RI LINE: LESALE D " '_' VER B. MONTY,- " CHANNELS Eh gree'b‘t NT WHO HA ALYSIS 0F MERC w,_~_- -wv. O 406 OD COM I ER G0 I sis for the DIST-R - -—- “'14--..I‘. The CONSUM N A N A . 94...... Est-”hréfi: . , d .. 3:3“. .. graffiti A... .Wnanm. a3 uni. . , 3.. a. . . .... hfiofll‘. J 33.,“ 4"" r. 32‘ - 5 2'3" {3"}. . J... o< . .- . «may ‘ . i b . r; o i .........,.H% i4 u If .- ,3" . #1:?” a: . .v .< , . » Z... . 7 c . Wmfiwflimué e _ .3... r“; I. h/lv. ”1... .n.’ IAN»... ...,..,.......,....... iii}. .2”. 3.x 2., ”3...“. ... w....y.....u.._x,...u_%fi ... at.» .jmfiw P. an” ' v . . 7.4-) a? ”W “7.... “WM”?- . as. . . I. we... ....5. r... am 0 “.3...“ . u... if... . a”? .2.’ :WW .1 ”5.3%“... , ... . . . 1/ I r .J ._ . . a 7.. w. ‘13:.74 .¢._~.Wwfluflmn.u_.fimfivm . . .. . . .. ., 44¢ .....u.......vu...% . (I. u VIM-.5 . 1 Io. v3.2... . . EOEGE MICHIGANSTATEUNI C , . , I . . . A .. o , . . an“.v(fl....t .9 i .o I. r .a..lv,....zv..ov.-.?p Viz...» . 90V; we 4 . .u .. ... FEEDER] . ... C”... .x. 1%.. ...i...x.... . . y 11...... 35...... .4 I; V... . . I k"... 5.5m.) "l JCfi-‘PA" -' IIIIIIIIIIIIIIIIIIIIIII IIIIIIIIIIII 31293 0100 This is to certify that the thesis entitled AN ANALYSIS OF MERCHANT WHOLESALER CONSUMER GOOD COMMODITY LINE DISTRIBUTION CHANNELS presented by Frederick George Bean has been accepted towards fulfillment of the requirements for Ph.D. degreein Marketing & Transportation Administration Date 0-7639 LIBRARY Midnigan Sm; University OCT 06 2001 _ ABSTRACT AN ANALYSIS OF MERCHANT WHOLESALER CONSUMER GOOD COMMODITY LINE DISTRIBUTION CHANNELS BY Frederick George Bean Statement of Purpose There were two purposes for this study. The first was to identify economic and distribution channel factors that are associated with certain characteristics of the merchant wholesaling sector of consumer good commodity line distribution channels. With knowledge of these factors, the individual merchant wholesaler might be better able to adjust his activities and services to meet the present and future needs of his suppliers and customers. The second purpose of this study was to analyze the application of techniques proposed to provide more efficient estimates of regression coefficients. The analysis of these techniques may aid in the identification of situations where they should or should not be applied. Frederick George Bean Procedure Four consumer good commodity line distribution channels were selected for analysis. These were the all commodity or aggregate channel and the drug, liquor, and lumber commodity line channels. The period 1948 to 1967 was studied using United States Bureau of Census data. The characteristics of the merchant wholesaling sector (de— pendent variables) selected were: sales per capita, establishments per capita, and sales per establishment. The economic and distribution channel factors (independent variables) selected were: personal income per capita, mean nonagricultural employment as a percentage of total em- ployment, pOpulation per square mile; all commodity retailers sales per capita, establishments per capita, sales per establishment; all commodity manufacturers value added per capita, establishments per capita, value added per establishment; all commodity merchant wholesalers sales per capita, establishments per capita, sales per establishment; commodity line retailers sales per capita, establishments per capita, sales per establishment; commodity line manufacturers value added per capita, establishments per capita, and value added per establish- ment. To provide the most efficient estimates of the regression coefficients computational techniques employing multiple sets of data were used. The results computed by \- .h mi» ‘- I . ~dI Frederick George Bean Van Tassell using the Efficient Estimator Fortran Program (EFFEST) were compared with the estimates in this study calculated using the Zellner-Aitken (ZA) and least- squares (LS) techniques. Findings In each channel, significant associations were found between combinations of merchant wholesaling sector characteristics and the economic and channel factors studied. These combinations varied from channel to channel. For example, there was a significant associ— ation between per capita drug merchant wholesaling sales and the independent variable, per capita all commodity merchant wholesaling sales. The association between the per capita liquor merchant wholesaling sales and the same independent variable, on the other hand, was only mar- ginally significant. The association between the per capita lumber merchant wholesaling sales and the same independent variable was not significant. These differ- ences in the significance of factors from channel to channel reflect the varied character of merchant whole- saling channels in the United States. Additionally, .shifts of regression coefficients between the dependent arui independent variables were found. Although not all were statistically significant over the period of the stmniy shifts were found in every channel. These findings ..n - o»' in- . bud 0.. I" v. 1..“ -¢. ll] ) .l fin. Frederick George Bean reveal the need for the merchant wholesalers in every channel to analyze their unique, dynamic situation. The regression coefficients by Van Tassel using EFFEST and those calculated in this study using ZA and LS were compared. It was concluded that the conditions neces- sary for applying EFFEST and ZA were not met. In addition, the regression coefficients calculated using these esti- mators resulted in biased estimates. Since the least- squares does not have these shortcomings, it was employed in this study. 1Charles E. Van Tassel, An Analysis of Factors Jurfluencing Retail Sales (East Lansing: Bureau of Business arui Economic Research, Michigan State University, 1966). AN ANALYSIS OF MERCHANT WHOLESALER CONSUMER GOOD COMMODITY LINE DISTRIBUTION CHANNELS BY Frederick George Bean A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Marketing and Transportation Administration 1971 (:7Copyright by FREDERICK GEORGE BEAN 1971 ACKNOWLEDGMENTS Sincere appreciation is extended to Dr. Richard J. Lewis for his interest, encouragement, and suggestions. He very generously fulfilled his role as chairman of my thesis committee. I would also like to thank Dr. Leo G. Erickson and Dr. Carl E. Liedholm for their assistance as committee members. Both aided this study by contributing valuable guidance. Most of all, I would like to acknowledge the immeasurable aid of my wife, Nancy. Typing many drafts of this study was the most tangible form of her aid. But more important was the support and love she provided dur- ing the long months of this study. I would also like to thank my parents and friends who provided encouragement. ii TABLE OF CONTENTS Chapter Page I. INTRODUCTION . . . . . . . . . . . 1 Statement of Purpose. . . . . 1 Terms and Definitions of Wholesaling . . 3 WhOIesaling O O O O O O O O O O 4 Wholesale Transaction. . . . . . . 4 Wholesaling Middleman. . . . . . . 5 Wholesaler . . . . . . . . 5 Wholesale Establishment . . . . 6 Types of Wholesale Establishments. . . 7 Operating Establishment . . . . . . 8 The Background of the Problem. . . . . 9 The Literature . . . . . . . . . 11 The Changing Economic Environment. . . 12 Changes Among the Types of Wholesale Operations. . . . 14 The Selection of the Merchant Whole- saler for Study . . . . . . . . 19 Selection of Data Sources . . . . . 21 Variable Selection . . . . . . . . 22 Characteristics (Dependent Variables) Selected to Analyze the Merchant Wholesaling Sector . . . . 22 Selection of Independent Variables . . 22 Problem Statement. . . . . . . . . 27 General Hypotheses . . . . . . . . 27 Limitations. . . . . . . . 28 Possible Contributions of Study . . . . 30 iii Chapter II. RESEARCH METHODOLOGY . . . . . . . . Measurement Units of Characteristics Used Selected for Study. . . . . . Geographic Control Units . . . . . Selection of Merchant Wholesaler Channels. . . . . . . . . . . Multiple Regression. . . . . . . . Zellner-Aitken Estimators. . . . . . Statistics Derived . . . . . . . . The Hypotheses Tested . . . . . . . Regression Coefficients. . . . . . Change in Regression Coefficients . . Significance of Findings in This Study . Regression Coefficient (B). . . . . Multiple Correlation Coefficients (r YXlX2 Changes in Regression Coefficients (1948-1967) 0 o o o o o o o 0 III. THE COMPARISON OF THE METHODOLOGY OF THIS STUDY AND THE VAN TASSEL RETAIL STUDY . . Introduction . . . . . . . . . . The Data Used. . . . . . . . . . Computational Techniques . . . Results of the Van Tassel Study and This Study for the Drug Retail Sector. . The Results of Comparing ZA to EFFEST Estimates . . . . . . . . . . Regression Coefficients (B1 and B 2). Standard Error of the Regression Coefficients (o and CB ) . . . . Bl 2 t Values, A/OAl'Bl/OBI'BZ/OBZ' . . . Conclusion on Comparison of EFFEST and ZA 0 O O C O O I O O O 0 iv Page 31 31 33 33 36 38 42 44 44 44 45 45 48 51 54 54 54 55 55 57 57 62 66 68 Chapter IV. FINDINGS: THE ALL COMMODITY MERCHANT WHOLESALER CHANNEL . . . . . . . . Merchant Wholesaler Characteristics studied. 0 O O O O 0 Characteristics of the Other Channel Sectors. . . . . . . . . . . Selected Economic Characteristics. . . Mean Per Capita Sales or Value Added and Establishments . . . Results of the All Commodity Channel Study 0 O O O O O O O I 0 Regression Coefficients (B) . . . . Multiple Correlation Coefficients. . . Changes in the Regression Coefficient. V. FINDINGS: THE DRUG CHANNEL . . . . . Drug Commodity Line Merchant Wholesaler Characteristics . . . . . . . . Drug Commodity Line Manufacturing Establishment Characteristics . . . Drug Commodity Line Retail Establish— ments' Characteristics. . . . . Mean Per Capita Drug Establishments' Sales or Value Added and Establishments Results of the Drug Channel Study. . . Regression Coefficients (B) . . . . Multiple Correlation Coefficients . Changes in the Regression Coefficients VI. FINDINGS: THE LIQUOR CHANNEL. . . . . An Analysis of the Effects of the Estimators on the Regression Coefficients . . . . . . . . . The Three-Dimensional Case . . . . . Three Cases of Efficient Estimators . . Testing the Significance of the Findings Conclusion . . . . . . . . . . Liquor Commodity Line Merchant Wholesaler Characteristics . . . . Liquor Commodity Line Manufacturing Sector Characteristics. . . . . . Page 68 68 69 77 82 84 84 85 87 87 88 88 94 97 103 103 106 106 107 108 108 115 115 120 120 123 at V... 0-,. ‘ I... Chapter Page Liquor Commodity Line Retail Establish- ments' Characteristics. . . . . 123 Mean Per Capita Liquor Establishments' Sales or Value Added and Establishments . 125 Results of the Liquor Channel Study . . . 125 Regression Coefficients (B) . . . . . 125 Multiple Correlation Coefficients . . . 130 Changes in the Regression Coefficients . 132 VII. FINDINGS: THE LUMBER CHANNEL. . . . . . 137 ' Lumber Commodity Line Merchant Wholesaler Characteristics . . . . . 137 Lumber Commodity Line Manufacturing Sector Characteristics. . . . 138 Lumber Commodity Line Retail Establish- ments' Characteristics. . . . 140 Mean Per Capita Lumber Establishments' Sales or Value Added and Establishments . 142 Results of the Lumber Channel Study . . . 142 Regression Coefficients (B) . . . . . 142 Multiple Correlation Coefficients . . . 148 Changes in the Regression Coefficient. . 148 VIII. SUMMARY, CONCLUSIONS, AND IMPLICATIONS. . . 155 Summary . . . . . . . . . . . . 155 Conclusions and Implication. . . . . . 156 Association of the Merchant Wholesaler Characteristics with Selected Independent Variables . . . . . . 156 Changes in Regression Coefficients Between the Dependent and Independent Variables . . . . . . . . . 157 Questions Concerning Selected Independent Variables . . . . 159 The Relative Changes in the Channel Sector Characteristics During the Period 1948- 1967 . . . . . 164 Possible Implications of the Preceding Findings. . . . . . . . . . 166 EFFEST and ZA Estimators . . . . . . 167 Geographic Control Units . . . . . . 168 Vi Chapter Page Suggested Areas for Future Research . . . 168 Academic . . . . . . . . . . . 168 Applied. . . . . . . . . . . . 169 SELECTED LITERATURE . . . . . . . . . . . 170 vii Table 1.4 1.7 LIST OF TABLES Aggregated Data for the Manufacturing, Wholesaling, and Retailing Sectors of the United States Economy in 1967 . . Number of Publications Listed in the Cumu- lative Book Index for the TOpics of Production, Wholesaling, and Retailing for the Period January 1959 to July 1969 . Level of National Income Origination in Selected Sectors of the United States Economy as Per cent of National Income (N.I.) in 1948 and 1967 and Per cent Change in Sector Share of N.I. . . . Sales and Percentage Total Sales of Whole- salers by Type of Operation in 1948 and 1967. . . . . . . . . . . . Establishments and Percentage of Total Whole- sale Establishments by Type of Operation in 1948 and 1967 and Per cent Increase in Number of Establishments by Type of Oper- ation from 1948 to 1967 . . . . . Sales per Establishment and Percentage Increase by Type of Wholesale Operation Between 1948 and 1967 . . . . . . Percentage Distribution of Sales by Types Wholesaler for Wholesale Establishments Reporting Class-of-Customer Information 1948 and 1963. . . . . . . . . of for Combinations of Single Dependent and Multiple Independent Variables Studied for Associ- ation in Terms of Multiple Correlation Coefficients . . . . . . . . . viii Page 10 ll 14 16 17 18 20 49 Table 3.1 3.4 4.3 The Values of the "Y-Axis" Intercept Regres— sion Coefficients, Standard Errors of the Estimates, and "t" Value Estimates for the Drug Store Retail Sales per Capita Using the EFFEST, LS, and ZA Programs for the Years 1948, 1954, 1958, 1963 (and 1967) . . Estimate of the Standard Deviation of the Regression Coefficients Using the Estimated Values of the Standard Errors of the Regres— sion Coefficients Calculated Using the LS Method and EFFEST and ZA Estimators for the Drug Retail Establishments' Sales per Capita for the Census Years 1948, 1954, 1958, 1963 (and 1967). . . . . . . . . . . . Range of Values for the Dependent and Inde- pendent Variables for the Individual Census Years and Total Census Period 1948-1967 . . The Coefficient of Correlation (r) Between the Independent Variable All Commodity per Capita Retail Sales and Dependent Variable Drug Merchant Wholesaler per Capita Sales for Least-Squares (LS) and Zellner—Aitken for the Census Years 1948, 1954, 1958, 1963, and 1967 Critical Values of Single and Multiple Coef- ficients of Determination (r2) and Coef- ficients of Correlation (r) at the .05 Level of Significance for Selected Sample Sizes and Numbers of Independent Variables . . Characteristics (and Index Numbers) of the All Commodity Merchant Wholesaler Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . Selected Characteristics (and Index Numbers) of the All Commodity Manufacturing Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967 . . . . Selected Characteristics (and Index Numbers) of the A11 Commodity Retail Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . ix Page 58 65 67 76 83 85 86 87 Table 4.4 Page Aggregate Selected Economic Characteristics (and Index Numbers) of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 88 Per Capita Sales or Value Added (and Index Numbers) for Manufacturing, Merchant Wholesaling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . . . . 89 Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population . 89 Correlation Coefficients (r) Between Mean per Capita All Commodity Merchant Whole- saler Sales and Selected Independent Vari- ables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 91 Correlation Coefficients (r) Between Mean per Capita All Commodity Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 93 Correlation Coefficients (r) Between Mean Sales per Merchant Wholesaling Establish- ment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . . . . 95 Multiple Independent Variable Combinations that Provided Correlation Coefficients Significantly Superior to Each of the Inde- pendent Variables Considered Separately for the All Commodity Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . 96 Estimated Regression Coefficients (B) and Standard Error of Coefficients (EB) Between Mean per Capita A11 Commodity Merchant Wholesaler Sales and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 99 Table 4.12 5.5 5.6 Estimated Regression Coefficients 1B) and Standard Error of Coefficients (03) Between Mean per Capita All Commodity Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . Estimated Regression Coefficients AB) and Standard Error of Coefficients (03) Between Mean Sales per A11 Commodity Merchant Wholesaling Establishmand and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . Selected Characteristics (and Index Numbers) of the Drug Merchant Wholesaler Sector of the Drug Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . The Index Numbers of Selected Characteristics of the Manufacturing, Merchant Wholesaling and Retailing Sectors of the All Commodity and Drug Merchant Wholesaler Distribution Channels for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Selected Characteristics (and Index Numbers) of the Drug Manufacturing Sector of the Drug Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . Selected Characteristics (and Index Numbers) of the Drug Retailing Sector of the Drug Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . Per Capita Sales or Value Added (and Index Number) for Drug Manufacturing, Merchant Wholesaling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . . . . Drug Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population xi Page 100 102 104 105 107 108 109 109 Table 5.11 5.12 6.1 6.2 Page Correlation Coefficients (r) Between Mean per Capita Drug Merchant Wholesaling Establishment Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . 111 Correlation Coefficients (r) Between Mean per Capita Drug Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 113 Correlation Coefficients (r) Between Mean Sales per Drug Merchant Wholesaler Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 114 Estimated Regression Coefficients (B) and Standard Error of Coefficients (03) Between Mean per Capita Drug Merchant Wholesaling Establishment Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . 117 Estimated Regression Coefficients (B) and Standard Error of Coefficients (c ) Between Mean per Capita Drug Merchant WhoIesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 117 Estimated Regression Coefficients (B) and Standard Error of Coefficients (03) Between Mean Sales per Drug Merchant Wholesaling Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 119 Selected Characteristics (and Index Numbers) of the Liquor Merchant Wholesaler Sector of The Liquor Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . 121 The Index Numbers of Selected Characteristics of the Manufacturing, Merchant Wholesaling, and Retailing Sectors of the A11 Commodity and Liquor Merchant Wholesaler Distribution Channels for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 122 xii Table 6.3 6.4 6.7 6.9 6.10 Page Selected Characteristics (and Index Numbers) of the Liquor Manufacturing Sector of the Liquor Distribution Chennel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . 124 Selected Characteristics (and Index Numbers) of the Liquor Retailing Sector of the Liquor Distribution Chennel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . 124 Per Capita Sales or Value Added (and Index Number) for Liquor Manufacturing, Merchant Wholesaling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . . . . 126 Liquor Manufacturing, Merchant Wholesaling and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population . 126 Correlation Coefficients (r) Between Mean per Capita Liquor Merchant Wholesaling Establishment Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . 127 Correlation Coefficients (r) Between Mean per Capita Liquor Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 129 Correlation Coefficients (r) Between Mean Sales per Liquor Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . 131 Estimated Regression Coefficients (B) and Standard Error of Coefficients (03) Between Mean per Capita Liquor Merchant Wholesaling Establishments' Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . 133 xiii Table 6.11 6.12 7.1 7.3 7.4 7.5 7.6 Estimated Regression Coefficients (B) and Standard Error of Coefficients (OB) Between Mean per Capita Liquor Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Estimated Regression Coefficients (B) and Standard Error of Coefficients (03) Between Mean Sales per Liquor Merchant Wholesaling Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Selected Characteristics (and Index Numbers) of the Lumber Merchant Wholesaler Sector of the Lumber Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . The Index Numbers of Selected Characteristics of the Manufacturing, Merchant Wholesaling, and Retailing Sectors of the A11 Commodity and Lumber Merchant Wholesaler Distribution Channels for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . Selected Characteristics (and Index Numbers) of the Lumber Manufacturing Sector of the Lumber Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . Selected Characteristics (and Index Numbers) of the Lumber Retailing Sector of the Lumber Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . Per Capita Sales or Value Added (and Index Number) for Lumber Manufacturing, Merchant Wholesaling and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . . . . Lumber Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population . xiv Page 135 135 138 139 140 141 143 143 Table 7.7 7.12 7.13 Correlation Coefficients (r) Between Mean per Capita Lumber Merchant Wholesaling Establishment Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . Correlation Coefficients (r) Between Mean per Capita Lumber Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Correlation Coefficients (r) Between Mean Sales per Lumber Merchant Wholesaling Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Multiple Independent Variable Combinations that Provided Correlation Coefficients Significantly Superior to Each of the Inde- pendent Variable Correlation Coefficients Considered Separately for the Lumber Channel for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Estimated Regression Coefficients 1B) and Standard Error of Coefficients (OB) Between Mean per Capita Lumber Merchant Wholesaling Establishment Sales and Selected Inde- pendent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . Estimated Regression Coefficients iB) and Standard Error of Coefficients (03) Between Mean per Capita Lumber Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . Estimated Regression Coefficients_(B) and Standard Error of Coefficient (03) Between Mean Sales per Lumber Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967 . . . . . . . . . . XV Page 144 146 147 149 151 151 153 Table 8.1 8.2 8.3 The Significant Associations Between Certain Characteristics of the Merchant Wholesaling Sectors and the Selected Independent Vari- ables of the Commodity Line Channels Studies for the Period 1948-1967 Noted Shifts (and t-values) of the Regression Coefficients Between Certain Characteris- tics of the Merchant Wholesaling Sectors and Selected Independent Variables of the Commodity Lines Studied for the Period 1948-1967. The Index Numbers of Selected Characteris- tics of the Manufacturing, Merchant Whole- saling, and Retailing Sectors of the All Commodity, Drug, Liquor, and Lumber Distribution Channels for the Years 1948, 1954, 1958, 1963, and 1967 xvi Page 158 160 165 Figure 1.1 LIST OF FIGURES Page The Relationships of the Merchant Whole- saler Distribution Channel Sectors for a Consumer Good Commodity Line . . . . 2 The National Income and Amount of National Income Origination in Selected Sectors of the United States Economy for the Census Years 1948 to 1967. . . . . . 13 The Merchant Wholesaler Type of Consumer Good Commodity Line Distribution Channel and Characteristics Selected for Study . 32 Geographic Divisions of the United States . 34 Test for Statistical Significance of the Change in the Regression Coefficient for the Period 1948 to 1967 . . . . . . 53 The Regression Coefficients (B ) Between the Independent Variable Personal Income per Capita and the Dependent Variable Drug Retail Establishments' Sales per Capita Using the Least-Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967) . . . 59 The Regression Coefficient (B2) Between the Independent Variable Mean Per cent Non— agricultural Employment as Per cent of Total Employment and the Dependent Vari- able Drug Retail Establishments' Sales per Capita Using the Least-Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967) . . . 