10 present ordinary regression results paralleling Tables 3, 4, and 5. The actual variables entering regression equations differ somewhat from those of Tables 3-5 due to differences in the eigenvectors for components ten and eighteen. No variable in any of these three regressions has a coefficient with both the expected sign and even marginal statistical significance. VI. Summary and Conclusions This paper has developed a model of market pricing of new debt securities issued by commercial banks. Two versions of the model were 21 tested, one including and the other excluding the new issue of debt. The two sets of results did display some differences, but overall statistical performance of the model was poor. It was found in both versions of the model that the first six principal components, encom passing about 77 percent of the variance of the independent variables, were not correlated with risk premium, the dependent variable. Only the tenth principal component, which accounted for just 2.2 percent of the variance of the independent variables, was significant in explaining risk premium. Based upon results of the principal components analysis, selected independent variables were entered into a standard multiple regression equation. This analysis was undertaken in the attempt to utilize the information gained from principal components analysis while avoiding some difficult problems of interpreting results. Results of re gression using ordinary least squares and the original independent variables were generally negative, a result not unexpected from the principal components analysis. The purpose of this paper was to address the question, Do securities markets make use of the information currently available in pricing new issues of debt securities by banks? Empirical inference was con siderably hampered by the small sample size. No clear answer to the question posed can be given based upon empirical results reported. If one considers, on an intuitive level, that the total variance of a data set is a measure of its informational content, then the results of principal components regression indicate 22 that not very much information is being used by the market. The results of ordinary multiple regression indicate that the useful information is not captured by just a few standard financial variables. Nonetheless, a single linear combination of eighteen variables did produce R-squares of 0.20 and 0.16, indicating that some information from the Reports of Condition and Income are used by financial markets. with a larger sample is warranted. Further testing FOOTNOTES *The author wishes to acknowledge helpful comments from several of his colleagues, especially from Bob Laurent. Thanks are due to Nancy J. Peterson for extensive research assistance and to Robert W. Keyt for data processing support. ■^George Tucker, The Theory of Money and Banks Investigated, Reprints of Economic Classics (New York: Augustus M. Kelley, Bookseller, 1974), originally published in 1839, p. 210. ^Frank Wille, "The FDIC Views Questions of Capital Adequacy", ad dress by the Chairman of the Federal Deposit Insurance Corporation before the National Correspondent Banking Convention of the American Bankers Association, San Francisco, California, November 6 , 1973, p. 1. ^John E. Sheehan, "Bank Capital Adequacy— Time to Pause and Reflect," Remarks of John E. Sheehan, Member, Board of Governors of the Federal Reserve System before the National Correspondent Banking Conference of the American Bankers Association, San Francisco, California, November 6, 1973, pp. 3, 11. ^Frank Wille, op. cit*, p. 2. ^Donald P. Jacobs, H. Prescott Beighley, and John H. Boyd, The Fi nancial Structure of Bank Holding Companies, A Study Prepared for the Trustees of the Banking Research Fund, Association of Reserve City Bankers, 1975. See also, by the same authors: "Financial Structure and the Market Value of Bank Holding Company Equities," Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1975, pp. 61-72; and "Bank Equities and Investor Risk Per ceptions: Some Entailments for Capital Adequacy Regulation," Banking Research Center, Northwestern University Graduate School of Management. ^Ideally, one would like to compare the risk security with a risk free security having the same coupon, as well as the same date of maturity, so that one bond does not sell at a substantial discount relative to the other. In practice it is not possible to achieve this comparability since one must frequently compare a newly issued bank security with a Treasury security issued several years previously but maturing on nearly the same day. ^By the provisions of Regulations Q and D that exempt capital notes from interest rate ceilings and reserve requirements, capital notes must be unsecured. Therefore, this characteristic, which on a priori grounds would be considered important, was excluded. o In this connection, it may be well to mention a controversy in the financial literature on banking. The argument concerns whether or not to include interest on deposits in fixed charges. The approach used here is to exclude interest on deposits, on the two related grounds that interest costs are quite flexible for all deposits except long-term certificates and that other costs of maintaining deposits are much more rigid than in terest costs• See David C. Cates, "Bank Analysis for Bond Buyers," Bankers Monthly, September 15, 1964, pp. 25 et seq. ^Richard V. Cotter, "Capital Ratios and Capital Adequacy," National Banking Review, Vol. 3 No. 3 (March 1966), p. 335. •^"Core Deposits" are demand deposits, apart from correspondent bal ances, and savings deposits, less investments. This measure of the sta bility of deposits to support lending is widely used by financial analysts; see, e.g., Harry V. Keefe, Jr., "Capital Funds in the Banking System— No More Free Lunches for Borrowers," an address before the association of Reserve City Bankers, New York, New York, February 3, 1975. ^'*‘Cates, op. cit., pp. 21-22. 12W. Braddock Hickman, Corporate Bond Quality and Investor Experience, National Bureau of Economic Research, Studies in Corporate Bond Financing, Volume 2, (Princeton: Princeton University Press, 1958), p. 396. •^David C. Cates, "Bank Debentures, Leverage, and Debt Capacity," Bankers Monthly, November 15, 1963, p. 48. •^Lawrence Fisher, "Determinants of Risk Premiums on Corporate Bonds," Journal of Political Economy, Vol. LXVII No. 3 (June 1959), pp. 217-237. “^Peter E. Sloane, "Determinants of Bond Yield Differentials— 19541959," Yale Economic Essays, Vol. 3 No. 1 (Spring 1963), pp. 3-55. -^Richard H. Pettway, "Market Tests of Capital Adequacy of Large Commercial Banks," Journal of Finance (forthcoming, June 1976). l^H. Prescott Beighly, John H. Boyd, and Donald P. Jacobs, "Financial Structure and the Market Value of Bank Holding Company Equities," Proceed ings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1975, pp. 61-72. l^Banks which retired an outstanding note and issued a new note of equal or smaller size will not be picked up by this procedure. •^Calculations were performed using the Biomedical Computer Programs, series BMDP. Regression on principal components is program BMDP4R; multi ple linear regression is program BMDP1R.
@FedFRASER