Prce Determnants of Show Qualty Quarter Horses Mykel R. Taylor Kevn C. Dhuyvetter Terry L. Kastens Megan Doutht and Thomas L. Marsh* The authors would lke to thank Professonal Aucton Servces, Inc. for provdng aucton data used n ths analyss. Selected Paper prepared for presentaton at the Western Agrcultural Economcs Assocaton Annual Meetng, Honolulu, Hawa, June 30-July 2, 2004. Copyrght 2004 by Taylor, Dhuyvetter, Kastens, Doutht, and Marsh. All rghts reserved. Readers may make verbatm copes for noncommercal purposes by any means, provded that ths copyrght notce appears on all such copes. * The authors are Extenson Assstant, Professor, Assocate Professor, former Graduate Student, and Assocate Professor n the Department of Agrcultural Economcs at Kansas State Unversty, respectvely.
Prce Determnants of Show Qualty Quarter Horses Abstract Ths study estmates the prce determnants of show qualty Quarter Horses sold at aucton. Several characterstcs ncludng genetc and physcal trats, the qualty of pedgree, and sale order affect prce. Sale prce s postvely affected by a strong performance record of the horse as well as the performance record of the horse s offsprng. A common practce at horse auctons s for the seller to reject the fnal bd offered and not sell the horse. The market prces predcted by the model for these horses ndcate that they are not undervalued by the fnal bds, based on ther characterstcs. Keywords: aucton, censored regresson, hedonc model, Quarter Horses 2
Prce Determnants of Show Qualty Quarter Horses There has been lmted economc research pertanng to the show horse ndustry. Researchers typcally have overlooked the show horse ndustry n favor of the racehorse ndustry. An attracton to researchers regardng Thoroughbred and Quarter Horse racehorses s the amount of money spent on the gamblng aspect of the sport. However, the show horse ndustry also has a sgnfcant economc mpact on socety. There are over 6.9 mllon horses n the Unted States and 7.1 mllon people nvolved n the horse ndustry. Of the $25.3 bllon n total goods and servces drectly produced by the horse ndustry, horse showng contrbutes over 25 percent (Barents Group). Typcal expenses nclude money spent on the horse, tack, hotel, food, entry fees, gas, vehcles, and the general care of the horse. In 2003, the Amercan Quarter Horse Assocaton (AQHA) sanctoned over 2,500 horse shows. Ponts earned at AQHA sanctoned shows allow rders to qualfy for the World Show held each November n Oklahoma Cty, Oklahoma. One of the major events at the World Show s the World Champonshp Sale. Ths consgnment sale of AQHA show horses regularly grosses over $3,000,000 n sales (Table 1). Horses are entered n the World Champonshp Sale as consgned anmals by the seller. The seller pays a $400 entry fee and agrees to pay 8% of the fnal sale prce as a commsson to the aucton company. The seller s responsble for provdng nformaton on the horse to be sold to the aucton company for use n the sale catalog. Sale catalogs typcally nclude detaled nformaton on the horse s performance record, pedgree, and genetc characterstcs. In addton to the sale catalog, whch s avalable approxmately one month pror to the sale, buyers and seller have the opportunty to nteract pror to the sale n the barns and rdng arenas located at the World Show. Many buyers use the days pror to the sale to see prospectve horses and nqure about the horses from owners and traners. A common practce at many horse sales, ncludng race horses, s the practce of usng 3
reserve prces or buyng back horses. Dependng on the aucton company, a seller may ether enter a mnmum (reservaton) prce for the horse wth the auctoneer or buy the horse back from the sale rng by enterng the fnal bd. In ether case, the seller determnes a mnmum acceptable prce for the horse and does not have to sell the horse f bddng does not meet or exceed ths mnmum prce. The World Champonshp Show uses the buy back method and requres the seller to enter the fnal bd for ther horse f they do not want the horse to sell for the last bd offered by a buyer. In ths case, the horse s referred to as a no-sale horse and there s no transfer of ownershp. The seller, however, s stll requred to pay the 8% commsson on the fnal bd. The average number of no-sale horses at ths sale s 20% per year over the perod 1995 to 2002. The frst objectve of ths study s to quantfy the prce determnants of show-qualty Quarter Horses sold at publc aucton. The factors that affect show horse prces nclude genetc trats of the horse, pedgree, performance n the show rng, and economc condtons. A second objectve s to determne f there s a market neffcency that causes sellers to buy back ther horses as opposed to lettng them sell at the fnal bd prce. Lterature Revew Rosen s hedonc prcng model s based on the hypothess that goods are valued based on ther attrbutes. Hedonc models have been wdely used to evaluate the mplct prces of many agrcultural commodtes, especally lvestock. Baley and Peterson estmated factors affectng feeder cattle prces at vdeo and tradtonal auctons. Dhuyvetter, et al. and Chvosta, Rucker, and Watts estmated purebred beef bull determnants and Mntert, et al. analyzed factors affectng the prce for cull cows. Lansford, et al. used a sem-log hedonc prcng model to estmate the prce of ndvdual and ancestral characterstcs of yearlng Quarter Horses bred for racng. They noted that there has 4
