ABSTRACT AN ENVIRONMENTAL MODEL FOR PERFORMANCE MEASUREMENT IN MULTI-OUTLET BUSINESSES by William R. Kinney, Jr. Standards for evaluating the performance of outlet managers in companies with numerous outlets are difficult to formulate because of the differences in the profit potentials of the various outlets. Dif- ferences in measured performance (profit or other appropriate measure) among outlets may be explained by differences in the characteristics of the locations and the facilities of the outlets as well as the differ- ences in the performance of the outlet managers. Since the location and facilities of an outlet are not controllable by the outlet manager, the effects of these "non-managerial" factors should be extracted before evaluating the performance of the manager. For a small number of outlets, central marketing officials can be familiar with the potentials of the outlets and can subjectively adjust for the potential differences. H0wever, for a large number of outlets, valid subjective adjustments for potential differences are not feasible. The environmental model is an objective method for measuring and evaluating the performance of outlet managers. In applying the method, a linear statistical model of outlet performance is constructed relating responsibility accounting measures of outlet performance to the levels of the non-managerial factors under which the outlets are operated. The effects of the significant non-managerial factors are removed to provide a measure of performance which is more relevant to the actions of the William R. Kinney, Jr. outlet manager. The ranking of the estimated managerial contributions forms a valid intraperiod standard. Through the separation of the effects of the managerial and the location and facility factors the relative importance of the three fac- tors can be determined. If managerial differences are important in determining outlet contribution, much effort and expense should be de- voted to manager selection and training. If the contribution of the outlets depends primarily on the particular location and facilities of the outlet, then relatively more resources should be devoted to location and facility selection with less emphasis on the selection of managers. In a broader sense, the environmental model can be used to maximize the profit of the multi-outlet business by examining relationships among the three broad factors explaining differences in performance among out- lets. The specific relationships are of interest in planning outlet 10- cations and facilities. While it is not feasible to rely upon only the model in making location and facility decisions, the model can be used to narrow the scope of detailed subjective investigations. The explor- atory use of the model for investigating numerous possible locations and various combinations of facilities may point up attractive oppor- tunities which would be overlooked by subjective research. The environmental model has been formulated and tested in a national firm offering a wide range of goods and services by catalogs and retail stores. The study was limited to one region of the catalog sales division of the company. Data were collected on factors such as: papulation characteristics (income, ages, socio-economic characteristics), competition, age and condition of outlet facilities, district in which William R. Kinney, Jr. located, and numbers of catalogs issued. Quadratic effects and first- order interactions were also tested. In the test company there were non-managerial factors which exe. plained statistically significant variation in the transforms of the net sales and the controllable expenses of the outlets. The location and facility differences of the outlets were of much greater importance in determining outlet performance than were the differences among the outlet managers. Thus the contribution of the outlet waslg poor indi- ggtgr‘gf the performance of the outlet managgg. If managers are evalu- ated on the basis of outlet performance, the potential of the outlet location and facilities is likely to be attributed (wrongly) to the outlet managers and significant errors in managerial performance evalu- ation will result. AN ENVIRONMENTAL MODEL FOR PERFORMANCE MEASUREMENT IN MULTI-OUTLET BUSINESSES By \\ \ . William Ri Kinney, Jr. A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Accounting and Financial Administration 1968 6 f; 307:1? ,i//§/(99 Copyright by WILLIAM R. KINNEY, JR. 1969 a ACKNOWLEDGMENTS I wish to express my sincere thanks to Dr. Gardner M. Jones for his encouragement and guidance during this study. Appreciation is also ex- pressed to Dr. James F. Hannan and Dr. Paul E. Smith who contributed valuable assistance. Special appreciation is expressed to the Ernst & Ernst Foundation, the American Accounting Association, Dr. Wilton T. Anderson and the H. B. Earhart Foundation, and to Dr. James Don Edwards and the Department of Accounting and Financial Administration at Michigan State university for financial support during my doctoral years. The cooperation of Montgomery Ward & Company through Vice President Of Operations, Harold F. Dysart, and Manager of Analysis and Special Studies, Kenneth Stow, is gratefully acknowledged. To the many teachers and fellow students, to my parents, to my children, Kristi, Jeff and Robby, who have all made my life richer, I eXpress sincere gratitude. And above all, I wish to express my deepest appreciation to my wife, Carolyn, whose sacrifice, devotion, faith and inspiration has made this period of our lives together one of happiness and fond memorieS. ii ACKNOWLEDGMENTS. . . . LIST OF TABLES . . . . . LIST OF FIGURES. . . . LIST OF APPENDICES . Chapter TABLE OF CONTENTS Page . . . . . . . ii . . . . . . . . . . . V . . . . . . . . vi . . . . . , . vii I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . 1 Purpose and Organization of the Study. . . . . . . . 1 Measurement and Performance Standards. . . . . . . . 2 The Problem. . . . . . . . . . . . . . . . . . . . . 7 II. THE MODEL . . . . . O O O O O O O O O O O O O O O 13 The Non-managerial Factors . . . . . . . . . . . . . . 14 The Statistical Model. . . . . . . . . . . . . . . . . 20 The Testing of Hypotheses. . . . . . . . . . . . . . 27 Revisions to the Accounting System . . . . . . . . . . 30 Implications . III. TESTING THE MODEL . O O O O O O O O O O O O O O O O O O O O 35 O O O O 0 O O O O O O O O O O O O O O O 38 The Test Company . . . . . . . . . . . . . . . . . . . 38 The Hypothesis . . . . . . . . . . . . . . . . . . . . 41 Formulating the Model for Testing. . . . . . . . . . . 41 The Model. . . O O O O O O O O O O O O O O O O O O O O 47 A Statistical Note . . . . . . . . . . . . . . . . . . 55 Testing the Primary Hypotheses . . . . . . . . . . . . 56 The "Carryover" Effect . . . . . . . . . . . . . . . . 57 Summary. . . . O O O O O O O O O O O O O O O O O O O 0 60 IV. IMPLICATIONS OF THE MODEL FOR LOCATION AND FACILITY SELECTION . . . . . O O O O O O O O O I O I O O O O O O O Q 62 The Profitability Measures . . . . . . . . . . . . . . 64 Determining Optimum Location and Facility combinations 0 O O I O O O O O O O O O O O O O O O O 65 iii TABLE OF CONTENTS (Continued) Chapter Page The Geometric Representation . . . . . . . . . . . . 68 Other Location and Facility Problems . . . . . . . . . 75 sumary O O O C O O I C O O O O O O O O I I O O O O O O 78 V. SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . . . . . 80 Summary of the Environmental Model . . . . . . . . . . 80 Potential Advantages and Disadvantages of the Model. . 83 Limitations of the Study . . . . . . . . . . . . . . . 85 Areas for Additional Research. . . . . . . . . . . . . 86 Conclusion . . . . . . . . . . . . . . . . . . . . . . 87 SELECTED BIBLIOGRAPHY. . . . . . . . . . . . . . . . . . . . . . . 89 APPENDICES . 0 0 Q 0 O O O O O O O O O O O O O O O O O O O O O O O 91 iv LIST OF TABLES Table Page 3.1 Analysis of Variance: Logarithm of Net Outlet Sales . . . . 53 3.2 Analysis of Variance: Square Root of Controllable Expenses. 54 Figure 1.1 2.1 2.2 3.1 3.2 4.1 4.2 4.3 4.4 LIST OF FIGURES Performance Standards. . . . . . . . . . Hypothetical Performance Report. . . . . . . Hypothetical Performance Reports for Selected Outlets for January, 1967. . . . . . . . . . . . . . . Primary Non-managerial Factors Considered in Formulation of the Model . . . . . . . . . Significant Non-managerial Factors . . . A Two-factor Surface . . . . . . . . . . . . Determining Optimum Conditions . . . . . . . Other Two-factor Surfaces. . . . . . . . . . A Three-factor Surface . . . . . . . . vi the Initial Page 31 33 43 49 69 72 74 76 LIST OF APPENDICES Appendix Page A An Illustrative Computation of Estimated Managerial contribution. O O O O O O C O O O O O O O O I O O O O O 9]- B Parameter Estimates - 1964, 1965, 1966. . . . . . . . . . 94 vii CHAPTER I INTRODUCTION Purpose and Organization.2£ the Study This dissertation is oriented to the special problems of measuring and evaluating the performance of outlet managers in the multi-outlet business. A theoretical solution to these problems will be presented and then tested in an actual business environment. The solution to be considered is an extension of the responsibility accounting concept to make intraperiod performance standards valid and operational. Specifically, the purposes of this study are to (1) develop a sta- tistical model which recognizes the effect of environmental and other non-managerial factors, as well as managerial contribution, in measuring the performance of outlets in multi-outlet businesses; (2) implement and test such a model; and (3) evaluate the implications of the model. This analysis is needed to provide a better understanding of the profitability relationships in multi-outlet businesses. The model will provide a basis of comparison for evaluating managerial performance and guidelines for resource allocation. Chapter One is a discussion of the concepts of measurement and per- formance standards in accounting. The inapplicability of traditional manager evaluation systems for the special problems of the multi-outlet organization is established. 2 In Chapter Two, the environmental model is presented in conceptual form with a general discussion of the uses, assumptions and limitations of linear statistical models. A plan for implementing the model, a dis- cussion of the non-managerial factors and the revisions of accounting reports are also presented. The construction and testing of the model in a multi-outlet busi- ness is the subject of,Chapter Three. (1) Two types of tests are involved: tests for the presence of the assumptions basic to the analysis of linear statistical models, and (2) tests of statistical significance of the factors related to the measured performance of the outlets in the test firm. The background of the test firm and the factors and relation- ships tested will be presented, as well as the results and significance of the tests. In Chapter Four, the implications of the environmental model for Selecting and evaluating locations and facilities, and the implications for capital budgeting decisions will be presented. The summary, conclusions and limitations of the study will be the subj ect of Chapter Five. Suggestions for additional research will be inc lud ed . Measurement and Performance Standards Accounting information facilitates large scale operations by making Possible a greater span of control of an individual over other individ- uals or groups. Through the use of accounting reports, a division of duties and specialization of activities can occur without loss of essen- tial control and direction. When the decision maker has available the i “formation which is necessary for decision making, he can make the 3 es s ential decisions covering a broader range of activities than he could i. f he gathered the information personally. The decision maker employs an information gathering specialist (in the present case, the accountant) and devotes his own attention to decision making. Each activity of each member in an organization cannot be reported easily and is, individually, of no particular interest to the decision maker. The accountant must summarize the results of members' activi- ties for reporting to the decision maker. Only the important results wi 11 be communicated if there is to be a saving of decision makers' titne. In accounting, activities are first reduced to quantitative terms and then summarized. The accounting problem thus becomes one of de- ciding how to express or represent activities quantitatively. Since an abstraction is being substituted for personal observation, there should be correspondence between the actual event or activity being measured and the impression formed in the mind of the decision maker as he reads the report. The report must contain not only the measurements which the decision maker would make if he could make the measurements personally, but it must also connote those subjective ob- servations and evaluations which the decision maker would make if he Qould personally observe the activity. This ideal is not often attainable--objective measurement and measurement rules may not allow expression of subjective observations. However, the net results of performance are eventually reflected in the accountant's measures. The accounting system may not reflect small de- Dartures from the performance which the decision maker would observe if he supervised personally, but hopefully, this small information loss will be more than offset by the economies of scale of the expanded 4 op erations made possible by the division of duties and specialization. Wid e departures from the performance expected by the decision maker will be detectable by reviewing the accountant's reports and the decision maker can act to correct these exceptional cases with a net gain in ef ficiency. After receiving a measure of performance, the decision maker eval- uates the performance and then exercises control over the activities through rewards or revisions of plans. The measure of performance by itself is of little value--there must be somebasis for comparing the measured performance with a norm or standard of performance. When the decision maker personally supervises all work, the performance standards may be entirely subjective and could even be subconscious. In the larger organization, however, the performance standards must be formal- iZed. There are three basic comparison standards in accounting. These all: e: (l) predetermined standards—-based on ideal or desired performance for the entity, individual or segment, (2) interperiod standards--based on past performance of the entity, individual or segment, and (3) M- .Eeriod standards--based on current performance of other entities, indi- v:lduals or segments. The choice of standard depends on the activities being measured and the reason for the evaluation of the performance of the particular seg- t1lent. All three types of standards have limitations; Figure 1.1 is a diagram of the three standards. It shows the chief difficulty which tRust be overcome or recognized in the use of each of the standards. The predetermined standards are of two basic types: the "absolute" Standard based on physical measurements of activities or quantities and A‘s.- . .uu..‘.. d "L EIéO 5 the "subjective" standard (budget) based on opinions, predictions or d esires of the standard setter. The limitations of the absolute standards are (1) the relatively EBIIJMELII area of possible application and (2) the difficulty in establish- ing the standards even where they are applicable. Setting of valid 8 tandards for any but the most menial tasks becomes difficult, particu- ]_£a;rrly in non-manufacturing work. Inputs in administrative, marketing .Errlci distributive activities can be determined and measured but outputs erlrea often not subject to objective measurement. When the actual output i-ES difficult to measure (or even determine), meaningful output standards €11?£e doubly difficult to formulate. Figure 1.1. Performance standards Actual Current Performance Unit 1 Subjectivity Relevance Uniqueness I Desired Actual Actual Current Performance Past Performance Current Performance Unit 1 Unit 1 Units 2, ..., m (I?redetermined Standard) (Interperiod Standard) (Intraperiod Standard) a l.l.l 55... V .0! I "' I Idfi r-«I liphl. snub : .3" ‘ng3 . ‘43-... -....3 m r) (l‘ (ti 'U (YD r' PM u n” - A" ~t 6 Subjective standards have wider applicability than do absolute s tandards but require the use of expert Opinion. The expert's biases and errors of estimates and predictions are difficult to determine. The obvious limitation of the interperiod standard is that of rele- vance. This limitation is so severe that this type of standard is of little value unless activities of segments cannot be reduced to a common - (ileztuxaminator (such as dollars) or the activities are so diverse that the C>I331qy'basis for comparison is simply an improvement over prior periods. oth er things being equal, improvement of the performance of the current F’eilriod over that of previous periods indicates progress; but even the iI‘D-proved performance may be unsatisfactory when compared with other seg- t“ents or on an absolute performance scale. The intraperiod standard probably has the most potential for a EEITeat many applications. The intraperiod standard is not a measure of absolute efficiency but a comparison of one individual's actual perform- ance with the actual performance of others faced with similar problems. Aha individual may be operating at only one-half of peak efficiency (nmasured on an absolute scale, if applicable) but if he is relatively t:he most efficient, he should receive whatever rewards go to the best P erformer . One intraperiod standard-~the rate of return on investment--has VVide application and is considered by many to be the best single overall Uneasure of performance of economic entities. The rate of return is a tneasure of success in earning a return on assets and success with respect ‘to the nature and amount of capital employed. The overall effectiveness of a management can be judged by a com- ‘parison of the rates of return among various companies. The basis for ,nbrw' t. ... b" :5 F.- 1!!» 9' IO\ tn» .5 . a" :1 ,9 sum .3 .L be In: .F- .Pb .