AN ANALYSIS or THE DEMAND FOR UQUOR m MICHIGAN Thesis for the Degree of Ph. D. MICHEGAN STATE UNEVERSiTY LED 1. NAVIN 1968 "PL‘n ‘- T“ \m um mm M uxxmumm \mummum u m L E: ‘~ 5 3 This is to certify that the thesis entitled An Analysis of the Demand for Liquor in Michigan presented by - Leo James Navin _ has been accepted towards fulfillment " of the requirements for Ph . D . degree in Economi cs fl/x Q/Q/ ' Major professor Date May 7, 1968 ABSTRACT AN ANALYSIS OF THE DEMAND FOR LIQUOR IN MICHIGAN by Leo J. Navin In view of the lack of information about liquor demand parameters in Michigan, the answers to policy questions which affect and are affected by demand var- iables have had to rely on "best guesses." Reliable revenue estimates of the yield of liquor excise taxes and profits from the State's liquor enterprise (ten per cent of State General Fund revenues in 1966) have been very difficult to make. These factors coupled with the excellent body of available data for a demand study were instrumental in promoting an analysis of the demand for liquor in Michigan. Through the use of quarterly time series data (1955- 1966) and multiple linear regression techniques, this study primarily identified price and income elasticity coefficients for "liquor," "distilled spirits" and "com— ponent" demand functions. Several variations of a basic demand model which included case sales, final price (a Laspeyres chain linked index), Michigan disposable income and dummy variables for seasonal adjustments were tested. Liquor, which included beverages having an alcoholic con— tent of sixteen per cent or more, was found to be slightly price and income inelastic with respect to physical Leo J. Navin volume sales. The price elasticity was not, however, significantly different from one. When actual total liquor expenditures (including taxes) were examined, a significant inverse price-total expenditures relationship emerged. A comparison of total expenditure, price and income elas- ticity coefficients arrived at through the direct regression of actual expenditures on the demand variables and the expenditure elasticities which would normally be inferred from the demand parameters revealed a significant discrep- ancy in the price elasticity coefficients. This was explained in terms of intra-basket substitution. A major portion of this substitution was accounted for when the analysis of demand for spirits over 22 per cent alcohol content was examined. A significant cross price elasticity of demand was revealed between fortified wines and the other liquors. An analysis of the demand for ten categories of liquor was performed. The resulting price and income elasticities were aggregated and compared with the composite liquor demand results as well as receiving brief individual treatment. The financial structure of the Michigan liquor Operation is subsequently analyzed with a view to the ident- ification of the impact changes in tax, mark-up and wholesale cost parameters would have on the price parameter and ulti-. mately State excise and monopoly revenue. A monOpoly profit model is presented along with an excise model. AN ANALYSIS OF THE DEMAND FOR LIQUOR IN MICHIGAN By Leo Jt‘Navin A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1968 ACKNOWLEDGMENTS The following study would not have been possible without the cooperation and assistance of many. I am indebted to the Budget Division of the Bureau of the Budget of the State of Michigan. Mr. Paul H. Wileden, Assistant State Budget Director, and Mr. Gerald Miller, Chief of Research, were extremely helpful in furnishing access to the Bureau's research facilities. Mr. George Burke, Jr., Business Manager for the Michigan Liquor Control Commission and Mr. John Q. Adamson, Comptroller, were COOperative in providing me with first hand knowledge of the merchandising operations of the Michigan Liquor Control Commission as well as assisting me in obtaining the data which were crucial in this study. The encouragement, time and effort of Doctor Paul E. Smith who has served as my thesis director and my major professor in the field of Public Finance, are appreciated. The suggestions and comments of Doctor John Moroney and Doctor Peter Lloyd have contributed to both the development of this topic and to its presentation. Errors and omissions, however, are the sole responsibility of the author. Last, but by no means least, I must thank my wife Joanne for her invaluable aid in preparing the data used in this study as well as ii performing much of the tedious work required in the prepar— ation of this manuscript. To the many other persons who have assisted me in this work and to the faculty of the Department of Economics at Michigan State University-- thank you. iii TABLE OF CONTENTS ACKNOWLEDGMENTS . . . . . . . LIST OF TABLES . . . . . . . . LIST OF APPENDICES . . . . . . LIST OF CHARTS . . . . . . . . INTRODUCTION . . . . . . . . Chapter I. EMPIRICAL STUDIES . . . . Foreign Studies Sten Malmquist A. R. Prest Richard Stone U. S. Studies Harold Wattel William Niskanen Julian Simon Concluding Comments II. MICHIGAN LIQUOR DEMAND . . Introduction The Models Market Defined General Function Composite Demand Distilled Spirits Demand Component Demand Conclusion iv Page ii vi xi 21 Chapter Page III. MICHIGAN LIQUOR REVENUE . . . . . . . . 63 Michigan's State Liquor Enterprise Receipts Expenditures Revenue Model Liquor Excise Taxes Conclusion APPENDICES O O O O O O O O O O. O O O O O 76 BIBLIOGRAPHY O O O O O O O O O O O O O O 121 Table l. 10. ll. 12. 13. 1“. LIST OF TABLES General fund alcoholic beverage revenue State of Michigan l9A6-l966 (in thousands). . . Regression coefficients, demand for liquor in SWGdGH 1923-19140 0 o o o o o o o o 0 Demand analysis for distilled spirits United Kingdom 1920-1938 . . . . . . Alcoholic beverage demand elasticities for the U. S. 193Ll-1960 o o o o o o o o 0 Price elasticities via the quasi—experimental methOd O O O O O O O O O 0 0 I Composite liquor demand equations . . . . . Price correlation coefficients between liquor categories indexes . . . . Composite liquor demand price and income elasticities . . . . . . . . . . A comparison of total expenditure price elasticities directly and indirectly computed O O O O O O O O O O O O O A comparison of income elasticities of the composite demand and the total expenditure functions . . . . . . . . . Distilled spirits demand price and income elasticities . . . . . . . . . . A comparison of distilled spirits total expenditure price elasticities directly and indirectly computed . . . . . . . . Income elasticities of the distilled spirits demand and the total expenditure functions Aggregated liquor demand price and income elasticities . . . . . . . . . vi Page 11 14 l8 19 34 37 42 A4 A6 47 48 52 Table 15. l6. 170 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. Page Liquor demand income elasticities composite and aggregated compared . . . . . . . . 53 Aggregated distilled spirits demand price and income elasticities . . . . . . . . . 55 Discounts as a per cent of total dollar retail sales fiscal 1955-1966 . . . . . . . . 65 Liquor demand equations linear composite . . . 77 Liquor demand equations linear composite using relative price and real income . . . 78 Liquor demand equations linear per-capita composite . . . . . . . . . . . . 79 Liquor demand equations log—linear composite . 80 Liquor demand equations log-linear composite using relative price and real income . . . 81 Liquor demand equations log-linear per- capita composite . . . . . . . . . . 82 Liquor demand equations linear composite using average price . . . . . . . . . 83 Gross dollar sales regressed on demand parameters of liquor and distilled Spirits . 8A Retail dollar sales regressed on demand parameters of liquor and distilled spirits . 85 Distilled spirits demand equations . . . . . 86 Component price and income elasticity coefficients . . . . . . . . . . . 87 Liquor demand equations for sub-categories linear hypothesis . . . . . . . . . . 88 Liquor demand equations for sub-categories linear hypothesis . . . . . . . . . . 89 Liquor demand equations for sub-categories log-linear hypothesis . . . . . . . . 9O vii Table 32. 33. 3A. 35. 36. 37. 38. 39. MO. 141. N2. 43. an. A5. 146. Liquor demand equations for sub-categories log—linear hypothesis . . . . . . Component demand equations blended whiskey log-linear hypothesis . . . . . . . Component demand equations bourbon whiskey log—linear hypothesis . . . . . . . Component demand equations Canadian whiskey log-linear hypothesis . . . . . . . Component demand equations Scotch whiskey log—linear hypothesis . . . . . . . Component demand equations gin log-linear hypothesis . . . . . . . . . . . Component demand equations vodka log-linear hypothesis . . . . . . . . . . . Component demand equations brandy log—linear hypothesis . . . . . . . . . . . Component demand equations cordials and liqueurs log-linear hypothesis . . . . Component demand equations rum log-linear hypothesis . . . . . . . . . . . Component demand equations wine log—linear hypothesis . . . . . . . . . . . Total case sales of liquor in Michigan 1955”].966 o o a o o o o o o o 0 Gross dollar sales of liquor in Michigan 1955-1966 c o o o o o o o o o 0 Composite liquor price index in Michigan l955-1966 o o o o o o o o o o 0 Total case sales of distilled spirits in MiChigan 1955-1966 0 o o o o o o 0 viii Page 91 92 93 9A 95 96 97 98 99 100 101 103 10“ 105 106 Table ”7. A8. A9. 50. 51. 52. 53. SN. Gross dollar sales of distilled spirits in MiChigan 1955-1966 c o o o o o o Distilled spirits price index for Michigan 1955- 1966 . . . . . . . . Michigan disposable income 1955—1966 . . Consumer price index Detroit 1955—1966 . Michigan population 21 years and over 1955-1966 0 o o o o o o o o 0 Beer price index U. S. 1955-1966 . . . Total beer purchased in Michigan 1955-1966 Simulations of liquor purchase revolving fund transfer model for fiscal 1967 (in millions of dollars) . . . . . . ix Page 107 108 109 110 111 112 113 118 LIST OF APPENDICES Appendix Page A. Selected Regression Equations and Elasticities . . . . . . . . B. Selected Time Series Data . . . . . . . 102 C. Revenue Estimate Simulations . . . . 11Ll LIST OF CHARTS Chart Page I. State of Michigan Liquor Distribution System . 25 II. Flow Chart of Costs in Michigan Liquor Distribution System . . . . . . . . . 70 xi INTRODUCTION When fiscal "crises" threaten to disrupt state budgetary processes, the executive and legislative bodies of State government usually turn their attention to the revenue side of the budget. While various measures are taken to assure the electorate that economy and efficiency are the overriding considerations with respect to state expenditures, it is usually assumed that the pressures which have been exerted to provide the various goods and services of state government reflect "the public's demand" or "needs."- Whether or not this is a correct assumption and whether or not the public values the level and allo- cation of governmental programs to such a degree that it is willing to assume consciously full financial respon- sibility for all such programs is uncertain. For the most part, the budgetary activities of governments are not clearly understood by the electorate. These activities do not seem to be of major concern to many except where there is a considerable degree of per- sonal involvement. On the revenue side of government finance, a lack of conscious involvement is, at times, a product of the type of tax legislation which is prOposed or enacted. The lack of broad public reaction to certain tax legislation may be due to ignorance of the complicated provisions of a law and its full implications or to the selective character of the tax. A group of taxes which fall into the latter classification are the sumptuary taxes. *Sumptuary taxes are purportedly levied to reduce or curtail the consumption of various goods and services on the grounds that some form of social control is necessary for moral or social reasons. Taxation is supposed to provide the economic vehicle for achieving the espoused objective by increasing the price and discouraging con- sumption. Under the disguise of control, the primary purpose of the tax may be to provide a new source of revenue for the government. If the sumptuary objective were to be successfully achieved, the revenue objective of the selective tax may be seriously frustrated. Thus it becomes quite relevant to attempt to identify the market behavior of a commodity which is likely to become subject to sumptuary taxation. Alcoholic beverages have historically been a favorite category of goods subject to selective taxation. In 1791 a specific excise was levied by the United States Govern- ment on distilled spirits. From that time on alcoholic beverages have been singled out as fair game for federal and state government tax policy makers. Following the ratification of the Twenty-first Amendment (repeal of Prohibition) in 1933, this group of commodities became even more vulnerable to selective taxation. When Prohi- bition was lifted, taxation of alcoholic beverages was particularly appealing. Here was a new source of state revenue at a time when state finances throughout the country were in a precarious condition. Taxation assumed such forms as state liquor monopoly profits--where the states actually went into the business of wholesaling and retailing beverages--and various license fees. Presently every state in the nation derives revenue from the taxation of alcoholic beverages. The growth in the pOpularity of these taxes is illustrated by the fact that between 193A and 196A state and local government revenue from alcoholic beverages rose from a level of $177 million to $1.8 billion.1 In Michigan, alcoholic beverage revenues have become an integral part of the state's financial structure. Since 1933 various fees, taxes, and changes in monOpoly profit margins have been employed to help alleviate periodic financial crises while functioning to "control" the con— sumption and distribution of alcoholic beverages. Excise taxes on distilled spirits, wine and beer, as well as lLicensed Beverage Industries, Inc., The Alcoholic Beverage Industry and the National Economy, A State by Analysis (New York: Licensed Beverages Industries, Inc., profits from the state liquor enterprise, constitute the sources for the bulk of Michigan alcoholic beverage revenue today. Various licenses and other fees and fines make up the remaining alcoholic beverage revenue. Local units of government receive the greatest portion of the latter types of revenue. In fiscal 1966 the State of Michigan had $97.9 million in revenue transferred to the State's General Fund from liquor excise collections, state liquor monopoly profits, and beer and wine excise taxes. This was in the neighborhood of eleven per cent of all General Fund revenue that year. An additional $9.6 million in liquor excise collections were deposited in the "special purpose" State School Aid Fund.2 In spite of the millions of dollars contributed to Michigan State Government by the various alcoholic beverage taxes, a number of gaps exist in the identification of basic economic behavioral parameters which are important in understanding the behavior of liquor revenue. It is the purpose of this study to attempt to identify some of the relevant parameters and to indicate through the construc- tion of a simple revenue model how this information can contribute to better informed policy decisions. Since 2State of Michigan, The Executive Bud et (Lansing, Michigan: State of Michigan Bureau of the Bu get, 1967), pp° 6: 7: 13- TABLE l.