Coffee Dfferentaton: Demand Analyss at Retal Level n the US Market Carmen Alamo and Jame Malaga Texas Tech Unversty Department of Agrculture and Appled Economcs Lubbock, Texas Phone: 806-742-1921 E-mals: carmen.alamo-gonzalez@ttu.edu, jame.malaga@ttu.edu, Selected Paper prepared for presentaton at the Southern Agrcultural Economcs Assocaton Annual Meetng, Brmngham, AL, February 4-7, 2012 Copyrght 2012 by Carmen Alamo and Jame Malaga. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.
Coffee Dfferentaton: Demand Analyss at Retal Level n the US Market Carmen Alamo and Jame Malaga Texas Tech Unversty carmen.alamo-gonzalez@ttu.edu, jame.malaga@ttu.edu Abstract Scanned data was used to estmate US coffee demand usng an AIDS model. The estmated elastctes have the expected sgns and magntude. Dfferentated coffees are complements for regular and unclassfed whle regular and unclassfed coffees are substtutes. These results could be useful n desgnng marketng strateges by coffee supplers. Introducton Coffee producton s regonally concentrated whle coffee demand s extended worldwde. In the largest consumng markets, the U.S., Germany, France, and Japan, whch together consume half of world exports, coffee s barely produced. Brazl, Vetnam, and Colomba are the world s largest producers and exporters. The caffenated agrcultural commodty s one of the most valuable prmary products traded n world markets. Sometmes t has been second only to ol exports as a source of foregn exchange for developng countres (ICO 2009). Its cultvaton, processng, tradng, transportaton, and marketng provde employment for mllons of people worldwde. Coffee s crucal to the economes and poltcs of many developng countres; for many of the world's Least Developed Countres, exports of coffee account for a substantal part of ther foregn exchange earnngs, n some cases over 80%. Coffee s a traded commodty on major futures and commodty exchanges, most mportantly n London and New York. The coffee prce crss assocated wth world supply
ncreases, and declnng per capta consumpton severely affected the economy of producer countres. New strateges are requred to boost coffee world prces. Producng countres are mplementng coffee dfferentaton strateges to ncrease ther profts and welfare. Coffee roasters and assocated companes n the consumng countres also are mplementng product dfferentaton strateges to enhance consumpton and profts. Dfferentaton n Agrcultural Products Producng countres and coffee roasters are pursung new marketng strateges that nvolve market segmentaton and product dfferentaton. When homogeneous commodtes are transformed nto dfferentated goods, unque, heterogeneous products are offered to consumers. The dfferentated goods are best descrbed as close but mperfect substtutes. They perform the same basc functons but have dfferences n attrbutes such as type, style, qualty, reputaton, appearance, and locaton that tend to dstngush them from each other. In the coffee case, the major dfferentaton process has been development n assocaton wth type, qualty, reputaton, and denomnaton of orgn. The transton from commodty to a dfferentated product changes the market structure from a perfect to a monopolstc competton model. Product dfferentaton leadng to dfferences n prces and market shares are explaned by theores of monopolstc competton (Chamberln, 1934) and love for varety (Dxt and Stegltz, 1977). These theores suggest that f a frm produces a product that s dstnct from others of the same type and f consumers are better off wth added varetes, market power takes place allowng the frm to set the prce that wll determne ts market share (Rakotoarsoa et al, 2003).