\. q: 7 det: ermining effectiveness is not actual performance compared with a measure of absolute performance but a comparison with how well others have been able to perform. An absolute standard isn't necessary for the investor--if he chooses the relatively most effective management (0 ther things being equal) he will maximize his return. The application of the intraperiod standard will thus direct the investor to an efficient a1 location of resources among companies. Efficient resource allocation is a primary problem within the firm as well as among firms. The internal allocation of resources has become more important with the consolidation and integration of businesses be- ccause many segments of modern businesses are now larger than whole compa- ni es of a few decades ago. The problem for corporate management is how t0 measure performance of segments of the business to determine which Segments are performing in a satisfactory manner and which are not. Unfortunately, many of the market mechanisms which insure efficient allocation of resources among firms do not exist within the firm. In t1loamy internal situations, valid intraperiod standards are not possible because no two departments do the same job or the output of the depart- HIents are too difficult to measure. In other situations, the job assign- ments may be similar but differences in the environments in which the assignment is performed make direct comparisons of results hazardous. The factors which must be considered in the determination of an appropriate standard by which to judge the performance of outlet managers in a multi-outlet business are now considered. The Problem In a production situation, an analysis of variations from budget or 8 physical or time standards may constitute an adequate performance measurement and evaluation system. Tasks are often repetitive and sub- j ec t to rather precise measurement. Attempts to determine responsibility and associate costs with responsibility for control of costs have been generally successful. Responsibility accounting for sub-unit performance has been developed, both theoretically and practically, in an attempt to retain some of the advantages of competition in the large organization. Responsibility accounting for performance is an improvement over full allocation accounting because the performance measure of an individual do es not include arbitrary allocations of costs which are not directly Controllable by the individual. The subjective nature of revenue standards and the lack of any meaningful absolute measure of sales potentials has prevented the de- VQIOpment of a rigorous theory of predetermined standards for the mar- IKeting aspects of business. However, subjective performance goals or budgets have been employed in many businesses with much success. Manag- ers in different divisions operate under different market conditions and félce different potentials and limitations. In setting divisions' goals these differences are taken into consideration by top management. Budgets for a division are admittedly subjective—-there is generally no absolute measure of what sales or profit for a particular division Should or could be for a particular period. But management can be inti- trlately familiar with the potentials and limitations of the limited number 0f divisions and make valid allowances for these differences. Moreover, they may develop budgeting skills which make reliable estimates possible. In summary, each division and each time period is different, but due to 9 the familiarity of the top marketing management, adjustments are made in setting the standards of performance. The same performance evaluation concepts which have been used suc- c essfully at the division level, have been extended to sub-divisions r esponsible for a particular product line or territory within divisions, and even to smaller operating units-~such as local outlets--within sub- d :L visions . The focus on individual outlet managers is entirely proper--"Indi- Viduals operate organizations. Any management accounting system, to be effective, must be designed around the responsibility centers 9}: indi~ w managers ."1 But analysis based upon contribution to unallocable c1<>sts and profits is not an adequate measure for evaluating the perform- ance of outlet managers in a large, multi-outlet organization because of a violation of the "controllability" criterion for both costs and revenue. Managers of outlets generally have no choice as to the outlet to which they are assigned, the size, location or facilities of the outlet, and have little control over the incomes, buying habits or other charac- teristics of the population. Yet, the performance measure (the net Contribution of the outlet) includes the effect of these factors which are not controllable by the outlet manager. Outlet manager performance measures should not reflect how well a particular combination of facili- ties, location and manager have done, but how well the manager has done with the resources at his disposal. The manager with more desirable re- sources at his disposal should naturally have larger reported outlet contributions . Charles T. Horngren, Accounting for Management Control: An Intro- duction, Englewood Cliffs, N. J., Prentice-Hall, Inc., 1965, p. 267. 10 As currently practiced, responsibility accounting for outlet manag- ea:c*53 is only a slightly improved measure of outlet performance. The use C) if outlet contributions as intraperiod standards for evaluating outlet managers' performance is still far from valid because the differences in [pttrcaufit potentials among the outlets are not recognized in the standard. It is relatively easy to compute the contribution of any operating unit to unallocable costs and profits but it is not easy to determine t211.ee extent to which each unit manager has effectively utilized the re- ssc>11rces at his disposal. The inherent environmental factors which are ‘FICDt: controllable by the outlet manager but which influence the perform- Eltlcze of an outlet must be considered in the evaluation of the perform- Eitlcze of the outlet manager. Only when the effects of the non-manages ITiMally controllable factors have been removed is it possible to deter- UniJne the results peculiarly attributable to the manager's action. Performance measures for retail outlets include: gross margin, con- trribution to unallocable costs and profits, outlet "profit," rate of re- tlurn on investment, share of market, and some other less comprehensive Uneasures. Present methods of performance evaluation for retail outlets en- CIompass the contribution statement and comparisons of actual results Vvith those of last year or with a subjective estimate (budget) of what sshould have taken place this year in a particular outlet. Estimates of ‘Vhat should occur under current conditions are generally considered to 13e superior to the results of past, non-comparable periods as a yard- stick for performance. Each outlet and each time period is different, éand adjusuments for those important differences must be made in oper- ‘ating plans if valid standards of performance are to be obtained. 11 Although the need for such adjustments is clear, the implementation of a valid "adjustment system" is difficult. Carefully thought out, detailed budgets for individual outlets for each week or month are often not economically feasible. One individual cannot be familiar with the potentials of several hundred outlets of a company. Also, standards of performance set by district supervisors of outlet managers reflect the personal biases of the supervisors and dif- ferences in budgeting skills among supervisors. Even if the outlet managers participate in the formulation of their budgets, the most per- suasive manager will receive the most favorable budget--and he may or may not be the best performer on the job. Since incentive systems and promotions are often tied to achieving quotas or budgets, this problen becomes quite important. Beyond these practical difficulties, traditionally conceived budgets estimate the wrong quantity from the point of view of outlet manager evaluation. The budgeted contribution is the desired performance of the combination of location, facilities and the manager. Success in meeting or exceeding such budgets is evidence of marginal efforts of the manager and is not indicative of the overall relative usefulness of the manager to the company. Since this "overall usefulness" type of budget is not in use today, it is questionable whether such budgeting is practical. Indeed, a priori, it seems that the setting of such "managerial perform- ance only" budgets would be much more difficult than the outlet perform- ance budget and subject to much greater errors, biases and inconsist- encies than the outlet performance budget. In addition to needing a basis for evaluating outlet managerial personnel, the top management of a multi-outlet business needs - ' l a u p.. 5w *5 in no ~C I. “- - n t n, kl rr» 12 information regarding the effectiveness of outlet locations and facili- ties apart from managerial contribution. The breakdown of profits among these three types of factors is important because excellence in one fac- tor can disguise a lack of excellence in another. The firm will maximize profits if its outlets are located in the ”best" locations with the optimal facility combinations and staffed with the "best" managers avail- able. To achieve this overall excellence, a given outlet can be expanded, contracted, remodeled, relocated, restaffed or eliminated. Thus, the need for objective data on location and facilities effectiveness is much like the need for managerial performance measurement. Top manage- ment also needs information on facilities and locations to aid in plan- ning expansions into new areas and new types of outlets. Past trends and relationships can be extracted much as the investor projects past earnings performance. This dissertation deals with the development of a statistical model approach to help solve many of the measurement and evaluation problems associated with the modern multi-outlet business by explicitly con- sidering the inherent environmental factors in the traditional accounting measurement and reporting system. The multi-outlet business is defined as one in which there is a large number of relatively homogeneous outlets in which the outlet managers' functions and responsibilities are virtually the same. The only differences in outlets are the location and facility differences-- none of which are controllable by the outlet manager. A "large number" of outlets is a number large enough for the limitations on subjective estimates of profit potentials discussed earlier to apply. CHAPTER II THE MODEL The performance measurement and evaluation system for multi-outlet businesses to be considered in this dissertation is an extension of the responsibility accounting concept to make intraperiod performance stand- ards valid and operational. The extended system will utilize responsi- bility accounting data in conjunction with statistical techniques for estimating the effects upon outlet performance of the non-managerially controllable environmental factors under which an outlet is operated and the effect of the actions of the outlet manager after considering the effects of the environmental factors. Statements of outlet contribution to unallocable costs and profit, alone, cannot serve as the basis of intraperiod standards for managerial performance because of differences in the environments among the outlets --differences in profit potentials. For contribution statements to be a valid measure of managerial performance, it must be assumed that the potentials of all outlets are equal; that all managers are subject to the same environment. Under these conditions, managers working at the same levels of efficiency and competence would yield equal contributions at the various locations. Such an assumption is clearly untenable; even to the casual observer, environments and potentials among the outlets are not the same. Outlet contributions must be adjusted for the dif- ferences in environments before managers can be compared. The relative 13 l4 importance of the environmental factors, levels of factors, and level combinations are also of interest in location and facility selection. The proposed "environmental model" is an objective method for esti- ‘mating the portion of the measured performance1 of an outlet attributable to the environmental factors and the portion attributable to the outlet manager. The basic requirement for the environmental model is that the measured performance of an outlet depend upon or be related to the levels of certain environmental factors under which the outlet is operated, in addition to depending upon the outlet manager's actions. Only the en- vironmental factors which are not controllable by the outlet manager will be considered in the model. These factors will be referred to as "non-managerial" factors. The manager should be held responsible for the effects of any factors controllable by him. No two outlets are ever exactly comparable, but through the use of statistical tools, the effects of the non-managerial factors can be extracted. Since the estimated managerial contribution does not depend upon the levels of the non-managerial factors (these effects having been stripped away) the managerial contributions are comparable and one manager's contribution can be compared directly with the contributions of others. The Non-managerial Factors The nature and number of non-managerial factors which are relevant for a particular business depends upon the nature of the business and 1"Measured performance" could be any measure of outlet performance such as sales, controllable expenses, net contribution, or a transfor- mation of one of these variables. 15 the products being offered. For example, the factors related to the performance of variety stores may be much different from the factors re- lated to the performance of restaurants. A classification scheme for a particular non-managerial factor is needed if the average performance of outlets differs among the various levels of the factor under which the outlets are operated.2 In general, as the complexities of the product being offered and the restrictions placed on the outlet managers increase, the complexity of the non- managerial factor structure increases. The only requirements for the inclusion of a factor in the model are: (l) the factor can be measured at each outlet or the outlets can be classified as being in some category of a nominal classification and (2) the classification can "explain" some of the variation in performance among outlets. A few examples of possible non-managerial factors are listed below. Some of the factors are overlapping and much confounding3 could result. This is a general list, however, and is not intended to be a list of factors for a specific firm. (1) Physical Facilities of the Outlet: Investment in the outlet Age of the outlet Square feet of floor space Feet of window (display) space Basic building model Service facilities Amount of available parking 2The familiar rate of return on investment is actually a "one-way" classification, and thus, the environmental model is just an extension of the same concept by adjusting simultaneously for other non-managerial factors. The effects of two factors are said to be "confounded” if it is impossible to separate the effects by statistical analysis. l6 (2) Location Characteristics of the Outlet: State City Neighborhood location (downtown, residential, suburban, etc.) Store complex (number and nature of other stores in the immediate shopping area) Proximity to traffic arteries Number and nature of competitors in trade area Socio-economic status of customers Total population in trade area Median family income in trade area Median age of population in trade area It is possible that two or more of these factors are related to outlet performance in the same way. Investment, floor space and basic building model may all be measures of the general physical attractive- ness of the outlet. Tests of colinearity can be made to avoid a "double counting" of factors which could logically be considered to be estimates of the same underlying phenomenon. The possible relationships of the facility factors to outlet per- formance are fairly straightforward. Investment, age, and footage are measurable on a cardinal scale while the basic building model could be nominally classified as type A, B, or C, or as the "new” or "old" model. The "new" model outlets might be expected to perform better than the "old" model outlets--the environmental model indicates how much better the "new" outlets should be, based on the average performance of the "new" outlets compared with the average performance of the "old" outlets. Different outlet model types may also perform at different levels. The location factors require more explanation. A nominal classifi- cation as to the state in which the outlet is located may be necessary because the outlets in one state may, on the average, perform at a dif- ferent level than outlets in other states. Differences in physical, political, economic and social climates may be underlying causes of 17 differences in potentials for the outlets among the states. As a simple example, the sales potentials of ice cream parlors located in Michigan may be less than the potentials of those located in Oklahoma because of the higher average temperatures and the longer "ice cream season" in Oklahoma. Furthermore, there may be only a small difference in the average performance of outlets in Michigan and Indiana during most years, but there may be a ”heat wave” in a particular year which is much more severe in Indiana than is experienced in Michigan. To the extent that the differences in the weather have affected the relative sales potentials for the outlets in the states (as measured by dif- ferences in the average sales of the outlets in the states), adjustments should be made. The adjustments in a particular year may be much larger than those necessary in the average year. A better basis for such weather adjustments might be a mean tem- perature classification or perhaps a district classification within states. A district effect could also arise because differences in skills of the district supervisors affect the average performance of outlets within the districts. A state classification is needed when average performance of out- lets differs among the states, whatever the reason. As indicated above, the state effect could perhaps be analyzed into components such as mean temperature which the state classification may only approximate. The question to be answered is: Does the added explanatory value of the more complex model justify the added cost of installing and operating the refined classification system? The approximate nature of some of the following classifications should be noted. 