Effects of state regulations on marketing ... - Wiley Online Library

Effects of State Regulations on Marketing Margins and Price Transmission Asymmetry: Evidence From the New York City and Upstate New York Fluid Milk Markets Robert Romain Maurice Doyon Mathieu Frigon Department of Economics of Agrifood and Consumer Sciences, Faculty of Food and Agriculture Science, Université Laval, Québec, Canada G1K 7P4. E-mail: [email protected], [email protected], [email protected]

ABSTRACT A marketing margin model that allows testing for constant returns to scale technology and asymmetric marketing costs and farm price transmissions is proposed+ Results indicate that a constant returns to scale technology cannot be rejected+ During the period prior to the enactment of the price gouging law in June 1991 by the New York State Legislature, significant short-run and long-run asymmetries in both marketing costs and farm price transmissions were identified+ After 1991, these asymmetries were no longer significant or were reduced substantially+ Finally, the legislative change that occurred in 1987, allowing Farmland Dairies’ entry into the New York City fluid milk market, contributed significantly to reducing marketing margins in the New York City fluid milk market+ @EconLit Citations: D400, C300# © 2002 Wiley Periodicals, Inc+

1. INTRODUCTION Hansen et al+ ~1994! differentiated two types of asymmetric price transmission: short-run and long-run+ Short-run asymmetry occurs when the immediate effect of a variation in the farm price on retail price is not the same when farm price is increasing as when it is decreasing; in the long run, the effects can be the same+ Long-run asymmetry occurs when an increase or a decrease in farm price is not fully transmitted to the retail price after a complete adjustment period+ In the short run, the impacts could be similar+ Thus, long-run price transmission asymmetry implies an irreversible adjustment in the marketing margin ~increase or decrease!, while short-run asymmetry reflects only a temporary effect+ Numerous studies have analyzed asymmetric price transmission+ Most have shown that increases in the farm price are transmitted more fully to the retail price than decreases ~Kinnucan & Forker, 1987; Novakovic, 1991; Emerick, 1994; Hansen et al+, 1994; von Cramon-Taubadel, 1998!+ This result led Hansen et al+ to conclude that asymmetric price Agribusiness, Vol. 18 (3) 301–315 (2002) © 2002 Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/agr.10019

301

302

ROMAIN, DOYON, AND FRIGON

transmission is one of the main factors that has driven the expansion of the farm-retail price spread since the late 1980s in the U+S+ fluid milk market+ The reasons underlying asymmetric price transmission have not been rigorously identified in these various studies+ Recently, Azzam ~1999! argued that “explanations of how and why asymmetry occurs have so far been casual” ~p+ 525!, and he demonstrated that intertemporal optimizing behavior among spatially competitive retailers, when repricing costs are involved, could be a possible explanation for asymmetry+ Using a spatial competitive model, Azzam also showed that an increase in local competition decreases the farm-retail markup, and vice versa+ The present article does not propose new explanations for asymmetry; however, it does provide empirical evidence that increased competition reduces the farm-retail price spread using the New York City and Upstate New York fluid milk markets+ We also show that the change in New York State dairy policy regarding maximum markups for fluid milk ~price gouging law! did have a significant impact on farm price transmission asymmetry+ The next section presents a short literature review as well as the theoretical framework of the study+ The subsequent section presents the empirical model and appropriate tests related to different types of asymmetric price transmission+ The description of the data and the empirical results follow, while the conclusions are presented in the last section+ 2. FARM-RETAIL PRICE SPREAD MODELS Kinnucan and Forker ~K&F! ~1987! combined the markup model proposed by Heien ~1980! and the asymmetric price transmission testing procedure originally proposed by Wolffram ~1971! and further improved by Houck ~1977! to specify and estimate nonreversible functions for major dairy products+ K&F assess the asymmetric response in the retail price to variations in the farm price by regressing the retail price on upward and downward movements in farm price+ This approach has also been used in the citrus industry ~Pick, Karrenbrock, & Carman, 1990!, in the beef, lamb, and pork industry ~Griffith & Piggott, 1994!, in the dairy industry ~Emerick, 1994!, and in the peanut industry ~Zhang, Fletcher, & Carley, 1995!+ One limitation of the above methodology is the assumption of constant returns to scale+ Indeed, the general form of the model does not include the level of output; it is given by R ⫽ f~F, MC!, where R is retail price, F is farm price, and MC is a vector of marketing costs+ Moreover, the empirical specification used in the above studies does not allow for direct testing for complete transmission of farm price increases and decreases+ Complete price transmission ~discussed further later! implies that the total variation in farm price is transmitted to consumers+ To test for this hypothesis, the retail and farm prices have to be measured in the same units ~i+e+, $0gallon or $0lb!, and the processing technology should be taken into account unless the product is characterized by a fixed unitary proportion technology+ This was not the case in these studies+ The marketing cost model offers an interesting alternative to the markup model+ This model reflects that the marginal cost of offering marketing services equals the marginal value of these services ~Wohlgenant & Mullen, 1987!+ It is expressed as follows: M ⫽ f ~Q, MC!

