How does ownership structure and manager wealth influence ... · ments system and financial...

Bankers are often characterized asbeing in the business of managing risk.To be successful, bank managers,stockholders, and directors must workclosely together in deciding what riskstheir bank will assume and how theywill control the bank’s overall riskexposure. Each of these participants,though, is likely to have different pref-erences and opinions regarding risk.In fact, the development of bank risk-management policies and procedurestypically involves bank owners, direc-tors, and managers in a process ofgive and take. Throughout thisprocess, differences in opinion mayarise because of differences in owner-ship positions, responsibilities formanaging the bank, and managementoversight functions. Consequently, ifmutually acceptable decisions are to bemade, each decision maker will have tounderstand the preferences of the oth-ers and develop policies that reflect allof their concerns.

Bank risk is also important from theperspective of public policy because ofthe corresponding risks to the pay-ments system and financial stability.Moreover, deposit insurance and thefederal safety net may provide incen-tives for banks to take on additionalrisk. Bank supervisors therefore focuson how well banks manage risk. Inmany cases, weaknesses in a bank’srisk management practices can betraced to weaknesses in its manage-ment and ownership structure. Thus,if a bank examiner is to suggest correc-tive steps for a problem institution, theexaminer should also understand the

basic components of a sound manage-ment and ownership structure.

This study examines the relationshipbetween ownership and managementstructure and a bank’s risk exposure.Previous research has addressed thistopic as well. This study, though, isable to take a more comprehensivelook at the factors influencing bankrisk taking by examining the portionof an owner’s or manager’s wealth thatis concentrated in bank stockholdings.This variable provides new insightsinto bank risk taking and helps clarifythe contribution of other factors.

In earlier work, we found ownershipand management structure to be animportant factor in the cost and profitefficiency of a sample of Tenth FederalReserve District banks.1 The samesample is used in this study to relateownership and management structureto bank risk. The data reveal that own-ership and wealth diversification ofbank owners and managers do influ-ence bank risk. These effects extendnot only to the overall risk of the bank,but are also reflected uniquely in assetquality measures, bank leverage, andother parts of a bank’s risk exposure.

The article first discusses how wealth,ownership structure, and managementposition might affect bank risk. It thenbriefly reviews other studies of thistopic, describes the data used in thisstudy, and presents characteristics ofmanager wealth and ownership at thesample banks. We next present a simpleanalysis of credit risk, other balance

15

1 Kenneth Spong,Richard J. Sullivan,and Robert DeYoung,“What Makes a BankEfficient?—A Look atFinancial Characteris-tics and Managementand Ownership Struc-ture,” Federal ReserveBank of Kansas City

Financial Industry

Perspectives(December 1995): 1-20.Robert DeYoung, Ken-neth Spong, and Rich-ard J. Sullivan,“Ownership, Control,and Performance atSmall Closely HeldFirms: The Case ofCommercial Banks”,working paper, FederalReserve Bank of KansasCity, 1998.

Richard J. Sullivanand Kenneth R. Spongare economists in the

Division of Bank Super-vision and Structure of

the Federal ReserveBank of Kansas City.

The authors are in-debted to the Federal

Deposit InsuranceCorporation and the

state banking depart-ments in the Tenth

Federal Reserve Dis-trict for their help in

collecting the dataused in this paper.

The views expressedin this article are those

of the authors and donot necessarily reflect

those of the FederalReserve Bank of Kan-

sas City or the FederalReserve System.

How does ownership structure and manager wealthinfluence risk? A look at ownership structure, managerwealth, and risk in commercial banks

Richard J. Sullivan and Kenneth R. Spong

sheet risks, and overall risk of hired-manager banks compared to owner-manager banks. The core of the articleis then presented, which is a regres-sion analysis of risk and ownershipstructure.

The relationship between wealth,ownership structure, and risk

Attitudes towards risk and the extentthat a person will accept risk will bedetermined by a number of personalcharacteristics. One important influ-ence on risk taking is the extent towhich an individual’s financialwealth is diversified. Diversificationcan reduce exposure to risk becausean investment with poor returns canbe offset by an investment with goodreturns. As long as financial wealthis diversified, an investor can be lessconcerned with the riskiness of anyindividual investment. But if theinvestor’s wealth is highly concen-trated in a particular investment,then he or she would be more moti-vated to monitor and control theinvestment to reduce its risk.

Owners and managers face differentopportunities as well as personallimitations that will influence theamount of risk that they might accept.Stockholders face a positive relation-ship between risk and return toinvestment projects, and must balancetheir desire for added returns withthe desire to avoid risk. Stockholderswishing to obtain higher returns maywant a business to operate withhigher levels of risk, while others arewilling to accept lower returns inexchange for lower risk.

Managers must also grapple with thelevel of risk and return appropriate tothe business they run. But managersoften have little ownership stake inthe business and therefore may nothave a direct claim on any added

profit earned by accepting added risk.Without an assurance that they wouldgain from accepting higher risk, themanagers may tend to avoid riskyinvestments. Moreover, much of thebackground and training necessary todo their jobs is specific to the businessfor which they work, the value ofwhich would disappear if the managerwere employed elsewhere. As a result,managers may be reluctant to takeadditional risks, since they may tar-nish their reputation and be out of ajob if their gambles fail.

These differences in risk behavior maycause a conflict between businessowners and hired managers.2 Manag-ers may operate the business in a lessrisky manner than that desired byowners. Thus, owners must devisemethods of encouraging managers totake reasonable risks. One method isto reward managers based on the prof-itability of the business. As long asprofitability rises as the level of riskrises, this would encourage managersto take on added risk. However, suchincentive schemes are difficult todesign properly, and in some cases,may cause managers to take on morerisk than owners desire. An alternativeis to give managers stock or stockoptions, thus aligning their interestswith that of other owners and encour-aging them to take more risks.

Another alternative is to monitor theactivities of managers. As representativesof stockholders, the board of directorsperforms the role of monitoring and con-trolling the performance of management.The board must design policies, proce-dures, and compensation schemes thatensure managers will act in the interestof stockholders while operating the bankin a safe and sound manner. Boardinvolvement may thus induce manag-ers to take appropriate risks whileavoiding highly risky ventures. However,this mechanism may not always be

16

FINANCIAL INDUSTRY PERSPECTIVES

2 Financial theory haslabeled this conflict a“principal-agent”problem. See MichaelJensen and WilliamMeckling, “Theory ofthe Firm: ManagerialBehavior, AgencyCosts, and Ownership

Structure,” Journal of

Financial Economics3 (October 1976):305-60, and EugeneFama, “AgencyProblems and theTheory of the Firm,”

Journal of Political

Economy 88 (April

1980): 288-307.

effective because careful monitoringrequires time and effort, and someboard members may not have suffi-cient motivation to bear these costs.3

Thus, the extent to which directorsengage in monitoring and controllingrisk may be tied to their ownershipcharacteristics. Owners with a signifi-cant share of ownership will be moremotivated to monitor the firm becausethey largely benefit from the firm’sgood performance. However, the riskpreference of these owners will dependon their wealth diversification, and astheir wealth becomes more concen-trated in the bank’s stock, the bankshould have less risk.

Finally, there are unique aspects ofthe banking industry that must alsobe considered regarding owner andmanager attitudes towards risk. First,banks are subject to supervision, anda major objective of this supervision isto ensure that banks are operated ina safe and sound manner. Supervi-sory agencies place this responsibilityon the bank’s management and boardof directors, so that bank managersand directors need to be particularlyconcerned with risk. Second, depositinsurance and other elements of thefederal safety net may encouragebank owners to take additional risk,because a risky bank can attractdeposits from insured depositors whowould be unconcerned about the levelof risk in the bank.

To summarize, this study focuses onthree hypotheses regarding risk,wealth, and ownership structure:

• As the manager’s or owner’s wealthbecomes more concentrated in thebank’s stock, the bank should haveless risk.

• Because owners and non-owner man-agers have differing risk preferences,

owner-managed and hired-managerbanks are likely to have differentlevels of risk. Furthermore, becausenon-owner managers are likely tobe more risk averse, an increase inhired-manager ownership of thebank should increase bank risk.