18 A city effect may arise because the average performance of outlets located in city A is different from the average performance of outlets in other cities. The effect may be related to population differences, family income differences or age differences. Or, one city could be primarily an industrial city while another is a commercial city. City A.may have limited consumer shopping while city B serves as a regional shopping city and have potential greater than that indicated by its population, family incomes and other characteristics. The neighborhood location classification may be needed because out- lets located downtown perform, on the average, at a level different from those located uptown or in suburban areas or shopping centers. The store complex (number and nature of other stores in the immediate shopping area) may considerably enhance the attraction of any store within the complex due to the variety of different products and services offered.4 The potential of an outlet in a planned complex may be greater than that of an outlet in an unplanned complex. Furthermore, outlets in regional shopping centers may have different potentials than those in neighbor- hood shopping centers. Available parking and proximity to traffic arteries can yield a difference in the potentials of outlets in other- wise similar shopping centers. The number and nature of competitors in the trade area of an outlet is probably of considerable importance. A manager facing a small number of competitors should do better than one facing a large number. And, even though the number of competitors may be the same at two locations, 4Bernard J. LaLonde, "Differentials in Supermarket Drawing Power and Per Capita Sales by Store Complex and Store Size," unpublished Ph.D. dissertation, Michigan State University, 1961, p. 12. 19 the nature of the competitor can be important--an outlet facing a com- petitor which has an extremely good location, large service staff, attractive facilities or well-established reputation should not be ex- pected to perform as well as another outlet in a location where the competitors have less desirable locations and facilities. The median age in the community or perhaps a breakdown into numbers of people in various age groups may be important for businesses whose outlets sell consumer durables, baby items, rocking chairs or other products for which sales may be related to the ages of the populace. The discussion of possible factors could be extended to cover many specific types of multi-outlet businesses. The above discussion is only an indication of a few, general, possibilities. From the viewpoint of evaluating outlet managers, the particular non-managerial factors ex~ plaining variation in performance among the outlets is not important. The only reason for considering the factors is to eliminate their ef- fects in order to compute a manager's contribution. From the vieWpoint of top management, both managerial contribution and the non-managerial factors are important because top management can, in the long run, control both of these profit-related factors. Within certain constraints, such as available funds, contract commitments and costs of abandonment, top management can expand, contract, remodel, re- locate or restaff any outlet. The use of the environmental model in making such decisions will be discussed in Chapter Four. Rather than continue the discussion of factors in general, the sub- ject will be dropped temporarily and taken up again in Chapter Three in the specific context of the test firm. The next topic is a presentation 20 of the statistical properties, assumptions and limitations of the pro- posed environmental model. The Statistical Model The environmental model is a linear statistical model of outlet performance in which it is assumed that the measured performances of an outlet can be described as the sum of: (l) the effects of the levels of the non-managerial factors under which the outlet is operated, (2) the effect of the outlet manager's actions after considering the effects of the non-managerial factors, and (3) a portion not explained by the model. The environmental model is in contrast to a more general model which is implied by the use of a responsibility accounting system in a multi-outlet business. Under a responsibility accounting system, the performance of an outlet is assumed to be attributable solely to the actions of the outlet manager. In a multi-outlet business the only standard with which to compare the measured performance of an outlet is the intraperiod standard of the current performance of the other outlets. Recall that in Chapter One intraperiod standards were rejected because of irrelevance, and valid predetermined standards were not operational in the multi-outlet case due to the large number of outlets for which potentials could not be adequately determined. In such a situation, the only comparison which can be made is the performance of a particular outlet with the average performance of all outlets. An implicit assump- tion is, then, that the potentials of the outlets are equal and there- fore performance should be the same at all outlets unless there is a 5See footnote 1 on page 14. 21 difference in the performances of the managers. The model implied by the use of such a responsibility accounting system is: Y =A*+M*j. In the model, Yj is the measured performance of outlet j for time period t, A* is the average performance of all outlets and M*j is the effect of the actions of the manager of the outlet. Thus, it can be seen that the manager of an outlet at which the measured performance is above the average will be deemed to be a "better-than-average'l manager and vice versa. The potential of the outlet above (or below) that of the average outlet (i.e., the effect of the non-managerial factors) is incorrectly attributed to the actions of the outlet manager. The environmental model makes use of concomitant information-~the levels of the non-managerial factors which are assumed to be related to outlet performance. By fitting the environmental model, the difference in performance which is assumed to be attributable to the concomitant variables is removed before considering the performance of the outlet manager. The environmental model is a variation of a ”covariance" model, so named because the covariance of the concomitant variables and measured performance is specifically considered. A variation from the usual co- variance model is required because the levels of many non-managerial factors remain constant over a relatively long period of time. 6Other non-managerial factors such as demographic characteristics do change quite often and are measurable but are not measured because the extra expense incurred in obtaining such measurements would likely outweigh their value. 22 To formulate the environmental model, consider a multi-outlet busi- ness in which there are "n" non-managerial factors which explain sig- nificant variation in the measured performance among the "m" outlets of the firm. It is desired to estimate the relative managerial performance among the m outlet managers (after considering the effect of the non- managerial factors) for a particular time period t. Managerial perform- ance is assumed to be constant throughout time period t. Reports of outlet performance are made for p sub-periods within time period t. The sub-period measurements of outlet performance are seasonally adjusted to remove the variation in performance which is due to seasonal influ- ences. The p measurements of outlet performance over time period t are p measurements of the performance of the same combination of non-manage- rial factors and outlet manager. The performance of a typical outlet j for sub-period of time k can be described as: Y. =A+BX.+...+BX.+M.+U.. J J nnJ J JR In this expression, Y is the seasonally adjusted measured performance jk of outlet j for sub-period k. The term A is the overall regression con- n stant or the Y-axis intercept and the E (Bixi ) is the sum of the ef- .- 3 1-1 fects of the levels of the non-managerial factors under which outlet j is operated. The "Bi"s are the regression coefficients which express the relationships between the non-managerial factors and measured outlet performance. Each Xij is the level of the non-managerial factor i at outlet j. M represents the effect of the actions of the manager of outlet j. .1 Mj is defined more explicitly on pages 24 and 25. For the multi-outlet 23 business as a whole, there are m ”Mj"s. The unexplained portion (Ujk) is due to chance or the inherently uncertain world in which the outlet is operated. The unexplained por- tion will be assumed to have an expectation of zero and be randomly distributed about that expectation. Thus, the expectation of measured performance (E(ij)) is equal to A +2 Bixij + Mj' More will be said about the unexplained term in a subsequent section concerning tests of hypotheses. The parameters A, B1’ ..., Bn’ M1, ..., Mm, expressing the relation- ship between the independent variables and the dependent variable are not known and must be estimated. There are several methods for esti- mating these parameters. The only estimation method to be considered in this dissertation is the method 2£_1east squares. The derivation of the method of least squares is available in many statistics texts and will not be repeated here.7 The definition of the managerial contribution (Mj) is the key to the special estimation problem posed by the environmental model. Once the nature of M has been clarified, the estimates of the parameters can 3 be determined by using the usual least squares method. 7For example, see Donald A. 8. Fraser, Statistics: ‘An Introduction, John Wiley & Sons, Inc., New YOrk, New York, 1958, pp. 228-240; or John E. Freund, Mathematical Statistics, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1962, pp. 321-325. 24 In vector notation, the expectation of the measured performance of the m outlets for time period t can be described as: ' ‘ F" 7 '1 ’ “ ‘ ' ’ ‘ P ‘ Yu 1 X11 an z1 Zm Y1k 1 x11 Xn1 2l 2m 1 = l E . . A + . B1 + . . + Bn + . M1 + . . . +.- ME ml xlm Xnm Zl zm Y 1 X X Z Z L.mk_ _._. ..1T_ firm: L.Em l-qL In this expression Z1 = 1 for all ij with j = l, and 0 for all other Y Z = l for all Y with j = 2, and 0 for all other ij, and so on. jk; 2 jk The expectation of outlet performance is assumed to be contained in the space generated by the ”one” vector, the n X vectors, and the m Zj i vectors. Examination of this assumption space reveals that the space generated by the one vector and the X1 vectors is a subspace of the space generated by the Z vectors. To illustrate, let Zij denote the J m Zj vector. The one vector is equal to E Zj and each Xi vector is m j=l equal to E ‘sz xij' Thus, the normal equations to be solved to esti- i=1 mate the parameters will be linearly dependent. Therefore, without some restriction, there is no unique solution to the normal equations. The number of possible solutions is, in fact, infinite. The determina- tion of a particular solution from this infinite number of possible solutions depends upon the definition of the managerial contribution. For the environmental model, a unique solution is determined by defining the managerial contribution (Mj) as the expectation of measured 25 performance of an outlet for sub-period k minus the sum of the regres- sion constant and the effect of the non-managerial factors. Thus: n M. = E Y. ‘— A + B. X.. . J (31.) < 21:1 1 13) ) is estimated by the average of the measured performance of k outlet j over all k sub-periods of t; A and the "Bi”s can be estimated The E(Yj using the usual least squares procedure. The estimated contribution of the manager of a typical outlet j (Mj) is the difference between the average of the measured performance of the outlet over time period t and the expected performance of the outlet considering only the levels of the non-managerial factors under which the outlet is operated. Thus: A —» ,A n A M,=Y,—(A+E B.X.). J J 1:]. 1 lj An example of the computation of Mj is included in Appendix A. HE is not solely a function of the actions of the outlet manager: the effect of any non-managerial factors not explicitly considered in the model and the effects of managerial action are confounded in M3. However, all non-managerial factors thought to explain significant vari- ation in performance among outlets can be tested for inclusion in the model. Furthermore, even if the effect of only a single non-managerial factor is removed, H5 is an improvement over a measure of outlet per- formance because the effect of one non-managerially controllable factor has been removed from the measure of performance. The sum of the estimated managerial contributions will be equal to zero. The average of the estimated managerial contributions will be 261': :1 .n a fit. VA Min TI TA ..-\ C. . . . . .C .1 A. u .C 1» a .1; ..L. 2. V. 1 e .C s t .C u. .1 .E I 2. . arc vi .1 at: he 91» .3 u n .C 2a a .. ..s 1: ..nl. alw- - 8. PA a 4 r C k 33 t u t «4‘ .4; .\u a n: u u 31 .1: u . A A his 5.. H p Y“ \FA .hs 26 zero: a manager with a zero contribution will be assumed to be an aver- age manager. The better than average manager will have a positive con- tribution and the below average manager will have a negative contribue tion. In the test company (and likely in other applications) the man- agerial contribution estimates were approximately normally distributed about the zero mean. The contribution estimates were converted to a standard normal distribution and percentile standings were determined using a cumulative standard normal table. For evaluating the perform- ances of managers, the estimated percentile standing of a manager is a measure of performance with a "built in" intraperiod standard of com- parison. The expression ”linear" is used in several senses in mathematics and a few words of explanation are in order. "Linear" in linear model means that the dependent variable can be expressed as a linear combi- gation of the parameters and unexplained terms. It does not mean that the relationship between variables is expressible only as a first-order equation. The relationship between variables may be a second, third, or ”n”th order polynomial. These higher order relationships can be ex- pressed in the linear (combination) model by letting independent variable X equal (X1)2, X equal (X1)3 and so on. 2 3 The use of cubic and higher order terms can be dangerous. As Fraser has stated, "A cubic term is a fast-changing function; if its co- efficient is in error, the estimate of the mean response can be very much in error, especially for values of the controllable variable away from those for which data were obtained."8 This limitation of linear 8Fraser, p. 296. 27 models should not be particularly troublesome in the multi-outlet busi- ness case since there are few factors which could logically be cubically related to the measured performance of outlets and virtually none of a fourth or higher order. In addition to the higher order effects of primary factors, inter- actions among factors can also be estimated in the linear model. An interaction effect exists when a combination of factor levels yields results different from the sum of the effects of individual factor levels considered alone. Estimates of the effect of interaction among various combinations of two primary factors say, X and X , can be generated by l 2 including an interaction term X where X is equal to X 3, 3 multiplied by 1 X2. When many primary factors are considered, all possible interactions cannot, in general, be estimated because of the limited number of outlets available. The number of possible interactions increases geometrically as primary factors are added to the model. However, many of the logi- cally important interactions can often be estimated. In the test com- pany, all interactions which were thought to be important were tested. The Testing 2f Hypotheses Many accounting reports are based, in part, upon estimates and yet carry no statement as to the variance of the reported amounts or even a caveat as to the fact that estimates are being used. In the environ- mental model, no statistical assumptions are necessary for a least squares estimate of the effect of any factor. However, it is possible and desirable to estimate the variance of the estimates of the parameters and thus obtain some assurance as to the precision of the parameter 28 estimates. It is desirable to have some assurance that the estimated effects of the factors-~both non-managerial and managerial--are statis- tically significant and cannot easily be explained by chance. To make inferences and tests of significance from the estimates, certain statistical assumptions must be met or the analysis must be ex- tended or revised to account for the lack of the required assumptions. There are three assumptions basic to the standard analysis of linear models. These are: (l) Homogeneity of variances--the variances of the un- explained terms do not depend on the levels of the independent variables. (2) Independence--the unexplained terms for the dif- ferent observations are statistically independent. (3) Normality--the unexplained terms are normally distributed. There are numerous tests available for determining whether these assump- tions are met. When the original data do not satisfy the assumptions, all is not lost. One common way of meeting the assumptions is to transform the data so that the transformed data meet the assumptions. A single trans- formation may suffice or perhaps a series of transformations--each for a separate deficiency--may meet the requirements. Fortunately, a trans- formation to correct one deficiency often helps to correct another. 9The specific tests used to test the assumptions for the test company will be presented on pages 55 and 56. 0Bernard Ostle, Statistics in Research, The Iowa State University Press, Ames, Iowa, 1963, p. 340. 29 In the company tested the assumption of homogeneity of variances of the original dependent variables (sales and controllable expenses) was not justified. Rather, the variability of these performance meas- ures increased as the magnitude of the variables increased. This same relationship would seem likely in other multi-outlet businesses. For the test company, a logarithmic transformation of sales and a square root transformation of controllable expenses satisfied the required assumption. Even when all of the assumptions cannot be met by transformation, the procedures can sometimes be revised so that a meaningful analysis is possible. Much of the work in the statistics literature in recent years has dealt with the analysis of experiments for which the basic assumptions are not met. If the assumptions are not met, the analysis is often complicated considerably but is not impossible. Under conditions of homogeneity of variances, independence and normality discussed above, hypotheses concerning the parameters can readily be tested by an analysis 2f variance. Specifically, the hypothe- sis that a parameter or group of parameters is equal to zero can be tested. Since the analysis of variance for the environmental model is no different from the analysis of other linear statistical models, a discussion of analysis of variance will not be presented here. For a discussion of analysis of variance in general, see Fraser11 or Freund.12 In summary, statistical inference is a means by which the validity of the environmental model approach to performance measurement in 11Fraser, pp. 261-269. 