Wha Deermines he Fuure Value of an Icon Wine? New Evidence from Ausralia Danielle Wood Produciviy Commission Melbourne dwood@pc.gov.au and Kym Anderson (corresponding auhor) School of Economics and Cenre for Inernaional Economic Sudies Universiy of Adelaide Adelaide SA 5005 kym.anderson@adelaide.edu.au May 2003 Paper for he Oenomerics IX Conference, Budapes, 22-24 May 2003. An earlier version was presened as an Invied Paper for he Ausralian Conference of Economiss, Adelaide, 1-3 Ocober 2002. Thanks are due o Orley Ashenfeler of Princeon Universiy for inspiring his research, o Langon s Fine Wine Aucions for access o heir daabase, and o he Ausralian Research Council for financial suppor.the work on his paper was compleed while Wood was a suden a he Universiy of Adelaide. The views are he auhors own and in no way reflec he opinions of he Produciviy Commission. www.vdqs.fed-eco.org
Wha Deermines he Fuure Value of an Icon Wine? New Evidence from Ausralia Absrac The Ausralian wine aucion marke is characerised by large variaions in price beween differen vinages of he same wine, relaive o he iniial release prices of hose wines. This paper addresses he quesion: o wha exen can we anicipae he fuure price of such icon wines from informaion available a he ime of release? A hedonic model is esimaed o explain he variaion in price beween differen vinages of four icon red wines. We show ha weaher variables and changes in producion echniques, along wih he age of he wine, have significan power in explaining he secondary marke price variaion across differen vinages of each wine. The resuls have implicaions for winemakers in deermining he prices hey pay for grapes and charge for heir wines, and for consumers/wine invesors as a guide o he qualiy of immaure icon wines.. Key words: wine qualiy, invesmen under uncerainy, hedonic pricing model JEL codes: C23, D12, D44, D80, G12 www.vdqs.fed-eco.org
Wha Deermines he Fuure Value of an Icon Wine? New Evidence from Ausralia I Inroducion The Ausralian wine aucion marke is characerised by large variaions in price beween differen vinages of he same wine. Ye he release prices of hese wines exhibi considerably less volailiy. Thus, here exiss he poenial for buyers o improve heir invesmen reurns by choosing o purchase hose vinages ha are under-priced a he ime of release, relaive o heir fuure secondary marke value. Similarly, wine producers could improve profis eiher by charging higher prices for he beer vinages (reflecive of he price hey will receive laer in he secondary marke), or by holding back some of he beer vinages o sell laer as he wine s fuure qualiy becomes more obvious. An imporan quesion for boh hese producers and buyers is: o wha exen can we anicipae he fuure prices of such icon wines from informaion available a he ime of purchase? Tasing he young wine, even by professionals, is unreliable because he high annin conen makes hem asringen o he palae in heir early years. Weaher condiions during he grape-growing season, long recognised by vignerons as a deerminan of he qualiy of a vinage, may provide an objecive and easily quanifiable guide (Gladsones 1992). Economericians have esed ha hypohesis for Bordeaux wines and found i is srongly suppored (Ashenfeler, Ashmore and Lalonde 1995). A morelimied es on jus one Ausralian wine (Grange) using only hree years of aucion daa gave promising resuls as well (Byron and Ashenfeler 1995). The purpose of he presen paper is o make use of he much larger daabase now available o es his hypohesis for a broader range of icon wines using up o 13 years of aucion prices. Specifically, a hedonic model is esimaed o explain he variaion in he secondary (aucion) marke price beween differen vinages of paricular wines, using several weaher variables plus dummy variables for capuring changes in winemaking and grape growing echniques over ime (based on inerviews wih he chief winemakers of he relevan wineries). The model is esimaed using aucion price daa for four Souh Ausralian icon red wines: hree by Penfolds (Grange, S Henri and Bin 707), and one by Henschke s (Hill of Grace). 1 This aemp o explain he variaion in price beween differen vinages of he same wine label is in conras o numerous sudies ha seek o explain he variaion in price beween he same vinages of differen wineries (see Oczkowski (2001), Schamel and Anderson (2003) and he references herein). 1 Some informaion on hese classic wines is provided in Appendix 1 of Wood and Anderson (2002). For more deails see Halliday (1998) and Read and Caillard (2000). www.vdqs.fed-eco.org
2 The paper is srucured as follow. We firs review previous sudies ha aemp o quanify he relaionship beween weaher condiions during he growing season and wine prices. We hen discuss our choice of variables for explaining he relaionship beween qualiy and weaher, producion echniques and wine age. The nex secion presens he empirical resuls, while he final secion draws ou conclusions, offers some implicaions for winemakers and consumers/invesors, and suggess areas for fuure research. II Previous Lieraure Ashenfeler, Ashmore and Lalonde (1995) were he firs o aemp an empirical explanaion of he variaion in price beween differen vinages of he same wine. They consider he variaion in price beween differen vinages of a represenaive sample of hireen Bordeaux wines (used o creae a vinage price index). The paper uses weaher condiions during he growing season ha produced he wine, widely recognised in he viiculure lieraure as a deerminan of wine qualiy. Ashenfeler e al. also include age as an explanaory variable o capure he effec of increasing scarciy and he opporuniy cos of holding wine. They find ha age alone can explain 21% of he variaion in he price index beween vinages. However, he inclusion of hree weaher variables in he model 2 R ), o 83%. The Bordeaux increases he model s explanaory power (as measured by Equaion as i is ermed, consiues a hedonic price equaion. The coefficiens esimae he implici marginal price of he aribues, in his case he weaher condiions ha produced he vinage and he age of he wine. Tha model was modified for a single wine by Byron and Ashenfeler (1995) in a sudy of Penfold s icon wine, Grange, and by Fogary (2000) in a sudy of he Wes Ausralian wine, Moss Wood. These sudies found a number of weaher variables and age o be 2 significan explanaors of variaion in price of hese wines (wih R of 0.83 and 0.88, respecively). In addiion, he models were demonsraed o have srong ou-of-sample predicive power. The findings of hese sudies suppor he Ashenfeler e al. hypohesis ha he secondary marke price of a given vinage depends on he weaher condiions ha produce he vinage. Jones and Sorchmann (1999) more clearly ariculae he relaionship beween weaher and wine qualiy. They adop a wo-sep approach o modelling he price variaion of Bordeaux wines, boh beween differen wines and across vinages. Firsly, hey esimae a model o explain variaion in sugar and acid conen a harves by climaic variables. Secondly, hey use hese wo endogenously deermined variables as explanaory variables in he price regression, hus highlighing he channel hrough which weaher influences qualiy and hence price. Anoher conribuion made by his paper is recogniion of he conribuion of winemaking echniques o qualiy variaion. III The Model I is possible o explain he exisence of secondary marke price variaion beween differen vinages of he same wine by adaping he hedonic price mehodology proposed by Rosen (1974). Rosen s model explains price variaion beween a differeniaed se of
3 commodiies via evaluaion of an implici or hedonic price associaed wih each characerisic. Rosen (1994) saes ha a paricular class of commodiies can be described by a vecor of n objecively measured characerisics as (1) z? z, z,..., z ) ( 1 2 n where z is he amoun of he i h characerisic conained in each good. i In he conex of his paper he class of commodiies refers o he se of vinages under consideraion for a paricular icon wine. The n characerisics which differ beween he vinages include he age of he wine and he qualiies of he wine once maure, which we include indirecly via objecively measured weaher and echnological change variables ha impar hese qualiies. The marke price of a paricular vinage reflecs he price of each of he characerisics embodied in ha vinage. Tha is, (2) p z)? p( z, z,..., z ) ( 1 2 n where p is an increasing funcion of all he characerisics. 2 The presen paper esimaes his hedonic price funcion in order o relae he price of a given vinage o is qualiy, as indicaed by he objecively measured characerisics of weaher, grape growing and winemaking echniques, and age. IV Explanaory Variables Wha facors explain he variaion in qualiy (and hence price) beween differen vinages of he four icon wines in his sudy? The poenial qualiy of a wine is a produc of he qualiy of he inpus (paricularly grapes) and he winemaking echnique used o ransform hese inpus ino he final produc. The qualiy of grapes in urn is deermined by he ineracion of soil, opography, climae and grape growing echniques. Given ha we are aemping o explain variaion in price of differen vinages of he same wine label, i is reasonable o rea soil qualiy, aspec, slope and aliude as consan beween vinages. Thus, his sudy focuses on variaions in weaher and changes in grape growing and winemaking echniques as explanaors of poenial qualiy differenials. However, he acual qualiy of he wine a any poin in ime depends on wheher i has ye reached or has passed is poenial, and hus we also discuss he imporance of age in explaining qualiy variaion across vinages a any poin in ime. 2 Deailed properies of his model are laid ou in Appendix 2 of Wood and Anderson (2002).
Weaher 4 The influence of weaher condiions during he growing period on grape qualiy has been well esablished. In recogniion of he imporance of climae, winemakers develop grape growing echniques o maximise he beneficial aspecs of climae while reducing weaher-based flucuaions in qualiy. Smar (2001) argues ha while all climae parameers can be imporan in influencing grape qualiy, emperaure is undoubedly he mos imporan. Gladsones (1999) suggess an average daily emperaure during he growing season (mid-sepember o March in souhern Ausralia) of 20-22ºC is opimal for he formaion of colour, flavour and aroma compounds in red able wines. Thus, we assume ha grapes grown under hese opimal condiions will be of he bes qualiy, and vinages produced from hese grapes will receive he highes prices. Ashenfeler, Ashmore and Lalonde (1995) repor a posiive linear relaionship beween average emperaure during he growing season and price for Bordeaux red wines. However, when considering he warmer Barossa region of Souh Ausralia where he average growing season emperaure regularly exceeds he suggesed opimum, Byron and Ashenfeler (1995) find a quadraic funcion o be he mos appropriae way o model he effec of emperaure on wine prices. The quadraic funcion hey esimae is concave wih a urning poin of 19.05 degrees, jus slighly below he emperaure range Gladsones pus forward as opimal. Temperaure also has he poenial o affec qualiy and yields hrough is variaion. Gladsones poins ou ha he biochemical processes of grape developmen are favoured by a low diurnal emperaure range (ie. he difference beween he daily maximum and he nighly minimum emperaures). His argumen is suppored by Byron and Ashenfeler, who find a significan negaive relaionship beween he price of Grange and he average emperaure differenial during is growing season. The fac ha diurnal emperaure variabiliy has he poenial o affec grape qualiy suggess ha average daily emperaure (he average of he daily maximum and minimum emperaures) may no be he mos appropriae index o es he affec of emperaure on grape qualiy. For example, even hough one vinage year may have a lower average emperaure han anoher, his may simply be because he minimum emperaures are lower (and hence he diurnal emperaure range larger). The average maximum emperaure during he growing season herefore would seem o be a more reliable index han simply he average emperaure o quanify he affec of emperaure on a paricular grape vinage. This is suppored by Happ (1999), who develops a hea work index which measures he hermal characer of a paricular day based on 20-minue emperaure observaions. He shows ha, as a less daa inensive approximaion, he average maximum emperaure provides a superior measure of he emperaure condiions affecing wine grapes o he average emperaure. Happ s (1999) index is based on he numbers of hours of he day wih emperaures beween 16ºC and 22ºC (a favourable zone for developmen of grape flavour compounds). While a simple daily maximum does no provide such an accurae measure of opimal condiions, Happ shows i provides a reasonable indicaor of a sie s endency o experience emperaures above 22ºC. The number of hours of sunshine is anoher variable imporan o grape qualiy, boh direcly and for is ineracion wih emperaure. Gladsones (1992) suggess ha sunshine hours during he growing season, paricularly in early spring, have a posiive influence on