been lttle research pertanng to genetc and ancestral characterstcs of Quarter Horses (.e., pedgree) despte vast record keepng of ancestral nformaton. The ancestral characterstcs of the yearlngs were descrbed by racng performance of the yearlng s sre and dam, as well as the racng performance of other offsprng of the sre and dam. Racng performance was descrbed as both number of races won and total race wnnngs. The authors concluded that several genetc and ancestral characterstcs nfluence the prce pad for race-bred yearlng Quarter Horses. Nebergs also used a sem-log hedonc prcng model to analyze Thoroughbred broodmare characterstcs. The characterstcs ncluded were descrbed as breedng, racng, genetc, and marketng factors. Breedng factors ncluded stud fee of coverng sre and earnngs of foals produced by the mare. Genetc factors n the model ncluded the racng record of sblngs and a qualty ndex for the mare s sre. The marketng factors consdered ncluded whether or not t was a dspersal sale and a bnary varable (RNA) f the horse faled to reach the reserve prce (.e., f t dd not sell). The model ndcated that horses that wn graded stakes races have the greatest purse earnng potental and the greatest value as a breedng prospect. The RNA bnary varable coeffcent was not statstcally sgnfcant and the author concluded that there s no evdence that the value of these no-sale horses justfes settng a reserve prce above the fnal bd. Hedonc Model Specfcaton The hedonc prcng functon used n ths study consders the nfluence of a vector of characterstcs of a horse on the sale prce at publc aucton. Sale prce s a functon of genetc and phenotypc (physcal) characterstcs, pedgree, performance, sale order, and economc condtons. Physcal characterstcs of a horse, such as conformaton, demeanor, and general appearance, are not easly recorded n a sale catalog and must be determned upon nspecton of the horse pror to or durng the sale. For that reason, many physcal characterstcs are not ncluded n the model. The general 5
specfcaton of the model s (1) ln[prce]=f(genetc and physcal trats, ndvdual performance, performance of offsprng, qualty of pedgree, sale order, year), where Prce s the sale prce of the horse and ln denotes natural logarthm. Genetc and physcal trats denotes a group of varables that descrbe the genetc makeup and physcal characterstcs of the horse ncludng age, color, sex, whether or not t s a bred mare (n foal), and the presence of genetc dseases. To allow for a nonlnear age effect by sex, age and age squared enter the emprcal model as nteracton terms wth sex (mare, stallon, or geldng). Ths allows for the dfferences n breedng potental between mares and stallons as well as the absence of breedng potental for geldngs. Age s expected to be postvely related to prce, but at a decreasng rate. Horse color s categorzed as bnary varables wth sorrel beng the default. There s no a pror expectaton of the effect of color on prce. A dummy varable for mares that are currently bred s ncluded. A bred mare s expected to brng a hgher value than a mare that s not currently n foal. A genetc dsease of concern to show horse owners and breeders s hyperkalemc perodc paralyss (HYPP). 1 Ths varable enters the model as a bnary varable nteracted wth the halter class bnary varable because the dsease s prmarly found n horses bred for halter classes. The nteracton term of halter and testng negatve for HYPP (n/n gene) was the default. Indvdual Performance s a group of varables that descrbe the show record of the horse beng sold. Each horse s classfed n one of fve prmary classes: western pleasure, hunter under saddle, halter, all-round (multple classes), or other (cuttng, regnng, or ropng). A bnary varable for each class s ncluded n the model wth the excepton of halter class whch s the default. There 1 HYPP s an nherted dsease of the muscle, whch s caused by a genetc defect. The gene occurs prmarly n horses bred for halter classes (where heavy musclng s desred) and can cause sudden paralyss or death n an anmal carryng the gene. Horses wll carry ether the n/n gene (no HYPP), the n/h gene (50 percent chance of passng on to offsprng), or the h/h gene (100 percent chance of passng HYPP on to offsprng). Testng for the gene has been requred on new foals by the AQHA snce 1998. 6
