Ann P. Bartel -- HRM and Organizational Performance - Evidence From Retail Banking

8/3/2019 Ann P. Bartel -- HRM and Organizational Performance - Evidence From Retail Banking

1/24

Human Resource Management and Organizational Performance: Evidence from Retail BankingAuthor(s): Ann P. BartelSource: Industrial and Labor Relations Review, Vol. 57, No. 2 (Jan., 2004), pp. 181-203Published by: Cornell University, School of Industrial & Labor RelationsStable URL: http://www.jstor.org/stable/4126616 .

Accessed: 17/08/2011 11:21

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

Cornell University, School of Industrial & Labor Relations is collaborating with JSTOR to digitize, preserve

and extend access toIndustrial and Labor Relations Review.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=cschoolhttp://www.jstor.org/stable/4126616?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/4126616?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=cschool


2/24

HUMAN RESOURCE MANAGEMENT AND ORGANIZATIONALPERFORMANCE: EVIDENCE FROM RETAIL BANKING

ANN P. BARTEL*

Studies of the relationship between human resource management and estab-lishment performance have heretofore focused on the manufacturing sector.Using a unique longitudinal dataset collected through site visits to branchoperations of a large bank, the author extends that research to the servicesector. Because branch managers had considerable discretion in managingtheir operations and employees, the HRM environment could vary greatly acrossbranches and over time. Site visits provided specific examples of managerialpractices that affected branch performance. An analysis of responses to thebank's employee attitude survey that controls for unobserved branch andmanager characteristics shows a positive relationship between branch perfor-mance and employees' satisfaction with the quality of performance evaluation,feedback, and recognition at the branch-the "incentives" dimension of a high-performance work system. In some fixed effects specifications, satisfaction withthe quality of communications at the branch was also important.

A growing body of research, includingboth industry-specific studies and cross-industry studies, investigates the impact ofhuman resource management (HRM) onfirm performance.' However, with few ex-ceptions, the prior industry studies focusonly on the manufacturing sector, despitethe fact that most employees work in ser-vice-producing industries. The HRM envi-ronment can be an even more importantdeterminant of productivity in the servicesector than in the manufacturing sector,given the much larger share of total pro-duction costs accounted for byemployment,and the much more extensive direct con-tact between employees and customers, inservices.

This paper extends the analysis of therelationship between the human resourcemanagement environment and establish-ment performance to the service sector byexamining the branch operations of a largeCanadian bank. Previous studies of pro-

*The author is A. Barton Hepburn Professor ofEconomics at the Columbia Business School, andResearch Associate at the National Bureau of Eco-nomic Research. Funding for this paper was receivedfrom the Alfred P. Sloan Foundation via a grant to theIndustrial Technology and Productivity Project of theNational Bureau of Economic Research. The authorgratefully acknowledges helpful comments and sug-gestions received from seminar participants at theColumbia Business School Finance Free Lunch, theNBERSummer Institute, and CIRANO. Sincere thanksto Adriana Lleras-Muney for excellent research assis-tance on this project.Because the data used in this study are proprietaryand were obtained only by signing a confidentialityagreement, the author is unable to release them.

'For a review of studies on HRM and manufactur-ing productivity, see Ichniowski, Kochan, Levine,Olson, and Strauss (1996).

Industrialand LaborRelationsReview,Vol. 57, No. 2 (January2004). ? by Cornell University.0019-7939/00/5702 $01.00

181


3/24

182 INDUSTRIAL AND LABOR RELATIONS REVIEW

ductivity in the banking industry indicatethe importance of getting "inside the blackbox" (Berger and Mester 1997), which canonly be done through detailed analysis atthe plant level, that is, the branch. Todevelop convincing estimates of the effectof HRM on performance, I collected aunique branch-level data set through sitevisits to the bank's headquarters and itsbranches.Several features of the data used for thisstudy help make estimates of the effects ofHRM on performance especially convinc-ing. First, since the branches are produc-ing the same products using the same pro-duction process, it is possible to estimatethe impact of the human resource manage-ment environment while greatly limitingthe confounding impact of unmeasuredattributes of the production process, a prob-lem that plagues cross-industry studies.Second, I collected performance and HRMdata for two different time periods in eachbranch, which allows for fixed effects esti-mation to control completely for any un-measured branch-specific effects that maybe correlated with the HRM environment.Third, the most likely time-varying factorthat may be correlated with HRM andthereby bias estimated effects of HRM is"manager quality" or "manager style." Ihave collected data on the identities of allbranch managers in both time periods andhave estimated models that control for thispotentially important time-varying factor.Fourth, branch managers are given consid-erable discretion in how to manage theirbranches and their employees, so the HRMenvironment can vary considerably acrossbranches and over time.To describe the human resource man-agement environment at each work site, Iuse employee perceptions of human re-source policies and work practices. Thisapproach has two advantages over that usedby most previous studies of human resourcemanagement and organizational perfor-mance. First, an employee survey providesinformation directly from the individualsworking at a site, rather than a manager'sdescription of what he or she perceives asthe environment, or, worse still, an ideal-

ized description of the environment. Thehigher the level of the manager who com-pletes the survey, the more limited his orher knowledge of what is actually happen-ing at the workplace. Second, since anemployee surveyfields many responses fromeach worksite, one person's idiosyncraticopinion or interpretation of the questionsis less likely to distort the results.

Prior LiteratureEconomic Literature onProductivity in the Banking IndustryThe economic literature on productivityin the banking industry has focused mainlyon how scale affects bank or branch effi-ciency. These studies typicallyuse the value-added or production approach, which viewsbanks as "producing"demand deposits, timeand savings deposits, commercial loans, realestate loans, and installment loans, usingcapital, labor, and materials to do so.2 Thebest example of a branch-level productivitystudy is Berger, Leusner, and Mingo(1997).3 Using data on 760 branches in a

2Theres also a literature nbranchproductivitythat uses dataenvelopment nalysis DEA). DEAcomparesachbranchwithallof the otherbranchesin the observation et and identifies he relativelyefficient(bestpractice) ubsetof branches ndthesubsetof brancheshatarerelativelynefficient.Inthesestudies,output s measured s the numberoftransactionsforexample,newaccounts, losedac-counts, oanapplications,heckscashed, ravelers'checks old)processed ythebranch, nd nputsarenumberof employees,office space,and supplies.Many tudies hat use DEAuse a smallnumberofobservationsforexample, herman ndGold 1985]used 14 branches,and Parkan[1987] used 35branches) elative o the number f inputsand out-putsandaretherefore redisposedo find thatmostbranches re efficient.Anexceptions the workbySchaffnit, osen, ndParadi1997),who tudied291Ontario-based ranchesof a largeCanadian ankand found that about50%of the branchesweretechnicallyfficient.3Other examples of parametric branch-level pro-ductivity studies are Murphy and Orgler (1982), whichestimated a Cobb-Douglas cost function for one year(1976) on 127 branches of an anonymous bank in asmall country; Doukas and Switzer (1991), which


4/24

HRM AND ORGANIZATIONAL PERFORMANCE IN RETAIL BANKING 183

large U.S. commercial bank for the timeperiod 1989-91, the authors found thatmost branches fell considerably short ofthe optimally efficient size but their aver-age cost curves were relatively flat. Onlyone paper by economists has consideredother correlates of efficiency, and it wasconducted at the bank (company) level,not the branch level. Berger and Mester(1997) used data from 6,000 U.S. commer-cial banks to estimate the performance ef-fects of bank size, bank age, organizationalform and governance, market characteris-tics, and state geographic restrictions oncompetition. They found that most of thevariance in measured efficiency remainedunexplained, and they attributed this tounmeasured factors such as differences inmanagerial ability; they concluded that thesources of the variation in bank efficiencyremain a "black box."Human Resource Managementand Organizational Performance

The "blackbox" in the banking industrymay indeed be the human resource man-agement policies and practices used bymanagers. A large body of research hasdocumented the link between human re-source management and organizationalperformance, primarily in the manufactur-ing sector. These studies typically fall intoone of two categories: (1) national cross-industry studies (Black and Lynch 2001;Bresnahan, Brynolfsson, and Hitt 2002;Cappelli and Neumark 2001; Huselid 1995);or (2) intra-industry studies (Kleiner,Leonard, and Pilarski 2002; Batt 1999, 2002;Boning, Ichniowski, and Shaw 2001;Appelbaum et al. 2000; Ichniowski, Shaw,and Prennushi 1997; Kelley 1996; Delery

and Doty 1996; Youndt et al. 1996; Dunlopand Weil 1996; MacDuffie 1995).In their reviews of this literature, Beckerand Gerhart (1996) and Delery (1998) cau-tioned researchers to develop a theoreticalbasis for the human resource managementconstructs they use in their empirical analy-sis. The widely accepted theoretical basisfor the relationship between human re-source management and organizationalperformance is the high-performance worksystem framework provided by Appelbaumet al. (2000). At the core of a high-perfor-mance work system, according to Appel-baum et al., is an organization that enablesnon-managerial employees to participatein substantive decisions. The high-perfor-mance work system also requires support-ive human resource practices that enhanceworker skills and that provide incentivesfor workers to use their skills and partici-pate in decisions. Appelbaum et al. (2000)showed how these three elements of a high-performance work system-opportunity toparticipate, skills, and incentives-contrib-uted to productivity in three manufactur-ing industries.The Service Sector Setting

With the exception of Batt (1999, 2002),Banker et al. (1996), and Delery and Doty(1996), all of the prior research on humanresource management and organizationalperformance has focused on the manufac-turing sector, despite the fact that todaymost employees work in service sector in-dustries. Services differ from goods in threeimportant ways: they are intangible, theytend to be produced and consumed simul-taneously, and they tend to involve theconsumer in their production and delivery(Bowen and Schneider 1988). The simulta-neous delivery and receipt of services in theface-to-face service sector brings employ-ees and customers close together, blurringthe boundary between the two groups(Parkington and Schneider 1979). Thedirect contact that exists between the em-ployee and the customer in the service sec-tor suggests that human resource manage-ment may be even more important in the

estimated a translog cost function using one calendarquarter of data (October 31, 1985, to January 31,1986) on 563 branches of an anonymous Canadianbank; and Zardkoohi and Kolari (1994), which esti-mated a translog cost function using 1988 data on 615branches of 43 Finnish savings banks.