• Monitoring the performance of amanager can assure appropriaterisk taking, but monitoring is acostly activity and will depend on amonitor’s motivation. Effectivemonitors would likely have a sig-nificant share of ownership. Thesemonitors’ risk preferences will de-pend on their wealth diversification,and as their wealth becomes moreconcentrated in a bank’s stock, thebank should have less risk.

Research on bank risk andownership structure

A number of studies examine the rela-tion between bank risk and owner-ship, and their findings have variedconsiderably. According to differentstudies, the relation between insiderownership and the risk of banks issometimes positive, sometimes nega-tive, sometimes U-shaped, and some-times inverse U-shaped (moreinformation on these studies are inBox 1, page 18). These inconsistentresults are due to a number of fac-tors, such as different measures ofrisk, different time periods for study,and different methods for analyzingthe relation between ownership andrisk. Another possible reason for theinconsistent results may be that noneof the studies had information on thewealth of owners or managers.

The sample used in this study containsinformation on the wealth of ownersand managers. The sample consists of270 banks randomly selected fromstate-chartered banks in the TenthFederal Reserve District.4 Bank

17

FEDERAL RESERVE BANK OF KANSAS CITY

3 One study found thatbusinesses with boardsof directors that havepoor attendance atboard meetings andlow ownership of thebusiness tend to beless efficient. SeeSpong, Sullivan, andDeYoung (1995), pp.8-11.4 The Tenth FederalReserve Districtincludes Colorado,Kansas, westernMissouri, Nebraska,northern New MexicoOklahoma, andWyoming.

18


Most studies that have related bank risk to ownership structurehave used measures of risk obtained from stock prices.1 In theearliest of these studies, Saunders, et. al. (1990) examine the linkbetween bank ownership structure and risk taking in a sample of38 large, publicly held bank holding companies. They use thestock prices of the banks to obtain several measures of the capitalmarket risk of the bank holding companies. Their measure ofownership structure was the combined share of outstanding stockand stock options held by all of the managers and directors at abank. They found a higher level of risk in banks that had higherlevels of ownership by managers and directors.

A number of similar studies followed that of Saunders and hiscoauthors by using risk measures derived from stock prices, butresults have been inconsistent. In a sample of 100 savings institu-tions, Brewer and Saidenberg (1996) find a U-shaped relationbetween risk and insider ownership, so that a positive relationmay occur only at higher levels of insider ownership. Knopf andTeall (1996) investigate a sample of 300 savings institutions overthe period from 1987 to 1992. They find that the relation betweenrisk and insider ownership changed over time and argue that therelationship may depend on the level of regulatory stringencyimposed on financial institutions. For the post-FIRREA period(after 1989), they find a negative relation between insider owner-ship and stock price risk. Demsetz, Saidenberg, and Strahan(1997) study a sample of 350 bank holding companies and find apositive relation between risk and insider ownership, but only forbank holding companies with a low franchise value. Chen, Steiner,and Whyte (1998) find a negative relation between insider owner-ship and risk in a sample of 302 banks and savings institutions.

Two studies look at risk in banks as measured by balance sheetindicators. Gorton and Rosen (1995) use the mix of loans in asample of 456 bank holding companies to measure risk, and findan inverse U-shaped relation between the proportion of riskyloans in the bank’s loan portfolio and insider ownership. Knopf

Box 1: Research on Bank Risk and Ownership Structure

1 Many studies of non-financial firms also show a relation between ownership structureand risk. For examples, see Amihud and Lev (1981), Agrawal and Mandelker (1987),Bagnani, et. al. (1994), and May (1995).

19


and Teall (1996) use such risk measures as real estate loans as aproportion of total assets, the equity-to-asset ratio, and brokeredCDs as a proportion of total assets and find an inconsistent rela-tion between insider ownership and risk.

References

Yakov Amihud and Baruch Lev, “Risk Reduction as a Managerial Motive forConglomerate Mergers,” Bell Journal of Economics 12 (Autumn 1981):605-17.

Anup Agrawal and Gershon Mandelker, “Managerial Incentives and Corpo-rate Investment and Financing Decisions,” The Journal of Finance 42(September 1987): 823-37.

Elizabeth Strock Bagnani, Nikolaos Milonas, Anthony Saunders, and Nicko-laos Travlos, “Managers, Owners, and the Pricing of Risky Debt: AnEmpirical Analysis,” The Journal of Finance 49 (June 1994): 453-77.

Elijah Brewer III and Marc Saidenberg, “Franchise Value, Ownership Struc-ture, and Risk at Savings Institutions,” Federal Reserve Bank of New YorkResearch Paper, number 9632 (September 1996).

Carl Chen, Thomas Steiner, and Ann Marie Whyte, “Risk-Taking Behaviorand Management Ownership in Depository Institutions,” The Journal ofFinancial Research 21(Spring 1998): 1-16.

Rebecca Demsetz, Marc Saidenberg, and Philip Strahan, “Agency Problemsand Risk Taking at Banks,” Federal Reserve Bank of New York StaffReport, number 29 (September 1997).

Gary Gorton and Richard Rosen, “Corporate Control, Portfolio Choice, andthe Decline of Banking,” The Journal of Finance 50 (December 1995):1377-1420.

John Knopf and John Teall, “Risk-Taking Behavior in the U.S. Thrift Indus-try: Ownership Structure and Regulatory Changes,” The Journal of Bank-ing and Finance 20 (September 1996): 1329-50.

Don O. May, “Do Managerial Motives Influence Firm Risk Reduction Strate-gies?” The Journal of Finance 50 (September 1995): 1291-1308.

Anthony Saunders, Elizabeth Strock, and Nickolaos Travlos, “OwnershipStructure, Deregulation, and Bank Risk Taking,” The Journal of Finance45 (June 1990): 643-654.

Box 1: Research on Bank Risk and Ownership Structure (continued)

examination reports are our primarysource of data on wealth and owner-ship among owners, directors, andmanagers. Bank examiners collect andreport information on the number ofshares held by officers, directors, andall major owners of the bank, theresponsibilities and compensation ofmanagers, and the net worth of allthe directors of the bank.5 Because abank’s chief executive officer or topmanager is typically on the board ofdirectors, the data set also containsinformation on his or her net worth.Ownership for the bank is identifiedfor the year 1994, and the samemajor ownership group must havecontrolled the bank from 1991 to1994.6 Financial data come fromquarterly call reports for 1990 to1994. More detail on the data can befound in Box 2 on the facing page.

Wealth and ownershipcharacteristics of managers

Sample banks can be in one of twogroups, based on the ownershipcharacteristics of the bank’s top

manager.7 In anowner-managerbank, the manageris a member of theownership groupwith the largeststake in the bank.In the sample, thereare 110 owner-manager banks(Table 1), and ownermanagers on aver-age held 37 percentof the stock in theirbank. Other mem-bers of the ownermanager’s familyoften owned a sig-nificant part of thebank as well, andthe combined familyownership in these

banks averaged 63 percent. Theowner-manager banks are a useful ref-erence group because there should beno conflict between their roles as own-ers and managers.

A hired-manager bank is any bank notin the owner-manager category, andthere are 160 hired-manager banks inour sample (Table 1). For most of thehired-manager banks, the managerhas little or no ownership stake: thisownership share averages only 3 per-cent.8 Moreover, including the mana-ger’s family raises the average share toonly 4 percent. In a portion of thehired-manager banks, the manager’sownership is more substantial—32 ofthe banks have managers with anownership of 5 percent or greater, witha maximum of 39 percent (not shownin Table 1). In all of these banks, how-ever, at least one other owner has alarger share of the bank. The hired-manager banks are an important focusof this study because of the hiredmanager’s potential for risk aversebehavior and the impact on bank risk

20


5 Details oncompensation, networth, and othersensitive informationare contained in aconfidential section ofthe examination report.The confidential sectionis for internal use bybank supervisors andis not part of theexamination report thatis provided to bankers.6 For independentbanks, ownership wasdetermined by theindividual’s proportionof bank common stock.For banks owned by abank holding company,individual ownershipwas calculated usingbank shares ownedindirectly throughownership of shares inthe holding companyplus any additionalbank shares that mightbe owned directly.7 We identify themanager as the personthat examiners list asresponsible on a dailybasis for directing theoperations of the bank.In most, but not all,cases this was thepresident or CEO.8 Often the ownershipwas in sharesnecessary for themanager to qualify as amember of the board ofdirectors.

