THE EFFECTS OF REGULATIONS UPON STATE LIQUOR PRICES Julian L. Simon, University of Maryland, College Park, and David M. Simon, Wildman, Harrold, Allen & Dixon Chicago, Illinois Send proofs and offprints to Julian L. Simon, 110 Primrose Street, Chevy Chase, Md., 20815, USA, phone 301 951 0922, fax 301 951 8468 page 1 liqrpm article8 April 11, 1996 Abstract Several estimation methods agree that state regulations such as retail price maintenance and retail price posting affected the prices of liquor brands up to the mid-1970s in the U. S. states in which the distribution system is privately owned; before- versus-after analysis using the quasi-experimental method provides the strongest evidence. The effects of particular regulations are not so clearcut, however. In the 1970s, the regulations supporting these practices began to be removed. The regulations that continued in effect seem to have lost their potency about that time. The effects of regulation no longer are seen. Key words: regulation, liquor, fair trade JEL codes: L3, L5 page 2 liqrpm article8 April 11, 1996 THE EFFECTS OF REGULATIONS UPON STATE LIQUOR PRICES Julian L. Simon and David M. Simon* INTRODUCTION As of 1963, there were 26 U. S. states in whose liquor systems the retail liquor stores are privately owned; in 16 other states (known in the industry as "monopoly" states), the state government owned the retail liquor stores, while the other 8 states are mostly privately owned but their data do not permit analysis here. This general pattern had continued with little change since the end of Prohibition in 1933 - state ownership was intended to protect the public from too-aggressive selling of alcohol - and the pattern continues in much the same form today. There are many variations in the natures of government ownership, including a single county (Montgomery, in Maryland, just outside of Washington, D. C.) where the government runs the stores), but these states are outside the scope of this paper except to note that their practices have evolved over the years, in part in response to growing freedom of competition in the "private" states. As of 1963, the various anti-competitive regulations - including "fair trade laws (resale price maintenance - RPM), mandatory markup laws, and price-posting laws1 - resulted in prices being generally higher in the "private" states than in the "government" (Simon, 1966a). But there was gradual reduction in state regulation; between 1961 and 1983, federal and state govrnments abolished most of the regulatory laws. In particular, a 1975 federal law eliminated the anti-trust law exemption for state fair-trade laws. Apparently as a consequence, as of 1983 (when the supply of data ended), prices had become generally lower in private states (J. Simon and D. Simon, 1995). This paper delves into the shifts over the years in those regulations, and into their effects. It compares liquor prices among states with private distribution systems having various types of rules or no rules, using a longer period and different methods than have been used in such earlier studies as those of Ferguson (1982), Luksetich (1975), and Ornstein and Hanssens (1985, 1987). We examine the patterns in various years, and note changes in the patterns over the examination period, during which the extent of rules changed greatly. The results cast light on the nature of regulatory effects, and also help explain the observed differences in prices between states with privately- owned and state-run systems. The results from this study cast light upon vertical restraints in general, and resale price maintencance in particular, in other industries in the United States as well as in Europe. Though some structural conditions in Europe differ from those in the U. S., the benefits of having a long data series for a large sample of states may make the conclusions of value nevertheless, especially because they seem to conflict with the broad sweep of the European literature. Our analysis ends with 1983 because the main data - the surveys of retailers' prices reported in the annual Gavin-Jobson Associates Liquor Handbook - end at that year.2 REGULATION VERSUS NO REGULATION Yearly Comparison The variety of regulations, the difficulty of deciding which regulations should be classed together, and the small numbers of states in some categories make many specific comparisons difficult. Nevertheless, Table 1 shows that for a representative brand, states with any price-regulating rules other than only price-posting at the wholesale level (column 2 in Table 1) have higher prices than do states with no rules other than wholesale price-posting (column 1; for the Old Crow brand, 20 out of 23 among the differences shown in column 3 are of the same sign, and a sign test (equivalent to asking the probability of getting 20 or more heads in 23 coin tosses) indicates that the likelihood of this happening by chance is far below one percent. (For Seagrams, our