The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
A Series of Occasional Papers in Draft Form Prepared by Members of the Research Department for Review and Comment. 76-2 Advertising for Demand Deposits Chayim Herzig-Marx Federal Reserve Bank of Chicago ♦♦♦♦♦♦♦♦♦ I I I x T i i r >♦♦♦♦♦♦♦♦<sitiius s tttm a R tttttfl ♦♦I Research Paper No. 76-2 ADVERTISING FOR DEMAND DEPOSITS By Chayim Herzig-Marx Department of Research Federal Reserve Bank of Chicago The views expressed herein are solely those of the authors and do not necessarily represent the views of the Federal Reserve Bank of Chicago or the Federal Reserve System. The material contained is of a preliminary nature, is circulated to stimulate discussion, and is not to be quoted without permission of the authors. I. INTRODUCTION* This paper explores the determinants of advertising for demand deposits. From a theoretical point of view, the market for demand deposits is particularly interesting. As an institutional feature, price competition for demand deposits is prohibited by statute and by regulation. Since prices cannot clear markets, some other mechanism must be found. Advertising could play an important role in adjusting supply of and demand for deposits. The lack of attention to advertising by commercial banks must be attributed to a distinct lack of data. Problems in defining inputs and outputs and the level of sales in the banking industry are quite severe, so that banks are excluded from inter-industry studies using Internal Revenue Service data. Most micro data until recently were held confidential, although even Report of Income statements do not contain a breakdown of expenses. This basic data problem was resolved for this study by using information obtained from the Functional Cost Analysis program sponsored by the Federal Reserve System. Functional Cost data have been used extensively in the past by researchers within the Federal Reserve System to study bank costs and production functions. The data themselves, however, are confidential and cannot be reported. Within the mainstream of research in industrial organization, advertising has been found to be an interesting feature of firm conduct. As such, this form of behavior may be conditioned by the firm’s environ ment. Empirical work to be reported below provides some support for the structure-conduct hypothesis. *Jan Gigstad and Robert Keyt supplied the extensive programming necessary to compile the data and prepare them for analysis. 2 The outline of this paper is as follows. Section X presents a model in which expenditures on advertising are adjusted to equate the supply of lendable funds to the demand for bank loans. Section II discusses the sample of banks used to estimate the relationship outlined in Section I and details the construction of the variables of the model. Section III reports empirical results, and section IV concludes the paper. II. A Model of Deposit Advertising Each bank has outstanding at any point in time certain commitments to extend credit to its clients. The maximum amount of credit and the rate at which credit will be extended are the most important character istics of the credit line, along with the fees or compensating balances which pay for the line. Since the credit line can be exercised at the discretion of the loan customer, each bank must forecast its expected loan "takedown” and be prepared to lend however much its clients want. The forecasting process can be thought of as follows. The bank estimates the probability that each credit line outstanding will be exercised. Then the bank estimates the expected size of the loan takedown, given that the credit line is exercised. pected size of the loan. The product of these two is the ex Summing over all credit lines outstanding gives the total expected loans by that bank at the future date. It is assumed that'banks forecast one year into the future.^ Notwithstanding the growing importance of non-deposit sources of funds, expansion of deposits will continue to be a major source of ■**It does not complicate the anlysis if it is also assumed that banks forecast some expected increase in loans from sources other than take-downs of credit lines, since other borrowers will respond to the same economic forces as owners of credit lines. 3 increased lendable funds for commercial banks ► Although the shift from demand to time deposits is certain to continue, demand deposits remain an important source of funds for banks. demand deposit advertising. This paper considers only Time deposits are considered to be suffi ciently different in nature to justify separate treatment.2 The loan forecast indicates to the bank the quantity of lendable funds it will need. By forecasting deposits the bank can estimate its expected lendable funds. Advertising enters the model through its effect on future deposits. We posit that advertising can be used to increase deposits at the bank, and that the level of advertising expen ditures can be determined so as to equate expected supply of lendable funds to expected