The transformaton of commodtes nto dfferentated products s also occurrng n the meat, dary, cacao and other agrcultural markets. The new trend on market structure organzaton due to heterogeneous supply and demand for dfferentated products brngs about questons that need to be studed: Do dfferentated products exhbt dfferent demand patters and elastctes than the undfferentated ones? In the coffee case: Do regular (undfferentated) coffee and dfferentated coffees behave as dfferent products? Do all dfferentated coffee groups behave smlarly? Do they exhbt dfferent demand patters and elastctes? These questons have been dffcult to answer for the US market because prevous coffee demand studes have focused on regular coffee undfferentated by country of orgn, flavor, blend, roast, and/or socal and envronmental causes. Recent studes estmated the demand for non-alcoholc beverages takng nto account the nterrelaton among several beverages and do not focus on coffee. Alvola, Capps Jr. and Wu (2010), Dharmasena and Capps Jr. (2009), and Pofahl, Capps, Jr., and Clauson (2005) have conducted demand system analyss of non-alcoholc beverages estmatng the regular coffee demand parameters usng panel data of household purchases. Other demand studes of nonalcoholc beverage ncluded coffee and tea combned (Zheng and Kaser (2008), and Yen et al (2004). Several demand studes related to coffee dfferentated products have focused on the estmaton of resdual demand elastctes, showng the role of the frms costs as a source of product dfferentaton (Rakotoarsoa et al. 2003), US demand for mld coffees from Colomba, Mexco,
and Brazl usng mports as consumpton (Houston, Santllan, and Marlowe, 2003) and demand of regular and soluble coffees (Huang et al. 1986). The objectve of ths study s to estmate the demand parameters at retal level of regular and dfferentated coffee for the US market. A unque data set was developed to classfy the dfferentated coffee n fve major groups. The dfferentated coffees were grouped nto the followng types: country of orgn, cause related (organcs, far trade, and ranforest), flavored, blends, and roasts types.. Methods and Procedures The data source used for the research s a weekly coffee sales scanned data at retal level n the US from 2001 to 2006. It was provded by Informaton Resources Inc. (IRI) Research and Developng Academc Data Set. The data source contans nformaton for US metropoltan areas n a tme perod of 313 weeks. Based on IRI data source, a unque data set was developed to classfy the dfferentated coffee n fve major groups. Informaton related to flavor, scent and descrpton of each coffee product lsted for the study perod was used for the groupng process. The dfferentated coffees were grouped nto the fallowng types: country of orgn, cause related (organcs, far trade and ran forest allance), flavored, blends, and roasts types. The regular coffees were grouped fallowng the flavor/scent descrpton of the IRI data set. Coffees whch do not belong to the regular and dfferentated categores were grouped on an unclassfed category. The quantty data of the retal
sales was standardzed n terms of pounds, the prces n terms of dollars per pound, and the expendtures n dollars. To nclude the effect of seasonalty n the studed perod the data source was decomposed n wnter, sprng, summer, and fall seasons. The 313 weeks were classfed accordngly wth the season n place at that partcular tme of the year. Informaton from the Weather Busness Bureau was accessed to determne the specfc startng and endng date for each season. Model The almost deal demand system (Full AIDS) model was selected to estmate the demand parameters at retal level of regular, dfferentated, and unclassfed coffee for the USA market. The AIDS model was developed by Deaton and Muelbauer (1980) and t was selected due to t several desrable propertes. Frst, t s a flexble functonal form but t has the added advantage of beng compatble wth aggregaton over consumers, thus can be nterpreted n terms of economc models of consumer behavor when estmated wth aggregated (macroeconomcs) or dsaggregated (household survey) data. Second, the AIDS model provdes an arbtrary frst-order approxmaton to any demand system. Thrd, t s derved from a specfc cost functon and therefore corresponds wth a well-defned preference structure. Fourth, homogenety and symmetry restrctons depend only on the estmated parameters and are easly tested and/or mposed. Ffth, t aggregates perfectly across consumers wthout nvokng parallel lnear Engle s curves. Fnally, t satsfes the axoms of choce exactly.