are no a pror expectatons for the class varables. Contnuous varables are ncluded for ponts earned at AQHA shows, ponts earned at non-aqha shows, number of World Show champonshps, number of World Show top placngs, number of futurtes won, and champonshps or placngs at non-aqha events. In addton to ponts earned at shows, horses can qualfy for awards based on the number of ponts earned n specfc events. Contnuous varables for number of regster of merts, whch requre ten ponts n a sngle event, and the number of superor ratngs, whch requre 50 ponts n a sngle event, are ncluded n the model. The varables for ponts and awards earned are expected to postvely nfluence a horse s value. Fnally, a bnary varable s ncluded for horses that are enrolled n or elgble for the AQHA Incentve Fund. If an ncentve fund horse wns at an AQHA show, the rder and owner wll receve a monetary award n addton to ponts. The expected sgn for ths varable s postve. Performance of offsprng denotes varables that descrbe the performance record of the offsprng of the horse beng sold. It ncludes contnuous varables for the number of offsprng that have earned AQHA ponts, won World Show champonshps, placed at the World Show, or won champonshps or placed at other events. Qualty of pedgree s a measure of the strength of a horse s lneage. Whle the sale catalogs provde detaled nformaton on the lneage of the horse, the strength of the pedgree s hard to determne wthout frst-hand knowledge of the reputaton of the varous sres and dams. Most breeders use rankngs of sres based on lfetme earnngs of offsprng to dstngush among the reputatons of varous sres. These rankngs are lsted by class (.e., western pleasure, hunter under saddle, and halter) and are ncluded for the sre of the horse, the sre of the horse s dam, and the servce sre s rankng. 2 Sre rankngs are calculated as both a contnuous varable of the actual rank and a bnary varable that equals one f the sre s ranked n the top 100 horses and zero otherwse. 2 Some of the mares sold at ths aucton are sold n foal or currently bred. The servce sre s the sre to whch the mare s bred. 7
Sale order s a contnuous varable correspondng to the order n whch the horses were sold at each year s sale. The horses are assgned a sale order or hp number by alphabetcal lstng of the frst dam s name wthn two groups, halter and all other performance horses. Therefore, the sale order varable s the random order n whch a horse was sold wthn ts group. To allow for a nonlnear effect by sale order, the contnuous varable enters the emprcal model as sale order and sale order squared. Due to the random sale order of the horses at the World Champonshp Sale, there are no pror expectatons as to the sgns of these varables. The varable Year represents year of sale and s modeled as a seres of bnary varables to capture the general effect of the overall economy (2001 s the default year). Varable names and descrptons are lsted n Table 2. Model Estmaton As mentoned prevously, some of the horses n the World Champonshp Sale dd not actually transfer ownershp due to the seller buyng back ther horses. For these no-sale horses, the sale catalog provdes all of the nformaton on the ndependent varables, but the only nformaton on prce s the fnal bd recorded. Although the fnal bd on a no-sale horse provdes some nformaton about the demand for that horse at aucton, t s not a market-clearng prce. The prce would have to be hgher than the fnal bd for a transfer of ownershp to occur, thus the fnal bds are essentally a censored value of the market-clearng prce. Nebergs estmated an OLS model usng the sale prce as the dependent varable for horses that sold and the fnal bd prce as the dependant varable for horses that dd not sell. To account for the horses that dd not sell, a bnary varable for the nosale horses (RNA) was ncluded n the model. However, usng an OLS model to estmate censored data wll generate based and nconsstent parameter estmates (Pndyck and Rubnfeld, pp. 325-327). Therefore, a censored regresson model allowng the no-sale observatons to be ncluded n 8
the dataset s a more approprate modelng approach. Cresp and Sexton used a censored regresson model to recover losng bds from an aucton for pens of fed cattle. Ther model allowed the censored value to adjust by observaton, rather than be set at a specfc value for all observatons. Followng ther model specfcaton, we assume that the natural log of the market clearng prce of horse * (ln ) P s determned by a vector of characterstcs, X, such that * ln 2 P = β Χ + ε. Further, and assumng that ε ~ N[0, σ ], the observed natural log of the sale prce s (1) ln P = ln P SP, f P * = P, and the censored natural log of the fnal bd s (2) ln P = ln P FB, f * P > P FB, where ln P s the natural log of the sale prce for horse (f the horse sold) and ln P FB s the SP natural log of the fnal bd for horse (f the horse dd not sell). For an observaton drawn randomly from the sample, whch may or may not be censored (Greene, p. 764), Χ β σ (3) E[ln P X ] =Φ ( Χ β + σλ ) where, (4) φ( Χ β / σ ) λ =. Φ( Χ β / σ ) Data Summary statstcs of the prces are reported n Table 1 and summary statstcs for the varables used n the model are lsted n Table 2. Sale prces and fnal bds were collected for the World Champonshp Sale from Professonal Aucton Servces, Inc., whch conducted the sale each of the years n the dataset. The sale data ncluded 3911 observatons from the tme perod 1995 to 2002. 9