loan demand. Implicitly the optimal level of adver tising will be a function of the determinants of expected loans and expected deposits. To estimate the conceptual model outlined above would require a com prehensive treatment of bank portfolio decision, complete with an ex plicit expectations generator. For the limited purposes of this study, such as undertaking is not worthwhile. Instead, a structure-conduct empirical model of the type commonly encountered in industrial organiza tion research will be used. An elementary forecasting procedure will be used for loans and deposits. o Some preliminary analysis of the type reported in this paper was conducted for time deposits with the basic result that this model had virtually no explanatory power. This probably results from two factors. First, most time and savings accounts are highly homogeneous commodities, so that price is the most important factor to the virtual exclusion of market structure or other characteristics. Second, for deposits which exceed the insurance limit, the adequacy of the bank’s capital, or depo sitor expectations of the probability of the bank’s failure, assume great significance and interact strongly with the rate paid on such deposits. Thus the present model is not adequate to explain advertising for time deposits. 4 Four factors are important for forecasting loan demand: (A) the scale of the bank’s operation, meaning the total number, and dollar value, of loan commitments outstanding; (2) the probability of loan commitments being exercised, which is a function of total financing needs of bank clients and alternative borrowing costs; (3) the likely size of loan takedowns, which is probably also a function of bank scale (large customers need to deal with large banks); and (4) expected in crease in loan demand other than from commitments. be used to represent these four factors: Three variables will deposits (as a measure of bank scale), current loan yields (net of cost of money), and the historical growth of loans in the bank’s market. The forecast of future deposits depends on three factors: (1) present deposits; (2) the expected growth rate of deposits; and (3) a confidence interval around the growth expectation. The second and third factors are represented by the historical growth rate of deposits and by the standard deviation of deposits around their historical growth trend. The efficacy of advertising in attracting deposits is conditioned by the nature of the individual bank and the characteristics of the market in which it is located. Four individual bank characteristics will be tested for their influence on advertising: market share, wholesale or retail orientation, age (years since founding), and holding company affiliation. The four aspects of market structure considered most impor tant are concentration, the conditions of demand for loans and supply of deposits (both discussed above), regulatory restrictions on branching (availability of substitutes for advertising), and the urban or rural nature of the market. 5 The larger the market share of a bank, the greater the proportion of the inter-industry effects of advertising the bank can expect to internalize; and consequently the larger its advertising budget can be expected to be. The relationship between market share and intra-industry effects is much less clear. It seems likely that firms with larger market shares feel more susceptible to inroads from other banks’ advertising. This may induce leading banks to spend more on advertising, as a defensive device. It may, on the other hand, induce leading banks to spend less on advertising, so as not to promote advertising by other banks. On the whole, it is likely that inter-industry effects outweigh intra-industry effects and that market share has a positive relationship to advertising. Wholesale and retail banking differ significantly in terms of the bundles of products offered. Wholesale banking is oriented toward large customers, with the provision of a credit line as the major product or service provided. The explicit or implicit price of the credit line (commitment fees or compensating balance requirements) is likely to be the most important means of competing for such accounts. The proliferation of personal banking packages in recent years attributes to the many ways of competing for the deposits of individuals. Non-price terms predominate here, and advertising is likely to play a much more important role. As a consequence, banks which are oriented toward wholesale customers are likely to advertise less. Even if both wholesale and retail banks do advertise, the media selected to carry advertising messages are unlikely to be the same for both. An important feature of this advertising process is that corporate treasurers will always be actively seeking the lowest prices for credit lines. This search for information on the part of bank customers will 6 also lower bank advertising expenditures. Because wholesale banks are likely to have considerably more large accounts, we will include average