5/24


service sector than in the manufacturingsector.In her study of telecommunications callcenters, Batt (2002) argued that the high-performance work system is likely to havean important impact on organizationalperformance in customer service settingsbecause "high involvement practices helpemployees develop the kind of firm-spe-cific human capital-knowledge of thefirm's products, customers, and work pro-cesses-that enables them to interact effec-tively with customers." Indeed, organiza-tions that compete in sales and service de-livery often use a "relationship manage-ment" strategy in which they seek to buildlong-term relationships with customers byproviding high-quality service.4 Heskett etal. (1997) provided evidence in support ofwhat they called a "service profit chain."Using data on six companies, they foundthat companies providing high-quality ser-vice have satisfied and loyal customers andsatisfied and loyal employees; they arguedthat satisfied customers lead to satisfiedemployees and vice versa. Further, theyfound that companies whose customers aresatisfied with service quality exhibit rev-enue growth-hence the "service profitchain." A key link in the "service profitchain" is a high-performance work system(Batt 1999).Human Resource Managementin the Banking Industry

A few scholars have studied the impact ofhuman resource management on perfor-mance in the banking industry, but thesestudies have important methodological limi-tations. Delery and Doty (1996) conducteda survey of senior human resource execu-tives in U.S. banks in order to obtain infor-

mation on the human resource policiesused by the banks for their loan officers.5Using a cross-sectional framework that ig-nored the role of bank fixed effects, theyfound a positive correlation between thebank's returns on assets and equity and theexistence of profit-sharing and employmentsecurity for loan officers, controlling forthe size and age of the bank. Frei, Harker,and Hunter (2000) have shown that X-efficiency, or how well management alignstechnology, human resources, and otherassets to produce a given level of output,plays an important role in the bankingindustry. It is important to note that bothof these studies used cross-sectional data,and the possibility thus remains open thata longitudinal study controlling for bank-specific fixed effects would produce differ-ent results. The use of cross-sectional dataalso characterizes the other work that hasbeen done on the impact of human re-source management in the service sector(for example, Batt 1999, 2002; Banker et al.1996).A second limitation of the two bankingstudies is that the analysis was done at thelevel of the bank. While the ability of thebank's managers at the firm or headquar-ters level can certainly affect the bank'sperformance, much of a bank's activitiesoccur at the branch level. In retail banking,customers have idiosyncratic needs, andthe interactions between these customersand bank employees take place at the branchlevel. Hence, the role that the managermight play in creating a high-performancework environment that will contribute toperformance is best studied at the branchlevel.One study that used branch-level data isSchneider and Bowen (1985), which ana-lyzed data from employees and customersin 28 branches of a U.S. bank and tested thehypothesis that branch employees' percep-tions of organizational human resourcespractices are positively correlated with

4Keltner (1995) found that a strategy of relation-ship banking coupled with cultivation of highly skilledand trained employees was a statistically significantfactor explaining why German banks outperformedU.S. banks in the 1980s. 5Their survey had a response rate of 11%, resultingin a sample of 216 banks.


6/24


branch customers' attitudes about service.Schneider and Bowen argued that the posi-tive correlation would exist because em-ployees who perceive their organization asone that facilitates performance, enhancescareer opportunities, and provides positivesupervision will be free to do the organi-zation's main work of serving customers.The study's main finding was that custom-ers' attitudes about overall service qualityat the branch were positively correlatedwith employees' ratings of the branch onthe quality of supervision, work facilitation,and career facilitation. While this studyhad the virtue of being conducted at thebranch level, its data were only cross-sec-tional, and the results could disappear in alongitudinal analysis that includes branch-specific fixed effects. Another limitation ofthe study is that it did not examine theimpact of the employee perceptions on theactual performance of the branch; the only"performance" measure that was used wasthe customers' intentions to leave thebranch. Hence, to date no study has usedlongitudinal data to analyze the impact ofthe human resource management environ-ment on branch-level performance in thebanking industry. The current paper fillsthis gap.

Getting Inside the "Black Box":How HR Practices AffectPerformance in Retail Bank BranchesThe New Environment in Banking

Although there are only five banks in theCanadian banking industry, Canada hasthe highest ratio of full-banking branchesto population of all the major industrial-ized nations (Canadian Bankers Associa-tion 1994). The availability of numerousretail branches, coupled with reforms thathave allowed banks to expand their prod-uct lines, has resulted in a very competitiveenvironment in which much attention ispaid to opportunities to increase the prof-itability of retail banking. In addition, tech-nological change has resulted in a majororganizational redesign in the Canadianbanking industry. Many paper-processing

tasks typically performed by branch per-sonnel have been moved offsite to "central-ized accounting units," thereby radicallychanging the tasks performed by branchpersonnel.6 For example, in the past, tell-ers simply processed customers' transac-tions. Today, they are evaluated on thebasis of their ability to sell various financialproducts or make referrals to the propersales personnel. In the words of the execu-tive vice-president of human resources atthe bank used for this study, "Sales is nowthe name of the game in this industry." Inthe new sales-oriented environment,branches are evaluated based on their salesof products.Insights from Branch Visits

Although the bank under study has aformal set of human resource policies re-garding job descriptions, salaries for par-ticular jobs, performance appraisals, andfeedback, the actual implementation ofthese policies differs across branch manag-ers. In my interviews with executives at thebank's headquarters, I learned that somebranches were considered to have excel-lent HRM environments (in the sense ofpositively affecting branch performance)whereas others were viewed less favorably.In order to understand how a branch man-ager might create a human resource man-agement environment that could affectbranch-level performance, I gathered datadirectly from managers and employees inten branches during the fall of 1995 andthe winter of 1996. I asked the bank head-quarters to select branches that representedthe range of HRM environments as per-ceived by headquarters. One day was spentin each branch, meeting first with the man-ager and then individually with five or sixemployees in different positions (tellers,

6SeeAutor, Levy, and Murnane (2000) and Hunter,Bernhardt, Hughes, and Skuratowicz (2000) for adiscussion of the various ways in which technologicalchange has affected job content and earnings at anumber of U.S. retail banks.


7/24


personal banking officers, customer ser-vice representatives, accounting clerks, andso on). These interviews proved to be aninvaluable component of the researchagenda, as they provided specific examplesof how managers could create a high-per-formance work environment that couldcontribute to the branch's performance.The interviews clarified the process bywhich branches make sales. The observedsales of a branch during time period t are afunction of the amount of contact the staffhas with customers and the probability thata given interaction with a customer leads toa sale. Customer contact depends on thevolume of customer traffic at the branch aswell as the number of contacts personalbankers have (both in person and by tele-phone) with existing and potential custom-ers. The probability of a sale given contactdepends both on the characteristics of thecustomer (for example wealth and age)and on the ability of the branch employeeto make a sale. The latter in turn is depen-dent on the employees' experience at thebranch (more branch-specific experienceleads to stronger relationships with cus-tomers) as well as their product knowledgeand motivation to sell.Below, I describe visits to two branchesthat show how the three dimensions of ahigh-performance work system postulatedby Appelbaum et al. (2000) contribute tothe productivity of branch employees. Iselected these two branches because theyprovide interesting contrasts in the way thebank's formal HRM policies are imple-mented, thereby creating different humanresource management environments at thebranch level.

Branch #1. As of the date of the branchvisit in January 1996, the branch managerhad been at the branch for almost twoyears. During the time I spent at the branch,the manager rarely left his office to moni-tor activities in the branch or to interactwith his employees. The manager com-plained that the employees were apatheticand that the branch was not operating atpotential.The interviews with the employees con-

firmed much ofwhat the manager discussed,and employees blamed him for the unsatis-factory work environment and the medio-cre performance of the branch. Specifi-cally, the employees complained about theprocess of receiving feedback and the re-ward and recognition system. For example,at least two of the employees complainedabout the manager's tendency to give nega-tive feedback in front of customers and hisencouraging employees to "snitch"on otheremployees. Some of them also complainedthat the manager did not provide real rec-ognition of employees who performed well.One teller discussed how she workedthrough her lunch hour to generate a re-ferral for a mortgage but received no recog-nition. There was a general sense that theemployees were not cooperating with eachother; for example, one employee recalleda recent occasion in which there was a longline and only one teller wasworking, but noone bothered to pitch in. This branchclearly lacked the attributes of a high-per-formance work system.