6O xvii Figure 3.3 3.7 3.9 3.10 The Standard Error of the Estimate of the Coefficient (03%) Between the Independent Variable Person 1 Income per Capita and Dependent Variable Drug Retail Establish- ments' Sales per Capita Using the Least- Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963, (and 1967) . . . . . . . . . . The Standard Error of the Estimate of the Regression Coefficient (33 ) Between the Independent Variable Mean er cent Non- agricultural Employment and Dependent Variable Drug Retail Establishments' Sales per Capita Using the Least-Squares (LS) Technique and EFFEST and Zellner- Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967) Non-Scale Representation for the Regres- sion Equation for the Drug Retail Establishments' Sales per Capita for the Census Years 1948, 1958, and 1967 . . Artificial Data to Show How the ZA Esti— mators May Improve the Estimates of the Regression Coefficients for the Indi- vidual Census Periods (1948-1967) . . Regression Equation for the Relationships Between the Independent Variable All Commodity Retail per Capita Sales and Dependent Variable Drug Merchant Whole- saler per Capita Sales Using LS and ZA Estimators for the Individual Census Periods (1948-1967) . . . . . . . Regression Equation for the Relationships Between the Independent Variable Mean Per cent Nonagricultural Employment and Dependent Variable Drug Retail Establish- ments per Capita Using LS and ZA Esti- mators for the Individual Census Periods (1948—1967) 0 o o o o o o o o o Two-Tail Test That Van Tassel Should Have used 0 O O O O O O O O O O xviii One-Tail Test as Hypothesized by Van Tassel. Page 63 64 70 72 74 75 79 80 CHAPTER I INTRODUCTION Statement of Purpose There were two basic purposes for this study. The first was to identify those economic and distribution channel factors that are associated with certain charac- teristics of the merchant wholesaler sector of consumer good commodity line distribution channels. The second purpose was to analyze the application of techniques pro- posed to improve the estimates of regression coefficients using multiple sets of data. The economic factors selected for study were: personal income per capita, mean nonagri- cultural employment as a percentage of total employment, and population per square mile. The distribution channel factors selected were: all commodity retailers sales per capita, establishments per capita, sales per establish- ment; all commodity manufacturers value added per capita, establishments per capita, value added per establishment; all commodity merchant wholesalers sales per capita, establishments per capita, sales per establishment; commodity line retailers sales per capita, establishments per capita, sales per establishment; commodity line manufacturers value added per capita, establishments per capita, and value added per establishment. Merchant wholesalers are wholesalers that take title to the goods they sell to other wholesalers, re- tailers, manufacturers, and various institutions. A simplified view of the merchant wholesaler distribution channel for a consumer good commodity line would have four basic sectors. They are the manufacturing, wholesale, retail, and consumer sectors. The basic relationships of a simplified consumer good commodity line distribution channel sectors are shown in Figure 1.1. It is this type of channel that was the basis for the study. Manufacturing Sector * V Merchant Wholesaler Sector * V Retail Sector * V Consumer *The arrows indicate the flow of ownership. {Figure 1.1. The Relationship of the Merchant Wholesaler Distribution Channel Sectors for a Consumer Good Commodity Line The individual merchant wholesaler should be aided by the identification of those environmental factors associ- ated with the characteristics of the merchant wholesale sector. With knowledge of these factors, the merchant wholesaler will then be better able to adjust his activi- ties and services to meet the present and future needs of his suppliers and customers. As noted earlier the second purpose of this study was to analyze the application of techniques proposed to improve the estimates of regression coefficients using multiple sets of data. The results obtained by Van Tassel (18:9) using the Efficient Estimator Fortran Program (EFFEST) were compared with the estimates in this study using the Zellner-Aikten (ZA) and least-squares (LS) techniques. The analysis of these techniques should aid in the identification of situations where they should or should not be applied. Terms and Definitions of Wholesaling A difficulty encountered in the study of whole- saling is the lack of clear nomenclature in this area of business activity. This situation is described by Schultz (17:175): No tOpic in Marketing is more booby-trapped with mis- understandings and misleading generalizations, for the advanced scholar as well as for the beginning student, than that of wholesaling. In part this is due to the grievous lack, until very recent years, of comprehensive information on the organization and activities of the many varieties of wholesaling middlemen. Wholesaling has been, and still is, the Dark Continent of American Marketing study. In part this is due to confusion of business nomenclature in this field, in part to the conflicting bases of classi- fication applied to such data as have been collected and published. Furthermore, the operations of whole- saling middlemen are characterized by extreme oper- ational and organizational flexibility, so that most generalizations about them require complicating qualifications and may hold true only temporarily. This section presents the terminology and defi- nitions used throughout this study. Unless otherwise noted, these definitions are the contribution of Beckman, 2 Engle, and Buzzell (11). Much of their effort is based on the work of Beckman who developed many definitions used in the Census of Business. Wholesaling Wholesaling includes all marketing transactions in which the purchaser is actuated solely by a profit or business motive in making the purchase and in which, if the goods are bought from a concern operating substanti— ally as a retail establishment and such goods are not intended for resale, the quantity is materially in excess of that which might reasonably be purchased by an ultimate consumer . Wholesale Transaction Three possible criteria for delineating between a wholesale sale and retail sale exist. They are: 1. Status or motivation of the purchaser, i.e., the position of the customer and his purpose in making the purchase. 2. The quantity of goods involved in the transaction. 3. Method of operation of the concern. From these possibilities the status or motivation of the purchaser was chosen as the most basic criterion. Therefore, the resulting definition of a wholesale trans- action is that it is one in which the purchaser does not buy for his own private or personal use or that of his family and friends, but is actuated instead by a profit or business motive in making the purchase. Wholesaling Middleman A business unit whose major activity is wholesaling on either a merchant or agent basis (17:642). Wholesaler A business unit which buys and resells merchandise to retailers and other merchants and/or to industrial, institutional, and commercial users but one which, at the same time, does not sell in significant amounts to con- sumers. In the fields of basic materials, semi-finished goods and tools and machinery, merchants of this type are commonly known as distributors or supply houses (28:23). Wholesale Establishment The definition of this term is probably the most flexible of any in the field of wholesaling. It seems to be a function of who is attempting to define it and for what purpose. Beckman, Engle, and Buzzell spend an entire chapter (Chapter 3) providing the various definitions used. The definitions and their origins are as follows: Judicial View.--The authors note that the courts have consistently used a functional basis in determining the character of an establishment. In 1941, in the case of Fleming vs. American Stores, Inc., the court held that an "establishment" must be a separate and distinct place of business. Generally the courts have held that the owner- ship of the establishment is irrelevant in determining its character; rather it is the functions which it performs that determine its status. Census View.--The United States Bureau of the Census hold that to be classified for census purposes as a wholesale establishment, a business unit must have over 50 per cent of its dollar volume at wholesale. Fair Labor Standard Act.--For purposes of adminis- tering the minimum wage laws (which do not apply to retail establishments), the law is so phrased that any establish- ment that has a "substantial" part of the total gross receipts in non-retail selling is considered to be a wholesale establishment and, hence, covered under the law. At the present time, 25 per cent of total gross receipts is considered "substantial." Robinson Patman Act.-—For purposes of administering the legality of "functional discounts," any establishment classified as a wholesale establishment is legally allowed a "functional 'discount'" on the portion of the business that it conducts on a wholesale basis. Types of Wholesale Establishments The Office of Statistical Standards classifies five principal types of wholesale establishments. They are (9:222): 1. Merchant wholesalers--wholesalers who take title to the goods they sell, such as wholesale merchants or jobbers, industrial distributors, voluntary group wholesalers, exporters, importers, cash-and-carry whole- salers, drop shippers, wagon distributors, retail c00per- ative warehouses, terminal elevators, and cooperative buy- ing associations; 2. sales branches and sales offices—-(but not retail stores) maintained by manufacturing or mining enterprises apart from their plant or mines for the pur- pose of marketing their products; 3. agents, merchandise, or commodity brokers and commission merchants; 4. petroleum bulk stations; and 5. assemblers, buyers, and associations--engaged in the cooperative marketing of farm products. Wholesaling Structure.--The wholesaling structure is defined as that complex of business establishments which is constantly functioning so as to move products of industry through the channels of trade from primary and other producers to the retail outlets or to industrial consumers. It includes all public and private agencies (with certain exceptions noted below) which contribute, as their maigg activity, to the physical flow of merchan- dise or to change in ownership up to the point where the goods reach the hands of retailers or industrial users. (It excludes railroads, trucks, and pipe lines plus the communications industry because these are highly differ- entiated, and their main activity is not exclusively con- fined to wholesaling functions.) Operating Establishment An operating establishment is an economic unit in a single physical location which produces only one or predominantly one good or service for which an industrial code is provided. If the location performs two or more activities for which separate Standard Industrial Classi- fications (SIC's) exist, an attempt is made by the Bureau of the Census to treat the location as two or more separate establishments. Each is assigned a SIC on the basis of the product or product line that is its major activity. The Background of the Problem Although the current study is concentrated on the merchant wholesaling type of wholesale establishment, this background section begins with a consideration of all whole— saler types. It then narrows to a consideration of the various types of wholesalers and, especially, the merchant wholesaler. A review of the literature of wholesaling indicates that very little effort has been made to study the wholesale sector compared to the manufacturing or retail sectors. A possible reason for the lack of literature in this area may be a prevalent belief that the wholesaling sector is very stable. However, a review of the Census data for the period 1948—1967 (l,2,3,4,5,6,7,8) seems to indicate that a considerable amount of change has taken place in the United States economy. Moreover, much of this change appears relevant to the wholesale sector and, especially, the merchant wholesaler. The data in Table 1.1 show the level of sales, employment, and number of establishments in each of the three channel sectors. The manufacturing sector is the largest on the basis of the value of shipments or sales, number of employees, and payroll. 10 mpmua Hfimuwm "mmmcflmsm mo msmcmo hmma "mmmcfimsm mo mamamu mama "GONZO m0 "mousom mucus madmmaozz n muonsuomwscmz mo msmcmo hmma ”mousomm maa.mm Hmm.m was amm.moa.a vam.osm omgflaflmumm mam.m~ mam.m mav.fl amq.aam one.mmv nmcuammmaocz Hm¢.mmaw mmv.ma amm.am ome.mom mmm.ammm mmcuusuOMMscmz loco.ooov looov Aoooc loco.ooov ucmsnmflanmumm uouomm aaoummm mmwonmEm mom mucmEmanm .mucmE mucmEmfinm Hmccmso . -nmuanmpmm mo msam> chum mo msam> Ho mm mm Ho mmamm H .hmma cw mEocoom mmumum omufics on» mo muouomm mafiaflmumm ocm .mcwammmaonz .mcwusuomwscmz on» MOM mama omummmummd .H.H manna 11 Considering only sales level, the second largest sector would be wholesaling. However, on the basis of payroll, the wholesaling sector is the smallest. And the comparison on the basis of payroll (Table 1.1) may be the most meaningful since this measurement may closely reflect the amount that the sectors are paid for their services. On this basis, the wholesale sector is less than one-fifth the size of the manufacturing sector and two-thirds of the retail sector. The Literature Considering the relative size of the sectors, the wholesale sector has received a relatively small amount of attention in the form of substantial publications. This limited attention is reflected by the data in Table 1.2. Less than 5 per cent of the publications directed toward the three channel sectors considered wholesaling as a major topic. Table 1.2. Number of Publications Listed in the Cumulative Book Index for the TOpiCS of Production, Wholesaling, and Retailing for the Period January 1959 to July 1969. T 'c Number of TOpic Publications as Cpl Publications Listed a Percentage of Total Production 167 40 Wholesaling 20 5 Retailing 231 55 Source: Cumulative Book Index (1959-1969) (New York: The H. W. Wilson Company). 12 This lack of interest in the literature of management and education is noted by Lopota (21:130): Although the wholesaling industry represents a major segment of the U.S. economy, it is little understood as an area, poorly defined, and considered by many to be dying or even dead. This paradox is reflected directly in the general literature of management, which contains little about wholesaling despite the fact that a large majority of our large companies have a stake in it. Graduate schools of business neglect wholesaling; witness the fact that the number of courses they offer which deal with it can be counted on the fingers of one hand. Business schools, uni- versity extensions, the American Management Associ- ation, and many other organizations very frequently sponsor seminars on manufacturing, for instance, and yet devote no attention to wholesaling, why? I sus- pect that for many people the answer to this question would be, "Well there's really nothing much to whole- saling these days—~it's a declining business function." Nothing could be more mistaken. The Changinngconomic Environment The growth rates for selected sectors of the United States economy are shown in Figure 1.2. Although the trend is upwards for all sectors, there are differences in the rate of increase. Table 1.3 presents the percentage or share of National Income (N.I.) contributed by each selected sector in 1948 and 1967. In addition, the change in the per- centage of National Income originating in each of the selected sectors through the period 1948-1967 is provided. There was a decrease in percentage of National Product contributed by each of the three channel sectors: manu- facturing, wholesaling, and retailing. The greatest soo - 250 - m H «3 H H O a 100 .. G .3 9-! -H m 8‘ 50 — r—‘l 25 — 10 Year 13 National Income Wholesaling r I I r I 1948 1954 1958 1963 1967 Source: Survey of Current Business, August, 1965, pp. Figure 1.2. The National Income and Amount of National Income Originating in Selected Sectors of the United States Economy for the Census Years 1948 to 1967. 14 Table 1.3. Level of National Income Originating in Selected Sectors of the United States Economy as Per cent of National Income (N.I.) in 1948 and 1967 and Per cent Change in Sector Share of N.I. Per cent Per cent Per cent Change Sector of 1948 of 1967 Of Sector Share N.I. N.I. Of N°I' 1948—1967 Manufacturing 30.6 29.9 - 2.3 Wholesaling 5.7 5.5 - 3.5 Retailing 12.0 9.5 -20.8 Services 8.9 12.0 +34.6 Source: Survey of Current Business, August, 1965, pp. 44-45; July, 1970, p. 21. decrease was in the retailing sector where the relative share contributed to N.I. decreased 20.8 compared to de- creases of 2.3 and 3.5 per cent for manufacturing and wholesaling respectively. During the same period the relative share contributed to N.I. by the service in— dustries increased 34.6 per cent. This service group in- cludes categories such as hotel, automotive, legal, and educational services. These shifts illustrate that changes were occurring in the economic environment during the period of the study. Changes Among the Types of Wholesale Operations On the basis of sales, the merchant wholesaler is the most important type of wholesale operation. It 15 accounted for 45 per cent of the 1967 sales of the whole- sale sector. The shares of the various types of wholesale Operations are shown in Table 1.4. Between 1948 and 1967 significant changes took place in the distribution of sales among the five types of wholesale operations analy- zed by the Bureau of the Census. It can be seen from Table 1.4 that the merchant wholesalers' and manufacturers' sales branches and offices increased their percentages of total whole sale sales at the expense of the other three wholesale types. During the period 1948 to 1967, the merchant wholesalers' sector grew from 129,117 to 212,993 establish- ments (Table 1.5) an increase of 65 per cent. This growth increased the merchant wholesalers' share of wholesale establishments to 68.4 per cent. Notable during these nineteen years was the fact that merchant wholesalers experienced both a combination of slight relative growth of total sales and a rapid growth in establishments com- pared to other types of wholesalers. Such a combination of trends has led to a slower growth rate for the merchant wholesaler on the basis of sales per establishment than the other types of wholesalers (Table 1.6). This differ- ence is especially significant with respect to the manu— facturer's sales comparison of these trends with the other wholesaler types was not possible because of the lack of data provided in the Census of Business on a commodity .momna mammmaonz ”mmmcflmsm mo msmsmo hwma .mvma “momsom .maflocoou ou map ucmo mom ooa Hmuou uoc >m2m 16 ~.N m.m mma.oa omm.m mposwoua EHMm mo manQEmmm¢ m.ma ~.ma aqm.ao ovm.mm mumxoua .macmmm mmHUQMAUHmZ v.m m.m mmw.vm mmv.os mamcuaump .mcoHDMUm xasn Esmaouumm m.sm H.mm smo.ama oom.om mmouuuo mmHMm .mmsoamun mmamm m.umusuommscmz m.¢e ¢.~e mmo.mo~m mmm.mam “mammmaoa3 hangoumz moms mesa head mema mmmamm mHMmmHogz loco.oooc couumumao 00 mass Hmuoe mo ucmo Hmm mm memm mmamm .hmma ocm mvma as coflumummo mo make >3 mumammmaozz mo mmamm Hmuoa mmmucmoumm paw mmamm .v.H magma 17 .momue mammmaon3 “mmmcflmsm mo msmcmo hmma .mvma "moudom .mcflpcsou on map ucmo mom ooa Hmuou no: >m2« m.ma amv.aam mmo.oam Hmuoe Am.mmc o.m m.a HOH.HH ama.ma muosmoua Show mo mumHnEwmm< o.mv m.m v.m mow.wm mma.ma mumxoun .mucmmm mmwocmnoumz m.m n.m H.mH mmm.om Hmm.mm mamcflauwu .mCOfipmum xaon Edwaouumm o.mm m.m o.HH wmm.om mon.mm mmoflmmo mmHMm .mmcocmun mwamm m.umusuommscmz o.mm v.mm n.mm mmm.mam naa.mma umHMmmaonz ucmnoumz homaumvma hood mwma hmma mvma musmenmwanmumm cofiumummo mmucmEQmHHQmumm coflumummo mo mmwa mammmaonz mucmanma amumm mo wows ca Ammmmuomov Hmuoe mo ucmo Hmm .H mmmmuocH ucmo Hmm mm mucmanmflanmumm .nmma ou mvma Eoum COHHMHmmo mo mama an mucmenmwanmumm mo umnasz ca Awmmmuomov wmmmuocH usmo mom can .hmma can mva CH coflumummo mo mama an muamenmwabmumm odomwaonz Hmuoa mo mmmusmonmm paw mucmenmaanmumm .m.H manna 18 .mpmue mammmaonz "mmmcflmsm mo mumcwu nmma .mvma “mousom as ooo.mev.a ooo.mmm mmasu Ham .Hmuoe mm ooo.mam ooo.amm muosuoum Eumm mo meHQEmmmfi mm ooo.omm.~ ooo.oam.a mumxoun .mucmmm mmflocmnoumz «NH ooo.omm ooo.osm mamcHEHmu .mcoflumum xasn Edmaouumm mma ooo.oma.m ooo.mqa.m mmoflmuo mmamm .mmnocmun mmamm m . HTMdfiUMmDCMZ mo ooo.amm w ooo.mmm m “mammmaonz ucmnoumz head on mvaa poms mama awe ”mama m GOwumummo u a .Hn u m mo mass mom mmamm mmmmuocH ucmo umm unmanmflanmumm mom mmamm .hwma paw mqma cmmzumm coflumuwmo odommaosz mo mama an mmmmuosH mmmusmoumm paw unmecmwabmumm umm mmamm .m.H magma 19 line basis. But a study analyzing the difference between types of wholesalers on an all commodity basis is currently under way at the University of California (22:91). The last comparison of the merchant wholesaler with the other types of wholesalers is made on the basis of class-of-customer (Table 1.7). The percentage of sales of the merchant wholesalers to retailers dropped from 46.8 to 40.8 per cent during the period 1948 to 1963. This "loss" of sales has been offset in part by an increase in level of sales to industrial and consumer users. These changes are also reflected in the change in the level of each sector's activities relative to G.N.P. presented in Table 1.3. The absolute and relative changes of the merchant wholesalers compared to other types of wholesalers and sectors of the economy are substantial. These changes include wholesale market share of the various types of wholesalers, number of establishments, sales per estab- lishment of the wholesaler types, and class—of—customers of wholesaler types. The Selection of the Merchant Wholesaler for Study This study concentrated on the merchant whole- saler type of distribution channel for a consumer good commodity line. There are three major reasons for this choice. First, this is the largest wholesaler type on .mpmna mammmaonz "mmmsflmsm mo momcmu mmma .mvma "mousom .mcwpcsou on moo ucmo mom ooa Hmuou uoc >828 20 o.N m.o m.bm m.Hm N.mH v.m 0.5 m.m m.mq w.mm muuspoum Show mo mumHQEmmm¢ m.v N.m w.Hm m.mm o.mH H.0H o.H ¢.o m.vv m.Hv mumxoun .muammm mmflocmzoumz m.m m.N N.mH o.mm «.mm m.Hm v.0 ¢.o H.vv h.Nv mwOHmmO .mmzocmun mmamm m.umusu06wscmz m.m o.m m.vH h.ma m.ov m.o¢ N.H w.H 0.5m m.Hm mHmHmmeO£3 Dawsonmz mmma mwma mmmH mvma mmmH mvma mwma mvma mmma mvma Hoax coflumu muons .oum u no u Iwcmmuo mnmawmumm mqudmsou .HmonmEEoo umHMmmHonz m mammmaosz . .HmanumsocH mo mama "on mmamm mo coausnfluumflo mommucmoumm .mmmH can mvma How sowumEH0mcH umEoumsunmoummmHU mcwuuommm mucmenmflanmumm mHMmmHonz Mom umammmaonz mo momma an mmamm mo coflusnflnumwo mmmusmoumm .>.H magma 21 the basis of total dollar sales and number of establish- ments. It therefore represents the majority of wholesale activity. Second, because of its independence, it is a type of wholesale operation with the ability to shift its efforts from one commodity line to another. Third, the merchant wholesale type of wholesaler has been given the most complete analysis by the Census of Business. The Census of Business often provides data on only two whole- saler categories. Those are total wholesale trade estab- lishments and merchant wholesalers. Such limitation pre— cludes the study of the other types of wholesale establish- ments because the necessary detailed data are not available. Selection of Data Sources As stated earlier, a purpose of this study was to analyze the association between the characteristics of the merchant wholesaling sector and the characteristics of the other channel sectors and the environment. Because it was also relevant to study the assiciations over a period of time so as to note changes occurring in channel structure, existing secondary data sources were required. The selected basic sources which provided the necessary data, were the Census of Business, Census of Manufacturing, and U.S. Census of Population. 