The seasonalty adjustment was ntroduced to the Full AIDS model as a dummy varable that represents wnter, sprng, summer, and fall seasons. The model s used to evaluate the mpact of cool, mld, and warm season on coffee consumpton. It s assumed that sprng and fall are mld seasons. The specfcaton of the Full AIDS model can be descrbed as: w t 3 * m j log ptj log d jsjt et, (1) j P j1 where P s prce ndex defned by 1 * log P 0 j ln p j j ln p p j (2) 2 j j and the parameters are defned by j 1 * ( j * j ) j 2 (3) where = (1, 2, and 3) ndexes of regular (1), dfferentated (2) and unclassfed (3) coffees n the system, t ndexes the tme n weeks (there are 313 weeks), p jt s weekly average nomnal prce for each coffee consderate n the study, m s the total expendtures calculatng usng nomnal prces p jt, and the total quantty of each coffee consumed per week s q t. The S jt s the seasonal dummy use to capture seasonalty of the four seasons of the year. The dsturbance term s represented by e t. The weekly budget share of each coffee consumed s: w t ptq m t (4)
The model was estmated usng SAS 9.2 statstcal software. To estmate the Full AIDS Model the fallowng theoretcal restrctons were mposed on the parameters: (1) Addng-up Restrcton requres, for all, 1, 0, k 0 k (5) (2) Homogenety s satsfed f and only f, for all, k k * 0 (6) (3) Symmetry s satsfed provded that k, k k (7) The Zellner s nteractve seemngly unrelated regresson (ITSUR) procedure was used to estmate the Full AIDS Model. The sgnfcance level used was a 10% (p-value 0.10). The mposton of the addng up restrctons and the fact that the expendture shares add to one resulted n the droppng of one budget share equaton (unclassfed coffees) to avod the sngularty error of the varance-covarance matrx. The parameters for unclassfed coffee share equaton were recovery usng the addng up restrcton. The own prce, cross prce, and expendture elastctes for the Full AIDS model were estmated employng the fallowng equatons:
The expendture elastctes for Full AIDS s as follows q m 1, (8) m q w The uncompensated own-prce elastctes for Full AIDS, e q p p q e j w j w w 9 k1 ln p k 1, (9) The uncompensated cross-prce elastctes for Full AIDS, e j q p j p q j e j j w j w w 9 k1 ln p, k j (10) The compensated own and cross prce elastctes for Full AIDS, e * e w (11) j j Emprcal Results and Dscusson Ths secton presents and dscusses the descrptve statstcs of the varables, the estmated AIDS model parameters, and the expendture and prce statstcs for regular, dfferentated, and unclassfed coffees. The summary of descrptve statstcs n table 1 shows that the unclassfed group presents the most expensve coffee ($5.23/lb.) followng by the dfferentated ($4.11/lb.) and regular ($3.10/lb.) groups. Unclassfed coffee had also the wdest range of prce wth a standard devaton of $0.76 per pound. The hgher prce of the unclassfed group mght be assocated to the fact that coffees wth one cup brewng system (pod) are ncluded n ths category. Usually the one cup brewng coffee s sold n a package that contans small quanttes
wth hgh prce per unt. Further studes are needed to evaluate n more deeply ths type of coffee that was ntroduced to the market durng the studed perod. Accordng to table 1, the dfferentated coffee was the most consumed durng the studed perod fallowng by the regular and unclassfed groups. The mean total quantty consumed of the dfferentated group was 60.80% and 98.62% hgher than regular and unclassfed, respectvely. The dfferentated coffee had the hghest budget share. Parameters estmated by the AIDS model are reported n table 2. Out of sx own and cross-prce coeffcents estmated fve were statstcal sgnfcant at 10% level. One of two ntercepts coeffcent (alphas) was statstcally sgnfcant at 10% level. The expendture coeffcents (betas) and the coeffcent assocated wth the dummy varable seasonalty (d s) were not statstcal sgnfcant at 10% level. The expendture, uncompensated, and compensated own and cross-prce elastctes were calculated and presented n table 3 and 4. Calculated expendture elastctes reveal that all coffee groups were normal goods whle unclassfed coffee s consdered a luxury good (table 3). The unclassfed good s the most elastc group followed by the regular and dfferentated groups. Uncompensated and compensated own and cross-prce elastctes are presented n table 4. All uncompensated and compensated prce elastctes have the theoretcal expected negatve sgn. The uncompensated own-prce elastcty for the unclassfed coffee s -1.850. Own-prce elastctes for dfferentated and regular coffees were -0.946 and -0.901 respectvely. The
uncompensated own-prce elastctes were lower than the ones estmated by Dharmacena and Capps Jr. (2009) for coffee undfferentated (-0.517) and Zheng and Kaser (2008) for coffee undfferentated and tea (-0.462). Compensate cross-prce elastctes reveal a complementary nteracton between regular and dfferentated coffees and a substtuton relatonshp between regular and unclassfed coffees (table 4). The dfferentated coffee has a complementary nteracton wth both the regular and unclassfed. Unclassfed coffee s a substtute for regular coffee and a complement for dfferentated coffee. The results revealed that unclassfed coffee has the hghest prce senstvty compared wth regular and dfferentated coffees, mplyng that when ts prce ncreases consumers substtute