Branch #2. Like the manager in Branch#1, the manager in Branch #2 had beenwith the branch for almost two years as ofthe date of the branch visit (November1995). Headquarters selected this branchbecause they felt this manager was a modelto which other managers should aspire.Interviews with the manager and with sev-eral branch employees as well as generalobservations confirmed that the humanresource management environment in thisbranch was a high-performance work sys-tem.One of the dimensions of a high-perfor-mance work system is the "opportunity toparticipate," and a key element ofAppelbaum et al.'s (2000) "opportunity toparticipate" scale is the extent of communi-cations with peers and with supervisors. Inthis branch, the manager held regular staffmeetings to encourage communicationsbetween peers and between employees atdifferent levels or in different functions inthe branch. At these meetings, the employ-ees were encouraged to identify areas forimprovement and to make suggestions for


8/24


change. The employees commented posi-tively about the weekly staff meetings. Asone employee put it, "We now know what isgoing on in all departments." Communica-tions between the staff and the managerwere described by the employees as excel-lent. "You can talk to her [the manager] ...she's one of us," commented an employee.The second dimension of a high-perfor-mance work system-skills-was also evi-dent in this branch. Unlike the manager inBranch #1, this manager appeared genu-inely interested in ensuring that her em-ployees had the proper skills to do theirjobs. In order to influence the actual skillsof her employees, at the staff meetings shewould teach her employees about new prod-ucts and how to sell them. She often usedgames to teach the employees about a newproduct, and this approach appealed to theemployees, who commented positivelyabout the party atmosphere at the meet-ings.Also evident at this branch was the thirddimension of a high-performance work sys-tem-incentives. For example, the man-ager held contests with small monetaryprizes to motivate her employees. Theemployees commented favorably on theperformance feedback and reward systemat the branch. "She keeps me posted onhow I am doing," said one employee.7Unlike the employees in Branch #1, theBranch #2 employees felt that the managerrecognized when they did a good job andrewarded them, even if only with a giftcertificate or a half-day off. The managerin Branch #2 also motivated her employeesto use their skills and to provide service tothe customers by being a hands-on man-ager who rarely sat in her office. While Iwas at the branch, she was frequently at thetellers' platform, either assisting the tellerswith questions or actually pitching in as ateller when the lines got long. Her ex-

ample was more effective at getting othersto pitch in than was any formal directive orpolicy issued by the manager at Branch #1.Finally, the personal bankers who were re-sponsible for generating new loan businessreported how they stayed after hours tocold-call potential clients; by contrast, inBranch #1, personal bankers worked "bythe book," leaving as soon as the regularbusiness day ended.Implications for the Impactof High-Performance WorkPractices on Branch PerformanceThese interviews indicated that althoughthe company has a set of formal humanresource policies for its branches, branchmanagers have discretion in their applica-tion of these policies. Some branches havehuman resource management environ-ments that can be characterized as high-performance work systems, while othershave more traditional systems.8 The branchvisits demonstrated how branch-level per-formance can be influenced by the threedimensions of a high-performance worksystem. Branch performance is measuredby the sales of deposit and loan products,and, as in Heskett et al.'s (1997) serviceprofit chain, branches will make more salesif customers are satisfied with the quality ofservice at the branch. The branch visitsdemonstrated that service quality will behigher in a high-performance work system.First, customer needs are more likely tobe satisfactorily addressed if employees havethe proper skills. Employees need a thor-ough understanding of the attributes of thebank's various products, and they also needbranch-level experience in order to under-

7This is a good example of an employee receivingfeedback from the supervisor. See Hollenbeck et al.(1998) on the impact of feedback on team perfor-mance.

8Given that the human resource management en-vironment varies across branches, a study of the im-pact of human resource management on performancein the banking industry must be done at the branchlevel, not at the bank level. Table 2 (below, in the"Results" section) documents the variation in thehuman resource management environment that ex-ists across the branches in the bank under study.


9/24


stand the specific needs of their customers.Second, an environment in which em-ployees can easily communicate with co-workers or managers will enable them touse their skills more effectively. Ease ofcommunication as well as a more coopera-tivework environmentwill enable customer-contact employees such as bank branchemployees to respond more quickly andmore effectively to customer demands.Third, employees need to have an incen-tive to devote effort to meeting customerneeds. They are more likely to make aneffort if they feel that their performance isevaluated accurately and that their effortsare recognized and rewarded.These three observable components ofthe HRM environments in the twobranches-skills, or specifically, productknowledge; quality of communications; andrecognition and reward-are very similarto the three dimensions of a high-perfor-mance worksystem described byAppelbaumet al. (2000). Recall thatAppelbaum et al.'sthree dimensions are skills, opportunity toparticipate in substantive decisions, andincentives to use skills and participate indecisions. The first and third factors Iidentified are very close matches to thecorresponding Appelbaum et al. factors,and my second item is closely related toAppelbaum et al.'s second item because akey element of Appelbaum et al.'s "oppor-tunity to participate" scale is the extent ofcommunications with peers and with su-pervisors. Hence, we would expect to seepositive relationships between the threeobservable components of the branch-levelHRM regimes and branch-level perfor-mance.

MethodologyRecent reviews of the literature on hu-man resource management and organiza-tional performance (Becker and Gerhart

1996; and Delery 1998) have identified fourmethodological issues that researchers inthis area need to consider: (1) the appro-priate measure of organizational perfor-mance given the context of the study; (2)whether human resource management

practices should be measured at the firmlevel or, instead, at the business unit orfacility level; (3) possible omitted variablesthat could bias the estimated relationshipbetween human resource management andorganizational performance; and (4) theextent to which the estimated coefficientson the human resource variables can beinterpreted as showing a causal relation-ship between human resource managementand organizational performance. In thissection, I describe my empirical methodol-ogy and indicate how my approach ad-dresses each of the four concerns.Defining Branch Output

The interviews I conducted with numer-ous branch managers and finance and ac-counting managers at headquarters indi-cated that, in the new sales-oriented envi-ronment, branches are evaluated based ontheir sales of products. In other words, agood branch is one that shows growth ofdeposits and loans.' This is because thelargest component of a branch's income isits "spread"income."' Each financial prod-uct a branch offers has a certain "spread"factor that equals the profit margin on theproduct." According to the managers in-

9When the performance of an entire bank is beingmeasured, a metric known as the efficiency ratio isoften used. This is defined as Non-interest expense/(Interest income + Non-interest income - Interestexpense). It is misleading to use the efficiency ratioto compare performance across individual branchesof the bank, because when, as often happens, custom-ers open accounts at one branch but use otherbranches to conduct subsequent business, the branchthat opens the account gets credit for the spreadincome even though the other branches incur theexpenses. Focusing on sales and the spread incomederived from sales focuses on the income that branchesderive for the bank.'0The other components of a branch's income areliability fees, such asfees from stop payments, bouncedchecks, low balances, wire transfers, and so on; asset

fees, such as fees from loan applications, loan pro-cessing, and late payments; transactional fees, such asfees for travelers' checks, safe deposit boxes, andATM transactions; and brokerage commissions.11Branch managers have no discretion in settingthe spread; the interest rates on the bank's productsare set at headquarters.


10/24


terviewed, branches are evaluated based ontheir sales of financial products becausegrowth in deposits and loans on a branch'sbalance sheet translates into an increase inspread income and, thereby, a larger finan-cial contribution to the bank's perfor-mance.12Specification of theBranch Output Equation

Branch-level sales of deposit and loanproducts are a function of capital and laborinputs, the characteristics of the neighbor-hood in which the branch is located and ofthe individuals who live there, the branchmanager's knowledge of the neighborhood,and the human resource management en-vironment in the branch. The specificequation that will be estimated is(1) SALESi = a + PlnK, + ylnL+ yXHRMit + TMKTi + 8MGRTENURE.ij+ YEAR + b. + m.+ "itwhere SALESit is the annual percentagechange in deposit or loan balances inbranch i at time period t, K.,is a measureof the branch's capital stock at time pe-riod t, Litis the number of employees atthe branch at time period t, HRMi, is avector describing the human resourcemanagement environment at the branchat time period t, MKTiis a vector of char-acteristics describing the neighborhoodin which the branch is located,MGRTENURE.. is the tenure of branch man-ager j in branch i at time t, and YEARs avector of time dummies that measuretime-varying effects that are common to

all branches.13 With the exception ofMKTi, all the variables in equation (1) varyover time.Since the dataset provides manager iden-tities, I am able to include two fixed effectsin equation (1); the first, b., is a branchfixed effect, and the second, m, is a man-ager fixed effect. This specification allowsfor the existence of permanent, unmea-sured branch characteristics that may af-fect performance, as well as permanent,unmeasured characteristics of individualmanagers that may be correlated with per-formance. The random, unobserved errorcomponent is denoted as i,,. Note that thespecification in equation (1) deals with thetwo remaining methodological concerns.First, omitted variable bias is unlikely to bea problem, because the specification isbased on discussions with professionals inthe banking industry. Second, inclusion ofbranch fixed effects and manager fixedeffects means that the coefficients on theHRM variables can be interpreted as esti-mates of the causal effect of HRM on orga-nizational performance.Finally, interviews with senior manage-ment at the bank indicated that bank man-agement was emphasizing loan productsales over deposit product sales for tworeasons. First, loan products have biggerspreads. Second, a lending relationshipwith a customer provides a natural oppor-tunity to discuss the sale of additional prod-ucts. Indeed, the importance of loan busi-ness in the banking industry is supportedby case studies of bank mergers that oc-curred in the United States in the 1990s.14This suggests that the human resource ac-

12Hunter and Hitt (2001) used a sample of 235branches from 101 different banks. Since their samplewas not restricted to one bank, their choice of mea-sures of branch-level performance was dictated by theneed for comparability across banks. The measuresthey chose were the number of checking accountsand the number of financial products held per cus-tomer. In order to avoid the comparability problem,the analysis of performance is best done acrossbranches within one bank.