22 Variable Selection Characteristics (Dependent Variables) Selected to Analyze the Merchant Wholesaling Sector The characteristics (or the dependent variables) selected to describe the merchant wholesaler sector were total dollar sales, establishments, and sales per estab- lishment. The dollar sales figure was probably the best single indicator of level of the activity in each of the sectors. The establishment figure was used because it has been changing considerably in the merchant wholesale sector (129,117 in 1948 and 212,993 in 1967). And, in order to provide consideration of the "average" merchant wholesaler, the sales per establishment figures were in- cluded. Selection of Independent Variables The independent variables were selected from two sources. First, the literature of marketing proposes a number of associations between the members of the distri- bution channel; that is, the association between the wholesaling sector and the manufacturing and retailing sectors. Second, other variables, economic and demographic in nature, were considered for their possible association with the wholesaling sector. The relationship between the retail sector and wholesale sector has long been considered one of the 23 principles of marketing. Presenting some of these principles in the marketing literature Bartels (10:62) states: Some (principles) have been taken verbatim from the writings; others have not been accredited to par- ticular writers because the generalizations are so common in marketing it would be unjust to attribute a statement of them to any one person. One of the principles in this latter category as outlined by Bartels was the following: "As changes occur in the retail structure, changes will also occur in the wholesale system" (10:64). In the light of this principle, it is interesting to note that a trend in decentralization of manufacturing and retailing activity in the United States has occurred. In the retail sector, Tallman and Blomstrom (25:130) assert: . . . retail outlets represents a catching-up of retail institutions with the vast changes that have taken place in consumer living habits, location, and buying power since World War II. These changes have generated a major revolution in the kinds of quanti- ties of goods purchased by consumers, and in the types and location of shopping which is convenient and attractive to them. And with respect to the manufacturing sector, Miller (23:156) notes: There seems to be present today a general agreement on the wisdom of the decentralization of industry. . . . A majority of the branch factories newly established by existing manufacturing concerns have been located away from their central areas in less concentrated industrial sectors of the country. Taking the above facts into consideration, the important question then becomes something like the following: when 24 shifting of activity occurred in the manufacturing and retailing sectors of the channel, did the merchant whole- salers shift locations to meet them? These shifts in the manufacturing and retailing sectors present a problem to the wholesaler, whose majority of sales are made to retailers. Does the wholesaler main- tain his location near the centers of high population con- centrations or move toward his retail customers who are decentralizing? If the wholesaler does follow the move- ment of the retailers, his sales should more closely follow the patterns of retailers and the population distri- bution. An argument has been presented, that the whole- salers should not follow these shifts. Miller contends that (23:156): It is an open question however, whether the whole- saling industry will or should follow this de- centralization of industrial employment and retail business. . . . There is good reason to believe that this general movement may not have much effect on the present structure of wholesale locations. Wholesaling is primarily a function suitable for large and medium-sized cities . . . To determine if wholesale activity has shifted with movements in population density, the relationship between the per capita merchant wholesale sales and population density was studied. The relationship between per capita merchant wholesale sales and per capita retail sales was also examined. Both relationships were studied for the period 1948-1967. 25 The decentralization of industrial employment was noted previously. If such a shift takes place in a com- modity line, does the merchant wholesaler shift the location of his activity? Possible changes in merchant wholesale sales, establishments, and sales per establish- ment may have been associated with the value added, establishments, and value added per establishment of the manufacturing sector. Because the merchant wholesalers depend upon retailers for the majority of their sales, the association of the wholesaling sector with the sales, establishments, and sales per establishment characteristics of the retail- ing sector were analyzed. Three economic or demographic variables were in- cluded to analyze their association with the dependent variables. The first, personal income, was selected be- cause it would seem to be a measurement of consumers' potential for economic activity. Ideally, a measurement such as disposable personal income should be used in a study of the channels for consumer goods. Unfortunately, these data are not available on the necessary geographic basis. Nevertheless, the inclusion of the independent variable, personal income, may indicate if there is an association between certain characteristics of the mer- chant wholesaling sectors of the channels studied and the amount of income available to the pOpulation of a geo- graphic unit. 26 The second economic variable selected was mean nonagricultural employment as a percentage of total employ- ment. This variable was included to provide a measurement of the association between the characteristics of the merchant wholesaling sectors studied and a variable that might reflect the level of economic development. The finding of significant associations may indicate relation- ships between the level of merchant wholesaling activity and one measure of general economic development. The third variable selected, population density, is a demographic measurement. Tallman and Blomstrom (25) have suggested that retailers have followed or adapted to changes in consumer locations. Because this change has been occurring in the retailing sector, it may have been necessary for the merchant wholesaling sector to adjust. They may have had to move their activities to provide the services required by the retailing sector. The inde- pendent variable, population per square mile, has been selected to provide a test of the association between the characteristics of the merchant wholesaling sector and consumer location. Two basic questions about the independent vari— ables selected for analysis in this section seem reason- able. First, do the selected characteristics have significant association with the characteristics of the merchant wholesaling channel sector? Second, do these 27 associations shift over time? In Chapter VIII the results of the study will be used in a consideration of these questions. Problem Statement As stated earlier, one purpose of this study was to analyze the association of selected independent charac- teristics of the merchant wholesaler's environment with selected characteristics of the merchant wholesaler. This study was also conducted to determine if these associ- ations shifted over a period of time. The independent characteristics data available from Census sources were compared to the values of the merchant wholesalers' total sales, establishments, and sales/establishment by geo- graphic division. These comparisons were made for the Census years 1948, 1954, 1958, 1963, and 1967. General Hypotheses l. The selected characteristics of the merchant wholesalers' sector of a distribution channel for a given commodity line are associated with charac- teristics describing the other sectors of the commodity line distribution channel, the all commodity distribution channel, and the economic environment. 28 The relationship between the selected charac- teristics of the merchant wholesalers' sector of the distribution channel and characteristics describing the other sectors of the distribution channel and economy will remain constant over a period of time. Limitations The limitations of the research are: The geographical areas selected were larger than may have been desired. This was because the smaller geographic units had less variation within the individual areas and more among them. However, when smaller geographic areas, such as states or standard metropolitan statistical areas, were considered, the data available were inadequate. Because of the small numbers of establishments in many categories, especially manufacturing, the United States Bureau of Census does not reveal sales or value added figures. To avoid disclosure limitations, then, geographic divisions were used in the analysis. The product (or commodity) line sales were not broken down by type of retailer and wholesaler in the 1954 and 1958 Census. This did not allow 29 a comparison of the merchant wholesalers with other forms of wholesalers. The value of the dollar was changing from period to period during the study. The statistics used in the analysis were not adjusted for this change because adequate commodity line price indexes do not exist. The product lines selected were limited to those lines where the necessary data detail was avail- able for the three major sectors of the channel. The commodity lines selected were limited to those where the commodity line sales were signifi- cant to all channel members. That is, one-half or more of their sales were of the commodity line. This limited the study to three commodity lines: drugs, liquor, and lumber. The macro approach used may not describe the position of the individual merchant wholesaler within the commodity line merchant wholesaler channel. Rather, the approach describes the mean averages which may not represent the situ- ations of individual group members. The first postwar Census of Business, including data on retailing and wholesaling activity, was 30 conducted in 1948. The first postwar Census of Manufacturers was conducted in 1947. For this census period it was necessary to use data that was not gathered in the same census period. Possible Contributions of Study This study may provide aid to the individual wholesaler constructing his plans for the future. Although the approach does not indicate trends for the individual wholesaler, it does indicate trends that affect his Operation. If the trend is toward large and fewer merchant wholesalers in a product line, the individual merchant wholesaler may realize his choice involves growing, changing product lines or leaving the wholesale trade. This study may help in the selection of the proper methodology in future analyses that use data from different time periods. It will compare the results obtained using least-squares regression analysis with the results obtained when using efficient coefficient estimators. And this com- parison may provide insights into the meaning of the results obtained when using the different methods. CHAPTER II RESEARCH METHODOLOGY Measurement Units of Characteristics Used Selected for Study The characteristics selected for study are pre- sented in Chapter I. Those selected are normally found in the form of totals for the geographic division. In the United States there is a wide variation of population from geographic division to geographic division. The Bureau of Census estimated that in 1967 the populations of the Mountain and East North Central geographic divisions, the smallest and largest, were 7,828,000 and 39,189,000, respectively. There is a relatively high association of economic activity and population in the United States. To reduce some of the association between variables that are corre- lated with population, many of the selected characteristics were used on a per capita basis. This procedure was employed for personal income, sales, and establishment totals. The variables selected for analysis are shown in Figure 2.1. 31 32 Manufacturing Sector 1. Value Added/Capita 2. Establishments/Capita 3. Value Added/Establishment l I Wholesalers, Other P than Merchant Whole- Merchant Wholesaling tutions, etc. 1. Sales/Capita 2. Establishments/Capita 3. Sales/Establishment ' I I salers, Other Manu- : ' facturers, Insti- I l l l _—m_.__.-—..—.-—.-_.-. »- | Other Wholesalers, I Manufacturers, | ' Institutions, etc. I Retail Sector 1. Sales/Capita 2. Establishments/Capita 3. Sales/Establishment l Economic Environment I 1. Personal Income/Capita I 2. POpulation/Capita 3. Mean Nonagricultural Employment as Per cent of Total I Employment ‘-- Flow of Commodity Ownership E] Channel Level Figure 2.1. The Merchant Wholesaler Type of Consumer Good Commodity Line Distribution Channel and Characteristics Selected for Study. 33 Geographic Control Units The geographic control unit used in this study is the ggographic division. The United States is divided into nine geographic divisions shown in Figure 2.2. Other geographic units considered for this study were states and standard metropolitan areas (SMA's). These control units, because of their smaller size, may have been preferable to the geographic divisions, but with important exceptions. These exceptions concerned the amount of useful data able to be obtained on a geographic unit basis. First, the census "withheld to avoid dis— closure" rule applied to the sales figure of the merchant wholesale for thirty out of fifty-one states (and the District of Columbia) for the wine, distilled spirits (SIC 5095 part) kind of business for 1963 (3:1963, Vol. V, 1-75). There were twenty "deletions" in the following merchant wholesalers kind of business: drugs, drugs prOprietaties, druggists' sundries (SIC 5095 part). Second, the small number of manufacturing establishments in many states made the disclosure problem severe in the manufacturing sector of the commodity lines selected. Selection of Merchant Wholesaler Channels Two types of merchant wholesaler channels were selected for analysis in this study. The first type is the all commodity merchant wholesale channel. This channel is comprised of the total activity of each channel 34 .mmumum omuHcD on» mo mcoflmfi>fla oanmmumomu .m.m musmflm .msmcmu mnu mo smmusm .moumEEoo mo ucmEuummmo .m.D "condom 35 sector. The all commodity manufacturing sector is com- prised of all manufacturing value added and establishments in the Census of Business. The all commodity merchant wholesaler sector is comprised of all merchant wholesaler sales and establishments, and the all commodity retail sector, of all retail sales and establishments. The all commodity channel is not a true channel in the sense that all goods of a particular type flow through it. However, it does represent the level of economic activity main— tained by the all commodity channel sectors. The study of the all commodity channel also allowed the establishment of a base line to compare the relationships of individual consumer good commodity line channels. The second type, the commodity line channels, were selected on the basis of the national market as sum- marized in the 1963 Census of Manufacturers (1) and Census of Business (2, 3). This census year was selected because it provided a breakdown on the national sales (or ship- ments) of manufacturers, wholesalers, and retailers by commodity line. This breakdown provision was made on a type of business basis. As noted in Chapter I, to be selected for this study, the commodity line must have been a significant portion of the business for the establishments at each level or the distribution channel. This was one-half of the sales or shipments. In addition, to provide relation- ships between each level of the channel, a significant 36 portion of the manufacturers' sales were to wholesalers and a significant portion of the merchant wholesalers' sales were to retailers. This was one-half of the sales, with the exception of the lumber merchant wholesaling establishments. Only 44.6 per cent of their sales were to retailers. Three commodity line channels met these require— ments. They are delineated in terms of the merchant wholesaler kind of business classifications (9): 1. Drugs, drug proprietaries, druggist sundries (SIC 5022) 2. Wines, distilled spirits (SIC 5095 part) 3. Lumber, millwork (SIC 5098 part) These commodity line channels are referred to as Drug, Liquor, and Lumber, respectively. Multiple Regression The measurements of association between the inde- pendent and dependent variables were determined by use of linear multiple regression or least squares. This pro— cedure finds that linear regression equation which best fits the data. It "yields those estimates of the param- eters which minimize for that form of regression equation the sum of squares of the deviation of the observations from the regression line, i.e., from the value of the obsemvation which it would have if it had coincided with the regression line" (13:86). 37 This method estimates an equation of the form: Y = a + b X + b X 1 1 2 2 where: Y = dependent variable a = constant bl = regression coefficient 1 b2 = regression coefficient 2 X1 = independent variable 1 X2 = dependent variable 2. Preliminary plots of some data used in this study indicate non-normal distributions exist. This fact de- manded a transformation to fulfill the requirements for the use of linear multiple regression. Van Tassel in his study of the retail sector used logarithmic transformation. He found that (13:10): The changes in the measurements of the extent of correlation between dependent and independent vari- ables were judged to be substantial enough to warrant the use of logarithms in the analysis, with the exception of one kind of retail business estab- lishment, "Lumber, Building Materials, Hardware, Farm Equipment Dealers." The relationship in this case appears to be best represented by arithmetic analysis. To provide outputs comparable to Van Tassel, logarithmic transformations were used in this study. The form of the regression equation using logarithmic trans- formations would be: 38 log Y = a + b log X + b log X 1 1 2 2 Thus, the values of the a and b terms differ from those found with the non-transformed data inputs. Zellner-Aitken Estimators Van Tassel used the Fortran program EFFEST to improve upon the "line of best fit" obtained by least- squares regression (13:8). To explain this procedure Van Tassel quotes Zellner (27:348-349): It is only under special conditions . . . that classi- cal least-squares applied equation-by-equation yields efficient coefficient estimators. For conditions generally encountered, (there is) an estimation pro- cedure which yields coefficient estimators at least asymptotically more efficient than single-equations least-squares estimators. In this procedure regres- sion coefficients in all equations are estimated simultaneously by applying Aitken's generalized least-squares to the whole system of equations. To construct such Aitken estimators, we employ estimates of the disturbance terms' variances and covariances based on the residuals derived from an equation-by- equation application of least-squares. Briefly, the purpose of the method is to use data from other time periods to estimate their regression coef- ficients. To be consistent with Van Tassel's notation, we have two time periods represented by the equations: it t t it it and 39 where Yit ,Yit = dependent variables 1 2 a , a = constant terms t t l 2 b , b = regression coefficient t t l 2 Xit ,Xit = independent variables 1 2 Uitl'UitZ = re51dual terms i = observation number t = first time period t = second time period. First the single equation least-square analysis for each time period is estimated. The implicit assumption is that residuals for each observation are equal to zero or: Then the regression equation for both equations are solved simultaneously assuming that: 40 Since neither value is equal to zero, it is assumed that the values have some relationship to one another. There- fore, the experiences of the two time periods are not independent, and the regression coefficient for the two periods are determined simultaneously. Two points should be made about EFFEST at this time. First, although the values of the regression coef- ficients are allowed to change from their values calcu- lated using the least-squares technique, the centroid (Y ) for each time period is held constant. Second, it'iit the Zellner-Aitken estimators are calculated using an iterative process, each iteration asymptotically approach- ing the "best fit." However, because of the cost of com- puter time and the small increments of improvement, not more than one iteration is performed. The ZA program developed by the Michigan State University Agricultural Experiment Station (30) is based on the same assumptions as the EFFEST program. However, the ZA program is a more powerful one than the EFFEST program, and has succeeded EFFEST in the Station's files because it provides a more efficient estimate. That is, its first estimate is closer to the "best fit" than the first estimate of the EFFEST program.l 1Discussion with Marylyn Donaldson, Michigan State University Agricultural Experiment Station, March, 1971. 41 An examination of Zellner (27) reveals that the EFFEST and ZA methods may not be appropriate in the Van Tassel study. Zellner states: "This gain in efficiency can be quite large if 'independent' variables in different equations are not highly correlated . . . " (27:348). Because all of the independent variables selected by Van Tassel are of the same economic variables measured at different times, they would tend to be correlated. Additionally, the applications suggested by Zellner (27:349) involve a system of dependent variables such as " . . . a single cross-sectional budget study regressions for several commodities are to be determined." In his study, Van Tassel has applied the method to a situation where the dependent variables are measurements of the same variable at different times. This does not appear to be a type of application suggested by Zellner. To determine if the application of the method by Van Tassel was correct, the following was done. In Chapter III the results obtained using least-squares and the ZA program are compared to the results obtained in Van Tassel's study using least-squares and the EFFEST program. The ZA run uses data from the Census of Business for the Retail sector on a geggraphic division basis. This was the same data source used by Van Tassel on a state basis. 42 Statistics Derived The following statistics will be calculated for each selected product line for each of the five Census years: Coefficient of determination (rZYX) which is the prOportion of the variance in the dependent variable that was associated with the independent variables. Degree of association is the relative amount of error reduced in the estimate of one variable because of knowledge about another (12). O O Y " y.x A = _ y.x o y Ay x = degree of association of dependent variable y with independent variable x. a? = standard deviation of variable y about its mean y. 5? x = standard deviation of variable y around conditional distribution of y. "Coefficient of regression which indicates the nature of the change in the dependent variable in response to a change in the independent variable" (18:11). Standard error of the estimate measures the extent of the deviation between the computed coefficient of :regression and the actual observations. 43 Cross-section elasticities of the independent variable with respect to the dependent variables can be determined in two ways. If the data are not transformed, the formula for the cross-section elasticity is: e = b x —— (18:12) e = cross section elasticity b = coefficient of regression D = mean value dependent variable I = mean value independent variable This formula determines the cross-section elasticity for only the point on the regression line where Dm and Im are the mean values of the dependent and inde— pendent variables. Unless the regression line passes through the origin, each point on the line will have a different value of cross-section elasticity. If the data is logarithmically transformed, the regression coefficients are the cross-sectional elastici- ties of the association. Since the data in this study were logarithmically transformed, it should be noted that the regression coefficients calculated are also the cross— sectional elasticities. 