5 qualiy. However, previous saisical analyses (Ashenfeler e al. 1995, Byron and Ashenfeler 1995, and Fogary 2000) fail o idenify any saisically significan relaionship beween hours of sunshine and icon wine prices. This failure is likely o be linked o he correlaion beween sunshine hours and emperaure, which makes isolaing heir separae effecs difficul. Afer emperaure, Smar (2001) ranks rainfall as he nex mos imporan climaic deerminan of grape qualiy. As Gladsones (1992) poins ou, i is he seasonal disribuion of rainfall which is imporan. Rainfall during winer and early spring aids grape developmen, paricularly since he four wines considered come from vineyards ha rely wholly or mosly on precipiaion. On he oher hand, rainfall in he period prior o harves can waerlog he soil and hus prove derimenal o boh grape yield and qualiy. These effecs find saisical suppor in he sudy by Ashenfeler e al. (1995), who repor evidence of a negaive relaionship beween rainfall prior o harves and Bordeaux wine prices, and a posiive relaionship beween rainfall during he winer preceding he vinage and price. While he sudy of Grange prices by Byron and Ashenfeler (1995) also finds saisical evidence of he derimenal effec of rainfall prior o harves, hey do no find any saisically significan relaionship beween winer rainfall and price. Gladsones (1992) also suggess here is a posiive relaionship beween wine qualiy and relaive humidiy in February, he las monh of he growing season. This relaionship is paricularly imporan in he relaively warm wine regions in Ausralia where afernoon humidiy is necessary o encourage ripening when February emperaures are high. The final climaic variable lised by boh Gladsones (1992) and Smar (2001) as imporan o grape qualiy is windiness. Wind can have boh a posiive and negaive influence on qualiy. On he posiive side, wind can help preven fross and provides air circulaion o he vines (which lowers humidiy). However, srong winds have he poenial o harm grape qualiy (Hamilon 1988). In Souh Ausralia, added dangers arise from ho, dry summer winds because hey can cause imperfec ripening. Vineyard managemen echniques In addiion o hese weaher influences, changes in vineyard managemen echniques can explain qualiy differences beween he grapes used o produce differen vinages. Gladsones (1992) deails a range of pracices imporan o boh grape yield and qualiy. The spacing of vines deermines he exposure of vines o sunligh, waer and soil nuriens and herefore affecs boh yield and qualiy. Also affecing sunligh exposure is he orienaion of rows. The heigh of vines is also imporan, because i deermines he amoun of hea he vine is exposed o via radiaion from he soil. Furher improvemen of he efficiency of ligh use can come from he adopion of a suiable rellising sysem and canopy managemen. Irrigaion, ferilisaion, arificial drainage and windbreaks are oher vineyard managemen echniques ha may be imporan influences on grape yield and qualiy.
Winemaking echniques 6 High-qualiy grapes are an essenial bu no sufficien condiion for producing highqualiy wines. I is only when qualiy grapes are combined wih superior winemaking echniques ha excellen wines are produced. The firs imporan face of winemaking is he selecion of he grapes, followed by any blending. For he hree Penfold wines considered in his sudy, he blends change wih each vinage along wih changes in grape qualiy from differen sies. For Henschke s Hill of Grace, a single-vineyard wine, he absence of he opion o blend o offse qualiy variaion means ha he choice of grapes is of umos imporance. Anoher imporan aspec of making icon wines is he oak in which he wine is maured. Changes in he ype of oak and he lengh of mauraion can aler he disincive qualiy of he wine and hus is marke price, as can wheher he barrels are new or used. Age of he wine A characerisic of icon wine of he sor considered in his sudy is heir abiliy o develop and improve wih age. These wines are characerised by a high conen of annins in heir youh, making hem unpleasan for early drinking. Then as he annin conen recedes, he qualiy gradually improves unil he maximum qualiy is reached. This sae can persis for a number of years or even decades before he qualiy begins o decline. Alhough previous sudies model he relaionship beween qualiy and age linearly (Ashenfeler e al., Byron and Ashenfeler, and Fogary), i seems reasonable, given he naure of he mauraion process, o model age as a quadraic. Age is also relaed o price because he scarciy of a given vinage is non-decreasing wih ime. As a wine ages, more of he given vinage sock is consumed so ha scarciy, and hence price, increases. V The Daa Price daa Wine aucions are he principal secondary marke for icon wines in Ausralia, aided by he fac ha he liquor licensing laws in many saes prohibi privae sales of wine. Aucion prices provide a comparaively high degree of price ransparency, and herefore provide he bes indicaion of he equilibrium value of a paricular vinage of wine a any poin in ime. In addiion, hese aucions are of he silen bid-wrien ender kind and hus circumven many of he problems of aucions as a price deerminaion mechanism. Langons represen over 70% of he wine aucion marke in Ausralia, 3 and hey have kindly provided he price daa for his sudy. 3 Deails are given in he annual repor of Langon s Ausralian Fine Wine Invesmen Guide, Sydney.