"3Thisspecification is derived from the Cobb-Dou-glas production function Q= AKELY, where Q s thedollar value of sales; K is the capital stock and EL iseffective labor, with EL = L(1 + XHRM); L is thenumber of employees; and HRM is the HRM environ-ment. HRM transforms employees into effective la-bor.

14Calomiris and Karceski (1998) discussed howopportunities for loan and mortgage originationsinfluenced banks' selection of acquisition targets.


11/24


tivities undertaken by the branch managerare likely to be targeted more at influenc-ing those employee activities that are corre-lated with the sale of loan products than atinfluencing deposit products. Hence, equa-tion (1) will be estimated separately fordeposits and loans.Sample

The sample that is used to estimate equa-tion (1) was constructed as follows. In1995, 333 branches were in operation inthe province of Ontario. These branchesoperate in very different environments,which can affect their ability to sell theirproducts. For example, some branchesoperate in downtown business areas andmany of their customers are large busi-nesses. Even within the group of branchesthat operate in residential areas, importantdifferences in customers' age and wealthcan affect performance.Based on my conversations with manag-ers in the bank's marketing department, Ieliminated branches that function as largecommercial banking centers'5 as well asbranches in rural areas, resulting in a samplerestricted to branches in metropolitan ar-eas.'6 Second, a branch was excluded if ithad not been in operation for at least oneyear prior to the start of the 1995 fiscal year.Third, in order to estimate the fixed effectsmodel, data from the employee opinionsurvey had to be available for at least twoyears (either 1995 and 1996, 1995 and i -7,1996 and 1997, or 1995, 1996, and 1997).Hence, branches for which only one year ofemployee opinion survey data were avail-able were excluded. The resulting sampleconsists of 160 branches, 150 of which havetwo years of data and 10 of which have three

years of data, for a total of 330 observa-tions.'7Definitions of Variables

Sales. A branch's annual sales are calcu-lated as either the percentage growth indeposits or the percentage growth in loansfrom the last day (October 31) of the previ-ous fiscal year to the last day (October 31)of the current fiscal year. Table 1 showsthat net sales of both deposits and loanswere declining over the 1995-97 time pe-riod, which the bank managers attributedto increased competitiveness in the indus-try. Year dummies are included in equa-tion (1) to control for this phenomenon.Capital stock (K). The branch's capitalstock is measured by the value of its prop-erty and premises in each fiscal year.Number of employees(L). The branch'slabor input is measured by the numbers offull-time and part-time employees in eachfiscal year. As shown in Table 1, the averagenumber of employees in a branch is ap-proximately 18.Characteristics f thebranch'smarket MKT).For each of the branches in my sample, Iwas able to obtain detailed informationabout the branch's location. In particular,the bank defines a branch's "market"as thearea within a 2.5-kilometer radius of thebranch, and it gathers data on the popula-tion residing within that circle. Five vari-ables were provided for each branch's mar-ket: total population, average dwellingvalue, education, household turnover, anda "lifestyles"vector that describes the typeof people living in the area. There are tenlifestyle categories-affluent, empty nest-ers, ethnic, low-income, middle-class, up-scale, working-class, young singles, youngcouples, and old/retired-and the mostcommon of these population "types" n thebranch's market is identified. All of thesemarket characteristics are measured in15Doukas and Switzer (1991) found that the pro-duction technologies of retail and commercialbranches are quite distinct.16The metropolitan areas are the cities of Toronto,Ottawa, London, Windsor, Hamilton, Kitchener,

Niagara Falls, and Peterborough, together with theirsurrounding communities.17Among the 150 branches with two years of data,102 have data for 1995 and 1997, 25 have data for1995 and 1996, and 23 have data for 1996 and 1997.


12/24


Table 1. Summary Statistics.1995 1996 1997

Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.Annual Growth Rate of Deposits 0.12 0.12 0.10 0.11 0.08 0.10Annual Growth Rate of Loans 0.19 0.13 0.10 0.08 0.05 0.10Average Dwelling Value 192,148 63,843 179,432 55,432 194,338 64,897Total Population 45,013 35,312 45,605 34,428 46,897 35,320Proportion with Post-Secondary Education 0.50 0.10 0.48 0.10 0.50 0.10HH Turnover Rate 0.06 0.02 0.06 0.02 0.06 0.02Affluent 0.03 0.17 0.02 0.13 0.03 0.17Empty Nesters 0.15 0.35 0.16 0.37 0.14 0.35Ethnic 0.07 0.25 0.10 0.31 0.07 0.26Low 0.06 0.23 0 0 0.06 0.23Middle 0.15 0.36 0.17 0.38 0.16 0.37Upscale 0.28 0.45 0.26 0.44 0.30 0.46Work 0.10 0.30 0.12 0.33 0.10 0.30Located in Mall 0.09 0.28 0.17 0.38 0.12 0.32Age of Branch (in years) 34.55 24.38 37.68 24.71 36.46 24.53Manager's Tenure in Branch 4.02 3.61 4.46 3.76 4.32 3.64Number of Full-Time Employees 8.94 4.22 10.77 4.92 9.94 4.18Number of Part-Time Employees 8.58 5.74 9.74 5.17 7.51 4.27Value of Property and Premises 131,017 91,809 142,322 81,656 143,175 93,881Average Education (Employees) 12.71 0.57 12.82 0.53 13.01 0.55Average Tenure (Employees) 4.64 2.30 4.55 2.22 4.43 2.13

N 137 58 135

1991. In addition to these variables, I cre-ated a dummy variable to indicate if thebranch is located in a shopping mall,'s andI also control for the age of the branch.'"The mean age of branches in the sample isapproximately 35 years.Manager tenure (MGRTENURE). his is thelength of time the current branch managerhas been managing this branch as of theend of the fiscal year. Table 1 shows thatthe average tenure of a branch manager isfour years.

Human resourcemanagementenvironment(HRM). As shown above ("Getting Insidethe 'Black Box"'"), he three dimensions ofa high-performance work system (highskills; opportunity to participate, or, spe-cifically, communications; and effectiveincentives) were observed in some of thebranches I visited, but not in others. Oneway to empirically measure these dimen-sions for all the branches would be to con-duct interviews of employees and managersat each of the branches. The size of thesample, however, makes that approach in-feasible. An alternative approach, which Ifollow here, is to use data from the em-ployee attitude survey.Non-managerial employees (both full-time and part-time) in each branch com-plete a survey once every year or two thatmeasures their assessment of a number ofdimensions of the human resource envi-ronment at their branch. The mean re-sponse rate on the employee survey was

18Executives at the bank suggested that thesebranches were likely to have high sales because of thelarge concentration of potential customers.19Itwas impossible to obtain accurate informationabout the number of competitors in each branch'smarket area. The bank's data source on number of

competitors automatically includes any credit unionthat is located within the defined market area; hencemany branches are shown to have twenty or morecompetitors.


13/24

Table 2. Human Resources Attitude Survey Questions.a1995

VariableName Definition Mean Std. Dev.(A) Overall1. Overall Rating Overall, how would you rate the bank as a place to work? 3.49 0.29(B) Communication2. Communications from Peers Rating of communications to you from others on same level. 5.05 0.643. Communication Upward Rating of opportunities to communicate upward. 5.10 0.494. Communications from

Superiors Rating of communications to you from superiors. 4.88 0.655. Overall Communications Overall ratings of communications. 4.91 0.61(C) Performance and Reward6. Understand Perf. Evaluation I have a clear understanding of how my performance isevaluated. 3.79 0.34

7. Contributions Recognized When things go well in your job, how often are yourcontributions recognized? 3.27 0.468. Frequency of Feedback How would you rate your supervisor on letting you knowhow you are doing your job on a regular basis? 3.99 0.37

(D) Climate9. Express Views I feel comfortable expressing my views/suggestions atbranch meetings. 3.71 0.4310.Morale Morale is high in my department. 3.23 0.4611.Cooperation How would you rate your branch on cooperation amongemployees? 2.91 0.5112.Supervisor Accessible How would you rate your supervisor on being easy to seewhen you have a problem? 3.60 0.54

(E) Skill13.Understand Products I have a good understanding of the bank's products andservices that I am expected to promote/sell. 4.16 0.25aResponses are scored from 7 (extremely good) to I (extremely poor) for questions 2-5; all other questions are scoreto 1 (strongly disagree, very poor, or never).