44 The Hypotheses Tested Regression Coefficients The first group of hypotheses tested were those related to the significance of the relationships between the dependent and independent variables for the relation- ships hypothesized in Chapter I--for example, the relation- ship between the dependent variable per capita drug mer- chant wholesalers' sales and the independent variable per capita personal income. The hypotheses for each set of relationships are: H : B = 0 O H B # 0 l: where B is the regression coefficient. Change in Regression Coefficients The second group of hypotheses to be tested were those related to the significant changes that have occurred in the relationship of the dependent and inde- pendent variables during the period of the study (1948— 1967). The hypotheses used for each set of relationships are: Ho‘ B1948 = B1967 H1: B1948 # B1967 or 45 o B1948 ‘ B1967 = B1948 ' B1967 # 0 Significance of Findings in This Study Regression Coefficient (B) The test for the significance of the regression coefficient can be made using the E test described in Chapter III. In the case of a single independent variable, significance of the regression coefficient can also be found by testing the correlation coefficient (r ). In YX this study, the correlation coefficient was so tested. This procedure provided an easier comparison of the find- ings where single and multiple variable correlation coef- ficients were determined. For example, there may be two independent variables that, individually, do not have a significant relationship with a common dependent variable. Yet the multiple correlation coefficient may be significant. The 5 test does not provide a way to examine this type of situation. A level of correlation required for significance was selected on the basis of two requirements. First, it should, at a minimum, be statistically significant (.6666). Second, the independent variable must explain a reasonable anwunt of the variation in the dependent variable. The level selected for this study was a coefficient of associ- ation (Ayx) of 0.5. This is translated into a correlation 46 coefficient (rY ) of .866. This means that the error in X estimating the value of the dependent variable is reduced by a factor 0.5 if knowledge of the value of the inde- pendent variable is available. To translate this .5 coefficient of association (Ayx) to coefficient of determination (rZYX) or coefficient of correlation (r ), the following relationships were YX used (12:512): rzyx = 1 — (1 - Ayz)2 r2YX = 1 - (1 - 5)2 r2YX = l - (.5)2 r2Yx = 1 - .25 r2YX = .75 r2YX = .866 In this study significant values of r and r2 are .866 XY XY and .760 respectively. The selection of an arbitrary value of correlation seems to be a contradiction of Elkblad's advice, "We think it is best not to have any arbitrary boundaries for a minimum degree of useful correlation" (12:518). However, such boundaries were selected for three reasons. First, 47 it was desired to find those associations that are useful indicators of merchant wholesaler activity. To achieve this purpose, there should be a "significant," not just a statistically significant, level of association. That is, the association should explain a substantial portion of the variance of the independent variable about its mean. This results in a conservative use of the term significant, com— pared to the .05 level of statistical significance. Second, different sample sizes were used with different method- ologies and the use of a significant level of association allowed a more meaningful comparison of results. Third, the values selected exceeded the statistically significant boundary at the .05 level. This situation is explored in more depth in Chapter III. The following classifications were used to evalu- ate the significance of the correlation coefficients. These standards were applied to only single independent variable relationships. Significant.--At least one year in which the correlation coefficient (rYX) was greater than [.866] and, for all other years tested, was greater than [.6666 Marginally Significant.--The mean correlation coef- ficient for the five years tested was greater than [.6666 Nonsignificant Relationship.-—The mean corre— lation coefficient for the five years tested was less than [.666]. 48 Multiple Correlation Coeff1c1ents (ryxlxz ) In addition to the single independent variable least-square analyses multiple regression analyses were run. This was done to determine if certain combinations of independent variables could provide significantly better estimates of the dependent variables than could the ipdi- vidual independent variables. The combinations of inde- pendent variables selected for testing were those found significant in Van Tassel's study along with others that appeared relevant to the merchant wholesaler's channel-- for example, the relationship of the dependent variable per capita merchant wholesaler sales and the independent variables per capita manufacturing value added and per capita retail sales. The combinations selected for this study are pre- sented in Table 2.1. In Chapters IV through VII the re- sults are reported for only those combinations of variables that provided significant improvement over the single independent variable. The criterion for selecting the significant multiple correlation coefficients were: 1. The multiple correlation coefficient was greater than [.866] for at least one year. This is equivalent to a coefficient of association of at least 0.5. 49 Table 2.1. Combinations of Single Dependent and Multiple Independent Variables Studied for Association in Terms of Multiple Correlation Coefficients. Dependent Variables Independent Variables *Per Capita MW Sales Per Capita MF Value Added Per Capita RT Sales *Per Capita MW Est. Per Capita MF Est. Per Capita RT Est. *Sales per MW Est. Value Added per MF Est. Sales per RT Est. Per Capita Comm. MW Sales Per Capita All MF Value Added Per Capita All MW Sales Per Capita All RT Sales Per Capita Comm. MW Est. Per Capita All MF Est. Per Capita All MW Est. Per Capita All RT Est. Sales per Comm. MW Est. Value Added per All MF Est. Sales per All MW Est. Sales per All RT Est. *Per Capita MW Sales Per Capita Personal Income Nonagricultural Employment Per Capita MW Est. Per Capita Personal Income Nonagricultural Employment Sales per Comm. MW Est. Per Capita Personal Income Nonagricultural Employment Per Capita Comm. MW Sales Per Capita Personal Income Population per Square Mile Per Capita Comm. MW Est. Per Capita Personal Income Population per Square Mile Sales per Comm. MW Est. Per Capita Personal Income Population per Square Mile 50 Table 2.1. (Continued) Dependent Variables Independent Variables Per Capita Comm. MW Sales Per Capita All MW Sales Per Capita Personal Income Per Capita Comm. MW Est. Per Capita All MW Est. Per Capita Personal Income Sales per. Comm. MW Est. Sales per All MW Est. Per Capita Personal Income *These combinations were studied for both the all commodity and commodity line channels. Key: All = All Commodity lines factor; Comm. = Commodity Line; Est. = Establishment; MW = Merchant Wholesaler; ME = Manufacturing; RT = Retailing. 51 2. The mean correlation coefficient must have .8744 averaged at least I.7943| or . These are the statistically significant levels of r, when there are two or three independent variables, respectively. 3. The multiple correlation coefficient must be at least 0.10 higher than the highest of the single independent variable correlation coefficients. Changes in Regression Coefficients (1948-1967) To test the changes that may have occurred in the regression coefficient, it was necessary to use a 5 test. Since the significance of the difference was being tested, the test was of the form: t = o (B1948 ’ 81967) t(Bl948 ‘ 81967) # 0 where: 81948 = regression coefficient in 1948 81967 = regression coefficient in 1967 381948 = standard error of B1948 3 = standard error of B1967 B1967 52 SE = /a’§ +62 1948-1967 1948 t = B1948 ’ B1967 (31948 ' B1967) SE 1948—1967 Graphically the test can be represented by Figure 2.3. 53 .hme on mva UOHHmm on» How ucmwoflmmmou scammmnmmm may cw mmcmsu ms» mo mommoHMHcmHm Hmowumflumum How umma .m.m musmflm .>.0Ahmmam mvmamvu omH.N u .>.olhmmam . mamamc ONH.NI u u jlfl m umwoo< om ummoofl mm umwoo< o v somam . mamam "mm o A smmam - mwmam "am 0 u ammflm - mvmam "om CHAPTER III THE COMPARISON OF THE METHODOLOGY OF THIS STUDY AND THE VAN TASSEL RETAIL STUDY Introduction The purpose of this chapter is to explore the methodology (ZA) used in this study and the methodology (EFFEST) used by Van Tassel in An Analysis of Factors Influencing Retail Sales. For purposes of comparison, a part of this study is a modified replication of Van Tassel's. This permits a more accurate appraisal of the similarities and differences between the methodologies in the two studies. Then, using the output obtained from this study, a decision concerning the appropriateness of the methodologies can be made. The Data Used The basic data source used for both studies was the United States Census of Business. The reports of 1948, 1954, 1958, and 1963 were used in Van Tassel's study. The reports of 1948, 1954, 1958, 1963, and 1967 were used in this study. 54 55 The goegraphic units used in Van Tassel's study were states, while those in this study were geographical divisions. Because Van Tassel was using data from only the retail sector of the distribution channel, he was able to use the smaller geographic unit, the state. But when the merchant wholesaler sector and manufacturing sector were included, much of the detail desired was not provided on a state basis. This was because the dis- closure rules prevent the Bureau of the Census from pro- viding sales data when only a few establishments are found in a geographic unit. Hence, it was necessary for this study to move to the next larger geographic unit, the gpographic division. Computational Techniques In both this study and Van Tassel's, the first stage of the process was a linear regression analysis using least-squares. The next stage in both studies was an attempt to provide regression coefficients that were more efficient estimators. Van Tassel used the EFFEST program while this study used the ZA program. The ZA technique is a more recently developed program and has succeeded the EFFEST program. Results of the Van Tassel Study and This Study for the Drug Retail Sector To replicate the Van Tassel study, a common set of dependent and independent variables were studied. The 56 dependent variable selected was drug retail establishment sales per capita. The drug retail establishment sector of the channel was the same as the Drug Stores and Proprietary Stores used in the Van Tassel study. The independent variables which were selected to duplicate the Van Tassel study were personal income per capita and mean per cent nonagricultural employment. The following statistics were calculated by Van Tassel using the EFFEST program ro by this study using least-squares and the ZA program: Van Tassel (EFFEST): B1 = regression coefficient of independent variable "per capita personal income." B2 = regression coefficient of independent variable "mean per cent nonagricultural employment." 3 = standard error of B . B1 1 3 = standard error of B . B2 2 t8 = t value of B1 or (Bl/oB ). l l th = t value of B2 or (BZ/OBZ). In addition to the above, the following were found by tflmis study using Least-Squares (LS) and (ZA): A0 = "Y - Axis" intercept of regression equation. 0A 0 standard error of A0. 57 _n n -' tA0 t value of A0, (Ao/OAO)' The results of the computation are shown in Table 3.1. The Results of Comparing ZA to EFFEST Estimates Regression Coefficients (B1 and B2) The three methods of calculating the coefficients of regression discussed in the previous sections were used. These methods are least-squares, EFFEST, and ZA. Calcu- lations were made for the 1948, 1954, 1958, 1963 (and 1967). The coefficients of regression calculated were the per capita personal income and mean nonagricultural employment as a per cent of total employment. The dependent variable in the comparison was per capita drug retailing establishments sales. The values obtained are plotted in Figures 3.1 and 3.2. Figure 3.1 shows the relationship between the independent variable, personal income per capita, and dependent variable, drug retail sales per capita. In both the case of the EFFEST and the ZA estimators, the coefficients calculated were less than the coefficients calculated using the least-squares technique. Figure 3.2 shows the relationship between the independent variable mean per cent nonagricultural employ— ment and independent variable drug retail sales per capita. 58 Table 3.1. The Values of the "Y-Axis" Intercept, Regres— sion Coefficients, Standard Errors of the Estimates, and "t" Value Estimates for the Drug Store Retail Sales per Capita Using the EFFEST Program, Least-Squares, and the ZA Program for the Years 1948, 1954, 1958, 1963 (and 1967). A B E Year 0 1 2 LS ZA EFFEST LS ZA EFFEST LS ZA 1948 1.33 1.86 .56 .89 .52 .31 - .36 - .03 1954 1.72 2.08 .66 .88 .51 - .10 - .55 - .11 1958 2.84 2.65 .58 .83 .41 - .05 -l.02 — .22 1963 2.72 2.45 .58 .76 .44 - .27 - .84 - .14 1967 3.67 2.84 a .67 .46 a -l.15 - .36 3 E 3 Year A0 B1 BZ LS ZA EFFEST LS ZA EFFEST LS ZA 1948 .78 .52 .ll .48 .23 .21 .56 .32 1954 .85 .67 .10 .58 .28 .25 .90 .45 1958 .85 .66 .09 .40 .21 .28 .80 .41 1963 1.24 .95 .ll .48 .27 .35 1.07 .62 1967 2.09 1.59 a .56 .36 a 1.61 1.05 t t t Year AO Bl 82 LS ZA EFFEST LS ZA EFFEST LS ZA 1948 1.71 3.60 5.63 1.87 2.30 1.41 - .65 - .09 1954 2.04 3.10 6.47 1.51 1.84 — .38 - .62 - .24 1948 3.34 4.02 6.29 2.07 2.01 - .18 -l.28 — .53 1963 2.20 2.58 5.09 1.60 1.62 - .77 - .79 — .23 1967 1.75 1.79 a 1.19 1.27 a — .71 - .35 aThe Van Tassel Study was concluded before the 1967 Census was conducted. EFFEST values are from the Van Tassel study. 59 (LS) N = 9 for \\\\ Each Year .6 .1 \ —— ‘—“ (EFFEST) N = 200 for Each Year S . \ \ \ = \ ’1’”; (ZA) Each43eiir \ /"’ / \x ’ .4 - I r f ’ 1 ' 48 54 48 63 6'7 Year 1950 1960 1970 Note: N is the sample size used to calculate 81' Figure 3.1. The Regression Coefficients (Bl) Between the Independent Variable Personal Income per Capita and the Dependent Variable Drug Retail Establish— ments' Sales per Capita Using the Least—Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967). Sample Size (N) for Each Year. 60 \.\ \\ /. \\ ‘/fl,’ \\ \1/’ \\\\ \\ (EFFEST) ,\ N = 200 for 0 Each Year (ZA) N = 45 for Each Year N = 9 for Each Year (LS) Year Note: N is the sample size used to calculate E Figure 3.2. T 48 I I T 1' 54 58 63 67 2. The Regression Coefficient (B ) Between the Independent Variable Mean Per cent Nonagricultural Employment as Per cent of Total Employment and the Dependent Variable Drug Retail Establishments' Sales per Capita Using the Least- Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967). Sample Size (N) for Each Year. 61 For this regression coefficient, the values calculated using the EFFEST and ZA estimators were greater than the values calculated using the LS estimators. Also for this coefficient, the values calculated for EFFEST and ZA are quite close. Although it could not be tested statistically, it appeared that the values of the regression coefficient calculated using the EFFEST and ZA methods were related. Three factors that may have contributed to the differences that did exist were identified. First, the ZA and EFFEST methods were not exactly identical. As discussed earlier, the computer programs used differ. Second, the geographi- cal units used for the techniques differ. States were used for the EFFEST program and geographic divisions for the ZA program. Third, the ZA estimates were calculated including 1967 census data; and since the observations from each time period affect all the other periods, when using the ZA and EFFEST estimators, some variations were expected. However, in spite of differences, both the estimating techniques (EFFEST and ZA) have a common origin, the LS analysis. And because both estimates were found to have a similar relationship with this similar origin, their relationship to each other seemed meaningful. 62 Standard Error of theRegression Coefficients (GB and 6B ) 1 2 The standard error of the regression coefficient was the estimate of sampling error that was calculated for each of the three techniques: least-squares, EFFEST, and Zellner—Aitken estimators. The values of these estimates are plotted in Figure 3.3 and Figure 3.4. The standard errors of the regression coefficients of drug retail establishments' sales per capita and person income per capita (331) are shown in Figure 3.3. The standard errors of the regression coefficients of retail drug establish- ments' sales per capita and mean per cent nonagricultural employment (3B2) are illustrated in Figure 3.4. In both cases, the size of the standard error estimates varied with the estimator used. The largest estimates were found using the least-squares technique, the second largest, using the ZA estimator, and the small— est, using the EFFEST estimator. This difference in estimates was expected considering the size of the sample used in each of the calculations. For the calculation using LS, the sample size was 9, for ZA, it was 45, and for EFFEST, it was 200. It was necessary to compare the estinmtes on the basis of standard deviation of the :regression coefficient. This comparison was done by Inultiplying the standard errors by the square root of 'their'respective degrees of freedom. These results are presented in Table 3.2. 63 El 07- 06 "‘ (LS) N = 9 for Each Year .5 d I / / / .4 — / I. / // \ /’ / \ / / \ / r” \\ \// 3 . (ZA) ' N = 45 for Each Year 0‘ x. /. —( \ ‘2 (EFFEST) \./ N = 200 for Each Year I l I I I 1 ' Year 48 54 58 63 67 1950 1960 Note: N is the sample size used to calculate EB . l 1Figure 3.3. The Standard Error of the Estimate of the Regression Coefficient (531) Between the Independent Variable Personal Income per Capita and Dependent Variable Drug Retail Establishments' Sales per Capita Using the Least-Squares (LS) Technique and EFFEST and Zellner-Aitken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967). Sample Size (N) for Each Year. 64 3 1.7 B2 1 1.6~ (LS) N = 9 for Each Year l.5~ 1.4. 1.3- 1.2‘ [(ZA) 1‘11 . / N = 45 for l 04 Each Year '9‘ .\\\\\\\ ./ .8« ° / 7- / .6. /' ' /’ 5« / . . / 4 / ’ ‘\ \ ./ o / / // /° (EFFEST) .3. ° , N = 200 for ..a—'°”’ ’I” Each Year .",,. .2‘ .1. Year 4'8 I 54 58 I 63 6T7 1950 1960 Note: Figure 3.4. N is the sample size used to calculate 682' The Standard Error of the Estimate of the Regression Coefficient (032) Between the Inde- pendent Variable Mean Per cent Nonagricultural Employment and Dependent Variable Drug Retail Establishments' Sales per Capita Using the Least-Squares (LS) Technique and EFFEST and Zellner-Aiken (ZA) Estimators for the Census Years 1948, 1954, 1958, 1963 (and 1967). Sample Size (N) for Each Year. 65 Table 3.2. Estimate of the Standard Deviation of the Regression Coefficients Using the Estimated Values of the Standard Errors of the Regression Coefficients Calculated Using the LS Method and EFFEST and ZA Estimators for the Drug Retail Establishments' Sales per Capita for the Census Years 1948, 1954, 1958, 1963 (and 1967). 5A 081 682 Year LS ZA EFFEST LS ZA EFFEST LS ZA 1948 1.91 3.03 1.52 1.08 1.31 2.90 1.37 1.87 1954 2.08 3.91 1.38 1.42 1.63 3.46 2.20 2.62 1958 2.08 3.85 1.24 .98 1.22 3.87 1.96 2.39 1963 3.04 5.54 1.52 1.08 1.57 4.84 2.62 3.62 1964 5.12 9.27 a 1.37 2.10 a 3.94 6.12 aThe Van Tassel study was concluded before the 1967 Census was conducted. Note: Degrees of freedom are equal to (n - k)s - l n = sample size k = number of independent variables 5 = number of samples For LS, d.f. = 6; ZA, d.f. = 34; EFFEST, d.f. = 191. 66 With minor exception, the standard deviation estimates using the ZA and EFFEST estimators are larger than those using the LS techniques. This was as expected because the ZA and EFFEST techniques use the observations from all census periods in the calculation of the regres- sion coefficients for each individual period. And a wider dispersion of values was found for the combined census years than for any one census year. These dispersions, or ranges, are presented in Table 3.3. For example, the independent variable, drug retail establishments' sales per capita, had ranges from $16 to $26 for the individual census years, while the range for the entire period, 1948— 1967, was $51. Therefore, the EFFEST and ZA methods, which used data from all time periods, had a larger vari— ance than the least-square method using one time period. When a significant difference in range of values for the individual year and the entire period was found, care in the application of techniques such as EFFEST and ZA had to be exercised. The problems encountered in this situation are presented in a later section of this chapter. "t" Values, A/EA , 81/3B , 132/6'B l l 2 Because of the difference in sample size, a com- parison of the "t" values was not made. These "t" values were used by Van Tassel to determine the significance of Imis findings (18:12-13). An evaluation of this part of ‘his pmocedure is made in a later section of this chapter. 67 Table 3.3. Range of Values for the Dependent and Independent Variables for the Individual Census Years and Total Census Period 1948- 1967. Dependent Independent Variable Variables Census Drug Retail Per Capita Mean Per cent Years Establishments Personal Nonagricultural Period per Capita Sales Income Employment Min. Max. Min. Max. Min. Max. 1948 $18.20 $34.60 $860 $1660 54.5 94.6 1954 22.40 41.00 1120 2150 63.5 95.3 1958 29.50 49.20 1340 2450 73.3 96.1 1963 34.90 57.10 1660 2870 77.5 97.2 1967 43.80 69.50 2240 3600 83.4 98.0 Total period (1948— 1967) 18.20 69.50 860 3600 54.5 98.0 68 Conclusion on Comparison of EFFEST and ZA Considering the limitations on the comparison, it was concluded that the EFFEST and ZA estimators do provide about the same result. As pointed out earlier, it was not possible to arrive at a conclusion concerning whether ZA is a more powerful estimator than EFFEST. An Analysis of the Effects of the Estimators on the Regression Coefficients The LS technique and ZA estimator give values of the regression coefficient that differ considerably. This section compares the differences to determine which was the correct value for consideration in this study. First, we looked at the drug retail establishments' sales per capita and two independent variables. When there are two independent variables, the correct graphical presentation of the regression equation is three—dimensional. Thus, a transition from numerical to graphical methods was made to provide a visual presentation of what occurs. Because of the difficulty in depicting the three-dimensional case, additional cases with only one independent variable were studied graphically. The Three-Dimensional Case If the regression equations for Van Tassel's drug retail establishments' sales per capita case were plotted, the results would be pictorially confusing. If 69 geometrically correct, this representation would require the plotting of ten planes (two for each of the five census years). To reduce some of the potential confusion caused by such complex geometric plotting, Figure 3.5 was drawn with only three pairs of equations represented. In addition, the relative position of the planes and their slopes were shifted to provide a clearer representation. Each of the equations is skewed to the left. This resulted in the left end of the 1967 ZA plane shifting downward and the left end of the 1948 plane shifting up- ward. (The centroids of the ZA planes remain constant.) But these shifted planes do not appear to provide the best description of the relationship for their respective years. The bias tended to pull the regression equations together in the direction of the skewness (to the left). This can be observed in the graphic presentation found in Figure 3.5. Three Cases of Efficient Estimators In order to overcome the graphical difficulty of ‘presenting three-dimension cases, a study of two- cthnensional cases was made. These cases have only one independent and one dependent variable. The first case presents what might have been expected of the ZA esti- mators. Because an actual example of this type was not :flound in the study of the channels examined, the variables used ftm'illustration were artificial. Even though it was Figure 3°5° (overlaY)“-Regression Equations Calculated Using ZA Estimators -——-..-. .._ ,-4l£u9153 anoiisupa noiaasxpsH--(ysixsv0) .a.