~1!

where M is the farm-retail price spread, Q the quantity of the agricultural commodity marketed, and MC a vector of marketing costs+

NY FLUID MILK MARKETS

303

The marketing cost model has been widely used in several industries to explain the farm-retail price spread ~Wohlgenant & Mullen, 1987 @beef#; Thompson & Lyon, 1989 @orange#; Faminow & Laubsher, 1991 @corn#; and Lyon & Thompson, 1993 @milk# !+ Although the performance of this model did vary significantly among studies, Lyon and Thompson ~1993! found it robust for the U+S+ fluid milk market+ They indicate that the marketing cost model performs especially well with monthly spatially disaggregated data+ Furthermore, and unlike the markup model, the marketing cost model does not assume constant returns to scale+1 The model used in this study is a general version of the markup model+ Consider the general form of price transmission as defined by Griffith and Piggott ~1994!: R ⫽ f ~F, MC, Q!,

~2!

which is the markup equation derived by Heien ~1980!, without imposing a constant returns to scale technology+ The middleman’s margin is expressed as: M ⫽ R ⫺ F ⫽ f ~F, MC, Q! ⫺ F ⫽ g~F, MC, Q!

~3!

This specification is a variant of the marketing cost model that allows testing several forms of asymmetric price transmission, as discussed in the next section+ 3. THE EMPIRICAL MODEL AND APPROPRIATE TESTS 3.1. The Empirical Model To test for both short-run and long-run asymmetry, the methodology proposed by Houck ~1977! to estimate irreversible functions is endorsed+ Considering that asymmetry can also be observed in marketing costs, both farm price and marketing costs transmissions are tested+2 Moreover, a lag structure is associated with the quantity variable, and the sum of its parameters allows testing for constant returns to scale technology+ The empirical specification of equation ~3! is given by equation ~4!: n1

n2

M ⫽ a 0 ⫹ a 1 * TREND ⫹ ( a 2A, i * INCMCt-i ⫹ ( a 2B, i * DECMCt-i i⫽0

k

⫹

m1

i⫽0

m2

11

( a 3, i * Qt-i ⫹ i⫽0 ( a 4, i * INCFt-i ⫹ i⫽0 ( a 5, i * DECFt-i ⫹ i⫽1 ( bi * Di i⫽0

~4!

where TREND is a trend variable that emanates from Houck’s approach, INC and DEC stand for increase and decrease, respectively, of marketing costs ~MC!, and farm prices 1 The use of this model for the U+S+ fluid milk industry is especially relevant+ The governmental regulation in the industry makes the Class I price of milk known at least 1 month before a sale is actually concluded+ At that price, processors can usually buy as much milk as they wish+ Thus, we can subtract farm price from both sides of the profit maximizing condition in a competitive market ~retail price ⫽ marginal cost! and obtain the following identity: farm-retail price spread ⫽ marginal marketing cost+ 2 Kinnucan and Forker did not test for asymmetry in marketing cost because their cost index did not exhibit declines+ This is likely due to their use of nominal values, while real values are used in this study+

304


~F!, while Di , i ⫽ 1–11, are monthly binary variables+ The number of periods needed for complete adjustment of the marketing margin to an increase or a decrease in farm prices are m 1 and m 2 respectively+ Similarly, k, n1 , and n2 are the number of periods needed for complete adjustment of the marketing margin to variations in Q, and increases and decreases in marketing costs, respectively+ All prices are expressed in real terms, using the consumer price index as the deflator ~Wohlgenant & Mullen, 1987!+

3.2. Testing Asymmetry in Marketing Costs and Farm Price Transmissions For long-run symmetric marketing costs transmission, the cumulative value of increases has to equal the cumulative value of decreases in marketing costs+ Therefore, the following test is performed 3 : n1

~H1!

n2

( a 2A, i ⫹ i⫽0 ( a 2, Bi ⫽ 0 i⫽0

Similarly, for farm price transmission, the test is: m1

~H2!

m2

( a 4, i ⫹ i⫽0 ( a 5, i ⫽ 0 i⫽0

If H1 ~or H2! is rejected, there is evidence of long-run asymmetry+ It is also possible to test for complete price transmission ~i+e+, to test if the entire variation in the farm price is transmitted to consumers!+ In the fluid milk sector, it can be assumed that processing technology is of fixed proportion and, furthermore, that it is of fixed unitary proportion+ This implies that 1 gallon of milk from the farm will produce 1 gallon of fluid milk to be sold to the consumer+ With a fixed unitary proportion technology, complete price transmission requires that both cumulative values of the parameters associated with increases and decreases in farm prices are not statistically different from zero+4 The appropriate tests are given by expressions H3 and H4+ m1

~H3!

( a 4, i ⫽ 0 for farm price increases i⫽0 m2

~H4!

( a 5, i ⫽ 0 for farm price decreases

i⫽0

3 Note that the parameters of DECMC ~a 2B, i ! should exhibit a negative sign because the variable is in absolute value+ 4 Complete price transmission cannot be tested for marketing costs+ In this case, complete transmission would imply that the margin is increased just enough to cover the marginal increase in marketing costs+ Such analysis would therefore require information about the marginal productivity of all inputs, and this information is not available+ In fact, an index is generally used to reflect marketing costs in applied analysis and since no units are associated with an index, the tests equivalent to H3 and H4 cannot be conducted+


305

Tests H1 to H4 formally assess the conduct of the market regarding price transmission in the long run+ However, as these tests are necessary but not sufficient to fully assess asymmetry; one also has to assess short-run asymmetry+ Short-run asymmetry, in marketing costs transmission, requires that at least one of the equalities in H5 be rejected: s

~H5!

s

( a 2A, i ⫹ i⫽0 ( a 2B, i ⫽ 0 where S ⫽ 0,1, + + + , max@n1-1; n2-1# i⫽0

Similarly, the tests for short-run asymmetry in farm price transmission are given by H6: j

