AN ECONOMIC EVALUATION OF THE HASS AVOCADO PROMOTION ORDER S FIRST FIVE YEARS A REPORT PREPARED FOR THE HASS AVOCADO BOARD BY Hoy F. Carman Lan Li Richard J. Sexon 1 March 30, 2009 1 Hoy F. Carman is Professor Emerius and Richard J. Sexon is Professor, Deparmen of Agriculural and Resource Economics, Universiy of California, Davis. Lan Li is Research Associae, Naional Insiue for Commodiy Promoion Research, Cornell Universiy.
Table of Conens 1. Inroducion...2 2. Avocado Promoion Programs...3 2.1. California Avocado Commission Programs...5 2.2. CAIA and MHAIA Programs...6 2.3. Hass Avocado Board Programs...8 3. Avocado Consumpion in he U.S....9 4. Modeling he Annual Demand for Avocados...11 4.1. Previous Sudies...12 4.2. Economeric Models of he Annual Demand for Avocados...13 4.3. Preliminary Daa Analysis...16 4.4. Srucural Breaks in Per Capia Consumpion...20 4.5. Esimaed Annual Demand Relaionships...20 4.6. Diagnosic Checks of Annual Demand Models...25 4.7. Two-Sage Leas Squares Esimaion...27 4.8. Summary...29 5. Benefi-Cos Analysis...31 5.1. Benefi-Cos Analysis in Promoion-Evaluaion Sudies...31 6. Demand Analysis a he Reail Level...39 6.1. The Daa...39 6.2. The Economeric Models...42 6.3. Resuls...45 6.4. The Effecs of he CAC s Promoions on Reail and Shipping-Poin Prices...55 7. Evaluaion of he HAB s Nework Markeing Cener Program...58 7.1. Variabiliy of Prices and Quaniies Over Time...58 7.2. Coss of HAB Informaion Programs...61 7.3. Esimaed Benefis from Informaion Programs...62 8. Daa Collecion Suggesions o Faciliae Fuure Evaluaions...65 9. Conclusions...67 9.1. Meeing he Impor Challenge...67 9.2. The Demand for Avocados...68 9.3. Economic Evaluaion of HAB Programs...70 9.4. Boom Line...72 References...73
Lis of Figures Figure 1: U.S. Per Capia Avocado Consumpion by Source, 1980...10 Figure 2: Per Capia Avocado Consumpion, U.S. Per Capia Disposable Income, and he Percenage of Hispanic Populaion...17 Figure 3: Per Capia Avocado Consumpion, he F.O.B. Price, and Promoion Expendiures...18 Figure 4: The U.S. Avocado Supply, Impors, and Domesic Producion...30 Figure 5: Avocado Promoion Simulaion Model...33 Lis of Tables Table 1: U.S. Avocado Promoional Expendiures in Dollars by Organizaion, 2003-2007....4 Table 2: Variable Definiions and Summary Saisics...15 Table 3: Correlaion Coefficiens for he Demand Model...19 Table 4: Esimaed Annual Demand Models: Ordinary Leas Squares...22 Table 5: The Firs-Sage Regression o Predic California f.o.b. Price...27 Table 6: Esimaed Annual Demand Models: Two-Sage Leas Squares...28 Table 7: Simulaion Model Resuls...37 Table 8: Summary Saisics for he Disaggregae Model...41 Table 9: Esimaion Resuls for he Reail Sales Model Wihin Model...47 Table 10: The Effecs of Promoions on Reail Sales from Panel I: 2003-2004...50 Table 11: The Effecs of Promoions on Reail Sales from Panel II: 2007-08...53 Table 12: The Effec of he CAC s Promoions on Reail Price and Shipping Price...57 Table 13: Sandard Deviaion of Weekly California and Toal Avocado Shipmens,...59 Table 14: Annual and Toal Coss of HAB Informaion Programs by Cos Caegory,...61 Table 15: Esimaed Toal Annual Changes in Gross Margins for Hass Avocados, Average Shipmens, Sandard Deviaion of Price, and Average Price, 2003-2007...63 2
AN ECONOMIC EVALUATION OF THE HASS AVOCADO PROMOTION ORDER S FIRST FIVE YEARS Hoy F. Carman Lan Li Richard J. Sexon 2 The U.S. Deparmen of Agriculure (USDA) requires periodic evaluaion of he effeciveness of naional promoion programs under is supervision. The Hass Avocado Board (HAB) has been direced by he Agriculural Markeing Service, USDA, o conduc an evaluaion of programs underaken during i is iniial five years of operaion, from markeing years 2002-03 hrough 2006-07. This repor provides he resuls of he USDA-mandaed evaluaion of HAB promoion programs. The evaluaion analyzes he impacs of hese expendiures and he overall reurns accruing o Hass avocado producers from all promoion programs. For some of he saisical mehods employed in his evaluaion, a five-year period provides insufficien daa. In hese siuaions, we evaluae he enire hisory of avocado promoion from he beginning of organized avocado promoion in California in 1962 o he presen. Following a brief inroducion, in secion 2 he repor discusses he major markeing programs conduced under he auspices of he HAB. Secion 3 provides an overview of rends in U.S. consumpion of avocados, while secion 4 conains a deailed analysis of annual demand for avocados in he U.S., wih he goal of deermining he impac ha promoions have had on avocado demand. Secion 5 inroduces and implemens a simulaion framework o esimae he impac of avocado promoions on grower price and income and on consumpion of avocados, based upon he resuls of he demand analysis. Secion 6 provides an analysis of avocado demand and he impac of promoions based upon reail scanner daa. Secion 7 presens an exploraory 2 Hoy F. Carman is Professor Emerius and Richard J. Sexon is Professor, Deparmen of Agriculural and Resource Economics, Universiy of California, Davis. Lan Li is Research Associae, Naional Insiue for Commodiy Promoion Research, Cornell Universiy. 1
analysis of he impacs of he HAB s informaion sharing and disseminaion programs. Secion 8 conains recommendaions for fuure daa-collecion programs ha he HAB should consider o faciliae fuure program evaluaions. Finally, secion 9 presens brief concluding remarks. 1. Inroducion The Unied Saes avocado indusry has evolved from an emphasis on seasonal domesic producion of a mix of avocado varieies o year-round availabiliy of domesic and impored Hass avocados. California avocado producers, who accoun for approximaely 90 percen of U.S. avocado producion and essenially all U.S. Hass avocado producion, have funded promoional programs for avocados since 1961. Wih few impors of avocados prior o he early 1990s, he benefis from hese demand-enhancing programs flowed direcly o California producers. Impors of avocados ino he U.S. have increased seadily since hen, resuling in a free-rider problem ha led ulimaely o creaion of he Hass Avocado Promoion, Research, and Informaion Ac of 2000 ha was signed ino law by Presiden Clinon on Ocober 23, 2000. This Ac esablished he auhorizing plaform and imeable for he creaion of he Hass Avocado Promoion, Research and Informaion Order (HAPRIO) ha was approved in a referendum of producers and imporers wih 86.6 percen suppor on July 29, 2002. The HAPRIO became effecive on Sepember 9, 2002, wih program assessmens becoming effecive on January 2, 2003. The 12-member Board ha adminisers he program under USDA supervision consiss of 7 domesic producers and 5 imporers. Appoinmen of he firs Hass Avocado Board (HAB) members on February 12, 2003 iniiaed aciviies under he HAPRIO. The mandaory assessmen rae is 2.5 cens per pound for all Hass avocados sold in he U.S., and he maximum permied assessmen is 5.0 cens per pound. The HAB is required o 2