£ 9: . —- —— aiojsmijea AS pniaU (xl'XZ’Y)1967 (Xl'XZ’Y)l958 (xl'XZ'Y)1948 2 C) Centroid 9f Ehe Data for Census Year (X Ix IY) 1 2 Least-Squares Equation x1 -’—- Zellner-Aitken Estimated Equation Figure 3.5. A Non-Scale Representation of the Regression Equations for the Drug Retailing Establish- ments' Sales per Capita for the Census Years 1948, 1958, and 1967. The Least-Squares (LS) and Zellner-Aitken (ZA) Estimators Were Used for Each Year. 71 not actually observed in the study, this case was included to provide a representation of the ZA method for the three conditions discussed. It reflects the advantage of the larger sample size employed when the estimators are used. If the small individual samples (years) from a common pOpulation were used in the estimate of the regression coefficient, it should have provided a better estimate of the population regression coefficient than the use of only one sample. In the second case the independent and de- pendent variables selected were all commodity per capita retail sales and per capita retail drug establishments' sales, respectively, while the independent and dependent variables in the third case were nonagricultural employ- ment and drug merchant wholesaler per capita sales, respectively. The length of the regression lines was determined by the range of values of the independent variable for individual census years. Case 1. In this case the regression coefficients for each set of data (time-period) were approximately the same. With this situation, it would appear that a more efficient estimator for each set of data would be found using all sets of data simultaneously. This case is shown in Figure 3.6. Each solid line represents the least-square (LS) regression equation for one set of data. The "0 s Figure 3.6. (Overlay)--Regression Equations Estimated Using ZA Estimators--—-— yd noiaaalpBH--( R Li (Ibij E3 1383 ‘L’bi‘IC-BVO) -aiojsnijed AS pniaU 1963 1948 Log Artificial 195:967 Dependent Variable 1954 Log Artificial Independent Variable C) Centroid for Data of Census Year LS Regression Equation _u—u——Regression Equation Calculated Using ZA Estimators Figure 3.6. Case 1. Artificial Data to Show How the ZA Estimator Might Improve the Estimates of the Regression Coefficients for the Individual Census Periods_(l948-l967). 73 represent the centroids of the individual sets of data, and each dashed line, the regression equation for a set of data after the application of the ZA routine. Case 2. In this case the least-squares equations were nearly parallel. This relationship is shown by Figure 3.7. The equations calculated using the ZA esti— mators are approximately parallel to each other, but not to the equations estimated by LS. Because the ZA equations were closer to the horizontal than the LS estimator, one reasonable explanation did exist. The ZA estimate was made using the data for all time periods. Because of this, a greater range of variables (Table 3.3) were used in the calculations of each ZA, compared to those employed for L8. These increased ranges, then, reduced the value of the estimators or slopes in this case. This is reflected by the lower coefficients of correlations (r2) as revealed in Table 3.4. Unless special circumstances exist, such as in Case 3, the ZA estimates tend to parallel the horizontal axis, rather than the slope of the individual equations calculated by LS. Case 3. In this case another type of undesired result was found while using both LS and ZA estimators. The pair of equations calculated by LS and ZA for each of the first three census periods were very similar (Figure 3.8). However, the two methods provided very different equations for 1963 and 1967. In this case, unlike Case 2, / // / / / fl / // / // ’q/ // / // / // fl/ ‘ / / / / / / / // /. / / / / / , / Figure 3.7. (Overlay)--Regression Equations Calculated Using ZA Estimators - —---- uajeituiLD anoijgspfi noieaejfofi-—(ysiiev0) .Y.E earpiq - - - eiojfimijafi AS pniaU O Centroid of Data for Year -—-—- LS Regression Equation 1967 ————- Regression Equation Calculated Using 2A Estimators 1963 1958 Log Drug Merchant Wholesaling Establishments per Capita Sales 1954 / Log All Commodity Retailing Establishments per Capita Sales Figure 3.7. Case 2. Regression Equations for the Relation- ships Between the Independent Variable, A11 Commodity Retail per Capita Sales, and Depend- ent Variable, Drug Merchant Wholesaling per Capita Sales, Using LS and ZA Estimators for the Individual Census Periods (1948-1967). bojsiuoisD noijsupB naiaaaipufl——(jsiievfi) - -—-.-'4r"‘4 2" -y 'F' w~ - -aiojbuijea A5 LHLeJ /e/ 1967 1963 Log Drug Merchant Wholesaling 1958 Establishments per Capita Sales 1954 1948 Log Mean Per cent Nonagricultural Employment 0 Centroid of Data for Year LS Regression Equation -—-—-—- Regression Equation Calculated Using ZA Estimators Figure 3.8. Case 3. Regression Equations of the Relation- ships Between the Independent Variable, Mean Per cent Nonagricultural Employment and Depend— ent Variable, Drug Merchant Wholesaling Estab- lishments per Capita Sales, Using LS and ZA Estimators for the Individual Census Years (1948-1967). 76 Table 3.4. The Coefficient of Correlation (r) Between the Independent Variable All Commodity per Capita Retail Sales and Dependent Variable Drug Mer— chant Wholesaler per Capita Sales for Least- Squares (LS) and Zellner-Aitken for the Census Years 1948, 1954, 1958, 1963, and 1967. Coefficient of Correlation (r) Year LS ZA 1948 .65 .61 1954 .55 .50 1958 .49 .46 1963 .32 .31 1967 .30 .28 the ZA equations have a much steeper slope (Bl) than the LS equations. The difference between the equations for each of these years was basically a result of the skewness of the data. Recalling that the centroid of the equations for the ZA and LS must be the same for each year, one notes the slope of the 1963 and 1967 ZA estimates shifting toward the left end of the 1948 and 1954 data ranges. Although this would make a better estimate for the entire period of the study, it did not improve the estimate of the individual census periods. Many varieties of 2 and 3 were found in a review of the computer output. However, none of the types described by Case 1 was found. Considering the limitations of Zellner included in the review of EFFEST and ZA in 77 Chapter II, it was concluded that the use of either EFFEST or ZA would be inappropriate. First, because the con- ditions stated by Zellner were not met and second because the actual application of EFFEST and ZA reviewed in this chapter obtained incorrect estimates with the two method- ologies. Therefore, the analyses for Chapters IV, V, VI, and VII were made using only the results of the least-squares calculations. Testing the Significance of the Findings Van Tassel used a p test to determine if the corre— lation between the dependent and independent variables occurred by chance. The hypotheses being tested were of the form: HO: B = 0 H1: B # 0 B = the regression coefficient. To perform this test the "t" values were used. {:ng— B 3B = standard error of B. For his study Van Tassel (18) selected a critical *value of p at the .05 level of significance. He calcu- .lated the degrees of freedom to be 46, and used a one-tail 78 test with a critical value of 1.67. Unfortunately, several errors of a technical or omission—type nature were com- mitted by Van Tassel on these points. First, the degrees of freedom for the ZA estimators as used by Van Tassel was 191 (see Table 3.2), not 46 as calculated by Van Tassel. Because he assumed a large sample size and used a E value, this procedural irregu— larity did not lead to an error in the selection of a critical value for testing. Second, the one-tail test was incorrectly used by Van Tassel. If prOperly employed, the one-tail test would determine those regression coefficients that were sta- tistically greater than 0. Or the test could also deter— mine if the value of E was large enough in the negative direction to be statistically significant. The test hypothesized by Van Tassel (18:13) is shown in Figure 3.9. He should have accepted Hl only when E was greater than 1.67. Yet, he accepted relationships as highly signifi- cant when their E values were negative. In these negative cases he stated that the relationships between dependent and independent variables were inverse. However, to test in both directions a two-tail test with Ec.v. = 1.96 was appropriate (Figure 3.10). Third, Van Tassel used the estimated standard error of the regression coefficient (EB) that was calcu- lated using the EFFEST program. Ruble has stated that 79 o A m . m I o o v m a m "Euom on» mo mommguommm ecu MOM HmmmMB cm> mp cmuwmmSDOQmm mm umma Hflmanmco .m.m musmflm fi'll. o m ummood m udmood H 8.1 80 "Euom may mo coma m>mm casocm Hmmmme cm> umce ummB HflmBIOBB .oa.m mucmflm 1A 1i H m ummood om ummood m ummood 81 uncertainty about the meaning of these terms remain (16:125): The square roots of the diagonal elements of the estimated coefficient variance-covariance matrix (i.e., the square roots of the estimated coef- ficient variances) are often used as approximate coefficient standard errors and the ratios of the coefficients to the square roots of the estimated coefficient variance are often used as approximate coefficient t—ratios; however, very little infor- mation is available on how well these computed values serve as approximate standard errors and approximate t-ratios. Unfortunately, however, Ruble does not provide a substitute for these values at this time. The statistical significance of the relationship between the dependent and independent variable(s) can be based on the E—ratio or the correlation coefficient. At a given level of significance both methods would reject or accept the same relationship. The critical values are calculated using the following formulation (14). R2 = (1 _ R2 ) (Fk,(n-k-l)'k) c.f. c.v. (n-k-l) where: R2 c.v. = critical value of the correlation coefficient F = "F" value for degrees of freedom k and (n-k-l) k = number of independent variables per establishment n = number of observations. 82 The critical values of the single (or multiple) coefficients of determination for this and the Van Tassel studies are presented in Table 3.5. A coefficient of determination .0193 would have been statistically signifi- cant in the Van Tassel study. However, on a practical basis the independent variable explained very little about the distribution of the dependent variable. Conclusion Based on the findings presented in this chapter, three methodological differences exist between Van Tassel's and this study. They are: 1. The efficient estimators (EFFEST and ZA) were not used in the remaining chapters of this study. 2. To test the significance of the relationships, the correlation coefficients, not the E-ratios, were tested in this study. 3. Conservative significance levels (as explained in Chapter II) were used to establish the criti- cal values for this study. ‘o 83 Bmmhmm m.ammmm9 cm> mmma. Nmna. mmma. mmmo. oomo. mmao. oom ¢N NA95m maze Nvav. mamm. mmmm. mama. nmma. memo. ma m4 mcsum mace vvnm. meme. meow. mvon. oamm. «eve. m m N H m m a moanmflum> mwabmflum> oNHm GOHDMOHHQQm ucmocmmwccH useccmmmccH mHQEmm coeumHmuuou coflumcflfiumumo mo unmeoflmmmou mo ucmeoemmmou .mmHQMHHm> useccmmmccH mo muonsoz cam mmuflm madamw cmuomamm you mocmoflmflcmflm mo Ho>mq mo. msu um Auv coHumHmuuoo mo mucmeoflmmmoo cam .m.m wanes ANHV coflumcflfiumumo mo mpcmwowmmwoo mamwuasz cam mamcflm mo mmzam> Hmoflueuu CHAPTER IV FINDINGS: THE ALL COMMODITY MERCHANT WHOLESALER CHANNEL The all commodity merchant wholesaler channel represents the sum of the characteristics (dollar sales and establishments) of the individual merchant wholesaler channels. This all commodity channel provided a base line to compare the individual commodity line distribution channels that were examined in this study. Merchant Wholesaler Characteristics Studied The characteristics of the merchant wholesaler sector selected for study were the total dollar sales, total establishments, and mean dollar sales per establish— ment. The levels of these characteristics for the all commodity channel are presented in Table 4.1 for the census years 1948 to 1967. The index numbers indicate the relative level of the individual characteristic with respect to the base year, 1948, a convenient reference base for this study. The use of index numbers provided a method of comparing 84 85 Table 4.1. Characteristics (and Index Numbers) of the All Commodity Merchant Wholesaler Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967. Total Total Mean Merchant Merchant Merchant Year Wholesaler Wholesaler Wholesaler Sales 000,000 Establishments Sales per Establishment (Index Number) (Index Number) (Index Number) 1948 $ 76,533 (1.00) 129,117 (1.00) $593,000 (1.00) 1954 101,436 (1.33) 165,698 (1.27) 612,000 (1.03) 1958 122,060 (1.60) 190,492 (1.47) 641,000 (1.08) 1963 157,392 (2.06) 208,997 (1.62) 753,000 (1.27) 1967 206,035 (2.69) 212,993 (1.65) 967,000 (1.63) the relative changes that took place between character- istics, channel sectors, and channels. It can be seen (from Table 4.1) that the merchant wholesaler activity grew during the study period (1948- 1967). Compared to the manufacturing and retailing sector, the most notable growth occurred in the number of merchant wholesaler establishments. This sector grew 65 per cent during the period, compared to 29 and 0 per cent for the manufacturing and retailing sectors, respectively. Characteristics of the Other Channel Sectors Manufacturing and retailing were the other two sectors of the all commodity merchant wholesaler channel 86 studied. The characteristics of these sectors selected for study are presented in Tables 4.2 and 4.3. Table 4.2. Selected Characteristics (and Index Numbers) of the All Commodity Manufacturing Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967. Total Total Mean Manufacturing Manufacturing Manufacturing Year Value Added Establishments Value Added 000,000 per Establishment (Index Number) (Index Number) (Index Number) 1948 $ 74,290 (1.00) 240,807 (1.00) $309,000 (1.00) 1954 117,032 (1.58) 286,814 (1.19) 408,000 (1.32) 1958 141,541 (1.90) 303,387 (1.26) 467,000 (1.51) 1963 192,103 (2.58) 311,921 (1.29) 616,000 (1.99) 1967 262,358 (3.54) 311,754 (1.29) 665,000 (2.15) The levels of total value added or sales rose during the period 1948-1967. However, the index numbers indicate that the manufacturers' value added was growing faster than the retailers' sales level (3.54 to 2.39). The total manufacturing establishments rose 29 per cent during the period, while the number of retail establishments remained stable. But the all commodity value added or sales per establishment of the two sectors grew at nearly the same rate during the period of the study. 87 Table 4.3. Selected Characteristics (and Index Numbers) of the All Commodity Retail Sector of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967. Total Total Mean Retail Retail Retail Year Sales Establishments Sales per 000,000 Establishment (Index Number) (Index Number) (Index Number) 1948 $130,521 (1.00) 1,796,540 (1.00) $ 76,000 (1.00) 1954 170,568 (1.31) 1,727,967 (1.00) 99,000 (1.30) 1958 200,365 (1.54) 1,794,744 (1.02) 112,000 (1.47) 1963 244,202 (1.89) 1,707,391 (0.96) 143,000 (1.88) 1967 310,214 (2.39) 1,763,324 (1.00) 176,000 (2.32) Selected Economic Characteristics The economic characteristics selected for study are presented in Table 4.4. These are mean per capita personal income, mean population per square mile, and mean per cent nonagricultural employment. The trend during the 1948—1967 period is not sur- prising. It included increasing personal income per capita and population density. The level of mean per cent non- agricultural also rose during this time. Mean Per Capita Sales or Value Added and Establishments To offset the differences in the levels of activity because of pOpulation variations between the geographic divisions, the analyses for most variables were conducted 88 Table 4.4. Aggregate Selected Economic Characteristics (and Index Numbers) of the United States Economy for the Years 1948, 1954, 1958, 1963, and 1967. Mean Mean Mean per Capita Population Nonagricultural Year Personal per Square Employment as Income Mile Per cent Total Employment (Index Number) (Index Number) (Index Number) 1948 $1435 (1.00) 49.0 (1.00) 80.0 (1.00) 1954 1770 (1.23) 54.2 (1.11) 84.7 (1.06) 1958 2057 (1.43) 58.5 (1.19) 86.9 (1.09) 1963 2449 (1.71) 63.4 (1.29) 89.6 (1.12) 1967 3159 (2.20) 66.4 (1.35) 93.1 (1.16) on a per capita or per 1,000,000 population basis. Tables 4.5 and 4.6 present the sales or value added and estab- lishments on this basis. The comments concerning the total sales and establishment levels, then, apply to per capita values. Obviously, magnitudes of the positive growth rates were reduced because of the pOpulation growth during the study period. Results of the All Commodity Channel Study Regression Coefficients (B) As was stated earlier (in Chapter II), the signifi- cance of the association between the dependent and inde- pendent variables was determined by analysis of the 89 Table 4.5. Per Capita Sales or Value Added (and Index Numbers) for Manufacturing, Merchant Whole- saling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967. Mean Mean per Mean per per Capita Capita Sales Capita Value Added by Merchant Sales by Year by Manufacturing Wholesaling Retailing Establishments Establishments Establishments (Index Number) (Index Number) (Index Number) 1948 $ 509 (1.00) $ 524 (1.00) $ 894 (1.00) 1954 726 (1.43) 629 (1.20) 1058 (1.18) 1958 813 (1.60) 701 (1.34) 1151 (1.29) 1963 1018 (2.00) 834 (1.59) 1294 (1.45) 1967 1326 (2.60) 1041 (1.99) 1568 (1.75) Table 4.6. Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 POpulation. Mean Mean Merchant Mean Manufacturing Wholesaling Retailing Year Establishments Establishments Establishments per 1,000,000 per 1,000,000 per 1,000,000 Population Population Population (Index Number) (Index Number) (Index Number) 1948 1649 (1.00) 884 (1.00) 12,116 (1.00) 1954 1779 (1.08) 1028 (1.16) 10,720 (0.88) 1958 1742 (1.06) 1094 (1.24) 10,358 (0.85) 1963 1653 (1.00) 1108 (1.25) 9,050 (0.75) 1967 1576 (0.96) 1076 (1.22) 8,912 (0.74) 90 correlation coefficients. This was done for two reasons. First, practical significance, not statistical signifi- cance, was used as a basis of analysis. Second, the find- ings in terms of correlation coefficients could more easily compare to the findings in the multiple correlation analysis. Per Capita All Commodity Merchant Wholesaler Sgl§§.--The simple correlation coefficients between the dependent variable per capita merchant wholesaler sales and the selected independent variables studied are pre- sented in Table 4.7. The relationships between the dependent variable, per capita all commodity merchant wholesaler sales, and the two independent variables, mean per capita all com— modity manufacturing value added, and mean per capita retail sales, were not significant. However, their correlation coefficients were of the same magnitude during the period studied (1948-1967). There was only one independent variable that had even a marginally significant association with per capita all commodity merchant wholesaler sales. It was per capita personal income. This correlation coefficient was slowly declining during the period. Neither of the other two independent variables selected, however, was significantly related to the dependent variable. 91 .ucmoHMHcmHm maamcflmumz« mmmm. meam. nemm. mmmm. emmm. nmma mmme. eeee. mmmm. mmmm. meme. mmma ommm. meme. mmem. oomm. emmm. mmma mmom. mame. memm. Hoam. meHm. emma momm. mmme. mums. mmmm. moom. meme ucmE>onEm mmamm cmccd mSHm> Hmuoa mo ucmo mom wwwww «Mmmmmw mafiaflmumm mcflusuommscmz mm ucmfimoamem umm m Hmuammum muflpoeaou Ham muwGOEEou Ham new» amusuasowummcoz coaum smom Hmm.cmmz muemmu muwmmu cmmz . H umm cmmz Hem cmmz .mmmH paw .mmmH .mmma .emma .memH mummy mnu MOM moanmwnm> pcmccmmmccH Umuomamm paw mmamm Hmammmaonz ucmzoumz mueCOEEou Ham muflmmu Hmm cmmz cmmBDmm AHV mucmeowwmmoo coHumHmuuou .m.e magma fi—_—fl__ 92 Per Capita All Commodity Merchant Wholesaler Establishments.--The simple correlation coefficients be- tween the dependent variable, per capita merchant whole- saler establishments, and the selected independent vari- ables studied are presented in Table 4.8. A significant association exists between the dependent variable and the independent variable, mean per capita personal income. Marginally significant relation- ships exist between the dependent variable and the inde- pendent variables, mean per capita manufacturing establish— ments and mean per cent nonagricultural employment. All three of the correlation coefficients decreased during the period of the study. Moreover, there is a notable similarity among the three independent variables associated with per capita merchant wholesale establish- ment. They are all measures of economic development. Thus, it would appear that the degree of association be- tween per capita wholesaling establishments and the three independent variables become weaker as the level of eco- nomic development increases. Perhaps this reflects no more than the economics of scale or a trend toward larger scale establishments with increasing levels of economic develOpment. Mean Sales per All Commodity Merchant Wholesaler §§tab1ishment.--The single correlation coefficients be- tween the dependent variable, mean sales per all commodity 93 .pCMOHMHcmHm«« .ucmoflwficmflm maamcflmumz« emmm. memm. Hmmm. maam. mmmm. mmma mmam. mmmm. ommm. mmmm. mmon. mmmH moms. mmmm. mmmm. Hmmm. mamm. mmma mmem. mmmm. momm. mmmm. Hmmh. emma mmmm. omoe. mmmm. omom. mmmm. mema «ucwEmOHdEm mad: wEoocH mucwenmflabmumm emucmESmeanmumm Hence no ucmo Hmm mummq *«mcomum mcflaemumm mcflusuommscmz mm ucmE>0HmEm mom m Hmuammom hufiposfioo Had muflpoaaoo Ham new» amusuasownmmcoz . muammu muemmo coaumadmom you new: . new: . Mom new: Hmm 2mm: .mmma paw .mmma .mmma .emma .memH mummy any How mmHQMfiHm> ucmpcmmmch cmuomamm Ucm mucmecmflanmumm mcflamwmaosz unencumz mufipoafiou Ham muwmmu you new: cmmBuwm Any mucmflowmmmoo coeumawuuoo .m.e manme 94 merchant wholesaler establishment, and the selected inde- pendent variables are presented in Table 4.9. None of the correlation coefficients was signifi- cant for the period of the study. Multiple Correlation Coefficients A selected group of multiple (linear) regression analyses were conducted. The combination of independent variables selected were presented in Chapter II. The pur— pose of the analyses was to find those combinations of independent variables that provided significantly better association with the dependent variables than did a single independent variable. In the all commodity channel only one combination of independent variables provided a significantly higher correlation coefficient than did the single correlation coefficients. The combination of per capita manufacturing establishments and per capita retail establishments had significantly higher correlation coefficients with per capita merchant wholesaler establishment than did either of the individual independent variables. These findings are shown in Table 4.10. Although not significant, by the definition out- lined in Chapter II, an additional relationship is pre- sented in Table 4.10. The combination of per capita manufacturer value added and per capita retail sales had a noticeably higher correlation coefficient with per 95 emmo. mmmm. mmee. manm. mmom. mmma mamo.| mmmm. emom. mmmo. . mmmm. mmma mmoo.| memm. Hmma. mnoo. Hmee. mmma momm.l mmmm. mmmo.| ommm.n meme. emma mmme.| puma. Hmom.l oemm.| mmma. mema ucmE>0amEm Hmuoe no name Hem mam: wEoocH acmenmwanmumm unmenmflanmumm mnmsvm Hchmuwm HHMpmm meansuomWscmz mm ucmeonmEm . Ham» Hem muwmmo mom Hem cmcpm amusuasoflummcoz Q a new: coHumac on no new: momma cmmz mcam> cam: .mmma can .mmmH .mmmH .emma .memH mummw ecu How mmHQmflHm> acmccmmmch cmpomawm can ucmEnmeHQMDmm mcHHmmmHosz ucmzoumz Hem mmamm cmmz cmmzumm Aug mucmfloflmmmou somumamnuou .m.e magma 96 Table 4.10. Multiple Independent Variable Combinations that Provide Correlation Coefficients Signifi- cantly Superior to Each of the Independent Variables Considered Separately for the All Commodity Channel for the Years 1948, 1954, 1958, 1963, and 1967. Multiple and Single Combination of Correlation CoeffiCients Dependent (Y) and Year Independent Vari- Multiple Single ables (x ,X ) . l 2 r r r Studied YXlX2 YXl sz Per Capita A11 MW Est (Y) 1948 .9561 .8599 .8080 Per Capita All MF 1954 .9132 .7931 .7887 ESt (x1) 1958 .8585 .7212 .6651 Per Capita A11 RT 1963 .7430 .7057 .2865 ESt (x2) 1967 .8225 .5995 .3113 Per Capita A11 MW Sales (Y) 1948 .7700 .6005 .6536 Per Capita All MF 1954 .7302 .6146 .6101 value Added (X1) 1958 .7235 .5374 .6300 Per Capita A11 RT 1963 .6152 .4679 .5356 sales (X2) 1967 .6676 .5564 .5355 97 capita merchant wholesaler establishments than either of them separately. Both of the findings in Table 4.10 indicate that the all commodity merchant wholesaler characteristics, per capita sales and establishments, were associated with the corresponding characteristics of both the manufacturing and retailing sectors of the channel. Changes in the Regression Coefficient As stated in Chapter I, the second hypothesis con- cerned changes over time of the relationships between the independent and dependent variables studied. In this section the regression coefficients and the coefficients' standard errors of the all commodity channel relationships are presented. These values were determined for each Census of Business year during the period 1948 through 1967. Also calculated with the E values for the differ- ences between the regression coefficients for the years 1958 and 1967. These t values provided a method to deter- mine if the regression coefficient (B) had changed signifi— cantly during the study period. The critical t value was 2.120, with 14 degrees of freedom at the .05 level of significance. Per Capita All Commoditererchant Wholesaler Sales.