Manager ownership and net worth(Variables with statistically different values are shown in bold type.)

Owner-ManagerBanks

Hired-ManagerBanks

Number of Banks 110 160

Manager Characteristics

Personal Ownership of Bank*** 37% 3%

Personal plus Family Ownership of Bank*** 63% 4%

Personal Net Worth (millions)*** $1.719 $0.472

Investment in Bank / Personal Net Worth+,*** 86% 21%

Notes: The sample consists of 270 state-chartered banks in the Tenth Federal Reserve District.The statistics are unweighted averages.+ Investment in bank = number of shares personally held in bank × book value of equity per share.*, **, *** indicate significant difference between owner-manager and hired-manager banks at the 10%,5%, or 1% significance level. Statistical difference based on a t-test for a difference in means.

Table 1

21


Ownership data come from a unique data set that was compiled frombank examination reports, and the financial variables are taken from callreport information. The sample of banks is from the Tenth FederalReserve District. To be considered for this sample, a bank had to be inexistence for at least five years prior to 1990 and remained in existencethrough 1994. A bank also had to be a full-service bank, offering loans,holding insured deposits, and generating noninterest income. All of thebanks had to have a complete set of financial data. A total of 1,421 banksmet these criteria.

As the second step of the data selection process, we randomly chose 304state-chartered banks from the group of banks identified in the first step.The data set on management and ownership was constructed by tran-scribing data from state agency, FDIC, and Federal Reserve examinationreports of state-chartered banks. The sample size was limited to 304banks because of the labor required to collect this examination data.Since data on ownership structure and wealth of directors were readilyavailable for state-chartered banks, the sample only included this charterclass. Exam reports have information on responsibilities, age, tenure andcompensation of bank officers; ownership, net worth, responsibilities, age,tenure and compensation of members of the boards of directors; owner-ship information on all major owners of the bank; family relationshipsamong officers and stockholders; and information on policymakingresponsibilities in the bank. On occasion we supplemented and verifiedthis data through a number of other sources, including Federal Reservebank holding company inspection reports and annual reports to the Fed-eral Reserve filed by bank holding companies.

In general, the ownership information comes from 1994 examinationreports, although some 1993 reports were used when the 1994 report wasunavailable. Our measures of the overall risk of the banks (the standarddeviation of adjusted net income and the survival likelihood index) usequarterly data from the five-year period from 1990 to 1994. In order tomatch risk performance and ownership structure over this period, weexcluded any bank that underwent a significant ownership change in1991, 1992, or 1993.1

A few banks had to be excluded because of information problems.Twenty-seven banks were excluded because they experienced a significantownership change (that is, the majority ownership of the bank changedhands). Missing data and other problems reduced the initial sample byanother seven banks, so that the final sample included 270 banks.Finally, the data used in the regressions reported below used 267 obser-vations because financial data were missing from some sample banks.

Box 2: Details on Sample Data

1 If a bank had an ownership change in 1994, data on the 1993 ownership structure wasused.

if they act in their own interest ratherthan that of owners.

Information on net worth and bankinvestment relative to net worthshows that, compared to hired man-agers, the average owner manager iswealthier and has more wealth con-centrated in the bank. The averagenet worth of owner managers is$1.719 million, compared to $0.472million for hired managers (Table 1).Wealth concentration in the bank iscalculated as the ratio of one’sinvestment in the bank to personalnet worth, with investment in thebank equal to the number of sharesof common stock held times the bookvalue of bank capital per share.9 Forowner managers, this wealth concen-tration averages 86 percent, while for

hired managers, it averages 21 percent(Table 1)10 For the majority of hired-managers, the ratio is under 20 per-cent (Chart 1). By contrast, Chart 1shows that a ratio greater than 20 per-cent is the rule for owner-managerbanks.11

These ratios and their distributionacross individual managers have anumber of implications for bank risktaking. When an owner-manager’swealth is well diversified, we expectowner-managed banks to be run in ariskier fashion than hired-managerbanks. In a number of the samplebanks, owner managers are well diver-sified, as Chart 1 shows. But Chart 1also shows that owner-manager wealthis often highly concentrated in theirbanks, which could lead them to run

22


9 Few of the banks hadstock that was activelytraded, so book valuewas used as a simpleproxy for the value ofstock in the samplebanks. Control sharesare typically worthmore than minorityshares, but book valuegave a consistentmeasure across allstockholders andsample banks.10 This ratio may besubject to measurementerror, since a great dealof time may not bespent in preparing orverifying the accuracyof some of the directornet worth statements.Also, book value is onlya proxy for the actualvalue of investment inthe bank. Overall,though, these ratiosshould provide a goodgeneral guide to thewealth concentration ofthe major players in abank’s operations.11 If the manager or theholding company useddebt financing in pur-chasing the bank, thenour wealth concentra-tion ratios could beoverstated and evenexceed one because wedid not adjust the bankinvestment measure forany underlying debt.On the other hand,lenders often ask majorstockholders to person-ally guarantee theirbank stock debt, whichwould suggest thatmore of their wealthcould be at stake thanjust their investment inthe bank. Underreport-ing of net worth couldalso lead to wealthconcentration ratiosabove one.

Distribution of manager bank investment / personal net worth

Chart 1

80% to100%

Ownermanagers

60

40

Number of banks

0

80

20

50

30

70

10

Hiredmanagers

Zero More thanzero to 20%

20% to40%

40% to60%

60% to80%

Bank investment / Personal net worth

More than100%

90

their banks conser-vatively. Moreover,many hired manag-ers have diversifiedwealth, which couldlead them to accepthigher levels of risk.These differencessuggest that weneed to account forboth ownership ofbank stock andwealth diversifica-tion to get a com-plete picture of howthe characteristicsof managers influ-ence risk at abank.12

Owner-managerbanks tend to besmaller, more rural,and somewhat lessprofitable thanhired-managerbanks in this sam-ple (Table 2). Aver-age assets are $38.6million for owner-managed banks and$56.7 million forhired-managerbanks. Eighty-one percent ofowner-managed banks are in non-metropolitan areas, compared to 72percent for hired-manager banks.Both return on average assets andreturn on equity in 1994 were higherfor hired-manager banks than forowner-manager banks. This is alsotrue when we calculate averageROAA and ROE for the period from1990 to 1994, although the differencefor average ROAA is not statisticallysignificant.

An alternative measure of profitabil-ity, and one more relevant for thestudy of risk, is operating net income.Operating net income is net income

before taxes, securities gains or los-ses, and extraordinary items.Operating net income is a betterreflection of a bank’s business riskbecause business risk is determinedby the variability of the demand forthe bank’s product and services,interest rate volatility, and the flexibil-ity of the bank’s asset and liabilitymanagement.13 Operating net incomefocuses on the core business of thebank and eliminates fluctuations inincome from variables that can bemanipulated on a short-run basis. AsTable 2 shows, hired-manager bankshave a higher average value of operat-ing return on average assets com-pared to owner-manager banks.

23


Basic characteristics of sample banks(Variables with statistically different values are shown in bold type.)

Table 2

Owner-Manager Banks Hired-Manager Banks

Basic information

Number of Banks 110 160Assets—Average (millions), 1994** $38.6 $56.7Assets—Median (millions), 1994 $26.3 $30.8Location—Percent nonmetropolitan* 81% 72%

Net income

Return on Average Assets (ROAA), 1994* 1.016% 1.138%Average ROAA, 1990-1994 1.077% 1.132%Return on Equity (ROE), 1994** 10.46% 12.33%Average ROE, 1990-1994* 11.47% 12.47%

Operating net income+

Operating ROAA+, 1994* 1.461% 1.619%Average Operating ROAA+, 1990-1994 1.428% 1.519%Operating ROE+, 1994** 14.93% 17.63%Average Operating ROE+, 1990-1994* 15.16% 17.05%

Notes: The sample consists of 270 state-chartered banks in the Tenth Federal Reserve District.The statistics are unweighted averages or median values for sample banks. Statistics covering the period from 1990 to 1994 arecalculated from quarterly data.*, **, *** indicate significant difference between owner-manager and hired-manager banks at the 10%, 5%, or 1% significance level.Statistical difference based on a t-test for a difference in means or a rank-order test for a difference in medians.+Operating net income measures profitability of banking operations, and is defined as net income + taxes + extraordinary Items – securitiesgains or losses. Operating ROAA = operating net income / average assets, and operating ROE = operating net income/equity capital.