12Freund, pp. 331-337. 4. 1:1. ES ERG It to: 30 multi-outlet businesses can be established. The conceptual model has been presented. The incorporation of the estimates of managerial performance into the accounting system will now be discussed. Revisions £9 the Accounting System The environmental model requires no changes in the traditional accounting system to generate the revised managerial performance reports. The only added features required are the non-managerial factor classifi- cations and the level of each factor under which each outlet is operated. In some applications of the environmental model (for example, in those applications involving no transformation) it may be possible to account for the dollar contribution of an outlet before occupancy costs as the sum of the estimated dollar effects of the non-managerial factors and managerial contribution and the dollars not explained by the model. It is questionable whether the estimated dollar effect of managerial contribution should be reported even in those (few) applications in which it is possible. It is questionable because the estimated manage- rial contribution is a measure of the relative value of a manager when compared to other managers and not a measure of the absolute value of a Imanager to a company. The percentile standings of managers are less likely to be misinterpreted and can be explained more easily to the out- let managers and other operating personnel. The only change in the outlet operating reports is the addition of tile percentile standing of the outlet manager which is based on the I'esults of his actions, given the potential of the resources at his disposal. Figure 2.1 is a hypothetical performance report for outlet 31 number one for the month of January, 1967. In the example, the environmental model analysis has been performed on sales and controllable expenses as well as the net contribution of the outlet. This division of net contribution into the two components was made because the factors associated with revenue differences may be considerably different from the factors associated with expense dif- ferences. By separating the components of the net contribution and per- forming a separate analysis on each component, the precision of the estimates may be improved if the non-managerial factors are related to the components in different ways. A factor which significantly influ- ences both sales and controllable expenses may have offsetting effects and thus, the factor bears no significant relationship to the net contri- bution. Furthermore, managerial excellence in selling may offset poor cost control or vice versa. Controllable expenses could be further divided into components if desired. Figure 2.1. Hypothetical performance report OUTLET NUMBER ONE CONTRIBUTION STATEMENT Month of January, 1967 Percentile Standing of Manager Sales: $10,000 86 Controllable Expenses: 7,000 45 Net Contribution before Occupancy Costs $ 3,000 70 Outlet Occupancy Costs: 500 Net Contribution to Unal- locable Costs and Profits $ 2,500 a: All ll! "14 CVGH 32 Occupancy costs are shown separately because the occupancy costs are generally not controllable by the outlet manager. The occupancy costs are included on the performance report because the costs are rel- evant when evaluating the performance of the combination of the outlet location, facilities and manager. During January, the manager of outlet number one did quite well in the selling aspects of the business, given the environment within which be operated, but a little below average in cost control. The estimated net managerial contribution is above average but if the outlet manager could exert more effort in the control of expenses and reduce expenses without reducing sales, he could improve his overall measure of perform- ance and percentile standing. To illustrate the possible usefulness of the environmental approach to performance measurement, consider Figure 2.2 in which contribution statements for three additional outlets are presented. The relationships shown are hypothetical but similar results were found in the test com- pany (see pages 53 and 54). The last line of Figure 2.2 is the per- centile standing of the outlet managers based upon the responsibility accounting model discussed on pages 20 and 21. Such a percentile stand- ing is not generally computed in a responsibility accounting system. The standing has been presented here to illustrate the changes in rank- ings of managerial performance when the non-managerial factors are con- sidered. When considering the net contribution of the outlets as a measure of managerial performance, the manager of outlet number one appears to perform on a par with the manager of outlet number two, to perform poorer than the manager of outlet number three and perform much poorer In: u‘na U7 33 than the manager of outlet number four. If the potential of the re- sources at the disposal of the outlet managers is considered (by using the environmental model to estimate the effects of the various non- managerial factors), it is discovered that the manager of outlet number one is performing much better than the managers of outlets two and four and on a par with the manager of outlet number three. Outlet contri- bution is thus not a good indicator of managerial performance. Figure 2.2. Hypothetical performance reports for selected outlets for January, 1967 Outlet 1 2 3 4 § PSM* $ PSM $ PSM $ PSM Sales 10,000 86 10,000 40 6,000 86 20,000 33 Controllable expenses 7,000 45 7,000 35 2,000 45 13,000 38 Net contribution before occupancy costs 3,000 70 3,000 37 4,000 70 7,000 35 Percentile based on Responsibility Accounting (40) (40) (55) (80) * Percentile standing of managerial performance based upon the en- vironmental model District supervisors and top management officials of multi-outlet businesses realize that there are differences in the profit potential among the various outlets. Management may be familiar with or suspect several factors as being important determinants of outlet performance. The questions which need to be answered are: How important are the known or suspected factors? Are there other non-managerial factors which are important? Are interactions among factors important? The environmental model is an objective method for obtaining answers to 34 these questions. To continue the illustration, assume that the firm is organized into districts and that the district supervisors have the authority to hire and dismiss outlet managers within their districts and can control the levels of some of the non-managerial factors under which the outlets are operated. The average estimated percentile standing of the managers in a particular district is a measure of the skills and performance of the district supervisor in hiring and supervising his subordinates. The magnitude of the non-managerial factors which the district supervisor can control (or perhaps the ratio of controllable to non-controllable factors) gives an indication of the sensitivity of the operations of the district to the district supervisor's control. It may be tempting to extend the measure of outlet manager perform- ance and evaluate the district supervisors' performance on the basis of the magnitude of the controllable non-managerial factors as an indication of a district supervisor's success in adjusting the controllable factors to a maximum expected contribution. Such an extension is hazardous, however, because there are differences in the relative potentials of the districts. An expected contribution analysis for districts would be necessary to evaluate their relative potentials for the extension to be valid. An environmental model for district supervisors' performance is not appropriate or necessary. There are not enough districts to generate precise estimates of the influences of all of the non-managerial factors affecting district contribution and the responsibilities of the district supervisors are not likely to be homogeneous among districts. Further- more, a subjective standard is probably adequate at the district level. 35 There are only a few districts and an experienced top management official can probably establish adequate predetermined standards for the small number of districts. Since one official can be familiar with all of the districts, this individual could set all the district budgets and there would be no bias caused by differences among the budget-setting officials. In short, the districts meet the requirements for a pre- determined standard system for performance evaluation. The environmental model is a performance measurement and evaluation tool to be used by top management and district supervisors to evaluate the performance of subordinates. The sum of the measured performances of all outlets is equal to mp(A +i E, ii.) --the sum of the esti- mated managerial contributions (fof-ihe company as a whole) is zero. It is only for segment analysis that the environmental model has mean- ing. Implications The implications of the model for evaluation of outlet manager per- formance have been discussed already. There are other implications, however, for manager, location and facilities selection. One possible conclusion from the analysis is that there is rela- tively little difference in performance among the managers. That is, most of the difference in performance among outlets is due to the dif- ferences in the levels of non-managerial factors such as outlet lo- cation and outlet facilities. In this case, the analysis would indicate that perhaps more resources should be devoted to outlet location and facilities selection and fewer resources to manager selection. Further- more, the relative magnitude of the non-managerial effects are (J 9—1- (J O ‘91 5‘. ti H rf‘ r0 36 indicative of which factor levels and combinations of factor levels are the most productive in terms of measured performance results. An alternative conclusion could be that the managerial effect is quite large: that is, non-managerial factors explain relatively little of the differences in the various outlets' performance. The evaluation of managers on the basis of the performance of their outlets may yield the same relative ranking as an evaluation based on the imputed manage- rial contributions, and if so, the outlet performance may serve as a valid intraperiod standard of performance. In this case, the model building and analysis of variance will have been worthwhile because it shows that the use of the performance of the outlets as an intraperiod standard is justified, and that considerable resources should be devoted to manager selection and training and perhaps fewer resources to lo- cation and facility selection. The outlet with the smallest reported dollar contribution to un- allocable costs and profits isn't necessarily the outlet which should be eliminated if an outlet is to be eliminated. Additional information and analysis is needed. The environmental model can provide some of the information needed for such an analysis. Poor reported contribution can be due to a poor locationigg poor facilities,g; poor management or some combination of these factors. A good location.with poor facilities and a poor manager should probably be remodeled and restaffed, not eliminated. A poor location with good facilities and a good manager should be eliminated and the funds and manager thus freed could be transferred to a better location. From the analysis of the effects of facilities, marginal rates of contribution could be developed for comparison with marginal investment 37 required in remodeling an outlet. Indeed, ”response surface" techniques could be incorporated to determine the optimum combination of levels of factors for location and facilities (that is, the combination yielding maximum expected contribution). The predictive value of such a system could be compared with selection techniques currently being used by the firm. The implications of the environmental model for location and facility selection will be discussed in more detail in Chapter Four. The implementation, testing and evaluation of the environmental model approach in an actual multi-outlet business is the subject of Chapter Three. CHAPTER III TESTING THE MODEL The construction and testing of an environmental model for perform- ance measurement in a multi-outlet business is the subject of this chap- ter. Two groups of tests will be conducted. In the first group are the tests for the presence of the underlying assumptions for the analy- sis of linear models. The second group consists of the tests of sig- nificance of the non-managerial factors which are being considered to explain the variation of the sales and controllable expenses of the outlets. The tests of the assumptions make possible the use of statistical inference in the analysis of the results of the environmental model. Tests of significance of the factors give the level of confidence in the results of the analysis. This chapter begins with a discussion of the background of the test company and the specific non-managerial factors and relationships tested. The Test Company A national company offering a wide range of consumer goods and services through retail stores and catalog order centers was selected as the firm on which to test the environmental model approach to perform- ance measurement. The study was limited to catalog order centers. Cata- log order desks in retail stores of the firm are included in the retail 38 39 store division and were not included in the study. The test firm satis- fied the requirements of the definition of a multi-outlet business as defined in Chapter One (page 12). The order centers were chosen because of several distinct advan- tages inherent in the nature of the catalog order business, including the overall comparability of the centers. The ease of classifying the centers as to the levels of the non-managerial factors is perhaps the greatest advantage of the company as a test firm. There is generally only one order center in a city and the problems related to location within cities are largely avoided. One complete region in which there are ten districts was the basis for the model building. The complete region preserves the operationality of the model in that a complete operating segment of the national organi- zation is included. There were 105 order centers operated in the region in 1964, 117 in 1965, and 116 in 1966. Of these outlets, 20 were lo- cated in the "core city" of large metropolitan areas and were excluded from the study primarily because of the difficulty of defining market areas, breaking down population characteristics and other problems of classifying the outlets.1 In addition, the metropolitan outlets ex- hibited considerably more variation in sales and controllable expenses so that even a "core city" effect was not meaningful. Regression co- efficients based on outlets located in towns and cities with from 5,000 to 50,000 inhabitants might be considerably distorted by the inclusion of a few outlets from a heterogeneous population of cities with as many 1This lack of detailed demographic data from secondary sources could be overcome with market analysis. 40 as 500,000 inhabitants. For these reasons the core city outlets were not included in this study. Outlet managers have primary responsibility for the hiring and re- tention of outlet employees. The typical outlet staff includes the out- let manager and from two to six employees. There is no complicated managerial hierarchy within the outlets which might influence operating results. All order centers are of approximately the same size. Because of this, managers are not generally promoted to larger outlets but are pro- moted to district supervisor and supervise from ten to fifteen outlets. The identification of managerial skills is particularly important in order to retain the excellent managers through promotion. The company already has a contribution accounting system in opera- tion and this makes the extension of the responsibility accounting con- cept by the use of the environmental model easier. Monthly operating statements are the most frequent comprehensive reports in the test company. Monthly inputs for the model were possible, but since the re- sources available for data collection were limited, it was deemed ad- vantageous to use a longer time period to allow an indication of the stability of the relationships over time rather than a more detailed estimate of performance for a shorter period. Quarterly data on outlet sales and controllable expenses were collected for the three-year period ending December 31, 1966. The portion of the company tested is admittedly a simplified situ- ation but more complicated organizations and factor structures can be introduced in other applications through appropriate revisions of the 41 model.2 The present objective is to determine whether the basic concept of the environmental model is implementable and potentially useful in multi-outlet businesses. To meet this objective, the following primary hypothesis was developed and tested. The Hypothesis The primary hypothesis of this study is: .EHEES.EEE non-managerial factors which explain significant variation $3 the measured performance 2f outlets‘ig‘g multi-outlet business. Non-managerial factors influ- encing revenues may be considerably different from those related to ex- pense differences and for this reason the two will be analyzed sepa- rately. Formulating the Model for Testing In formulating the model for testing, corporate marketing officials were asked to identify the non-managerial factors which they believed might cause differences in measured performance among the outlets. From the discussions with corporate officials and a review of the related literature, the initial formulation of the model was derived. The dependent variables analyzed were net outlet sales and the out- let expenses which were controllable by the outlet manager. As mentioned in Chapter Two, the analysis could be performed on individual components of net sales and controllable expenses to gain additional insight into the operations of the outlets. In this study, however, the analysis 2A complete model of the test firm including both retail store and catalog divisions could be developed. 