--General fund alcoholic beverage revenue State of Michigan 1996—1966 (in thousands). Fiscal Liquor Purchase Liquor Beer-Wine Alcoholic Year Revolving Excise Excise Beverage Revenue Total 1966 $50,220 $9,575 $38,098 $1,160 $99,053 1965 93.932 8,615 37,702 1,181 91,930 1969 29,607 7,805 36,052 1,161 69,625 1963 39,931 7,357 39,005 1,210 82,003 1962 39,293 193 6,829 1,922 97,687 1961 36.906 6,703 13,601 1,389 58,599 1960 39,095 3,678 11,399 1,909 55,976 1957 “9,963 - 7.367 1,337 53,167 1956 33.798 - 7,797 1,397 92,892 1959 37,988 - 7,385 1,316 96,689 1953 35.0““ - 7,392 1,315 93,751 1952 35,561 - 6,861 1,203 93,625 1951 28,322 - 6,975 988 36,285 1950 30,279 - 6,832 990 38,101 1999 39,863 _ 6,909 1,072 92,899 1998 33,988 - 6,980 561 91,529 1997 13,005 - 6,787 990 20,282 1996 23,790 — 6,686 908 30,889 Source: State of Michigan, The Executive Budget for selected years. liquor monopoly profits and excise taxes are the major source of alcoholic beverage revenue, specific attention is devoted to the demand for liquor. The restriction of the inquiry to liquor, herein defined to mean distilled spirits of at least 22 per cent alcohol and "fortified" wines, is due to its importance as a revenue source and to the limitations on reliable data for beer and wines not handled by the state enterprise. The basic format found in the following chapters is as follows. (a) A brief survey of various liquor demand studies appears in Chapter I. These studies are of foreign markets as well as United States markets. The significance of these inquiries lies in the scope of their analysis, the statistical techniques employed, and the character of the parameters identified. (b) Chapter II presents the major results of our statistical investigation. Models are constructed on three levels of aggregation to identify significant parameters. These are principally the income and price elasticities. The analysis of a composite demand function for liquor is presented. This function aggregates total liquor case sales in Michigan and regresses them on seasonal, price and income variables. The price-quantity relationship is found to be slightly inelastic. However, price elastic demand behavior is "identified" according to the movements of total expenditures; that is, when actual dollar sales data are regressed on the demand parameters a significant inverse relationship emerges. The subsequent level of analysis, which concentrates on an aggregated demand function for distilled spirits reveals that the price elasticity of the composite demand function can be partially explained by the strong substitution of fortified wines for distilled spirits. This intra-basket substitution is con— firmed by a strong positive cross price elasticity between fortified wines and distilled spirits. The price elas- ticity of distilled spirits case sales can be closely reconciled with the behavior of actual dollar sales on distilled spirits during the sample period. Finally there is presented a discussion of ten demand functions which are "components" of the composite function. These component functions are aggregated and compared with the foregoing results. A brief discussion of each component follows. [The basic statistical tool used throughout the analysis of the demand for liquor was the direct least squares multiple regression technique. The equations were estimated with the use of basic data that were obtained as part of this inquiry.] (0) Chapter III deveIOps a sketch of Michigan's state liquor revenue structure. Through the use of a model for computing profits from the state liquor monopoly and a simple excise tax model, the information obtained from the preceding chapter is shown to be useful in assessing the revenue implications of variations in such factors as excise tax rates, profit margins and increased liquor costs from the manufacturers. The models presented are very meaningful for revenue estimating. The liquor demand variables would play the primary role in the determination of such estimates. Following the conclusion of Chapter III, three appendices appear. The first contains a number of selected regression equations and demand elasticities which are the product of Chapter II's inquiry. Selected data are pre- sented in the following appendix. The last appendix presents a simulation of the revenue models deve10ped in Chapter III to estimate liquor revenues for fiscal 1967. These results are compared with reported 1967 returns. CHAPTER I EMPIRICAL STUDIES In 1998 Sten Malmquist, a Swedish economist, under— took a statistical demand analysis for liquor in Sweden.1 Malmquist approached the analysis by using a log linear model of the demand equation. He utilized annual time series data which ran from 1923 through 1990. The basic equation employed the three basic economic variables: quantity (Q), price (P) and income (Y). A fourth variable, the average ration during the year (Z), was also employed as an explicit explanatory variable. The constraint imposed by rationing was of central concern to the author. Malmquist constructed three separate price statis— tics. One was the average price per liter of liquor (distilled and fermented spirits); another was the average price per liter of distilled 90% (80 proof) spirits; and the last was an "average price per litre of liquor cal— culated from the distribution of purchases among the main types of liquor in 1930."2 These statistics are identi- fied as P', P"', P" respectively. lSten Malmquist, A Statistical Analysis of the Demand for Liquor in Sweden (Uppsala, Sweden: University of Uppsala, 1998). 2 Ibid., p. 18f. 10 Seventeen regressions were performed using various combinations of the variables mentioned plus a consumer price index for forming a relative price statistic as well as "real" income data. The results are found in Table 2. The price and income elasticities are very low in the above regressions. Overall, "(t)hese results would appear to show that the demand for liquor is decidedly under-elastic (inelastic)."3 This however is due to the rationed characteristic of the commodities under consid- eration. How should a rationed commodity affect elasticity estimates? Malmquist indicates that the price elasticity is not the elasticity which can prOperly be associated with a "normal" demand function. The function under study involves the elasticity for only a select group of consumers. In the case of a price hike those who had a "normal" consumption level less than or equal to the rationed quantity at the original market price, curtail consumption as would be expected. Others who at the original price level would have consumed a much larger quantity of the produce than they were able to purchase, do not respond in a normal fashion. They do not react by reducing their consumption of the commodity and hence 3Ibid., p. 9. 11 TABLE 2.--Regression coefficients, demand for liquor in Sweden 1923-1990.1 log Q = a + log Pé log Z log Yn log C R 1. -0.923 - - - 0.881 2. —0.336 0.957 - - 0.926 3. —0.967 - 0.173 - 0.906 9. -0.353 - - 0.711 0.977 5. -0.372 0.705 0.316 - 0.992 6. -0.373 0.868 0.379 0.173 0.993 log P; log P; log P;' log Z log Y R 7. -0.906 - — - - 0.959 8. -0.369 - - 0.179 - 0.969 9. —0.369 - - 0.651 0.300 0.992 10. - -0.376 - - - 0.999 11. - -0.320 - 0.782 0.368 0.995 12. - - -0.282 — - 0.969 13. - - -0.279 0.531 0.309 0.993 19. —0.956 — - — - 0.969* 15. - - -0.326 - - 0.961* 16. -0.993 - - 0.538 0.286 0.982* 17. - — —0.390 0.338 0.271 0.973* lRegressions 1-13 used data to 1939 and 19-17 used data through 1990 (*); subscripts n, r signify nominal and deflated data respectively. Source: Sten Malmquist, A Statistical Analysis of the Demand for Liquor in Sweden (Uppsala, Sweden: University of Appsala, 1998), Chapter I, Tables I, II, III, IV. 12 the elasticity coefficient understates "normal" behavior. An analogous explanation can be made for the low income elasticity.Ll Malmquist's study drew upon the basic causal rela- tionships established by economic theory. In formulating the relationship in a log linear form using multiple regression techniques, he attempted to discover the elas- ticity coefficients in accordance with popular econo- metric methods. Malmquist's principal concern, however, was directed toward the behavior of a rationed commodity. The year after Malmquist published his analysis, A. R.-Prest wrote on "Some Experiments in Demand Analysis."5 While investigating general consumer expenditure patterns in the United Kingdom (1870-1938), he performed an analysis of the demand for spirits. Prest followed a pattern quite similar to Malmquist in utilizing a logarithmic form of the demand function. He added two variables to the three basic variables of quantity, real (national) income and prices. These were population and a variable identifying the given time period (year). The latter variable was intended to explicitly identify a given state of tastes and consequently a change in tastes over time. His basic equation was of uIbid., Chapter II. 5A. R. Prest, "Some Experiments in Demand Analysis," The Review of Economics and Statistics, XXI (February, 1999). l3 bl b2 b3T the following form: Q/O = A(Y/O) (P/C) e quantity demanded (proof gallon) population "real" (deflated by C) income price of commodity index of all other prices base of the natural logarithm given time period (a given state of tastes) HQOWKOO II II II II II II II The price elasticity (b2) equaled -.8365 and the income elasticity (b1) was 1.0996. While various alter- native formulations were also tested, the above results represent the basic findings of the inquiry. In 1959 Richard Stone published the results of an extensive study of consumer expenditures and behavior in the United Kingdom.6 In the course of this inquiry he performed an analysis of demand for distilled spirits. This analysis used annual time series data (1920-1938). This statis- tical model for spirits deviated from his general time series regression model for demand functions. The modi- fications consisted of eliminating per capita income and quantity variables, the exclusion of prices of related variables and the introduction of a trend variable. The regressions were run in first differences in a log-linear form. The results are as follows: 6 Richard Stone, The Measurement of Consumers' Expenditure and Behavior in the United Kingdom,_l920-l938 (Cambridge, England: University Press, 195“). 19 TABLE 3.--Demand analysis for distilled spirits--United Kingdom 1920—1938. log (d0) = a + b log (dY) + b lob [d(P/C)] + b3(dT)* 1 2 Income Residual 2 Elasticity Elasticity Trend R d 0.80 -0.053 97 1.22 (0.21) (0.010) - -0.020 55 2.58 ' (0.009) 0.60 -0.57 —0.033 79 2.98 (0.19) (0.12) (0.006) *Notation is essentially the same as that used for Prest, c.f. p. 9. Source: Richard Stone, The Measurement of Consumers' Expenditure and Behavior in the United Kingdom, 1920-1938 (Cambridge, England: University Press, 1959), Table 110, p. 390. These results are reasonably consistent with the work of Prest as far as the relatively inelastic price behavior is concerned. U. S. Studies While in the process of analyzing the whiskey industry in the United States, Harold Wattel discussed the demand characteristics of the Pennsylvania market during the period extending from 1936 to 1951.7 He formulated a series 7Harold L. Wattel, The Whiskey Industry: An Economic Anal sis (New York: New School for Social Research, 1953), p. 296 ff. 15 of point elasticities of demand for each year using a demand function derived from "multiple correlation" tech- niques. He employed the simple linear hypothesis with consumption, price and income data. Price was related to the consumer price index for the nation as a whole° Unfor- tunately the results of the analysis did not report enough statistical information to evaluate the coefficients or the overall performance of the equation. Wattel's "basic formulae" were: Q = 2.0169 - 0.2999 P + 0.0001098 Y et = Pt/Qt (-0.2999035) Q = consumption P = relative price Y = income et = elasticity in year t The resulting point elasticities ranged in value from -l.29 to -.69 with the mean value being approximately -9.1. Wattel did not have considerable interest in the demand characteristics pg: s9. He seemed to accept these results and alluded in selected references to various opinions and at what can probably best be described as "guesstimates" of the demand characteristics of distilled spirits.8 8Ibid., p. 290ff. 16 In January, 1960 William A. Niskanen published a monograph entitled Taxation and the Demand for Alcoholic Beverages.9 This publication was directed toward describing an aggregate demand function for alcoholic beverages in the United States. The model used the total national market for alcoholic beverages. It essentially broke this into three markets for the aggregated commodities of distilled spirits, beer and wine. Niskanen utilized annual time series data from the years 1939-1959. He formulated a system of demand and supply equations. Least squares estimates of the reduce-form coefficients were transformed to yield the estimates of the structural parameters of his demand functions. He obtained the following elasticity coefficients as the product of the ratio of the arithmetic means of the "explained" and "unexplained" variables and the relevant demand coefficient. price elasticities: distilled spirits -l.79 wine -2.27 beer - .99 Interestingly, rather than using income as an exogenous variable in the system, "real monetary assets" (demand deposits adjusted plus currency outside banks) was used as a measure of general purchasing power. This produced a "purchasing power" elasticity of 1.99 for spirits, .57 for beer and .78 for wine. 9William A. Niskanen, Taxation and the Demand for Alcoholic Beverages (Santa Monica, California: The Rand Corporation, 1960). 