12 There is a positivecorrelation betweenownership and the ratioof bank investment topersonal net worth. Thecorrelation is .944 forhired-manager banks,although if you only con-sider hired managerswith an ownership stakeover 1 percent, the corre-lation drops to .531. Forowner-manager banks,the correlation is only.441. The relatively lowcorrelation implies thatthere can be consider-able error in assumingthat a higher percentageof ownership also im-plies less diversificationof wealth.13 Joseph F. Sinkey,

Jr., Commercial Bank

To gain insight into the range of risksthat a bank must control, we studieda number of measures of bank risk,and this section of the article intro-duces these risk measures. This sec-tion also compares the values of therisk measures at owner-managedbanks with those for hired-managerbanks to analyze the effect of man-agement structure on risk. Thiscomparison yields some interestingfeatures of risk-taking and manage-ment structure, but is incompleteand sometimes contradictory. Thesecontradictions occur becauseunderlying characteristics of hired-and owner-manager banks, such aswealth diversification, asset size, andlocation, each have their own effecton bank risk. Tables 1 and 2 showthat these underlying characteristics

can be very differentacross banks. The nextsection will present amore complete analysisof ownership structure,diversification of themanager’s wealth, andother factors that affectbank risk, and willreconcile any contradic-tions uncovered in thissection.

Credit risk. The first setof risk measures exam-ines a bank’s exposure torisk through its lendingactivities. The loan port-folio is a major source ofrisk that the board ofdirectors and manage-ment control by estab-lishing policies regard-ing lending limits forloan officers, limiting theloan-to-asset ratio, andlimiting credit concen-

trations among industries, loan cate-gories, or geographic locations. Seniormanagement and the board of direc-tors may also approve major loans asan added level of control.

Compared to hired-manager banks,owner-managed banks have a higherlevel of risk in their loan portfolios, asreflected in higher rates of loan losses,higher levels of past due loans, andhigher ratios of noncurrent assets tothe loan loss reserve (see Table 3).These differences are due to a numberof different factors. Owner managersmay be willing to set less conservativestandards for acceptable loans becausethey can benefit from the higherexpected return associated with riskierloans. Hired managers may set moreconservative standards because badloans are highly visible, and hiredmanagers may want to avoid reportinglosses to the board of directors. On the

24


Financial

Management (NewYork: MacMillanPublishing Company,1989), p. 408.

Manager ownership and credit risk(Variables with statistically different values are shown in bold type.)Year-end 1994

Table 3


Total Loans / Total Assets 51.90% 54.80%

Net Loan Losses / Total Loans* 0.227% 0.157%

Nonperforming Assets / Total Loans** 1.360% 0.807%

Past Due Loans / Total Loans* 2.818% 2.093%

Noncurrent Assets / Total Assets* 0.929% 0.616%

Allowance for Loan Losses / Total Loans 1.866% 1.899%

Provision for Loan Losses / Total Loans 0.193% 0.216%

Other Real Estate Owned / Total Assets 0.236% 0.162%

Noncurrent Assets / Allowance for Loan Losses * 114% 65%

Notes: The sample consists of 270 state-chartered banks in the Tenth Federal Reserve District. Statistics are unweightedaverages. Nonperforming assets = loans that do not accrue interest plus loans that are past due by 90 days or longer. Past dueloans = loans that do not accrue interest plus loans that are past due by 30 days or longer. Noncurrent assets = nonperformingassets plus other real estate owned.*, ** indicate significant difference between owner-manager and hired-manager banks at a 10% or 5% significance level. Statisticaldifference based on a t-test for a difference in means.

Risk characteristics of owner-and hired-manager banks

other hand, this resultmay be somewhatsurprising, becausemany owner managershave much of theirpersonal wealth at stakein their bank and mightbe expected to controlcredit risk more carefullythan hired managers.This latter effect isimportant, as will beseen in the next section.However, consideringthis risk in isolation, thenet effect of all thedeterminants of creditrisk is to leave owner-manager banks in oursample with more creditrisk, on average, comparedto hired-manager banks.

Other balance sheet

risk. Banks must alsocontrol risk associatedwith other balance sheetitems. The bank isexposed to risk associ-ated with access to funds, commit-ments to the cost of fixed assets, andinterest rate fluctuations. Bank man-agers and owners must also make afundamental decision about howmuch equity to hold in the bank. Thisdecision is important because equityprovides a cushion to absorb loanlosses or unexpected drops in netincome.

We find that owner-manager bankshave lower levels of other balancesheet risk compared to hired-managerbanks. Sample data show thatowner-manager banks have higherlevels of capitalization and lower lev-els of operating leverage and market(interest rate) risk compared tohired-manager banks (see Table 4).This result, which is somewhat atvariance with measures of credit risk,

may reflect the concentration ofwealth that owner managers have intheir bank and the interest this willgive them in the long-run stability oftheir banks. Additional determinantsof other balance sheet risk also play arole, as we will see in the next sectionof the article.

Overall risk. Because higher risk inthe loan portfolio is offset to someextent by lower risk in other balancesheet accounts for owner-managerbanks (and vice-versa for hired-manager banks), it is useful to exam-ine measures of the overall risk of thebank. The commonly used market-based measures of bank risk (such asfluctuations in stock returns) areunavailable for this study becausemost of the sample banks do not havepublicly traded common stock.14

25


Manager ownership and other balance sheet risk(Variables with statistically different values are shown in bold type.)Year-end 1994

Table 4


Capitalization—leverage risk

Equity / Total Assets** 10.18% 9.36%

Operating leverage—fixed asset risk

Premises and Fixed Assets / Total Assets** 1.133% 1.419%

Non-core funding risk

Non-core Liabilities / Total Deposits1 10.24% 10.83%

Market risk—interest rate risk

Absolute value (Asset-Liability Mismatch / Total Assets)**,2 13.48% 16.47%

Notes: The sample consists of 270 state-chartered banks in the Tenth Federal Reserve District. Statistics are unweighted averagesfor sample banks.*, ** indicate significant difference between owner-manager and hired-manager banks at a 10% or 5% significance level. Statisticaldifference based on a t-test for a difference in means.1 Non-core liabilities include time deposits or certificates of deposits over $100,000, federal funds purchased, repurchaseagreements, and foreign deposits.2 The asset-liability mismatch measures the gap between short-term (under one year) assets and liabilities. We calculate gap bysubtracting fixed and floating rate short-term deposits (both time and certificates of deposits) from fixed and floating rate short-termearning assets (loans and securities). Risk rises as the absolute value of the gap rises, because the bank’s earnings are thensubject to greater fluctuations due to changes in interest rates.

14 Only five of thebanks in the studyare in bankingorganizations that havestock traded on majorexchanges.

Instead, accounting-based measuresof risk will serve as a substitute.15

The first measure we consider isvariation in income.16 A bank withrisky loans and investments will bemore likely to experience large gainsand losses, which will be reflected inextensive income variability. Table 5shows information on the variation ofincome, as measured by the variationin total revenue or the variation inoperating return on assets. Hired-manager banks have greater variationin income compared to owner-managed banks, although the differ-ence is not statistically significant.

A second measure of overall risk is abank’s survival likelihood index.While income fluctuations are impor-tant indicators of the riskiness of abank, their ability to measure risk isincomplete. A bank may have astable income stream, but it could bemore risky than some other banks

due to lower levels of average profit-ability or less capital protection.

Income fluctuation, capitalization, andaverage profitability combine in anumber of different ways to produce aunique level of risk. In the results pre-sented above, the average owner-manager bank was less profitablecompared to hired-manager banks,which implies more risk, but had ahigher capital-to-asset ratio and lowerfluctuations in income, both of whichimply less risk. An advantage to thesurvival likelihood index as a measureof risk is that it incorporates all three ofthese factors together.17

The survival likelihood index is defined as

The higher the value of the survivallikelihood index, the lower the risk ofthe bank. An increase in the capital-

26


Manager ownership and measures of overall bank risk(Variables with statistically different values are shown in bold type.)