7 The daa provided are he high and low sale price and he dae for every occasion on which each of he four icon wines was raded over he period 1988-2000. From his, he unweighed average of he sale price for each vinage in each aucion year is calculaed for each of he four wines. The prices are unweighed because daa on he volume of wine raded a each dae are, unforunaely, no available. However, in so far as differences in volumes raded are no large and are randomly disribued for a given vinage, his should no unduly affec he analysis. Because he wine aucion marke in Ausralia only really developed in he las decade, many of he vinages were no raded every year, paricularly in he firs few years. Thus, only he aucion years which provide a sufficien number of observaions are considered for each wine. They are 1988-2000 for Grange, 1994-2000 for S Henri and Bin 707, and 1995-2000 for Hill of Grace. In addiion, some of he earlier vinages of each wine are excluded from he analysis eiher because hey were oo infrequenly raded or because heir price reflecs collecor value raher han qualiy. These older vinages belong in a differen marke, where price is driven purely by scarciy raher han qualiy. The vinages ha are included are: 1960-1995 for Grange, 1965-1995 for S Henri, 1971-1994 for Hill of Grace, and 1976-1995 for Bin 707. Despie, narrowing he daa ses here sill remains a number of missing observaions, when a paricular vinage is no raded in an aucion year wihin he sample range. The unweighed average nominal price in each aucion year is convered ino real erms using he Consumer Price Index wih base year 2000. Weaher daa The weaher daa are sourced from he Ausralian Bureau of Meeorology s Nurioopa and Coonawarra Climae Saions. A major issue is deermining he weaher condiions during he growing season for Penfolds grapes, becausehey are sourced from various regions bu in differen proporions each vinage. Because he exac percenages of grapes used from each region o produce a given vinage are no publicly available, i is no possible o creae a weighed average weaher index for each vinage. Mike Farmilo, he Chief Winemaker a Penfolds, suggesed he Nurioopa Climae Saion would be he mos represenaive sie for Grange, and indeed Byron and Ashenfeler (1995) found ha he daa from his saion provided he bes fi. Hence his sudy will also use Nurioopa as a represenaive Climae Saion for Grange. S Henri, like Grange, also uses grapes primarily sourced from he Barossa region and so Nurioopa will be used as he represenaive saion for his wine also. For Bin 707, he majoriy of grapes are sourced from he Coonawarra region and so Coonawarra is chosen as he represenaive climae saion. However, daa from Coonawarra are no available for all he variables of ineres and hus Nurioopa daa are used for he variables no colleced a Coonawarra. For Hill of Grace, a single sie wine, Nurioopa is he closes Climae Saion wih sufficien daa available.
Technological change daa 8 Daa on he major changes in viiculural managemen and winemaking echniques were colleced by inerviewing he Chief Winemakers a Souhcorp (he producer of Penfolds) and Henschke. In he inerviews John Duval from Penfolds and Sephen Henschke discussed wha hey considered o be he major changes in heir grape growing and winemaking echniques over he sample period. In recogniion of he imporance of grape qualiy o all is blended wines, Penfolds inroduced he sar qualiy sysem in 1983. This sysem helps overcome he possible principal-agen problem, by offering subsanial bonuses o grape growers in line wih he number of sars (a measure of qualiy) ha heir grapes achieve. Penfolds managemen adoped his sysem o mainain he inegriy of heir op wines, by inroducing a minimum sar requiremen for he grapes used o produce each wine. However, while i is likely he plan has improved qualiy (and reduced qualiy risks) since is inroducion, he fac ha i was phased in over a number of years makes i less likely o show up as significan han if i was inroduced overnigh. Anoher innovaion inroduced across all hree Penfolds wines was a change in grapepressing echniques in 1990, again aimed a improving he qualiy of he wine produced. For Bin 707, a suble change in syle was inroduced in 1985, from which vinage he wines became more focused. Thus, we also es for his change in syle using a dummy variable o see if i is posiively recognised in he secondary marke price. Langons 1998 Fine Wine Invesmen Guide suggess ha early vinages of Hill of Grace (1971-1977) are no considered as disinguished as vinages produced from 1978 when Sephen and Pru Henschke ook over he family winery. Thus, we include a dummy variable wih a break a 1978 o deermine wheher his is indeed he case. In addiion, Sephen Henschke deailed a number of changes inroduced during his period of managing he winery ha may have improved he qualiy and hence he demand for he wines. In 1983, refrigeraion was inroduced ino he winery. In he same year, he wine was maured in new French Oak barrels for he firs ime. And in a furher aemp o improve grape qualiy, a new rellising sysem was inroduced in 1990. VI Esimaion Mehodology The price daa se for his sudy is more exensive (in he number of boh vinages and aucion years) han considered in he previous sudies of single wines by Byron and Ashenfeler (1995) and Fogary (2000). Boh hose analyses use only hree years of aucion daa, and boh sudies esimae separae hedonic price funcions for each year for jus one wine. However, given ha we have a number of years of aucion daa for each wine, panel daa analysis is a more appropriae mehodology o adop. Panel daa analysis uilises longiudinal daa ses, examining a se of he same individuals a various poins in ime. The daa se for his sudy is unique in ha boh he cross secional (aucion year) and ime series (vinage year) componens of he daa se