17 In 1962 Niskanen published his doctoral dissertation in which he altered somewhat the form of his earlier work as well as updated his data.10 He included the 1955-1960 period and excluded the 1992-1996 period. He added the income variable. He used the direct least squares and the two-stage techniques to estimate the relevant parameters. Niskanen also employed the two-stage estimators with a transformation of price data (basically producers' price data) to reflect a hypothesized constant absolute distrib- utors' profit margin. His results are found in Table 9. In 1966 Julian Simon published a paper in Econometrica which was devoted to the development of a method of deter- mining the price elasticity of liquor in the United States.ll Simon's "quasi-experimental method" was essentially an effort "to examine the 'before' and 'after' sales of a given state, sandwiched around a price change and standardized with the sales figures of states that did not have a price change . . . then pool the results of as many quasi- experimental 'trial' events as are available."12 10William Arthur Niskanen, Jr., The Demand for Alcoholic Beverages (Chicago, Illinois: The University of Chicago, 1962). llJulian L. Simon, "The Price Elasticity of Liquor in the U. S. and a Simple Method of Determination," Econometrica, XXXIV (January, 1966). 12 Ibid., p. 196. 18 TABLE 9.--Alcoholic beverage demand elasticities for the U. S. 1939—1960. Elasticities 2 Technique price income monetary H assets (a) Direct spirits: -1.920 .219 .565 .929 least ( .235) .153) (.197) squares beer: - .626 .332 .991 .961 ( .169) .139) (.073 wine: - .563 .687 —.098 .993 ( .227) .172) (1.90) (b) Two spirits: —l.901 .327 .957 .999 stage ( .290) .80) (.173) least squares beer: - .696 -.380 .922 .959 ( .236) .180) (.090) wine: - .681 .606 -.102 .992 ( .229) .192) (.155) (c) Two spirits: -2.l35 .327 .957 stage ( .367) .180) (1.73) constant distributor beer: - .696 .380 .922 margin ( .326) .180) (.090) hypothesis wine: - .981 .606 -.103 ( .329) .192) (.155) Source: William A. Niskanen, Jr., The Demand for Alcoholic Beverages (Chicago, Illinois: Chicago, 1962), Table 8, Table 9, Table 10, pp. 56-58. The University of l9 Simon's analysis obtains one "trial" over a horizon of thirty—one months. He assumed one brand (Segrams Seven Crown) 13 as being representative of all spirits. Simon computed the "price elasticities" for a number of states at various times for given price changes. The results of some of these computations are found in the following table. The states selected are those for which he reported two or more "trials." The "trials" are arranged in chronological order. TABLE 5.--Price elasticities via the quasi-experimental method. Trial State #1 #2 #3 Idaho 0.85 -0.89 Ohio 0.83 -1.32 Washington 0.25 -0.02 Oregon 0.06 90.03 -l.00 Maine -0.12 -0.89 Montana -0.15 ‘-3.73 Iowa -1.03 -l.90 Virginia -l.90 -2.25 Source: Julian L. Simon, "The Price Elasticity of Liquor in the U. S. and a Simple Method of Determination," Econometrica, XXXIV (January, 1966). 13In our own inquiry using a logarithmic form of the function and including seasonal and income variables, these price elasticities differed considerably with Seagrams Seven Crown having a price elasticity of -2.27 (std. error .276) and the overall demand price elasticity of -.925 (std. error .231). 20 Concluding Comments Each set of circumstances surrounding the various liquor demand studies have, of course, led to results peculiar to those circumstances. The identification of behavioral parameters in each instance can only be relied upon for dependable information about the relationships directly measured. Inference to different circumstances or to different behavioral patterns not directly identi- fied may lead to misinformation. With these thoughts in mind, we will now proceed to analyze the demand for liquor in Michigan and attempt to identify the parameters which are relevant for a clear understanding of that market. CHAPTER II MICHIGAN LIQUOR DEMAND The basic postulates of economic theory set out a general functional relationship between the quantity of a good purchased and a vector of determining factors. A select few of these factors or variables have attained special importance in economic analysis. The price of the commodity under consideration and the overall level of income of the consumer are singled out as principal economic determinants of demand. Prices of complementary and substitute goods, the state of tastes, and a host of variables which can be related to the commodity round out the list of determining factors. The definition of the period of time during which the flow of demand is to be considered plays an important role in determining whether a seasonal variable needs to be explicitly identified in the demand function, and in influencing the characteristics assumed by the other behavioral parameters. The general demand function for a commodity assumes the following form: (2.0) Q F (PX,Y,PC,PST,Z,S) x where: Qx = the quantity of good x demanded PX = the price of good x Y = the income level of the consumer(s) 21 22 Pc = vector of prices of complementary goods Ps'= vector of prices of substitute goods T = a state of tastes Z >5 vector of all other factors not elsewhere identified 8 = seasonal variable (when required) Such a general function, while logically complete, is not empirically implementable. However, the identifi— cation of economically relevant variables establishes the point of departure for productive analysis. Similarly the identification of the nature of the influence of the deter- mining variables on quantity demanded provides a logical framework within which to examine empirical demand results. For example, we would normally expect an inverse relation- ship between the price and the quantity demanded of a good. The empirical problems associated with the general demand function are not due to the lack of recognition of relevant variables, but rather to its all-encompassing nature. It is impossible to empirically identify, let alone quantify in a meaningful fashion, all of the variables which in any way influence the quantity demanded of a good. Consequently market demand studies must rely on the distil- lates theory offers, namely, the variables which rank high among the determinants of demand. Likewise such investi— gations can only record results within the limits of the data and the caveats surrounding the statistical techniques used. 23 The Models The market defined.-—The first task of our inquiry is to delineate the perimeter within which we will operate. The market under consideration has been defined by the policy orientation of this study. The Michigan liquor market is the subject of our analysis. Liquor as employed in this study connotes the entire range of distilled alcoholic beverages with an alcohol content of 22 per cent or more, plus the whole range of "fortified" wines having an alcohol content of 16 per cent or more. This assortment of alcoholic beverages constitutes the merchandise which is wholesaled solely through the facilities of the Michigan Liquor Control Commission. The market for liquor is con- fined to legal sales and does not attempt to incorporate an analysis of illegal liquor traffic.1 Within the general confines of the market as used in this study we have dealt primarily with wholesale sales of liquor to licensed retail outlets. These include "specially designated distributors," taverns, hotels, and selected organizations such as social clubs, etc. By far the largest retailers of liquor are the "specially designated distributors" or "S.D.D."s. They account for almost three- quarters of the state's liquor sales. S.D.D.s sell pack- aged liquor for consumption off premises. The Liquor 1The attempt to measure illegal traffic must rely on. rough proxy variables for measurement. Such variables are dependent upon a number of factors other than the amount of illegal traffic. The most significant is enforcement effort. 29 Control Commission also operates packaged retail outlets which handle from two to three per cent of gross sales in the state.2 Bars and hotel sales make up the bulk of the remaining sales. In this study it is assumed that whole- sale sales are final sales. The problem of stocks of inventories in the retail outlets as well as the consumers' stocks are ignored. The basic liquor distribution system encompasses the central warehousing, wholesale, and some retail operations. Chart I provides an overview of the physical structure of the distribution system. It also indicates the magnitude of the flow of traffic between districts. As can be seen, the system operates through three major warehouses and approximately ninety state operated "stores." The latter are for the most part combination stores which supply the various retail licensees with liquor as well as maintain a very limited retail trade. Close to eight hundred persons were involved in this system's Operations in fiscal 1966 not counting the non-state contractural help. Common carriers and railway transportation provided the basic means for shipments from point to point throughout the system. The heaviest concentration of activity was in the metropolitan Detroit area which accounted for close to seventy per cent of total annual sales. 2Michigan Liquor Control Commission, Financial Report, selected.years. 25 .sppmsm soapspahpmao hoseflq sameness mo ppmsmuu.H 696:0 Anm.mv spasmomm w M Aao.:mv “no.0mv wcfimcmq CO ; moomcooaq Hampmm poapso Hamumm /\ ooq . mpmapso \\ HampomlmammmaonB DOA ARH.mzv oaa ARm.mwv momzocmpmz OLOum xhwm CHOOCHA DOA .HV mthSBommscmz 26 In defining the market it is important to identify the time span under consideration as well as the unit of time in which the data are formed. The market for liquor was examined over an eleven-year period beginning in July of 1955 and ending in June of 1966. This time span was chosen due to data limitations as well as considerations for reasonably consistent population parameter determin- ’ ation for current policy decisions. The unit of time considered during this period was the quarter. This was considered as optimal both for maximizing the number of observations obtainable for time series analysis, and for aiding in identifying points in time when significant variables changed. It likewise is a useful form for obtaining information for budgetary purposes. Within the framework of the above market description we have found it informative to attempt three different approaches to the liquor market due to each's contribution to the identification of significant behavioral statistics and due to their interest to economic analysis. At one level the market is studied in its completely aggregated form. This entails the attempt to identify the parameters for a composite demand function for liquor. A second approach singles out distilled spirits. It attempts to identify behavior in this market both for analytical and policy purposes. Finally the market is subdivided into ten categories or "components" to reveal the reactions of 27 each subgroup to the independent demand variables as well as to compare the composite demand parameters with Weighted parameters of the subfunctions. The market for liquor is thus delineated by its geographical, legal, distributional and temporal charac- teristics for analysis on three distinct planes. The General Function The general functional relationship between variables draws upon the prominent economic factors identified in the general demand function previously discussed. It assumes the form: (2.1) Q = f (P,Y,S) where: Q = quarterly case sales P = price expressed in index form Y = quarterly estimates of "Michigan Disposable Income" S = seasonal factor identified by three dummy variables Variations of this function were tested in the composite, distilled spirits and component markets. Composite Demand.--The composite function assumed the following form: (2.2) 0% = F (PgY,S) 28 In this demand function liquor was treated as a composite commodity leaving no room for distinction of type, brand or price category. It basically assumed a homogeneity of the commodity and demonstrated the behavior of the total quarterly volume of case sales of liquor responding to the determining variables--price, income and the seasonal factor. In identifying the price of liquor, a Laspeyres chain linked price index was used.3 It employed seventy individual brand prices and quantities as a representative sample of the commodity, liquor. Seven brands from each of ten categories of liquor were chosen.“ They were selected from the most popular low, medium and high priced brands within each category. The formula used to form each link of the index was: 3The chain linked index is conceptually similar to that used in Niskanen's study. It differs principally in so far as it does not use average prices of liquors in the various categories but a Laspeyres type index to determine these also. Furthermore, the time span per observation is shorter. This type of index satisfies the proportionality criterion, i.e. as all prices within the components vary by some proportion k, the index will also vary by k. This property is important when comparing the composite and aggregated component results later in this chapter. ”These categories--blends, bourbon, Canadian, scotch, gin, vodka, rum, cordials and]iqueurs,brandy and wines-- are discussed below in the section on component demand functions. 29 7 t 10 2 Pt . Qt‘l _ i 1 1-1 2 , Qt-l 7w J _ i i 1.1 It = 3 1 7 t‘ t t-l 10 2 P - Q i=1 i 1 . Qt-l Z J 7 i=1 2 PE-l - QE-l i=1 3 where: Q: (j) = quarterly seasonally adjusted volume of case sales for the i (j)th brand (category) in time period t. P1 = Ehihfénal price to the gonsumer of the - rand in period t. The income statistic was computed by reducing quar- terly estimates of Michigan Personal Income by the amount of federal personal income taxes paid by Michigan residents. The adjustment was deemed expedient in light of the passage of the Revenue Act of 1969 which changed significantly the federal income tax rates for 1969 and 1965. Thus the income statistic more closely reflected Disposable Income behavior. 