42 was limited to the overall measures of performance because of the explor- atory nature of the study. Data on seventeen non-managerial factors (see Figure 3.1) were gathered for all outlets for consideration in the initial formulation of the model. The demographic characteristics of the communities in which the outlets are located were obtained from various publications of the Bureau of the Census of the United States Department of Commerce. Information about the other location characteristics, the physical fa- cilities of the outlets and the other factors were obtained from the financial, personnel and property records of the test company. From this group of seventeen primary factors, forty-four additional variables were created and tested to determine whether any quadratic effects or certain logically important first-order interaction effects could explain significant variation in the dependent variables. The same factors which were considered for sales were considered for controllable expenses except that sales was included as an inde- pendent variable in the analysis of controllable expenses. It might appear circular to remove the effect of sales, which is partially con- trollable by the manager, as a non-managerial factor in the analysis of controllable expenses. HOwever, to a large extent the expenses of an outlet are determined by the volume of sales due to the direct nature of commissions and order processing costs. If the manager keeps his ex- penses at a lower level for a given volume of sales than the average manager does, he should be congratulated. The deletion of sales from the controllable expense equation would merely emphasize the other fac- tors (other than sales) which are related to the sales of the outlet. 43 Figure 3.1. Primary non-managerial factors considered in the initial formulation of the model Location Characteristics of the outlet: Demographic characteristics: Population 3+1-X-X-X- Lnwar-d . Median age Other: Median family income Median years of schooling Percentage of population which is non-white 6. District in which outlet is located 7. Location within city (i.e. in a shopping center or other location) 8. Distance from distribution center (warehouse) 9 . Competition - store 10. Competition - outlet 11. Competition - 12. Competition - 13. Competition - primary catalog competitor - retail primary catalog competitor - catalog secondary catalog competitor #l secondary catalog competitor #2 secondary catalog competitor #3 Physical Facilities of the outlet: * 14. Years since last remodeling of the outlet * 15. Years since outlet first opened in community Other Factors: * l6. Managerial tenure in years * 17. Number of catalogs issued by the outlet * Indicates quadratic effect tested The population of the cities varied from approximately 5,000 to 200,000, with only about ten percent above 75,000. It would be expected that average sales would increase with increasing population but at a decreasing rate due to the greater retail competition in the larger towns and thus, the greater availability of goods from alternative SOUI‘C BS 0 44 The relationship of median family income to sales might be expected to be an inverted "U” shape due to income and substitution effects (to the detriment of catalog order center because of the nature of its prod- uct offering) which occur as family incomes rise. The median years of schooling classification is intended to approxi- mate possible differences in buying habits of different socio-economic classes of consumers. Office workers and factory workers may earn equal wages but have different buying habits. The median years of schooling may provide useful insight into the socio-economic makeup of the com- munity. The classification, "percentage of population which is non-white" arises because of considerations similar to the median years of school- ing classification. Non-whites may shop in different establishments and types of establishments than do whites. The effect of the difference in potential due to race should be reflected in the performance measure of the outlet manager. The reasons for the inclusion of the median age, district and lo- cation within cities factors have been discussed in Chapter Two. Increasing distances from the warehouse should be a detriment to the outlets for two reasons. First, as distance increases, the effec- tive price to the customer increases due to the increased transportation charge which is added to the order price. Second, as distance increases, the delivery time increases and the relative advantage of local retail competitors increases. A third detriment does not arise in the region studied but might arise in other regions. The third detriment occurs when another catalog competitor has a warehouse closer to the city in question than the firm does and the competitor can thus offer faster 45 service and reduced transportation charges. In this study, all catalog competitors' warehouses are located in the same city. Since there is little difference in facilities among the outlets there are few facility differences which can be tested. The average outlet has approximately twenty-five hundred square feet of space with very little deviation among outlets. The size variable was ignored be- cause of the lack of significant size differences. The firm does not have definite outlet model types such as exist in service station chains and many food chains. Outlet designs change gradually over time and are tailored to the store complex in which the individual outlet is located. Thus, the length of time since the last remodeling of the outlet is perhaps the best measure of the general attractiveness of the outlet facilities. The years since the outlet was first opened in the community is a complex factor. The age of the outlet facilities may give a measure of the general surroundings in which the outlet is located (other stores in the neighborhood may not have been remodeled). This effect would presumably be negatively related to sales. On the positive side, the age gives a measure of the length of exposure of the community to the firm. The algebraic sign of the age effect can perhaps indicate which of these aspects prevails if indeed the age factor is significant. The competitors of a catalog order center are difficult to define and classify. Due to the complete line of goods and services offered by the firm, almost any business in the community could be considered as a competitor of the outlet. Total square feet of retail selling space in the city was considered as a possible factor basis but data were not readily available for the small communities. Furthermore, 46 different mixes of retail competition would likely have more influence on the outlet than the total space available. For practical reasons, the presence of other catalog order firms' outlets in the community were the only competitive factors tested. The presence of the primary catalog competitor was further distinguished as to the type of facility--that is, a retail store or a catalog order center. TWO of the secondary catalog competitors are well established catalog firms which offer goods only through the catalog. The third is a national dry goods chain which has only recently entered the durable goods market and established a mail order department in many of the local retail stores that do not offer a complete line of durable goods. This third secondary competitor has had an excellent reputation for its medium-priced soft goods lines for many years. For application of the environmental model to other types of multi-outlet businesses, better measures of competitive factors can undoubtedly be developed and the precision of the analysis increased accordingly. The effects of many of the location and facility factors discussed above may be of specific interest in planning new locations, store re- modelings and expansions. Managerial tenure was tested because years of experience may be related to outlet performance in several ways. As tenure increases, customer loyalties may be developed resulting in increased sales, while experience in managing employees and cost control may help reduce ex- penses. The effect of tenure should be removed to avoid underestimating the capabilities of an inexperienced but otherwise excellent manager. The nature of the tenure relationship may also be of interest in plan- ning managerial tenure periods and job rotation policies. 47 The number of catalogs issued by the outlet is probably the most important single non-managerial factor related to sales because it is proof of past purchases from the firm. Catalogs are issued after a customer has ordered a minimum number of times from the company. The manager has limited discretion in issuing catalogs for promotional pur- poses. The number of catalogs is important because outlets in two cities with equal populations and other location factors and facility factors may have different profit potentials because the number of past customers (and thus the number of current catalog holders) is different. The catalogs are virtually the only form of local advertising utilized by the outlets. With this background, the construction of the final model, the tests performed and the results of those tests begins. The Model All of the primary non-managerial factors listed in Figure 3.1 and the derived factors (quadratic effects and interactions) were thought to be potentially significant in explaining variation in the measured performance among outlets. Some were thought to be highly significant while others were thought to be of only minimum significance. To de- termine the relative importance of individual non-managerial factors a "stepwise addition of variables" was performed. In the stepwise addition of variables, the independent variable which reduces the unexplained sum of squares the most is the first independent variable added to the regression equation. The procedure is then repeated until the variable considered for addition to the equation is not significant at some prespecified significance level. For the test company the .05 48 significance level was used. With the individually significant non-managerial factors thus de- termined, a procedure fitting all of the significant factors simulta- neously was conducted. Figure 3.2 is a list of the non-managerial fac- tors which were significant, the algebraic sign of the coefficient of the factor and the number of years in which the factors were significant. The factors are listed in approximately the order in which they were added to the equations in the stepwise additions of variables. In Appendix B, the factors are listed (by year) in the order in which they were added to the equations. As was anticipated, the number of catalogs issued by an outlet was the most important factor related to the logarithm of net outlet sales for all three years. The negative sign of the coefficient of the quad- ratic term (which was significant in two of the three years) indicates that, as the absolute number of catalogs issued by an outlet increases, the sales per additional catalog decreases. For the range tested, the relationship of sales to the number of catalogs was constantly in- creasing. That is, the stationary point (the point at which the first derivative is equal to zero), beyond which total sales decrease with each additional catalog issued, was far to the right of the range of the numbers of catalogs which.were actually issued by the outlets. The years since the opening of the outlet in a community was also significant in all three years. The positive sign of the linear term indicates that the exposure of the firm to the community was associated with increased sales. As with catalogs, sales increase at a decreasing rate until a maximum is reached. The stationary point is at approxi- mately twenty-five years which was well within the range tested. In a 49 study of six supermarket chains, Applebaum found a similar age-sales relationship. Figure 3.2. Significant non-managerial factors Logarithm.2£ Net Outlet Sales Significant“; Years Significant‘g Years Significant 1 Year + Number of catalogs - Number of catalogs + Median age (quadratic issued issued (quadratic effect) + Years since opening effect) + Population X.Median + Median family income - Median age family income + Years since remodel- - Years since opening + Shopping center X ing (quadratic effect) Population + Years since remodel- - Secondary competitor - Population X Secondary ing (quadratic #3 competitor #3 effect) + Median years of * District schooling + Median years of schooling (quad- ratic effect) Square Root 2f Controllable Expenses Significant 3 Years Significant,2 Years Significant,l Year + Sales - Managerial tenure + Distance from distri- bution center * The sign of the district effect for a particular district may be positive or negative. The sum of the district effects is zero. Median family income was positively related to sales and there were no significant quadratic effects during any year. Thus, the higher the median family income of a community the higher the expected sales of an outlet located in the community. As with some of the other non- managerial factors, the relationships may not hold for median family 3William Applebaum, "Store Performance in Relation to Location and Other Characteristics," Chain Store Age, Executive Edition, v. 41, November, 1965, p. E16. 50 incomes beyond the range tested. The highest median family income for the communities in the study was under $10,500. The shape of the "years since remodeling--logarithm of sales” curve was similar to the ”years since opening--logarithm of sales" curve, but the years since remodeling curve reaches a maximum much earlier than does the years since opening curve. It may appear unusual that sales continue to rise for a few years after remodeling rather than tapering off as the remodeled facilities age. A possible explanation of this is that part of the decision to remodel is based on a higher than average rate of growth of the community in which an outlet is located. This higher than average rate of growth disguises for a time the effect of the aging of the remodeled facilities. Differences in average outlet performance among the districts were significant for all three years. Since the evaluation of the performance of district supervisors was not of interest, the underlying cause of the differences among districts was not investigated. Whether the dif- ferences in potential are due to differences among the supervisors, dif- ferences in climate or other factors is of no consequence in evaluating the performance of the outlet managers. The district factor is con- sidered only because the effect of factors beyond the control of the outlet manager should be removed from the measure of his performance. The remainder of the non-managerial factors listed in Figure 3.2 were significant in only one or two of the three years. Thus, less re- liance can be placed on the results of any analysis based on their in- clusion. However, the nature of the relationships indicated by these factors is of interest in understanding the profitability factors within the firm. For this reason, a brief discussion of the possible 51 significance of these factors is included. Median age in the test region ranged from twenty-one to thirty- nine. The younger median ages were associated with higher sales and the older median ages were associated with somewhat lower sales. The posi- tive sign of the quadratic effect (significant in only one year) indi- cates that the "median age--logarithm of sales” curve is ”U" shaped. However, the indicated stationary point is very near the right bound of the median ages in the communities tested. In general, the higher the median age in a community the smaller the expected sales of an outlet. The median years of schooling factor seems to indicate that the outlets located in "white collar" communities are likely to have higher sales than those located in "blue collar” communities. The correlation of median family income with median years of schooling was near zero. The presence in a community of what was thought to be the primary catalog competitor of the test firm had no significant impact on outlet sales for the region and time periods tested. The presence of the third secondary competitor did have a significant effect on outlet sales in two of the three years, however. This can possibly be explained by the competition of the medium-priced soft goods lines which the com- petitor can offer from stock. Fashion merchandise makes up a large part of the sales of most of the outlets of the test firm. 4 Three interaction effects were significant. The Population X Median family income interaction indicates that large populations and high median family incomes, ig combination, yield sales that are higher than Can be explained by the levels of population and median family income alone. Similarly, in large communities, outlets located in shopping <2enters seem to do much better than those in other locations within 52 cities. The presence of the third secondary competitor in a large com- munity may be particularly disadvantageous to an outlet. The presence of the third secondary competitor was not highly correlated with popu- lation (correlation coefficient was approximately .18). As was to be expected, sales was the most important independent variable related to controllable expenses. Managerial tenure was sig- nificant in two of the three years and was negatively related to ex- penseS. As tenure increased, controllable expenses decreased; the effect of tenure should be considered in the environmental model so that the inexperienced managers will not be underevaluated. The presumed distance effect was significant in only one year and thus, does not appear to be of particular importance, at least for the test region. The partitions of the sum of squares, the related degrees of free- dom and the computation of the test statistic for testing the signifi- cance of differences among the managers for the three test years are ' presented in Table 3.1 and Table 3.2. The multiple coefficient of de- termination (R2) is the ratio of the sum of squares explained by the model (after the mean) to the total sum of squares (after the mean). The multiple coefficient of determination can be interpreted as the proportion of the variation of the dependent variable which is "explained" by the model. The variation explained by the environmental model is much smaller for controllable expenses than is explained for sales. Part of the unexplained variation in expenses is due to accounting adjustments of prior quarters operating results for many of the outlets. If these ad- justments could be analyzed and redistributed to the appropriate periods, more of the variation of the controllable expenses could be explained by 3 5 eon-uauamuam soc-ouuucuam do. coconuuucuqm «nowoc. -o~n~.o coonuo. moonsu. canon M. “mummw.a~oz ssONOO. 5N00¢N.n~ nae—HO. soonee. ado-x M unusmw coo: coouoo. somon~.n~ woo—no. oowmnq. ofiuom M ou-amw ado: mofimm umfiuoo um: mo aooouum mo noomuvn acuoohh mm i a man now as o~ 6230 um wlo. g Esuwumwofi Mao I an unnNNn.n HNNmoo nahnnn.~ ocoaun.n mwmummm.ml.llw mam I N! ocwenn.uu mcqaco. oonmn¢.N snuNHn.o n~mnn~.~a Nmomoo n~u¢~c.~ onon~n.¢ mmmummw.mm.mmw "moamwum> wo mwmzfimc< .a.m oHan Acne! osu nauwnv uduoa toad-Hauuafi main-cl: 7308 05 .3qu uncuuau "awuuwuanIIOI scauawua> mm.mmmmmw «cod Adios ucu uuumav “Inca voaadanuuc: qumdaul Asia! ecu unumdv uncuuaw uuwuuwdaulunon_ coauowuu> mm.mmmmmm mood Acne! onu “dunno anuoa 83:90.5 chum-ail Anne! ozu know-v cacao-u H-wuqucilouoz .533»; wlo. a cow" 54 do. oucdouuumu«m Ho. oucauafifla nNo. cocoowuumuum ~62 oen.