--The regression coefficients and the coefficients' standard errors between the dependent variable per capita 98 merchant wholesaler sales and the selected independent variables are presented in Table 4.11. None of the coef- ficients changed significantly during the period of the study. The low E values indicate the low level of change that took place. The greatest change occurred in the case of mean per cent nonagricultural employment. This change may be attributed, in part at least, to the narrowing range of values for this independent variable. An example of this narrowing was observed in Figure 3.8. Per Cgpita Merchant Wholesaler Establishments.-- In Table 4.12 the regression coefficients and coefficients' standard errors are presented for the dependent variable, per capita merchant wholesaler establishments, and selected independent variables. One statistically significant change in regression coefficient took place. This was between the dependent variable and per capita retail establishments. Although not statistically significant, other changes in regression coefficients were noted. The regression coefficients of independent variables, per capita manufacturing establish- ments, per capita personal income, and population per square mile, decreased during the study period. This would tend to indicate that the dependent variable was varying less with changes in the level of the independent variables toward the end of the study period. The lower levels of regression coefficients (or lepes of the linear 99 mm.+umm|meu mo.|nmm|meu eH.|ummnmeu mm.+umm|meu ma.+unm|meu mme.a mmm.H mmm. mmm. mmm. moo.a mom. omH.H eom. Hmm. nmma Hem. mmm.H eno. nmo. omm. mmm. oam. mmo.H cam. mmm. mmma mmm. Hoe.a mno. mmo. mam. mmm. mmm. mmH.H mmm. mmm. mmma Hom. HHN.H omo. mmm. mmm. mmm. mmm. omo.H mmm. mmm. emma ome. Hmo.a mmm. mmm. eom. mmo.H mme. mmm. mmm. . mam. mema mm m mm m mm m mm m mm m ucmfimoamam mad: OEOOCH mmamm Uwpcm msam> Hmuoe mo ucmo Mom wummqm HMCOmuwm mcflaflmumm mcmusuommscmz new» we ucmE>OHmEm “we muammo muHcOEEOO Had muHOOEEOU Had amusuadowummcoz cod m :90 Hmd.cmm muflmmo muflmmo saw: .u a m 2 Add new: mom cam: ANH.N u u m0 mmHm> Hmo Iwuwuov .ucmflommmmou COHmmmummm am uwflnm nmma ou mema mo msam> u .mmma mam .mmma .mmma .emma .mema mummy may MOM mmabmwum> usmccmmmch cmuomamm cam mmamm nmmmmmaonz ucchHmz >UHOOEEOO Had MUHQMO umm cmmz cwm3umm Amov mucmwOflmmmoo mo uounm cwmccmum paw Amy mucmflowmmmou conmmHmmm OmumEHDMM .HH.e magma 100 .eeaoaeaeeam maaaoaemaemum. mH.+nmmImeu mm.lunm|meu mm.Huummumeu «om.m+u>mnmeu me.anummnmeu eom. eom.a omo. Heo. mmm. eom. mmm. mmm. HmH. mmm. nmma mmm. eme.H mmo. mmo. emm. mmm. moa.a mmm. mmm. Noe. mmma mam. moe.H emo. mmo. mmm. mmm. mom. mmm.H mmH. mme. mmma mmm. mmm.a mmo. emo. emm. moo.H mmm. mmm.a mmm. mem. emmH mom. mmm.a mmo. mom. mmm. m>H.H omm. mmm.m mmm. can. memH mm m mm m mm m mm m mm m ucmE>OHmEm mucmezmeanmumm mucmenmeabmumm Hmuoe mo ucmo Hem OWMMW Mmmmmw mamammumm mcflusuomwdcmz Hmmw mm ucmamoamsm Hem m Hmuadmum muHcOEEOU Ham muHcOEEOO Had amusuHOOflHmmcoz . muammo muammo coaumasmom umm cmmz . . cmmz . mom cmmz umm cmmz ANH.~ n 0 Mo 05H8> Hmoeueeoe .ucmeHmmmou GOHmmmummm CH amanm mmmH ou memm mo msHm> u .nmma paw .mmmm .mmma .emma .mema mummy may MOM mOHQMeHm> ucmpcwmmccH cwuomamm cam mucmE Ismfimnmumm mCHHMmmaonz ucmnonmz mumcofiaou Had meadow Hmm cam: cmm3umm Amov mucmwommmmou mo uouum pumwcmum paw Amy mucmfloflmmmoo coemmmummm cmumEHDMm .ma.e OHQMB 101 regression equation) indicate the dependent variables were less responsive to changes in the independent variables. Mean Sales per Merchant Wholesaler Establishment.-— The regression coefficients and coefficients' standard errors between the dependent variable, sales per merchant wholesaler establishment, and selected independent vari- ables are shown in Table 4.13. None of the changes was statistically significant for the period of the study. However, four of the relationships underwent changes that deserve attention. The correlation coefficients of value added per manufacturing establishment, sales per retail establishment, per capita income, and pOpulation per square mile increased. This latter finding appeared to be related to the changes noted for the previous dependent variables, per capita merchant wholesaler sales and establishments. 'While the per capita sales relationships remained stable, the number of establishment relationships decreased and the sales per establishment relationships increased. This suggested that while the merchant wholesaler establish- ments had become more uniformly distributed on a per capita basis, the variation in the scale of operations (sales per establishment) became greater. 102 hmlmep mmlmeu omlmew omlme omlmeu me.+u em.+u He.H+u mm.+u u am.a+u omm. Hma. meo. mmo. mmm. mmm. hoe. eom. mom. eae. omma mmm. meo.n mmo. meo. mmm. mom. mmm. mmo. Ham. oma. mmma eme. noo.| mmo. mmo. mom. NHH. mmm. moo. ema. mem. mmmH mmm. oam.n Heo. mmo. omm. meo.| mme. mmm.: mmm. Hmm. emma mmm. mmm.: oeo. oao. mom. mmm.: mmm. oom.| emH. mmo. memH mm m mm m mm m mm m mm m acmemoamfim Hmuoe mo ucmo mom mam: OEOOGH ucmESmenmumm ucmEnmenmumm mm cwEmo as mumsvm Hchmumm mcHHfimumm meansuomwscwz Ham» p H m Hem muflmmo Hmm Hon pmcp< Hmnsuasowummcoz a a new: :owumas om um cmmz mwamm new: 05Hm> new: ANH.~ u p Mo msam> Hmoauauoo .mcmaoaemmou scammmaemm ea ueaem emme op mama mo maaa> 0 .emme can .mmma .mmma .emma .meme mummw may 90m mmanmeum> ucmmcmmmmcH omuomamm mam ucmfinmflanmumm mcflHMmmHOSZ ucmaoumz muHUOEEOU Had you mmamm smmz cmmzumm Amov mucmwowmmmou mo Houum Unaccmum cam Amy mucwHOflmmwou cofimmmummm cmumEHHMm .mH.e OHQMB CHAPTER V FINDINGS: THE DRUG CHANNEL In this chapter characteristics of the drug mer- chant wholesaler channel sector are compared to traits in selected drug manufacturing and retailing sectors' charac- teristics. In addition, the merchant wholesalers' charac— teristics are compared to selected economic characteristics. Drug Commodity Line Merchant Wholesaler Characteristics The characteristics of the drug merchant whole- saler sector selected for study were the total dollar sales, total establishments, and mean dollar sales per establishment. The levels of these characteristics are presented in Table 5.1 for the census years in the period 1948 to 1967. The index numbers indicate the relative level of the individual characteristics with respect to the base year, 1948. The use of index numbers provided a method of comparing the relative changes that took place between characteristics, channel sectors, and channels. A11 characteristics of the drug merchant wholesaler sector increased during the study period, 1948 to 1967. 103 104 Table 5.1. Selected Characteristics (and Index Numbers) of the Drug Merchant Wholesaler Sector of the Drug Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Drug Total Drug Mean Drug Merchant Merchant Merchant Year Wholesaler Sales Wholesaler Wholesaler 000,000 Establishments Sales per Establishment (Index Number) (Index Number) (Index Number) 1948 $1370 (1.00) 2205 (1.00) $ 621,315 (1.00) 1954 2173 (1.58) 2801 (1.27) 775,794 (1.25) 1958 2826 (2.06) 3042 (1.38) 928,994 (1.50) 1963 3581 (2.62) 3321 (1.51) 1,078,289 (1.74) 1967 4749 (3.47) 3053 (1.38) 1,555,519 (2.50) Comparing these findings to the all commodity channel on the basis of index numbers (Table 5.2), one notes several differences. First, the total sales level of the drug merchant wholesaler grew faster than the rate for all mer- chant wholesalers. Second, until the 1ast census year, 1967, the number of drug merchant wholesalers establish- ments was proportional (on the basis of index number) to the number of merchant wholesaler establishments. Third, the sales per establishment of the drug merchant whole- salers increased much faster than the sales per establish- ments of the all commodity merchant wholesaler group. 105 Table 5.2. The Index Numbers of Selected Characteristics of the Manufacturing, Merchant Wholesaling, and Retailing Sectors of the A11 Commodity and Drug Merchant Wholesaler Distribution Channels for the Years 1948, 1954, 1963, and 1967. Manufacturing Total Total Value Added per Year Value Added Establishments Establishment All Dru A11 Dru All Dru Commodity g Commodity g Commodity g 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.58 1.92 1.19 1.00 1.32 1.90 1958 1.90 3.10 1.26 0.96 1.51 3.22 1963 2.58 4.27 1.29 0.87 1.99 4.90 1967 3.54 6.14 1.29 0.75 2.15 8.10 Merchant Wholesaling Sector Total Total Sales per Year Sales Establishments Establishment All All All Commodity Drug Commodity Drug Commodity Drug 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.33 1.58 1.27 1.27 1.03 1.25 1958 1.60 2.06 1.47 1.38 1.08 1.50 1963 2.06 2.62 1.62 1.51 1.27 1.74 1967 2.69 3.47 1.65 1.38 1.63 2.50 Retailing Sector Total Total Sales per Year Sales Establishments Establishment All Dru All Dru All Dru Commodity g Commodity g Commodity g 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.31 1.31 0.98 1.00 1.30 1.30 1958 1.54 1.69 1.02 0.91 1.47 1.81 1963 1.89 2.11 0.96 0.98 1.88 2.16 1967 2.39 2.72 1.00 0.96 2.32 2.83 106 Drug Commodity Line Manufacturing Establishment Characteristics The characteristics of the drug manufacturing sector are presented in Table 5.3. This sector reflected two major trends during the 1948-1967 period: a rapid growth in the amount of value added, and a concentration of activity into fewer manufacturing establishments. These trends compound into a very substantial increase in value added per establishment. A comparison of the drug and all commodity manu- facturing sectors' index numbers (in Table 5.2) indicates the magnitude of the above mentioned trends. The level of the drug manufacturing value added increased by 514 per cent compared to 254 per cent for the all commodity manu- facturing sector. And while the number of all commodity manufacturing establishments was growing 29 per cent, the number of drug manufacturing establishments decreased by 25 per cent. But the greatest difference was in value added per establishment. While the all commodity manu- facturing value added per capita increased only 115 per cent, the drug value added per capita increased 710 per cent. Drug Commodity Line Retail Establish- ments' Characteristics The drug retail establishments' sales and sales per establishment grew about the same amount during the period of the study. The number of drug retail 107 Table 5.3. Selected Characteristics (and Index Numbers) of the Drug Manufacturing Sector of the Drug Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Drug Total Mean Drug Manufacturing Drug Manufacturing Year Establishments' Manufacturing Value Added per Value Added Establishments Establishment 000,000 (Index Number) (Index Number) (Index Number) 1948 $ 607 (1.00) 1158 (1.00) $ 524,179 (1.00) 1954 1162 (1.92) 1163 (1.00) 999,140 (1.90) 1958 1882 (3.10) 1114 (0.96) 1,689,407 (3.22) 1963 2595 (4.27) 1011 (0.87) 2,566,765 (4.90) 1967 3721 (6.14) 875 (0.75) 4,252,571 (8.10) establishments remained about constant during the period. These figures are shown in Table 5.4. The levels of the selected characteristics for the drug and all commodity retail sectors are shown in Table 5.2. The levels of drug retail establishments' sales and sales per establishments grew slightly faster than the levels for the all commodity retail establishments. Thus, there are only slight differences between the index numbers of the drug and all commodity retail establishments. Mean per Capita Drug Establishments' Sales or Value Added and Establishments To offset the differences in the total levels of activity among the various geographic divisions, the 108 Table 5.4. Selected Characteristics (and Index Numbers) of the Drug Retailing Sector of the Drug Distri- bution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Drug Total Mean Drug Retailing Drug Retailing Year Establishments' Retailing Sales per Sales Establishments Establishment 000.000 (Index Number) (Index Number) (Index Number) 1948 $ 4,012 (1.00) 55,796 (1.00) $ 71,904 (1.00) 1954 5,250 (1.31) 56,009 (1.00) 93,734 (1.30) 1958 6,778 (1.69) 50,792 (0.91) 133,446 (1.81) 1963 8,487 (2.11) 54,732 (0.98) 155,064 (2.16) 1967 10,930 (2.72) 53,722 (0.96) 203,454 (2.83) least-squares analysis of variables was conducted on a per capita basis. Tables 5.5 and 5.6 show the sales or value added and establishment on this basis. And the comments concerning the total sales and establishments' levels apply to the per capita values. Obviously, the magnitudes of the positive growth rates were reduced because of the population growth during the study period. Results of the DruggChannel Study Regression Coefficients (B) As was stated earlier (in Chapter II), the signifi- cance of the association between the dependent and Table 5.5. 109 Per Capita Sales or Value Added (and Index Number) for Drug Manufacturing, Merchant Whole- saling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967. Mean Mean per Mean per per Capita Capita Sales Capita Value Added by Drug Sales by Year by Drug Merchant Drug Manufacturing Wholesaling‘ Retailing Establishments Establishments Establishments (Index Number) (Index Number) (Index Number) 1948 $ 4.16 (1.00) $ 9.83 (1.00) $27.48 (1.00) 1954 7.22 (1.74) 13.50 (1.37) 32.61 (1.19) 1958 10.82 (2.60) 16.24 (1.65) 38.95 (1.42) 1963 13.73 (3.30) 18.95 (1.93) 44.90 (1.63) 1967 18.79 (4.52) 23.98 (2.44) 55.20 (2.01) Table 5.6. Drug Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population. Mean Drug Mean Drug Mean Drug Manufacturing Merchant Retailing Year Establishments Wholesaling Establishments per 1,000,000 Establishments per 1,000,000 Population per 1,000,000 Population Population (Index Number) (Index Number) (Index Number) 1948 7.93 (1.00) 15.10 (1.00) 38.22 (1.00) 1954 7.22 (0.91) 17.40 (1.15) 34.79 (0.91) 1958 6.40 (0.81) 17.48 (1.16) 29.19 (0.76) 1963 5.35 (0.67) 17.57 (1.16) 28.96 (0.76) 1967 4.42 (0.56) 15.42 (1.02) 27.13 (0.71) 110 independent variables was determined by analysis of the correlation coefficients. Per Capita Drug Merchant Wholesaler Sales.--The simple correlation coefficients between the dependent variable, per capita drug merchant wholesaler sales, and the selected independent variables are presented in Table 5.7. Two significant associations and one marginally significant association were noted for the study period. First, the dependent variable was significantly associ- ated with the independent variables, per capita all com— modity merchant wholesaler sales and per capita personal income. The level of the first correlation coefficient was increasing, while the second was decreasing. Then, an additional correlation coefficient was marginally significant. This was the association between the dependent variable and the independent variable, per capita all commodity retail sales. All three of the above associations were between the dependent variable and variables that are indicators of general economic activity, that is, wholesale sales, personal income, and retail sales. The only other independent variable that came close to a significant correlation coefficient was per cent nonagricultural employment. This was also related to the general economic indicators mentioned above. 111 Table 5.7. Correlation Coefficients (r) Between Mean per Capita Drug Merchant Wholesaling Establishment Sales and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean per Capita Mean per Mean per . Capita Drug Capita Drug Alieggggiilty Year Manufacturing Retailing Wholesalin Establishments Establishments' Establishmegts' Value Added Sales Sales** 1948 .3843 .5652 .8169 1954 .3934 .3953 .8757 1958 .1927 .3106 .9130 1963 .3817 .0690 .9128 1967 .4582 —.0952 .9107 . Mean Mgznigzr Populztion Nonagricultural Year Pergonal Sqfiare Employment Income** Mile as Per cent of Total Employment 1948 .8879 .2546 .8080 1954 .8444 .3812 .7446 1958 .7547 .2843 .6017 1963 .7083 .4847 .5043 1967 .7282 .5347 .5285 Mean per Mean per Capita Capita A11 Commodity All Commodity Year Manufacturing Retailing Establishments Establishments' Value Added Sales* 1948 .4975 .8091 1954 .5847 .7395 1958 .3575 .6997 1963 .4440 .5640 1967 .4651 .5477 *Marginally significant. **Significant. 112 Per Capita Drug Merchant Wholesaler Establish- mgpt§.--The simple correlation coefficients between the dependent variable, mean per capita drug merchant whole- saler establishments, and the selected independent vari- ables are presented in Table 5.8. A significant association existed between the dependent variable and independent variable, per capita all commodity merchant wholesaler establishments. This finding tied in with a related result on a per capita sales basis reported in the previous section. One marginally significant correlation coefficient was also found. This was between the dependent variable and the independent variable, per capita drug manufactur- ing establishments. It indicated association between the per capita number of establishments in the manufacturing and merchant wholesale sectors of the drug channel. But such a relationship did not exist with respect to the drug merchant wholesaling and retail sectors. None of the other independent variables had a correlation coefficient that was significant. Mean Sales per Drug_Merchant Wholesaler Establish- mgpt.--The single correlation coefficients between the dependent variable, mean sales per drug merchant whole- saler establishment, and the selected independent vari- ables are presented in Table 5.9. 113 Table 5.8. Correlation Coefficients (r) Between Mean per Capita Drug Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean per Capita Mean per Mean per . Capita Drug Capita Drug Aléegggggilty Year Manufacturing Retailing Wholesaling . * . Establishments Establishments Establishments** 1948 .6745 .4242 .9106 1954 .7531 .1471 .8732 1958 .6554 .0297 .8371 1963 .6548 -.0612 .7458 1967 .6151 -.0622 .7980 Mean er Po ulation Mean .p p Nonagricultural Year Capita per Employment Personal Square as Per cent of Income Mlle Total Employment 1948 .7484 .3425 .7042 1954 .6900 .5268 .6620 1958 .6178 .4358 .5178 1963 .4408 .4839 .4038 1967 .4573 .4622 .4070 Mean per Capita A11 Commodity Year Mean per Capita All Commodity Manufacturing Retailing Establishments Establishments 1948 .6571 .6890 1954 .6635 .5470 1958 .5535 .4356 1963 .5292 .0545 1967 .4830 .0971 *Marginally significant. **Significant Table 5.9. 114 Correlation Coefficients (r) Between Mean Sales per Drug Merchant Wholesaler Establish- ment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean Value Mean Sales Mean Sales per Added per All Commodity Year Drug gziaiiig Merchant Manufacturing Establishmgnt Wholesaling Establishment Establishment 1948 .1645 .3701 -.2165 1954 -.0323 .7630 -.2317 1958 -.5118 .6059 —.6968 1963 -.1167 .3075 -.2405 1967 .1067 .2218 -.0079 Mean per Population Nonaggizfiltural Year Capita per Employment Personal Square as Per cent of Income Mlle Total Employment 1948 -.2393 -.3528 -.2693 1954 .4557 -.3004 .2760 1958 -.0164 -.6199 —.0883 1963 .3297 -.l423 .0531 1967 .3903 .0174 .1233 Mean Value Mean Sales Year Added per per All Commodity All Commodity Manufacturing Retailing Establishment Establishment 1948 .1420 .0352 1954 .1928 .7552 1958 -.3484 .4433 1963 -.0051 .4338 1967 .1212 .4783 115 None of the correlation coefficients was signifi- cant for the study period 1948-1967. Multiple Correlation Coefficients A selected group of multiple (linear) regression analyses were conducted. The combinations of independent variables selected were presented in Chapter II. The pur- pose of the analyses was to find those combinations of independent variables that provided significantly better association with the dependent variables than did single independent variables. In the drug channel none of the combinations of independent variables provided significantly higher (multiple) correlation coefficients. Changes in the Regression Coefficients As stated in Chapter I, the second hypothesis concerned changes over time of the relationship between the independent and dependent variables studied. In this section, the regression coefficients and coefficients' standard errors of the drug channel relationships are analyzed. Per Capita Drug Merchant Wholesaler Establish- ments' Sales.--The regression coefficients and the coef- ficients' standard errors between the dependent variable, 116 per capita drug merchant wholesaler sales, and the selected independent variables are presented in Table 5.10. None of the E values was statistically signifi- cant. Only one exceeded Il.00l. The t value of the change in regression coefficient between the dependent variable and the independent variable, per capita drug retail establishments' sales, equaled -1.43. This decrease in regression coefficient might have reflected the "scrambled merchandising" in the retail sector, that is, commodity lines being handled by more types of retail outlets. The figures in Table 5.2 support this belief. While the value added and sales of the drug manufacturing and retail sectors increased 514 and 247 per cent respectively, the sales of the retail sector increased only 172 per cent. The increased output of the manufacturing and wholesaling sectors had to be disposed of somewhere. Per Cgpita Drug Merchant Wholesaler Establish- mg§t§.--In Table 5.11 the regression coefficients and coefficients' standard errors are presented for the dependent variable, per capita drug merchant wholesaler establishments, and selected independent variables. None of the changes in the regression coefficients was statistically significant. However, four of the regression coefficients indicated changes with E values that were close to or greater than 1.00. These regression 117 Table 5.10. Estimated Regression Coefficients (B) and Standard Error of Coefficients (3h) Between Mean per Capita Drug Merchant Wholesaling Establishment Sales and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. _t_ Value of 19“ to 1967 Shift in Regression Coefficient. (Critical Value of _t_ - 2.12) 3 . -.-4 z a» u e u e - e o e e o —:c c e e . u n e .vu\e a! e-u u 8 u a ~4 u e u e -auu “cu «fidfl « a 0 0 -uaou fix a do: an: no fi' no « QUCC nu c as 2 ed em 0 e: e have new 3 ea U: . 0: 0: :1 6. s: :25: U s a U . Yur been kg: “32”” an 3: Egon u 3:: u 3: :::< 1.. 8.... a: .. 2 . a .. a..- DO A. UACI 0 MN Hfld Duns UADD some case a uwu ca :3 some a as c see 32:: 32:: ~:~:= :: 8. ':.: 3:5:: :2::: ZOH> 0mm 2‘30: :0. mm 2004!? 2 II> t‘dmm - 6' — — - B 03 a a a 0B 8 a B aa 5 08 B 08 3 aa 1948 .072 .065 .778 .429 .734 .196 1.064 .208 .059 .085 1.139 .314 .236 .156 1.108 .304 1954 .065 .057 .560 .492 .829 .173 1.039 .249 .087 .080 1.409 .478 .302 .159 1.171 .403 1958 .025 .047 .446 .516 .761 .128 .790 .259 .054 .069 1.259 .631 .186 .184 1.109 .428 1963 .046 .042 .083 .452 .740 .125 .733 .276 .086 .059 1.165 .754 .226 .172 .873 .483 1967 .062 .046 -.117 .463 .771 .132 .889 .316 -099 .059 1.854 1.126 .255 .184 1.022 .590 :48-67'-'13 t‘8_67--1.43 ‘48-67-+’16 t‘8_67--.46 ‘48-67- +.39 :48-67-+'61 t48-67-7'08 ‘48-67--'13 Table 5.11. Estimated Regression Coefficients (a) and Standard Error of Coefficients (33) Between Mean per Capita Drug Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954. 1958, 1963, and 1967. 5 Value of 1948 to 1967 Shift in Regression Coefficients. (Critical Value of t - 2.12) m c 7 '7 e u E e 1.: 2 e e 31;“ 3:». U 0~4 H g H 5 U U D U D -daJu d CaJ w Arid -d a 0 O a aura d > u on: mac muofi do a mac: mu : “°§ ..§ 6”§. as .2 82:2 gang :8 ° N -« ye“: Us on s 04 20.0 3 m5 use nun u an ua a: so u an u an a 7 8““ 1 ‘ a” “ 2 X ”7 a 7“ 2a me one a as u see OHM n once 0 an ace we UMD ctrfi c(re : deu c e o g : ea.e c a c s a 8’“ :3: ::::= a: 8 33.: :25: 32:: u e tom tom sexes In 23 gush x Hmowuwuuv .mucwwowmmmou coflmmwumwm cw uuficm road 0» mvma mo wsam> M .homa can .moma .mmma .