9 have a ime dimension. However, his by no means limis he applicaion of he normal panel esimaion echniques. The advanage of employing he panel daa mehodology in his conex is ha we have a greaer number of observaions for a given number of esimable parameers, and herefore we achieve more efficien esimaes of hese parameers. A key consideraion in panel daa analysis is differeniaing beween homogeneous (same for each cross secion) and heerogeneous (cross secion specific) parameers. I is necessary o provide a heoreical basis for he choice in each case. We would expec o see variaion in price beween aucion years, reflecing underlying macroeconomic influences or changes in ases or prices of subsiues or complemens ha are no included in he model. Thus, we conclude ha he pooled OLS echnique (which simply pools daa across cross-secional unis) is inappropriae. However, heoreical consideraions sugges ha he coefficiens of he explanaory variables would be consan across aucion years. Tha is, we expec implici prices associaed wih he weaher and echnique variable o be independen of facors such as macroeconomic condiions. Thus, we only need o esimae one parameer for each explanaory variable. These heoreical consideraions sugges ha fixed or random effecs models would mos appropriaely capure he heerogeneiy in he daa. The fixed effecs approach includes an aucion year specific consan erm o capure he heerogeneiy, whereas he random effecs model includes an independenly and idenically disribued aucion year specific disurbance. Given he likely sysemaic naure of he aucion year specific effecs, i is more appropriae o employ he fixed effec esimaion mehodology here. Given all he above, we evaluae he implici prices associaed wih he characerisics impared hrough each of he explanaory variables by specifying he following hedonic price funcion: P? i F( Ai, W, I ) where P i = he aucion price of vinage in aucion year i, A = he age of vinage in aucion year i, i W = he weaher condiions during he growing season of vinage, I = he echniques used by winemakers o produce vinage, More specifically, he explanaory variables, as discussed above, are: Age i : Age of vinage in aucion year i ; RainH : Toal rainfall in January, February and March, (he period jus prior o harves) in he year ha produced vinage ; MaxTemp : Average daily maximum growing season emperaure, calculaed by averaging he daily maximum emperaures over he monhs Ocober- March; MaxTempS : MaxTemp squared,
TempD : Humid : Wind : WindS : Sun : Press : Syle : SP : Fridge : Oak : 10 Average daily growing season emperaure variaion, calculaed as (average monhly max average monhly min) hen averaged over he monhs Ocober- March, Average daily 3p.m. relaive humidiy in February (las growing monh), Average daily wind speed in November and December, Wind Squared, Toal hours of brigh sunshine during spring (Sepember, Ocober, November). A dummy variable for Penfolds wines wih break a 1990, he year in which Penfolds changed heir pressing echnique, A dummy variable for Bin 707 wih a break a 1985, he year Penfolds changed he syle o creae a more focused wine, A dummy variable for Hill of Grace wih a break a 1978, he year Sephen and Prue Henschke ook over he family winery, A dummy variable for Hill of Grace wih break a 1983, he year refrigeraion was inroduced in he winery, A dummy variable for Hill of Grace wih break a 1990, he year a new rellising sysem was inroduced and also he year from which he wine was maured in new French Oak. VII Resuls The esimaes of he hedonic price funcions for each of he four wines are presened in Tables 1 and 2. I is eviden ha for each wine here is a saisically significan relaionship beween he age, weaher and echnological change variables and he secondary marke price of he wines. Each variable is considered in urn. Age of he wine The effec of age on price is no direcly comparable across all four wines in he sudy because of he differen funcional forms adoped. For Grange and S Henri, age is esimaed o effec he secondary marke price in a linear fashion. For boh wines, he coefficien on age is posiive and significan as expeced. However, he magniude of he coefficiens differ markedly. The coefficien for Grange, 0.025, suggess ha an exra year of aging corresponds o an average 2.5% increase in he real price, ceeris paribus. This is in line wih previous sudies (Ashenfeler e al., Byron and Ashenfeler, and Fogary) which repor esimaed coefficiens (based on nominal prices) of 0.035, 0.041 and 0.03, respecively. S Henri s coefficien on age, 0.003, is significanly differen from zero bu much lower. We suggesed wo facors ha would deermine he coefficien on age: qualiy improvemen and increasing scarciy over ime. In his case he difference in he esimaed coefficiens may be due o he difference in he aging poenial of he wo wines, wih Grange exhibiing a much longer erm cellaring poenial han S Henri. According o Read and Caillard (1994), all bu four of he vinages of Grange in our sample were sill a heir peak or were ye o reach heir peak in he aucion years considered. For S Henri, by
11 conras, many of he earlier vinages were considered pas heir peak, in which cases an addiional year of aging would no be valued for S Henri. Also, like many of he firs growhs from Bordeaux ha achieved a similar coefficien on age in he Ashenfeler e al. sudy, Grange has collecor value well beyond is poenial for drinking. Tha is, he vinages of Grange sill in circulaion bu possibly pas heir prime for drinking are mainly in he marke as aniques, raher han as par of he marke for fine wine, wihconsumers implicily valuing highly heir scarciy. For Hill of Grace and Bin 707, age produces he bes fi when modelled as a quadraic. For boh wines he funcion is concave and significan. The urning poins of he funcions are approximaely 13 years for Bin 707 and 24 years for Hill of Grace. These are esimaes of he average peak drinking imes for hese wines. (Drinking recommendaions in Read and Caillard (1994) have 13 years as he lower end of he average cellaring recommendaions for Bin 707.) For boh hese wines, heir price is esimaed o decline a a slow rae afer he peak qualiy is reached, consisen wih he idea ha he qualiy of wellcellared icon wines plaeau and hen decline only slowly afer reaching heir peak. This adds suppor o he hypohesis ha he relaionship beween price and age for hese wo wines is primarily driven by qualiy. Temperaure Temperaure is found o have a significan effec on he secondary marke price variaion of he vinages for he icon wines considered. For he hree Penfold s wines, he average maximum emperaure during he growing season was shown o be significan in explaining he secondary marke price variaion beween he vinages. For Grange, and S Henri, his relaionship is found o be bes modelled as a linear one, wih coefficiens of 0.094 for Grange and 0.024 for S Henri. This suggess ha a one degree increase in he average maximum growing emperaure, leads o a 9.4% increase in he price of Grange bu only a 2.4% increase in he price of S Henri. For Bin 707, he relaionship beween secondary marke price and average maximum growing season emperaure is found o be bes approximaed by a quadraic. Tha is, higher average maximum emperaures lead o higher secondary marke prices up o he 24.5 C opimal level, bu for emperaures higher han 24.5 C, he opposie is rue. These resuls canno be direcly compared o he previous sudies, because we consider he average maximum emperaure raher han average emperaure. The diurnal emperaure range is also found o be significan in explaining he variaion in price beween vinages for all of he wines considered. Consisen wih he viiculure lieraure, he esimaed relaionship in each case is negaive, suggesing a large emperaure range has a negaive affec on he qualiy of he grapes in a paricular vinage year. Rainfall The esimaed effec of oal rainfall in he period prior o harves was remarkably consisen across all four icon wines. In line wih viiculural expecaions, higher rainfall during his period has a negaive effec on grape qualiy and hereby on secondary marke