Sx observatons were dropped because the horses dd not show up for the sale. Eght horses n the dataset were ranked themselves on the all-tme sre lst for one of the three classes. These horses were consdered outlers and were dropped from the dataset. Of the 3897 observatons remanng, 3093 horses sold and 804 were no-sale horses. Data on the top 100 sres ranked by lfetme earnngs of offsprng were collected for each sale year from Equ-Stat. The rankng data are assgned to each observaton based on the sale year. Ths s meant to reflect the current nformaton on sre rankngs avalable to buyers and sellers pror to the sale. All other data used n the model were collected from the sale catalogs for the respectve sale years. Results The hedonc prcng functon s modeled as a Tobt model and was estmated usng Lmdep. The coeffcent estmates and margnal effects of the Tobt model are shown n Table 3. The margnal effects of the model are ln P (5) = ρnon lm X [ ] β, where β s the vector of estmated coeffcents and ρ non lm s the probablty of an observaton not beng censored such that (6) ρ = { Φ [1 λ ( α + λ )] + φ ( α + λ )} non lm, where α = Χ β, Φ = Φ( α ), and λ = φ / Φ, wth Φ and φ denotng the cumulatve and densty functons, respectvely, of the standard normal dstrbuton. Greene (pp. 674-676) recommends computng the margnal effects at each observaton and reportng the sample average of the ndvdual margnal effects due to the nonlneartes of dscrete choce models. The sample average of the margnal effect for each parameter s reported n Table 3. The 10
nterpretaton of the margnal effect for each coeffcent s the proportonal change n prce for a one unt change n the parameter, gven that some sellers wll not sell ther horses. Takng nto account the probablty that a horse may not sell, the margnal effects are slghtly smaller n magntude than the estmated coeffcents of the Tobt model. The followng dscusson of the dfferent model varables s based on the margnal effects. Genetc and Physcal Characterstcs The coeffcents for age and age squared of mares (M*Age, M*Age2) and stallons (S*Age, S*Age2) are sgnfcant. The postve sgn on the lnear term and negatve sgn on the squared term ndcate that prce ncreases as mares and stallons get older, but at a decreasng rate. Fgure 1 shows the model predcted effect of age on market prce by sex (a more detaled dscusson of the model predcted prces s presented n the followng secton). The sgns of the coeffcents may be ndcatng that the value of mares and stallons ncreases as ther show careers progress, but wll eventually fall off when they are used only for breedng later n lfe. The coeffcents for Geldng and Stallon were both statstcally dfferent from zero and ndcate that mares receve a premum of 24.46 percent and 20.56 percent over geldngs and stallons, respectvely. All of the coeffcents for color were sgnfcant, except Chestnut, and had a postve sgn suggestng the default color (Sorrel) s less preferred to other colors. The coeffcent for Bred was not statstcally dfferent from zero. The model predcts that horses regstered n or elgble for the ncentve fund (Incentve) receve a premum of 6.90 percent over horses that are not elgble. Ths program allows rders and owners/breeders to receve money for ponts earned at AQHA shows. Therefore, the postve effect on the sale prce of a horse s expected. The only nteracton term between the halter class and the HYPP gene that was sgnfcant was the term descrbng a halter class horse that tested n/h for HYPP (see footnote 1). The margnal effect of the H*NH coeffcent ndcates that a halter horse wth the n/h gene wll brng 10.12 percent more than a halter horse that 11
tested negatve for the HYPP gene. Ths margnal effect may be the result of breeders or owners who contnue to take the rsk of a horse gettng HYPP n return for heaver musclng, whch s hghly valued n halter classes. Indvdual Performance Of the bnary varables denotng the prmary class of the horse, only the ClassOther coeffcent was sgnfcant. Horses n western pleasure, hunter under saddle, or some combnaton of these classes do not have a sgnfcant premum or dscount relatve to halter horses. The sgnfcant and postve sgn on the ClassOther varable ndcates the possblty of a dfferent set of buyers for the performance horses (cuttng, regnng, or ropng). Several of the ndvdual performance varables descrbng the horse s record were sgnfcant. Specfcally, the number of awards (ROM, Superor), the