~H6!

j

( a 4, i ⫹ i⫽0 ( a 5, i ⫽ 0 where j ⫽ 0,1, + + + ,max@m 1-1; m 2-1# i⫽0

Note that a case of short-run asymmetry can occur without long-run asymmetry, and vice versa+ 4. DATA 4.1. Regions The model given by equation ~4! is estimated for Upstate New York ~UNY! and New York City ~NYC! fluid milk markets+5 NYC is analyzed separately from the rest of the state because two major regulatory changes occurred over the period of analysis, and they affected the two fluid milk markets differently+ One regulatory change affected only the NYC market, while the other, in all likelihood, affected more the NYC than the UNY market significantly+ First, the deregulation of the distribution of milk in NYC permitted Farmland Dairies ~a major New Jersey milk dealer! to penetrate the NYC market in January 1987+ Before 1987, the NYC milk market was protected from outside competition by a system of licenses that granted the right to distribute milk to a limited number of milk dealers+ The impact of that regulation is accounted for by adding a binary variable ~DD1987 ! to the NYC equation+ Second, a major change to New York State dairy policy was the introduction in June 1991 of the “200% law+” The price gouging law, as it became known, imposed a ceiling on the retail price of fluid milk: it could not be more than twice the farm price+ This law was enacted by the New York State Legislature following the perceived lack of response in retail milk prices to decreases in farm prices in NYC+ Even though mainly directed toward retailers in NYC ~Emerick, 1994!, the price gouging law’s effect on UNY is also tested in this study+ To perform this analysis, an intercept shifter has been included in the equations starting in 1991, as well as several slope shifters on the marketing costs and farm price variables+ 4.2. Data Description The models are estimated using monthly data for the 1980 to 1997 period, and all economic variables are expressed in real terms using the consumer price index for the Northeast U+S+ as deflator ~1982 to 1984 ⫽ 100!+ A description of the variables follows: 5

New York City designates the 11 counties of the New York City metropolitan region+ Upstate New York includes the rest of the counties in the state+

306


M:

TREND: Q: MC:

INCMCLi: DECMCLi: INCFLi: DECFLi: D1–D11: DD1987 : DD1991 :

Marketing margin, i+e+, the difference between the retail price and the farm price of fluid milk+ Farm prices for UNY and NYC are simply the Class I price of the New York-New Jersey Federal Milk Marketing Order ~FMMO! for 1 half-gallon of milk with 3+5% butterfat ~source: New York State Department of Agriculture and Markets @NYSDAM# !+ Retail prices in UNY and NYC are prices of 1 half-gallon of whole milk ~source: NYSDAM!+ Trend variable; this variable takes the values 1 to 215 ~215 monthly observations!+ Quantity of milk used by processors for the fluid market+ For both UNY and NYC, this variable reflects Class I milk in the New York-New Jersey FMMO+ The Li suffix indicates that the variable has been lagged by “i” month~s!+6 A marketing cost index+ Its construction is based on the estimated share of total marketing costs of the main intermediaries in the industry, i+e+, processors, retailers, and haulers+ Labor, energy, and packaging are the marketing costs that were taken into account ~BLS!+ The INC and DEC prefixes indicate that this variable has been split in rising and falling marketing cost phases+ The suffix “1991” added at the end of the variable denotes a slope shifter to test if the marginal effect of that variable has changed since the imposition of the price gouging law in 1991 Cumulative marketing costs increase as defined previously+ The Li suffix indicates a “ith” month lag+ Cumulative marketing costs decrease as defined previously+ Cumulative farm price increases as defined previously ~per cwt!+ Cumulative farm price decreases as defined previously ~per cwt!+ Monthly binary variables+ Di takes the value 1 in month i, and 0 otherwise+ Binary variable that reflects the deregulation of the NYC milk market as of February 1987+ It is equal to 0 prior to February 1987 and 1 thereafter+ Binary variable that reflects the imposition of the price gouging law in June 1991+ It is equal to 0 prior to June 1991 and 1 thereafter+

5. EMPIRICAL RESULTS Table 1 presents the estimated parameter values for the two regions, NYC and UNY+ Considering that margins are likely to be influenced by similar forces in both regions, both equations are estimated simultaneously using seemingly unrelated regression+7 Since autocorrelation was endemic in both regions, the equations are corrected for autocorrelation+ Because of the controversy surrounding the use of the R 2 when OLS is not used, the correlation coefficients ~CC! between the dependent variable and the estimated values of the regressions are reported+8 The CC range from 0+942 to 0+960, depending upon region+ This suggests that both models have a good explanatory power+

5.1. Lag Length The maximum lag length was first assumed to be 12 months for the quantity of milk variable ~Q!, and 8 months for both marketing costs increases ~INCMC! and decreases ~DECMC!, as well as for farm price increases ~INCF! and decreases ~DECF!+ Then, the Akaike Information Criterion ~AIC! was used to select the appropriate lag structure+ The lagged variables were withdrawn sequentially, starting with the less significant param6 Unfortunately, the quantity of fluid milk is not available for UNY and NYC alone+ However, using the same variable as a proxy is not a severe assumption because shortcomings or surpluses of milk in the region are likely to affect both regions in a similar way+ 7 The equations have been estimated using LIMDEP Version 7+0+ See Greene ~1998!+ 8 Judge, Griffiths, Hill, & Lee ~1980!+