t for regular coffee but may contnue buyng coffee from the dfferentated group. The dfferentated group ncludes coffees by orgn of denomnaton, cause related (organc, ran forest, and far trade) and dfferent flavors, roasts and blends. The varety n the dfferentated group appeals preferences of a wde group of consumers, whch may explan ts complementary nteracton wth the regular and the unclassfed coffees. We consder the performed demand analyss for US dfferentated coffee at retal level as an ntal exploratory study, and so, future analyses are recommended. Based on the study results, the followng recommendatons are made for future analyss: 1) to measure more accurately the demand of the dfferentated group by performng a separate demand analyss for each one of the coffees that compose the group (dfferent country orgns, cause related, flavors, blends, and
roast), 2) to revse the unclassfed group and estmate separately the demand for the coffee used on the one cup brewng system, and 3) to analyze and compare the demand of regular, dfferentated, and unclassfed coffees for dfferent regons n the US. Summary and Conclusons The almost deal demand system (AIDS) model was selected to estmate the demand parameters at retal level of regular and dfferentated coffee for the US market. The data source used n the research was a weekly (313) coffee sales scanned data at retal level n the USA from 2001 to 2006. It was provded by Informaton Resources Inc. (IRI). Based on IRI data source a unque data set was developed to classfy the coffee n three major groups: regular, dfferentated, and unclassfed. The dfferentated coffees were grouped nto the followng types: country of orgn, cause related (organcs, far trade and ranforest), flavored, blends, and roasted types. The regular coffees were group followng the flavor/scent descrpton of the IRI data set. Coffees whch do not belong to the regular and/or dfferentated groups were grouped on unclassfed category. The results revealed that the unclassfed group was the most expensve coffee ($5.22/lb.) followng by the dfferentated ($4.14/lb.) and regular ($3.09/lb.) groups. The hgher prce of the unclassfed group mght be assocated to the fact that coffees wth one cup brewng system (pod) are ncluded n ths group. Calculated expendture elastctes showed that all coffee groups were normal goods, whle the unclassfed coffee should be consdered a luxury good.
All uncompensated and compensated own-prce elastctes have the theoretcally expected negatve sgn. The uncompensated own-prce elastctes were found as follows: unclassfed; (- 1.850), dfferentated; (-0.946), and regular; (-0.901). Compensated cross-prce elastctes reveal a complementary nteracton between regular and dfferentated coffees and a substtute relaton between regular and unclassfed coffees. The dfferentated coffee has a complementary nteracton wth both regular and unclassfed coffee. Unclassfed coffee s a substtute for the regular one. The performed study assumed an aggregated consumpton at retal level. Consderng ths, the results suggest that household s members mght have dfferent preferences and, thus mght purchase coffees that belong to the dfferent groups. The varety n the dfferentated group appeals preferences of a wde group of consumers, whch may explan ts complementary nteracton wth the regular and the unclassfed coffees. Future Work: We consder the performed demand analyss for US dfferentated coffee at retal level as an ntal exploratory study, future analyses are recommended. Based on the study results the followng recommendatons are made for future analyss: 1) to measure more accurately the demand of the dfferentated group by performng a separate demand analyss for each one of the coffees that compose the group (dfferent country orgns, cause related, flavors, blends, and roast), 2) to revse the unclassfed groups and estmated separately the demand for the coffee use for the one cup brewng system, and 3) to analyze and compare the demand of regular, dfferentated, and unclassfed coffees for dfferent regons n the US.
References Alvola, Pedro A., Capps, Oral Jr. & Xmng Wu. Mcro-Demand Systems Analyss of Non- Alcoholc Beverages n the Unted States: An Applcaton of Econometrc Technques Dealng Wth Censorng. Paper prepared for presentaton at the Conference of Agrcultural & Appled Economcs Assocaton, Jont Annual Meetng, Denver, Colorado, July 25-27, 2010. Chamberln, E.H. The Theory of Monopolstc Competton. Cambrdge: Harvard Unversty Press, 1934. Deaton, A. S. and J. Muellbauer. An Almost Ideal Demand System. Amercan Economc Revew 70(1980):312-326. Dharmasena, Senarath and Oral Capps Jr. Demand Interrelatonshps of At-Home Nonalcoholc Beverage Consumpton n the Unted States. Agrcultural and Appled Economcs Assocaton s 2009 AAEA & ACCI Jont Annual Meetng, Mlwaukee, Wsconsn, July 26-29, 2009. Dxt, A. K. & J. E. Stgltz. Monopolstc Competton and Optmum Product Dversty. Amercan Economc Revew Vol. 3: 297-308, 1977.