account size in the regression equation to control for the wholesale/retail dimension. New firms in general must advertise to make known their existence and to attract a loyal clientele. In commercial banking, the importance of the customer relationship reinforces this motive to advertise and makes it very likely that younger banks will advertise considerably more than older ones. In order to use a continuous variable, the age of the bank (years, expressed in decimals, since the bank opened for business) in reciprocal form will be entered into the regression model. Unfortunately, this variable is unlikely to have any statistical significance for our sample of banks. The data requirements for the deposit growth and de posit variability variables forced the exclusion of any banks fewer than eight years old. This will probably render the age variable insignificant. The effects of holding company affiliation on bank decision making are still quite unclear. Available evidence on changes in bank operations following holding company affiliation are only tangentially related 3 to advertising. "Other operating expenses," which include advertising, appear to rise significantly for banks after they are acquired by holding companies. On the other hand, service charges on demand deposit accounts fall slightly but not significantly. The main effect expected is the centralization in the parent company of the advertising function, on the assumption that bank subsidiaries can benefit from association with the parent name. Such centralization can also be seen as eliminating o Samuel H. Talley, "The Effect of Holding Company Acquisitions on Bank Performance," Staff Economic Studies #69, Board of Governors of the Federal Reserve System. 7 potentially wasteful duplication of effort and taking advantage of what ever economies of scale may exist in advertising. On the whole, we expect banks affiliated with holding companies to advertise somewhat less than other banks. As in any structure-conduct-performance test, the definition of the market is crucial to the analysis. bank in its local market. In this paper, we focus on the This viewpoint is similar to that adopted by the Board of Governors of the Federal Reserve System in its deliberations on mergers and holding company acquisitions and is also in close keeping with the nature of the sample of banks (see section II). Specifically, the market for any bank is assumed to be exactly coterminous with the county in which the bank or its head office is located. While this usage is fairly common, the restrictiveness of this assumption should not be underestimated. Appendix I discusses the applicability of this definition for the sample employed. Concentration is measured by the numbers equivalent, the reciprocal of the Herfindahl index. To take account of the possibility that local market shares are distorted due to large demand deposit balances of large firms located in other banking markets, a Herfindahl index was computed only for accounts of less than $1,000. This concentration index accords better with the local nature of the banking market and consistently yields better regression results. All results reported below, therefore, use the numbers equivalent based on accounts of less than $1 ,000. The conditions of supply of lendable funds and demand for bank loans were discussed above in conjunction with the forecasting process. Establishing branch offices is one of the more important alterna tives to advertising as a means -of increasing lendable funds. Branches 8 enable banks to compete by offering locational convenience and by enabling the bank to follow shifts of population. With the exception of Illinois, all states within the Seventh Federal Reserve District allow some form of branching within the county, and some states allow branching into con tiguous counties. Branching restrictions in Illinois are thus consider ably more stringent than in other district states, and this qualitative difference will be taken into account. The urban or rural nature of the bank’s market may be important for three reasons. First, there may be important cost differences between the two types of regions. Especially, some advertising costs can be expected to be functions of the distance over which messages are propagated. Advertising will then be cheaper in urban areas since population densities are higher. Second, more developed transportation systems in urban areas may reduce the economic distance between banks, heightening competition and promoting greater advertising expenditures. Third, the significance of location may differ considerably. Conven ience and accessibility are important motifs in bank advertising. Lo cational differences may be more important in cities, with their locationally specialized transportation networks (e.g., in most cities the best means of transportation are those which go into the downtown area.) Traffic congestion is another factor increasing the importance of loca tion in urban areas. In addition to these conflicting aspects of urban- rural location, it is likely that different advertising media predominate in the different areas, making cost comparisons extremely hazardous. Thus the net effect of the urban or rural nature of the market cannot be pre dicted, but is likely to be insignificant. 