rebae 85 percen of domesic assessmens o he California Avocado Commission (CAC) and up o 85 percen of imporer assessmens o imporer associaions, which use he funds for heir own promoion programs. The HAB uses he remaining 15 percen of assessmens for is operaions, promoion, and informaion echnology programs. During is firs five years of operaion, he HAB colleced assessmens oaling $98.67 million and rebaed $77.6 million o counry producer organizaions, including $38.64 million o he CAC, $20.54 million o he Chilean Avocado Imporers Associaion (CAIA), and $18.42 million o he Mexican Hass Avocado Imporers Associaion (MHAIA). Toal five-year promoional expendiures were as follows: CAC, $50.98 million; CAIA, $16.71 million; MHAIA, $14.35 million; and HAB, $9.27 million, for an overall oal of $91.3 million spen on Hass avocado promoion in he U.S. marke. 2. Avocado Promoion Programs Even hough he HAPRIO is only five years old, promoion of avocados in he U.S. marke by California producers provides significan program experience o build upon. Using a sae markeing order, he California avocado indusry conduced generic adverising and promoion programs from 1962 hrough 1977 and has operaed under provisions of he CAC since Sepember 1978. Review of annual repors of he markeing order and commission programs indicaes ha he California indusry spen over $334 million (in 2007 dollars) on adverising, promoion, and relaed services from iniiaion of he program in 1962 hrough 2002. 3 Toal promoion expendiures by he California indusry, including he 2003 hrough 2007 HAPRIO 3 The U.S. avocado markeing year runs from November 1 hrough Ocober 31 of he following year. We use he convenion of referring o he markeing year as he laer year, e.g., we refer o November 1, 2002 hrough Ocober 31, 2003 as 2003. 3
allocaions, were almos $398 million (in 2007 dollars). Addiional promoional expendiures in he U.S. marke by he HAB, MHAIA, and CAIA from iniiaion of HAPRIO assessmens in 2003 hrough he end of he 2006-07 markeing year oaled almos 42 million (in 2007 dollars). Hass avocado promoion programs ake many forms. The CAC allocaes he majoriy of is funds for consumer adverising and merchandising/rade promoions. 4 Significan expendiures are also made on foodservice, public relaions, nuriion and inerne markeing programs. The CAIA conraced wih CAC o conduc is markeing/promoion programs for he firs four years from incepion hrough he 2006 markeing year. In 2007 he CAIA began conducing is own programs. The MHAIA did no conduc any promoion programs in 2003 bu has been acive since 2004. HAB expendiures have emphasized naional marke communicaions and indusry informaion programs. 5 The annual promoional expendiures by counry organizaion are shown in able 1. Noe ha he California daa are for markeing programs only (indusry programs and adminisraion are excluded), and HAB daa are for markeing communicaions only (informaion programs and adminisraion are excluded). All organizaional expendiures are repored for Chile and Mexico. Table 1: U.S. Avocado Promoional Expendiures in Dollars by Organizaion, 2003-2007. Year CAC CAIA MHAIA HAB Toal 2003 8,682,060 1,427,000 0 146,499 10,255,559 2004 10,756,130 3,010,060 700,000 859,284 15,325,474 2005 11,838,029 5,742,600 2,900,000 2,603,124 23,083,753 2006 10,498,717 2,660,763 4,500,000 2,562,140 20,221,620 2007 9,205,138 3,864,637 6,246,500 3,096,859 22,413,134 Toal 50,980,074 16,705,060 14,346,500 9,267,906 91,299,540 * Noe: The avocado markeing year runs from he prior November 1 o Ocober 31 of he year lised. 4 The CAC also collecs addiional funds from he producion of California Hass as well as oher ypes of avocados o suppor is indusry programs and oher aciviies. 5 Informaion in his secion is based upon each of he organizaions annual business or markeing plans and budges, as available. 4
2.1. California Avocado Commission Programs The CAC ypically allocaes abou 70 percen of is revenues from producer assessmens o is markeing programs. Consumer adverising is he leading aciviy in erms of expendiures, wih programs conduced in major markes across he U.S. Consumer adverising messages and iming are ailored o individual markes. The demographics for CAC radiional consumer arges include: women and aduls 25 o 54 years old, aended or graduaed from college, household income $50,000 plus, employed, and Hispanic. In erms of psychographics, he arge audience is advenurous and open-minded; ineresed in home, syle and food; healh conscious; and someone who enjoys a variey of cuisines. CAC geographic emphasis coninues o be on core markes in Wesern saes and developing markes in oher regions. 6 Consumer adverising is focused during he period from February hrough Augus coinciding wih he California harves. Radio has been he main medium for consumer adverising. Billboards, newspapers and cable elevision are also used, depending on he marke and message. Naional adverising programs have used cable TV (Discovery Nework, Food Nework and Fine Living), prin (Food and Wine, Saveur) and he inerne. Merchandising and rade promoion programs ake a variey of forms, including poin-ofpurchase maerials, display coness, produce programs (AvoInfo/RipeMax), co-markeing, rade adverising, reail ie-in evens, and cooperaive adverising. While he main arge is supermarke chains and mass reailers, smaller specialy reailers and wholesalers are no overlooked. The CAC s foodservice markeing effors focus on resauran operaors (including upscale independens and chains, as well as fas-food and casual-dining esablishmens), 6 The CAC s core markes in 2007 included Los Angeles, San Francisco, San Diego, Sacrameno, Phoenix, Seale, Porland, Dallas, San Anonio, and Houson. Is single selec developing marke was Alana. 5