Income variation

Standard Deviation of Total Revenue / Average Assets .286% .340%

Standard Deviation of Operating Return on Average Assets .787% .842%

Survival likelihood index+

(Equity/Assets + Average operating ROAA++) /Standard Deviation of operating ROAA

20.58 18.96

Notes: The sample consists of 270 state-chartered banks in the Tenth Federal Reserve District.Statistics are calculated from quarterly data for the period from 1990 to 1994, and are unweighted averages for sample banks.*, **, *** indicate significant difference between owner-manager and hired-manager banks at least at a 10%, 5%, or 1% significance level.Statistical difference based on the Wilcoxon test for a difference rank order of the data.+ The survival likelihood index measures the number of standards deviations that ROAA or operating ROAA would need to fall in order toexhaust equity and force a bank failure. A large value of the survival likelihood index implies low risk, that is, the larger the value of theindex the lower the probability of bank failure.++ Operating net income measures profitability of banking operations, and is defined as net income + taxes + extraordinary items –Securities Gains or Losses. Operating ROAA = (operating net income) / (average assets).

Table 5

15 Research has shownthat there is a positiveand significantcorrelation betweenaccounting- andmarket-basedmeasures of risk; seeWilliam Beaver, PaulKettler, and MyronScholes, “TheAssociation BetweenMarket Determined andAccounting Determined

Risk Measures,” The

Accounting Review(October 1970): 654-82and William Beaverand James Manegold,“The AssociationBetween MarketDetermined andAccounting DeterminedRisk Measures ofSystematic Risk: SomeFurther Evidence,”

Journal of Financial

and Quantitative

Analysis (June 1975):231-84.16 Variation in incomeis measured using thestandard deviation ofincome over the 20quarters in the periodfrom 1990 to 1994.17 The index is basedon the Z score in JohnH. Boyd and Stanley L.Graham, “Bank HoldingCompany Risk,”chapter 10 in Benton

Gup, editor, Bank

Mergers: Current

Issues and

Perspectives. (Boston:Kluwer AcademicPublishers, 1989), pp.200-1. This representsthe number of standarddeviations below themean that operatingreturn on assets wouldhave to fall in order toeliminate capital.

capital-to-asset ratio+ average value of operating return on assets

standard deviation of operating returnon assets.

to-asset ratio would raise the index,as would an increase in the meanvalue of operating return on assets,both of which imply less risk. A risein the standard deviation of operatingreturn on assets would lower theindex, which implies more risk.

A second advantage is that the sur-vival likelihood index can be viewedas a measure of the likelihood of fail-ure. The smaller the value of the sur-vival likelihood index, the more likelya bank will fail. This is of particularimportance to stockholders and regu-lators since bank failure will wipe outa stockholder’s investment, whileexposing the bank insurance fund toloss.18 Table 5 shows that the survivallikelihood index is higher for owner-manager banks, although the differ-ence is not statistically significant.Thus the differences in profitability,capitalization, and income fluctua-tions combine into a slightly highersurvival likelihood index for owner-managed banks.

A regression analysis of bankrisk, ownership structure, andmanager wealth

Dividing banks into owner-managerand hired-manager categories capturesa notable difference in the manager’sownership status but is a crudemethod for understanding the manyvariables that can influence bankrisk. The differences in risk revealedin Tables 3, 4, and 5 reflect not onlythe manager’s ownership status, butalso other underlying characteristics,such as the manager’s wealthdiversification, the amount ofownership of the hired manager, thewealth diversification of the majorowners of hired-manager banks, andthe size and location of the bank. Aclear understanding of how bankrisk responds to differences inownership structure, wealth diversifi-

cation, and other characteristicsrequires that their effects beaccounted for simultaneously.

Multiple regression provides theappropriate statistical tool for thistype of analysis. We specify an equa-tion that makes risk a mathematicalfunction of several explanatory vari-ables: the manager’s ratio of bankinvestment to personal net worth, theownership share of the hired manager(when the bank has a hired manager),and a “monitor’s” ratio of bank invest-ment to personal net worth (forhired-manager banks only). In thisequation, the “monitor” is defined asthe director who holds the mostshares of any director and is also amember of the largest ownershipgroup. The equation also includesvariables to indicate the asset size ofthe bank and whether the bank is ina nonmetropolitan location.

The regression technique allows us toestimate the function and see how riskresponds to an individual explanatoryvariable after accounting for theother explanatory variables. Forexample, we can estimate the changein risk associated with a change inthe ownership of a hired manager,while holding the other explanatoryvariables at certain, specified values.As a result, we will have a cleanermeasure of how changes in hiredmanager ownership might affect riskor how changes in any of the othervariables might affect risk separately.

We estimate the regression equationusing all of the measures describedabove for credit risk, other balancesheet risk, and overall bank risk. Inorder to focus the following discussionon essential results, technical detailsof the analysis are in an appendix. Toillustrate the results, we present graphsfor various risk measures plottedagainst various explanatory variables.19

27


18 Boyd and Graham(1989), pp. 221-2,consider the question ofwhether the survivallikelihood indexcomputed usingaccounting data orstock market data is abetter measure ofbankruptcy risk. Theyconclude that theaccounting-basedsurvival likelihoodindex conveys “much ofthe same informationthat is in commercialpaper ratings. Themarket [survivallikelihood indices] donot. To the extent,therefore, thatcommercial paperratings are usefulmeasures ofbankruptcy risk, thesefindings favor the useof [a survival likelihoodindex]computed withaccounting data.”19 The graphs useestimated regressionequations and assumedvalues for variables asspecified in each figure.

Wealth concentration. Under ourfirst hypothesis, managers and own-ers are assumed to become moreconservative as more of their wealthbecomes tied up in the bank. Forboth hired and owner managers, thisrelationship appears to be trueacross a wide range of risk measures.The manager’s ratio of bank invest-ment to personal net worth is statis-tically significant in explaining assetquality (as measured by the ratios ofnet loan losses to total loans, non-performing assets to total loans, pastdue loans to total loans, and noncur-rent assets to total assets; see TableA1.1).20 Estimates show that as amanager has more of his or herwealth concentrated in the bank,asset quality improves, suggestingthat the manager is reducing therisk in the loan portfolio.

Owner managers typically have amuch higher concentration of wealthin their bank investment comparedto hired managers, and by itself, thiswould imply a lower average creditrisk in owner-managed banks. Table3, however, shows that the averagehired-manager bank had lower levelsof credit risk. The apparent contradic-tion is because other factors that influ-ence credit risk are not equal acrossowner- and hired-manager banks, andthe differences are sufficient to causeaverage credit risk to be higher atowner-managed banks.

The manager’s ratio of bank invest-ment to personal net worth is alsostatistically significant in explain-ing the equity-to-asset ratio (TableA2) and the survival likelihoodindex (Table A3). As the manager’sconcentration of wealth in the bankrises, the equity-to-asset ratio risesand the survival likelihood ratiorises, thus indicating a decline inrisk.

Chart 2 illustrates how the rate of loanlosses, the equity-to-asset ratio, and thesurvival likelihood index each respondsto changes in wealth concentration.21

Other factors held constant, risk declinesas the manager’s wealth becomes moreconcentrated in the bank’s stock for allthree measures of risk.

Hired managers and their owner-

ship. In our second hypothesis, hiredmanagers are assumed to be more riskaverse than otherwise comparableowner managers. Our regressionresults show that banks with hiredmanagers have lower credit risk asreflected in lower rates of loan losses,nonperforming assets, past due loans,and noncurrent assets (see the hiredmanager indicator variable, TableA1.1). One reason hired managers maybe avoiding risky loans is becausereporting bad loans to the bank’sboard of directors is a highly visibleblemish on the performance record ofthe manager. Moreover, a hired-manager could also be concernedabout a reputation for making badloans because it might have implica-tions for the manager’s future jobprospects. In contrast to owner man-agers, who are committed to their ownbank, a hired manager may have aspi-rations to move to other banks as pro-fessional opportunities arise.