«moe .mvma mama» «so you mmHQMeum> acmocmawcce omuomamm can unmeameabmumm sceammmaozs neurone: memo you modem coo: comzuom Ammv mucwfloflmuoov mo wouum oumocmuw pom Amy mucmfiowwwooo coflmmoumwm oquEMumm .ma.m manna CHAPTER VI FINDINGS: THE LIQUOR CHANNEL In this chapter characteristics of the liquor merchant wholesaler channel sector were compared to selected characteristics of the liquor manufacturing and retail sectors. The merchant wholesalers' characteristics are also compared to selected economic characteristics. Liquor Commodity Line Merchant Wholesaler Characteristics The characteristics of the liquor merchant whole- saler sector selected for study were the total dollar sales, total establishments, and mean dollar sales per establishment. The levels of each are presented in Table 6.1 for the census years during the period 1948 to 1967. The index numbers indicate the relative level of the individual characteristics with respect to the base year, 1948. Their use provided a method of comparing the relative changes that took place between characteristics, channel sectors, and channels. The level of liquor merchant wholesale dollar sales grew 146 per cent during the study period. While 120 121 Table 6.1. Selected Characteristics (and Index Numbers) of the Liquor Merchant Wholesaler Sector of the Liquor Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Liquor Total Liquor Mean Liquor Merchant Merchant Merchant Wholesaler Wholesaler Wholesaler Year Sales Establishments Sales per 000,000 Establishment (Index Number) (Index Number) (Index Number) 1948 $2518 (1.00) 1527 (1.00) $1,648,984 (1.00) 1954 3376 (1.34) 1518 (0.99) 2,223,978 (1.35) 1958 3799 (1.51) 1446 (0.95) 2,627,247 (1.59) 1963 4810 (1.91) 1470 (0.96) 3,272,108 (1.98) 1967 6119 (2.46) 1486 (0.97) 4,117,765 (2.50) the number of establishments remained about constant, the sales per establishment grew 150 per cent during the period. A comparison of the liquor and all commodity mer- chant wholesaling sector index numbers revealed a trend of relative concentration. The all commodity and liquor merchant wholesalers had approximately equal sales index numbers (Table 6.2). But, while the number of all com- modity merchant wholesalers increased 65 per cent, the number of liquor merchant wholesalers decreased 3 per cent. Consequently, the sales per liquor merchant wholesaler increased 150 per cent compared to 63 per cent for the all commodity merchant wholesaler. 122 Table 6.2. The Index Numbers of Selected Characteristics of the Manufacturing, Merchant Wholesaling, and Retailing Sectors of the A11 Commodity and Liquor Merchant Wholesaler Distribution Channels for the Years 1948, 1954, 1958, 1963, and 1967. Manufacturing Total Total Value Added per Year Value Added Establishments Establishment All . All . All . Commodity Liquor Commodity Liquor Commodity Liquor 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.58 0.78 1.19 0.61 1.32 1.27 1958 1.90 0.96 1.26 0.56 1.51 1.71 1963 2.58 1.30 1.29 0.51 1.99 2.54 1967 3.54 1.59 1.29 0.49 2.15 3.24 Merchant Wholesaling Sector Total Total Sales per Year Sales Establishments Establishment All Li uor All Li uor All Li uor Commodity q Commodity q Commodity q 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.33 1.34 1.27 0.99 1.03 1.35 1958 1.60 1.51 1.47 0.95 1.08 1.59 1963 2.06 1.91 1.62 0.96 1.27 1.98 1967 2.69 2.46 1.65 0.97 1.63 2.50 Retailing Sector Total Total Sales per Year Sales Establishments Establishment All . All . All . Commodity Liquor Commodity Liquor Commodity Liquor 1948 1.00 1.00 1.00 1.00 1.00 1.00 1954 1.31 1.23 0.98 0.94 1.30 1.32 1958 1.54 1.63 1.02 1.11 1.47 1.47 1963 1.89 2.01 0.96 1.20 1.88 1.68 1967 2.39 2.58 1.00 1.19 2.32 2.18 123 Liquor Commodity Line Manufacturing Sector Characteristics The characteristics of the liquor manufacturing sector are presented in Table 6.3. This was the only manufacturing sector studied that showed a drop in the level of manufacturing value added from the 1948 to the 1954 census. Between 1948 and 1954 the level of value added dropped 22 per cent compared to an increase of 58 per cent for all manufacturing. And in 1967, the level of value added was only 59 per cent above the 1948 level. The number of liquor manufacturing establishments in 1967 were less than one-half the number in 1948. Be- cause of the drOp in number of establishments, the value added per establishment rose 224 per cent during the study period, compared to 115 per cent for all manufacturing establishments. Liquor Commoditnyine Retail Establish- ments' Characteristics The characteristics of the liquor retail sector are presented in Table 6.4. The level of sales increased throughout the study period. Although the number of establishments fluctuated during the period, the trend seemed to be slightly upward. In addition, the level of sales per retail establishment for this commodity line rose. Comparison of the liquor retail sector with the all commodity retail sector revealed that both had about 124 Table 6.3. Selected Characteristics (and Index Numbers) of the Liquor Manufacturing Sector of the Liquor Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Liquor Total Mean Liquor Manufacturing Li uor Manufacturing Year Establishments q . Value Added Manufacturing Value Added Establishments per 000,000 Establishment (Index Number) (Index Number) (Index Number) 1948 $586 (1.00) 644 (1.00) $ 909,937 (1.00) 1954 455 (0.78) 395 (0.61) 1,151,898 (1.27) 1958 560 (0.96) 361 (0.56) 1,551,246 (1.71) 1963 761 (1.30) 329 (0.51) 2,313,069 (2.54) 1967 934 (1.59) 317 (0.49) 2,946,372 (3.24) Table 6.4. Selected Characteristics (and Index Numbers) of the Liquor Retailing Sector of the Liquor Distribution Channel for the Years 1948, 1954, 1958, 1963, and 1967. Total Liquor Mean Liquor Retailing {itiir Retailing Year Establishments' 9 . Sales Retailing Sales Establishments per 000,000 Establishment (Index Number) (Index Number) (Index Number) 1948 1954 1958 1963 1967 $2580 3171 4202 5191 6662 (1.00) (1.23) (1.63) (2.01) (2.58) 33,422 31,240 37,068 40,188 39,619 (1.00) (0.94) (1.11) (1.20) (1.19) $ 77,194 101,504 113,359 129,167 168,151 (1.00) (1.32) (1.47) (1.68) (2.18) 125 the same rate of sales increase. The number of liquor retail establishments increased compared to the number of all commodity retail establishments. Therefore, the relative level of sales per establishment rose faster for the all commodity retail sector. Mean per Capita Liquor Establishments' Sales or Value Added and Establishments To offset the differences in the total levels of activity among the various geographic divisions, the least-squares analysis of the variables was conducted on a per capita basis. Tables 6.5 and 6.6 show the sales or value added and establishments arrived at through LS. Results of the Liquor Channel Study Regression Coefficients (B) As was stated earlier (in Chapter II), the signifi— cance of the association between the dependent and inde— pendent variables was determined by analysis of the corre- lation coefficients. Per Capita Liquor Merchant Wholesaler Sales.--The simple correlation coefficients between the dependent variable, per capita liquor merchant wholesaler sales, and the selected independent variables are presented in Table 6.7. 126 Table 6.5. Per Capita Sales or Value Added (and Index Number) for Liquor Manufacturing, Merchant Wholesaling, and Retailing Establishments for the Years 1948, 1954, 1958, 1963, and 1967. Mean Mean per Mean per per Capita Capita Sales Capita Value Added by Liquor Sales by by Liquor Merchant Liquor Year Manufacturing Wholesaling Retailing Establishments Establishments EstabliShments (Index Number) (Index Number) (Index Number) 1948 $4.01 (1.00) $17.25 (1.00) $17.67 (1.00) 1954 2.83 (0.71) 20.97 (1.22) 19.70 (1.11) 1958 3.22 (0.80) 21.83 (1.27) 24.15 (1.37) 1963 4.03 (1.01) 25.45 (1.48) 27.47 (1.55) 1967 4.72 (1.18) 30.90 (1.79) 33.65 (1.91) Table 6.6. Liquor Manufacturing, Merchant Wholesaling, and Retailing Establishments (and Index Numbers) for the Years 1948, 1954, 1958, 1963, and 1967, per 1,000,000 Population. . Mean Liquor . Mean Liquor Mean Liquor . Merchant . . Manufacturing . Retailing . Wholesaling . Establishments . Establishments Year Establishments per 1,000,000 er 1 000 000 per 1,000,000 Population pP ’ .’ Population opulation (Index Number) (Index Number) (Index Number) 1948 4.41 (1.00) 10.46 (1.00) 228.92 (1.00) 1954 2.45 (0.56) 9.43 (0.90) 194.04 (0.85) 1958 2.08 (0.47) 8.31 (0.79) 213.03 (0.93) 1963 1.74 (0.39) 7.78 (0.74) 212.63 (0.93) 1967 1.60 (0.36) 7.51 (0.72) 200.10 (0.87) 127 Table 6.7. Correlation Coefficients (r) Between Mean per Capita Liquor Merchant Wholesaling Establish- ment Sales and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean per Capita ”fan per M5an P9r A11 Commodity Capita Liquor Capita Liquor Year . . . Merchant Manufacturing Retailing . . . . Wholesaling Establishments Establishments Establishments' Value Added Sales* Sales* 1948 -.0048 .6208 .7398 1954 .0263 .7220 .7995 1958 .0866 .7402 .8746 1963 .0198 .8176 .7536 1967 -.0793 .7444 .6157 M an er Po ulation Mean e .p p Nonagricultural Capita per Year Employment Personal Square t f Income** Mile as Per cen 0 Total Employment** 1948 .9018 .3941 .9119 1954 .9585 .5177 .9195 1958 .9478 .4666 .8011 1963 .9269 .4822 .8356 1967 .8988 .4584 .8231 Mean per Mean per Capita Capita Year All Commodity All Commodity Manufacturing Retailing Establishments Establishments' Value Added Sales* 1948 .6594 .8078 1954 .7545 .8265 1958 .6346 .8260 1963 .5344 .7996 1967 .4268 .7942 *Marginally significant. **Significant. 128 Two significant associations were noted for the study period. The dependent variable exhibited a meaning- ful association with the independent variables, per capita personal income and per cent nonagricultural employment. And these correlations were positive, indicating a positive association between the variables involved. Three marginally significant associations were also noted for the period. The dependent variable was margin— ally associated with per capita liquor retail sales, all commodity merchant wholesaler sales, and all commodity retail sales. In addition to the above, a large number of posi- tive correlation coefficients between per capita liquor merchant wholesaler sales and indicators of economic activity were found. This would seem to indicate a strong association between the two. Per Capita Liquor Merchant Wholesaler Establish- ment§.--The simple correlation coefficients between the dependent variable, mean per capita liquor merchant whole- saler establishments, and the selected independent vari— ables are presented in Table 6.8. Though there were no significant correlation coefficients for the study period, there was one margin- ally significant association. This was the association between the dependent variable and the independent variable, per capita income. Standing alone, this finding 129 Table 6.8. Correlation Coefficients (r) Between Mean per Capita Liquor Merchant Wholesaling Establish- ments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean per Capita Mean per Mean per . Capita Liquor Capita Liquor Al; Commodity . . . erchant Year Manufacturing Retailing . . . Wholesaling Establishments Establishments . Establishments 1948 .3645 .5629 .5236 1954 .1967 .3782 .5528 1958 .0215 .2842 .3207 1963 —.0230 .3691 .4328 1967 -.0917 .1980 .4951 Mean er Po ulation Mean Y .p p Nonagricultural ear Capita per Employment Personal Square t f Income* Mile as Per cen 0 Total Employment 1948 .7743 -.0551 .7669 1954 .7774 -.0356 .5872 1958 .5835 -.1891 .3585 1963 .6331 -.1880 .4375 1967 .7291 -.0080 .4251 Mean per Capita Mean per Capita Year All Commodity All Commodity Manufacturing Retailing Establishments Establishments 1948 .5122 .3458 1954 .5511 .4828 1958 .2938 .0719 1963 .3493 .0069 1967 .3130 -.0172 *Marginally significant. 130 did not merit significance. However, it did seem to be related to the findings that were reported in the previous and subsequent sections of this chapter. Simply stated, this finding indicated a positive association between liquor channel related economic activities and the level of economic activity in a region. Mean Sales per Liquor Merchant Wholesaler Estab- lishment.--The single correlation coefficients between the dependent variable, mean sales per liquor merchant whole- saler establishment, and the selected independent variables are presented in Table 6.9. None of the correlation coefficients was significant for the study period. Multiple Correlation Coefficients A selected group of multiple (linear) regression analyses were conducted. The combinations of independent variables selected were presented in Chapter II. The pur- pose was to find those combinations of independent vari— ables that provided significantly better association with the dependent variables than did single independent variables. In the liquor channel, however, none of the combinations of independent variables provided signifi— cantly higher (multiple) correlation coefficients. Table 6.9. 131 Correlation Coefficients (r) Between Mean Sales per Liquor Merchant Wholesaling Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. Mean Value Mean Sales Mean Sales per Added per L' All Commodity Year Liquor pgr iquor Merchant . etailing . Manufacturing Establishment Wholesaling Establishment Establishment 1948 -.0325 .0084 .4640 1954 -.3891 -.0177 .0439 1958 -.3015 .0807 .2976 1963 -.0384 .1283 .2320 1967 -.0625 .2872 .0546 Mean per Population Mean . Nonagricultural Year Capita per Personal Square Employment Income Mile as Per cent of Total Employment 1948 -.l497 .4990 -.1284 1954 .3928 .6557 .5313 1958 .4755 .6026 .5061 1963 .3480 .6788 .4837 1967 .6040 .5807 .7272 Mean Value Mean Sales Added per per Year All Commodity All Commodity Manufacturing Retailing Establishment Establishment 1948 .1848 -.4296 1954 .2774 -.0653 1958 -.0165 .0600 1963 -.1104 .1008 1967 -.l621 .4334 132 Changes in the Regression Coefficients As stated in Chapter I, the second hypothesis concerned changes over time of the relationship between the independent and dependent variables studied. In this section the regression coefficients and coefficients' standard errors of the liquor channel relationships are analyzed. Per Capita Liquor Merchant Wholesaler Establish- ment Sales.--The regression coefficients and the coefficients' standard errors between the dependent vari- able, per capita liquor merchant wholesaler sales, and the selected independent variables are presented in Table 6.10. None of the E values was statistically significant. However, three of them did have E values close to 2.00. These were for the independent variables per capita per- sonal income, per cent nonagricultural employment, and per capita all commodity retail sales. All three of the regression coefficients were increasing during the study period. The level of all commodity and liquor retail establishment sales rose 91 and 75 per cent, respectively (Tables 4.5 and 6.5). The relatively slight difference in growth indicated that changes were occurring in the liquor merchant wholesaler sales patterns. 1L33 cam. mem.~ mmm. mqv. Hmv.e am¢.m was. man. «mm. mmo. mas. cam. 3mm. one. mmo. meo.- some 5mm. Hmm.a mvm. mac. mmh. omm.~ omo. and. vmm. oov.a mom. mmm. mom. mmm. one. coo. moma nmv. hmm. vmm. wmv. 5mm. ow¢.~ vmo. and. oma. oov.a mmm. mmo.a mmm. ems. moo. mac. mmma mmo. omm.a med. mvm. Ham. -v.~ moa. moa. mma. avo.a mmm. mmo.a cam. 5mm. mmo. moo. vmma mvm. Hm~.H va. vmm. new. vmv.a Hmo. voa. mmm. mmm.H mmm. mmm. vom. wmw. mmo. Hoe.) mvma mm m mm m mm mm m mm m mh. m mm m mm m SHHVN ASWVW 193W Sd dw msWVW 331W ATTW 958.13 959.18 OSDB b0 88 090.18 EST.8 PUTS 7.3.47.9 Iq+u In: 3 1 e n.o 1 e u 1.37.? Iqab e Tabb e 899 u no.“ U Pdeu en su 3T.D U aenu n nU SQTD SQJD TOT 11 o squ SQO 030 T.Inud 4L9 OAu .1 N a 9 u.a .6 enud 171d Send T:I.W.e v.14uw a H coo 3 9 a U m.e I. .u v.4 a csu J Pe.3 J mqsmnu W.} 1.3 c534+m.4 sud: Pn.w.1 0.60 0.an dado. To 99 0 Us 0.0.? m 0.3 SNIPS IUTB Tu IO T.3MD.3 m33 STUD umm> a 1:9 0.81.1.9 0301 8 UP 89.0.1.9 89? 0.1.“? U 3.0 UU3d rA .AT. .0 3d 50.0.4.6 UI.d 53d 3 KI. 36K? moms a o? IIKT. 31? HE? S 3 s 3 31:8.“ 1 m.4 Ila 3 8 14+ mnu3 o P P U UT. 89 S. 9 .UP 832 3 3. U. 5 UH _ .41 s _ .NH.~ n m no msam> Hmueueao. .mucmaoeoemoo coemmmuomm an Steam some 02 meme do maem> m .noae new .mema .mmma .emme .mvma meow» one HON mmflbmeum> ucmncmamocH omuumamm 6cm woemm ucmezmeanmumm mcaemmmaors ucmguumz wooded muaamu Hod coo: cwwzuom Ammv mucmfiofimwwoo mo nouum pmuocmum pom Amy mucofloflwumoo coflmmmummm Uwumeflumm .oa.w manme 134 Per Capita Liquor Merchant Wholesaler Establish- ment§.--In Table 6.11 the regression coefficients and coefficients' standard errors are presented for the dependent variable, per capita liquor merchant wholesaler establishments and selected independent variables. None of the changes in the regression coefficients (or E values) was statistically significant. Two of the E values were greater than [1.00]. These, the regression coefficients of the dependent variable with respect to the dependent variables, per capita liquor retail and manufacturing establishments, were decreasing. Mean Sales per Liquor Merchant Wholesaler Establishment.--The regression coefficients and coef— ficients' standard errors between the dependent variable, sales per liquor merchant wholesaler establishment, and the selected independent variables are shown in Table 6.12. One statistically significant E value was found. The regression coefficients between the dependent variable and the independent variable, per cent nonagricultural employment, rose from -.l92 in 1948 to 3.865 in 1967. In addition, large positive increases, though not statistically significant changes, occurred in the regression coef- ficients of the independent variables, per capita personal income and sales per all commodity retail establishment. The above changes indicate that the dependent variable, mean sales per liquor merchant wholesaling 135 Table 6.11. Estimated Regression Coefficients (B) and Standard Error of Coefficients (3b) Between Mean per Capita Liquor Merchant Wholesaling Establishments and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. 5 Value of 1948 to 1967 Shift in Regression Coefficients. (Critical Value of g - 2.12) u H: u :1: o 1: nu u s.» s c s o e e o .4 c c e 3 :23 3:: 2.3: :3 g 32:; a..: .. 3 0.0.... 0.“: 0.430 0.0 --4>. (Lac: Q...) : .25-d es ew-a ‘3 e: 0 1.000 own-«0 0*! U H 005 0 £44 UH EH tau-«ca 0‘09: 0 E E e e o- s acid a u 8 MK. 1.. an: Lav-4 ~41: 6:505 in an N an a a a a a c x6 u o a ow & a“ ~- "c °2z. s :2 =-z. Us; 8:“ c3“ :85 coouu so as came 5 as c 33 ... v“ ‘:::= a: g” 33.3 :=: "2.: 235 is; 2...: .. .3 .... 2.2. i... B as B on B on B as B 03 B 03 B on B 08 1948 .171 .165 .569 .316 .720 .443 1.413 .437 -.019 .134 1.646 .520 .582 .369 1.622 1 664 1954 .047 .088 .288 .267 .659 .375 1.080 .330 -.009 .098 1.256 .654 .454 .260 1.384 .949 1958 .004 .077 .199 .253 .400 .446 .711 .374 -.042 .082 .87] .859 .232 .286 .223 1.170 1963 -.005 .079 .326 .311 .655 .515 .943 .371 -.044 .087 1.329 1.033 .301 .305 .032 1.752 1967 -.017 .070 .141 .264 .725 .481 .968 .343 -.002 .076 1.622 1.305 .275 .315 -.046 1.004 t48-67"1'°5 t‘8_67--l.04 t4a-e7'*'°° t48-67'I'80 t‘._67-+.ll t‘8_67--.01 t‘._67--.65 t48-67"'86 Table 6.12. Estimated Regression Coefficients (B) and Standard Error of Coefficients (33) Between Mean Sales per Liquor Merchant Wholesaling Establishment and Selected Independent Variables for the Years 1948, 1954, 1958, 1963, and 1967. 5 Value of 1948 to 1967 Shift in Regression Coefficients. (Critical Value of E - 2.12) ' u 5 3 3 g u f e 0 L c C g-u ” .. ~:. 3.2: :3 ~ BSEE 3 .. ~. . 3 cc 8.: egg Q0 8 a» e c: a» c a o e o -« e c o u 01:0 .4 s -« 0 Year 01.35 was no u put ca vac-4 0 H2 at! E a o s o e e ow. e 0 5L): 0 a days as» .4 an: Lav-4 w: c503 us an '4 cu: “0'0" Q 1'4 8g: 8': U g “'40... ‘3‘.“ >we~ mud so so u >~s~ m «d Aug 0% on: 0 an HUM (“fl Ufig : a c a 5 o-«.: c e a 5 ea.s : a S e z~53 3F: .2::: a: 8; .:.: “~=: .23: :3. u :31.) t wvmm. mmoo. mva.I thm. mva va umm 32 HmQEDA Mom mmamm m N H m N H xwu xmu xMH x x xwu m N Howflosum mamcflm mamwuasz ucmwcmmmbcH can va ucmbcmmwo mo coaumcflneou mucmfloflmwmou coflumamuuou mammam can meanness .bmma can .mmma .mmma .vmma .mvma muse» 0:» 00m Hmccmno goofing on» “Om mamumummwm bmumofimcoo mucmNOHmmmoo coaumamunoo mannanm> ucwbcmmmwcH may mo comm on Howummsm mauQMUNMHcmflm mquHOHmwooo coaumHmuuoo mUH>oum ance chHumcHQEou magmaum> uchcmmmbcH mamwuasz .oa.n manna 150 section the regression coefficients and coefficients' standard errors of the lumber channel relationships are analyzed. Per Capita Lumber Merchant Wholesaler Establish- ment Sales.--The regression coefficients and the coef- ficients' standard errors between the dependent variable, per capita lumber merchant wholesaler sales, and the selected independent variables are presented in Table 7.11. None of the E values was statistically signifi— cant. However, one E value was -1.52. This was the relationship between the dependent variable and the inde- pendent variable, mean per capita lumber retail establish- ments' sales. This fluctuating regression coefficient was +.330 in 1948 and —.751 in 1967. During the study period small decreases in the regression coefficients for relationship between the independent variable and the dependent variables, per capita all commodity merchant wholesaler establishments' sales, and per capita personal income, were noted. Such declines may have been indica— tive of the relative change occurring in the lumber channel. Per Capita Lumber Merchant Wholesaler Establish- ments.--In Table 7.12 the regression coefficients and coefficients' standard errors are present for the dependent 151 Table 7.11. Estimated Regression Coefficients (B) and Standard Error of Coefficients (38) Between Mean per Capita Lumber Merchant Wholesaling Establishment Sales and Selected Independent Variables for the Years 1948. 1954, 1958, 1963, and 1967. 5 Value of 1948 to 1967 Shift in Regression Coefficients. (Critical Value of t - 2.12) n e NU 1 s c u c u u u s+20 s c- s o e s 2 ~ c c s s 110% awn U 6314 al Is 304.3 a! D a, m .41 w-QHU «ca—4.4 «no 8- 050 «>011 -—4>~ u mum ow: muons do ~4> x guns as c “2:8 83" 8"‘f32 85 CS 8222 88'2” 86 3 Year Use-o 0.5 as o-« 00.00. :15 0nd 03:01! Nil.“ La 3mm 5'; 3:: £505 ragga 3%.” 8.5.3“ 8.3.: 8.000" 0.: so g h dos—4 GOA-4 «me one once 0 an HOH nun own cg : a so c ow» :e as uses a as : ss '° 2‘: 33:: 3:32;: 2%.: a” 33.2; :25: 3:2: £4-4) 1.4mm t‘ZME to. AS £0564 x~~ «o a 0:0 -~>mu w» u Qua aw: ngfi do -a x Qucc do a saw sso sa sc 0 nauo mews we 0 06H UuE U‘U u ow CH D—‘CH trous- 00 E Year «9.3 or: e o---« scion. :! 0 HI emu: Inge-lid tar-4 -—4]: C506 Boga-003 pace 0 U a «d 8. C all 0 o m 8- out 0,-4.4 one ea oso : so 2 u os~ QOHH em on once 0 an ~04 Dug veg cg cgs : uwu cm as asas c a c s s m s u adudc su as as u sac» sauu o c e a sauna so 0v some case mass 24“ In!“ 24:40 :0- sun tuss- I‘lld I¢¢N B 08 B GB B 08 B 08 8 ca B ”B 8 08 B 08 1948 .028 .145 -.260 .230 .460 .372 .456 .517 .120 .095 .794 .566 .428 .298 - .533 1.386 1954 .099 .110 -.175 .197 .248 .321 .313 .371 .049 .071 .525 .565 .263 .208 - .079 .803 1958 .063 .095 -.291 .163 .035 .318 .170 .305 .027 .056 .496 .593 .148 .194 -1.289 .625 1963 .028 .082 -.164 .137 .132 .277 .210 .240 -.004 .043 .462 .537 .090 .156 -1.560 .627 1967 .036 .079 -.206 .111 .113 .293 .157 .262 .012 .041 .896 .693 .130 .170 - .502 .503 t‘a_67-§.05 t‘8'67-+.21 t‘a-67--e73 £48-67--051 t‘8-67--1.05 t‘a-67-+ell tr1-67--.87 t‘8-67.+e2° 152 variable, per capita lumber merchant wholesaler establish- ments, and selected independent variables. None of the changes in the regression coefficients (or E values) during the study period was statistically significant. Only one of the E values was greater than [1.00 . The shift of the regression coefficient between the dependent variable and the independent variable, popu- lation per square mile, and a E-value of —1.05. Mean Sales per Lumber Merchant Wholesaler Estab— lishment.--The regression coefficients and the coefficients' standard errors between the dependent variable, sales per lumber merchant wholesaler, and the selected independent variables are presented in Table 7.13. None of the shifts in the regression coefficients was statistically significant. Two of the E values were greater than |l.00 . These were the regression coef- ficients between the dependent variable and the independent variables, pOpulation per square mile (E = 1.48) and sales per all commodity retail establishments (3 = -l.39). The increase of the regression coefficient, in the case of population per square mile, left it with a very small value (.036), indicating that the importance of the independent variable, with respect to sales per lumber merchant whole— saler establishment, was much reduced. The decrease in the level of the regression coefficient, sales per all commodity retail establishment, suggested that this am.H-uae-mvu mm.+nho-mvu ao.+uamumeu we.a+u u ma.-uaeumeu Nm.+ueoumvu av.+uao-mvu Hm.