TABLE 1+ York

Parameters of the Fluid Milk Margin Equations in New York City and Upstate New

Variables

INTERCEPT QL1 QL2 QL3 QL4 QL5 QL6 QL7 INCMC INCMCL1 INCMCL2 INCMCL3 INCMC1991 INCMCL11991 DECMC DECMCL1 DECMCL2 DECMCL3 DECMCL4 DECMC1991 DECMCL11991 INCF INCFL1 INCFL2 INCFL3 INCFL4 INCFL5 INCF1991 INCFL21991 INCFL41991 INCFL51991 DECF DECFL1 DECFL2 DECFL3 DECFL4 DECF1991 DECFL11991 DECFL21991 DECFL31991 DECFL41991 DD1987 DD1991 CC

307

New York City

Upstate New York

Parameter

p Value

Parameter

p value

6+53E⫺01 ⫺1+51E⫺04 1+12E⫺04 4+86E⫺06 1+36E⫺05 ⫺1+04E⫺04 1+06E⫺04 ⫺1+18E⫺04 2+63E⫺03 3+09E⫺03 ⫺7+67E⫺03 ⫺3+79E⫺03 ⫺3+54E⫺03 ⫺3+57E⫺03 ⫺5+05E⫺03 1+77E⫺03 — — — 2+96E⫺03 2+00E⫺03 1+28E⫺03 6+30E⫺03 7+71E⫺03 1+02E⫺03 ⫺6+40E⫺03 6+66E⫺03 — ⫺2+00E⫺03 ⫺4+53E⫺03 ⫺2+54E⫺03 3+34E⫺02 ⫺5+91E⫺03 ⫺3+15E⫺03 ⫺1+79E⫺03 ⫺5+41E⫺03 ⫺1+34E⫺02 — — ⫺1+90E⫺03 — ⫺7+34E⫺02 0+202278

0+000 0+006 0+013 0+924 0+755 0+060 0+034 0+041 0+190 0+132 0+000 0+030 0+218 0+119 0+002 0+265 — — — 0+153 0+196 0+807 0+341 0+270 0+887 0+374 0+256 — 0+066 0+000 0+019 0+000 0+095 0+372 0+603 0+070 0+001 — — 0+000 — 0+000 0+000

5+42E⫺01 ⫺2+52E⫺05 ⫺6+90E⫺05 ⫺5+70E⫺05 4+16E⫺05 9+12E⫺06 1+59E⫺05 7+97E⫺05 5+15E⫺03 ⫺3+28E⫺03 ⫺3+29E⫺03 — — — ⫺1+75E⫺04 ⫺3+40E⫺03 ⫺3+71E⫺03 ⫺3+04E⫺04 3+27E⫺03 ⫺1+14E⫺03 4+05E⫺03 ⫺5+46E⫺03 6+83E⫺03 ⫺6+49E⫺03 1+22E⫺02 3+26E⫺03 — 6+26E⫺03 — ⫺1+45E⫺02 — 1+79E⫺02 ⫺5+13E⫺03 1+47E⫺03 ⫺2+64E⫺03 ⫺6+69E⫺03 — ⫺1+04E⫺02 — — 4+04E⫺03 — —

0+000 0+614 0+272 0+375 0+430 0+878 0+801 0+160 0+016 0+118 0+122 — — — 0+929 0+101 0+049 0+862 0+053 0+196 0+011 0+382 0+356 0+400 0+125 0+665 — 0+176 — 0+006 — 0+000 0+232 0+699 0+475 0+047 — 0+008 — — 0+104 — —

0+960

0+942

Note+ The cumulative values of the lagged variables and their associated p values are reported in Table 3+ Coefficients of monthly dummy variables are not reported in order to ease the presentation of the results+ The binary variables that entered the final equations were D1 , D3 , D7 , D9 , and D11 in the NYC equation and D7 , D8 , D9 , D10 , and D11 in the UNY equation+ All these variables were negative+

308


eter; and when the AIC no longer decreased, all shorter lags were kept in the equations+ Note that for the slope shifters, all those that contributed to decrease in the AIC were withdrawn from the equations+ For example, only the slope shifters associated with DECF and DECFL3 ~i+e+, DECF1991 and DECFL31991 !, remained in the final equation in NYC+ The slope shifters associated with the other lagged decreasing farm price variables DECFL2 and DECFL4 did not contribute to decrease the AIC and were discarded+ Hence, the parameters associated with DECFL2 and DECFL4 are the same prior to and after 1991+ In order to decrease the level of multicollinearity, one could plea for the use of an Almon polynomial lag scheme, as was done in most previous studies+ However, the lag lengths were found to be relatively short for farm prices and marketing costs, while tests performed on the degree of the polynomial for the quantity variable rejected the Almon lag scheme for second-, third-, and fourth-degree polynomials+9 Therefore, any imposition of parameter weights would have been arbitrary, and no restrictions were imposed+ Results presented in Table 1 show that there are seven lags associated with Q, both in NYC and UNY+10 Hence, processors take several months to adjust their margins to variations in the quantity of milk+ With respect to marketing costs increases and decreases ~INCMC and DECMC!, the number of lags differ between them and between regions+ For NYC, three lags were necessary to reflect the dynamic pattern of INCMC, while only one was sufficient for DECMC+ In UNY, two lags were necessary for INCMC, but four lags were required for DECMC+ Farm price increases ~INCF! were lagged 5 months in the NYC equation, while for the UNY equation, four lags were sufficient+ With respect to farm price decreases ~DECF!, the AIC suggested a lag of 4 months for both NYC and UNY equations+ The lag lengths on farm prices are different than those found in previous studies+ Kinnucan & Forker ~1987! found that a 3-month lag was necessary for both farm price increases and decreases to adequately model the fluid milk market+ K&F reached this result by adding additional lags until a nonsignificant parameter was found+ Thus, the discrepancy can first be explained by the different methodology used by K&F to determine the lag length+ Also, their use of national data as compared with regional data ~as in this study! is likely another source of the dissimilarity+ Emerick ~1994! specified a 2-month lag for both farm price increases and decreases+ However, her decision was subjective, based on “a priori knowledge” of the U+S+ fluid milk market+