other representative brand, the pattern is even more one-sided.) This conclusion must be qualified, however, by the facts that a) the no-rules sample is too small for those observations to be given much weight prior to 1976, at which time the presence or absence of a single state may be seen to alter the sign of comparisons, and b) RPM regulations disappeared rapidly after 1975. Another way of looking at the matter is to combine into a single index all the seven brands for which there are publicly- published data. This is a comparison of the mean ranks3 for states with and without either RPM or retail price posting (the numbers in the groups being roughly those shown in columns 1 and 2 in Table 1): Comparison of States With and Without RPM and/or Price Posting Mean Rank Mean Rank With Without Difference 1961 28.7 11.3 17.4 1964 24.9 12.6 12.3 1969 25.5 18.9 6.6 1974 21.5 21.6 - 0.1 1978 24.0 21.0 3.0 1983 21.1 23.1 - 2.0 This pattern, too, shows that regulation raised prices in the earlier years. The effect of regulation was declining even before RPM was made illegal, however. By the end of the study period -- when RPM no longer existed and there was only retail price posting -- states with regulation had mean prices as low or lower than non-regulation states. This result is qualified by our only having data for national brands, rather than also including unadvertised and retailer's brands. Nevertheless, it seems conclusive that after 1978, regulations were not effective in maintaining prices above a non-regulated level. For greater precision we adjust for differences in excise taxes. Regressions run with the states' prices as a function of various combinations of a) taxes, b) presence of rules, and c) both variables, suggested that the tax affects price roughly penny for penny as the simplest theoretical analysis suggests. We therefore adjusted all retail prices accordingly, to a base of no tax, and re-did Table 1 for Old Crow, as seen in Table 2. The effects of rules are not seen much more sharply than in Table 1, however, if indeed there is any difference at all. Table 2 Before-and-After Comparisons Comparing a state's own prices before and after a change in rules circumvents the troublesome heterogeneity among the states -- heterogeneity that adding "background" variables in an equation cannot cure. Using this device, Ferguson (1982) studied the change in New York State retail bottle prices following the 1964 abolition of "non-signer RPM" (fair-trade prices mandatory for non-signers as well as for signers of fair-trade agreements, with a weaker "non-mandatory" signers-only form of RPM was in force afterwards) together with removal of limits on the number of package store licenses. Using two sets of data -- fair trade prices, and store survey prices (both lowest and average) -- he found "substantial declines" from the fair trade prices, especially for quart bottles (in comparison to fifths), from 1965 to 1967. From 1967 to 1969, the lowest survey prices declined again, even though recommended fair trade prices went up somewhat.4 Ferguson's study is useful. But it has the drawbacks that a) the increase in the number of outlets throws into question the extent to which the changes in RPM and in outlets are responsible for the observed results, and b) the observed results could be caused by other factors operating over that period in New York. In a similar study, Luksetich (1975) found that after Minnesota dropped non-signer RPM in 1969, retail prices in privately-owned stores did not rise as much as did wholesale list prices, which implies that the markup in retail stores fell due to the abolition of non-signer RPM. Luksetich also found that after the abolition of non-signer RPM, prices in municipally-run liquor stores were higher than in private stores, as well as there being much less variation in prices among municipal stores than among private stores. He writes that there is no evidence of additional revenue being produced for the municipalities that would explain the higher prices, and he suggests that the difference is due to poorer management in the municipally owned stores. But Luksetich's work, too, has the drawback that his study covered only one state and is without controls. The quasi-experimental method is a systematic device for avoiding heterogeneity by making controlled before-and-after comparisons of RPM rule changes. (The formula for this method apparently was first set forth in Simon, 1966b; for analysis of the method, see Lyon and Simon, 1968; see Baltagi and Goel (1990) for a recent application to liquor price elasticity). The computation formula is Trial state price "after" minus trial state price "before" trial state price "before" minus Comparison states prices "after" minus comparison states prices "before" comparison states prices "before" Concerning the choice of comparison states: State-owned distribution systems had no RPM changes, of course, and they