A lower level of credit risk at hired-manager banks may further indicateconflict between the risk preferencesof hired managers and bank owners.As we argued above, stock ownershipon the part of a hired-manager mayovercome the tendency for the man-ager to avoid risk. The regressionresults suggest that this is true withregard to net loan losses, noncurrentassets, bank equity, variation ofoperating earnings, and the survivallikelihood index.22 In each of thesecases, an increase in a hired manager’sownership is associated with greater

28


20 The manager’s ratioof bank investment topersonal net worth wasnot statisticallysignificant in explainingallowance or provisionfor loan losses.21 Panels A, B, and Cof Chart 2 are based onregression equations(2), (10), and (16) fromAppendix Tables A1.1,A2, and A3.22 See the variable forownership share ofhired manager inregressions (2), (5), (10),(15) and (16) in TablesA1.1, A2, and A3.

29


Net

loan

loss

es/t

otal

loan

s

Percent

.50

.40

.30

.20

.10

00 0.5 1.0 1.5

Manager’s bank investment / net worth

Panel ACredit risk falls as the manager’s wealthbecomes concentrated in bank equity

Manager’s bank investment / net worth1.51.00.5

6

8

10

12

Percent

Equ

ity/a

sset

s

Panel BLeverage risk falls (equity/assets rises) as the manager’s wealth

becomes concentrated in bank equity

0

Manager’s bank investment / net worth

1.00.5010

15

20

25

30

Sur

viva

llik

elih

ood

inde

x

Panel CLikelihood of survival rises as the manager’s wealth


1.5

Chart 2

Note: Calculations assume the monitor’s investment relative to net worth is 50%, the bank has $50 million in assets,, and is in a nonmetro-politan area. In Panels A, B, and C, calculations assume that the hired manager owns, respectively, 20.5%, 6.7%, or 4.8% of the bank. Thesurvival likelihood index is defined as (equity / assets + average operating ROAA) / standard deviation of operating ROAA.

risk. Chart 3 illustrates this effect:holding other variables constant,credit risk rises (as reflected in therate of net loan losses) and the sur-vival likelihood index falls, as thehired manager’s ownership in thebank increases.23

Monitor’s wealth concentration.Under our third hypothesis, indi-viduals in a position to monitorhired managers should influence

bank risk in accordance with theirinvestment concentration in thebank. Results show that the monitor’sratio of bank investment to personalnet worth is positively related to theequity-to-asset ratio and to the sur-vival likelihood index, and negativelyrelated to variation in total revenue.24

Thus as bank monitors have more oftheir wealth tied up in the bank, theybecome more careful about the risks

30


Chart 3

Note: Calculations assume the monitor’s investment relative to net worth is 50%, the bank has $50 million in assets, is in a nonmetropolitanarea, and bank investment/personal net worth for the manager is .5. The survival likelihood index is defined as (equity / assets + averageoperating ROAA) / standard deviation of operating ROAA.

Net

loan

loss

es/t

otal

loan

s

Percent

.50

.40

.30

.20

.10

00 0.1 0.2 0.3

Ownership proportion of hired manager

Panel ACredit risk increases as ownership of the bank

by the hired manager rises

Ownership proportion of hired manager

10

15

30

Sur

viva

llik

elih

ood

inde

x

Panel BLikelihood of survival falls as ownershipof the bank by the hired manager rises

0.250.150.05

0.30.250.20.150.10.050

20

22

Hired-manager banks

Average forowner-manager banks


Hired-manager banks

23 Panels A and B ofChart 3 are based onregressions (2) and (16)of Table A1.1 and A3.24 See the variable formonitor’s bankinvestment/net worthin regressions (10), (14),and (16) of Tables A2and A3.

that managers and the bank areallowed to assume. Chart 4 illustratesthis effect.25 When the ratio of themonitor’s bank investment to per-sonal net worth is .10, the equity-to-asset ratio would be roughly one per-centage point below comparableowner-managed banks. If the ratio ofthe monitor’s bank investment to networth rises to .60, the equity-to-assetratio would be at a higher level,implying not only less risk, but alsoan equity-to-asset ratio close to thatof comparable owner-managedbanks.

The effect of the monitor’s wealthconcentration on the equity-to-assetratio helps to explain why hired-manager banks, on average, have alower equity-to-asset ratio comparedto owner-manager banks (see Table4). Among hired-manager banks, themonitor’s investment in the bankrelative to personal net worth averages51 percent, while Table 1 shows thatthe figure for owner managers is 86

percent. Thus, the wealth of themajor owners of hired-managerbanks is more diversified than that ofowner managers. As a result, thelower ratio of equity to assets amonghired-manager banks seen in Table 4is not due solely to the manager’sownership status, but also due to thewealth diversification of the majorowner of the bank (and to otherunderlying characteristics).

Since the monitor is a major ownerand is also on the board of directors,he or she has the power to influencethe financial position of the bank.Our results show that much of thisinfluence is reflected in the equity-to-asset ratio, fluctuations in totalincome, and the bank’s survival like-lihood index. The results did notshow that this monitoring individualhad as strong of an influence overcredit risk in the bank, which mayindicate a lesser involvement in dailycredit decisions.

31


Chart 4

Note: The “monitor” is defined as the director who holds the most shares of any director and is also a member of the largest ownershipgroup. Calculations assume that the hired manager owns 10% of the bank, bank investment / net worth for the hired manager is .2, thebank has $50 million in assets, and is in a nonmetropolitan area.

Equ

ity/a

sset

s

Percent

12

10

8

60 0.4 1.0 1.4

Monitor’s bank investment / net worth

Leverage risk falls (equity / assets rises)as the monitor’s wealth


0.2 0.6 0.8 1.2

Hired-manager banks


25 Chart 4 is based onregression (10) of TableA2.

Location and bank size. Ourresults suggest that banks in non-metropolitan areas are less riskycompared to those in metropolitanareas. Banks in nonmetropolitan areashad lower loan loss rates, lower ratesof noncurrent assets, lower holdingsof other real estate owned, higherequity-to-asset ratios, lower ratios ofpremises and fixed assets to totalassets, and a higher survival likeli-hood index, all of which suggest lessrisk for nonmetropolitan banks.26

Our results show that bank size waseither insignificant or had conflictinginfluences on bank risk. Results didnot show that there was a relation-ship between asset size and creditrisk (Tables A1.1 and A1.2). Otherrisk measures were sometimes posi-tively and sometimes negativelyrelated to bank size (Tables A2 andA3), and so it may be best to turn toan overall measure of bank risk tosummarize these conflicting results.

A good summary measure is providedby the survival likelihood index. Ourresults show that the survival likeli-hood index increases with asset size,as illustrated in Chart 5.27 The figurealso shows that the relationship is notlinear, with the curve becoming flatterat higher asset sizes. Thus, banks thatgrow from $10 million to $15 millionin assets will have a larger increase inthe survival likelihood index thanbanks that grow from $95 million to$100 million in assets.

Summary and conclusions

Controlling risk in banks is a challeng-ing responsibility, but the task can bemade easier through an understand-ing of how ownership structure andthe diversification of wealth influencesthe risk preferences and risk-takingbehavior of bank managers and owners.This study looks at a sample of TenthFederal Reserve District banks to exam-ine the relationship between bank risk,ownership of the bank by managers,

32


Chart 5

Note: Calculations assume the bank investment/personal net worth for the manager is .5, the hired-manager owns 0% of the bank, themonitor’s bank investment/personal net worth is .88, and the bank is in a nonmetropolitan area. The survival likelihood index is defined as(equity/assets + average operating ROAA) / standard deviation of operating ROAA.

Sur

viva

llik

elih

ood

inde

x30

25

20

15

10$0 $50 $100 $150

Assets (millions)

Likelihood of survival risesas the bank becomes larger

26 See regressions (2),(5), (9), (10), (11), and(16) of the Appendixtables. Counter to theseresults is a higher ratioof volatile deposits tototal deposits fornonmetropolitan banks(regression (12), TableA2).27 Chart 5 is based onregression (16) of TableA3.

and the degree to which managersand owners have their wealth concen-trated in their bank stockholdings.The major results are that:

• Concentration of financial wealthin the bank has a clear role indetermining bank risk: banks areless risky when bank managers(whether he or she is a hired or anowner manager) have a higher con-centration of wealth in their bank.This effect is multifaceted, throughimpacts on credit risk, the equity-to-asset ratio, variation in earnings,and average profitability (as reflectedin the survival likelihood index).