5In its composite form the price statistic included all excise taxes. When the tax component was not included the resulting price index did not perform well in statis- tical analysis. The standard errors of the price coeffi- cients in the equations tested were too high to be statistically significant. 30 The use of quarterly unadjusted quantity data required that a seasonal variable be introduced into the equation. This was accomplished by the use of dummy variables.6 Variation of the basic model used selected additional variables. These included population (0), consumer price index (C), the price of beer (B) and a trend factor (T) as a proxy for a state of tastes. Annual estimates of population twenty-one years of age and over were employed to attempt to identify if possible the impact of the number of eligible consumers. The Detroit Consumer Price Index was introduced into the equations as an explicit variable. It was also used in the formation of relative price and "real" income statistics. Beer was assumed to be a substitute commodity and its price was used in some of the variations of the equations tested. Due to the lack of reliable beer price data specifically for the Michigan market, the beer price index provided by the Bureau of Labor Statistics for the U. S. was employed as a proxy.7 6The employment of dummy variables was intended to avoid the possibility of introducing spurious behavior into the quantity data through deseasonalization. The introduc- tion of the dummy variables involves three such variables with the remaining quarter being identified in the constant term of a linear regression. For the first calendar quarter, the variables D(l), D(2), D(3) assumed the values 0; in the second quarter 1,0,0 respectively; in the third quarter 0,1,0; and in the fourth quarter 0,0,1. See the brief discussion in Lawrence R. Klein, An Introduction to Econo- metrics (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 19625: P. 35. 7Tables of all the statistics discussed in this section are found in Appendix B. 31 The five basic variables (three dummy seasonal variables, the composite price index and income) were employed in various equations alone and in combination with the remaining four variables. A linear equation and exponential form using natural logarithms were used. Under these hypotheses the multiple linear regression technique was employed to identify price and income elasticity coefficients as well as to help evaluate the explanatory value of the other variables. This technique was expected to provide the best linear unbiased estimates of the demand parameters. The disturbance terms in the linear regression were assumed to reflect specification errors in the basic equations since the basic data which were used were con- sidered reliable. The regression equations took the basic forms: (2.2a) Q = a + le(l) + b2D(2) + b3D(3) + buP + bSY + ... +E (2.26) an = a + le(l) + b2D(2) + b3D(3) + bulnP + bSlnY + . . .,+ E where: Q = total case sales of liquor D(i)= seasonal dummy variables (1 = k,2,3) P = composite price index Y = income E = disturbance term assumed N(O,l) 32 The above equations assume that the statistics being used in their formulation identify the demand for liquor. This critical premise is based on the definition of the relevant market, the character of the variation in and specification of the demand and supply functions and the causal relationship established within the body of economic theory. Since Michigan constitutes a relatively small percentage of the national market for liquor and since the consumers may obtain any quantity of liquor they wish at the given price, the assumption is made that within the unit of time specified the supply function for liquor was perfectly price elastic. Thus variations in the quantity of liquor purchased were determined by the demand variables. Variations in price due to the vertical shifting of the supply function aided in the identification of the "price— quantity demanded" relationship. While income was considered an explicit variable in the demand function, there was no reason to believe that it had any direct explanatory value in the assumed supply function. Thus the income coefficient could be expected to identify the relationship between income and quantity demanded.8 The 8The price variations during the sample period can for the most part be directly attributed to supply-price changes associated with alterations in the excise tax rate. Thus the demand function can be statistically "identified," that is, var (Et) < k var (Ut) where 0 < k < l and Ut is the supply function disturbance. The absence of income in the supply function indicates its independence within the 33 line of causality between the dependent and independent variables was based on the postulates of demand theory as presented at the beginning of this chapter. Selected results of the regressions performed are reported in Table 6. All of the equations presented in Table 6 exhibited high values of the F statistic. The simplest linear form of the demand equation performed statistically as well if not better than all of the other equations tested. This equation was also the frontrunner when additional variables were added to the equation under both the linear and logarithmic hypotheses.9 The introduction of population, beer prices and the trend variables proved to be of little explanatory value. This is clearly demonstrated by the actual decreases in the R-bar square statistic as these 10 variables were added. This may partly be explained assumed system of equations and thus its coefficient in the demand equation statistically identifies its relationship to the quantity of liquor sold. See Klein, pp. 13-19. 9The statistical results of the variations tested are found in Appendix A, Tables 18-29. Also found in the Appendix is a more detailed description of the equations presented in Table 6 above. 10The R-bar square statistic is the coefficient of multiple determination adjusted for the degrees of freedom. It is formed according to the following formula: R 2 = 1 - N ' l (1-R2) where N is the number of obser- N - K - l vations, K is the number of independent variables and R2 is the coefficient of multiple determination. 39 .pmms mOHpmemum mEoocfi samba: new means m>HpmHomm .emms moapmdpmum esoocfi use mpfipcmsu deadwotpmm m .eomeHocH ma scene opmecmum on» mmoac: o>onm no He>oa mmm. can as uanHMficwflm one mucoaoammooo HHm mo mmo.m mmm.o mmwa.o mawa.o| wmmo.o mmao.o seao.o mmom.o mm NHH.N mom.c mma.o Hmm.mpwl mao.mmm mmm.mm wam.am Ham.ozm < in mg m as + 33 mp + SE up + 33 as + e n a comps: . Infinnso mm .mfimonpoazn nmmcaq a.mc0fipmduo escape heaved mpam0a80011.m mqm coca mmm Hmm .omm mam mam new coca omm mom mwm mwm copoom QQQH mmm mmm mom cmfipmcmo coca mom Ham mconpson QQQH mom mocman oooH opHmOQEoo mcflz Esp Upoo moan mxp> cfiw Spam Ucmo anon moan QEoo .moxmocfl mofiaowopmo nozvfia comzuon mucoHOHummoo COHumHopsoo moapmll.w mqm , s <0 F“ l 50 Tables 29 through 92 in Appendix A present the more sig- nificant logarithmic results obtained for each category. Aggregation: While the investigation of the component demand equations was partially intended to reveal some specific difference with respect to the behavioral param- eters within the group, it was also intended to provide components to be aggregated into overall price and income elasticities. These aggregated statistics could then be compared with the results of the composite and distilled spirits functions. Aggregation of elasticities was accomplished by taking the quantity weighted average of the income and price elasticities of the ten component * it 01 = fi[(Pi/C) , (Y/C) , s, T] - r [( c * c.* * * _ * * it- * 0i - fi[(Pi/C) , (Y/C) , s, (B/C) , o , T1 * i6 *- (Qi/O)_ = fitPi, (Y/o) , S] 9(- * ‘1! (01/0) = fitpi, (7/0) , s, T] * 9(- * (01/0) = fi[(Pi/C) , (Y/C/o) , s] it it 96 (01/0) = fi[(Pi/C) , (Y/C/O) , s, T] *signifies variables tested using their natural logarithms as well as linear form. 51 functions. The quantity weights were the mean quantities of each of the sub-categories.19 The elasticity coeffi- cients.from each of the components were selected under three criteria: most significant value for the coeffi— cient, coefficient from the best fitted-equation and coefficients from identically structured equations.20 The results are reported in Table 19. .While the aggregate elasticity coefficients were close to each other when the first two criteria were used, there were considerable differences between these results and the parameters arrived at by aggregating the elas- ticities of the equations which were from the identically structured equations, Qi = fi(P1, Y, S). The major deviant was the perennial trouble maker--fortified wines. The price and income coefficients in the wine category were influenced to a high degree by the introduction of a trend variable to the simple equation. Both the price and income elasticities became negative and increased in significance 19 10 W n = Z ni 1 1-1 Q1 10 where W1 = T— and if]. W1 = l 2 Q1 1-1 This technique is similar to that in Wold and Jureen, pp. 112-119. 20The component elasticities are found in Table 28 in Appendix A. 52 TABLE l9.--Aggregated liquor demand price and income elasticities. Price Elasticity Income Elasticity Criterion linear log linear log identically - structured -0.2091 -0.l229 0.7818 0.8197' equation (0.999) (0.639) (0.089) (0.102) most significant -2.7903 -2.7055 0.5828 0.5619 coefficient (0.582) (0.699) (0.085) (0.106) best fitted -2.7328 -2.6075 equation (0.589) (0.633) 8998 as above with its introduction. The Durbin-Watson statistic for the wine equations increased significantly to an acceptable level as did the coefficient of multiple determination. Thus the wine component was the major cause for the drastic shift in the aggregated elasticity coefficients under the various criteria tested. This influence is due to the relatively large physical volume of this component and its large elasticity values. Wine constituted 16 per cent of the physical volume and had price elasticities ranging from 7.1922 to -8.5311. The high cross elasticity factor referred to in the discussion of the composite function as well as weakly correlated and non-proportional price move- ments within the wine category itself resulted in a price coefficient which cannot be considered reliable for the above type of statistical aggregation. 53 In spite of the poor performance of the aggregated price elasticities, income behavior was consistent through- out the sub-categories. The aggregation of the income elasticities could therefore be expected to provide reasonably good results. When the weighted average income elasticity for like structured equations is compared with the composite demand functions income elasticity coeffi- cients, the results are very similar. See Table 15. TABLE 15.—-Liquor demand income elasticities composite and aggregated compared. Aggregated Demand HypotheSis CompOSite Demand identically most significant structured coefficient, linear 0.7725 0.7818 0.5828 (0.095) (0.089) (0.085) logarithmic 0.8196 0.8197 0.5619 (0.099) (0.102) (0.106) When the coefficients from the best fitted equation of the sub-categories as well as the most significant regres- sion coefficients were compared, they proved to be the same coefficients as those in the identically structured equation [Q = f(Pi, Y,S)] except for the wine coefficient. As mentioned above, the wine income elasticity coefficient reversed its sign (became negative) and improved in 59 significance when-the trend variable was added. The presence of the trend variable in the composite function, however, did not produce results which had any significant effect on the composite income elasticity (nor price for that matter) though a slight change in income (and price) elasticities could be noted. These changes influenced the elasticities in the direction of the aggregated results--though not significantly.21 Next in our analysis, aggregation of the sub— categories' price and income elasticities was undertaken with the exclusion of wine. These aggregated results were distilled spirits parameters. These are presented in Table 16. These statistics were fairly compatible with the results obtained from the distilled spirits demand func- tion discussed above. The absolute values of the coeffi- cients were somewhat less than were obtained from the direct distilled spirits demand regressions. A strong positive vodka price elasticity appears as a likely candi- date which might account for a large prOportion of the difference.22 21See equation B, Table 33 in Appendix A. 22When vodka price and income were divided by the Detroit Consumer Price Index and a trend variable added, a fairly large significant price elasticity wasnobtained. See Table 38 in Appendix A. 55 TABLE l6.--Aggregated distilled Spirits demand price and income elasticities Price Elasticity Income Elasticity Criterion linear log linear log identically structured —l.6150 -l:5580 0.8696 0.9289 equationl (0.906) (0.335) (0.077) (0.093) most 68 6 significant -1. 37 -1. 