14/24


80% and the mean number of responses(per branch) was 14 in each of the threefiscal years under study. The bank pro-vided me with the responses to 13 of the 68questions on the employee attitude survey;the excluded questions concerned satisfac-tion with employee benefits, training pro-grams, job security, and physical condi-tions in the branch. The included ques-tions, listed in Table 2, focused on theemployee's assessment of performance andrecognition at the branch, the nature ofcommunication flows between the man-ager and staff and between co-workers,morale, the level of cooperation, and acces-sibility of the supervisor.The responses to the employee surveymeasure the workers' perceptions of vari-ous dimensions of the human resourcemanagement environment at the branch,rather than the incidence of specific HRpractices.20 For example, the questions on"Communications" ask the employees toevaluate the quality of communications atthe branch, rather than to indicate thenumber of times they speak with their man-ager or co-workers during the week. As Ishow above (in the Results section), thenumeric responses to the employee surveyare consistent with the observations madeduring the branch visits, suggesting thatthere is a close relationship between theresponses to the employee attitude surveyand actual practices at the branches.The responses to the employee surveywere used to construct four measures (seeTable 2) to proxy the three dimensions ofAppelbaum et al.'s (2000) high-perfor-mance work system. The first constructedmeasure, Communication, s an index basedon four communications-related questionsfrom the employee survey (communications

from peers, communication upward, com-munications from superiors, and overallcommunications).21 Recall that communi-cation is a key element of Appelbaum etal.'s (2000) "opportunity to participate"scale, one of the dimensions of a high-performance work system. To create theindex, I transformed the 1-7 Likert re-sponse format for each question (where 7 isthe best) to a 0-100 scale, and calculatedthe mean value of the four variables.A second constructed index, Climate,isused to measure the extent to which theenvironment in the branch encouragesparticipation. Climate s an index based onanswers to four questions, concerning com-fort in expressing views/suggestions, levelof morale, degree of cooperation amongemployees, and accessibility of the supervi-sor.22 To create this index, I transformedthe 1-5 Likert response format for eachquestion (where 5 is the best) to a 0-100scale, and calculated the mean value of thefour variables.In order to measure the "incentives" di-mension of a high-performance work sys-tem, an index called Performance nd Rewardis created. Performanceand Reward s basedon three questions from the employee sur-vey concerning, respectively, understand-ing of how performance is evaluated, howoften contributions are recognized, andfrequency of feedback from the supervisor.The procedure for constructing this indexis identical to that used for Climate.23Finally, the last dimension of a high-performance work system, namely high rela-tive skill requirements, is proxied by anindex called Skill that uses the response toa question on the employee survey regard-

20Thisapproach is superior to using the sole sourcemanager survey response used in most previous re-search. An example of a prior study that used em-ployee survey responses is Schneider and Bowen(1985), which studied the correlation between branchemployees' perceptions of HR practices and branchcustomers' attitudes about service.

21Cronbach's alpha is .92 in 1995, .88 in 1996, and.90 in 1997.22Cronbach's alpha is .67 in 1995, .72 in 1996, and.75 in 1997. Accessibility of supervisor appears tobelong in this index, not in the Performanceand Re-ward index. When it was included in the Performanceand Reward ndex, the Cronbach alphas were lower.23Cronbach's alpha for Performanceand Reward is.72 in 1995, .62 in 1996, and .32 in 1997.


15/24

Table 3. Pairwise Correlations of Employee Attitude Survey Questions1 2 3 4 5 6 7 8 9 10 11 12

1 Overall 1.00002 Perf./Reward 0.5346 1.00003 Climate 0.5508 0.6017 1.00004 Communication 0.4259 0.6645 0.7472 1.00005 Skill 0.2121 0.1620 0.1390 0.0542 1.00006 Recog. 0.3579 0.7291 0.5900 0.7483 -0.0305 1.00007 Perf. Eval. 0.4637 0.7089 0.5125 0.5439 0.1940 0.5012 1.00008 Freq. Feedback 0.2535 0.5199 0.0948 0.0180 0.1864 -0.1031 0.0443 1.00009 Express 0.4018 0.5234 0.7550 0.7180 0.0670 0.5567 0.4058 0.0641 1.000010 Coop. 0.4587 0.4286 0.8185 0.5205 0.1799 0.3984 0.3743 0.0841 0.4680 1.000011 Sup. Access 0.3509 0.3003 0.7040 0.3350 0.0846 0.2423 0.3010 0.0712 0.3158 0.4713 1.000012 Morale 0.4911 0.6287 0.8198 0.7764 0.0868 0.6566 0.5127 0.0724 0.6490 0.5429 0.3606 1.000013 Comm. Peer 0.4366 0.5834 0.7214 0.8899 0.0417 0.6664 0.4860 -0.0000 0.6651 0.5538 0.3429 0.697814 Comm. Up 0.3898 0.6735 0.6047 0.8189 0.0957 0.5599 0.4369 0.3056 0.5983 0.3894 0.2840 0.635615 Comm. Down 0.3303 0.5282 0.6644 0.9213 0.0001 0.7373 0.4866 -0.1729 0.6670 0.4441 0.2669 0.718216 Overall Comm. 0.3832 0.6262 0.6769 0.9334 0.0753 0.6828 0.5289 0.0292 0.6307 0.4685 0.3110 0.714917 Products 0.1920 0.1114 -0.0346 -0.2059 0.4648 -0.3602 -0.0917 0.6394 -0.0809 0.0653 0.0553 -0.1636

Key: Performance and Reward is based on Recog., Perf. Eval., and Freq. Feedback; Climate is based on Express, Coop., Sup. AccComm. Peer, Comm. Up, Comm. Down, and Overall Comm.; Skill is based on Products and employee tenure and education. See t


16/24


ing the employee's assessment of how wellhe or she understands the bank's products,and information on the average educationlevel and average branch tenure of theemployees in the branch in the particularyear. Table 1 shows that the average tenureof a branch employee was about 4-1/2 yearsand the average educational attainmentwas about 13 years. The product knowl-edge response and the tenure and educa-tion variables were each transformed to a zscore, and the mean value of the three z-scores was calculated to create the index.Table 2 shows that for the most part,employee opinions of the human resourceactivities of their managers improved be-tween 1995 and 1997. This is likely due tothe fact that during this time period seniormanagement at the bank put much greateremphasis on the importance of human re-source management and tried to addressthe serious deficiencies in some branchesthat had been revealed by the employeeopinion surveys. Including year dummiesin equation (1) enables me to control forthis change. Table 3 shows the pairwisecorrelations of the thirteen individual itemsfrom the employee opinion survey(each con-verted to a 0-100 scale) and the four con-structed measures of the HRM environment.

ResultsOLS Estimates

Equation (1) was first estimated exclud-ing the branch and manager fixed effects.The results are shown in Tables 4 and 5.24These OLS specifications were obtainedwith the HRM variables first excluded(Table 4) and then included (Table 5).25The non-HRM variables. Table 4 showsthat together these variables explain 23%of the variance across branches in theirsales of deposit products and 43% of thecross-branch variation in loan sales. Both

market size and the rate of household turn-over have statistically significant positiverelationships with sales of deposits andloans. Another market predictor of thesale of loans is whether the branch operatesin a market where the residents are identi-fied as affluent or upscale.Size of the branch is statistically signifi-cant only when measured by the number ofpart-time employees in the loans equation.An alternative specification of equation (1)includes the lagged value of deposits orloans in order to better control for the sizeof the branch. When the lagged value isincluded, the coefficients on the numbersof full-time and part-time employees areboth positive and statistically significant inboth equations, and the full-time elastici-ties are larger than the part-time elastici-ties.26The HRM variables. Table 2, which re-ports the means and standard deviations ofthe employee attitude variables, shows thatthere is variation across branches in theemployee responses to the survey ques-tions.27 Responses pertaining to percep-tions of specific human resource practiceshave more variability than responses toquestion 1, which solicits an overall ratingof the bank. Question 1 is similar to ques-tions of the type that are typically used instudies of employee attitudes.28It was argued in the previous section thatthe responses to the employee attitude sur-

24In order to be included in the sample, the branchhad to have two years of employee survey responses.25Thecoefficients on the non-HRM variables shownin Table 2 were largely unaffected by the inclusion ofthe HRM variables.

26The number of employees and lagged balancesare positively correlated. Since the lagged value hasa negative coefficient, the coefficients on number ofemployees were reduced when the lagged value wasexcluded.27It is possible that part of the variation acrossbranches is due to branch-level variation in employeecharacteristics. Since I include education and tenureof the branch employees in the regressions (throughthe index Skill), I argue that the variation in theresponses to the employee attitude questions is cap-turing the branch effect, unless there is substantial

cross-branch variation in unmeasured employee char-acteristics.28Fora recent review of the literature on employeeattitudes and performance, see Judge et al. (2001).With the exception of Ostroff (1992), which usedorganizational-level data, virtually the entire litera-ture used individual-level data.