number of champonshps or top placngs at the World Show, and the number of futurty champonshps or placngs (WorldC, WorldP, Futurty) were sgnfcant and postve. An addtonal regster of mert ncreases the sale prce of a horse by 15.20 percent, whle an addtonal superor ratng ncreases sale prce by 14.88 percent. A World Show champonshp (top placng) ncreased prce by 8.63 (7.88) percent and wnnng or placng at a futurty ncreased prce by 7.97 percent. These margnal effects ndcate that the show record of a horse postvely mpacts t value. Performance of Offsprng All of the varables measurng the performance of the horses offsprng (f they had any) were postve and sgnfcant. A horse havng an addtonal offsprng that has earned AQHA ponts ncreases the sale prce by 5.04 percent, whle an addtonal offsprng that has a World Show champonshp (top placng) ncreases sale prce by 7.10 (3.26) percent. Each offsprng that has receved an award (regster of mert, superor ratng) or won a champonshp at a futurty or other event ncreases the sale prce of a horse by 2.50 percent. 12
Qualty of Pedgree The rankng of a horse s sre was broken out by class: western pleasure, hunter under saddle, and halter. For western pleasure, a sre ranked n the top 100 (SreWPRankBV) adds 28.28 percent to the sale prce. The contnuous varable for sre rank (SreWPRank) ndcates that the sale prce falls by 0.23 percent for a one unt ncrease n rank (the best rank possble s 1 and the worst s 100). Ths relatonshp ndcates that the premum of havng a ranked sre n western pleasure s reduced to almost 5.28 percent from 28.28 percent as the level at whch the sre s ranked falls from 1 to 100. For hunter under saddle, the bnary varable (SreHUSRankBV) was sgnfcant and added 12.65 percent to the sale prce. The contnuous varable was not statstcally sgnfcant. For horses wth sres ranked n the halter class (SreHALTRankBV), the added value s 13.10 percent. The contnuous varable (SreHALTRank) ndcates that the premum from havng a ranked sre n halter s decreased by 0.13 percent for each declne n rank from 1 to 100. Ths relatonshp ndcates that the premum s reduced to less than 1 percent as the level at whch the sre s ranked falls to 100. Fgure 2 shows the change n the predcted premum for a ranked sre n western pleasure or halter classes as the rank declnes from 1 to 100. For horses whose dam s sre was ranked n the western pleasure class (DSreWPRankBV), the sale prce s ncreased by 10.34 percent. For a dam s sre ranked n the halter class (DSreHALTRankBV), the sale prce s ncreased by 10.14 percent. The other varables for dam s sre rankng were not statstcally dfferent from zero. For bred mares wth servce sres that were ranked n the western pleasure class (SSreWPRankBV), the sale prce s ncreased by 22.66 percent and declnes by 0.41 percent for each fall n rank from 1 to 100 (SSreWPRank). The premum from beng ranked n western pleasure s reduced to zero by the 55 th ranked horse and a negatve 18.34 percent at the 100 th rank. The rankng of a servce sre n the hunter under saddle class (SSreHUSRank, SSreWPRankBV) 13
was not sgnfcantly dfferent from zero. For the mares wth servce sres ranked n the halter class (SSreHALTRankBV), the sale prce s 41.30 percent hgher and the prce declnes by 0.40 percent for each fall n rank from 1 to 100 (SSreHALTRank). The premum s reduced to 1.3 percent for a servce sre n the halter class at the last rankng (100 th ). Sale Order The coeffcent for sale order by class (SOClass) was statstcally sgnfcant and postve. The coeffenct for sale order by class squared (SOClass2) was sgnfcant and negatve. The postve sgn on the lnear term and negatve sgn on the squared term ndcate that prce ncreases the farther nto a sale a horse s sold, but at a decreasng rate. Ths quadratc relatonshp may descrbe the change n atttude of buyers over the duraton of the sale. Fgure 3 presents the effect of sale order on prce. Year The bnary varables for year were ncluded to account for general economc condtons. The coeffcent for 1999 (Year1999) was sgnfcant and postve. The base year for comparson s 2001, mplyng that horses wth dentcal characterstcs sold for 11.17 percent more n 1999 than n 2001. The coeffcent for 2002 (Year2002) was also sgnfcant, but the sgn ndcates that horses sold n 2002 went for 9.47 percent less than an dentcal horse sold n 2001. All other year coeffcents were not sgnfcant n explanng the varaton n prce for horses sold. Predctng Sales Prces Whle t s approprate to evaluate ndvdual characterstcs for show horses usng the margnal effects from the Tobt model, ths may not be the best model to use for predctng sale prces. To predct the sale prce of a horse, we used the parameter estmates β from the Tobt model (.e., lnˆ[ P ] = Χ ), whch assumes that any random horse selected wll sell (.e., the uncensored model). β 14