751 equation (0.923) (0.399) same as above 1 These were also the best fitting equations in which the same form of the price and income statistics were used, ViZo’ Q = f(P’ Y, S). It should be recalled that while price movements between the various categories are highly correlated, they are not perfectly correlated. Small price changes exper- ienced within the categories which might not evoke a strong response may well attenuate the overall measure of respon- siveness or "price elasticity" for the category. The design of the sample of seven brands making up the price index may also not be large enough and_may introduce a downward bias on the elasticity_coefficient by overstating the pervasive character of price movements within a given category. One interesting observation is that while the income elasticity coefficient resulting from the sub-categories being aggregated is smaller than the income elasticity 56 directly computed in the distilled Spirits demand equation, it is almost identical with the income elasticity of total expenditures on distilled spirits. Overall, the aggregation of the component functions to form demand parameters for liquor does not appear to be productive. The increased sensitivity of the disaggregated functions may well be a liability rather than an asset when broad based parameters need to be identified. Component demands: The component demand functions reveal some interesting differences in the behavior of the sub-categories of liquor. The most popular type of liquor sold in the Michigan market was blended whiskey. This group of whiskeys accounts for almost 91 per cent of the liquor sold and close to 99 per cent of distilled spirits sales. Blended whiskey on the whole is the least expensive of the whiskey group. It is widely used in popular mixed drinks. The demand parameters which were identified in the regression performed on "blends" showed the commodity to be fairly price elastic and income inelastic. The dominant explanatory variable in terms of the quarterly case sales was the fourth quarter seasonal factor. This is reflected in the very high beta weight associated with the dummy variable D (3). The introduction of the trend variable into the equations as well as the forming of relative price and "real" income statistics increased the income elasticity 57 and had a depressing effect on the price coefficient. The high price elasticity may indicate the possibility of a fairly high substitutability between blends and the more neutral spirits such as gin or vodka. Bourbon whiskey which was another category including bonded bourbon, displayed an inelastic price elasticity and an income elasticity of approximately unity. Bourbon, which is predominantly a corn distillate, has a more dis- tinct character than blended whiskey. If the distinct physical characteristics do indeed reduce the suitability of cheaper substitutes, this might explain the elasticities which were obtained. Canadian whiskey is for the most part more expensive than bourbon though it is usally a bourbon type liquor. Its fairly high price elasticity may reflect the consumers' willingness to substitute the somewhat less expensive bourbon. The high income elasticity would tend to confirm the "superior" character of the commodity and help rein- force the expectation of a higher price elasticity. Scotch whiskey has a rather special character that would tend to isolate its market from other whiskeys. In spite of this factor, Scotch appeared to be slightly price elastic. This might suggest that given price changes some Scotch drinkers may--at least in the short run--prefer to switch to higher quality liquor among the other categories than switch to cheaper grades of Scotch. 58 Gin, which is predominantly a neutral spirit, was one of the lowest priced types of liquor. For this reason, it would be one of the distilled spirits which would be a substitute good for other less expensive types of alcoholic beverages. This would contribute to an elastic price coefficient. On the other hand, it would also be substi- tuted for more expensive distilled spirits when their prices increased. This aspect would dampen, completely offset or more than offset the elastic property less expensive types of beverages would tend to create. Our investigation indicates that overall gin was price elastic (in the neighborhood of -2.0). The heavy seasonal character of gin sales was reflected in the highly significant summer dummy variable [D (2)] and the high beta weight for this variable during the sample period. The elastic income coefficient for gin indicated both its "superior" character with respect to alcoholically weaker and less eXpensive competitors as well as more affluent alcoholic consumption patterns associated with a rising standard of living. Vodka, like gin, is a relatively inexpensive distilled spirit which is predominantly "neutral" in character. The statistical behavior of vodka was dominated by its increased popularity. Most of the equations tested indicated some autocorrelation in residuals. One of the better equations which minimizes the problem appeared when "real" income, relative prices, beer prices, population 59 and the trend variable were included. The trend variable would presumably account for some of the changes in tastes directed in favor of vodka. This should be expected to reduce autocorrelation in disturbances. The price variable was dwarfed in terms of its explanatory value by trend, seasonal, pepulation and income variables. Under this formulation the price elasticity which was not statistically too significant, was more elastic than gin 23 in a similarly structured equation. Vodka's income elasticity was negative under this hypothesis indicating that while there was a movement which favors vodka, it is-- technically speaking--an inferior good. Brandy is a rather select type of liquor which is popular during the winter holidays. It was somewhat price elastic. Its income elasticity hovered near unity. v Cordials and liqueurs form a rather heterogeneous type of liquor. They are more or less Specialty items and are quite seasonal in character. They displayed a slightly price elastic behavior. Their reputation of being luxury goods was confirmed by their elastic income parameter. Rum is a type of liquor which is popular both in the summer and during the winter holiday season. It is a very inexpensive liquor and a likely candidate for substitutions 23See Table 37, equation B and Table 38, equation C in Appendix A. 60 with cheaper alcoholic beverages as well as higher priced distilled spirits. It had a fairly strong income elasticity-- in the neighborhood of two. It may as well qualify as a "luxury" good more for specialty drinks in which it is used than as a ready substitute for other alcoholic beverages as suggested above. Price does not perform well statistically and the price elasticity seems to be impre— cisely estimated. Fortified wines have been one of the most elusive types of liquor studied. The trend variable had the most significant'explanatorw'statistical value. If this was a true measure of changing tastes or habits, this variable had some analytical explanatory value. Income ranked second in importance in terms of explaining wine consump- tion. Interestingly enough, the income elasticity coeffi- cient for wine was negative in sign and greater than one in absolute value. This definitely indicated that forti- fied wine was technically speaking, economically "inferior." As consumers' incomes rose, less and less fortified wines were purchased due to this rise. The significant wine price elasticity coefficients were negative and fairly high in absolute value. The high degree of substituta- bility between fortified wines and non-fortified wines is the first line of explanation. It is possible to purchase the same physical volume of wine at a somewhat lower price 61 but with a loss in alcoholic content. Other alcoholic and even non-alcoholic beverages may serve as other fairly close substitutes. Overall, our investigation of the liquor market in Michigan revealed the following. Because of non-proportional and imperfectly correlated price changes for the commodity "liquor" and intra-basket substitutions during the period, a slightly inelastic price parameter was derived from the composite liquor demand equations. This inelasticity was restricted to the price-quantity relationship. An examina- tion of the behavior of total expenditures on liquor revealed a definite inverse relationship between the prices and total expenditures. The subsequent distilled spirits demand analysis led to the identification of a significant cause of the conflict between the composite demand and total expenditure results by separating fortified wines from the other spirits. A high cross price elasticity between distilled spirits and fortified wines was then revealed. The aggregation of component demand functions revealed that. using the most significant parameters for the demand equa- tions and aggregating them, provided an overall price elastic demand for liquor. Income elasticities were fairly consistent between the composite, the distilled spirits and the aggregated component functions. A brief look into the component demands revealed certain other interesting points 62 about the behavior of each. The higher component price elasticities simply pointed out the fact that for any given type of liquor there are generally more substitutes than for liquor as a whole. CHAPTER III MICHIGAN LIQUOR REVENUE In view of the material discussed above, it is now possible to discuss Michigan's liquor revenues and to better understand the impact the major economic determinants have on them. Furthermore, we can eXplore the implications that changes in selected parameters will have on revenue as well as provide a basic framework for productive revenue estimation. Michigan's liquor revenue flows from two basic sources--profits from the state liquor enterprise and excise tax collections. Currently excise taxes are channeled into both the General Fund and the School Aid Fund of the State of Michigan. Each fund received close to ten million dollars in excise revenues in fiscal 1966. In that same year, the amount of revenue transferred to the General Fund from state liquor monopoly earnings was two and one- half times that amount or approxaimtely fifty million dollars. The transferred earnings were from the state's Liquor Purchases Revolving Fund. The LPRF is the operating fund of the state liquor monopoly. 63 69 Michigan's State Liquor Enterprise There are three basic sources Of receipts for the liquor monopoly. Ths most significant is the total dollar sales of liquor.1 Two minor sources of revenue are the return from the sale of liquor tax stamps to manufacturers and the "mark-up" received when liquor is brought into the State of Michigan by consumers.2 The two latter sources are relatively insignificant in size and can be ignored. The total retail dollar sales of liquor do not con- stitute the receipts of the state enterprise since this amount includes the discounts allowed the various licensees.3 These discounts are aggregated into a total discount figure which is the weighted average of the lExcise taxes do not constitute part of the receipts of the monopoly. They are merely collected by the monopoly and are deposited with the State Treasurer in the General Fund. These are discussed later in this chapter. 2The fact that certain out-of-state purchases of liquor are subject to the payment of the "mark-up" sub- stantiates the contention that the liquor Operations are involved in indirect taxation. The exceptions to the mark—up payment are found in Section 936.3 and Section 936.9 of the Liquor Control Act. 3The discount on the retail price is different for different categories of licensees. From 1955-1966 it was 10%, 129% and 22% for S.D.D.s, class "C" licensees, and the military and hospitals respectively. separate discounts. 65 The weights in this instance are primarily the value of the sales made to the separate cate— gories Of licensees, that is: where: If both D and total discount retail value of sales to the i-th category of licensees discount rate allowed the i-th category of licensees were divided by total retail sales, the resulting expression would be the overall average discount rate. Interestingly, when the overall actual average discount rate was computed for the period 1955-1966 on an annual basis, it was found to be fairly stable and just slightly above the discount rate for S.D.D.s, the dominant retailer of liquor. (See Table 17.) TABLE l7.--Discounts as a per cent of total dollar retail sales fiscal 1955-1966. l955--10.l8 1956--10.22 1957--10.23 1958--10.21 l959--10.25 1960--10.28 1961--10.29 1962--10.3l 1963--10.39 1969--10.37 1965--10.37 1066--10.39 Source: Computed from Michigan Liquor Control Commission, Financial Report for years indicated. 66 The total retail dollar sales, as we have seen in the previous chapter, depend in part upon price.. The retail price of liquor is determined in a fairly uniform fashion. Most liquor is marked up forty-six per cent of its delivered cost to the state warehouses.“ Basically retail price is determined according to the following formula: P = C(l + M) where: retail price of liquor delivered cost Of liquor "mark-up" P C M Although the foregoing formula describes how the retail price is determined, it ignores the pg valorem excise taxes on spirits. These were included in the price index used in the analysis in the previous chapter. It can be included in the preceding formulation quite easily along with a weighting factor. The latter is necessary since these excise taxes do not apply to spirits with less than twenty-two per cent alcohol. Thus we have: PT = C(l + M)(l + tw) “Special order items which are not found on the published "Price List" are charged a forty—eight per cent mark—up. These orders constitute a very small proportion of sales and need not be separated from the regular codes on the revenue side. 67 where: PT total price of liquor Cf. ll excise tax rate ratio of taxable liquor sales to total dollar liquor sales. w < l. 22 II This formula now gives us the basic pricing structure in the Michigan market as well as a tool for approximating price responses to parametric changes.5 Consequently the impact of such changes can be traced through the price elasticity coefficients developed in the previous chapter and provide a measure Of the response on total dollar retail sales. For example, if the question arose concerning the impact an increase in the mark-up from forty-six to fifty per cent might have on net liquor receipts, we could procede as follows. Total price would respond by increasing approximately 3.29 per cent (using the arc formula above) and consequently total retail dollar sales (excluding excise 5The elasticity of total price of liquor with respect to the variables C, M or t is Obtained from the following formulas if we treat w as a constant: point elasticity, arc elasticipy E( PT 0) = 1 1 _ M (M + M')/2 E( PT M) ” l + M l + (M + M')/2 _ ‘tw [(t + t')/2]w E< PT t) ‘ I‘I‘tfi l + [(t + t')/2]w An additional bit of information revealed by these formulae is that as long as M > tw, we know E(PT C) > E(PT M) > E(PT t). 