5.2. Regression Results The results presented in Table 1 generally conform to expectations+ The trend variable does not appear in the final estimated equations because its removal contributed to decreasing the AIC+ The quantity variable was lagged 7 months in both equations, which reflects a relatively long period of adjustment+ However, a test performed on the sum of the parameters shows that after complete adjustment, the quantity variable does not sig9 The degrees of the polynomial were tested by imposing the linear restrictions implied by the Almon polynomial lag scheme, and the null hypothesis was always rejected at the +05 level+ 10 The quantity variable for the current period is not included in the equations+ This assumes that processors are not able to react immediately to variations in the quantity demanded and that they set the margin according to an expected quantity, and the expectation is assumed to be naive in this case ~i+e, E~Qt! ⫽ Qt-1!+ Not including Qt in the equation also eliminates potential simultaneous equation bias+ Note that when included, Qt was not significant in both equations+


309

nificantly affect the margin in both regions ~Table 2, H0!+ Therefore, the hypothesis of constant returns to scale technology cannot be rejected+ The variable reflecting deregulation in the distribution of milk in NYC ~DD1987 ! is highly significant+ Farmland Dairies’ entry in the market has significantly reduced marketing margins by roughly $0+07 per half-gallon of whole milk ~constant dollar!+ Thus, the NYC milk market deregulation led to a lower farm-retail price spread+ This result supports the theoretical argument of Azzam ~1999! regarding the impact of increasing the retailer’s market radius on aggregate spatial demand+11 5.2.1. Long-run asymmetry. Prior to the introduction in June 1991 of the price gouging law, there was significant long-run asymmetric price transmission in marketing costs, in both NYC and UNY ~Table 2, H1!+ It is interesting to note that for both regions, the summations of the parameters associated with INCMC, DECMC, and their lags, are negative ~⫺5+74E-03 and ⫺3+28E-03 in NYC; ⫺1+43E-03 and ⫺4+31E-03 in UNY; see Table 3!+ This indicates that increases in marketing costs were not, at least completely, transmitted to consumers while decreases were, at least partly, transmitted+12 Finally, no significant difference in the long-run asymmetry between the two regions can be established during this period ~ p ⫽ +107!+ After 1991, long-run asymmetry in marketing costs is no longer significant in UNY+ Considering that the price gouging law was not aimed at reducing asymmetry in marketing costs transmission, this result was unexpected+ However, asymmetry in marketing costs remains significant in NYC+ The summation of the parameters associated with INCMC and its lags in NYC is still negative, statistically significant ~ p ⫽ +0001!, and is more important, in absolute terms, than the summation of the parameters associated with DECMC and its lags, which is positive but not significant ~ p ⫽ +458; ⫺1+29E-02 vs+ 1+67E-03, respectively!+ Also, the long-run asymmetry in NYC after 1991 is not statistically different from that before 1991 ~ p ⫽ +106!+ Therefore, the asymmetry in marketing costs transmission in NYC was not significantly affected by the price gouging law+ Prior to 1991, long-run asymmetry in farm price transmission is statistically significant at the +05 level in both NYC and UNY ~Table 2, H2!+ In fact, prior to the price gouging law, both increases and decreases in farm prices caused an increase in the farm-retail price spread; the summation of the parameters associated with INCF is 1+66E-02 and that associated with DECF is 1+72E-02 in the NYC equation and, respectively, 1+04E-02 and 4+93E-03 in the UNY equation ~Table 3!+ Both parameter summations are highly significant in NYC ~H3 and H4!, while the asymmetry in farm price transmission in UNY is due to an over-transmission of farm price increases because the summation of the parameters associated with farm price decreases is not significant+ These results suggest that an under-allocation of milk in Class I occurred before 1991 because of the permanent abovecompetitive-level selling price of milk+ After 1991 ~price gauging law!, in both regions, there is no statistical evidence supporting the hypothesis that farm price increases and decreases are not completely transmitted to retail price ~H3 and H4!+ The price gouging law was, therefore, beneficial to consumers+ However, in NYC, there is still statistical evidence of long-run asymmetry at 11

Actually, this result reflects more than an increase in competition, because the larger NYC milk dealers were convicted of price fixing following an investigation in the early 1980s ~NYSDAM, 1988!+ 12 Considering that marketing costs are measured by an index ~see footnote 4!, the portion of increases or decreases in marketing costs that is transmitted cannot be determined+

Constant returns to scale No long-run asymmetry in marketing cost transmission No long-run asymmetry in farm price transmission Farm price increases completely transmitted Farm price decreases completely transmitted No asymmetry after month 3 in marketing costs transmission No asymmetry after month 2 in marketing costs transmission No asymmetry after month 1 in marketing costs transmission No asymmetry after month 0 in marketing costs transmission No asymmetry after month 4 in farm price transmission No asymmetry after month 3 in farm price transmission No asymmetry after month 2 in farm price transmission No asymmetry after month 1 in farm price transmission No asymmetry after month 0 in farm price transmission

*The values in this table are the p values of the Chi-square statistics ~Wald Test!+

~H0! ~H1! ~H2! ~H3! ~H4! ~H5! ~H5! ~H5! ~H5! ~H6! ~H6! ~H6! ~H6! ~H6!