The ft of the model s descrbed by calculatng a correlaton coeffcent for the natural log of the sale prces of horses that sold wth the natural log of ther predcted values. Ths coeffcent cannot nclude the no-sale horses because there s no observed sale prce to use as a comparson. The correlaton coeffcent s 0.294 for the log prces of the sale horses. The average dfference between the log of the observed sale prce and the predcted log sale prce for a horse that sold was negatve 0.073. The average dfference between the log of the observed fnal bd and the predcted log sale prce for no-sale horses was negatve 0.007. Ths means that, on average, the predcted sale prce for a no-sale horse was 0.7 percent hgher than the fnal bd. Table 4 lsts the summary statstcs for the predctve model for both sale and no-sale horses. It may be easer to understand these results f the predcted log prces are transformed to prce for comparson to the fnal sale bds. Due to bas n the detransformaton of a sem-log lnear model, an adjustment s appled to the transformaton (Mller). The transformaton s as follows (7) ˆ βχ 2 0.5 ˆ σ E(lnˆ P ) = e e, 2 where ˆ σ s the model root mean squared error. Once the predcted log sale prces for the sale and no-sale horses are transformed, the average dfference between the observed sale prce (fnal bd) and the predcted sale prce s negatve $478.57 for sale horses and negatve $552.49 for no-sale horses. The percentage of predcted sale prces that are above the observed sale prce s 69.9 percent for sale horses. When usng the model to predct the sale prces of no-sale horses, the percentage of predcted prces that are hgher than the fnal bd s a comparable 67.0 percent. The relatvely small dfference n the percentages of predcted prces that are hgher than the observed prce (fnal bd) suggests that nosale horses are not consstently undervalued by the fnal bd, based on ther characterstcs. The results also ndcate that whether or not a horse sells at aucton appears to be a random event. There are several explanatons for why sellers may choose not to sell ther horses at aucton. 15
Some sellers may have nformaton on the horses expected show or breedng performance that s dffcult to express to potental buyers through the catalog or pre-sale vewng. Ths neffcency n the flow of nformaton could cause buyers to undervalue a horse relatve to the seller s reservaton prce. Another possble explanaton for no-sale horses s overvaluaton by sellers. The seller may smply gnore the market sgnals from buyers at the aucton and decde the horse s too valuable to sell at the fnal bd prce. Conclusons Knowng how ndvdual characterstcs of horses, rangng from genetc characterstcs to performance dscplne to pedgree, mpact prces s crtcal nformaton for both buyers and sellers of Quarter Horses. Buyers desre ths nformaton so they can make nformed purchase decsons possbly reducng the rsk assocated wth ther nvestments. Lkewse, sellers desre ths nformaton so they can make breedng and show decsons to capture the trats most demanded by buyers. Several of the genetc trats, ncludng age, color, and sex mpacted sale prce. For mares and stallons, the postve relatonshp between age and prce declnes as the horse ages. The coeffcents on sex revealed that mares receve a premum relatve to both geldngs and stallons. Ths lkely s due to both ther breedng potental, as compared to geldngs, and ther tendency to be easer to handle n the show rng after they have started ther breedng career. Stallons tend to be much harder to work wth after ther breedng lfe has begun. Each of the statstcally sgnfcant varables measurng a horse s performance postvely mpacted sale prce. Ths ndcates that horses wth dstngushed show records are valuable as show horses and possbly as breedng anmals. Enrollment n or elgblty for the AQHA Incentve Fund also ncreases the sale prces of horses. 16
The postve effect of the performance of offsprng and the rankng of sres, dams sres, or servce sres all ndcate that a strong pedgree s valuable for show horses. Pedgree s lkely to be a sgnfcant factor n many breedng programs because t s a valuable trat desred by buyers n the market. Sale order does affect the sale prce of horses, wth horses sellng at the begnnng and end of the sale recevng a slght dscount relatve to horses sold n between. Although horses are consdered a luxury good and expendtures n the horses ndustry may be affected by the condton of the economy, the bnary varables used for each sale year were generally not statstcally sgnfcant. In addton to understandng the ndvdual characterstcs that affect show horse value, ths model also allowed the predcton of market-clearng prces for the no-sale horses. The results suggest that no-sale horses are not undervalued by the fnal bd at aucton and that whether or not a horse sells appears to be a random event. Some explanatons for why sellers may choose not to sell ther horse at aucton nclude neffcency n the flow of nformaton between buyers and sellers or overvaluaton of the horse by the seller. Future research wll address the possble neffcences n the flow of nformaton regardng the characterstcs of no-sale versus sale horses. Ths wll allow further nvestgaton nto the practce of the no-sale horses at auctons for show-qualty Quarter Horses. 17