68 taxes) would be depressed by approximately 6.29 per cent.6 If it is assumed that there was little or no effect on the distributional parameters between the retail licensees, the net receipts of the state enterprise would be attenuated prOportionately--note, actual sales need not fall since they may still increase if other factors such as increases in income offset the depressing price effect. Also note that this depressing effect does not mean that the state will necessarily lose money since it will now be getting a larger slice of a smaller pie. We find two general determinants of the monopoly Operation's receipts, namely, total retail dollar sales of liquor and the total amount of dollar discounts allowed against these sales. The former is determined by the major economic variables income and price. The price is anatomically determined by delivered cost of liquor, mark-up rates, and excise taxes. Discounts amount to the weighted average of three basic discount rates. This figure has been fairly stable with a very slight movement upward. The costs incurred in the distribution of liquor are basically outlined in Chart II. In view of the retail pricing structure discussed above, the costs incurred in 6The value of the sales elasticity with respect to price (-1.9909) is taken from equation B, Table 26 in Appendix A. 69 Qanufacturea Expenses: Cost of liquor Cost of shipment Phase I Warehouse Expenses: Cost of handling in and out Capital and administrative costs Cost of shipment to stores Expenses: Cost of handling in Phase III w and out Expenses: Cost of delivery Phase IV Cost of handling in and out @see Phase 11 Chart II.--Flow Chart of Costs in Michigan Liquor Distribution System. 70 Phase I on the flow chart can be expected to bear a direct proportional relationship to total dollar retail sales. If all items were marked up forty-six per cent, the cost of liquor sold would be 68.993% of total sales. The actual mark-up during the 1955-1966 period was approximately 96.3187% or 68.399% of retail dollar sales. The slight discrepancy is attributable to the special order items mentioned above. In so far as Phase I costs form the base for price determination, they affect the absolute value of the mark-up on a particular item when they change. An increase in delivered costs, for example, increases the gross profit margin on a particular code. However, since such variations in the base cause a decrease in the value of total retail sales through the price variable, increased delivered costs would have an adverse impact on revenue. Therefore it is in the interests of the Michigan Liquor Control Commission to resist price increases from the manufacturers of liquor and attempt to keep transportation costs experienced in delivery to the warehouses at a minimum. Costs incurred in Phases II and III of the liquor operation are relatively small in magnitude, ranging from $3.8 million in 1956 to $9.3 million in 1966. These costs are predominantly "variable" in character, that is, they are related to the physical volume of liquor sales rather than the dollar value of sales. This factor would tend to 71 reduce the variation of such costs more than Phase I costs when price changes occur. This follows from the relatively inelastic character of the liquor price-quantity relation- ship identified in the foregoing chapter. These costs, which are predominantly handling costs, are strongly influenced by labor costs and the overall efficiency of the distribution system. Labor costs reflect changes in wage rates, fringe benefits, and productivity emerging from more efficient handling techniques and equipment. The overall efficiency of the system is influenced by all of the foregoing factors being integrated into an optimal combination of capital facilities and flow patterns. This optimal system must Operate within the constraints and objectives set forth in the Liquor Control Act. When considering projections of these costs, changes in quantity sales, labor costs, equipment mix and any major changes in the system structure must be considered. Merchandizing costs incurred in Phase IV are not incurred by the state except in its very limited retail Operations. They are presented primarily to round out the picture of the total distributional system. Costs of the state liquor enterprise can be broken into two major categories--the cost Of the merchandise and total handling costs. The former classification is directly related to the retail dollar sales and the latter is geared to physical sales volume and input productivity and factor market conditions. 72 So far we have outlined the basic factors which must be considered to identify the revenue which accrues to the State of Michigan from the Liquor Purchase Revolving Fund. One key factor, however, needs to be identified before our model is complete. This is the increase in the State's equity in the liquor monOpoly. This primarily consists of the net changes in the value of liquor inventories over the fiscal year. The amount of such changes directly affects the size of the transfer from the LPRF to the General Fund. While our major concern is with the market parameters and their impact on revenue through sales, the inventory variable cannot be completely ignored. Since there is not at this time a systematic inventory control policy which can be relied upon for estimating purposes, an estimate of the timing of the demand parameters in the second calendar quarter may help somewhat in estimating inventory behavior. Gathering together the major elements of the receipt and cost factors involved in the State liquor monopoly Operation in Michigan, we form the following simple LPRF transfer model: LPRFt = S - D - Cl - C2 - i where: S = total retail value of liquor sales D = total discounts C = cost of liquor sold 73 O M II handling costs change in the value of inventories Sales can be determined on the basis of the demand parameters with due consideration being given to any changes in cost of liquor, tax rates or mark-up policies. Discounts can be considered a function of sales unless the discount rates are altered. Cost of liquor sold can be expected to vary prOportionately with sales. Cost of handling must be estimated in terms of projected quantity growth, labor costs, etc. Change in inventories may be partially projected on the basis Of estimated fourth quarter timing of variation in demand parameters. This model in its simplicity performed quite well when tested against actual historical data. The major problem area, the inventory behavior, notwithstanding.7 Liquor Excise Taxes Liquor excise taxes are levied on both distilled spirits and fortified wines. The former, however, are taxed on an 3d valorem basis whereas the latter are taxed at varying rates on a physical volume base.8 Since forti- fied wines constitute less than a third of the volume of wine consumed in Michigan, the amount of excise revenue 7 8Wines are taxed on a discriminatory basis. Wine made with Michigan-grown grapes receives favorable treat- ment vis-a-vis "imported wines." See Appendix C for a simulation of this model. 79 which they bring in is relatively small, less than a half million dollars in 1966. Their revenue growth is directly prOportional to their sales behavior. Thus it can be identified from the demand parameters for wine. Currently distilled spirits are taxed at a rate of eight per cent. Half of the revenue is deposited in the State General Fund and the other half in the State School Aid Fund. Since the taxes are levied on the retail sale price of distilled spirits, their yield is determined by the rate of taxation (t) and the base, namely, total dollar retail sales of distilled spirits (S). Excise taxes = t8 A change in the rate would directly affect the yield of the taxes and would affect it indirectly through its impact on price and ultimately retail dollar sales. Since liquor excise is a function of retail dollar sales, it is also affected by any variation in the non-tax price parameters as well as Michigan Disposable Income-- a major determinant of sales. It should be noted that due to the price elastic character of distilled spirits sales, excise tax yields would, ceteris paribus, be adversely affected by increases in the non-tax parameters. Excise yields would be the same proportion of a reduced base.9 It 9See Table 26 in Appendix A for regressions of demand parameters on retail dollar sales of distilled spirits. 75 is possible that the behavior of sales due to the income elasticity may outweigh and thus "hide" the loss incurred. This basic model performed very satisfactorily when tested against excise returns.lO From the foregoing material it is clear that liquor revenue is very intimately bound up with the movements of liquor demand. Both Michigan's liquor monopoly profits and excise yields are directly determined by the liquor demand parameters. The information generated by the demand analysis throws some light on the implications changes in policy parameters and the demand variables have on state revenue as well as statewide alcoholic beverage consumption patterns. While there is room for more extensive study of some of the areas of the alcoholic beverage market, at least some of the guesswork associated with liquor demand and liquor revenue has been removed. 10See Appendix c for a simulation of this model. APPENDIX A Selected Regression Equations and Elasticities 76 77 TABLE 18.--Liquor demand equations linear composite. Equation Q . " A B c D a 990.6116* 1023.7039 1222.8383 823.9995 + bl 0(1) 81.5176* 81.5063* 81.6317* 81.8226* 4. 02 D(2) 59.1982* 59.3619* 59.3088* 58.8039* 4. b3 D(3) 399.0921* 399.0189* 397.2113* 395.8905* + bu P —788.9509* —898.2993 -1088.8320* -806.6926 + (998.96) (963.27) b5 Y 0.1922* 0.1880* 0.1891* 0.2197* + b6 0 -0.l900 —0.2611 + (0.1612 (0.2027) b 0 +7 b8 B 699.9519 1278.1902 + (703.93) (919.92) b9 T 0.3991 -3.0929 (2.1881) R2 0.9658 0.9699 0.9653 0.9652 D.W. 2.1117 2.0995 2.1386 2.2659 F, Sig 0.000 0.0005 0.0005 0.0005 (overall Regression) *Coefficient significant at the .995 level or above. 78 TABLE l9.--Liquor demand equations linear composite using relative price and real income. Equation Q N c a 970.3651 833.8587 289.7213 289.2710 + 61 0(1) 81.8500* 82.3932* 83.8862* 83.8891* 4. b2 D(2) 59.0273* 59.1962* 61.3029* 61.3076* + 03 D(3) 309.6696* 309.5708* 397.5920* 397.5919* + bu P/C -891.9787 -750.8519 —993.8159 —995.9199 + (339.32) (302.71) (913.03) (977.79) b5 Y/C 0.2015* 0.2190* 0.2079* 0.2078* + b6 0 —0.0015 —0.0013 + (0.20) (0.21) 67 B/C 769.1917 771.3802 + (795.19) (818.57) b8 T -0.5395 0.0087 (1.13) (1.25) 62 0.9669 0.9653 0.9655 0.9695 D.W. 2.1162 2.1593 2.1919 2.1915 F Sig 0.000 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 79 TABLE 20.-—Liquor demand equations linear per-capita composite. Equation ‘52 A B C1 D1 a 0.2098* 0.2256 0.1966 0.1621 + bl 0(1) 0.0177* 0.0177* 0.0178* 0.0179* + b2 D(2) 0.0128* 0.0128* 0.0128* 0.0128* + b3 D(3) 0.0866* 0.0866* 0.0866* 0.0865* + buP —0.1715* -0.1892 -0.1663 -0.1958 + (0.0992) (0.0716) (0.0782) b5 Y/O 0.1889* 0.1897* 0.1979* 0.2159* + b6 T 0.0001 —0.0002 fie 0.9632 0.9622 0.9635 0.9630 D.W.. 2.0963 2.0853 2.1237 2.1796 F Sig. 0.000 0.0005 0.000 0.0005 (Overall regression) *Coefficient significant at the .995 level or above. lRelative price and "real" income statistics used. 80 TABLE 2l.-—Liquor demand equations log—linear composite. Equation ln Q A B c 0 a 0.0983* 0.1762 9.8208 7.6959 + bl 0(1) 0.0896* 0 0896* 0.0896* 0.0899* + b2 0(2) 0.0569* 0 0569* 0.0567* 0.0561* + 63 0(3) 0.3531* 0 3532* 0.3519* 0.3999* + bu In P -0.9251* -0 9796 -1.2687* -0.9807 + (0.2311) (0.9380) (0.3509) (0.9972) b5 ln Y 0.8196* 0.8039* 0.7817* 0.9362* + 66 ln B —0.5272 -1.0160 + (0.6530) (0.8099) b7 ln 0 0.8065 1.3232 + (0.6619) (0.8276) b8 T 0.0003 -0.0032 (0.0022) (0.0030) B2 0.9627 0.9617 0.9629 0.9625 D.W. 2.5363 2.5217 2.5959 2.7281 F Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 81 TABLE 22.--Liquor demand equations log-linear composite using relative price and real income. Equation 1n Q A B C D a 0.2160 -0.l389 -0.5933 -0.3992 + bl D(l) 0.0893* 0.0897* 0.0867* 0.0866* + b2 D(2) 0.0563* 0.0569* 0.0590* 0.0592* + b3 D(3) 0.3539* 0.3532* 0.3522* 0.3522* + bu 1n P/C -0.8713 -0.8075 -0.9881 -1.0599 + (0.3292) (0.3665) (0.3966) (0.9562) b5 ln Y/C 0.7991* 0.8923* 0.8295* 0.7880* + b6 ln 0 0.0651 0.0839 + (0.8390) (0.8518) 07 1n B/C 0.8850 0.9860 + (0.6866) (0.8799) b8 T -0.0009 0.0009 (0.0011) (0.0012) 32 0.9629 0.9616 0.9626 0.9616 D.W. 2,5307 2.5679 2.6301 2.6092 F Sig. 0.000 0.0005 0.000 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 82 TABLE 23.--Liquor demand equations log-linear per-capita composite. Equation 1n Q/O A B 01 01 a -1.9796* 1.9789* 1.9893* —l.9680* 61 0(1) 0.0895* 0.0899* 0.0899* 0.0899* 62 0(2) 0.0567 0.0568* 0.0568* 0.0568* E3 0(3) 0.3539* 0.3539* 0.3538* 0.3535* bu ln P -0.9205* -0.9570 -0.7990 -0.7251 + (0.2353) (0.9509) (0.3201) (0.3938) b5 1n Y/O 0.8061* 0.7928* 0.7872* 0.8536* b6 T 78:88:92.) 28:889.?) R2 0.9595 0.9589 0.9591 0.9589 0.w. 2.5097 2.9933 2.5133 2.5712 F Sig. 0.0005 0.0005 0.0005 0.0005 *Coefficient significant at the .995 level or above. lRelative price and "real" income statistics used. 83 TABLE 29.--Liquor demand equations linear composite using average price. Equation Q N D a 392.1310 507.5150 1276.6269 719.9820 + 01 0(1) 89.1163* 88.3219* 85.9879* 91.3789* 4. b2 0(2) 61.5391* 69.9310* -6l.5359* 65.9610* 4. b3 0(3) 912.9062? 930.1795' 918.6920* 936.8782* 4. bu AP -9.0255 -ll.l737 -6.1669 —15.9019 + (7.9209) (7.0750) (7.3693) (7.0886) b5 Y 0.1707* 0.2916* 0.2065* 0.2996* 4. b6 0 0.0109 -0.9179 + (0.1989) b7 B -970.8817 1756.0395 + (510.78) (935.39) b8 T -3.7670 , -8.6058 _2. R 0.9556 0.9639 0.9575 0.9667 0.w. 1.6767 1 1.9627 0.8398 2.2157 F Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) price inclusive of all excise taxes. *Coefficient significant at the .995 level or above. 1Average price is used here to mean the average final 89 TABLE 25.--Gross dollar sales regressed on demand parameters of liquor and distilled spirits.l Liquor Distilled spirits Equation: A B2 A B2 s H a 96010.852* 7.5895* 93890.988* 7.9769* + 01 0(1) 9979.616* 0.0976* 9561.972* 0.1030* + b2 0(2) 3290.920* 0.0675* 3236.668* 0.0697* + b3 0(3) 22399.138* 0.9109( 22181.962* 0.9217* + bu P -99589.810* —1.0863* -93097.932* —1.0731* + (12537.28) (0.260) (12227.159) (0.263) b5 Y 10.559* 0.9303* 10.308* 0.9386* (0.61) (0.055) - (0.697) (0.061) B2 0.9667 0.9692 0.9600 0.9566 F Sig. 0.000 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. lGross dollar sales include all excise taxes. 