0+4689 0+0000 0+0000 0+0000 0+0000 — 0+0188 0+3195 0+3908 0+0000 0+0000 0+0000 0+0000 0+0000

0+4689 0+0002 0+0444 0+0672 0+6701 — 0+0253 0+9384 0+3726 0+4604 0+0610 0+0062 0+0054 0+0022

After the price gouging law ~ p value!

New York City Before the price gouging law ~ p value!

Marketing Costs and Farm Price Transmissions Test Results*

Null Hypothesis

TABLE 2+

0+9804 0+0003 0+0000 0+0092 0+2464 0+0001 0+0023 0+6583 0+1416 — 0+0269 0+3110 0+1064 0+1046

Before the price gouging law ~ p value!

0+9804 0+1452 0+8530 0+5112 0+7618 0+0193 0+0551 0+7604 0+2666 — 0+0964 0+5875 0+2301 0+0108

After the price gouging law ~ p value!

Upstate New York

310 ROMAIN, DOYON, AND FRIGON

⫺1+29E⫺02 ~0+000! 1+67E⫺03 ~0+458! 7+51E⫺03 ~0+0672! 1+84E⫺03 ~0+670! 0+52 0+43 ⫺1+12E⫺03 ~0+000!

0+522 0+244 0+301

After the price gouging law

⫺1+43E⫺03 ~0+473! ⫺4+31E⫺03 ~0+003! 1+04E⫺02 ~0+009! 4+93E⫺03 ~0+246! 0+62 0+49 ⫺3+43E⫺04 ~0+000!

0+488 0+166 0+182

Before the price gouging law

⫺1+43E⫺03 ~0+473! ⫺1+39E⫺03 ~0+454! 2+10E⫺03 ~0+511! ⫺1+40E⫺03 ~0+762! 0+52 0+51 ⫺6+66E⫺04 ~0+000!

0+430 0+244 0+301

After the price gouging law

Upstate New York

Note+ The values in parentheses are the p values+ a The mean increase and the mean decrease in the farm price are expressed on a per cwt basis+ b The price transmission elasticities are evaluated at the mean value of the F0R price ratios, which are, before and after the price gouging law, respectively, 0+51 and 0+45 in NYC, and 0+55 and 0+50 in UNY+

L0R increasing price elasticities b L0R decreasing price elasticities b Trend in marketing margin * $0half-gallon0month

DECF

INCF

DECMC

⫺5+74E⫺03 ~0+014! ⫺3+28E⫺03 ~0+040! 1+66E⫺02 ~0+000! 1+72E⫺02 ~0+000! 0+70 0+30 ⫺6+47E⫺05 ~0+513!

0+581 0+166 0+182

Before the price gouging law

New York City

Descriptive Statistics, Cumulative Parameter Values, Elasticities of Farm-Retail Price Transmission, and Trend in Marketing Margins

Mean marketing margin Mean increase in farm price a Mean decrease in farm price a Cumulative parameter values INCMC

TABLE 3+


311

312


the +05 level+ This is due to the additive positive effects on the marketing margin of both increases and decreases in farm prices+ The magnitude of the asymmetry has, however, been reduced significantly between the two periods ~3+38E-2 vs+ 0+93E-02!+ The information presented in Table 3 allows further analysis of the above results+ Note that the mean marketing margin is higher in NYC than in UNY in both subperiods, and it has decreased in both regions after 1991+ Consider the NYC market prior to 1991 and assume a $0+166 increase in the farm price of milk+ The estimated long-run impact on the marketing margin is an increase of $0+00275+ If we assume a $0+182 decrease in farm price, the long-run impact is $0+00312+ When compared with the mean margin of $0+581, these numbers seem modest+ However, since they are monthly averages, they reflect a continuous upward trend in processors’ margin+ Note that for that period, the estimated trend in the margin, reported in Table 3, is slightly negative, but not significant+ This is due to the previous result related to the transmission of marketing costs; increases are not completely transmitted to consumers, while decreases are partly transmitted+ However, a similar marginal analysis cannot be conducted for marketing costs, because an index is used+ After 1991 in the NYC market, the positive long-run impact of an increase in farm price has been divided by 2, while that of a decrease in farm price has been divided by almost 10, and neither are statistically significant+ This drastic change, accompanied by no significant change in the asymmetry of marketing costs transmission, as observed previously, strongly contributed to the significant negative trend in the marketing margin since 1991—a $0+001120half-gallon0month decline+ In the UNY market, before 1991, the same increase and decrease in farm price as those mentioned above would have contributed to increases of $0+00172 and $0+00090, respectively+ These increases are significantly less than those in NYC, which explains, in part, the significant downward trend in margin observed in this market ~i+e+, $0+000340halfgallon0month!+ After 1991, this trend doubled in magnitude because the positive impact on the margin of farm price increases has been divided by 5 and is no longer significant, and because farm price decreases are still statistically insignificant+ The degree of farm price transmission can also be seen through the elasticities of price transmission+13 Table 3 also reports the long-run price transmission elasticities, before and after the price gouging law for both NYC and UNY+ The elasticity of price transmission is given by ~ b ⫹ 1!~F0R!, where F is the farm price, R is the retail price and b is the appropriate estimated parameter, or sum of parameters for the long-run effect+14 Note that the extent to which the elasticity differs from the farm to retail price ratio indicates the degree of price transmission; the closer the elasticity to the price ratio, the more complete the transmission+ The elasticity results in Table 3 reflect the previous conclusions+ Prior to the price gouging law, long-run asymmetry in farm price transmission was established in both regions+ A 1% increase in the farm price transmitted to 0+7% and 0+62% increases in the retail price in NYC and UNY, respectively, while a 1% decrease in the farm price transmitted to 0+30% and 0+49% decreases in NYC and UNY+ Considering that the farm to retail price ratios were +51 and +55, respectively, long-run asymmetry is significant because the increasing and decreasing price elasticities are different from these values+ Af13