References Baley, D. and M.C. Peterson. A Comparson of Prcng Structures at Vdeo and Tradtonal Cattle Auctons. Western Journal of Agrcultural Economcs 16(December,1991):392-403. Barents Group, LLC. The Economc Impact of the Horse Industry n the Unted States. Techncal report prepared for The Amercan Horse Councl Foundaton, December, Washngton DC, 1996. Chvosta, J., R.R. Rucker, and M.J. Watts. Transacton Costs and Cattle Marketng: The Informaton Content of Seller-Provded Presale Data at Bull Auctons. Amercan Journal of Agrcultural Economcs. 83(May, 2001):286-301. Cresp, J.M., and R.J. Sexton. Bddng for Cattle n the Texas Panhandle. Amercan Journal of Agrcultural Economcs. 86(August, 2004):660-674. Dhuyvetter, K.C., T.C. Schroeder, D.D. Smms, R.P. Bolze, Jr., and J. Geske. Determnants of Purebred Beef Bull Prce Dfferentals. Journal of Agrcultural and Resource Economcs. 21(December, 1996):396-410. Equ-Stat. Fort Worth, Texas. Personal communcaton wth Donna Tolson, June, 2004. Greene, W.H. Econometrc Analyss, 5th ed. Upper Saddle Rver, New Jersey: Prentce Hall, 2003. Lansford, Jr., N.H., D.W. Freeman, D.R. Toplff, and Odell L. Walker. Hedonc Prcng of Race- Bred Yearlng Quarter Horses Produced by Quarter Horse Sres and Dams. Journal of Agrbusness. 16(Fall, 1998):169-185. Lmdep Reference Gude Verson 8.0. Planvew, New York: Econometrc Software, Inc, 2002. Mller, D.M. Reducng Transformaton Bas n Curve Fttng. The Amercan Statstcan. 38(May, 1984):124-126. Mntert, J., J. Blar, T. Schroeder, and F. Brazle. Analyss of Factors Affectng Cow Aucton Prce Dfferentals. Southern Journal of Agrcultural Economcs. 22(December, 1990):23-30. Nebergs, J.S. A Hedonc Prce Analyss of Thoroughbred Broodmare Characterstcs. Agrbusness. 17(Sprng, 2001):299-314. Pndyck, R.S., and D.L. Rubnfeld. Econometrc Models and Economc Forecasts, 4th ed. Boston: Irwn McGraw-Hll, 1997. Professonal Aucton Servces, Inc. Leesburg, Vrgna. Sale data from World Champonshp Show, 1995 to 2002. Rosen, S. Hedonc Prces and Implct Markets: Product Dfferentaton n Pure Competton. Journal of Poltcal Economy. 82(Jan.-Feb., 1974):34-55. 18
Table 1. Summary Statstcs of AQHA World Champonshp Sale Year Sale Horses 2002 2001 2000 1999 1998 1997 1996 1995 No-Sale Horses 2002 2001 2000 1999 1998 1997 1996 1995 Sale Prce Gross Average Mnmum Maxmum Count % Sold $2,993,850 $7,338 $800 $170,000 408 87.0% $3,128,400 $8,063 $700 $77,000 388 79.6% $3,424,200 $8,291 $900 $85,000 413 77.9% $3,289,700 $8,328 $550 $90,000 395 78.3% $3,214,500 $8,159 $750 $77,000 394 78.5% $3,193,325 $8,084 $1,000 $73,500 395 80.5% $2,769,650 $6,959 $700 $145,000 398 75.9% $1,792,200 $5,934 $800 $45,000 302 79.5% n/a $6,969 $500 $58,000 105 13.0% n/a $9,393 $1,100 $103,000 123 20.4% n/a $9,583 $1,500 $49,000 100 22.1% n/a $7,539 $1,400 $26,500 108 21.7% n/a $7,787 $1,400 $27,000 109 21.5% n/a $10,614 $1,200 $80,000 112 19.5% n/a $6,581 $900 $50,000 102 24.1% n/a $5,200 $1,000 $14,500 45 20.5% 19
Table 2. Varable Descrptons and Summary Statstcs a Varable Name Descrpton Mean Standard Devaton Mnmum Value Maxmum Value lnp Log of sale prce (fnal bd prce) 8.63 0.77 6.21 12.04 Geldng Bnary varable equal to 1 f horse s a 0.14 0.35 0 1 geldng Mare Bnary varable equal to 1 f horse s a mare 0.70 0.46 0 1 Stallon Bnary varable equal to 1 f horse s a 0.16 0.37 0 1 stallon Age Age of horse 4.61 4.64 0 23 G*Age Geldng and age nteracton term 0.31 1.08 0 14 G*Age2 Geldng and age squared nteracton term 1.26 8.07 0 196 M*Age Mare and age nteracton term 3.96 4.90 0 23 M*Age2 Mare and age squared nteracton term 39.68 73.65 0 529 S*Age Stallon and age nteracton term 0.35 1.31 0 15 S*Age2 Stallon and age squared nteracton term 1.83 12.20 0 225 b Color Bnary varable for color of horse -- -- 0 1 Bred Bnary varable equal to 1 f horse s 0.38 0.49 0 1 marketed as breedng stock Incentve Enrolled n or elgble for AQHA Incentve 0.70 0.46 0 1 Fund H*NoTest Halter class and horse not tested for HYPP 0.07 0.26 0 1 nteracton term H*NN Halter class and horse s homozygous 0.23 0.42 0 1 negatve for HYPP nteracton term H*NH Halter class and horse s heterozygous for 0.11 0.32 0 1 HYPP nteracton term H*HH Halter class and horse s homozygous 0.00 0.03 0 1 postve for HYPP nteracton term Halter Halter class 0.41 0.49 0 1 HUS Hunter under saddle class 0.11 0.32 0 1 WP Western pleasure class 0.36 0.48 0 1 Allaround One or more classes 0.07 0.26 0 1 ClassOther Other class 0.04 0.19 0 1 Ponts AQHA ponts earned n lfetme 13.82 45.96 0 837 NonPonts Non-AQHA ponts earned n lfetme 1.19 25.40 0 915 ROM Regster of mert 0.19 0.57 0 5 Superor Superor ratng 0.08 0.38 0 5 WorldC AQHA World Show champon 0.05 0.41 0 12 WorldP AQHA World Show placng 0.15 0.73 0 10 Futurty Champonshp or placng at AQHA futurty 0.14 0.69 0 9 NonCP Champonshp or placng at non-aqha 0.02 0.18 0 6 show OffsprngP Offsprng that have won ponts 0.34 1.19 0 20 OffsprngWC Offsprng wth AQHA World Show 0.04 0.52 0 15 champonshp OffsprngWP Offsprng wth AQHA World Show placng 0.12 0.73 0 13 OffsprngOther Offsprng wth ROM, SUP, or futurty champonshp or placng 0.30 1.31 0 25 a Total sample sze n=3,897 b Color categores are Bay, Black, Brown, Chestnut, Gray, Palomno, Redroan, Sorrel, and ColorOther 20