insiuions such as universiies, and foodservice influencers including ediors, food wriers, and rend-seing chefs. The CAC s healh and nuriion research program is designed o work in synergy wih is public relaions effors o esablish and communicae he healh and nuriional benefis of consuming avocados. Research focused iniially on a deailed analysis of he composiion and nurien conen of avocados, including fay acids, viamins, and minerals. More recenly, emphasis has shifed o quanifying and qualifying various phyochemicals (i.e. pyoserols, caroenoids, gluahione), as well as heir healh benefis and effecs on disease processes. The CAC made a sraegic decision o focus on he use of public relaions o disseminae he healh and nuriional message for avocados raher han using paid adverising and promoion. Consumers place much more crediabiliy on a news sory abou healh and nuriion benefis of consuming a produc han hey do on adverising wih he same message. The CAC s healh and nuriion messages are also feaured on is websie (hp://www.avocado.org/healh-nuriion/). 2.2. CAIA and MHAIA Programs The CAIA and MHAIA have a shor hisory relaive o he CAC, and informaion on heir programs is limied. The MHAIA conduced adverising/promoion programs for four of he five years in he evaluaion period while he CAIA s firs independenly operaed program was conduced in 2007. Beginning wih $700,000 of expendiures in 2004, he MHAIA expanded is markeing budge and aciviies each year hrough 2007. The MHAIA spen abou wo-hirds of is 2004 markeing funds on consumer adverising, wih 55 percen of oal expendiures for radio adverising. Trade adverising, including co-op markeing funds and websie expendiures accouned for anoher 26 percen of oal funds. The 2005 MHAIA budge increased o $2.9 6
million wih $1.5 million (52 percen) spen on radio adverising. Imporer co-op markeing and public relaions accouned for anoher 31 percen of oal funds wih he remainder being spen for websies, rade adverising, and adminisraive funcions. Wih a oal 2006 budge of $4.5 million, MHAIA spending on radio adverising increased o over $2.0 million, bu radio adverising s oal budge share decreased o 45 percen. Imporer co-op markeing, public relaions, and rade adverising accouned for almos 32 percen of oal expendiures in 2006. The oher major new expendiure of $498,000 was for a NASCAR sponsorship. Wih a oal 2007 budge of $6,246,500, MHAIA spending on consumer radio adverising increased slighly o $2,040,000 (32.7 percen). Wih oal spending of $1.89 million, he share for imporer co-op markeing, public relaions, websie and in-sore adverising was jus over 27 percen. Spending on he NASCAR sponsorship and promoion increased o $1,423,500 (22.8 percen). There is some annual informaion on CAIA markeing programs while i conraced wih he CAC bu only aggregae expendiure daa for is independen 2007 program. Noe also ha he CAIA daa available are for he period each season when avocados from Chile are expored o he U.S. raher han for he California markeing year. Informaion for he period Augus 2003 hrough February 2004 indicaes ha he CAIA markeing program conduced by he CAC included radio adverising, public relaions, and merchandising. Radio adverising accouned for abou 85% of expendiure on all programs. Four 3-week radio campaigns were conduced in 12 seleced markes: 8/18-9/8 including Labor Day, 9/29-10/20, 11/10-12/1 including Thanksgiving, and 12/29-1/19/2004 leading up o he Super Bowl. 7 The CAIA conduced he same markeing programs in he same markes he following season (April 2004 hrough February 2005). Toal spending for radio programs during his period was $3.3 million. CAIA 7 Thanksgiving, Super Bowl, and Labor Day are holidays shown by Li (2007) o represen periods of high avocado consumpion in he U.S. 7
markeing programs from April 2005 hrough February 2006 were similar o he previous wo years wih an emphasis on consumer radio adverising. Raher han four 3-week radio campaigns as previously, he 2005-2006 campaign consised of hree 3-week and wo 2-week programs. In 2007, CAIA also included TV ads for boh he general and he Hispanic markes, as well as spor and media promoion programs. 2.3. Hass Avocado Board Programs HAB markeing programs fall under wo major caegories, Informaion Technology (InfoTech) and Markeing Communicaions (MarCom). The informaion echnology consiss of AvoHQ.com Inrane and he Nework Markeing Cener (NMC), designed o exchange markeing and sraegic informaion from all suppliers of Hass avocados o he U.S. marke in order o improve he flow of frui and mainain orderly markeing condiions. Markeing communicaions consis of consumer communicaions, online markeing, rade communicaions, indusry communicaions and markeing research. Markeing communicaions share of he HAB budge has increased over ime, as he oal HAB budge has increased and as he InfoTech program has become esablished. While geing esablished in 2003, HAB spen $340,179 on InfoTech and $146,499 on MarCom. Expendiures grew for boh programs in 2004, wih $1,090,228 spen on InfoTech and $859,284 spen on MarCom. Consumer communicaions included Super Bowl and Cinco de Mayo promoions and public relaions effors ha resuled in news releases wih high visibiliy. The web sie avocadocenral.com was esablished. MarCom became HAB s major expendiure caegory in 2005, wih a budge of almos $2.7million, wih $746,000 expended for InfoTech projecs. MarCom s budge increased o almos $3.5million in 2006. Parnering wih he Beef 8