We thank an anonymous referee for suggesting this analysis+ When farm prices are decreasing, the negative values of the b parameters are used in the calculations because the variable reflecting decreases in farm prices is defined in absolute values+ 14


313

ter 1991, long-run asymmetry remained significant only in NYC, and this is reflected by different price transmission elasticities when farm prices are increasing versus when they are decreasing ~+52 vs+ +43!+ In UNY, the increasing farm price elasticity is the same as the decreasing farm price elasticity ~+52 vs +51!+ Moreover, these elasticity coefficients are close to the farm to retail price ratio ~+50!, which reflects that both farm price increases and decreases are fully transmitted ~H3 and H4!+ 5.2.2. Short-run asymmetry. When long-run asymmetric price transmission cannot be statistically sustained, it is still possible to have short-run asymmetry+ Recall that there is no statistical evidence of long-run asymmetry in marketing costs, nor in farm prices transmissions in UNY after 1991+ It is also possible that asymmetry shows up only in the long run, without significant impacts in the short run+ Further analysis has been conducted to assess short-run asymmetry+ Prior to the price gouging law, long-run asymmetry was established for both farm prices and marketing costs transmissions in both regions+ Results in Table 2 show that, in both regions, asymmetry in marketing costs transmission appears only after 2 months; no immediate short-run asymmetry is statistically significant ~H5!+ A similar pattern is observed in farm price transmission in UNY; asymmetry becomes significant only after 3 months+ Such delayed impacts could reflect the marketing strategy used by processors during this period+ Interestingly, short-run asymmetry in farm price transmission is immediate ~month 0! in NYC and remains significant thereafter+ In particular, the high parameter value associated with a decrease in farm price at time 0 suggests little transmission of a farm price decrease to the consumer+ This explains, in part, why the price gouging law was mainly directed toward retailers in NYC+ After 1991 in UNY, there is little evidence of short-run asymmetry in marketing costs transmission; it is significant only for 1 month ~after 3 months!+ Similarly, short-run asymmetry in farm price transmission lasts only 1 month—the first+ Both short-run asymmetries are not sufficiently strong to create significant long-run asymmetries+ In NYC, the pattern of short-run asymmetry in marketing costs transmission after 1991 is similar to the pattern that existed prior to 1991; it only becomes significant after 2 months and lasts in the long run+ With regards to farm price transmission, asymmetry is immediate and lasts for 2 months+ Even though it is no longer significant after 3 to 4 months, the short run is sufficiently strong to persist in the long run, as discussed previously+ 6. SUMMARY AND CONCLUSION A marketing margin model that allows testing for constant returns to scale technology and asymmetric marketing costs and farm price transmissions is proposed+ Consistent with the theoretical findings of Azzam ~1999! regarding increased local competition, results indicate that Farmland Dairies’ entry into the NYC fluid milk market has significantly reduced marketing margins, measured by the farm-retail price spread+ The results also shows that the price gouging law, enacted in June 1991 by the New York State Legislature, has had a statistically significant impact on farm price transmission in both Upstate New York ~UNY! and New York City ~NYC! markets+ However, the magnitude of the impact has been more important in NYC+ While farm price increases were not fully transmitted in both regions prior to 1991, this hypothesis could not be rejected after 1991+ There is no statistical evidence that farm price decreases were not fully transmitted in UNY before 1991, nor after 1991+ In NYC, farm price decreases were