• Hired managers typically operatetheir banks with lower credit riskthan banks with owner managers,reflecting the different incentivesthat these managers face. Stockownership by hired managers canprovide an incentive to bring the levelof credit risk in their bank closer tothat of owner-manager banks.

• Active monitoring by an owner ordirector can guide hired-managerbanks towards a level of riskdesired by bank owners.

Because of these combined effects,there is no simple way to characterizewhether a hired-manager bank islikely to be more or less risky com-pared to an owner-managed bank.Credit risk may be higher at owner-manager banks, but owner managersoften take steps to offset this risk,such as holding more capital in thebank. Results of this study do, how-ever, allow some generalizations:

• owner-manager banks tend to bemore risky when the owner’s wealthis well-diversified; and

• hired-manager banks tend to bemore risky when the hired managerhas a significant ownership stakein the bank but has wealth that iswell-diversified, and where anymajor owners that are likely to ful-fill a monitoring role have wealththat is highly diversified.

These results have implications fortrends in the banking industry andfor the risk-focused examinationprocess. Consolidation will have aprofound effect on the ownershipstructure of the banking industry byreducing the proportion of owner-managed banks relative to hired-manager banks and by creating largerbanks with more diversified investors.According to the results of this study,more hired managers could cause thelevel of risk in bank loan portfolios tofall, but this lower credit risk may beoffset by higher risks in other aspectsof a bank’s operations. Diversifiedinvestors with smaller blocks of stockin larger banks could mean that otherrisk control mechanisms will becomemore important, such as managerialcompensation schemes, boards ofdirectors, and equity markets. Bankexaminers and supervisors need to beaware of management and ownershipstructures that could lead to exces-sive risk taking—such as a bankwhere major owners have diversifiedwealth and where other controlmechanisms are weak.

For individual banks, results of thisstudy show that ownership structureand concentration of wealth in bankequity have significant influence onbank risk. Understanding how riskpreferences depend on ownership andwealth diversification can be valuableinformation to managers and ownersas they grapple with the level andtype of risk to take in their banks.

33


34


Specification

The statistical model is designed to test hypotheses related to theeffect of ownership structure and diversification of wealth on bankrisk. The basic statistical model makes bank risk a function of char-acteristics of ownership and management and two control variables:

Risk Measure = f (manager’s bank investment/personal net worth,hired manager indicator variable,ownership share of hired manager,monitor’s bank investment/net worth ×

hired manager indicator variable,nonmetropolitan indicator variable,assets, assets squared)

The first and second hypotheses discussed above (p. 17) suggeststhat the relation between risk and a manager’s bank investment/personal net worth should be negative, while the relation betweenrisk and the ownership share of hired managers should be positive.

Because of the differences in the preferences towards risk betweenowners and hired-managers, monitoring of hired-managers can bean important influence on bank risk taking. To account for thisaspect of hired-manager banks, we had to identify a person whowould most likely take on the responsibility of monitoring the bank’smanager. We chose the director who holds the most shares of anydirector and is also a member of the largest ownership group.1 Forthis monitor, we calculate the value of his or her investment in thebank relative to personal net worth. Among hired-manager banks,the average value of the monitor’s bank investment relative to per-sonal net worth was 51 percent. As this variable rises, and themonitor’s wealth becomes less diversified, we would expect to see themonitor become more risk averse and more interested in limitingbank risk taking.

The regression equation also includes variables for the bank’s loca-tion and size. Location is measured by whether or not the bank is ina nonmetropolitan area. Banks in metropolitan areas face a differentmarket for loans compared to banks in nonmetropolitan areas. Mar-ket characteristics can influence the mix of loans that a bank can

Appendix: Regression Analysis of Risk, Ownership Structure, and Managerand Ownership Wealth

1 We also considered two other definitions of monitor: the director who holds the mostshares of any director regardless of belonging to the largest ownership group and the chairof the board of directors. The results using these individuals as potential monitors are notstatistically significant in the regression equations and are not presented here.

35


make, and since different types of loans have different risks, locationcan influence the amount of risk at a bank. Bank size may alsoinfluence risk, since larger banks often have more opportunities todiversify their loans and investments and thus lower portfolio risk.While banks can have some control over their loan mix and assetsize, for many of the banks in our sample the degree of control islimited. Location and asset size are included in the regression equa-tion in order to account for some factors influencing risk that arebeyond the control of bank management.

Tests revealed that the error terms for regressions with raw values ofthe standard deviation of earnings or the survival likelihood indexwere not normally distributed. A log transformation of these depend-ent variables corrected this problem (Table A3 has the results). Inthese equations we used the log of assets (rather than assets andassets-squared) to allow for a potential nonlinear relation betweenthe dependent variables and asset size.

We experimented with the variable “manager’s bank investment/networth” for appropriate functional forms. First, we looked for a non-linear relationship by entering the variable as a set of dummy vari-ables. Results did not suggest that this specification was superior toentering “manager’s bank investment/net worth” as a continuousvariable. We also looked for a nonlinear relation by making the riskmeasure a quadratic function of “manager’s bank investment/networth” (by entered the variable itself and the variable squared). Wedid find instances where the squared term was statistically signifi-cant, and these are presented in the results. Finally, the variable“manager’s bank investment/net worth” was interacted with thehired-manager indicator variable to see if risk responded to this vari-able differently for hired managers compared to owner managers. Inno instance, however, were these additional terms significant(results are available upon request).

Estimation method and issues

The model equation was estimated using ordinary least squares regres-sion. Tests for heteroscedasticity revealed a problem only when theratio of volatile liabilities to total deposits was the dependent variable.To correct, we used White’s heteroscedastic-consistent standarderror in regression (12) of Table A2.2

Appendix: Regression Analysis of Risk, Ownership Structure, and Managerand Ownership Wealth (continued)

2 Hal White, “A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test

for Heteroscedasticity,” Econometrica 48 (1980): 817-38.

36


In this formulation, a measure of risk is the dependent variable,which implies that risk adjusts to ownership characteristics. Othershave argued that ownership structure adjusts to the risk of a firm orindustry, where closer ownership control may be expected in riskierenvironments.3 In the case of banking, however, financial portfoliosare relatively flexible compared to ownership structure. As Saundersand his coauthors note, the transaction cost of changing ownershipstructure would be large relative to the costs of altering a financialportfolio.4 For example, regulatory oversight and geographic restric-tions have raised the cost of hostile takeovers in banking, and havemade them rare relative to nonbank industries.5 We recognize that,in the long run, risk and ownership structure would be determinedsimultaneously but assume that, for the relatively short-run periodwe analyze, risk adjusts to a fixed ownership and managementstructure.

Results

Regression results show that our measures of overall bank risk areinfluenced in varying degrees by ownership structure and the diver-sification of wealth. Credit risk as measured by loan losses and non-performing assets was clearly tied to ownership structure anddiversification of wealth (Table A1.1), but measures of bank prepara-tion for bad loans such as allowances and provisions for loans losseswere not (Table A1.2). Of other balance sheet risks, only the equity-to-asset ratio was significantly related to ownership structure anddiversification of wealth (Table A2). Finally, variation in total incomewas not statistically related to either manager’s bank investment/networth or the ownership share of the hired manager, but was nega-tively related to the monitor’s bank investment/net worth (regression(14), Table A3). On the other hand, the survival likelihood index wassignificantly related to all three of these variables. Variation in oper-ating earnings was statistically related to the ownership share of thehired manager, but not to the manager’s or monitor’s bank invest-ment/net worth (regression (15), Table A3).

3 Harold Demsetz and Kenneth Lehn, “ The Structure of Corporate Ownership: Causes and

Consequences,” Journal of Political Economy 93 (1985): 1155-77.4 Saunders et. al., (1990): 645.5 Stephen D. Prowse, “Alternate Methods of Corporate Control in Commercial Banks,”

Federal Reserve Bank of Dallas Economic Review (Third Quarter 1995): 24-36.