2Logarithmic form of the equation: lnS = a + bl 0 (l) + b2 0(2) + 03 D(3) + b9 lnP + b lnY. 5 85 TABLE 26.--Retail dollar sales regressed on demand parameters of liquor and distilled spirits.l Liquor Distilled Spirits Equation: 2 A B A B2 S It a 89293.137* 2.9828* 78792.859* 7.9738* + bl 0(1) 9316.975* 0.0987* 9382.835* 0.1092* 4. b2 0(2) 3086.109* 0.0676* 319l.9l7* 0.0721* + b3 0(3) 21291.359* 0.9099* 21071.151* 0.9217* + bu P -796l9.659* -l.9909* -79893.282* -l.8815 + (11929.81) (0.250) (11293.81) (0.259) b5 Y 9.859* 0.9306* 9.617* 0.9395* (0.56) (0.053) (0.70) (0.060) Ba 0.9670 0.9693 0.9599 0.9553 F Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. lTotal dollar sales excluding gg_valorem excise taxes. 2Logarithmic form of the equation: 1nS = a + bl 0(1) + b2 0(2) + b3 0(3) + b9 lnP + b lnY. 5 86 TABLE 27.—-Distilled spirits demand equations. Equation 8 A B 01 01 a 19895.9773 2388.631 9.3922* -5.6836 + (17297.63) (11.312) bl 0(1) 299.299 238.197 0.0506 0.0999 + (316.032) (322.09) (0.030) (0.031) 02 0(2) 221.258 229.998 0.0395 0.0901 + (316.398) (322.83) (0.030) (0.031) b3 0(3) 3916.985* 3911.929* 0.3663* 0.3658* + bu P -15658.368* -18179.728 -1.9059* -2.1985* + (9112.18) (6373.92) (0.929) (0.635) 65 Y 2 399* 1.988* 1.1161* 0.9236* + b6 0 1.520 0.7511 + (3.16) (1.313) b7 B 95.089 1.1619 (155.32) (1.539) 132 0.8881 0.8609 0.8881 0.8850 F Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. lLogarithmic form of the equation: an = a + 01 0(1) = 02 0(2) + b3 D(3) + b9 lnP + b lnY + b6 an + b lnB. 5 7 TABLE 28.--Component price and income elasticity coefficients. 87 Price Income Category linear logarithmic linear logarithmic same equationl Blends —2.3197 -2.3978 0.3899 0.9995 (0.317) (0.351) (0.072) (0.086) Bourbon -0.7751 -0.9265 0.9591 1.0189 (0.272) (0.309) (0.051) (0.061) Canadian -2.6708 -2.8997 1.9593 1.9671 (0.977) (0.911) (0.093) (0.087) Scotch -0.7116 —0.6098 2.0933 2.0673 (0.362) (0.905) (0.066) (0.080) Gin -1.8863 -2.0838 1.0236 1.1007 (0.362) (0.306) (0.070) (0.066) Vodka +2.9252 +9.9598 1.6328 1.9038 (0.739) (1.237) (0.192) (0.277) Brandy —0.3655 -0.3730 1.3389 1.3716 (0.231) (0.275) (0.058) (0.075) Cordials -1.2576 -1.9576 1.9171 1.9800 (0.396) (0.919) (0.085) (0.096) Rum +0.1101 +0.152O 1.7609 1.7669 (0.599) (0.535) (0.109) (0.109) Wine +6.9709 +7.1922 0.1900 0.2273 (1.699) (1.863) (0.123) (0.197) coefficient of best fit equation2 Wine -8.5311 -8.0723 -l.0329 -1.3252 (1.661) (1.829) (0.129) (0.179) most significant coefficient2 Scotch -1.2997 -1.6809 (0.529) (0. 539) Brandy -l.159l -1.3553 (0.353) (0.900) Rum -0.7687 -l.0518 (0.797) (0.659) Wine -8.5311 -8.0723 -1.0329 -1.3252 3 (1.661) (1.829) (0.129) (0.179)‘ lBasic equation(s) being: b3 D(3) + bu (1n)P + b (an = a + bl D(l) + b2 D(2)+ 5 (1n)Y. 2Coefficients exactly the same as those in "same equation" category except for coefficients indicated. 88 .m>onm no Hm>ma mom. on» 00 pcmofimaswfim pcmaofimmmoo* ACOHmmmnwmm Hampm>ov 000.0 0000.0 000.0 0000.0 000.0 .000 .0 0000.0 0000.0 0000.0 0000.0 0000.0 .3.0 0000.0 0000.0 0000.0 0000.0 0000.0 0m 00000.00 00000.00 00000.00 00000.00 00000.00 0 00 000.000000 00.000000 0+ 00000.000000- 0000.00000- 00000.000000- 0000.00000- 00000.000000- 0 0 + 00000.00000 00000.00000 00000.00000 00000.00000 00000.000000 0000 00 + 00000.00000 00000.0000 0000.0000 0000.0000 0000.0000 A000 00 + 00000.00000 00000.0000 00000.00000 0000.0000 00000.00000 A000 00 + 00000.000000 0000.00000- 00000.000000 0000.00000 00000.0000000 0 m CHO nopoom cmfiumcmo mCOQLdom mocmam "coapmddm .mfimmnpoazs hmmcwa mmfiaowmpMOIQSm Log mCOHumsum 000000 00:000un.00 00000 89 .m>onm no Hm>m0 mam. on» as unwOHMHCwfim pcmfiofimmooox Acoammmhwmu Hamho>ov mooo.o mooo.o mooo.o mooo.o ooo.o .wfim m 0mmo.o 00mm.0 0mom.0 omoo.0 mmmw.o .3.o 0mmo.o m«mm.o mumm.o womm.o 0mom.o mm 00000.00 . _.K 0 5033.5 *mzom.: *Nmoa.ma *mamm.w *ommm.om w 9 00.000mv A0.m00w0v 00.00000 0 *0zam.0000m00 oa:m.mm00 m0m0.m0m0m| m:mm.mw0ml *mowm.mmwmm0 m n + moo0.mmmm0 *wmam.0mmm *mmmm.oomm0 -xmwam.m0wm *mmww.mmmw0 Amva mg + oamo.mmmml *omon.mmmm *oumo.mm00| *mom0.mmoml *mmwm.mmaoa Amvm mp + 0000.0000 00000.000 00000.0000- 0000.000- 00000.0000 0000 00 + mmmma.mo~m:oal oomm.0mm00| momm.mmaaa Hm0m.0:ml 000mm.mm:m00I m mmcaz Esm Mummswwq zpcmmm mxoo> G 0 0o 0 "GOHpmsum .mfimmsuoqmc mecfia mm0yowmmeIn5m mom mCOHumzdm 000500 000000uu.00 00000 9O .o>onm no Hm>ma mom. 0:» pm pcwofimacw0m pcmfioammmoo 000.0 0000.0 0000.0 0000.0 0000.0 .000 .0 0000.0 0000.0 0000.0 0000.0 0000.0 .3.0 mwzmé :mwmé mmwmé mmmmé wowmé m-m 00000.0 00000.0 00000.0 00000.0 00000.0 000 00 0000.00 00 00000.0- 0000.0- 00000.0- 00000.0- 00000.0- 000 0 + 00000.0 00000.0 00000.0 00000.0 00000.0 A000 00 + 00000.0 00000.0 0000.0 0000.0 0000.0 0000 00 . + 00000.0 00000.0 00000.0 0000.0 00000.0 0000 00 + 00000.0 00000.0- 0000.0- 00000.0 00000.0 0 = 000 £00 nouoom swapmcmo mconpsom mUQmHm "mcoHumSUm mHmepomzn 000:001woa mmfihowomeIQSm hon 020000330 ocMEmp posvaqll.am m0m<9 91 .m>opm no 0m>m0 000. 000 00 00000000000 0:0000000000 0:000000w00 00000>ov 0000.0 0000.0 000.0 0000.0 0000.0 .000 .0 0000.0 0000.0 mmoo.m m0oo.m 0000.0 .3.Q m0m0.o 0000.0 0000.0 0000.0 0mm0.o mm 0000.0v m m0mm.o 0m000.0 00000.0 0000m.0 0mmoz.0 0:0 2 0000.00 0000.00 0000.00 0+ 0mmmH.0 omm0.o 000m0.0: om0m.o: 000m0.0 m20 n + :m0o.o mmmmm.o 00mmz.o 0ozmm.o 0mmom.o Amvo mp + 0000.0: 00000.0 00000.0: 00000.0: 0000.0 0000 00 0? 0000.0 00000.0 00000.0: 0000.0: . 0000.0 0000 00 + 00000.0 00000.0: 0000.0: 0000.0: 0000.0: 0 00203 Esm mmzmzuaq mucmmm mxwo> 020 00000000 "0CO0pmzum .0000500902 mmeHleoa mmHLmeumolndm pom 0C00u03dm vcwsmn 003d00::.mm mqm¢H 92 TABLE 33.--Component demand equations blended whiskey log-linear hypothesis. Equation: A B2 C1 D1,2 an a 0.6081* 7.6302 1.5167 9.7692* + bl 0(1) 0.1011* 0.1082* 0.1003* 0.1039* 4. b2 D(2) '0.0235 0.0306 0.0267 0.02u0 4. b3 D(3) 0.3853* 0.3825* 0.3873* 0.3853* + 6Ll In P -1.2312 —1.1u95 -2.2757* -o.5039 + (0.618) (0.607) (0.512) b5 1n Y 0.89u9* 0.9997* 0.3629* 1.1683* + b6 1n B 2.u830 + (1.22u) b7 1n 0 —0.3uu7 + (1.337) b8 T -0.0082 -0.0085* -0.0113* (0.009) fiz 0.8910 0.9048 0.87u5 0.8932 D.W. 1.7703 2.0893 1.u805 1.8999 F. Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .955 level or above. 1Per capita” income and quantity variables. 2Relative price and "real" income variables. 93 TABLE 3M.--Component demand equations bourbon whiskey log-linear hypothesis. Equation: A B2 01 01,2 an a 3.3882 -u.6113 -2.9508* 2.9358* + bl 0(1) 0.0565 0.0695* 0.0563 0.0583* + b2 D(2) 0.0257 0.0363 0.0268 0.0274 + b3 D(3) 0.u167* 0.9155* 0.9179* 0.9171* . bu in P -1.1196 -0.9606 —0.8575 -0.6589 + (0.483) (0.953) (0.9342) (0.397) b5 ln Y 0.9uu2* 0.8978* 0.93u2* 1.0760* + b6 ln B 2.8880* + b7 ln 0 0.9899 + (0.973) b8 T 0.0013 0.0019 (0.002) (0.001) E2 0.9587 0.9675 0.9523 0 9581 D.W. 2.0128 2.9102 0.8200 2.1096 F. Sig. 0.0005 0.0005 0.0005 0.000 (overall regression) .995 level or above. *Coefficient significant at the 1Per capita income and quantity variables. 2Relative price and "real" income variables. 9“ TABLE 35.--Component demand equations Canadian whiskey log-linear hypothesis. Equation: A2 B2 Cl 1,2 ln Q a -u.9296* -22.0876 -1u.0361* 3.0952* + . bl 0(1) 0.1521* 0.1su6* 0.151u* 0.1513* + 02 0(2) 0.0716* 0.0782* 0.0721* 0.0696 + b3 D(3) 0.6285* 0.6312* 0.6307* 0.6269* + bu ln P -1.9053* -1.u775 -0.9977 -2.2990* + (0.609) (0.594) b5 ln Y 1.9u5u* 1.78u3* 2.0287* 1.9176* + b6 ln B 0.9509 + (1.151) b7 ln 0 2.1892 + (1.235) b8 T -0.0079* -0.0075* -0.0117* -0.0071* P2 0.9712 0.9721 0.9733 0.9671 D.W. 1.5829 1.7107 0.9518 0.4116 F. Sig. 0.0005 0.0005 0.0005 0 0005 (overall regression) *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2Relative price and "real" income variables. 95 TABLE 36.--Component demand equations Scotch whiskey log-linear hypothesis. Equation: A B2 Cl 1,2 ln Q N a -3.0909 -u.7771* -10.2283* 2.956u* + bl 0(1) 0.1u78* 0.1538* 0.1469* 0.1537* + b2 0(2) 0.1927! 0.1u7i* 0.1439* 0.1951* + b3 D(3) 0.9981fi 0.u999* 0.9993* 0.9981* + bu ln P -1.6809* -0.7209 -1.8533* 1.9586* + (0.“96) b5 ln Y 1.65uu* 1.85u6* 1.4965* 0.9586* + b6 T 0.0072 0.0059 0.0086* 0.0056* (0.003) 32 0.9792 0.0911 0.9747 0.9819 0.w. 1.8386 2.1833 0.5791 2.3793 F. Sig. 0.0005 0 0005 0.0005 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2 Relative price and "real" income variables. 96 TABLE 37.-—Component demand equations gin log-linear hypothesis. Equation: A B2 c1 01:2 ln Q a 1.0951 28.6107* -5.7602* 3.0016* + bl 0(1) 0.3103* 0.3123* 0.3097* 0.3137* + b2 D(2) 0.5210* 0.3123* 0.3097* 0.3138* + b3 D(3) 0.3351* 0.3312* 0.336“* 0.3351* + bu 1n P -1.8716* -1.2587* -2.0752* -0.7853 + (O.u28) b5 1n Y 1.2163* 1.8601* 1.0247* 1.6974* + b6 111 B ~0.7166 + (0.979) b7 ln 0 -3.8915* + b8 T -0.0018 0.0062* -0.0066* (0.003) 92 0.9481 0.9649 0.9375 0.9522 0.w. 1.3642 2.1536 1.1681 1.7693 F. Sig. 0.0005 0.0005 -0.000 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2Relative price and "real" income variables. 97 TABLE 38.--Component demand equations vodka log-linear hypothesis Equation: A B2 C2 1,2 ln Q a 8.6759 17.8476* -112.6384* 1.4705* + 81 0(1) 0.1703 0.1518 0.1816* 0.1511 + b2 D(2) 0.2069 0.2003* 0.2624* 0.2033* + 03 D(3) 0.3038* 0.2974* 0.3131* 0.3118* + bu ln P 0.6208 —4.4019 -1.9251 -4.0091 + (1.947) (1.523) (1.164) (1.373) b5 ln Y 0.1666 -0.9311 -2.3959* -1.2924 + (0.568) (0.498) (0.494) b6 ln B 11.5757* + b7 1n 0 16.8807* + b8 T 0.0242 0.0353* 0.0410* 0.0367* (0.010) 92 0.8175 0.8520 0.9371 0.8613 D.W. 0.3433 0.3917 1.0189 0.4423 F. Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) lPer capital income and quantity variables. 2Relative price and "real" income variables. *Coefficient significant at the .995 level or above. TABLE 39.--Component demand equations 98 brandy log-linear hypothesis. Equation: A B2 C2 1,2 an a 1.6415 1.1665 —o.4909 1.4074* + bl 0(1) —0.0444* -0.0442* -0.0387 -0.0040 + b2 D(2) -0.1291* -0.1290* -0.1208* -0.1289* + b3 D(3) 0.3957“ 0.3956* 0.3948* 0.3955” + bu ln P -1.3553* -1.1276 -1.2038* —1.1081 + (0.410) b5 in Y 0.9693* 1.0280* 0.9167* 1.0518* + b6 In B 2.2779 + (0.927) b7 ln 0 0.8959 + (0.956) b8 T 0.0077* 0.0062* 0.0075* 0.0060* (0.397) P2 0.9765 0.9761 0.9787 0.9749 D.W. 2.2837 2.2829 2.4840 2.2926 P. Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) f j *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2Relative price and "real" income variables. TABLE 40.--Component demand equations cordials and 99 liqueurs log-linear hypothesis. Equation: A B2 C1,2 1,2 1n Q a —1.2035 1.7345 2.1593* 2.2965* + bl 0(1) -0.1025* -0.0935* -0.0998* —0.0983* + 02 D(2). -0.1599* -0.1524* -0.1596* -0.1593 4.. b3 D(3) 0.4655* 0.4624* 0.4663* 0.0654* + 6LI 1n P -1.7095 -1.2912 -0.8640 -0.6751 + (0.756) (0.600) (1.521) (0.555) b5 ln Y 1.3986* 1.6669* 1.5781* 1.7688* 4. b6 ln B 2.4650 + (1.151) b7 -0.6139 + (1.367) b8 T 0.0016 -0.0018 (0.004) (0.002) §2 0.9600 0.9678 0.9628 0.9628 D.W. 1.9962 2.5621 2.2197 2.3256 P. Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) fit *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2 Relative price and "real" income variables. 100 TABLE 41.--Component demand equations rum log-linear hypothesis. Equation: A B2 C1,2 D1,2 in Q N a -2,9265 2h.8609 ‘ O.8“S2* 0.6095* + bl D(l) 0.0940* 0.0928* 0.1003* 0.0996* + b2 D(2) 0.2261* 0.2125” 0.2255* 0.2287* + . b3 D(3) 0.5251* 0.5184* 0.5253“ + b4 ln P —1.0518 ~1.0808 -O.853“ -O.5624 + (0.65“) (0.700) (0.628) (0.501) b5 ln Y 1.0187* 2.8510* 2.310“* 1.5950* b6 1n B -2.6225 + (1.242) b7 ln 0 -M.2938* + b8 T 0.0071 0.0071* (0.003) E2 0.9737 0.9675 0.9621 0.9763 D.W. 1.8641 0.712“ 1.9136 2.2359 F. Sig. 0.0005 0.0005 —0.000 0.0005 (overall regression) *Coefficient significant at the .995 level or above. 1Per capita income and quantity variables. 2Relative price and "real" income variables. TABLE 42.--Component demand equations 101 wine log-linear hypothesis. Equation: A 32 Cl 01’2 ln Q a 22.5389* 9.9509 19.8903* 3.0250* + bl D(l) 0.0748* 0.0619 0.0730* 0.0581 + b2 D(2) -0.0l60 —0.0l72 ~0.0150 —0.0193 + b3 D(3) 0.0820* 0.079" 0.0833* 0.0825* 4. b4 ln P -8.0723* -0.u386 -7.3294* 1.0288 + (2.105) (0.686) b5 in Y -l.3252* -1.02U5* -l.N2OM* —l.0668* + b6 ln B 1.7997 + (2.034) b7 ln 0 1.2082 4. b8 T 0.0318” 0.0180* 0.0315* 0.0194* 92 0.8940 0.8324 0.8845 0.8329 D.W. 1.6545 0.9946 1.6304 1.0360 F. Sig. 0.0005 0.0005 0.0005 0.0005 (overall regression) *Coefficient significant at the 1Per capita income and quantity variables. 2Relative price and "real" income variables. .995 level or above. APPENDIX B Selected Time Series Data 102 103 TABLE 43.--Tota1 case sales of liquor in Michigan 1955-1966. ,_ Quarter Year - ' 1 _1. . - I II III IV - 1955 883,786 1,296,228 1956 864,155 924,440 932,176 1,287,689 1957 901,642 1,022,389 829,412 1,221,741 1958 818,798 915,073 883,291 1,173,887 1959 865,059 953,285 892,379 1,246,693 1960 905,734 1,047,623 950,093 1,288,134 1061 867,950 906,437 988,333 1,304,749 1962 902,387 1,003,855 979,110 1,349,146‘ 1963 937,106 1,044,483 1,052,688 1,398,224 1964 1,027,277 1,073,845 1,183,855 1516,229 1965 1,110,514 1,190,193 1,269,967 1,672,612 1966 1,187,639 1,298,549 Source: Michigan Liquor Control Commission, Financial Report, 1955-1966. 104 TABLE 44.--Gross dollar sales of liquor in Michigan 1955-1966. T“, —__f _, V 17 Quarter Year T 7777 1 I .I II III IV 1955 41,644,466 63,964,689 1956 40,353,647 43,314,895 44,010,852 63,347,308 1957 42,302,584 48,847,318 38,543,202 59,379,094 1958 36,658,220 40,436,881 39,310,550 54,840,899 1959 38,394,078 42,498,478 40,451,977 59,101,265 1960 40,891,479 46,648,973 42,352,978 59,567,685 1961 36,687,551 38,954,270 43,755,890 60,546,508 1962 39,373,982 45,593,536 ' 42,723,471 61,731,971 1963 40,465,294 45,381,908 46,000,919 64,384,371 1964 44,616,159 50,917,996 52,389,234 70,724,141 1965 49,361,819 53,422,096 57,202,808 79,555,124 1966 53,793,191 59,324,178 Source: Michigan Liquor Control Commission, Financial Report, 1955-1966. 105 TABLE 45.--Composite liquor price index in Michigan 1955-1966. (100:1955) Quarter Year ‘ ' 7 ' I II III IV 1955 100.000 100.008 1956 100.008 100.360 100.360 100.469 1957 100.469 101.874 105.391 105.564 1958 105.564 105.454 105.454 105.022 1959 105.022 105.284 105.284 105.326 1960 108.754 108.714 108.714 108.651 1961 109.738 109.694 106,376 106.322 1962 106.322 106,372 109.704 109.640 1963 109.640 109.629 109.629 109.904 1964 109.904 109.802 109.802 109.666 1965 109.666 109.679 109.679 109.606 1066 109.606 109.837 Sources: Michigan Liquor Control Commission, Financial Re ort, 1955-1966; Michigan Liquor Control Commission, Retail Price List, 1955-1966. 