314


not completely transmitted before 1991, but the price transmission mechanism became more effective thereafter+ Long-run asymmetry in marketing costs transmission was also identified in both regions prior to 1991+ While this asymmetry vanished after 1991 in UNY, no significant difference is noted between the two periods in NYC+ Long-run asymmetry in farm price transmission was significant in the first period in both regions+ The asymmetry disappeared after the implementation of the price gauging law in UNY, but remained statistically significant in NYC, though at a much lower level+ Short-run asymmetry was important in NYC before 1991 and remained significant in the second period+ Short-run asymmetry was less evident in UNY during the first period and almost disappeared during the second period+ The above results suggest that middlemen in the fluid milk market were exercising market power before the price gouging law+ However, to rigorously address the issue of a noncompetitive market, an alternative model would have to be developed+ Finally, a caveat of this study is the use of the Class I price as a proxy for the farm price because it does not account for over-order payments+ Not accounting for over-order payments could contribute to bias upward the actual level of asymmetry, but the data were not available for the period under study+ Nonetheless, discussions with New York State stakeholders lead us to believe that during the period of analysis, this omission would have had a limited impact+ ACKNOWLEDGMENTS The authors wish to thank one anonymous referee as well as the editor for helpful suggestions on an earlier version of this manuscript+ We would also like to thank the Food Marketing Policy Center of the University of Connecticut for the financial assistance provided to visiting graduate student Mathieu Frigon+ This project has been partially funded by a grant from the International Dairy Federation+ Data source: New York State Department of Agriculture and Markets ~NYSDAM!+ New York State Dairy Statistics Database System+ Bureau of Labor Statistics ~BLS!+ Internet web site ~http:00www+bls+gov0!+ REFERENCES Azzam, M+A+ ~1999!+ Asymmetry and rigidity in farm-retail price transmission+ American Journal of Agricultural Economics, 8, 525–533+ Emerick, P+A+ ~1994!+ An econometrical analysis of dairy market price transmission processes+ Unpublished master’s thesis, Cornell University+ Faminow, M+D+, & Laubscher, J+M+ ~1991!+ Empirical testing of alternative price spread models in the South African maize market+ Agricultural Economics, 6, 49– 66+ Greene, W+H+ ~1998!+ LIMDEP Version 7+0 user’s manual, revised edition+ Plainview, New York: Econometric Software Inc+ Griffith, G+R+, & Piggott, N+E+ ~1994!+ Asymmetry in beef, lamb and pork farm-retail price transmission in Australia+ Agricultural Economics, 10, 307–316+ Hansen, B+, Hahn, W+, & Weimar, M+ ~1994!+ Determinants of the farm-to-retail milk price spread+ ERS-USDA, Agriculture Information Bulletin, No+ 693+ Heien, D+M+ ~1980!+ Markup pricing in a dynamic model of the food industry+ American Journal of Agricultural Economics, 62, 10–18+ Houck, J+P+ ~1977!+ An approach to specifying and estimating nonreversible functions+ American Journal of Agricultural Economics, 59, 570–572+ Judge, J+G+, Griffiths, W+E+, Hill, R+C+, & Lee, T+-C+ ~1980!+ The theory and practice of econometrics+ New York: Wiley+


315

Kinnucan, H+W+, & Forker, O+D+ ~1987!+ Asymmetry in farm-retail price transmission for major dairy products+ American Journal of Agricultural Economics, 69, 285–292+ Lyon, C+C+, & Thompson, G+D+ ~1993!+ Temporal and spatial aggregation: Alternative marketing margin models+ American Journal of Agricultural Economics, 75, 523–536+ New York State Department of Agriculture and Markets @NYSDAM#+ ~1988!+ Review of dairy regulation+ State Milk Control in New York and Contiguous States, New York State Legislative Commission on Dairy Industry Development+ Novakovic, A+M+ ~1991!+ Price formation and the transmission of prices across levels of dairy markets ~Staff Paper No+ 91–8!+ Ithaca, NY: Department of Agricultural Economics, Cornell University+ Pick, D+H+, Karrenbrock, J+, & Carman, H+F+ ~1990!+ Price asymmetry and marketing margin behavior: An example for California-Arizona citrus+ Agribusiness, 6, 75–84+ Thompson, G+D+, & Lyon, C+C+ ~1989!+ Marketing order impacts on farm-to-retail price spreads: The suspension of prorates on California-Arizona navel oranges+ American Journal of Agricultural Economics, 73, 647– 660+ von Cramon-Taubadel, S+ ~1998!+ Estimating asymmetric price transmission with the error correction representation: An application to the German pork market+ European Review of Agricultural Economics, 25, 1–18+ Wohlgenant, M+K+, & Mullen, J+D+ ~1987!+ Modeling the farm-retail price spread for beef+ Western Journal of Agricultural Economics, 12, 119–125+ Wolffram, R+ ~1971!+ Positivistic measures of aggregate supply elasticities: Some new approaches— some critical notes+ American Journal of Agricultural Economics, 53, 356–359+ Zhang, P+, Fletcher, S+M+, & Carley, D+H+ ~1995!+ Peanut price transmission asymmetry in peanut butter+ Agribusiness, 11, 13–20+

Robert Romain is a professor in the Department of Economics of Agrifood and Consumer Sciences at Laval University, Quebec. He also serves as director of the Centre for Research in the Economics of Agrifood (CREA) at Laval University. Dr. Romain’s academic degrees include a bachelor in applied science (1977) from Laval, an M.Sc. in agricultural economics (1980) from the University of Manitoba, and a Ph.D. in agricultural economics (1983) from the Texas A&M University. Production economics and agricultural policy are among the subjects of his research. Maurice Doyon is assistant professor in the Department of Economics of Agrifood and Consumer Sciences and assistant director of the Groupe de Recherche en Économie et Politique Agricoles at Laval University, Quebec. He received a bachelor in applied science in 1991 from Laval University. He then studied at Cornell University in Ithaca, New York, where he earned master and doctoral degrees in agricultural economics. Dr. Doyon’s current research interests include policy analysis, international trade, and strategic marketing applied to the dairy, hog and forest sectors. Mathieu Frigon is currently assistant director of policy and economics with the Dairy Farmers of Canada. He holds a bachelor of science from Montreal University (1995), and an M.Sc. in agricultural economics from Laval University, Quebec City (1999). His research concentrates on dairy policy analysis.