9 Size-related bias is to be expected in any econometric relationship estimated with micro data. Larger banks will spend more on advertising because their total budgets are larger. To correct for this bias, the dependent variable will be deflated by thousands of dollars of demand deposits and specified as an expenditure intensity rather than as dollars of expenditure. The form of the estimating equation is (1) A/D = bn + b-LYLD + boLGR0 + b DGRO + b DVAR + b ^ + b l , U 1 Z 3 4 — 0— — 6— where D is demand deposits, LYLD is the net yield on loans, LGRO is the historical growth rate of loans in the county, DGRO is the historical rate of growth of deposits of the bank, DVAR is the variability of deposits around their growth trend, S is a vector of market structure character istics (other than LGRO), and I is a vector of individual bank charac teristics. III. The Sample and Construction of Variables The advertising relationship discussed in section I was esti mated with micro data drawn from a sample of 160 Seventh Federal Reserve District commercial banks which participated in the 1972 Functional Cost Analysis program sponsored by the Federal Reserve Bank of Chicago. Since participation in the program is voluntary and the data are held con fidential, disclosure of information on a bank by bank basis is not possible. Expenditure data are for the year 1972. The purpose of the Functional Cost Analysis program is to assist banks in maintaining accurate and useful cost accounting of their operations. For the most part, participating banks are too small to maintain cost accounting departments of their own, but it will be seen IQ that a considerable range of firm size is still encompassed in the sample. Because the program is voluntary, and because the output of the program is of such benefit to the participating banks, we can have considerable confidence in the overall quality of the data. The cost accounting framework breaks down all bank operations into separate functions, such as demand deposits, time deposits, real estate loans, personal loans, trust services, data processing services, etc. For each function, participating banks allocate their expenses to the best of their abilities. Any expenses which cannot be allocated directly to functions are reported as a residual. This residual, by type of expense, is allocated to functions indirectly by the functional cost program itself. Since these indirectly allocated expenses tend to be overhead or fixed costs, attribution of them to specific functions is not completely accurate. In this paper, the advertising costs we seek to explain are only those which are allocated directly by participating banks. The sample of banks displays considerable diversity, given that all are located within the Seventh Federal Reserve District. Size as measured by total deposits ranges from under $5 million to well over $1 billion, with a mean size of $32 million and a median size of $45 million. Sixty-six banks are chartered in Illinois, 20 in Indiana, 29 in Iowa, 21 in Michigan, and 24 in Wisconsin. Ninety-four of the banks are located in Standard Metropolitan Statistical Areas, and twenty three are affiliated with bank holding companies. The oldest bank was chartered in 1848, and the youngest opened for business in 1964. This sample thus displays considerably more diversity of size 11 than most samples of industrial or commercial firms, yet because most banks are rather small there is considerable homogeneity in terms of products and services. In addition, we are not forced to define the firmfs market as the entire country, since information (from Reports of Condition and Income and Dividends) is available on virtually the entire universe of commercial banks from which to construct market structure variables. Net yield on loans (LYLD) is calculated by summing gross earnings on all loan categories and substracting total expenses for all loan categories (direct and indirect expenses). Dividing this difference by total loans gives gross yield on loans. From gross yield is sub tracted what is termed "cost of money," or an average cost of all lendable funds. Multiplying the result by 100 gives net yield on loans. Historical rate of growth of loans in the market (LGR0) is calculated as the 1971 to 1968 ratio of all loans made by all banks in the county. Historical rate of growth of deposits (DGR0) was calculated separately for each bank in the sample. A compound growth rate was fitted to annual observations on total demand deposits by simple re gression. Deposit variability (DVAR, multiplied by 100 for scaling purposes) was calculated as the standard error of the estimate divided by the mean value of deposits. Data for years 1964, 1965, 1968-1971 were used. . All data on market shares (SHARE) were calculated using the 1970 Summary of Deposits survey. Average account size (ACSIZ) was calculated as total demand deposits (in thousands of dollars) divided by total number of demand deposit accounts. 