12 prices. We esimae a one-millilire increase in he oal rainfall prior o harves leads o an average 0.4% decrease in he secondary marke price of Grange, 0.5% for S Henri and 0.2% for boh Hill of Grace and Bin 707. These esimaes are also remarkably similar o hose esimaed in oher sudies (0.4% for Ashenfeler e al. and 0.3% for Byron and Ashenfeler). On he oher hand we find winer rainfall, esimaed in he Ashenfeler e al. sudy o be significanly posiively relaed o he secondary marke price of Bordeaux wines, o be posiive bu insignifican for all four Ausralian icon wines. Ineresingly, neiher of he wo previous hedonic price sudies of Ausralian wine deeced a significan winer rainfall effec. Presumably his is because of he greaer use of irrigaion in Ausralia, which can compensae for below-average winer rain. Oher weaher variables Humidiy, windiness and sunligh were all discussed above as poenially imporan o wine qualiy. Alhough hey were no found o be so in previous saisical sudies, in his sudy hey are all found o be significan for a leas one of he wines considered. Humidiy in he las monh of he growing season is significan and of he expeced sign and magniude for boh Grange and Bin 707. Windiness in November and December, esimaed o be significan in explaining he price variaion of all of he wines considered, is in each case bes modelled as a quadraic, consisen wih viiculure heory. The opimal wind speeds were esimaed o be 11.3 kmph for Bin 707, 12.0 for Hill of Grace and slighly higher a 16.7 and 18.2 for Grange and S Henri, respecively. The model suggess ha here are negaive implici prices associaed wih higher winds once hese opimal speeds are reached. The difference in he opimal levels is likely o be relaed o he ineracion of wind wih he oher weaher variables and he underlying syle of he wine. While sunshine during spring is only found o be significan in explaining price variaion for Hill of Grace, i is esimaed o have a large affec for his wine: one addiional hour of brigh sunligh during spring is esimaed o increase he secondary marke price of he vinage by 2.7%. However, correlaion beween hours of sunligh and average emperaure may explain he absence of a significan sunligh relaionship for he oher wines. Presumably his is why hese wo variables are never boh significan in explaining price variaion for he same wine. Technological change Changes inroduced by he viiculuriss and oenologiss are shown o have a posiive effec on he secondary marke prices of heir wines wherever hey show up as being saisically significan. The inroducion of a new pressing echnique for Penfolds wines in 1990 has had a posiive effec on he price of Grange. The change in syle by winemakers o produce a more focused Bin 707 in 1985 is also well received by consumers, who increased heir willingness o pay by an average of 17.5% for he new syle vinages. For Hill of Grace, consumers are increasingly willing o pay for vinages produced since he ime ha Sephen and Prue Henschke ook over heir family winery in 1978. Alhough i is no possible o separae ou he effecs of he inroducion of refrigeraion and he swich o
13 new French Oak in 1983, joinly hese changes are shown o have furher increased he secondary marke prices of ha wine. Furhermore, he improvemen in grape qualiy as a resul of a new rellising sysem in 1990 also is refleced in he secondary marke prices of vinages produced from ha year onwards. Aucion price dummies The esimaed aucion year specific consan erm for each wine is presened in Table 2. Alhough hese consans have limied inerpreaion, we noe ha for each wine he consans end o increase in each successive aucion year. These consans are designed o capure he effecs of exogenous influences on prices across aucion years. Therefore, he increase over ime in he aucion year dummies may be a reflecion of income growh or shifs in preferences owards consumpion of Ausralian icon wines in he sample period (e.g. greaer awareness of hese wines in Asia as a consequence of publiciy via wine wriers such as Rober Parker), causing prices o rise independen of qualiy over he periods of aucion years considered. This rend reversed slighly in 1999 or 2000 for all of he wines, perhaps reflecing he influence on consumer and invesor confidence of he Asian financial crisis of he lae 1990s. Goodness of fi The signs of all he coefficiens in our model are in line wih viiculural expecaions. In addiion, here are oher indicaions ha our model is explaining well he price variaion 2 beween differen vinages of he same wine. Firsly, he unweighed adjused R for every model is high, a 0.77, 0.59, 0.70 and 0.78 for Grange, S Henri, Bin 707 and Hill of Grace, respecively. This indicaes ha he weaher and changes in producion echniques, along wih he age of he wine, explain around 60% of he price variaion beween vinages for S Henri, and 70-78% for he oher wines. In addiion, he models in each case demonsrae significanly superior explanaory power compared o models esimaed wih age as he only explanaory variable. The unweighed adjused R 2 for he models wih weaher and echnological change variables excluded are jus 0.62, 0.45, 0.45 and 0.67 for Grange, S Henri, Bin 707 and Hill of Grace, respecively. Furhermore, he model is found o be paricularly robus o alernaive specificaions in each case. For each wine he regression was re-esimaed, sysemaically dropping one of he vinage observaions each ime. The coefficiens in each of he models exhibied minimal variaion, and here was shown o be no saisically significan change in he models explanaory power (as measured by adjused R 2 ) any of he imes ha each model was re-esimaed. The coefficien values aribuable o each of he explanaory variables were also shown o be remarkably robus o he inclusion of insignifican explanaory variables in each model. Also, he acual price was almos always wihin he 95 per cen confidence bands around he esimaed price, and up o four-fifhs of he price urning poins were prediced be he model (see Figure 1 for he Hill of Grace equaion for aucion prices in 2000, for example). Thus, given he heoreical plausibiliy of our models, along wih heir eviden saisical robusness, we conclude ha here is suppor for our hypohesis ha he variaion in consumer s willingness o pay for differen vinages is