106 TABLE 46.~-Total case sales of distilled spirits in Michigan 1955—1966. Quarter Year 757 I II III IV 1955 773,844 1,156,667 1956 734,002 792,824 808,021 1,146,602 1957 769,772 883,805 698,860 1,064,242 1958 663,661 732,580 717,295 995,871 1959 701,938 778,604 747,842 1,072,297 1960 748,015 857,061 786,486 1,115,593 1961 712,381 734,677 817,125 1,119,061 1962 732,296 851,349 797,411 1,141,689 1963 752,971 847,607 862,405 1,194,181 1964 830,103 852,599 1,000,180 1,315,165 1965 920,463 1,003,699 1,079,497 1,484,754 1966 1,005,213 1,115,942 vr—w f vv . v. v— v ‘7 Source: Michigan Liquor Control Commission, Financial Report, 1955-1966. 107 TABLE 47.-—Gross dollar sales of distilled spirits in Michigan 1955-1966. Quarter Year I II III IV 1955 40,360,060 62,347,980 1956 38,891,910 41,840,740 42,743,180 61,689,570 1957 40,758,820 47,283,480 37,077,120 57,528,040 1958 34,929,810 38,412,980 37,469,070 52,725,890 1959 36,555,320 40,557,950 38,831,830 57,026,490 1960 39,060,600 46,433,050 40,479,260 57,517,860 1961 34,892,300 37,032,690 41,889,080 58,391,880 1962 37,462,090 43,562,050 40,702,260 59,451,390 1963 38,425,600 43,214,580 43,900,160 62,037,670 1964 42,432,160 48,584,000 50,211,990 68,417,600 1965 47,240,580 51,341,520 55,253,120 77,379,190 1066 51,663,250 57,484,200 T—' Source: Report, 1955 -1966. Michigan Liquor Control Commission, Financial 108 TABLE 48.--Distilled spirits price index for Michigan 1955-1966 (100 = 1955). $5 w —v— y r w fi—r Quarter Year ' I II III IV 1955 100.000 100.009 1956 100.009 100.281 100.281 100.385 1957 100.385 102.012 106.092 106.297 1958 106.297 106.280 106.280 105.745 1959 105.745 105.579 105.579 105.629 1960 109.691 109.643 109.643 109.567 1961 110.582 110.527 106.408 106.342 1962 106.342 106.404 110.496 110.418 1063 110.418 110.404 110.404 110.741 1064 110.741 110.614 110.614 110.452 1965 110.452 110.468 110.468 110.382 1966 110.382 110.535. wv a Sources: Michigan Liquor Control Commission, Financial Report, 1955-1966; Michigan Liquor Control Commission, Retail Price List, 1955—1966. 109 TABLE 49.--Michigan Disposable Incomel 1955-1966. (Millions of dollars.) Quarter Year I II III IV 1955 3593.9 3668. 1956 3603.6 3616.6 3669.3 3800. 1957 3788.1 3743.6 3752.6 3739. 1958 3690.2 3633.7 3786.0 3760. 1959 3805.6 3952.0 3949.9 3954. 1960 4097.7 4089.6 4084.5 4023. 1961 3941.0 4037.1 4067.2 4194. 1962 4180.7 4277.9 4321.4 4433. 1963 4488.5 4547.0 4626.1 4811. 1964 4920.8 5027.7 5133.0 5183. 1965 5422.8 5578.1 5659.0 4920. 1966 5899.8 6012.0 1 11 Michigan Disposable Income figure is arrived by by deducting Michigan Personal Income Taxes from Michigan Personal Income. Sources: Direct correspondence, Regional Economics Division, Office of Business Economics, U.S. Department of Commerce, June 20, 1967; U. S. treasury Department, Internal Revenue Service, Individual Income Tax Returns, selected years 110 TABLE 50.--Consumer Price Index Detroit 1955—1966 (100 = 1957-1959). 1" Quarter Year ’ If I II III IV 1955 94.7 94.6 1956 94.5 95.7 97.2 96.5 1957 98.0 98.9 99.7 99.9 1958 100.4 100.8 100.5 100.0 1959 100.0 100.1 100.8 100.8 1960 100.4 101.0 101.9 101.9 1961 102.3 101.9 101.7 101.4 1962 101.7 102.0 102.3 102.6 1963 102.6 102.7 103.9 103.6 1964 103.5 103.5 104.4 104.8 1965 104.8 106.2 106.9 107.7 1966 108.9 110.7 Source: U. S. Department of Labor, Bureau of Labor Statistics, Consumer Price Index, Bulletin No. 1351.1953-1962. 111 TABLE 51. --Michigan population 21 years and over 1955-1966 (quarterly estimates in thousands). Quarter Year I 7 I II III IV 1955 4,471 4,502 1956 4,533 4,563 4,594 4,596 1957 4,598 4,600 4,602 4,607 1958 4,613 4,618 4,623 4,617 1959 4,611 4,605 4,599 4,593 1960 4,586 4,580 4,578 4,575 1961 4,573 4,570 4,567 4,563 1962 4,558 4,554 4,550 4,561 1963 4,573 4,584 4,596 4,613 1964 4,630 4,647 4,665 4,687 1965 4,709 4,731 4,752 4,758 1966 4,764 4,770 _,—.*—v v—f Source: U. S. Department of Commerce, POpulation Estimates, Series P-25, No. 151, 172, 194, 214, 254. 112 TABLE 52.--Beer Price Index U. s. 19—5-1966 (100 = 1957- 1959). Quarter Year I II III IV 1955 96.4 96.6 1955 96.8 97.5 98.8 99.7 1957 100.1 99.8 99.9 99.9 1958 99.8 99.7 99.5 99.7 1959 99.7 99.9 101.3 101.4 1960 101.5 101.9 102.6 102.3 1961 101.9 102.0 102.4 102.5 1962 102.3 102.8 102.8 103.2 1963 103.2 103.5 103.8 104.2 1964 104.2* 104.2 104.5” 104.8 1965 104.9* 105.0 105.6* 106.2 1966 106.9 107.5 *Interpolated by Straight Line Method. Source: U. S. Department of Labor, Bureau of Labor Statistics, Consumer Price Index, Price Indexes for Selected Items and Groups, selected years. I ‘ 113 TABLE 53.--Total beer purchased in Michigan 1955-1966 (full barrels of 31 gallons) Quarter Year ‘ 7 7 I II III IV 1955 1,617,947 1,238,873 1956 1,179,971 1,483,380 1,458,378 1,262,964 1957 1,100,974 1,468,947 1,471,390 1,178,042 1958 1,061,993 1,389,367 1,399,900 1,160,697 1959 1,040,324 1,432,567 1,547,074 940,367 _ 1960 1,264,225 1,461,311 1,482,302 1,183,758 1961 1,127,473 1,416,944 1,545,066 1,150,741 1962 1,134,301 1,529,717 1,407,028 1,147,527 1963 1,124,318 1,152,400 1,466,049 1,216,981 1964 1,194,126 1,485,554 1,540,897 1,250,754 1965 1,211,214 1,538,158 1,523,482 1,248,537 1966 1,278,753 1,553,684 Source: Report, 1955-1966. Michigan Liquor Control Commission, Financial APPENDIX C Revenue Estimate Simulations 114 115 Liquor Purchase Revolving Fund The simple revenue model deve10ped in Chapter III states: - C - i LPRF = S - D — C 2 t 1 To help evaluate this model the following simulation was run based on recent data which described the behavior of economic parameters identified in Chapter II. The results are compared with actual revenue yields. Sales for the folowing fiscal year are estimated according to the following procedure: * - St (1+rla) _ 31 St (1+rla)(l+r2a) = 32 St (1+rla)(l+r2a)(1+r3a) = 83 S (1+rla)(1+r2a)(1+r3a)(1+rua) = S“ t 4 E S S i=1 1 t+l ’ 4 where:, St(+1) = total retail liquor sales in year t(+l) r1 = rate of growth of disposable income in quarter 1 a = the income elasticity of sales *If measurable price changes are to take place, the bracketed term for that quarter is modified as follows: (1+r1a+niz) where: n1 is the percentage change in price in the ith quarter and z is the price ealsticity of sales. 116 Actual quarterly growth rates of Michigan Personal Income as reported in the Surveygof Current Business (October, 1967) were used as proxies for disposable income growth during fiscal 1967. The two liquor income elasti— cities derived from equations A and B in Table 26, Appendix A and the approximate retail liquor sales figure for fiscal 1966 ($250 million) were also employed to compute the estimates of 1967 sales. = 3.860 = 0.952 r = —l.055 r“ = 2.395 r'1 r2 3 a = (linear) 0.8675 a' = (log) 0.9306 It should be noted that the total dollar sales of the preceding year are used and not the last quarter's sales. This approach avoids the need for seasonal adjustments. Discounts would be estimated on the basis of the model presented in Chapter III and relies on the foregoing sales estimate. During fiscal 1967 the discount allowed Specially Designated Distributors was changed from 10% to 11.5% effective February 26, 1967 (enrolled House Bill No. 3236). Thus to estimate discounts, 1966 sales distribution weights were assigned the discount rates for the respective categories 117 of customers and the resulting overall discount rate proportioned to the sales estimated to be made during the respective periods of time. d1 = 10.0% (11.5%) Wl = .799 d2 = 12.5% W2 = .229 = = * d3 22.0% W3 .004 The weighted average discount rate prior to February would be 10.39% and after the change approximately 11.51%. Assuming approximately 2/3 sales would take place prior to the change and 1/3 after, the overall weighted average rate would be approxaimtely 10.75% of sales. 2222 of liquor sold (C1) would be eXpected to approximate the same proportion of sales as 1966 costs since there was not reason to anticipate changes in the percentage mark-up, i.e. C1 = 0.68344 x Sales. Costs of distribution (C2) would be contingent upon expectations of growth in physical volume and any extra- ordinary expenses which might be tied to the efficienty of the distribution system. Since it is not the purpose of the demand analysis to account for the latter cost factors, *These do not total 1.000 due to retail sales by the Michigan Liquor Control Commission. 118 we will use the actual costs incurred to arrive at the model's estimate. Increase 13 inventories (i) would have to be estimated outside the framework of the demand model. We will simply use the actual figures for this component. Our results are presented in Table 54. TABLE 54.-~Simu1ations of liquor purchase revolving fund transfer model for fiscal 1967 (in millions of dollars). Linear Logarithmic Actual Hypothesis Income Income Reported Elasticity Elasticity Piguresl Sales $260.12 260.86 260.13 Discounts -27.96 -28.04 -27.95 Cost (1) -l77.78 ‘ -178.28 -177.39 Cost (2) -1.67 -1.67 -1.67 Inventory -2.79 -2.79 -2.79 Adjustment “9.92 50,08 50.33 LPRFt (actual less estimates) 0.41 0.25 ———— 1 Source: Michigan Liquor Control Commission, Financial Report (June, 1967). 119 Liquor Excise Taxes: The simple revenue model developed in Chapter III for excise taxes states: T = t8 Sales of liquor subject to the excise tax during the fiscal year are estimated according to the following procedure: (St (1+rlb)* = Sl St (1+rlb)(l+r2b) = S2 St (1+rlb)(1+r2b)(1+r3b) = S3 St (1+rlb)(1+r2b)(1+r3b)(1+rub) = 84 a 1:1 Si S =._..__.....____... t+l 4 where: = total retail distilled spirits sales S t<+l> in year t(+l) r1 = rate of growth of disposable income in quarter 1 .b = the income elasticity of sales Actual quarterly growth rates used in the LPRFt model were employed with the two distilled spirits income elasticities derived from equations A and B in Table 9, Appendix A, and the approximate retail sales figure for fiscal 1966 ($241.8 million) to compute 1967 sales estimates. _7 fi *If measurable price changes are to take place the bracketed term for that quarter is modified as follows: (1+r1b+mix) where: m1 is the percentage change in price in the i-th quarter and x is the price elasticity of sales. 120 b = (linear) 0.8801 b' = (log) 0.9306 When the four per cent tax rate was applied to the sales figures the following results were obtained: b = $10.07 million b' = $10.10 million actual: $10 million BIBLIOGRAPHY Articles Burton, Edith T. "Quarterly Estimates of State Personal Income: A New Series," Surveypof Current Business, XLVI (December, 1966), 13-15. V Cline, Denzel D. "Tobacco and Alcoholic Beverage Taxation," Michigan Tax Study Staff Papers, 1958. Lansing, Michigan: State of Michigan, 431-443. Prest, A. R. "Some Experiments in Demand Analysis, " The Review of Economics and Statistics, XXI (February, 1949), 33- 99. Simon, Julian L. "The Price Elasticity of Liquor in the U. S. and a Simple Method of Determination," Econometrica, XXXIV (January, 1966), 193-205. Books Baumol, William J. Economic Theory and Operations Anal sis. Englewood Cliffs, New Jersey: 'Prentice- HalI, Inc., 1965. Ezekiel, Mordecai and Fox, Karl A. Methods of Correlation and Regression Analysis. New York: John Wiley and Sons, Inc., 1959. Fisher, Irving. The Making of Index Numbers. New York: Houghton Mifflin Company, 1927. Friedman, Milton. Essays in Positive Economics. Chicago: The University of Chicago Press, 1953. Hicks, J. R. A Revision of Demand Theory. Oxford: At the Clarendon Press, 1956. Klein, Lawrence R. An Introduction to Econometrics. Englewood Cliffs, NewfiJersey: Prentice-HalI, Inc., 1962° 121 122 Johnston, J. Econometric Methods. New York: McGraw- Hill Book Company, Inc., 1963. Licensed Beverage Industries, Inc. The ABC's of Alcoholic Beverages. New York: Licensed Beverage Industries, Inc. . The Alcoholic Beverage Industry and the National Economy;, A State byTState Analysis. New York: Licensed Beverage Industries, Inc., 1965. . The Story of the Distilled Spirits Industry. New York: LIcensed Beverage Industries, Inc., 1964. Malmquist, Sten. A Statistical Analysis of the Demand for Liquor in Sweden.7 Uppsala, Sweden:7 University of Uppsala, 1958. Michigan Tax Reports,1967. Chicago, Illinois: Commerce Clearing House, Inc. Niskanen, William Arthur, Jr. The Demand for Alcoholic Beverages. Chicago: The University of Chicago, 1962. . Taxation and the Demand for Alcoholic Beverages. Santa Monica, California: The Rand Corporation, 1960. Schultz, Henry. The Theory and Measurement of Demand. Chicago: The UnIVersity of Chicago Press, 1938. Stone, Richard. The Measurement of Consumers' Expenditure and Behavior in the United Kingdom, 1920-1938. Cambridge, England: University Press, 1954. Wattel, Harold L. The Whiskey Industry: An Economic Analysis. New York: New School for Social Research, 1953. Wold, Herman and Jureen, Lars. Demand Analysis. New York: John Wiley & Sons, Inc., 1953. Yamane, Taro. Statistics, An Introductory Analysis. New York: Harper & Row, 1964; 123 Public Documents Liquor Control Commission. The Michigan Liquor Control Act and Rules and Regulations Governing'the‘SaIEIof AIcéholic Beverages at Retail. Lansing, Michigan: State 0? Michigan, 1964. Michigan Liquor Control Commission. Financial Report. Lansing, Michigan: State of Michigan, 1955-1967. Michigan Liquor Control Commission. Retail Price List. Lansing, Michigan: State of Michigan, 1955-1967. State of Michigan. The Executive Bud et. Lansing, Michigan: State of Michigan, 1946—1967. U. S. Department of Commerce, Bureau of the Census. Po ulation Estimates, Series P-25, Numbers 151, I73 I94 214 254. Washington, D. C.: Government 3 0 0 Printing Office. U. S. Department of Commerce, Office of Business Economics. "Table A--Quarter1y Total Personal Income, by States and Regions," Survey of Current Business, XLVII (October, 1967), 9. I U. S. Department of Labor, Bureau of Labor Statistics. Consumer Price Index, Bulletin Number 1351. Washington, D. C.: Government Printing Office, 1953-1962. . Consumer Price Index, Price Indexes for Selected Items and Groups. Washington, D. C.: Government Printing Office, 1955-1966. U. S. Treasury Department, Internal Revenue Service. Individual Income Tax Returns. Washington D. C.: Government Printing Office. Direct Correspondence. Regional Economics Division, Office of Business Economics, U. S. Department of Commerce, June 20, 1967. MICHIGAN STATE UNIV. LIBRQRIES llHI“WII”WIWIWNUIHII‘IM‘MNHIWWI 31293101378812