17/24


Table4. Determinants of Branch-LevelProductivity(OLS).aGrowthRate of Deposits GrowthRate of Loans

Independent Variables b t b tMarket CharacteristicsLn (Avg. Dwelling Value) 0.026 (0.56) 0.003 (0.06)Ln (Total Population) 0.017 (1.70)* 0.031 (3.37)***% Post-High School Education -0.002 (-1.46) -0.002 (-1.73)*Household Turnover 0.011 (2.60)*** 0.017 (4.15)***Affluent 0.042 (0.97) .100 (2.47)***Empty Nesters -0.045 (-1.80)* -0.016 (-0.67)Ethnic -0.043 (-1.27) -0.006 (-0.18)Low -0.015 (-0.45) 0.051 (1.59)Middle -.022 (-0.78) 0.017 (0.67)Upscale 0.015 (0.61) 0.046 (1.99)**Working -0.014 (-0.50) 0.023 (0.87)Branch CharacteristicsLocated in Mall -0.017 (-0.78) -0.017 (-0.85)Age of Branch -0.020 (-2.56)*** -0.001 (-0.12)Manager Tenure -0.0001 (-0.56) 0.0001 (0.47)Ln (No. Full-Time Employees) -0.01 (-0.49) -0.028 (-1.40)Ln (No. Part-Time Employees) 0.012 (0.80) .023 (1.68)*Ln (Property and Premises) 0.009 (0.63) -0.026 (-2.02)**Year Dummies1995 .034 (2.63)*** .132 (10.84)***1996 .017 (1.01) .055 (3.47)***N 330 330R2 0.225 0.428

aRegressions include city dummy variables. Each observation is a branch-year, where year is either 1995, 1996,or 1997.*Significant at the .10 level; **at the .05 level; ***at the .01 level.

vey can be used as proxies for the dimen-sions of a high-performance work system.This is borne out by a comparison of thevalues of the HRM indices for Branch #1and Branch #2. Recall from the discussionof the branch visits that, in contrast to themanager at Branch #1, the manager atBranch #2 was observed to undertake ac-tivities that were consistent with a high-performance work system. The 1995 re-sponses to the employee attitude surveysfor these two branches support the claimthat the HRM environments at thesebranches differ in a way that is consistentwith my observations.29

For example, at Branch #2, the calcu-lated value for the index Communication s76, while at Branch #1 it is 57, and across allbranches the mean for the index is 66.Employees at Branch #2 rate communica-tions at their branch as being significantlybetter than the communications ratingsgiven by employees at Branch #1 (p < .05).Similarly, at Branch #2, the calculated valuefor the index Performance nd Rewardis 71,while at Branch #1 it is 53, and across allbranches the mean for the index is 67. Onthis dimension, employees at Branch #2rate their branch as being significantly bet-ter than the ratings given by Branch #1employees (p < .01). For the index Climate,the value for Branch #2 is 68, the value forBranch #1 is 45, and the overall mean is 59;the Branch #2 rating is significantly higherthan the Branch #1 rating (p < .01). Finally,the Skillindex, which by construction has a

29Since I visited the branches at the end of 1995and the beginning of 1996, the responses to the 1995employee attitude survey would most closely matchthe environment I observed.


18/24


Table5. Effects of HR Indices on Branch Performance (OLS).aGrowthRate of Deposits GrowthRate of Loans

Index b t R2 b t R2Each HR Index Entered Separately1. Overall 0.000 (0.01) .225 0.020 (2.55)*** .4412. Skill -0.062 (-0.50) .226 0.070 (0.61) .4293. Performance and Reward 0.013 (1.55) .231 0.028 (3.45)*** .4504. Climate 0.002 (0.24) .225 0.005 (0.78) .4305. Communication 0.003 (0.41) .226 0.023 (3.13)*** .447HR Indices Entered Simultaneously .237 .4666. Overall -0.007 (-0.63) 0.016 (1.61)7. Skill -0.093 (-0.72) 0.014 (0.13)8. Performance and Reward 0.023 (1.87)* 0.021 (1.83)*

9. Climate -0.002 (-0.20) -0.025 (-2.60)**10. Communication -0.007 (-0.63) 0.021 (1.96)**aThese are coefficients and t-values from complete regressions as specified in Table 4.*Significantat the .10 level; **at the .05 level; ***at the .01 level.

zero mean across all branches, has a valueof.07 for Branch #2 and a value of -0.42 forBranch #1, a statistically significant differ-ence (p < .05). Hence, all four indices showa significantly higher branch rating byBranch #2 employees than by Branch #1employees.The top panel of Table 5 shows that theOverall rating of the bank as well as thePerformance nd Rewardand the Communica-tion indices have positive and statisticallysignificant effects on the branch's loangrowth rate. In the bottom panel of Table5, all four indices and the Overallrating areincluded in the regression together. Thefact that the Performanceand Reward andCommunication ndices are statistically sig-nificant even when we control for the Over-allrating indicates that the employees' per-ceptions of specific HRM activities containimportant information quite distinct fromtheir overall assessment of the bank.30 Notefurther that, as expected, the HRM vari-

ables have weaker effects in the depositsequation, with only Performance nd Rewardbeing statistically significant. The magni-tudes of the statistically significant HRMindices are rather large. A one standarddeviation improvement in either the Perfor-mance and Reward ndex or the Communica-tion index corresponds to a 2 point increasein the loan growth rate, or 18% of theaverage annual loan growth rate of 11.4percentage points. The statistical insignifi-cance of the Skill index is consistent withBatt's (1999) finding that skill level did notaffect the sales productivity of customerservice workers in a telecommunicationscompany.31Estimates with Branchand Manager Fixed Effects

The correlations in Table 3 show that theOverall rating and the Performanceand Re-ward, Climate, and Communication indicesare highly correlated, with correlationsranging from .4 to .7. Positive correlations

SoWhen the indices are used together, the coeffi-cient on Climateactually has the wrong sign. Somepossible explanations for this finding are discussed inconnection with the specification that includes thebranch and manager fixed effects."lThe individual items in the Skill index werestatistically insignificant when entered as separatevariables.


19/24


Table6. Effects of HR Indices on BranchPerformance.a(Includes Branch and ManagerFixed Effects)GrowthRate of Deposits GrowthRate of Loans

Index b t R2 b t R2Each HR Index Entered Separately1. Overall -0.002 (-0.15) .220 0.022 (1.91)* .4892. Skill -0.282 (-1.33) .228 0.189 (0.97) .4813. Performance and Reward 0.014 (1.19) .226 0.033 (3.21)*** .5104. Climate -0.006 (-0.57) .221 -0.011 (-1.14) .4825. Communication 0.004 (0.34) .220 0.017 (1.72)* .487HR Indices Entered Simultaneously .245 .5606. Overall -0.007 (-0.45) 0.016 (1.23)7. Skill -0.261 (-1.20) 0.243 (1.31)8. Performance and Reward 0.028 (1.67)* 0.043 (3.00)***9. Climate -0.013 (-0.94) -0.047 (-4.03)***10. Communication -0.002 (-0.16) 0.014 (1.14)

aThese are coefficients and t-values from complete regressions as specified in Table 4, with the exclusion ofmarket characteristics, location in mall, age of branch, and In(property and premises), all of which do not varyover time for a given branch.*Significant at the .10 level; **at the .05 level; ***at the .01 level.

among the HRM characteristics could bedue either to some other factor that simul-taneously results in employees evaluatingall of the dimensions highly, or, to theextent that the attitudes measure HRMprac-tices, to complementarities in the use ofvarious HRM practices, as suggested byIchniowski, Shaw, and Prennushi (1997).It is possible, therefore, that the resultsin Table 5 simply reflect the effect of someomitted branch-specific or manager-spe-cific factor that leads to high levels of salesand more favorable employee attitudesabout the branch's HRM environment,rather than a true improvement in perfor-mance that is stimulated by better commu-nications and employee recognition andperformance feedback. To test this com-peting explanation, I re-estimated equa-tion (1) including branch dummy variablesand manager dummy variables.32

The results, shown in Table 6, indicatethat, controlling for fixed unobservedbranch and manager characteristics andthe Overall rating, the indices Performanceand Rewardand Communication re still posi-tive and statistically significant when en-tered separately in the loan equation. Thefact that these two HRM dimensions re-main statistically significant even whenbranch dummy variables and managerdummyvariables are included indicates thatthe effects of the HRM variables that wereobserved in the OLS specification were notdue to unobserved branch characteristicsor unobserved fixed characteristics of par-ticular managers (charisma, for example).Rather, the evidence presented here isconsistent with the argument that, overtime, managers experiment with differentHRM activities that create different HRMenvironments, and some of these activitiespositively affect branch performance. Thebranch visits provided concrete evidenceof these productive managerial activitiesthat contribute to the creation of a high-performance work system at the branchlevel.An important question, of course, is whyall managers do not engage in these pro-

32Ideleted the value of the branch's property andpremises from this specification because the onlynon-zero year-to-year changes in this variable that didexist were very small in magnitude. Equation (1) wasalso estimated with the branch dummy variables only.The results were virtually the same as those reportedin Table 6.