14 relaed o qualiy differenials across he vinages brough abou by variaion in age, weaher and producion echniques. VIII Summary and Conclusions This sudy adds o he exising empirical lieraure on wine pricing by considering he deerminans of secondary marke price variaion beween differen vinages of he same wine. We assume ha he price of a given vinage on he aucion marke is deermined by he qualiy of he wine. Therefore, he price variaion beween vinages can be explained by differences in qualiy, which affecs he consumers willingness o pay (and sellers o sell). In paricular, exending he framework of Ashenfeler, Ashmore and Lalonde (1995), we are able o explain hese qualiy differences on he basis of weaher condiions during he growing season and improvemens in grape growing and winemaking echniques, in addiion o age of he wine, via esimaion of a hedonic price funcion using panel daa echniques. The heoreical plausibiliy of our model is suppored by he fac ha our esimaes of he implici prices impared hrough superior weaher condiions and winemaking echniques are in line wih viiculural expecaions. Our model is furher validaed by is robusness o alernaive specificaions, and by is high values for R 2. The findings of his sudy have possible implicaions for viiculuriss, wine makers, wine consumers and invesors. For viiculuriss, he sudy provides an indicaion of he mos appropriae climaic regions for growing ulra premium wine grapes. In he pas, his decision has been based mainly on he undersanding of he relaionship beween erroir (physical and hydrological aspecs of soil, macro- and meso-climaes, opography, ec.) and qualiy. However, by quanifying he relaionship beween weaher and price, our sudy may allow a more deailed cos-benefi analysis of he decision o plan in differen areas, given heir climaic hisory. 4 For winemakers buying grapes from independen growers, he price paid could be se in par on he basis of he weaher variables using he equaions presened in his sudy, pending he developmen of more-reliable quaniaive indicaors of grape qualiy such as grape colour, baume, ph using NIRS echnology. 5 For he wine consumer, our sudy provides an objecive guide o qualiy of immaure icon wines. While he consumer has a number of avenues for esablishing he qualiy of he wine once maure, such as exper asers opinions (see, for example, Schamel and Anderson 2003), less informaion is available he furher away are he peak drinking years of he wine. Since weaher variables evidenly provide a reliable indicaor of fuure qualiy, our model provides a useful guide o he wine invesor/consumer. Alhough no explored in his sudy, due o difficuly in obaining he release price daa, wine invesors and/or producers may be able o benefi from a sudy of he efficiency of he primary (release) markes for icon wines. If he release price of a wine does no fully 4 See Ashenfeler (1997) on how his hedonic approach can be applied o vineyard sie selecion. 5 See Golan and Shali (1993). For deails of he progress being made in Near Infra Red Specroscopy (NIRS) in Ausralia, see GWRDC (2001).
15 reflec all he weaher feaures of is vinage, hen wine invesors could exploi he publicly available weaher informaion and make economic profis by choosing only o inves in hose vinages ha are under-priced a he ime of heir release. Alernaively winemakers could use ha same informaion no only in seing heir grape purchase price bu also he wine release prices and/or quaniies. For example, if he winemaker knows he weaher condiions were excepionally good for her/his grapes relaive o ohers grapes in he same region in a paricular year, s/he would benefi from wihholding some of ha vinage for laer sale once consumers realize how excepional is ha vinage. We have ye o go back hrough criics raings books o see he exen o which heir raings when hese wines were firs released provided beer or worse guidance o consumers han he weaher variables idenified in he equaions esimaed in his sudy. Anoher ineresing avenue for fuure research would be backcasing, o show he exen o which our model can rack pas aucion price movemens. Does he acual price approach he price prediced by our model (from below in good vinages and from above in poor vinages) as he wine ages? Is he pace of ha convergence increasing over ime as invesors/consumers/producers learn more abou he weaher and oher deerminans of qualiy (and in paricular as viiculuraliss find more-precise ways o compensae for adverse weaher condiions)? And why do producers vary he release price of heir icon wines so much less from one vinage o he nex relaive o he flucuaions in weaher condiions from year o year?
16 REFERENCES Arguea, N. and C. Hsiao (1993), Economeric Issues of Esimaing Hedonic Price Funcions, Journal of Economerics 56: 243-67. Ashenfeler, O. (1997), A Hedonic Approach o Vineyard Sie Selecion, Presenaion o Fifh Annual Meeing of he Vineyard Daa Quanificaion Sociey. Ashenfeler, O., D. Ashmore and R. Lalonde (1995), Bordeaux Wine Vinage Qualiy and he Weaher, Chance 8(4): 7-14. Byron, R.P. and O. Ashenfeler (1995), Predicing he Qualiy of an Unborn Grange, Economic Record 71: 400-14. Conbris, P., S. Lecoq and M. Visser (1997), Esimaion of a Hedonic Price Equaion for Bordeaux Wine: Does Qualiy Maer? The Economic Journal 107: 390-402. Fogary, J. (2000), Prices, Age and Weaher, Unpublished Honours hesis, School of Economics, Universiy of Wesern Ausralia, Perh. Gergaud, O. and V. Ginsburgh (2001), Endowmens, Producion Technologies and he Qualiy of Wines: Is i Possible o Produce Wine on Paved Roads? mimeo, Brussels: ECARES, Universié Libre be Bruxelles. Gladsones, J. (1992), Viiculure and he Environmen, Adelaide: Wineiles. Golan, A, and H. Shali (1993), Wine Qualiy Differenials in Hedonic Grape Pricing, Journal of Agriculural Economics 44: 311-21. Grape and Wine Research and Developmen Corporaion (GWRDC) (2001), R & D Highlighs, Adelaide: GWRDC. Halliday, J. (1998), Classic Wines of Ausralia, Sydney: Harper Collins, 2 nd Ediion. Hamilon, R.P. (1988), Wind Effecs on Grape Vines, pp 65-68 in he Proceedings of he Second Inernaional Symposium for Cool Climae Viiculure and Oeneology, Auckland. Jones, G.V. and K.H. Sorchmann (2000), Wine Marke Prices and Invesmen under Uncerainy: An Economeric Model for Bordeaux Crus Classés, Agriculural Economics 76(1): 1-18. Lancaser, K.J. (1966), A New Approach o Consumer Theory, Journal of Poliical Economy 74(2): 132-57. Landon, S, and C. Smih (1998), Qualiy Expecaions, Repuaion and Price, Souhern Economic Journal 64(3): 628-47. Oczkowski, E. (2001), Hedonic Wine Price Funcions and Measuremen Error, Economic Record 77(239): 374-82, December.