constitute a set of goodly size. But state-store prices as a group rose relative to Private prices over the entire period, the very phenomenon we seek to understand (see Simon and Simon, 1995). Therefore, comparison with state-owned systems introduces a bias in the direction of the observed difference. We therefore performed the comparisons against private-distribution states where there were no changes in RPM-type rules for two years before or three years after the "trial" state's RPM change.5 The median of the comparisons is -1.22% and the mean is - 1.38%, as seen in Table 3. Among the 30 comparisons6, 24 show the expected negative result, a number far greater than chance; by the sign test, this is highly significant, the probability of it occurring by chance being far less than one percent (see the notes below Table 3 for details). This is solid evidence that RPM-type regulations raise prices. (Further details and interpretation are given below in Table 3, and the same data are grouped by type of regulation in Table 4.) Tables 3 and 4 Other Evidence Indirect evidence corroborating that RPM-type rules affect price comes from the pattern of consumption related to the rules. Ornstein and Hanssens found that "The presence of RPM is related to a decrease in per capita consumption of about 8 per cent" (Ornstein and Hanssens, 1987, p. 5, italics in original). This estimate comes from a pooled regression for 1974-1978 that included price as a regressor, and hence this is a partial rather than a full estimate of RPM's effect. Regressions for the individual years (results not shown here) do not show a consistent pattern, apparently because the sample sizes are so unbalanced at both ends of the period. And pooled regression is not likely to be reliable for the 1961-1983 period (much longer than Ornstein and Hanssens used), even with individual year dummies and other adjustments, because during the early years almost all states had RPM-type regulations, and in the later years practically none did; furthermore, there is much evidence for taste changes over that period. But the quasi- experimental method can also provide reliable evidence on consumption, because it allows for trend effects in addition to not being affected by state heterogeneity. Here we can use all states, including state-store systems, for comparison. Table 5 provides solid evidence that dropping RPM-like rules leads to an increase in consumption, from which we can infer that RPM raises price. (If dropping RPM resulted only in a shift of consumption among brands, with no difference in properly weighted price structure, then no differences in total consumption should be seen.) If we apply an elasticity of .8 (Simon, 1966b; Baltagi and Goel, 1990) to the median observation of 2.5% increase in consumption, we infer an associated price effect of 3.1% for the three-year period, which is even larger than the directly-observed quasi-experimental estimate shown in Table 3. Hence this indirect evidence from consumption supports the previous price analyses. Table 5 Ornstein and Hanssens (1987) found that the prices of liquor-store licenses in California fell between 23% and 25% following the repeal of RPM-type rules in 1978. This provides additional strong corroboration that RPM raises prices. SPECIFIC REGULATIONS The only conclusion about specific regulations that can be drawn from Table 1 is that retail price posting (RPP) increased price. This conclusion is based on our comparisons of RPP plus RPM versus RPM alone from 1961 to 1975, and of RPP versus no rules starting in 1976 (though note the reversals in the last few years).7 DISCUSSION We investigated the relationship of RPM-type rules to the number of outlets. A pooled regression indicated more outlets in states with RPM, and the quasi-experimental method showed a reduction in outlets after dropping rules. But because entry is regulated almost everywhere, and changes in the number of licenses are subject to many political forces as well as economic forces, an independent study in depth would be necessary to understand this phenomenon well. See Smith (1982) for a public- choice perspective on related matters. SUMMARY Several estimation methods agree that state regulations such as retail price maintenance and retail price posting affected the prices of liquor brands up to the mid-1970s; before-versus-after analysis using the quasi-experimental method provides the strongest evidence. The effects of particular regulations are not so clearcut, however. In the 1970s, the regulations supporting these practices began to be removed. The regulations that continued in effect seem to have lost their potency about that time. The effects of regulation no longer are seen. page 1 liqrpm article8 April 11, 1996 REFERENCES Baltagi, Badi H. and Rajeev K. Goel, "Quasi-Experimental Price Elasticity of Liquor Demand in the United States: 1960- 83", American Journal of Agricultural Economics, May, 1990, pp. 