~ «.3.» 06?.3 Sana M "modal g nénw 3min n.3n 0.03.: 0.33. M "a use: ~.oh~ n3; o.no~ 0.0¢o.nn on; M a g momCoaxo mfinmaaouucoo wo uoou gag a... cmN nm N gvooum MN M013 mumsvm ”mocmwum> mo mflmmamc< can I Nu 2.3.2: amine 3a.! 3....3 gum.” man I N! «3.92 .fimJo: _ 8%.: «9.8 mflflfiflwwmflw was I NI nn~.nan nuh_°m N0§.Q~ 006.cnu and” .N. m mHan 2.3- 2.» hum-v 38 338.... anon-alt A8... 2.. .33 :59: and; gnaw 30a Anal of 833 ~38 1333 Pang :33 1: hum-v Faun-u dial-III! flaws—4% clan of 83: ~38. 3:11.88 PUD-III cl!- 3... flaw: 9:3qu and: inflow 30.— 55 the model. The total variation explained by the model is the sum of the amount explained by the non-managerial factors and the amount explained by the differences among managers after considering the non-managerial factors. For sales, the non-managerial factors explained 74.3%, 73.7%, and 71.4% of the variation for 1966, 1965, and 1964 respectively, while differences among managers eXplained 20.6%, 21.1%, and 19.9% of the total. For controllable expenses, the non-managerial factors explained 59.4%, 23.0%, and 27.0% of the variation and 13.2%, 34.5%, and 31.4% was explained by differences among managers. Nearly four times as much of the variation in the logarithm of sales among outlets is explained by the levels of the non-managerial factors under which the outlets were operated than by differences among managers after considering the non-managerial factors! [A Statistical Note To determine whether the three assumptions basic to the analysis of linear models were justified in the test company, three tests of the assumptions were conducted. Bartlett's Test for Homogeneity of Variances4 was used to test for the homogeneity of variances assumption. The hypothesis of homogeneity had to be rejected for the seasonally ad- justed sales and controllable expenses. However, when sales were sub- jected to a logarithmic transformation and expenses subjected to a square root transformation, the hypothesis could not be rejected at the 4M. S. Bartlett, "Some Examples of Statistical Methods of Research in Agriculture and Applied Biology," Journal gf Royal Statistical Society, (Supplement), v. 4, 1937, p. 137. 56 .05 level of significance. The standard deviation of sales and the variance of controllable expenses were apparently proportional to the means of these variables for the outlets in question. These transfor- mations are "order preserving": the relative rank order of the sales and controllable expenses of the outlets is preserved for the transformed values of these variables. The Kolmogorov-Smirnov Test for Goodness of Fit5 was conducted to test the assumption of normality. The normality assumption was justi- fied at the .05 significance level for the transforms of sales and con- trollable expenses. To test for independence or random unexplained terms, the Durbin-Watson ”d" statistic6 was computed. The computed ”d” statistic exceeded the approximate upper bound for the test statistic in all six analyses and thus the hypothesis of random unexplained terms was not rejected. Testing the Primary Hypothesis The primary hypothesis, that there are non-managerial factors which explain significant variation in measured performance of outlets of a multi-outlet business, has been accepted for the test company. The logarithms of net outlet sales and the square roots of controllable expense meet the assumptions basic to the analysis of linear models and there are non-managerial factors which explain significant variation in 5F. J. Massey, Jr., "The Kolmogorov-Smirnov Test for Goodness of Fit," Journal 2; the American Statistical Association, v. 46, 1951, pp. 68-78. 6J. Durbin and G. S. Watson, ”Testing for Serial Correlation in Least Squares Regression," Biometrika, v. 37, 1950, pp. 409-428, and v. 38, 1951, pp. 159-178. 57 these measures of outlet performance. Starting from the logical need for adjustments for differences in potentials of outlets of multi-outlet businesses, it has been shown that through the use of the environmental model significant variation in performance among outlets can be explained by factors not under the con- trol of the outlet managers. Also, it has been shown that managerial differences (after considering the effects of the non-managerial factors) explain significant variation in the measured performance among outlets. These findings have important implications for performance evalu- ation. If managers in the test company were evaluated on the basis of the contributions of the outlets to which they had been assigned, an average manager fortunate enough to be assigned to an outlet with a high potential would be very much overrated. Likewise, an excellent manager assigned to an outlet with low potential would receive a rela- tively low evaluation. In other words, the potential of the outlet is likely to be attributed (wrongly) to the performance of the outlet manager. Since significant managerial differences seem to exist, sub- stantial resources should be devoted to manager selection and training. The "Carryover" Effect A problem which has received little attention in the accounting literature is the importance of a residual or "carryover" effect of the actions of previous managers of an entity on the current performance of the entity. This problem is of particular interest when using the environmental model because the goal of the model (for performance measurement) is to make intraperiod standards valid by removing the effect of significant factors which the outlet manager cannot control 58 from the measure of his performance. Certainly the effect upon current operations of previous managers is not under the control of the present manager. Furthermore, after all significant, measurable, non-managerial factors have been considered, there may still remain some non-managerial influences which affect the performance of a particular outlet which cannot be identified or measured on a company-wide basis. The effects of these factors are confounded with the manager's contribution. Even though the effects of such carryover and non-quantifiable factors cannot be segregated for individual outlets, it is possible to test whether such effects seem to exist for the region as a whole over time. If such effects are not significant for the region as a whole over time, increased confidence in the ability of the environmental model to remove the effects of non-managerial influences from the managerial performance measure is justified. The expected contribution of the average manager is zero and his percentile standing is the 50th percentile. To illustrate the problem of the carryover effect, assume that manager number one of a particular outlet has been performing at the 95th percentile according to the en- vironmental model and is promoted or resigns. If manager number two (the successor of manager one at the outlet) is an average manager, the percentile standing of the second manager (which has an expectation of 50) may also be quite high if there is a strong carryover and/or non- quantifiable effect. This high ranking would diminish over time to the 50th percentile as the carryover effect became less important and the true contribution of manager number two became apparent. The percentile standing of a manager who stays at one outlet for several years will probably not change drastically over the periods; it 59 would be expected that the performance of a manager in year one would be about the same in year two and year three. However, if there are no strong carryover and non-quantifiable effects, there should be a con- siderable difference in the percentile standings of three different managers each assigned to an outlet for a period of one year. To test the importance of such carryover and non-quantifiable ef- fects, the eighty-four outlets which operated all three years were divided into three groups according to the number of managers which the outlet had over the three-year period. Thirty-five outlets were in group one which had a single manager over the period. Thirty-eight outlets were in group two which were the outlets with two managers. Only eleven outlets were in group three which had three managers over the period. The variance of the estimated managerial contribution at each out- let was computed and the average variance for the three groups compared. The hypotheses tested were: The average variance of group three is greater than the average variance of group one or group two and, the average variance of group two is greater than the average variance of group one. The results of the tests were not conclusive. The average variance of group two was significantly larger than the average variance of group one. However, the hypotheses concerning group three had to be rejected; the average variance of group three was not significantly larger than the average variance of either group one or two. The result of one test indicates little carryover while the other tests indicate that a strong carryover effect may exist. Group three was a small group and may not be representative of the outlets with three managers in three years for 60 the company as a whole. However, additional testing will be necessary to verify the effect of managerial carryover and non-quantifiable fac- tors on the results of the environmental model. Summary The results of the tests of the primary hypothesis presented in this chapter do not "prove” that the environmental model is the best way to evaluate the performance of the managers of the outlets of the test company or even that the model is worth its cost of operation. It has been demonstrated that, in the test case, there were identifiable and measurable non-managerial factors which explained significant vari- ation in the results of operations of the outlets. The environmental model is an objective method for comparing the performances of all managers in all districts. District supervisors may be better able to evaluate the performance of the managers in their own particular districts because of the expert knowledge and familiarity with the potentials which exist at each outlet. Some of the "potential" characteristics may not be quantifiable and can be accounted for only by subjective evaluations. However, even if the district supervisors can rank their own managers adequately, the problem of comparisons among districts would remain. Through the use of the environmental model, the performance of the outlet 22g its manager has been separated into components attributable to the various location and facility factors and the contribution of the manager. The location and facility factors, from a managerial per- formance measurement point of view, are of interest only so that the effect of these factors can be removed. However, these factors which 61 are not controllable by the outlet manager are controllable at some level of higher management. The implications of the environmental model for the location and facility selection function of management is the subject of Chapter Four. CHAPTER IV IMPLICATIONS OF THE MODEL FOR LOCATION AND FACILITY SELECTION The use of the environmental model for estimating outlet contri- bution to unallocable costs and profits which will result from various combinations of the location and facility factors is the subject of this chapter. The net income of a multi-outlet business is equal to the sum of the contributions of the outlets to unallocable costs and profits minus the unallocable costs. Company net income can be increased by actions which will increase the sum of outlet contributions relatively more than unallocable costs or by actions which will reduce the unallocable costs relatively more than sum of the outlet contributions. The requirements of multi-outlet businesses for control of unallocable costs are essen- tially no different from other large organizations and such control sys- tems will not be discussed here. The maximization of the sum of the contributions of the outlets is the route to higher profits which will be considered. Differences in the contributions of outlets arise from differences in three broad factors: the locations of the outlets, the facilities of the outlets and the outlet managers. To maximize the sum of the out- let contributions, top management can (within certain constraints such as available funds, contract commitments, and costs of abandonments) 62 63 expand, contract, remodel, relocate or restaff and outlet or build new outlets. In the previous chapters the primary emphasis has been on measuring and evaluating the performance of outlet managers. The use of the en- vironmental model to gain a better understanding of the relationships among the various location and facility factors is yet to be considered. With a better understanding of these relationships, locations and facili- ties can be manipulated to increase the net income of the company. The specific non-managerial factors and relationships among factors were of no interest when evaluating outlet managers, because the levels of the factors were not controllable by the outlet managers. The only reason for considering the factors was to remove their effects from the outlet operating results. The scope of the investigation is now expanded to include specific non-managerial factor---outlet contribution relation- ships. Since top management can (within constraints) control the factors which are not controllable by the outlet manager, the specific factors and relationships among factors which are related to outlet performance are of interest when making location and facility decisions. Levels of factors can be adjusted to yield the maximum expected contribution from an outlet or group of outlets. The question which must be answered before such a maximization process can take place is ”How is outlet performance related to the levels of the location and facility factors?" The environmental model will be used to answer this question. The estimates of contribution relationships by the use of the en- vironmental model are based on an analysis of the performance of the outlets in operation at the present time. Thus when using the 64 environmental model to predict the performance of future outlets or the effect of future changes in present outlets, it is implicitly assumed that the present relationships will extend into the (near) future. Since it is known that relationships change over time, it may be de- sirable to base predictions of future performance on the trends in re- lationships instead of the relationships which exist at the present time. With objective information as to how contribution is related to present combinations of location and facility factors (and trends in such re- lationships) management can reduce the uncertainty in making decisions concerning future combinations of factors. A major benefit of using the environmental model is the separation of the effects of the three major determinants of outlet contributions so that informed "piecemeal" decisions as to the desirability of changing the levels of one or more of the factors can be made. Since an outlet may have many managers over its life, location and facility decisions should not be biased by the inclusion of the contri- bution of a different-from-average manager. For location and facility decisions, therefore, the estimated managerial contribution term in the outlet performance equation should be the expectation of estimated managerial contribution which is zero. The Profitability Measures In evaluating the performance of outlet managers it is desirable to separate net contribution of the outlet into its component parts of sales and controllable expenses to give separate evaluations of the manager on his performance in selling and cost control. When making location and facility decisions, such a division is not necessary: the 65 net contribution of an outlet is all that is of interest. Under a responsibility accounting system, outlet occupancy costs would be an expense allocated to the outlets but not charged against the outlet managers because the occupancy costs cannot be controlled by the outlet managers. When making location and facility decisions such costs are, of course, relevant. Occupancy costs could be subtracted from out- let contribution before occupancy costs and the analysis performed on the resulting net contribution to unallocable costs and profits. How- ever, it seems logical that the factors determining occupancy costs can be best estimated by considering such costs separately from sales and controllable expenses because occupancy costs are likely to be related to different factors than are sales and controllable expenses. Sales may be closely related to population and median family income while con- struction costs or rental payments are related to other factors not closely related to population or median family income. Estimates of occupancy costs for outlets could be determined by an analysis of the relationship of construction costs (or value of rentals) in the communities in which the present outlets of the firm are lo- cated. These relationships could then be applied to the locations and facilities being considered. The two measures to be used in estimating outlet performance in this chapter are (1) outlet contribution before occupancy costs (i.e., outlet sales minus controllable expenses), and (2) outlet occupancy COStS 0 Determining Optimum Location and Facility Combinations The determination of optimum location and facility combinations for 66 present outlets and potential outlets will now be discussed. The ap- proach taken will parallel the presentation of O. L. Daviesl in a chap- ter entitled, "The Determination of Optimum Conditions." Davies was concerned with yield in a chemical process while the yield in the present application is outlet contribution before occupancy costs. An optimum combination for a given outlet is that combination which yields the maximum expected contribution to unallocable costs and profits of the company. The determination of and adjustment to optimum combinations of fac- tors for given outlets is not equivalent to the maximization of the net income of a multi-outlet business. The maximization of the net income of a multi-outlet business is an extremely complex subject; it involves the determination of optimum firm size and capital funds market con- siderations as well as the determination of optimum location and facil- ity combinations for given outlets. A detailed analysis of the maxi- mization of company net income is clearly beyond the scope of this paper. However, the results of the environmental model might be used to ad- vantage in such an analysis. 1Davies, 0. L. (Ed.), The Design and Analysis gf Industrial Experiments, Hafner Publishing Company, New York, 1956, pp. 495-551. 