20/24


ductive activities. While this question isbeyond the scope of this paper, one specu-lative explanation is that some managerslack sufficient training in how to imple-ment the formal human resource policiesso as to effectively create a high-perfor-mance work system.The bottom panel of Table 6 shows theresults of including all of the HR indices inthe equations at the same time. Performanceand Rewardis the only index that remainspositive and statistically significant in theloans equation, and it is even statisticallysignificant in the deposits equation. Thisindicates that the "incentives" dimensionof a high-performance work system is themost important predictor of performancein the banking industry.One surprising finding is the negativeand statistically significant effect of the Cli-mate ndex when it is included in the regres-sion with all of the other indices. By itself,Climate was statistically insignificant be-cause it was capturing the positive effects ofthe Performance nd Rewardand Communica-tion indices. It is possible that the wrongsign on Climate in the bottom panel ofTable 6 occurs because the Climate ndex isnot a good measure of the "opportunity toparticipate." Therefore, I estimated fixedeffects regressions that eliminated Climateand used Communication alone to proxy"opportunity to participate." The coeffi-cient on Performanceand Rewardwas robustwith respect to this change; branches thatrank high on the incentives dimension arebetter performers.33Alternatively, if we accept the Climateindex as valid and the specification in thebottom panel of Table 6 as correct, thenthe results imply that performance is en-hanced when employees take a favorableview of the incentives they are provided butan unfavorable view of morale, coopera-tion, employee expression, and supervisoraccessibility in their workplace. Although

such a pattern would contradict the predic-tions of the high-performance work systemframework, it is not inconceivable in a highlycompetitive industry in which employeesare encouraged to use their time to sellproducts.While the inherent problem in OLS thatan omitted variable may be correlated withboth performance and the HRM environ-ment is not relevant to the fixed effectsestimates shown in Table 6, the latter esti-mates are not without their own potentialproblems. One is that they may be inconsis-tent if a branch's improvement in its HRMenvironment is correlated with its perfor-mance in a period prior to the change inHRM. This could happen if, for example,senior management decides to replace apoorly performing branch manager or urgesthe existing branch manager to improvehis or her human resource managementenvironment.There is indeed evidence of some suchinfluence at the bank under study. In aregression of the change in surveyresponsescontrolling for the branch's performancein the initial year as well as the vectors ofmarket and branch characteristics, I findthat branches with low sales of loans in theinitial year showed subsequent improve-ments in employees' assessment of commu-nications from superiors and overall com-munications at the branch. Note that thisfinding is exactly the opposite of Beckerand Gerhart's (1996) description of thereverse causation problem in the literatureon human resource management and orga-nizational performance. Those authorsworried that a positive correlation betweena human resource management practiceand organizational performance could re-sult from better performance making theorganization more likely to implement theHR practice.In any event, my finding that there is arelationship between initial performanceand subsequent changes in the HRM envi-ronment indicates that prior performanceof the branch could be an important vari-able omitted from the specification in equa-tion (1). I therefore re-estimated equation(1) with the lagged growth rate of deposits

33Instead of deleting Climateentirely, I also triedeliminating some branches that appeared to be outli-ers on the Climatedimension. The results were unaf-fected.


21/24


Table7. Effectsof HR Indices on BranchPerformance.a(Includes Branch and ManagerFixed Effects and LaggedPerformance)GrowthRate of Deposits GrowthRate of Loans

Index b t R2 b t R2Each HR Index Entered Separately1. Overall -0.005 (-0.46) .346 0.020 (1.81)* .4992. Skill -0.326 (-1.68)* .357 0.127 (0.64) .4893. Performance/ Reward 0.005 (0.47) .347 0.033 (3.16)*** .5194. Climate -0.010 (-1.02) .350 -0.011 (-1.16) .4925. Communication -0.003 (-0.35) .346 0.016 (1.66)* .497HR Indices Entered Simultaneously .367 .5656. Overall -0.004 (-0.32) 0.015 (1.17)7. Skill -0.300 (-1.52) 0.019 (1.02)8. Performance/Reward 0.019 (1.29) 0.043 (3.01)***9. Climate -0.011 (-0.85) -0.047 (-3.92)***10. Communication -0.007 (-0.48) 0.013 (1.07)

aThese are coefficients and t-values from complete regressions as specified in Table 4, with the exclusions ofmarket characteristics, location in mall, age of branch, and In(property and premises), all of which do not varyover time for a given branch.*Significant at the .10 level; **at the .05 level; ***at the .01 level.

or loans added to the equation. The re-sults, shown in Table 7,34are virtually iden-tical to those shown in Table 6, indicatingthat the omission of prior performance wasnot biasing the results.A further consideration is that the fixedeffects specification assumes that the unob-servable attributes of the employees andthe market environment, as well as anyunmeasured employee attitudes, are fixedover the 1995-97 time period. It is possiblethat the unobservable attributes of employ-ees changed over time within branches andthat these changes are correlated with thechanges in the responses to the employeesurvey. But given the fact that the observedemployee characteristics (incorporated inthe Skill index) were statistically insignifi-cant, it is unlikely that time-varying unob-servable employee characteristics are bias-ing the results.

The fixed effects specification also ig-nores time-varying changes in the branch'smarket environment (for example, the in-flux or outflux of residents or businesses).If important changes did take place overthe sample period, this could bias the re-sults if these changes are correlated withchanges in the responses to the employeesurvey. Unfortunately, the data do notenable me to rule out the possible role oftime-varying market factors.Finally, there may be unmeasured attitu-dinal variables that co-vary with attitudestoward communications, performanceevaluation, and recognition. The evidencefrom the branch visits, however, would seemto indicate that communications, perfor-mance evaluation, feedback, and recogni-tion are the critical factors that contributeto performance.

ConclusionsEmpirical research on the relationshipbetween human resource management andestablishment performance has focused onblue-collar workers in manufacturing. Thispaper extends the analysis to the servicesector-where, in fact, most employees

34For iscal year 1995, the lagged value (1994) onlyrefers to the last six months of fiscal 1994. The bankwas unwilling to release data for an earlier timeperiod.


22/24


work-by examining the retail branch op-erations of a large Canadian bank. A uniquelongitudinal data set collected through sitevisits was used to estimate the determinantsof branch-level performance and specifi-cally to consider if the dimensions of ahigh-performance work system contributeto performance. Previous studies of branchperformance have largely focused on therole played by scale in determining theefficiency of a bank branch, leaving most ofthe variance in measured efficiency unex-plained.

Interviews with managers and employeeswere used to guide the specification of thebranch-level production function. Follow-ing the lead of Appelbaum et al. (2000), Iused three dimensions of a high-perfor-mance work system to characterize the hu-man resource management environmentat the branch: opportunity to participate,skills, and incentives. To measure thesedimensions, I used data from the bank'semployee attitude survey--an improvementover the sole source manager survey re-sponses on which other studies have relied.The econometric analysis showed that,controlling for the characteristics of themarket in which the branch is located, aswell as unobserved fixed branch and man-ager characteristics, employees' perceptionsof the performance feedback and recogni-tion system at their branch-that is, theincentives dimension of a high-performancework system-had a positive and statisti-cally significant relationship with branchperformance, as measured by its sales ofloans, that was robust under alternativespecifications. Some of the fixed effectsresults showed a positive effect of the qual-ity of communications between the man-ager and the staff and among staff mem-bers, a component of the "opportunity toparticipate" dimension of a high-perfor-mance work system. The fact that the HRMvariables remained statistically significanteven when manager dummy variables wereincluded in the regressions indicates thatthe results are not due to unobserved per-sonality characteristics of particular man-agers. The results were also unaffected by

the inclusion of a variable measuring thebranch's performance prior to any changein its human resource management envi-ronment.How confident should we be in inter-preting these results as evidence that ahigh-performance work system can influ-ence performance in the banking sector?Ideally, in order to answer this question, wewould want to have an experimental designin which human resource managementpractices are randomly assigned across thebranches. In this way, the treatment andcontrol groups would not differ in terms ofother organizational characteristics thataffect performance. As Ichniowski et al.(1996) observed in their review of the lit-erature on the effects of management prac-tices on organizational performance, ex-perimental designs in this arena are typi-cally infeasible. The alternative approach,which is used here, is to control for vari-ables that affect performance and that arelikely to be correlated with a high-perfor-mance work system. Specifically, controlsfor worker quality (education and ten-ure), manager quality (tenure), marketcharacteristics, prior performance of thebranch, and fixed unobserved branch andmanager characteristics were used here.One remaining caveat is that unobservedtime-varying market attributes could playa role.The results from the econometric analy-sis should not be viewed in isolation. Animportant component of this research isthe branch visits, which provided concreteevidence of specific actions taken by man-agers that created real differences in thehuman resource management environ-ments at the branches that, in turn, re-sulted in variation in performance acrossbranches. Observations made during thebranch visits lend credence to the dataobtained from the employee surveys. Thecombination of the branch visits and theeconometric results supports the notionthat branch-level performance in the bank-ing industry can be influenced by specifichuman resource management-related ac-tions.