Checkoff and a Napa Valley winery, HAB collaboraed o develop a 30-minue elevision program called Ho Trends in Tailgaing ha was aired on hree cable neworks (The Food Nework, HGTV and Fine Living Nework). Six airings of he 30-minue show on each of he hree neworks was supplemened wih 30-second consumer adverising spos ha ran more han 100 imes on four consecuive weeks on he Food Nework, creaing more han 27 million media impressions. The ailgaing heme was coninued in 2007 wih a new parner (Miller Brewing) and a coordinaed program wih reailers, inerne markeing and public relaions. MarCom expendiures increased o almos $3.2 million while InfoTech expendiures were $750,000. 3. Avocado Consumpion in he U.S. U.S. consumpion of avocados before 1990 was largely from California producion; expors and impors were very small. Esimaed U.S. avocado consumpion remained below a pound per capia unil 1975 when a large California crop pushed consumpion o 1.2 pounds per capia. Avocado impors slowly increased during he las half of he 1980s and firs exceeded 25.0 million pounds in 1990. Impors coninued o expand hrough he 1990s and hen exploded as Mexico gained access incremenally o he U.S. marke. Impored avocados accouned for 26.1 percen of U.S. consumpion in 1998 and reached 73.1 percen in 2007. Per capia consumpion expanded wih increased impors, reaching a record high of 3.45 pounds in 2006 and again in 2007. Esimaed U.S. per capia avocado consumpion by source of he avocados is shown in Figure 1 from 1980-2007, demonsraing he sriking growh of impors in oal and as a share of U.S. consumpion. Several facors are associaed wih increased U.S. avocado consumpion. No doub a key facor is he availabiliy of good-qualiy avocados in he U.S. year around. Qualiy has been 9
improved in par by indusry-sponsored merchandising programs for produce personnel in supermarkes ha sress he imporance of proper ripening and having differen mauriy levels available for consumers. As noed, indusry promoion budges have included nuriion programs for several years. Some esimae of impac of hese programs is available from CAC racking sudies which show for example Healhfulness/Good for You as a reason o purchase avocados increased from 35 o 73% of survey respondens beween 1994 and 2004. The indusry has made Figure 1: U.S. Per Capia Avocado Consumpion by Source, 1980. successful effors o have avocados menioned specifically as a recommended frui in die plans and pyramids, such as he Medierranean die pyramid and he Adkins Lifesyle Food Guide Pyramid. 10
Indusry sudies have also examined demographic characerisics of avocado consumers. Cook (2003) described he ypical U.S. avocado purchaser as a women 25-54 years of age, $50,000 plus income, upscale, college educaed, working full/par ime, and healh conscious. The mos frequen uses for avocados are guacamole, par of a Mexican side dish, in a salad, eaen plain, in a sandwich/burger, and as par of a non-mexican enrée (Cook 2003). The HAB has sponsored research examining U.S. Hispanic consumers ha ends o confirm some widely held indusry percepions, including ha Hispanics are heavy users of avocados. The HAB repors ha Hispanics buy significanly more avocados han he average consumer (2007, p. 12). For he ime period racked, 97 percen of Hispanic shoppers bough avocados as compared o 49 percen of he general populaion. In addiion, 60 percen of Hispanic shoppers purchased avocados weekly, and heir average purchase of 4.8 avocados was 58 percen greaer han he average overall purchase of 2.8 avocados. Research reveals wo disinc segmens of he Hispanic marke: U.S.-born Hispanics who speak English in he home and foreign-born Hispanics in Spanish-language-dominan households. U.S.-born Hispanics are more aware of he Hass variey bu purchase fewer avocados han do heir foreign-born counerpars (p. 13). 4. Modeling he Annual Demand for Avocados This analysis benefis from a considerable base of prior research on he avocado marke and on avocado promoion. Previous sudies provide analyical models and empirical esimaes for avocado demand parameers, demand response o promoion programs, and acreage response o price changes. We discuss his work and provide an updaed analysis in his secion. 11
4.1. Previous Sudies Carman and Green (1993) esimaed price and acreage-response equaions as he major componens of a simulaion model of he California avocado indusry ha was used o esimae he impac of generic adverising on acreage and reurns over ime. The price equaion yielded a price flexibiliy of demand of -1.16 and an adverising flexibiliy of demand of 0.15, calculaed a mean values. 8 Carman and Cook (1996) used a revised and updaed version of he Carman and Green model o examine possible impacs of avocado impors from Mexico on he California indusry. The price equaion yielded a price flexibiliy of demand of -1.53 and an adverising flexibiliy of demand of 0.28, calculaed a mean values. Carman and Craf (1998) esimaed boh annual and monhly price equaions for California avocados in a sudy of he reurns o CAC promoional programs. The esimaed annual flexibiliies of demand for price and adverising were -1.33 and 0.13, respecively. Esimaed monhly flexibiliies of demand for price and adverising were -1.54 and 0.137, respecively. 9 Carman and Craf esimaed ha California avocado producers achieved an annual average benefi-cos raio of 2.84 for he 34-year period covered by heir analysis. Shor-erm reurns, based upon assumpion of fixed supply, ranged from $5.25 o $6.35 per dollar spen on adverising. The USDA s Animal and Plan Healh Inspecion Service (APHIS) included an economic analysis of he poenial economic impac of increased Hass avocado impors from Mexico in 8 Thus, a one percen increase in adverising expendiures was esimaed o increase he California f.o.b. price by 0.15 percen. 9 The responsiveness of avocado demand o generic adverising found in hese prior sudies is consisen in magniude o ha found for several oher California commodiies. Alson e al. in heir summary of commodiy promoion programs, lised adverising flexibiliies of demand of 0.13 and 0.16 for eggs and srawberries, respecively, and adverising elasiciies of demand as follows: able grapes, 0.16; dried plums, 0.05; almonds, 0.13; walnus, 0.005; and raisins, 0.029 in Japan and 0.133 in UK (pp. 406-07). Kinnucan and Zheng summarize esimaed elasiciies and benefi-cos esimaes for he dairy, beef, pork and coon promoion programs (pp. 262-63, 270). 12