12 Age of the bank (AGE) was calculated relative to December, 19J2, which is the reporting date for FCA data. Years and months since the bank opened for business, or since the bank was chartered if the opening date was not known, were converted to years and decimals. The multiplicative inverse was taken to yield the age variable. Holding company status (HC) is represented by a simple dummy variable taking on the value of unity for any bank affiliated with a holding company, zero otherwise. Herfindahl indices of concentration (CONCE) were computed using market shares based on the 1970 Summary of Deposits survey for total deposits and for deposits in accounts of less than $1,000. To represent branching restrictions (BRNCH) a simple dummy variable was used, taking the value of unity for all banks located in Illinois, zero otherwise. Several variables were tried to represent the urban or rural nature of the bank’s market. A dummy variable (SMSA) taking the value of unity if the bank was located in an SMSA and zero otherwise, population density (PDEN) in thousands of persons per square mile, and percent of the popula tion living in urban areas (PURB) were all tried. The total number of Seventh District banks participating in FCA in 1972 was 213. Of this number, 34 banks made no direct allocation of their advertising and were excluded from the sample on this ground. Of the remaining banks, 19 had to be excluded for various reasons, primarily lack of time series data on deposits for the construction of DGRO and DVAR. Mergers and consolidations accounted for the lack of consistent time series data. One or two banks were excluded because 13 no non-bank financial intermediaries were operating in the market. Since the presence of non-bank financial intermediaries is required for any inter-industry effects to be possible, and because so very few potential sample banks did not fulfill this requirement, it was de termined to drop them from the sample rather than take account of this status with a separate independent variable. IV. Empirical Results Some estimates of the advertising relationship are given in Table 1. Equation 1 includes all variables specified above. Signs of three variables, DGRO, HC, and AGE, are counter to predictions, but none of these coefficients is significant. In fact, the only variable whose coefficient is significantly different from zero is DVAR, deposit vari ability. The only other variable whose coefficient exceeds its standard error is BENCH, the branch banking dummy. Equation 2 deletes three individual characteristics and one market structure variable which added virtually nothing to the explanatory power of the model. (LYLD, which also adds almost nothing to the model, is retained because of its role in the forecasting process.) Market concen tration is now significant at the 10 percent level, and the BRNCH and LGRO coefficients exceed their standard errors. Most standard errors are smaller than in equation 1, indicating a reduction in collinearity. Equation 3 further deletes DGRO and ACSIZ, with the result that LGRO achieves 10 percent significance and most standard errors fall. Finally, deleting LYLD in equation 4 raises CONGE to 5 percent significance and BRNCH to 10 percent. 14 The the coefficients four equations. siderably, No Further 4. Each SMSA, of was continued pected to b e equation SMSA is of PURB is PURB high zero. is SHARE, 2, see if in one the HC, or more the are age con altered. ACSIZ, of on equation AGE, and them might All across excluded conducted based equation. although 4, using Table In PURB population be these variables variable 1, and PDEN density as took on is entered Its sign but the coefficient CONCE is alternatives 1. addition, entered is into negative, variables are specified. is h i g h l y measured with other the m o d e l stable collinearity with probably equation: DGRO, increase the specifications variables, 1 equation icance while most really collinearity with was to stability and LGRO here, variables Table equation from other its CONCE reported conducted 1 of coefficient what reasonable the ex unexpectedly SMSA. along with negative, is to as w a s not the the sign significantly switches to the sign. When its of occur when insignificant, in Table different "wrong" show sign. variables of not fewer BRNCH greater individually Testing was In to excluded with and coefficients changes analysis, added significant due sign the DVAR The probably variables. of the less can be size, The other coefficients collinear in (equations independent become from it PURB Table with insignificant. following 2) regardless (and PDEN) variables. emerges the 3, significant problem with error, seen and 2 and high That Since signif PURB regression 15 PURB - 43.7 - .386 D G R O (9.27) (..357) - 1.62 L G R O (3.27) + .820 D V A R (2.34) + 1.94 L Y L D (1-81) - + .846 C O N C E 0190) 2.89 B R N C H (2.73) + .201 S H A R E OQ27) - 4.45 H C (3.24) +70.8 A G E +22.7 SMSA (76.3) (2.36) To judge by regression PURB and t-statistics equation SMSA the results and SMSA would tation of are, is m o s t likely improve results. While the .973 A C S I Z (.644) R-square this in addition, obtained. the from + equation, including to CONCE, good affect substitutes. PURB SHARE, These in = .666 the a n d HC. are, including PURB in place of CONCE, statistical fit, it w o u l d confound in fact, SHARE, interpre 16 V. Summary and Mixed results advertising The to growth and and whose of should yield loans prising described on is for loans this m o d e l lendable empirical is interval enter expected the coefficient confidence obtained process from net growth were equate forecasting confirmation in Conclusions in Section testing. fail to consistently so strongly result. Further the in study I does Both of the not forecast. That deposit of The DVAR, equation strong deposit significance, one variable representing this is variability demand. receive rate is use loan statistical significant regression banks expected significant. deposit the to exhibit only marginally around funds in w h i c h the variable a rather appears sur to b e order.^ The from the fundamental the m a rket are the In little sum, globally, representing condition of v e r y hypothesis concentration variable dummy variable senting structure-conduct the of performance however, the inability demand. assistance and, explaining this empirical model results a lesser to b r a n c h Individual in of to receives bank are extent, and LGR0, from repre characteristics demand is support deposit spotty. advertising. Judged sufficiently good 2 'V/ (R ^ .25) accounting ^In 76-3 for this another individual paper rate of multicollinearity is a reasonable complex start such an As those market share, growth poses a results deposit of been made Variability," investigation representing market that has in phenomenon. ("Long-Run Deposit just banks. finding concentration, especially that highly on variables important by consider (forthcoming)) gressed of to is Staff Memoranda undertaken. structure and relate to this variability is explained holding deposits. substantial company problem in study, it the is is the quite well affiliation, Therefore, DVAR characteristics and clear that present work. re APPENDIX I COUNTIES Three broad a firm's market theory must be classes for taken effects (3) availability types of To firms put of view, to b e the the these area attempts of to susceptible a to A the grave sample Achieving to deal with tively which tion definition might large this a are less are less sample problem. The individuals in the have sample; definition It adequate for not two are and is Therefore, therefore in economic (non-bank) not clear, subject to calculation, market the definitions preferred. are are only preferred, In be geared case agents therefore, them. is quickly firms, of which a market this both to b e geographically of point requires transmitted must other possible are Theory types certain biases as a practical investigation, and respects. is from definitions deposits. exhaustive extensive perspective under size economic definitions the market influence incorporate not is that that apparent. practical way which are addition, since rela markets aggregation likely. Availability of in case. the into one which market of of sample. Market is implications both influences in many definition included competitive advertising, the can be important. demand and the any different conditions the effect. geographically biases as firms influence least competitive risk, used which is over which significantly single market of factors theory MARKETS (1) (2) strongly three attract purposes: account; comprise substantial type differ types which economic and with to on into GEOGRAPHIC considerations empirical profound data of AS usual data is, unfortunately, Normally, either the m o s t the m a x i m u m important considera sample known is 1-2 in advance and accommodate one or a are as a the banks local criteria tions the definition sample, or few definitions Most kets the set are in this of to requirements While small, counties and 80 percent originate, service service (o r some an is that size of its area. made application to the regulatory culties of service areas a need any This and for the political and a simple to locate feature Another demographic is other by only is geographic other a bank, market In be their m a r arbitrary banks for defini defined figure) of definition. itself only must provide which had a merger addition, the considerable, smallness, all of counties are well-defined decision and the b a n k s especially welcome data which of the b a n k would clear-cut benefit two authorities They is latter to or diffi since boundaries. relative as m a r k e t s . matter area sample. variables the that smallness, Therefore, b a n k ’s m a r k e t centration measures. economic follow exhaustiveness relatively