Appendix: Regression Analysis of Risk, Ownership Structure, and Managerand Ownership Wealth (continued)

37


Regression analysis of manager wealth, ownership, and credit risk

Table A1.1

Dependent Variable

Total Loans /Total Assets

Net Loan Losses /Total Loans

NonperformingAssets /

Total LoansPast Due Loans /

Total LoansNoncurrent Assets /

Total Assets

Independent Variable (1) (2) (3) (4) (5)

Constant term 0.526618***

(.02126)

0.003915***

(.00091)

0.017888***

(.00109)

0.033843***

(.00417)

0.011236***

(.00168)

Manager’s bank

investment/net worth

-0.016548

(.01619)

-0.002771**

(.00131)

-0.004556**

(.00206)

-0.006201*

(.00318)

-0.002681**

(.00128)

Manager’s (bank

investment/net worth)2

not entered† 0.000651**

(.00033)

not entered† not entered† not entered†

Hired manager indicator variable 0.016365

(.02440)

-0.003151***

(.00094)

-0.008709***

(.00311)

-0.010435**

(.00479)

-0.005182***

(.00193)

Ownership share of hired manager 0.082268

(.19132)

0.014681**

(.00696)

0.036378

(.02438)

0.039030

(.03756)

0.029017*

(.01514)

Monitor’s bank investment / net worth

× hired manager indicator variable

-0.000695

(.02152)

0.000278

(.00071)

-0.002377

(.00274)

-0.005366

(.00422)

-0.002046

(.00170)

Nonmetropolitan indicator variable -0.035599*

(.02224)

-0.001309*

(.00074)

-0.001377

(.00283)

-0.003792

(.00437)

-0.003389*

(.00176)

Assets 0.000034

(.00029)

0.000003

(.00001)

-0.000018

(.00004)

-0.000028

(.00005)

-0.000008

(.00002)

Assets2 -0.000000

(.00000)

-0.000000

(.00000)

0.000000

(.00000)

0.000000

(.00000)

0.000000

(.00000)

R2 .0301 .0570 .0301 .0507 .0555

Notes: The sample consists of 267 state-chartered banks in the Tenth Federal Reserve District. Reported statistics are coefficient estimates and associatedstandard errors.***, **, and * indicate statistically different from zero at a 1%, 5%, or 10% significance level.† Tests indicated that the coefficient on this variable was not statistically different from zero, and so the variable was not entered for this regression.

38


Regression analysis of manager wealth, ownership, and credit risk

Dependent Variable

Allowance for LoanLosses / Total Loans

Provision for LoanLosses / Total Loans

Other Real EstateOwned / Total Assets

Noncurrent Assets /Allowance for Loan

Losses

Independent Variable (6) (7) (8) (9)


(.00154)

0.000607

(.00153)

0.002082***

(.00078)

1.367990***

(.26572)

Manager’s bank investment/net worth 0.000714

(.00117)

0.000687

(.00116)

-0.000211

(.00059)

-0.294672

(.20232)

Manager’s (bank investment/net worth)2 not entered† not entered† not entered† not entered†


(.00177)

0.000187

(.00175)

-0.001159

(.00089)

-0.678058**

(.30498)

Ownership share of hired manager 0.004233

(.01384)

0.002097

(.01375)

-0.012224*

(.00700)

2.203199

(2.3909)

Monitor’s bank investment/net worth


-0.001192

(.00156)

-0.000022

(.00155)

-0.000755

(.00079)

-0.226297

(.26901)

Nonmetropolitan indicator variable 0.003668**

(.00161)

-0.001143

(.00160)

-0.001928**

(.00081)

-0.269063

(.27793)

Assets 0.000033

(.00002)

0.000016

(.00002)

0.000002

(.00001)

-0.000542

(.00348)

Assets2 -0.000000*

(.00000)

-0.000000

(.00000)

-0.000000

(.00000)

0.000001

(.00001)

R2 .0437 .0079 .0407 .0317

Notes: The sample consists of 267 state-chartered banks in the Tenth Federal Reserve District. Reported statistics are coefficient estimates and associatedstandard errors.***, **, and * indicate statistically different from zero at a 1%, 5%, or 10% significance level.†Tests indicated that the coefficient on this variable was not statistically different from zero, and so the variable was not entered for this regression.

Table A1.2

39


Regression analysis of manager wealth, ownership, and other balance sheet risk

Dependent Variable

Equity / Total assetsPremises and fixed assets

/ Total assetsVolatile liabilities/ Total deposits†

Absolute value(asset-liability mismatch) /

Total assets

Independent Variable (10) (11) (12) (13)


(.00576)

0.010778***

(.01455)

0.090889***

(.01432)

0.128589***

(.01661)

Manager’s bank investment/net worth 0.019801**

(.01015)

-0.000888

(.00111)

-0.003513

(.00842)

-0.006256

(.01265)

Manager’s (bank investment/net worth)2 -0.003413*

(.00243)

not entered‡ not entered‡ not entered‡


(.00592)

0.002804*

(.00167)

-0.011572

(.01138)

0.015255

(.01907)

Ownership share of hired manager -0.084461*

(.04387)

-0.020826

(.01309)

-0.017577

(.08065)

0.018623

(.14948)



0.016560***

(.00451)

-0.000880

(.00147)

0.028009***

(.01005)

0.014589

(.01682)

Nonmetropolitan indicator variable 0.010073**

(.00466)

-0.005330***

(.00908)

0.022827**

(.00908)

0.009995

(.01738)

Assets -0.000093

(.00006)

0.000008

(.00002)

-0.000501*

(.00028)

0.000370*

(.00022)

Assets2 -0.0000002**

(.000001)

-0.000000

(.00000)

0.000000

(.00000)-0.000000(.00000)

R2 .1239 .1018 .1092 .0418

Notes: The sample consists of 267 state-chartered banks in the Tenth Federal Reserve District. Reported statistics are coefficient estimates and associatedstandard errors.***, **, and * indicate statistically different from zero at a 1%, 5%, or 10% significance level.† Tests indicated that the error terms for this regression were heteroscedastic. We corrected for this by using White’s (1980) heteroscedastic-consistent standarderrors.‡ Tests indicated that the coefficient on this variable was not statistically different from zero, and so the variable was not entered for this regression.

Table A2

40


Regression analysis of manager wealth, ownership, and measures of overall bank risk

Table A3

Dependent Variable

Total Revenue Variation: Log(Standard Deviation of TotalRevenue / Average Assets)

Net Earnings Variation: Log(Standard Deviation of Operating

Return on Average Assets)

Survival Index: Log [(Equity +Average Operating Return on

Assets) / (Standard Deviation ofOperating Return on Average

Assets)]

Independent Variable (14) (15) (16)

Constant term -5.791945***

(.23610)

-3.180676***

(.41888)

0.752156

(.46112)

Manager’s bank investment/net worth -0.003614

(.03563)

-0.100298

(.06321)

0.145045**

(.06958)

Manager’s (bank investment/net worth)2 not entered† not entered† not entered†

Hired manager indicator variable 0.139355**

(.05399)

-0.049743

(.09578)

-0.013011

(.10544)

Ownership share of hired manager -0.000582

(.31749)

1.083792*

(.56328)

-1.337034**

(.62007)



-0.179664***

(.04745)

-0.016636

(.08417)

0.155403*

(.09266)

Nonmetropolitan indicator variable -0.038601

(.04864)

-0.124302

(.08630)

0.247042***

(.09500)

Log (Assets) -0.010791

(.02320)

-0.174922***

(.04115)

0.195719***

(.04531)

R2 .0667 .0875 .1125

Notes: The sample consists of 267 state-chartered banks in the Tenth Federal Reserve District. Reported statistics are coefficient estimates and associatedstandard errors.***,**, and * indicate statistically different from zero at a 1%, 5%, or 10% significance level.† Tests indicated that the coefficient on this variable was not statistically different from zero, and so the variable was not entered for this regression.

Date post:	12-Aug-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

How does ownership structure and manager wealth influence ... · ments system and financial...

Documents