23/24


REFERENCESAppelbaum, Eileen, Thomas Bailey, Peter Berg, andArne L. Kalleberg. 2000. Manufacturing Advantage:WhyHigh-PerformanceWorkSystemsPay Off. Ithaca,N.Y.: ILR Press (an imprint of Cornell UniversityPress).Autor, David, Frank Levy, and Richard Murnane.2001. "Upstairs, Downstairs: Computers and Skillson Two Floors of a Large Bank" Industrial and LaborRelations Review,Vol. 53, No. 3 (April), pp. 432-47.Banker, Rajiv D., Seok-Young Lee, Gordon Potter,and Dhinu Srinivasan. 1996. "Contextual Analysisof Performance Effects of Outcome-Based Incen-tive Compensation." AcademyofManagementJournal,Vol. 39, No. 4 (August), pp. 920-48.Batt, Rosemary. 1999. "WorkOrganization, Technol-ogy, and Performance in Customer Service andSales." Industrial and LaborRelationsReview,Vol. 52,No. 4 (July), pp. 539-64.. 2002. "Managing Customer Services: HumanResource Practices, Quit Rates, and Sales Growth."AcademyofManagementJournal,Vol. 45, No. 3 (June),pp. 587-97.Becker, Brian, and Barry Gerhart. 1996. "The Impactof Human Resource Management on OrganizationalPerformance: Progress and Prospects." AcademyofManagement Journal, Vol. 39, No. 4 (August), pp.779-801.Berger, Allen N.,John H. Leusner, andJohnJ. Mingo.1994. "The Efficiency of Bank Branches." Board ofGovernors of the Federal Reserve System WorkingPaper, August.Berger, Allen N., and LorettaJ. Mester. 1997. "Insidethe Black Box: What Explains Differences in theEfficiencies of Financial Institutions?" WhartonFinancial Institutions Center, Working Paper 97-04,January.Black, Sandra E., and Lisa M. Lynch. 2001. "How toCompete: The Impact of Workplace Practices andInformation Technology on Productivity." ReviewofEconomicsand Statistics,Vol. 83, No. 3 (August), pp.434-45.Boning, Brent, Casey Ichniowski, and Kathryn Shaw.2001. "Opportunity Counts: Teams and the Effec-tiveness of Production Incentives." NBER WorkingPaper No. 8306, May.Bowen, David E., and Benjamin Schneider. 1988."Services Marketing and Management: Implica-tions for Organizational Behavior." In Barry Stawand Lawrence Cummings, eds., Research n Organiza-tional Behavior, Vol. 10. Greenwich, Conn.: JAIPress, pp. 43-80.Bresnahan, Timothy F., Erik Brynjolfsson, and LorinM. Hitt. 2002. "Information Technology, Work-place Organization, and the Demand for Skilled

Labor: Firm-Level Evidence." Quarterly ournal ofEconomics,Vol. 117, No. 1 (February), pp. 339-76.Calomiris, Charles W., and Jason Karceski. 1998. Isthe Bank MergerWaveof the 1990s Efficient? Washing-ton, D.C.: AEI Press.Canadian Bankers Association. 1994. "Bank Facts1994." Mimeo, Toronto.

Cappelli, Peter, and David Neumark. 2001. "Do'High-Performance' Work Practices Improve Estab-lishment-Level Outcomes?" Industrial and LaborRelations Review,Vol. 54, No. 4 (July), pp. 737-75.Colwell, R. J., and E. P. Davis. 1992. "Output andProductivity in Banking." Scandinavian Journal ofEconomics,Vol. 94, Supplement, pp. 111-29.Delery, John E. 1998. "Issues of Fit in StrategicHuman Resource Management: Implications forResearch." Human ResourceManagementReview,Vol.8, No. 3, pp. 289-309.Delery, John E., and D. Harold Doty. 1996. "Modesof Theorizing in Strategic Human Resource Man-agement: Tests of Universalistic, Contingency, andConfigurational Performance Predictions." Acad-emy of ManagementJournal, Vol. 39, No. 4 (Decem-ber), pp. 802-35.Doukas, John, and Lorne N. Switzer. 1991. "Econo-mies of Scale and Scope in Canadian Branch Bank-ing." Journal of International Financial Markets,Insti-tutions and Money,Vol. 1, No. 2, pp. 61-84.Dunlop, John T., and David Weil. 1996. "Diffusionand Performance of Modular Production in theU.S. Apparel Industry." Industrial Relations,Vol. 35,No. 3 (July), pp. 334-55.Frei, Frances Z., Patrick T. Harker, and Larry W.Hunter. 2000. "Inside the Black Box: What Makesa Bank Efficient?" In Patrick T. Harker and StaviosA. Zenios, eds., ThePerformanceof Financial Institu-tions. Cambridge: Cambridge University Press.Helper, Susan, David Levine, and Elliot Bendoly.1999. "Employee Involvement and Pay at U.S. andCanadian Auto Suppliers." Institute of IndustrialRelations Working Paper.Heskett,James L., W. Earl Sasser,Jr., and Leonard A.Schlesinger. 1997. The ServiceProfit Chain. NewYork: Free Press.

Hollenbeck,John R., Daniel R.Ilgen,JeffreyA. LePine,Jason A. Colquitt, and Jennifer Hedlund. 1998."Extending the Multilevel Theory of Team DecisionMaking: Effects of Feedback and Experience inHierarchical Teams." Academy of ManagementJour-nal, Vol. 41, No. 3 (June), pp. 269-82.Hunter, Larry, Annette Bernhardt, Katherine L.Hughes, and Eva Skuratowicz. 2001. "It's NotJustthe ATMs: Technology, Firm Strategies, Jobs, andEarnings in Retail Banking." Industrial and LaborRelations Review, Vol. 53, No. 2A (special issue,March), pp. 402-24.Hunter, Larry W., and Lorin M. Hitt. 2001. "WhatMakes a High-Performance Workplace? Evidencefrom Retail Bank Branches." Mimeo, WhartonSchool, May.Huselid, Mark. 1995. "The Impact of Human Re-source Management Practices on Turnover, Pro-ductivity, and Corporate Financial Performance."AcademyofManagementJournal,Vol. 38, No. 3 (June),pp. 635-72.Ichniowski, Casey, Thomas Kochan, David Levine,Craig Olson, and George Strauss. 1996. "WhatWorks at Work: Overview and Assessment." Indus-


24/24


trial Relations, Vol. 35, No. 3 (July), pp. 299-333.Ichniowski, Casey, Kathryn Shaw, and GiovannaPrennushi. 1997. "The Effects of Human ResourceManagement Practices on Productivity: A Study ofSteel Finishing Lines." American Economic Review,Vol. 87, No. 3 (June), pp. 291-313.Judge, Timothy,Joyce E. Bono, CarlJ. Thoresen, andGregory K. Patton. 2001. "TheJob Satisfaction-JobPerformance Relationship: A Qualitative and Quan-titative Review." PsychologicalBulletin, Vol. 127, No.3, pp. 376-407.Kelley, Maryellen R. 1996. "Participative Bureau-cracy and Productivity in the Machined ProductsSector." Industrial Relations, Vol. 35, No. 3 (July),pp. 374-99.Keltner, Brent. 1995. "Relationship Banking andCompetitive Advantage: Evidence from the U.S.and Germany." California Management Review,Vol.37, No. 4 (Summer), pp. 45-73.Kleiner, Morris M., Jonathan S. Leonard, and AdamM. Pilarski. 2002. "How Industrial Relations AffectsPlant Performance: The Case of Commercial Air-craft Manufacturing." Industrial and LaborRelationsReview,Vol. 54, No. 2 (January), pp. 195-218.MacDuffie, John Paul. 1995. "Human ResourceBundles and Manufacturing Performance: Organi-zational Logic and Flexible Production Systems inthe World Auto Industry." Industrial and LaborRelations Review,Vol. 48, No. 2 (January), pp. 197-221.

Murphy, N. B., and Y. E. Orgler. 1982. "CostAnalysisfor Branching Systems: Methodology, Test Results,and Implications for Management." Journal of Fi-nancial Research,Vol. 5, No. 2-3, pp. 181-88.

Ostroff, Cheri. 1992. "The Relationship betweenSatisfaction, Attitudes, and Performance: An Orga-nizational Level Analysis." Journal ofAppliedPsychol-ogy,Vol. 77, No. 6 (December), pp. 963-74.Parkan, Celik. 1987. "Measuring the Efficiency ofService Operations: An Application to BankBranches." Engineering Costs and Production Econom-ics, Vol. 12, No. 1-4 (July), pp. 237-42.Parkington, J. J., and Benjamin Schneider. 1979."Some Correlates of Experienced Job Stress: ABoundary Role Study." AcademyofManagementJour-nal, Vol. 22, No. 2 (June), pp. 270-81.Schaffnit, Claire, Dan Rosen, and Joseph C. Paradi.1997. "Best Practice Analysis of Bank Branches: AnApplication of DEA in a Large Canadian Bank."EuropeanJournal of OperationalResearch,Vol. 98, No.2 (April), pp. 269-89.Schneider, Benjamin, and David E. Bowen. 1985."Employee and Customer Perceptions of Service inBanks: Replication and Extension." Journal of Ap-plied Psychology,Vol. 70, No. 3 (June), pp. 423-33.Sherman, H. David, and Franklin Gold. 1985. "BankBranch Operating Efficiency: Evaluation with DataEnvelopment Analysis." Journal of Banking and Fi-nance, Vol. 9, No. 2 (June), pp. 297-311.Youndt, Mark A., Scott A. Snell, James W. Dean, Jr.,and David P. Lepak. 1996. "Human ResourceManagement, Manufacturing Strategy, and FirmPerformance." Academyof ManagementJournal, Vol.39, No. 4 (August), pp. 836-66.Zardkoohi, Asghar, andJames Kolari. 1994. "BranchOffice Economies of Scale and Scope: Evidencefrom Savings Banks in Finland." Journal of Bankingand Finance, Vol. 18, No. 3 (May), pp. 421-32.

Documents

Ann P. Bartel -- HRM and Organizational Performance - Evidence From Retail Banking