hree repors issued on proposals o increase he number of saes and ime period for shipmens of avocados from Mexico. In he 2001 and 2003 repors, APHIS used a price elasiciy of demand of 0.86 (USDA 2001, 2003). For he 2004 sudy, hey used a price elasiciy of demand of 0.57 (USDA 2004). 10 USDA APHIS did no consider he possible impacs of adverising and promoion on he demand for avocados. The mos recen analysis of he impac of promoion on U.S avocado demand, based on annual daa from 1962 hrough 2003, esimaed ha imporers could realize reurns ranging from $2.09 o $3.26 per dollar of HAPRIO expendiures, wih ne reurns decreasing as impors increased (Carman 2006, p. 476). This sudy assumed ha he effeciveness of imporer promoional expendiures would be equivalen o he effeciveness of CAC expendiures. 4.2. Economeric Models of he Annual Demand for Avocados Esimaed U.S. avocado demand equaions in he sudies cied in he preceding subsecion included variables for per capia sales, real prices, income, promoional expendiures, and share of Hispanic consumers. Oher possible demand-shif variables, such as he prices of possible subsiues and complemens for avocados, and facors associaed wih rends in demand, including increased seasonal availabiliy of avocados due o impors, changing demographics, and he growing populariy of Mexican foods have also been invesigaed, bu wih limied success. Carman inroduced a variable for he percenage of Hispanics in he U.S. populaion as a measure of he impac of demographic changes and he increased demand for Mexican foods (2006, p.472). He also examined use of variables o measure increased impors (and, hence, 10 The price elasiciy of demand is he inverse of he flexibiliy of demand esimaed by Carman (2006) and ohers. Thus, all of hese esimaes are broadly consisen demand for avocados is inelasic (or flexible), meaning ha a one percen increase in price causes less han a one percen reducion in he quaniy demanded. 13
increased seasonal availabiliy) and accoun for possible subsiues bu none had a measurable impac on avocado demand. Using he resuls of he previous sudies and economic heory, we specified annual demand for avocados in he Unied Saes as a funcion of several explanaory variables: (1) Qa = f (Pa, A,Y,H ), where he variables for a given crop year are defined as follows: Qa is per capia U.S. sales (pounds per person) of avocados from all sources (California, Florida and all impors), less expors from he U.S., Pa is he average real f.o.b. (farm) price of California avocados, Y is real per capia U.S. disposable income, A is he real oal value of avocado adverising and promoion expendiures. H is he percenage of Hispanics in he oal U.S. populaion. The consumer price index for all iems (1982-1984=1.00) was used o deflae prices, income and promoion expendiures o a consan-dollar basis. 11 Deailed informaion on each of he variables used in he analysis, including means and sandard deviaions, is included in able 2. Prior o he creaion of he HAPRIO, A consised mainly of expendiures by he CAC. Hispanics share of U.S. populaion increased from 3.7 percen o 15.0 percen during he period of analysis, and U.S. Census Bureau populaion projecions have Hispanics share of oal populaion seadily increasing o 24.4 percen in 2050. Mexico is he world s larges avocado producer and consumer, wih annual per capia avocado consumpion recenly repored a 8 kg or 11 Based on he resuls of previous sudies, we do no include prices of subsiues or complemens in he demand model for avocados. Some of he previous variables ha have been invesigaed include prices of fresh omaoes, fresh peppers, and leuce (Carman 2006). In essence, he lack of imporance of hese producs as a facor in influencing avocado demand means ha avocados are a unique produc in he die of mos consumers, who are no readily willing o subsiue oher fresh ingrediens in place of avocados in heir dies. 14
abou 17.6 pounds, as compared o abou 3.5 pounds per capia in he U.S. (USDA, FAS, 2005). Wih approximaely wo-hirds of he U.S. Hispanic populaion from Mexico, he Hispanic populaion variable is inended o measure demographic change ha may be relaed o avocado demand. The increasing Hispanic share of U.S. populaion may also ac as a proxy for he populariy of Mexican food and Mexican resaurans in he U.S. Consumer demand is inversely relaed o he price paid based upon he law of demand. A basic quesion is which price o use in he demand analysis. Reail prices differ across sores, and grower prices differ by avocado ype and counry of origin e.g., California avocados generally receive a higher price han avocados from Chile or Mexico. We used he California grower or f.o.b. price because i is he only price series ha is available for he enire period of he daa analysis. Prices for avocados from differen origins (California, Florida, Chile, or Table 2: Variable Definiions and Summary Saisics Variable Definiion Unis Range of Values Qa Pa Annual average per capia U.S. sales of all avocados, (California, Florida and all impors) Average annual f.o.b. price of California avocados deflaed by he consumer price index (CPI) for all iems (1982-1984=1.00) Y U.S. per capia disposable income, deflaed by he CPI for all iems (1982-1984=1.00) A Annual adverising and promoion expendiures by he CAC, HAB, CAIA, and MHAIA deflaed by he CPI for all iems (1982-1984=1.00) H Hispanic populaion as a percenage of oal U.S. populaion pounds per capia real cens per pound housands of real dollars millions of real dollars 0.39 o 3.48 14.33 o 113.88 7.20 o 16.26 0.60-7.55 (CAC); 0.86-6.37 (HAB, CAIA, MHAIA) percen 3.67 o 15.01 Mean S Dev Value 1.49 0.797 50.85 21.656 11.83 2.477 4.17 (CAC); 4.05 (HAB, CAIA, MHAIA) 1.744 (CAC); 2.333 (HAB, CAIA, MHAIA) 7.85 3.487 15
Mexico) and a differen sages of he marke chain (farm, wholesale, reail) should move in unison as a consequence of he law of one price, so he specific choice of price series should be of lile consequence. 12 4.3. Preliminary Daa Analysis Before conducing formal economeric analysis, i is imporan o underake descripive sudies of he variables being considered for he economeric analysis. A basic problem in conducing aggregae ime series analysis of economic relaionships is ha many variables end o move ogeher over ime (o be coinegraed), making i difficul o isolae he unique impacs of each variable. From equaion (1), we seek o explain changes over ime in per capia consumpion of avocados in he U.S. as a funcion of variables such as real per capia U.S. disposable income (Y ), real expendiures promoing avocados (A ), Hispanic share of he U.S. populaion ( H ), and real California grower price (Pa ). Figures 2 and 3 illusrae a fundamenal challenge. Hispanic populaion share and per capia disposable income (figure 2) have increased over he ime period of he daase, 1962-2007, in a raher smooh, coninuous fashion. Alhough promoion expendiure and per capia avocado consumpion (figure 3) have been somewha more volaile, hey, oo, have rended upward over ime. Among he explanaory variables conained in equaion (1), he real California f.o.b price (figure 3) is he only one ha does no exhibi a significan upward rend. 12 The law of one price follows from basic arbirage. For example, prices for avocados from differen counries of origin may differ for a variey of reasons, such as percepions of differences in qualiy, bu as long as buyers are willing o subsiue among he various offerings, heir prices should move in unison. The same argumen perains o prices a differen sages of he marke chain. 16