17 Read, A. and A. Caillard (2000), The Rewards of Paience: A Drinking and Cellaring Guide o Penfolds Wines, Sydney: Penfolds Wines, 4 h ediion. Rosen (1974), Hedonic Prices and Implici Markes: Produc Differeniaion in Pure Compeiion, Journal of Poliical Economy 82(1): 34-55. Schamel, G. (2000), Individual and Collecive Repuaion Indicaors, CIES Discussion Paper 00-09, Cenre for Inernaional Economic Sudies, Universiy of Adelaide. Schamel, G. and K. Anderson (2003), Wine Qualiy and Varieal, Regional and Winery Repuaions: Hedonic Prices for Ausralia and New Zealand, Economic Record 79(246), Sepember (forhcoming). Smar, R. (2001), Where o Plan and Wha o Plan, ANZ Wine Indusry Journal 16(4): 48-50. Wood, D. and K. Anderson (2002), Wha Deermines he Fuure Value of an Icon Wine? Evidence from Ausralia, CIES Discussion Paper 0233, Cenre for Inernaional Economic Sudies, Universiy of Adelaide, November.
Table 1: Regression esimaes (a) Penfold s Grange Regression Esimaes Age RainH Maxemp TempD Humid Wind WindS Press p i 0.025-0.004 0.094-0.210 0.007 0.261-0.008 0.561 sa isic 9.92** 9.17** 4.99** 5.80** 3.12** 7.38** -7.63** 8.08** (b) Penfold s S Henri Regression Esimaes Age RainH Maxemp TempD Wind WindS p i 0.003-0.001 0.024-0.031 0.073-0.002 sa isic 2.36* 6.71** 2.74** 1.68 4.56** 4.25** (c) Penfold s Bin 707 Regression Esimaes Age AgeS RainH # Maxemp# MaxempS # TempD# Humid# Wind WindS Syle p 0.072-0.003 0.005 3.33-0.068-0.143 0.008 1.220-0.054 0.175 i sa isic 3.86* * 4.10** 5.23** 2.28* 2.27* 5.16** 2.52** 5.22** 5.05** 2.51* * (d) Henschke s Hill of Grace Regression Esimaes Age AgeS RainH SunS Temp Wind WindS SP Fridge Oak D p i 0.043-0.001-0.002 0.027-0.109 0.184-0.008 0.086 0.113 0.077 sa isic 3.96** 3.71** 6.17** 5.41** 4.39** 4.05** 4.02** 2.63** 3.06* 2.32**
Table 2: Aucion price dummies (a) Penfold s Grange 88 89 90 91 92 93 94 95 96 97 98 99 00 Coefficien 2.83 2.74 2.62 2.61 2.71 2.82 2.96 3.03 3.24 3.37 3.39 3.34 3.36 (b) Penfold s S Henri 94 95 96 97 98 99 00 Coefficien 0.69 0.75 0.81 0.90 0.94 0.96 0.92 (c) Penfold s Bin 707 96 97 98 99 00 Coefficien -42.36-42.09-41.93-41.95-41.92 (d) Henschke s Hill of Grace 95 96 97 98 99 00 Coefficien 1.22 1.33 1.41 1.46 1.46 1.44
5.56 5.48 5.4 5.32 5.24 5.16 5.08 ln aucion price 5 4.92 4.84 4.76 4.68 log fied pric log price 95% CI 95% CI 4.6 4.52 4.44 4.36 4.28 4.2 1971 1973 1975 1977 1979 1981vinage 1983 1985 1987 1989 1991 1993 Figure 1: Acual and fied aucion prices in 2000 for pas vinages of Henschke s Hill of Grace
Appendix 1: The four icon wines Penfolds Grange Penfolds Grange is Ausralia s mos-recognised icon wine. The firs vinage of Grange was released in 1951 by he lae Max Schuber. Grange is made predominaely from ulrapremium Shiraz grapes (85-100%), bu may also include a proporion of Caberne Sauvignon (up o 15%). I is a muli-disric blend, wih hree source regions: Barossa, Clare Valley and McLaren Vale. The percenage of grapes sourced from each region varies according o vinage condiions. Since is incepion, Grange has been maured in new American oak (wih he excepion of he 1957, 1958 and 1959 vinages which were produced in secre and maured in used barrels). Alhough Grange is released four years afer i is boled, i only begins o reach is peak afer 12-15 years. However, Grange has much longer erm cellaring poenial, wih he bes vinages sill a heir peak afer fory years. Penfolds S Henri John Davoren produced he firs vinage of Penfolds S Henri in 1956. Like Grange, i is primarily made from Shiraz (85-100%), bu can include up o 15% Caberne Sauvignon. The Shiraz grapes are sourced predominaely from Barossa Valley, wih smaller conribuions from McLaren Vale, Clare Valley and Eden Valley. The Caberne Sauvignon, when included, is sourced from Coonawarra and Barossa Valley. The wine is maured mosly in 2000 lire old oak vas called ovals. S Henri is a medium erm cellaring wine (generally peaking a around 12-15 years), alhough excepional vinages exhibi longer cellaring poenial. Penfolds Bin 707 Alhough firs released in 1964, Bin 707 was disconinued in 1969 bu hen reinroduced in 1976 and released on an annual basis (wih he excepion of 1981,) since hen. I is produced from Caberne Sauvignon grapes, he majoriy of which are sourced from Coonawarra bu he Barossa conribues up o one-hird. The wine is maured in new American oak, and exhibis medium o long-erm cellaring poenial. Henschke s Hill of Grace The Henschke s mos influenial icon wine, Hill of Grace, was pioneered by Cyril Henschke in 1958. Hill of Grace is produced from Shiraz grapes, from a single Keynon vineyard in he Eden Hills Region, from vines 70-130 years old. The wine is maured in new oak and has long-erm cellaring poenial.