-uam-mvu va. th. chm. va. ham.d vvv. mmo. mmo. va. mew. mmv. mam. mmm. mac. vmo. mno. mama Nun. mmN.H own. Hom.l baa. NMF. mno. HNo.I hem. awe. mam. mmm. mac. omN. mno. ova. mead mam. ovm.a Hhv. va.I mmm. mmm. mmo. mmo.l vmm. Nam. NNQ. th. who. th. moa. nma. mmwa mom. CNN.N «cc. mmo.l mvm. NNw. Noa. mmo.: mmv. mum. mam. th. one. oav.l mma. wma. vmma mmo. mon.a mmm. mNm.I mmm. com. Hmo. mNH.| «mm. mmm. mom. mac. vvm. oom.| «ma. moa. mvma 153 mo n an m an m mo m mo m mo m mo m mu m BHVW EWdW 193W Sd dW mswvw 31W HJdW s.u7.a nee e.u .ue.n a b o a a a eruvla s 3 ea? aid 3319 3uxe 3 :9 nd 19 uTJIs 3mg 3019 e e u p.n u e.dp.u p.n s u .41.o u e u e.+ 0 0.110 "H1.V 1.8.1 1.l o u Ugo 0.8 nun 1 .17.0 S .Le_T.A .1 N a e u.d Aup.o S 1.1 S .11. A I4rw P 1.3719 3 ago 3 9 e "um e .E P Iéim e su T. 83 T. momu W1. 1.1 33 T. SET. SUQI use» "H.508 U.n3n dad? 1.0 S 09 v.88 urban m 0.9 mloa 10.15 Tu ID 3MPS mews m 18 a nd wmmu. mnmu ad win. “new... as a nu. . 3d UT.d u 3 Re 36 o. moms a or. ITKa 319 «4 up I Pay .eqaa n 1 m 3 11¢ J 1.: u p 1.8 u "er. a e s . u n a 3nv 3 3 . q 6 . P 'A _ ANH.N u u no 0§Hs> HMUNHMMUV .muc0HUwawou Godmwmumwm CH uwwzm hmma Ou wvma mo 05Hm> M .hwmd can .mmma .mmma .emma .mvma muse» ecu HON moanswus> ucmbcwdmucH pneumawm was ucmEQmHHnsumm mcflasmmaocz Beacons: umnEdA use mmHsm cs0: cmm3umm Ammo muc0w0«wwmou mo uouum busvcsum 0cm Amy mucmfiowwmmoo coflmmmummm bmuseflumm .ma.h manna 154 variable was also decreasing in importance as an indicator of trends in the dependent variable, sales per lumber merchant wholesaler establishment. CHAPTER VIII SUMMARY, CONCLUSIONS, AND IMPLICATIONS Summary In Chapter I certain characteristics (sales, establishments, sales/establishment) of wholesalers in the United States were reviewed for the period 1948-1967. Special emphasis was placed on the merchant wholesaler type of wholesale establishments because they are the most significant type of wholesale establishment on the basis of dollar sales. In addition, the Census provided the most detailed data on the merchant wholesaler type of wholesale establishment. Selected characteristics of the consumer good commodity line distribution channel and economy were proposed for possible association with cer- tain characteristics of the merchant wholesaling sector. It was then hypothesized that these selected independent characteristics are associated with the characteristics of the commodity line merchant wholesaler. And, in addition, it was hypothesized that these associations will change over time. The identification of such inde— pendent variables may aid the merchant wholesale manager in planning strategy for the future. Given his unique 155 156 position in service to both manufacturers and retailers, the merchant wholesaler should be aware of any changes in either sector and their implications for his Operation. In Chapter II the geographic division was selected as the geographic control unit. The methodology selected was the linear least-squares analysis and the Zellner- Aitken (ZA) estimators to provide the "best fit" line. The criteria to judge the significance of the findings were also developed. Conclusions and Implication The conclusions and implication were derived from the findings in Chapters III, IV, V, VI, and VII. The conclusions are divided into seven sections. The first two concern the hypotheses proposed for this study, the third, the questions raised in Chapter I in regards to the selected independent variables. The fourth regards the observed differences between channels, and the fifth, the possible implication following from the conclusion in the first four sections. Finally, the last two sections deal with the methodology used in this study. Association of the Merchant Wholesaler Characteristics with Selected Independent Variables Significant associations between certain charac— teristics of the merchant wholesaler and the selected characteristics of the manufacturing and retail channel 157 sectors were found. Also, significant relationships be- tween the merchant wholesaler characteristics and the selected economic characteristics were discovered. The combinations of significant characteristics varied from commodity line to commodity line (Table 8.1). For example, there was a significant association between the per capita drug merchant wholesaling sales and the inde- pendent variable, per capita all commodity merchant wholesaling sales. The association between the per capita liquor merchant wholesaling sales and the same independent variable, per capita all commodity merchant wholesaling sales, on the other hand, was only marginally significant. And the association between the per capita lumber merchant wholesaling sales and the same independent variable was not significant. These different combinations reflect the varied character of the distribution channels in the United States. Such findings also reveal that the merchant whole- salers in each of the commodity line channels studied face a unique situation. Changes in Regression Coefficients Between the Dependent and Independent Variables Shifts (though not all were statistically signifi- cant) were found in the regression coefficients between the dependent and independent variables (characteristics) in every channel studied. The independent variables included the selected characteristics of the manufacturing and 158 u m: “mauasmsaocz ucssousz u 32 “ucsecmuansumm n .umm “snug xuuUOEEOU : .mcflaususm n 8m “mauuuuosmscsx I .EEOU “snug xuubOEEOU Had HH< “xsm .ucmouuucoum.. .ucssuuucmum xaascfiqus2« 1mm.unv..emm em Had usm msasm Aom.nvcseoocu HMCOmusd suudso use A>5.H+Vsmsasm em Had sauaso use Avv.+vsmsasm 3: HH< suudso usm Aon.+vsms~sm 9m uosquq suuasu use Aoo.~+c syncssondEm asusuaso uuumscoz ucso usm Amm.H+vassEOOCH ascomusm suudso usm Aou.u+v ..umm a: msuo spamsu usm RON.+Vss.umm 32 Had suudso use Ama.:vsmsasm Hm HH< suudso use Awe.nvaaseoocH HsCOmusd suudsu use noa.+vssmsamm 32 Had suudso usm ucsEcmansumm use msasm Amu.+v «ucsENOHmEm asusuazo qummcoz ucso use Aoq.uuv..umm u: Had suuasu use Ano.anva«sfioocH ascomusd spudsu use mucsecmuunmumm suumsu use Ava.uvsseoocH HsCOmusd suuQsU use msasm suuasu use usnfisq uosquq msuo >uquEEou Had oflumuusuosusco uouosm sauq >uu©0EEou mauasmsaocz ucscousz Anomaumvma oOuusd qu mucsuouuusou COummsumsm saw no muuucm qu mssas>uuv .hwmalmvma UONusm snu qu usucSum masccsno snug >uu©0EEou scu mo msabsuus> ucsocsdsbcH usussasw scu 0cm muouosm mcuusmsaozz unscousz scu uo msuumuusuosuszo cusuuso css3usm mc0uusu00mm4 ucsouwucmum s59 .H.m sanse 159 retail channel sectors and the economy. The greatest number were found in per capita establishments and sales per establishments. This finding emphasized the shifts of these characteristics in the merchant wholesaler sector relative to the retail and manufacturing sectors. The degree and direction of the shifts varied from channel to channel (Table 8.2). For example, the regression coefficient between per capita drug merchant wholesaling establishments and per capita drug manufacturing establish- ments increased. However, the regression coefficient be- tween per capita liquor merchant wholesaling establishments and per capita liquor manufacturing establishments de- creased. Such variation indicates that an individual analysis would be required for each channel selected for study. Questions Concerning Selected Independent Variables There is diversity and sparseness in the signifi- cant findings discussed in the preceding two sections. This situation makes it impossible to adequately resolve the questions raised in Chapter I concerning the inde- pendent variable selected for the study. Briefly, the various authors suggested that the activity in the whole— saling sector should or should not shift with changes in the manufacturing and retailing sectors. In addition, the relationships between the wholesaling sector and several economic variables were included for consideration. .mCuHususm u Hm noduusuosmscsz u m: “mauasmsaonz ucssousz u 3: uucsfinmuansumm u .umm “snug huuUOEEOU n .5600 “snug xuuUOEEOU Had u Had "xsx 160 ima.u+. meoocu HsCOmusm suwasu usm Ame.u+v .umm em Hum use msasm Amm.anv .umm 9m Aoo.an .umm m2 AHM.H+V .umm u: Had use msasm Had use ususd scas> Had use oscvd ssHs> Amv.a+v sad: sussqm Amn.N+v ucseaoHQEm asusu AHN.H+V sEoocH AHv.H+v sEoocH ucsEcmansumm usm c0uusH500d nasouumscoz ucso usm asc0musm suumsu usm HsCOmusm suHQsU usm use msasm Ron.auv .umm 9m Anm.anv seoocH HH< suuasu usd HsCOmusd suumso use Avo.unv .umm um Auo.uuc .umm em Aom.Nuv .nmm uosquq spudso use msuo suudmo use am suudsu use “mo.alv sauz sussvm Amo.anv .umm m: Aoa.a+v .umm m: on.auv .umm m2 mucsEnmansumm usm c0uusadmom uosquq sawmsu use msuo spudsu use Had suudso usm suumso use A>R.H+v msumm em uu< suudso use loo.~+c unmeuouaem umusu (Hsouumscoz ucso use ANm.HuV msasm am Amm.a+v sEoocH Amv.HnV msasm .umm em usnEsq suumsu use ascomusm suumsu use qsuo suudsu use msasm suumsu use usnESA uosquq msuo xuuUOEEou Add ouumuusuosusno uoussm mcu scuq xuuUOEEoo Iasmsaonz ucszousz .nmma Imvma OOAusm s2» qu msupsum mscuq xuuUOEEou szu mo msansuus> unsucsdsUcH usuosasm was muouosm ocuasmsaonz ucscousz say NO mouumuusuosusnu susuuso css3usm mucsuOuwwsoo COummsumsm scu mo AmsSHs>uu ccsv mumunm usuoz .N.m sanse 161 The following questions were presented in Chapter I with respect to the independent variables selected: 1. Do the selected variables have significant associ- ations with the characteristics of the merchant wholesaling sector? 2. Do these associations shift over time? The first question is answered by the data pre- sented in Table 8.1. Of the seventy-eight associations calculated, six were significant and ten were marginally significant. The sixteen significant or marginally significant were associated with nearly all of the pro- posed independent variables. Two were associated with the all commodity manufacturing sector, three with the all commodity merchant wholesaling sector, and four with the retailing sectors. In addition, five were associated with per capita personal income and two with per cent nonagricultural employment. Therefore, the existence of associations between the characteristics of the merchant wholesaling sectors and the selected independent charac- teristics was established. As can be seen in Table 8.1, the combinations of significant independent and dependent variables varied from channel to channel. As noted earlier, the second question concerned the shifts in the association between the independent and dependent variables. The combinations of independent and 162 dependent variables studied for level of shift were those that had a significant level association with each other. This meant there was a meaningful association between the variables during the period of the study. The Efvalues of these shifts are presented in Table 8.1 along with the significant associations. In Chapter I a principle of marketing was pre- sented. It was: "As changes occur in the retail struc- ture, changes occur in the wholesaling system" (10:64). The findings in the study (Table 8.1) verify this prin- ciple, however, these changes do not follow a single pat- tern. While the E value between the per capita drug merchant wholesaling establishments' sales and per capita all commodity retailing sector sales decreased .13, the association between per capita liquor merchant wholesaling establishments' sales and per capita all commodity retail- ing establishments' sales increased 1.77. As in the case above, the shifts in the regression coefficients between the characteristics of the merchant wholesaling sector and both the characteristics of the manufacturing sector and the economic sector are both positive and negative in direction. Because of the limited number of channels examined, generalizations concerning the shifts in associations could not be made. The economic variable, per capita personal income was the independent variable most closely associated with 163 the characteristics of the merchant wholesaling sector. In three of the four channels studied, this variable had a significant or marginally significant association with per capita merchant wholesaling establishment sales. This finding certainly indicates that there is an association between the level of merchant wholesaling activity and the purchasing power of the consumer population. The economic variable, per cent nonagricultural employment, had a significant association with only one dependent variable, per capita liquor merchant wholesaling establishment sales. This finding does not support the proposition that this measure of economic development is associated with the level of activity in the merchant wholesaling sector. This should not be too surprising for the United States, where in recent years the per cent nonagricultural employment has nearly reached or exceeded 90 per cent in each geographic division. With such a limited variance, it is unlikely that this variable will provide significant levels of associations with the dependent variables. The demographic variable, population per square mile was not significantly associated with any of the selected characteristics of the merchant wholesaling sectors studied. To determine any existing associations between the consumers' location and the characteristics of 164 merchant wholesaling sectors, additional measurements would be needed. The Relative Changes in the Channel Sector Character- istics During the Period 1948-1967 Although it was not a purpose of this study, differences in the growth rate of the channel sectors' characteristics were noted. These differences occurred between channel sectors and commodity lines (Table 8.3). For example, drug manufacturing establishments' value added increased 710 per cent during the study period, while the drug merchant wholesaling sales increased only 150 per cent during the same time. And the number of liquor merchant wholesaling establishments decreased 3 per cent during the period, while overall the number of merchant wholesaling establishments increased 65 per cent. These differences in growth (or decline) rates indicate the magnitude of the differences that exist be- tween the various merchant wholesaling distribution channels. Reasons for these differences may be assigned to changes in the retailing structure. Other changes may follow efforts of large manufacturing and retailing firms to engage in vertical intergration, and often take the form of manufacturers' establishing sales branches or sales offices, thus taking commodity lines away from merchant wholesalers. .muOuosm enuusuoswscss qu Uspps ssam> .muouosm mauusmsaocB pas mauaususu qu msamm asuoes 165 mm.H mm.a vo.a oa.a oo.H mm.o HH.H vm.H mH.H oo.H mm.a em.a mm.H hm.m oo.a usnESA ma.~ mm.a hv.a ~m.a oo.d mH.H om.~ HH.H vm.o oo.H mm.~ Ho.m mo.a m~.H oo.a uoswuq mm.N oa.~ Hm.H om.H oo.~ om.o mm.o Hm.o oo.a oo.~ mn.~ HH.N mo.H Hm.a oo.H msuo ~m.~ mm.H hv.H om.H oo.H oo.H om.o No.a mm.o oo.H mm.~ mm.a wm.~ Hm.a oo.H euuGOEEou Had ocuaumuse mm.a MH.H oo.a mo.H oo.a om.a om.a Hm.a om.H oo.H no.m mo.m Hm.H mm.a oo.H usnEdA om.~ mm.a mm.a mm.H oo.H >¢.o mm.o mm.o mm.o oc.H wv.~ Hm.a Hm.a vm.H oo.H uosvuq om.~ vb.H om.H m~.H oo.H mm.H Hm.a mm.H h~.H oo.H nv.m mm.~ wo.~ mm.H oo.H msuo mw.a h~.H mo.H mo.H oo.H mw.H mo.H hv.H bN.H oo.H mm.~ oo.~ oo.H mm.H oo.a euupoesou HH< ocuasmsaonx ucszous: vo.m vm.a mo.a mm.o oo.H em.o nw.o mm.o vo.H oo.H mH.H vo.a mm.o mo.H oo.H usnEsq vm.m vm.~ HF.H hm.a oo.H mv.o Hm.o wm.o Ho.o oo.H mm.H om.H mm.o mh.o oo.a uosvuq oa.m om.q -.m om.H oo.a mn.o hw.o om.o oo.a oo.a va.w e~.e oa.m mm.a oo.H msuo mH.~ mm.H Hm.a Nm.H oo.H m~.H m~.H m~.H ma.a oo.a vm.m mm.~ om.a mm.a oo.H euuoOEEoo Haé onwuzuumwscsz head moma mmma vmma mvma head mead mmma vmma mvma mama moau mmma vmma mvma stud euupOEEou uss» uss» uss» uouosm unsenmuunmumm mucssrmuunmumm umuos omega ssum> uo msumm umuoe use msasm uo psppa ssas> m .hoaa was .mmma .wmma .vmma .mvma muss» snu u0w masccsnu COuusnuuumuo usnE:q pas .uozquq .msuo .euHUOEEOU HH< snu Mo muouosm mauuususe pas vauasmsaocz ucsnousz .ecflusuomMscsz sap mo mouumuusuosumnu psuosasm mo musnfidz xsch s59 .m.m sauce 166 Possible Implications of the Preceding Findings The associations between the dependent and inde- pendent variables that were significant varied from com- modity line channel to commodity line channel in this study. Several possible explanations for these differ— ences are outlined below. 1. Groups of Commodity Line Channels. There may be groups of commodity line distribution channels whose characteristics have significant associations with the same independent variables. A possible grouping might be made on the basis of the type of goods handled in the channels. For example, the commodity lines might be classified as convenience, shOpping, or specialty goods. The merchant wholesalers handling the commodity lines within a specific classification may share significant associations with the same independent factors. Because of the limited number of commodity lines analyzed in this study, however, it was not possible to test this approach. 2. Unique Commodity Line Channels. It may be that each commodity line distribution channel is unique. That would mean that either each commodity line merchant wholesaling sector has a unique combination of independent factors with which it is associated or that it has such a complex association with the independent factors that it cannot be discovered with present methodology. 167 3. The Independent Variables Selected for Analysis. There is always the possibility that when a finite number of independent variables are selected for analysis, those having significant associations with the dependent variable were not included in the selection. This type of omission can only be discovered by the inclusion of additional factors in future analyses. EFFEST and ZA Estimators EFFEST estimates from the Van Tassel study and the ZA estimates from this study were analyzed. The tech- nique was simply a comparison of the results from the two studies with the least-squares approach on approximately the same data. It was concluded that the results, using these methods, were not correct for two reasons. First, they resulted in biased estimates of the computed values of the regression coefficients. That is, the methods did not provide a better estimate of the pOpulation regression coefficients than was obtained by using least-squares analysis. This does not imply, however, that EFFEST or ZA are inappropriate for all situations. Nevertheless, in situations where the independent variables are highly correlated and/or there is only a single dependent vari- able, the methods should not be applied. In other situ- ations, even when the necessary conditions are met, careful review of the estimates is strongly advised. 168 Second, the conditions for the application of the methodology outlined by Zellner were not met. For these reasons, it was concluded that it is not correct to apply the EFFEST or ZA methods to the linear regression analysis in this study. Geographic Control Units The geographic units used were too large for pre- cise analysis. Larger units were necessary, however, so that some of the desired variables could be included in the analysis. The use of smaller geographic units might provide less variability within units and more variability between units. Suggested Areas for Future Research Academic The differences between the various commodity line merchant wholesaling sectors have been documented in this study. Additional research might be directed toward determining if these differences can be grouped on other bases, such as classification of goods. Additional independent variables might be proposed, and a study could be directed toward determining if sig- nificant associations exist between these variables and the merchant wholesaling sector. Further research might be conducted considering two other changes from the methods used in this study. 169 First, the geographical control units could be reduced in size to provide more variability between geographical units. Second, the manufacturing sector might be dropped so that additional commodity line distribution channels could be analyzed. Because of the limited number of manufacturing establishments in many channels, these channels were not included in this study. Applied At the current state of development, the methodology investigated in this study has limited application to "real world" problems. However, the research does indicate the need for the individual merchant wholesaler to recognize the uniqueness of the channel sector to which he belongs. SELECTED LITERATURE l. SELECTED LITERATURE Books U.S. Bureau of the Census. 1942! 1954, 1958, 1963. ment Printing Office, 1951, 1957, U.S. Bureau of the Census. 1954, 1958, 1963. Retail Trade. Census of Manufacturers, Washington, D.C.: Govern- 1961, 1966. Business, 1948, Washington, D.C.: Census of Government Printing Office, 1952, 1966. U.S. Bureau of the Census. 1954, 1958, 1963. Retail Trade. 1957, 1961, Business, 1948, Washington, D.C.: Census of Government Printing Office, 1952, U.S. Bureau of the Census. 1967. Industry Series. 1957, 1961, 1966. Census of Manufacturers, Washington, D.C.: Government Printing Office, 1970. U.S. Bureau of the Census. 1940, 1950, 1960. Washington, D.C.: U.S. Census of Population, Government Printing OffiCe, 1943, 1952, 1963. U.S. Bureau of the Census. Retail Trade: United States Summary. Census of Business, 1967. Washington, D.C.: U.S. Bureau of the Census. Wholesale Trade: Washington, D.C.: 1970. U.S. Bureau of the Census. the United States: Washington, D.C.: 1948-1970. U.S. Bureau of the Budget. Standards. Washington, D.C.: 1967. 170 1948-1976. Government Printing Office, Government Printing Office, 1970. Census of Business, 1967. United States Summary. Government Printing Office, Statistical Abstract of 69th-9lst eds. Office of Statistical Standard Industrial Classification. Government Printing Office, 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 171 Bartels, Robert. Marketing Theory and Methethgory. Homewood, I11.: Richard D. Irwin, Inc., 1970. Beckman, Theodore N.; Engle, Nathanael H.; and Buzzell, Robert D. Wholesaling. New York: The Ronald Press Co., 1959. Elkblad, Frederick A. The Statistical Method in Business. New York: John Wiley & Sons, Inc., 1962. Ferber, Robert, and Verdoorn, P. F. Research Methods in Economics and Business. New York: The Macmillan Company, 1962. Huang, David S. Regression and Economic Methods. New York: John Wiley & Sons, Inc., 1970. Paramountain, Joseph C. The Politics of Distri- bution. Cambridge: Harvard University Press, 1955. Ruble, William Lewis. "Improving the Computation of Simultaneous Stochastic Linear Equation Estimators.‘ Unpublished Ph.D. dissertation, Michigan State University, 1968. Shultz, William J. American Marketing. San Francisco: Wadsworth Publishing Co., Inc., 1961. Van Tassel, Charles E. An Analysis of Factors Influencing Retail Sales. East Lansing: Bureau of Business and Economic Research, Michigan State University, 1966. Periodicals Cumulative Book Index, (1959-1969). New York: The H. W. Wilson Company. Emmerson, Richard M. "Power-Dependence Relations." American Sociological Review, XXVII (February, 1962), 31-40. LOpata, Richard S. "Faster Pace in Wholesaling." Harvard Business Review, XLVII, No. 4 (July— August, 1969), 130-43. McVey, Phillip. "Are Channels of Distribution What the Textbooks Say?" The Journal of Marketing, XXIV (January, 1960), 61-65. 23. 24. 25. 26. 27. 28. 29. 30. 172 Miller, Nelson A. "Wholesale Centers and Marketing Areas in the United States." The Journal of Marketing, XIV (September, 1949), 156-68. Stigler, George J. "The Division of Labor is Limited by the Extent of the Market." The Journal of Political Economy, LIX, No. 3 (June, 1951), Tallman, Gerald B., and Blomstrom, Bruce. "Retail- ing Innovations Challenge Manufacturers." Harvard Business Review, XL (September-October, 1962), 136-41. U.S. Department of Commerce. "National Income by Industrial Origin." Survey of Current Business, xxx, XXXVI, XLI, XLVI, L (July, 1950; July, 1957; July, 1961; July, 1966; July, 1970). Zellner, Arnold. "An Efficient Method of Estimating Seemingly Unrelated Regressions and Test for Aggregate Bias." Journal of American Statistical Association (June, 1962), 348-75. Miscellaneous Materials Alexander, Ralph 8., ed. Marketing Definitions: A Glossary of Marketing Terms. Chicago: American Marketing Association, 1960. Bucklin, Louis P. "Models of Wholesale Market Structure." Broadening the Concept of Marketing. Boston: American Marketing ASsociation, 1970 Fall Conference, p. 91. Michigan State University Agricultural Experiment Station. Summary of Stat Control Cards and Codes. STAT Series Description No. 4. East Lan51ng: Michigan State University, January, 1970. "‘ssulfiss