Figure 2: Per Capia Avocado Consumpion, U.S. Per Capia Disposable Income, and he Percenage of Hispanic Populaion Simple correlaion coefficiens are a good way o measure co-movemen over ime of economic variables. Correlaion coefficiens vary from -1.0 (perfec negaive or inverse correlaion) o + 1.0 (perfec posiive correlaion). A value near zero indicaes variables ha are essenially uncorrelaed. Of course, a high posiive or negaive correlaion does no necessarily imply ha changes in one variable are causing he movemens observed in he oher variable. I 17
Figure 3: Per Capia Avocado Consumpion, he F.O.B. Price, and Promoion Expendiures merely means ha hey move ogeher joinly over ime. Examinaion of able 3 (column 2) reveals ha per capia consumpion, he variable we seek o explain, is highly correlaed wih A,H, and Y hese correlaion coefficiens all range from 0.83 o 0.86. In fac, a simple rend variable, YEAR, consruced by seing is value o 1.0 in he iniial year of he daa and increasing i by 1.0 each successive year, has a similar correlaion wih per capia consumpion. Also imporan o observe is ha Y and H are boh highly correlaed wih YEAR he correlaion coefficiens are 0.97 and 0.99 respecively. In essence, his means ha over he period of our daa, 1962-2007, H and Y each have increased over ime in a manner ha can be almos 18
perfecly prediced wih a linear rend variable. Of course, H and Y are also highly correlaed wih each oher. Real promoion expendiures, A, while quie highly correlaed wih H, Y, and YEAR, exhibi some independen co-movemen, as manifes by correlaion coefficiens wih hese hree variables ranging from 0.74 o 0.77. 13 This is favorable informaion, given he purposes of he sudy, because he independence of movemen of Table 3: Correlaion Coefficiens for he Demand Model Per capia Consumpion (Qa ) Qa 1.0 Pa YEAR A Real CA price (Pa ) -0.50 1.0 YEAR 0.83-0.15 1.0 Real oal promoion Expendiures (A ) 0.86-0.25 0.74 1.0 Hispanic share of Populaion (H ) Real per capia disposable income (Y ) A creaes he poenial o H 0.84-0.15 0.97 0.75 1.0 Y 0.83-0.12 0.99 0.77 0.95 1.0 isolae he impac of promoions on avocado consumpion relaive o he impacs of he oher variables. However, he exremely high correlaions among H, Y, and YEAR mean ha here is no opporuniy o isolae heir individual impacs. 13 Formal ess for he ime-series properies of he model variables were also conduced. These include ess of he null hypohesis ha a variable is saionary (i.e., is a variable ha revers o a consan mean and does no exhibi a deerminisic rend) agains he alernaive hypohesis ha he variable has a uni roo (i.e., he variable has no mean and follows a random walk ). Deailed resuls of hese ess are available from he auhors. Briefly, he California grower price, Pa, has no significan rend and is co-variance saionary, i.e., saionary wihou a deerminisic rend. All oher variables, per capia consumpion, real promoion expendiure, real per capia disposable income, and percen Hispanic populaion have a saisically significan rend as is apparen from examinaion of figures 2 and 3. Per capia consumpion and real promoion expendiure are rend-saionary (saionary afer removal of a linear rend), while real per capia disposable income and percen Hispanic populaion each conain a uni roo. 19
4.4. Srucural Breaks in Per Capia Consumpion Among he challenges in esimaing an annual demand model is ha he fundamenal economic relaionships linking he variable of ineres, per capia consumpion in our case, o he poenial explanaory variables may change over ime. In oher words here may be srucural breaks in he daa. This possibiliy is especially relevan for avocados given he significan changes ha have occurred in he indusry over he 1962-2007 period of analysis in erms of rapidly escalaing impors and he availabiliy of produc year around. Examinaion of he plo of per capia consumpion over ime in figure 3 reveals wo poenial disrupions in he upward rend in per capia consumpion an upward shif in consumpion beween 1980 and 1981 and a downward shif in consumpion beween 1993 and 1994. Furhermore, following he decline in per capia consumpion beween 1993 and 1994, he upward rend in consumpion from 1994 onward is a a higher rajecory han in he preceding years, no doub reflecing a leas in par he progressive opening of U.S. markes o Mexican impors ha began in 1997. A fundamenal issue in he demand modeling is how o handle hese srucural shifs in consumpion and he revised rend in consumpion ha began in 1994. If we do no accoun for hese shifs hrough separae inercepshif and rend variables and insead allow he changes in per capia consumpion o be explained by he variables in equaion (1), he esimaed impac of he promoion variable is much greaer, as is is saisical significance. These resuls are presened nex. 4.5. Esimaed Annual Demand Relaionships Various demand funcions based upon equaion (1) were esimaed using 46 annual observaions for he markeing years 1962 hrough 2007. The key objecive is o deermine he impac of oal adverising and promoion programs on annual U.S. demand for avocados. Resuls from he 20