2The environmental model may also be useful in estimating cash flows for use in many of the capital budgeting models available today. In the test company, outlet sales depended on the length of time the outlet had been in the community and a definite pattern of cash flows followed a remodeling of the outlet facilities. The exact capital budgeting analysis which should be used by the multi-outlet business is beyond the scope of this study. The purpose here is merely to indicate how the data necessary for many of the decision models available today might be generated. 67 The adjustment to optimum combinations involves the expansion, con- traction, remodeling, relocation or discontinuance of existing outlets and the planning of new outlets. Facility changes are piecemeal revisions and the effect of a facil- ity change is not difficult to visualize. The procedure is similar to partial differentiation in that the change in the level of only one factor is being considered while the levels of the other factor(s) are being held constant. To estimate the change in contribution resulting from a change in a factor such as store size of the addition or deletion of a service, it is only necessary to compute the expected contribution under the two conditions. By substituting the new values of the variable being considered for change into the outlet performance equation, the total expected contribution after the change can be computed. The ex- pected contribution of the present combination would then be subtracted from the contribution of the revised outlet and the difference compared with the estimated cost of the revision in making the capital expendi- ture decision. When making location and facility decisions for increasing the number of outlets, numerous combinations of facility and location fac- tors are possible and the effect of changes in several factors at once is difficult to visualize. To help visualize the environmental model approach, a two and three non-managerial factor geometric representation will be utilized. In a two factor model the relationship between the factors and out- let performance is a surface. The surface represents the performance response of the outlets to the various combinations of the location and 68 facility factors. In a three factor model, the response "surface" is a solid of varying density. A contour map of the response surface can aid in the visualization of the expected effect of changes in up to three location and/or facility factors. While in many applications (as in the test company) there will be more than three significant non-managerial factors, the plotting of the contour of a two or three factor surface can still be useful because the two or three most important factors will often account for most of the variation in performance among the outlets. The algebraic model can, of course, be used for an "n" factor model and rank all possible locations and combinations of facilities which may be considered. The Geometric Representation Assume that the relationship shown in Figure 4.1 is the true, but unknown, relationship between the population of the city in which an outlet is located, the size of the outlet and the contribution of the outlet. The contour lines are the same as the lines of a contour map of geographic terrain--the lines on a map of terrain represent the locus of all combinations of the factors which yield equal contributions. The rate of ascent can be judged by the closeness of the contour lines. In Figure 4.1, the optimum combination would be to locate outlets in cities with populations of 50,000 people and to build outlets with 70,000 square feet of floor space. This relationship can be estimated by the use of the environmental model and the performance of the present outlets. If the present outlets are located in cities ranging from 5,000 to 100,000 in population and are sized from 40,000 to 100,000 Figure 4.1. A two-factor surface Population 100,000 d 50,000,_ - _- fl $9, 000 $8,000 $7,000 5 000 T I | L l E—IN . re 1 40,000 70,000 100,000 Outlet Size in Square Feet 70 square feet, the true response surface can probably be estimated fairly accurately. That is, if the present outlets are a representative sample of all of the possible combinations, then the estimation of the response surface from the performance of the present outlets should give good predictions of the contributions of the future outlets. Once the expected contributions of the possible future locations and facilities have been estimated, the estimated occupancy costs can be subtracted and the outlets ranked as to profitability. The entire exploration process can be programmed for a computer and the computer could develOp the data showing the most attractive location and facility combinations. It is not reasonable to assume that the environmental model can be relied upon as the only location and facility selection mechanism. How- ever, the model could be used to scan a large number of locations and compute the expected contribution of all facility combinations for these locations. The Rayco Company has used a similar computer pre- diction and scanning system with much success. Some combinations which would likely be unprofitable could be eliminated from consideration by the location and research staff and some quite attractive combinations which may have been overlooked by the staff will be considered and noted. Some relationships considered to be of little importance by the research department may prove to be highly correlated with outlet contribution. The use of the model would replace intuition with objective estimates of expected performance based on the performance of the present outlets. Once the number of possible 3"Can a Computer Tell You Where to Locate Stores?", Chain Store Age, Executive Edition, January, 1961, pp. E27-E28, E38. 71 locations and facility combinations has been reduced, specific estimates of construction costs from local contractors or estimates of rentals from local realtors can be substituted for the index of construction or rental costs from secondary sources. This substitution will improve the estimate of the expected net benefits from the new outlet and possi- bly help to further decide the desirable course of action. When the present outlets are not representative of the possible combinations, the expected contributions of outlets with factor combi- nations far from the levels tested may be very much in error. For example, assume that the present outlets of a firm are at combinations of factors near point A in Figure 4.2. From this sample of all possible X1 and X2 combinations, the linear effects of the factors would dominate. Even if the quadratic effects were significant, the effects would be small and the estimates of the quadratic coefficients would likely be very much in error. If the company projected contributions based on the regression co- efficients computed from the outlets near point A, it would locate new outlets at combinations of large values of X1 and X2--say near point B. By so doing, the contributions of the new outlets would be very little different from the contributions of the old outlets. Company profit would likely be reduced if one of the factors was square feet of floor space because of higher occupancy costs of the new larger outlets. While such a dramatic underestimate of contributions may not be likely in practical situations, significant (and costly) errors from extrap- olation are possible. If changes are made in the direction of the steepest ascent of the contribution "hill" as indicated by the model, and if the changes are 72 relatively small so that the relationships expressed by the model still apply (at least approximately), the profit of the company will rise. By repeating the process of estimating the coefficients and locating the new outlets in the direction of steepest ascent, the contribution hill can be climbed and the optimum conditions eventually determined. The path taken would be approximately perpendicular to the equi-contribution contour lines. The path from point C to the optimum combination in three steps is also illustrated in Figure 4.2. Figure 4.2. Determining optimum conditions 73 In some multi-outlet businesses, there may be no unique maximum expected contribution as existed in Figure 4.1 and Figure 4.2. Davies4 presents three examples in which no such maximum exists (see Figure 4.3). The shape of the contribution contour can have significant impli- cations for location and facility selection. For example, in Figure 4.3 (a) there is a strong interaction between X1 and X2. If the factors , are outlet size and median age in the neighborhood, the company could keep outlet contributions constant and reduce occupancy costs by shifting to small outlets in neighborhoods with many older citizens. Figure 4.3 (b) illustrates the desirability of testing more factors than are expected to be significant to be sure that all, possibly sig- nificant factors are investigated--not just the ones which are logically the most important. Suppose that the firm's present outlets are located near point A and the firm considers expanding its operations through the addition of outlets. From a one-way analysis based only on factor X1, it appears that high contributions are associated with large values of X . If the additions are made in that direction, company profits will 1 rise. By considering only factor X however, point B is mistaken for 1’ the optimum point. Point B is an optimum but only for the given value of X I If the original outlets were located near C and only X 'was 2 2 varied, the same point B would have been reached. Only by considering the two factors simultaneously would the true relationship between X1, X2 and outlet contribution be discovered. The accuracy of the predictions of the model may signal the need for considering additional factors. If the variance of the estimates of 4Davies, p. 504. 74 Figure 4.3. Other two-factor surfaces (a) Stationary Ridge (b) Rising Ridge 2 $10 \$8 $9 \\\\\3 (f/’——~\\\\\\\$10 8 9 (c) Minimax 75 expected outlet performance is large, a factor which has not been con- sidered may explain much of the variance of the estimates. The example in Figure 4.1 could be expanded by considering a third factor, say median age in a community. By adding a third dimension for age, the locus of the equi-contribution lines might be represented by the con- centric shells shown in Figure 4.4. The relationships for store size and population are the same as in Figure 4.1: only the third, previously unconsidered factor is added. Other Location and Facility Problems In the test company, all non-managerial data were gathered from the financial, property and personnel records of the firm or from published secondary data. In other applications such data will not be adequate, especially for scanning new sites. The needed data may require first- hand knowledge and observation. For most multi-outlet businesses, several outlets may be located in one community or metropolitan area. In these companies the location of the outlets within the city becomes much more important than was true for the test firm with only one outlet per community. When there are several outlets in one locale, the problem of overlapping trade areas may reduce the efficiency of the outlets. That is, a new outlet may simply take away the customers from the present outlets of the company. Profits of the company will be reduced because combined sales have re- mained approximately constant while occupancy costs and salaries have risen. For such firms, a "proximity to other outlets" factor should be investigated and possibly included in the model. The shopping habits of consumers in the community would be of interest in such an investigation. 76 Figure 4.4. A three-factor Median Age surface -*:—-a\\\50,000 \ 7 \ . . Population 70,000 ..W/ /.. Outlet Size in Square Feet 77 Montgomery Ward & Company has had some success in defining trade areas by tracing automobile license tags observed in shopping centers to the addresses of the owners of the automobiles on the automobile registra- tions.5 Such an analysis could be more informative than a mere spatial analysis because of differences in access roads and traveling habits. Store saturation6 for the type of outlet in question could be in- vestigated by an analysis of competition and a breakdown of retail sales in the area. Not only existing facilities and competition but also planned additions of competitors should be considered. The types of goods being offered by the outlets help determine the number and location of outlets within a community. For outlets offering "convenience" goods (low in unit value, quickly consumed and standard- ized in nature) the primary consideration should be the accessibility of the outlet.7 Proximity to traffic arteries, volume and nature of traffic, and ease of entry and exit for parking may be very important considerations for locating "convenience" outlets. Even the side of the street on which the outlet is located may be important in attracting the segment of traffic important to the outlet. A doughnut shop may find it desirable to locate to the right side of work-bound traffic while a convenience grocery may locate to the right of home-bound 5"Area Research Gives Ward Detailed Basis for Growth," Chain Store Age, Executive Edition, December, 1964, p. E28. 6"Saturation implies a balance between the amount of existing re- tail store facilities and their use (which in turn is a reflection of need)." (William Applebaum and Saul B. Cohen, "Trading Area Networks and Problems of Store Saturation," Journal,g£ Retailing, Winter 1961-62, p. 35). 7John E. Mertes, "A Retail Structural Theory for Site Analysis," Journal 2f Retailing, Summer 1964, p. 19. 78 traffic. The environmental model can be used to measure the importance of these factors. For "comparison" goods (characterized by high unit value and rela- tive durability) fewer outlets may be needed but the location of the few may be quite important. Stores offering the same type of comparison goods are often located together. If an outlet offering the same type of goods is located elsewhere, it may not even be considered by some potential customers or gain such consideration only by offering sizeable discounts from the usual price. Neither situation is desirable unless a large difference in occupancy cost can offset the disadvantage of the location. Again, the use of the environmental might reveal such a re- lationship. Summary The use of the environmental model in location and facility plan- ning is potentially of great benefit. The outlets of a multi-outlet business are a large number of homogeneous units which, individually, are an insignificant part of the whole organization. There are far too many outlets for a detailed subjective analysis of each possible change for each outlet. The environmental model provides an objective method for estimating the effect of numerous possible changes. If present relationships in a company between the location and facility factors and outlet contribution hold for future outlets, good predictions should result. The identification and measurement of the relevant factors and trends in factors may be much more difficult for other applications of the model than it was for the test company. How- ever, the saving of real estate research investigation costs and the 79 improved predictive power of the model may be well worth the effort and expense of the use of the model. CHAPTER V SUMMARY AND CONCLUSIONS In this chapter, the environmental model is summarized and the con- clusions from the testing of the model presented. The limitations of the study and areas for further research are indicated. Summary 2: the Environmental Model Any measure of performance must be compared with a standard or norm of performance before the measure is meaningful. In accounting, there are three types of standards: (1) interperiod standards, (2) intraperiod standards, and (3) predetermined standards. Standards for evaluating the performance of outlet managers in companies with numerous outlets are difficult to formulate because of the differences in the profit potentials of the various outlets. Dif- ferences in measured performance among outlets may be explained by dif- ferences in the locations and facilities of the outlets as well as the differences in the performance of the outlet managers. Since the lo- cation and facilities of an outlet are not controllable by the outlet manager, the effects of these "non-managerial" factors should be ex- tracted before evaluating the performance of the outlet manager. Interperiod standards have limited applicability and are of very little use in multi-outlet organizations. The use of predetermined standards is difficult because of the large number of outlets. 80 81 Predetermined standards set by central corporate officials cannot re- flect the differences in profit potentials of the outlets and standards set by district supervisors don't allow a valid comparison of managers among districts. The environmental model is an objective method for estimating and extracting the effects of the non-managerial factors causing significant differences in the measured performance of the outlets. In applying the method, a linear statistical model of the performance of the outlets is constructed using accounting data on outlet sales and controllable ex- penses and data on the levels of the non-managerial factors under which the outlets are operated. The effects of significant non-managerial factors are removed from the outlet performance to provide a measure of performance which is more related to the actions of the outlet manager. By removing the effects of the non-managerial factors from the performance measure of the outlet manager, the use of the environmental model makes the measures comparable among the managers. The ranking of estimated managerial contributions forms a valid intraperiod standard. The model utilizes responsibility accounting performance data al- ready gathered in the usual course of business and relates this data to the levels of the environmental factors under which the outlets are operated. Through the separation of the effects of the three broad factors, the relative importance of the manager and the location and facility factors can be determined. If managerial differences are important in determining outlet contribution, much effort and expense should be de- voted to manager selection and training. If the contribution of the outlets depends primarily on the particular location and facilities of 82 the outlet, then relatively more resources should be devoted to location and facility selection with less emphasis on the selection of managers. In a broader sense, the environmental model can be used to maximize the profit of the multi-outlet business by examining interrelationships among the three broad factors explaining differences in performance among outlets. The specific non-managerial factors and relationships among the factors are of interest in planning outlet locations and facilities. It is not feasible to rely on only the model to make lo- cation and facility decisions, but the model can be used to narrow the scope of detailed subjective investigations. The exploratory use of the model for investigating numerous possible locations and various combi- nations of facilities may point up attractive opportunities which would be overlooked by subjective research. Furthermore, the model could help quantify certain concepts and relationships which have been discussed in only qualitative terms. The environmental model has been formulated and tested in a national firm offering a wide range of goods and services through catalogs and retail stores. The study was limited to the catalog sales division of the company. The factor structure for the firm was a relatively simple one; complications will undoubtedly arise in other applications of the model. In the test company there were non-managerial factors which ex- plained significant variation in the transforms of net sales and con- trollable expenses of the outlets. The location and facility differ- ences among the outlets were of much greater importance in determining outlet contribution than were the differences among the outlet managers. 