market. not characteristics determining is candidates constructing market Besides desirable be can be met often-used on acquisition would implicitly implies satisfy the is areas which relative information formal the are In particular, deposits drawback data is m a d e the market. sample circle within which t h e b a n k ’s the m a r k e t exhaustiveness possible. The basic other in nature. above, of counties available in for is for in any have areas, so addition that it particular calculating the w e a l t h such other con of regions. Against these advantages for data collection and sample determination, one must balance the possibility that economic activity does not follow county lines. This appendix presents evidence that, for the Seventh Federal Reserve District, counties are reasonable approximations to economically relevant markets for banking activities both of individuals and of firms. 1-3 A reasonable business at supposition a bank office residence. County markets people work in the m e a n force and sented in county standard residing calculated work the in for the a person either near his are of county counties sample. near where that they then their of in These place suitable state its the data transact of the show and a of strong Residing Counties Seventh in the two those his extent table below The of that shows the labor statistics counties tendency Percentage in or ne a r percentage for banking for are repre people to live. Number State to The employment. state his of w o r k only residence. deviation by the all is District sample in total in of Labor County Force of Employment sample mean total s.d. mean s. d . Illinois 29 58 74.6 11.9 68.6 12.3 Indiana 13 64 79.2 13.6 66.8 15.7 Iowa 19 97 83.5 9.9 81.2 8.1 Michigan 20 67 75.4 13.8 70.1 13.7 Wisconsin 17 46 77.8 11.5 76.8 9.6 SOURCE: 15 Table (Illinois), As pal 89 pal the banking While this of firms, by the all most same banks county, relationship survey was •'■Robert F. Ware 17 the I, data Federal a bank that and are firms in the all less in E. direct. A 1973 maintained (87 of banking that relation changed in the same which is not in Survey their percent bank A Parts (Wisconsin). Ohio Cleveland^- indicates county Reserve 51 study of firms Duro, in O h i o , F e d e r a l 1970, and Bank same another in Ohio, Lorraine Population: (Michigan), a principal nearly selected of 24 Reserve added conducted and Census (Iowa), respondent Firm-Bank Relationships 1974. Volume relation with Furthermore, in firms percent banking bank 119, (Indiana), for business manufacturing that 16 their princi in 1969). chose a princi county. the D i s t r i c t of of M a n u f a c t u r i n g Bank of Cleveland, June 1-4 our sample and does haye somewhat different economic features from the Seventh District (notably the lack of a money-market the size of Chicago), it is considered that the results are representative of Seventh District experience also. APPENDIX II Simple Correlation Coefficients for Independent Variables DGRO DVAR DVAR CONCE SHARE LGRO LYLD BRANCH HC AGE SMSA .684 CONCE -.014 - .1 0 2 SHARE -.254 -.176 -.652 LGRO .072 -.029 -.258 .176 LYLD .003 - .1 1 2 -.344 .409 .107 -.051 -.147 .528 -.387 .032 -.004 .016 -.024 -.062 .128 -.177 .022 -.307 -.106 - .1 1 2 .216 .036 -.167 .110 .092 .202 BRANCH HC ACCTSIZE - ACCTSIZE AGE .720 .504 .145 -.148 -.108 - .1 2 0 .040 .020 -.191 SMSA .002 .047 .414 -.372 -.236 -.158 .186 .054 .154 .160 MEAN 5.496 .5256 9.784 22.72 2.065 2.249 .4125 .1438 2241. .0212 .5875 ST. DEV. 5.158 .6366 8.881 16.62 .3510 .6658 .4923 .3508 1788. .0217 .4923 Table 1 Advertising Intensity Regressed on Selected Variables Standard Errors in Parentheses EQUATION 1 2 3 4 DGRO .008 (.010) .005 (.008) DVAR .304*** (.067) .300*** (.066) •330*** (.047) .328*** (.046) CONCE -.004 (.005) -.006* (.005) -.006* (.004) -.007** (.004) SHARE .001 (.003) LGRO .085 (.094) .098 (.089) .114* (.088) .113* (.087) LYLD .0 1 1 (.052) .018 (.050) .015 (.048) BRNCH .10 1 (.079) .092 (.073) .094 (.073) .099* (.071) HC .009 (.093) ACSIZ -.017 (.019) AGE -.883 (2 .20) SMSA -.013 (.068) fl f .066 (.267) .056 (.239) .015 (.234) .054 (.195) R-sq .280 .277 .273 .272 R-sq .226 .244 .249 .253 F 5.23 -.013 (.017) 8.33 11.5 ^Denotes significance at the 10 percent level. **Denotes significance at the 5 percent level, ***Denotes significance at the 1 percent level. 14.4 Table 2 Advertising Intensity Regressed on Selected Variables Standard Errors in Parentheses EQUATION 1 2 DGRO 3 .007 (.010) DVAR •329*** (.046) .335*** (.046) CONCE .003 (.009) - .0 0 1 (.005) -.0003 (.006) -.00002 (.003) SHARE .100 (.088) LGRO .307*** (.066) .094 (.087) LYLD .079 (.092) .018 (.052) BRNCH .091 (.071) .088 (.071) .091 (.078) HC .031 (.092) ACSIZ - .0 1 1 (.018) AGE -.627 (2.17) PDEN -.052 (.044) PURB '1 * R-sq R-sq F .031 (.196) .278 .255 11.9 -.004** (.002) -.004** (.002) .294 (.223) .284 (.282) .293 .270 12 .8 .299 .247 5.73 ^Denotes significance at the 10 percent level. **Denotes significance at the 5 percent level. ***Denotes significance at the 1 percent level.