alernaive esimaions are presened in able 3. All esimaions repored in able 3 were conduced via ordinary leas squares. Several conclusions immediaely follow from examinaion of able 3. Firs, overall explanaory power of he model is very high, as measured by he adjused 2 R saisic, which measures he proporion of oal variaion in per capia consumpion from 1962-2007 ha is explained by he variables included in he model. Adjused 0.99 for he alernaive models presened in he able. 2 R varies from abou 0.92 o nearly Second, price is inversely relaed wih per capia consumpion in a way ha is significan saisically and robus o alernaive model specificaions. 14 This resul is, of course, consisen wih prior sudies and is simply an affirmaion of he law of demand, bu is graifying noneheless as an indicaion of an economeric model ha is working properly. The esimaed price elasiciy of demand, evaluaed a he daa means, ranges from -0.41 o 0.46, depending upon model specificaion. Thus, demand is in he inelasic range, meaning ha a one percen increase in producion causes roughly a wo percen decrease in he f.o.b. price, oher facors consan. 15 14 Conclusions abou saisical significance are based upon he absolue saisics, which measure he precision of esimaes. A larger value for he saisic indicaes greaer precision in esimaion. For sufficienly large samples, an absolue saisic greaer han abou 1.96 allows he researcher o rejec wih 95 percen confidence a null hypohesis ha he rue effec is in fac zero. 15 The esimae of price-inelasic demand is consisen wih resul from prior sudies of avocado demand. The esimae from his sudy is lower (more inelasic) han prior sudies, bu his resul is consisen wih he higher raes of avocado consumpion in he U.S., beginning in he 1990s and coninuing o he presen moving consumpion down he demand curve ino is more inelasic porions. 21
Variable California f.o.b. price (Pa ) Per capia disposable income (Y ) Hispanic share of populaion (H ) Toal promoions (A ) Table 4: Esimaed Annual Demand Models: Ordinary Leas Squares Model 1: Base model -0.0120 (7.48) [-0.414] 0.0739 (1.56) [0.592] 0.0609 (1.87) 0.1192 (5.79) [0.372] Model 2: Base model + rend -0.0125 (7.32) [-0.431] 0.1782 (1.43) [1.429] 0.0798 (2.06) 0.1110 (4.93) [0.347] Model 3: Model 2 + srucural break for 1994-2007 -0.0132 (8.69) [-0.455] 0.1904 (2.88) [1.526] -0.0103 (0.15) 0.0475 (2.04) [0.148] -0.0001 (0.92) Model 4: Model 3 wihou YEAR -0.0131 (8.66) [-0.451] 0.1389 (3.97) [1.114] 0.2878 (0.56) 0.0562 (2.66) [0.176] YEAR -0.0230 (0.91) YEAR 1994-2007 0.0902 0.0795 (3.59) (3.58) D1994-2007 -180.2921-158.9093 (3.60) (3.59) Consan 0.1850 44.485 (0.58) (0.91) 2 Adjused R 0.9192 0.9188 0.9879 0.9879 Noe: Absolue saisics are indicaed in parenheses; elasiciies evaluaed a daa means are in brackes. Third, high correlaion (mulicollineariy) among he variables Y,H, and YEAR makes i impossible o esimae he individual effecs of hese variables on consumpion. Recall from he prior discussion ha each of hese hree variables is almos perfecly correlaed. Thus, alhough we know from economic heory, pas research, and basic informaion on he avocado indusry ha per capia consumpion is likely o be posiively relaed o per capia income and o he percen Hispanic share of he U.S. populaion, i is no possible o isolae hese wo effecs, or 22
for ha maer o separae heir effecs from a simple rend variable ha could be capuring he effecs of boh Y and H, as well as oher omied variables affecing consumpion. This mulicollineariy among Y,H, and YEAR manifess iself in erms of esimaed effecs for each of hese variables being unsable and highly sensiive o model specificaion. For example, Model 3 shows an inverse (and saisically insignifican) effec beween H and Qa and also beween YEAR and Qa, bu hese resuls are merely due o he saisical program impuing all of he impac of hese hree upward-rending variables o Y in his model. In essence, due o heir high mulicollineariy, his aribuion of impac is almos arbirary, as can be seen in erms of he sensiiviy of resuls for hese variables o minor changes in he model specificaion. Imporanly, because our main ineres is in evaluaing he effec of promoions, his inabiliy o separae impacs due o Y,H, and YEAR does no consiue a significan limiaion on he analysis. Fourh, promoions have had a posiive effec on demand ha is saisically significan for all models presened in able 4. However, he magniude of he promoion impac hinges imporanly upon wheher we accoun separaely for he shif in per capia consumpion ha occurred beween 1993 and 1994 and he increasing rend line for per capia consumpion ha began in 1994 and coninues hrough he daase. The downward shif in per capia consumpion is capured by he dummy variable D1994-2007, which is negaive and saisically significan in Models 3 and 4. The greaer rend upward in consumpion ha begins in 1994 is capured by he rend variable YEAR 1994-2007, which is se o 1.0 in 1994 and increases by 1.0 for each subsequen year, and is zero for all years preceding 1994. Imporanly, his change in rend canno be explained by he oher hree variables in he model, Y,H, and YEAR, ha are each 23
rending upward smoohly hrough ime. All hree of hose variables are included in Model 3 and Y and H are included in Model 4. YEAR 1994-2007 is posiive and saisically significan despie he presence of hese oher variables. The one variable in he model ha can accoun, a leas parially, for his increase in he consumpion rend line is oal promoions, which also exhibis an increasing rae of rend beginning abou his same ime, escalaing especially rapidly wih he creaion of he HAPRIO. Thus, when we inroduce separae shif (D1994-2007) and rend (YEAR 1994-2007) variables o accoun for his eviden change in per capia consumpion, i eliminaes roughly half of he esimaed impac of he promoion variable, as comparison of resuls for Models 1 and 2 wih Models 3 and 4 demonsraes. Promoion effecs, however, are noneheless posiive and saisically significan even in Models 3 and 4. The esimaed elasiciy of demand wih respec o promoions ranges from 0.148 o 0.372, depending upon model specificaion. The upper end of his range is high relaive o prior esimaes for he avocado indusry and relaive o oher promoion sudies generally. The lower end of he range is very consisen wih prior esimaes for avocados and in general. We do no consider i possible o obain a reliable separae esimae of he impac of promoion expendiures funded by he HAB (i.e., expendiures in he las five years) using an annual model. The main problem is ha he creaion of he HAB was simulaed in large par by he rapid increase of avocado impors ino he U.S. and he growing free rider problem caused by imporers no conribuing o adverising programs funded hrough he CAC. Thus, HAB s naissance and commencemen of funding programs under is auspices are associaed wih rapidly rising per capia consumpion of avocados in he U.S. Any variables creaed o measure HAB s influence on consumpion apar from he general influence over ime of promoions as measured 24