83 The environmental model approach to performance measurement is applicable to only a limited number of businesses. There must be a large number of relatively homogeneous outlets. The large number of outlets is necessary to ”average out” chance fluctuations to generate the overall effects of the factors. With only a few heterogeneous out- lets, considerable confounding of factor effects will result because of too few observations to be able to separate the effects of all of the factors. Fortunately the limitations of predetermined standards are less critical when used in firms with a few outlets. Potential Advantages and Disadvantages f the Model The environmental model for performance measurement can have the following advantages as a tool for evaluating managerial performance in multi-outlet businesses. Relevance: The environments in which outlets are operated do not have equal potentials for profits. It is the effect of the actions of the outlet manager and not the effect of the combination of lo- cation, facilities and the manager that is needed in evaluating the performance of an outlet manager. The environmental model pro- vides a relevant measure of performance and a valid intraperiod standard with which to judge the measured performance of a manager. Objectivity: In a responsibility accounting system, arbitrary allocations of certain common costs are eliminated from the perform- ance measure of outlets. In the environmental model approach, the effects of all significant non-managerial influences are objec- tively estimated and removed from the measure of performance of the outlet manager. '84 Implementability: The environmental model is in agreement with current responsibility accounting concepts and can be implemented with presently generated accounting data and certain other infor- mation concerning the levels of the environmental factors under which outlets are operated. One limitation of the use of the environmental model could be the lack of acceptance of the model results as a valid measure of perform- ance by the outlet managers. For example, it may be difficult to con- vince a manager whose outlet may have the highest contribution in all districts, that his performance is not satisfactory considering the po- tential of the outlet as estimated by the model. Such a dramatic change in ranking is possible if managers are assigned to outlets with dif- ferent potentials. Some of the better managers will be assigned to the poorer outlets and some of the poorer managers will be assigned to the better outlets. This is an education problem, however, and should be overcome if the value of the environmental model approach can be demon- strated. Most managers will agree that the profit potentials of the outlets are not equal and the environmental model is an objective method of making adjustments for the effects of classifiable differences in potentials. Other possible disadvantages of the model are the difficulty in identifying and measuring the relevant non-managerial factors and the difficulty of classifying the outlets according to these factors. Addi- tional empirical testing will show whether these practical problems can be overcome. 85 Limitations 2f the Study In the tests conducted in this study, an environmental model was constructed and the results of the model analyzed. The model was not used in the day-to-day operations of the test firm. Thus, the behav- ioral effects of the model as a motivating and control device have not been investigated. This problem may have considerable importance and should be investigated on an individual firm basis before installing the environmental model in the accounting system. As stated earlier, the test company was selected for its character- istics which simplify the formulation and testing of the model. Most other multi-outlet businesses will have a more complicated non-managerial factor structure. There are no problems in theory in extending the en- vironmental model approach to companies with a complex factor structure. Another limitation of the study concerns the predictive power of the environmental model over time. For planning new outlets, it is future relationships and not current relationships that are of interest. Populations, products and buying habits change over time. Store 10- cations and facilities mature and decline. The usefulness of the en- vironmental model for location and facility selection may be in esti- mating trends in relationships and not the relationships at any par- ticular point in time. The inputs for the environmental model are past performance data and thus if the type of location or facility being con- sidered is substantially different from present and past outlets the potential of the new locations and facilities cannot be accurately esti- mated by the model. 86 Areas for Additional Research The areas for additional research are related to the limitations of the present study. The behavioral implications of all standards used in accounting is an area in need of much research. Adverse reaction of subordinates may completely outweigh the advantages of a standard. Numerous cases of practical problems with standards and motivation could be cited. Most of these cases involve production or time standards. The use of standards in marketing (and particularly standards for sales) is a much more recent development in accounting and very little is known about the effects of such standards on marketing personnel. An investigation of the effects of the environmental model on outlet managers may well provide insight into the broader area of marketing standards. The environmental model as a test mechanism has the advantage of being objective and can even be used to test the effects of factors which the outlet managers themselves consider to be causes of differences in profit potentials. The application of the environmental model to companies with more complicated factor structures will present practical problems. However, the complex firm is the one in which mathematical tools like the en- vironmental model are most needed. In a small firm, the owner can direct the operations of the entire firm by personal observation and subjective standards of performance. In a large firm, the number of outlets alone makes a more structured evaluation system necessary. The adaptation of linear statistical models to complex problems of analysis has been taking place for several decades in the natural sciences and more recently in the social sciences. There are many 87 statistical tools now available for use in constructing an environmental model for a very complex multi-outlet business. Such applications should be attempted. The environmental model has been tested over only one three-year period. The time period was relatively short and no attempt was made to extract trends in relationships over time. The use of the environ- mental model to extract trends in location and facility relationships should be attempted and then followed by an analysis of the predictive power of the model. Conclusion The environmental model offers many potential benefits to the multi- outlet business with a small addition to the costs of the operation of a responsibility accounting system. Theoretically, the model offers a solution to the problem of objective measurement of the performance of outlet managers and provides guides to resource allocation in selecting managers, locations and facilities. The empirical testing of the model indicates that the effects of the non-managerial factors are significant and can be much more impor- tant than the manager in determining outlet contribution to unallocable costs and profits. It has been shown that the rankings of managerial performance based on the environmental model can be grossly different from the rankings of outlet performance. If the assumptions of the en- vironmental model are valid, the test results indicate that intraperiod standards based on outlet contributions may be a very poor approximation to rankings based upon the estimated effects of managerial actions. Poor managers may be given preference over excellent managers if the 88 outlet contributions are taken as the measure of managerial performance. The error is due to the confounding of the effects of the location and facility factors with the effects of the manager's actions. Generalization to other multi-outlet businesses is hazardous. How- ever, on the basis of the test results, the environmental model seems to offer much insight into the profitability relationships in the multi- outlet business. SELECTED BIBLIOGRAPHY Anthony, Robert N. Management Accounting, Robert D. Irwin, Inc., Homewood, Illinois, Third Ed., 1964. Applebaum, William. "Store Location Research - A Survey by Retailing Chains," Journal 9; Retailing, Vol. 40, Summer, 1964, pp. 53-56. Applebaum, William. "Store Performance in Relation to Location and Other Characteristics," Chain Store Age: Executive Edition, vol. 41, November, 1965, pp. El4-El7. Applebaum, William, and Saul B. Cohen. "Trading Area Networks and Problems of Store Saturation," Journal 2f Retailing, Vol. 37, Winter 1961-62, pp. 35-43. ”Area Research Gives Ward Detailed Basis for Growth," Chain Store Age: Executive Edition, Vbl. 40, December, 1964, pp. E28-E3l. Bartlett, M. S. "Some Examples of Statistical Methods of Research in Agriculture and Applied Biology," Journal 9f Royal Statistical Society, (Supplement), Vol. 4, 1937, p. 137. Cohen, Saul B. "Facing Today's Store Location Challenges," Chain Store Age: Executive Edition, Vol. 39, November, 1963, pp. E23, E27- E28, E38. . "Can a Computer Tell You Where to Locate Stores?" Chain Store Age: Executive Edition, Vol. 37, January, 1961, pp. E27- E30. Dalrymple, Douglas J. Measuring Merchandising Performances $3 Depar - ment Stores, Retail Research Institute, New York, New York, 1964. Davies, 0. L. (Ed.). The Design and Analysis gf Industrial Experiments, Hafner Publishing Company, New York, New York, 1956, pp. 495-551. Durbin, J., and G. S. Watson. "Testing for Serial Correlation in Least Squares Regression," Biometrika, vol. 37, 1950, pp. 409-428, and Vol. 38, 1951, pp. 159-178. Fraser, Donald A. S. Statistics: ‘An Introduction, John Wiley & Sons, New York, New YCrk, 1958. Freund, John E. Mathematical Statistics, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1962. 89 90 Graybill, Franklin A. 'An Introduction to Linear Statistical Models, Vol. I, McGraw-Hill Book Company, Inc., New York, New York, 1961. Horngren, Charles T. Accounting for Management Control: .52 Introduc- tion, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1965. Huff, David L. "A Programmed Solution for Approximating an Optimum Retail Location," Land Economics, Vol. 42, August, 1962, pp. 293- 303. Ijiri, Yuji and Robert K. Jaedicke. ''Reliability and Objectivity of Accounting Measurements," Accounting Review, Vol. 41, July 1966, pp. 474-483. Kempthorne, Oscar. The Design and Analysis of Experiments, John Wiley & Sons, Inc., New York, New York, 1960. La Londe, Bernard J. "Differentials in Supermarket Drawing Power and Per Capita Sales by Store Complex and Store Size," unpublished Ph.D. dissertation, Michigan State University, 1961. Massey, F. J., Jr. ”The Kolmogorov-Smirnov Test for Goodness of Fit," Journal of the American Statistical Association, V61. 46, 1951, pp. 68-780 Mertes, John E. "A Retail Structural Theory for Site Analysis," Journal gthetailing, Summer 1964, pp. 19-24. Ostle, Bernard. Statistics in Research, The Iowa State University Press, Ames, Iowa, 1963, p. 340. . Performance Appraisal and Review, The Foundation for Research on Human Behavior, Ann Arbor, Michigan, 1958. Ritland, Ross W. "Intra-Chain-Store Competition," Journal of Retailing, Summer 1962, pp. 15-21. Steinberg, Martin D. "Predicting Dealer Success," Journal'gf Marketing, Vol. 26, April 1962, pp. 75-76. APPENDIX A AN ILLUSTRATIVE COMPUTATION OF ESTIMATED MANAGERIAL CONTRIBUTION The purpose of this appendix is to illustrate the computation of the estimated contribution of the manager to the measured performance of an outlet and the estimated percentile standing of the manager's per- formance. The computation of the estimated contribution and percentile standing of the manager of outlet number one of the test company to the logarithm of sales for 1966 will be illustrated. Outlet number one issued 2,000 catalogs during 1966. The outlet was first opened in the community in 1938 and last remodeled in 1947. The community in which the outlet is located has a median age of 33 years, a median family income of $5,500, and a population of 13,100. The third secondary competitor operates an outlet in the community. The seasonally adjusted net sales by quarter for 1966 for the out- let were: $8l,423, $73,712, $78,946 and $71,225. The logarithms of was net sales were: 4.910747, 4.867538, 4.897330 and 4.852632. 3'11 thus equal to 4.882061. Table A.1 illustrates the computation of the non-managerial factor effects (outlet one is located in district one and the district one effect for 1966 was - .100226). Substituting into the formula for the estimate of managerial con- tribution, it is seen that M is equal to .094930. 1 91 92 Table A.1. Computation of the non-managerial factor effects for outlet number one for 1966 Factor Parameter x Factor = Estimated Estimatea Level Factor Effect A A (Bi) (x11) (Bixip Number of catalogs issued .010499 20 .209980 Number of catalogs issued (quadratic effect) '-.000001 400 ‘3000400 Years since opening .010837 28 .303436 Median age '-.006972 33 -3230076 Median family income .002983 55 .164065 District 1 ‘—.100216 1 '-.100216 District 2 xxx 0 0 District 10 xxx 0 0 Populationb X Secondary Competitor #3 '-.000144 131 ‘-.018864 Years since remodeling .016374 19 .311106 Years since remodeling (quadratic effect) ‘-.000423 361 ‘-.152703 Years since opening (quadratic effect) ‘—.000199 784 ‘-.156016 Populationb X median family incomeb .000002 7205 .014410 20 Z 131xi1 .344712 8See Appendix Table B.3. b . Factor level measured in one-hundreds. fi -71' — A+§20BX) 1 ‘ 1 ( i=1 1 11 4.882061 - (4.442419a + .344712) .094930. As was mentioned on page 26, the estimated managerial contributions for the test company were approximately normally distributed. The esti- mated standard deviation of the managerial contribution estimates for 1966 was .083246. The distribution of estimated managerial contributions was converted to a standard normal distribution by dividing each contri- bution by the standard deviation. The manager of outlet one had a "standardized" estimated contri- bution of l.l40--(.094930/.083246)--that is, the manager of outlet one was performing at a level 1.140 standard deviations above the average for all managers. By referring to a cumulative standard normal table, it is seen that 1.140 is at the 87th percentile in a standard normal distribution. Thus, it is estimated that the manager of outlet number one per- formed at a level above that of 87 percent of his fellow managers for 1966. 8See Appendix Table B.3. APPENDIX B PARAMETER ESTIMATES - 1964, 1965, 1966 94 95 Table B.l. Parameter estimatesa 1964 Logarithm of Net Outlet Sales Factor Estimate Regression constant b 6.028347 Number of catalogs issued .003656 Years since remodeling .017253 Years since remodeling (quadratic effect) '- .000609 Median age -’ .008571 Median age (quadratic effect) .001260 Secondary competitor #3 - .336354 Years since opening .001903 Median family income .004326 Median years of schooling b .004962 Shopping center X population .000093 Median years of schooling (quadratic effect) .000237 Square Root 2f Controllable Expenses Factor Estimate Regression constant 53.307352 Sales .026709 Distance from distribution center 4.420873 aNon-managerial factors are listed in the order in which they were added to the equations in the stepwise addition of variables. bFactor level measured in one-hundreds. 96 Table B.2. Parameter estimatesa 1965 Logarithmlgf Net Outlet Sales Factor Estimate Regression constant b 4.462961 Number of catalogs issued .004280 Years since remodeling .014347 Years since remodeling (quadratic effect) ‘— .000524 Numbers of catalogs issued (quadratic effect) ‘- .000001 Years since opening .010399 Years since opening (quadratic effect) -' .000208 Secondary competitor #3 ‘- .340534 Median family incomeb .003801 Median years of schooling .004325 Median years of schooling (quadratic effect) .000143 Square Root 2: Controllable Expenses Factor Estimate Regression constant 55.222405 Sales .039237 Managerial tenure - 1.532677 aNon-managerial factors are listed in the order in which they were added to the equations in the stepwise addition of variables. bFactor level measured in one-hundreds. 97 Table B.3. Parameter estimatesa 1966 Logarithm pf Net Outlet Sales Factor Estimate Regression constant b 4.442419 Number of catalogs issued .010499 Number of catalogs issued (quadratic effect) '- .000001 Years since opening .010837 Median age b ‘- .006972 Median family income .002983 Population X secondary competitor #3 '- .000144 Years since remodeling .016374 Years since remodeling (quadratic effect) '- .000423 Years since opening (quadratic effect) '- .000199 Populationb-X median family incomeb .000002 Square Root‘pf Controllable Expenses Factor Estimate Regression constant 58.899033 Sales .053328 Managerial tenure -'l.718602 aNon-managerial factors are listed in the order in which they were added to the equations in the stepwise addition of variables. bFactor level measured in one-hundreds.