Houston, Jack E., Santllan, Manlo, and Jula Marlowe. US Demand for Mld Coffees: Implcatons for Mexcan Coffee. Journal of Food Dstrbuton Research 34 (1): 92-98, 2003. Huang, C.J., J.J., Segfred and F. Zardoshty. The Demand of Coffee n the Unted States: 1963-77. Quart. Rev. Econ. and Bus., 20: 36-50, 1986. Internatonal Coffee Organzaton (ICO), Hstorcal Statstc Data. Internet ste: http://www.co.org/new_hstorcal.asp (Accessed at October 1, 2009) Pofahl, Geoffrey M., Capps, Jr., and Annette Clauson. Demand for Non-Alcoholc Beverages: Evdence from the ACNelsen Home Scan Panel. Amercan Agrcultural Economcs Assocaton Annual Meetng, Provdence, Rhode Island, July 24-27, 2005. Rakotoarsoa, Mantra A., Henneberry, Shda, Shapour, Shanhla & Trueblood, Mchael A. The Export for Dfferentated Processed Agrcultural Products; The Role of Factor Prces and Fxed Cost. Paper Presented at Amercan Assocaton Annual Meetng, Montreal, Canada, July 27, 2003. Yen, S.T., B Ln, D.M. Smallwood, and M. Andrews. Demand for Nonalcoholc Beverages: The Case of Low Income Households. Agrbusness, 20(3): 309-321, 2004. Zheng Y. and H. M. Kaser. Advertsng and USA Nonalcoholc Beverage Demand, Agrcultural and Resource Economcs Revew, 37(2): 147-159, 2008.
Table 1. Descrptve Statstcs for Each Coffee Group n the US by (n=313 weeks) Mean Std. Devaton Mnmum Maxmum Prce ($/lb) Regular 3.10 0.30 2.30 3.84 Dfferentated 4.11 0.51 3.21 5.30 Unclassfed 5.29 0.76 3.05 7.99 Total Weekly Quantty (lb) Regular 208630.50 41391.65 143362.04 372003.52 Dfferentated 532045.48 83893.86 398715.14 958284.62 Unclassfed 7298.65 2452.10 4340.81 25392.34 Total Expendture ($) Regular 638014.52 89198.68 482348.91 1014794.88 Dfferentated 2169402.98 336434.45 1570905.17 3728647.00 Unclassfed 37258.40 7029.16 27367.94 88203.63 Budget Shares Regular 0.230 0.026 0.160 0.290 Dfferentated 0.760 0.030 0.700 0.820 Unclassfed 0.010 0.003 0.0070 0.0310
Table 2. Parameter Estmates of AIDS Model for US Coffee 1 Parameter Estmate Std Error t Value Pr> t y 11 0.02249 0.0189 1.19 0.2338 y 12-0.03647 0.0196-1.86 0.0637 y 13 0.013976 0.0061 8.70 <.0001 y 21-0.03647 0.0196-1.86 0.0637 y 22 0.039155 0.0205 1.91 0.0573 y 23-0.00269 0.0022-1.22 0.2222 α 1 0.20885 0.1807 1.16 0.2488 α 2 0.790255 0.1877 4.21 <.0001 b 1 0.00112 0.0132 0.08 0.9325 b 2-0.00257 0.0137-0.19 0.8517 d 11 0.004526 0.00446 1.02 0.3104 d 12 0.006812 0.00472 1.44 0.1500 d 13 0.006488 0.00498 1.30 0.1997 d 21-0.00422 0.00462-0.91 0.361 d 22-0.00629 0.00489-1.29 0.1997 d 23-0.0056 0.00516-1.09 0.2762 1 Coffee group: Regular (1), Dfferentate (2), and Unclassfed (3). Note: all estmated coeffcents n bold are sgnfcant at 10%
Table 3. Expendture Elastctes for Coffee* Usng Full-AIDS System Item Elastcty Group Regular e1 1.005 Dfferentated e2 0.997 Unclassfed e3 1.109 *Coffee groups: Regular (1), Dfferentated (2), and Unclassfed (3).
Table 4. Uncompensated and Compensated Elastctes for Coffees* Usng Full-AIDS System Item Elastcty Uncompensated e11-0.901 e12-0.165 e13 0.062 e21-0.047 e22-0.946 e23-0.003 e31 1.026 e32-0.285 e33-1.850 Compensated e11-0.675 e12 0.599 e23 0.075 e21 0.177 e22-0.187 e23 0.009 e31 1.276 e32 0.559 e33-1.83 *Coffee groups: Regular (1), Dfferentated (2), Unclassfed (3).