by A will necessarily capure he rising rend in consumpion, much in he way i is capured presenly by he variable YEAR 1994-2007 in models 3 and 4. Because he creaion of he HAB and he rapid increase in impors and, hence, domesic consumpion are, in essence, simulaneous evens, we canno impue a causal relaionship from HAB s creaion and commencemen of is funding programs o he increase in per capia consumpion. Again, he indusry s abiliy o wihsand he rapid escalaion of impors wihou enduring drasic decreases in real price demonsraes ha demand grew subsanially over his period, a leas parially in our view due o promoions. 4.6. Diagnosic Checks of Annual Demand Models Here we repor briefly on diagnosic ess of he models repored in able 4. These ess are imporan in deermining he confidence we can have in he esimaed resuls. Deails on various diagnosic ess are available from he auhors. The abiliy o aach confidence inervals o esimaed effecs and conduc saisical ess (such as wheher he effec of promoions is saisically differen from zero) depends upon he properies of he esimaed residuals (acual per capia consumpion in year minus prediced per capia consumpion in year ). Tess confirmed ha he esimaed residuals are homoskedasic (i.e., hey have a consan variance) and ha hey are disribued normally. Boh are desirable properies ha suppor he use of he model for hypohesis ess. The residuals do, however, reveal some evidence of serial correlaion, i.e., he expeced value of he error in period is a funcion of he error in period -1. The esimaed coefficiens are consisen in he presence of serial correlaion, bu he esimaed sandard errors (used o consruc he saisics) need o be adjused o correc for is presence. A more insidious 25
problem is ha serial correlaion of he errors may creae problems of endogeneiy of he explanaory variables. 16 Eliminaion of serial correlaion in he errors may, hus, eliminae some endogeneiy problems. Among he explanaory variables in he model, wo are candidaes o be endogenous, Pa and A. The California f.o.b. price may be endogenous because i could be deermined conemporaneously wih consumpion hrough he ordinary workings of he marke. Promoion expendiures could be endogenous because he oal budge for promoion is deermined by he check-off rae muliplied by he oal producion of avocados subjec o he check off. Realisically, here is a lag beween he realizaion of producion and consumpion, he collecion of he check-off funds, he preparaion of markeing budges, and he acual expendiures on promoion, making i reasonable ha promoion expendiures during year were deermined for he mos par by producion in year -1, making A exogenous unless errors are serially correlaed. Our sraegy was o address he endogeneiy of he California f.o.b. price using he wosage leas squares (2SLS) esimaion procedure because good insrumens for i are available, as described in he nex subsecion. Unless here is srong evidence of auocorrelaed residuals in he 2SLS esimaion, we can be relaively confiden ha A is exogenous for he reasons noed. 16 An economeric model suffers from endogeneiy if one or more of he explanaory variables is no exogenous, in which case i is correlaed wih he error erm, making he esimaed coefficiens be inconsisen. Endogeneiy is more likely in he presence of serially correlaed errors because an explanaory variable whose value in period was, say, deermined by evens in period -1 and would, hus, be uncorrelaed wih he error erm in period in he absence of serial correlaion, becomes correlaed wih he error erm in period and, hence, endogenous if he error erm is serially correlaed. 26
4.7. Two-Sage Leas Squares Esimaion The firs-sage esimaion in he 2SLS procedure involves regressing insrumens. The second-sage involves using he prediced values, re-esimaing he annual demand model. Using correlaion wih he error erm. ˆPa in place of Pa on a se of exogenous ˆPa, in place of acual Pa and Pa should eliminae he variable s The key crieria for selecion of insrumens are ha (a) he insrumens mus be exogenous, and (b) hey are correlaed wih Pa. Insrumens seleced for he sage 1 regression Table 5: The Firs-Sage Regression o Predic California f.o.b. Price Variable Hispanic share of populaion (H ) Toal promoions (A ) Coefficien (absolue saisic) -5.7004 (0.46) -5.4186 (2.33) YEAR 1.8234 (0.58) YEAR 1994-2007 0.8776 (0.18) D1994-2007 -1758.741 (0.18) World impor price 0.5355 (2.10) U.S. avocado acreage -0.5175 (1.40) Chilean avocado acreage 0.8209 (0.61) Mexican avocado acreage 0.0377 (0.15) Consan -3523.319 (0.57) 27
were oal U.S. avocado acreage harvesed, Chilean avocado acreage harvesed, Mexican avocado acreage harvesed, and impor price for avocados in he world marke. 17 Sage 1 regression resuls are conained in able 5. The second-sage esimaion resuls are repored in able 6. Given he high mulicollineariy among Y,H, and YEAR, Model 1 includes only YEAR o capure he linear rend effec presen in all hree variables. Model 2 adds Y as an explanaory variable, bu, as expeced, is inclusion does no improve he esimaion, and he esimaed coefficien on income is negaive (couner o he expeced posiive relaionship beween avocado consumpion and income) and is no saisically significan. Table 6: Esimaed Annual Demand Models: Two-Sage Leas Squares Variable Model 1 Model 1 + Income Prediced California -0.0117-0.0109 farm price (Pa ˆ ) (5.64) (3.86) Per capia disposable income (Y ) Toal promoions (A ) 0.0492 (2.34) -0.0698 (0.53) 0.0527 (2.32) YEAR 0.0362 (8.20) 0.494 (1.92) YEAR 1994-2007 0.0964 (5.99) 0.988 (5.86) D1994-2007 -192.778 (5.99) -197.587 (5.86) Consan -69.899 (1.69) -95.385 (1.93) 2 Cenered R 0.96 0.96 The model shows a small increasing rend in consumpion of abou 0.036 lbs/year up o 1994, when he rend line spikes upward, increasing a abou 0.036 + 0.096 = 0.132 lbs. per 17 Harvesed acreage is expeced o be a good insrumen for price because i is deermined in advance of prices and, hrough he link beween acreage and oal producion, should be correlaed wih price. 28