451-464 Ferguson, James M., "Retail Prices and Liquor Regulation in New York," xerox, November, 1982. Luksetich, William A. "A Study of Regulation: The Minnesota Liquor Case." Southern Economic Journal, January 1975, 457-465. Lyon, Herbert C., and Julian L. Simon, "The Price Elasticity of Demand for Cigarettes in the United States," American Journal of Agricultural Economics, Vol. 50, November, 1968, pp. 888-895. Ornstein, Stanley I., and Dominique M. Hanssens, "Alcohol Control Laws and the Consumption of Distilled Spirits and Beer." Journal of Consumer Research, Vol. 12, September 1985, 200-213. Ornstein, Stanley I., and Dominique M. Hanssens, "Resale Price Maintenance: Output Increasing or Restricting? The Case of Distilled Spirits in the United States." The Journal of Industrial Economics, Vol. XXXVI, September, 1987, pp. 1-18. Simon, Julian L., "The Economic Effects of State Monopoly of Packaged-Liquor Retailing." Journal of Political Economy, April, 1966a, 188-94. _____, "The Demand for Liquor in the U.S. and a Simple Method of Determination." Econometrica, January, 1966b. _____, and David M. Simon, "Public Ownership Versus Private Enterprise: State Liquor Distribution Systems Revisited", unpublished, 1995. Smith, J.K. "An Analysis of State Regulations Governing Liquor Store Licensees." Journal of Law & Economics, October 1982, 301-319. page 2 liqrpm article8 April 11, 1996 ENDNOTES *This paper is an outgrowth of a senior thesis by the second author. We appreciate useful information from Steve Barsby and James M. Ferguson, and helpful comments from Stanley Ornstein and Dennis W. Carlton. Gary L. Marshall of the Distilled Spirits Council of the U. S. provided the information that "After 1984, DISCUS no longer published the retailer's prices for distilled spirits sold at retail establishments" (correspondence of October 25, 1993). 1Ferguson also found additional price reductions after removal of the ban on retail price advertising in 1969 even though prices had already declined substantially after the removal of mandatory fair trade. Because retail prices declined even though supplier and wholesaler posted prices did not decline, Ferguson interpreted the data as being more consistent with the hypothesis of a retail cartel than with the supplier- price-discrimination hypothesis. The increase in the number of liquor stores when the limit was lifted in 1965 also fits with his view that the retailer cartel had held prices high before 1965. 2 The varieties of RPM included fair-trade laws applicable only to signers of fair-trade contracts, applicable to both signers and non-signers, and mandatory RPM. Price-posting requirements were at the wholesale and/or retail levels. More specific definitions and the details of the individual states' rules are available upon request of the second author. 3 The means were derived by ranking the prices of each of the brands in each of the states, then averaging the ranks for the brands for each state, and then averaging the states in each group. The ranks total diffeent sums in different years because data were not available for every state for every brand for every year. 4For the three years from 1984-1986, prices obtained from the Distilled Spirits Council (a trade association) were published. These prices, however, are not usable for several reasons: 1) The entries for each state are identical for the three years. Unlike the previous data, these are not retailer's prices but rather are recommended prices, which are far less meaningful than transactions prices in this context. 2) No data are given for Old Crow, one of the two main series available earlier. 3) The data are not continuous with the earlier data set, and show several discrepancies with them. 5The 3 year period was chosen because analysis using State- run systems for comparison showed a large rise from the one-year comparison to the three-year comparison, but no further rise to the 5 year period. The means and medians were for l year: +.049%, -.0.42%; 3 years: -2.27%, -2.22%; 5 years: -2.79%, - 2.04%; these estimates are biased in a negative direction, but have the virtue of consistency by reason of the same states being included in all comparisons, which is not feasible with Private- state comparisons. 6Because of differences in comparison-sample sizes, the observations differ in the weights they deserve and in whether other observations are being made in the same year. But none of this is likely to affect the estimate. 7Price posting at the wholesale level has no observable effect in Table 1, which is consistent with Ornstein's and Hanssens's finding for 1974-1978 (1985, p. 209, and letter from S. Ornstein of Dec. 21, 1987). (It would be imprudent to conclude that it has no effect; it certainly was intended to raise prices). No difference between non-signer RPM and signer-only RPM is apparent in the small samples. page 3 liqrpm article8 April 11, 1996 page 4 liqrpm article8 April 11, 1996