Table 2. Varable Descrptons and Summary Statstcs, cont. Varable Name Descrpton Mean Standard Devaton Mnmum Value Maxmum Value SreWPRank Rank of sre for western pleasure 9.73 20.72 0 100 SreWPRankBV Bnary varable equal to 1 f sre s ranked for 0.33 0.47 0 1 western pleasure SreHUSRank Rank of sre for hunter under saddle 7.37 20.50 0 99 SreHUSRankBV Bnary varable equal to 1 f sre s ranked for 0.18 0.38 0 1 hunter under saddle SreHALTRank Rank of sre for halter 5.68 16.19 0 100 SreHALTRankBV Bnary varable equal to 1 f sre s ranked for 0.25 0.43 0 1 halter DSreWPRank Rank of dam's sre for western pleasure 3.62 13.60 0 99 DSreWPRankBV Bnary varable equal to 1 f dam's sre s 0.13 0.33 0 1 ranked for western pleasure DSreHUSRank Rank of dam's sre for hunter under saddle 3.19 13.91 0 97 DSreHUSRankBV Bnary varable equal to 1 f dam's sre s 0.07 0.25 0 1 ranked for hunter under saddle DSreHALTRank Rank of dam's sre for halter 1.63 9.79 0 100 DSreHALTRankBV Bnary varable equal to 1 f dam's sre s 0.07 0.25 0 1 ranked for halter SSreWPRank Rank of servce sre for western pleasure 2.29 10.55 0 99 SSreWPRankBV Bnary varable equal to 1 f servce sre s 0.08 0.27 0 1 ranked for western pleasure SSreHUSRank Rank of servce sre for hunter under saddle 1.78 10.29 0 97 SSreHUSRankBV Bnary varable equal to 1 f servce sre s 0.04 0.20 0 1 ranked for hunter under saddle SSreHALTRank Rank of servce sre for halter 3.21 12.07 0 98 SSreHALTRankBV Bnary varable equal to 1 f servce sre s 0.13 0.33 0 1 ranked for halter SOClass Sale order wthn class 129 80 1 327 SOClass2 Sale order wthn class, squred 22,952 23,624 1 106,929 YEAR Bnary varable for each sale year -- -- 0 1 21
Table 3. Hedonc Model Regresson Results Varables Parameter Estmate Standard Error t-statstc p-value Average Margnal Effect Constant 8.0163 0.0678 118.1770 0.0000 7.1735 G*Age 0.0211 0.0411 0.5130 0.6077 0.0189 G*Age2-0.0032 0.0042-0.7660 0.4437-0.0029 M*Age 0.0781 0.0133 5.8920 0.0000 0.0699 M*Age2-0.0057 0.0007-8.4390 0.0000-0.0051 S*Age 0.1431 0.0318 4.5000 0.0000 0.1280 S*Age2-0.0123 0.0030-4.1490 0.0000-0.0110 Geldng -0.2734 0.0689-3.9700 0.0001-0.2446 Stallon -0.2297 0.0541-4.2460 0.0000-0.2056 Bay 0.0775 0.0310 2.4960 0.0126 0.0693 Black 0.3302 0.0610 5.4140 0.0000 0.2955 Brown 0.2411 0.0517 4.6590 0.0000 0.2157 Chestnut -0.0091 0.0316-0.2880 0.7732-0.0082 Gray 0.3070 0.0529 5.8000 0.0000 0.2747 Palomno 0.1818 0.0864 2.1030 0.0355 0.1627 Redroan 0.1970 0.0858 2.2970 0.0216 0.1763 ColorOther 0.1481 0.0682 2.1730 0.0298 0.1326 Bred 0.0689 0.0493 1.3990 0.1618 0.0617 Incentve 0.0771 0.0296 2.6080 0.0091 0.0690 H*NoTest -0.0325 0.0478-0.6790 0.4974-0.0290 H*NH 0.1131 0.0403 2.8080 0.0050 0.1012 H*HH 0.0278 0.3909 0.0710 0.9433 0.0249 HUS 0.0352 0.0506 0.6950 0.4870 0.0315 WP 0.0089 0.0443 0.2010 0.8407 0.0080 Allaround -0.0678 0.0525-1.2900 0.1970-0.0607 ClassOther 0.4168 0.0656 6.3530 0.0000 0.3730 Ponts 0.0007 0.0006 1.1570 0.2474 0.0007 NonPonts 0.0002 0.0005 0.4890 0.6246 0.0002 ROM 0.1698 0.0249 6.8310 0.0000 0.1520 Superor 0.1662 0.0607 2.7400 0.0061 0.1488 WorldC 0.0965 0.0397 2.4300 0.0151 0.0863 WorldP 0.0881 0.0195 4.5090 0.0000 0.0788 Futurty 0.0890 0.0172 5.1830 0.0000 0.0797 NonCP 0.0857 0.0742 1.1550 0.2483 0.0767 OffsprngP 0.0563 0.0169 3.3340 0.0009 0.0504 OffsprngWC 0.0793 0.0241 3.2980 0.0010 0.0710 OffsprngWP 0.0365 0.0199 1.8290 0.0673 0.0326 OffsprngOther 0.0279 0.0143 1.9590 0.0502 0.0250 SreWPRank -0.0026 0.0008-3.3500 0.0008-0.0023 SreWPRankBV 0.3161 0.0411 7.6840 0.0000 0.2828 SreHUSRank 0.0002 0.0009 0.2100 0.8340 0.0002 SreHUSRankBV 0.1413 0.0506 2.7930 0.0052 0.1265 SreHALTRank -0.0014 0.0009-1.6780 0.0933-0.0013 SreHALTRankBV 0.1464 0.0395 3.7060 0.0002 0.1310 22
Table 3. Hedonc Model Regresson Results, cont. Varables Parameter Estmate Standard Error t-statstc p-value Average Margnal Effect DSreWPRank -0.0005 0.0012-0.4550 0.6489-0.0005 DSreWPRankBV 0.1155 0.0553 2.0880 0.0368 0.1034 DSreHUSRank -0.0014 0.0016-0.8700 0.3845-0.0013 DSreHUSRankBV 0.0644 0.0965 0.6680 0.5041 0.0577 DSreHALTRank -0.0011 0.0014-0.8090 0.4187-0.0010 DSreHALTRankBV 0.1134 0.0593 1.9130 0.0557 0.1014 SSreWPRank -0.0046 0.0017-2.6400 0.0083-0.0041 SSreWPRankBV 0.2533 0.0862 2.9370 0.0033 0.2266 SSreHUSRank -0.0004 0.0018-0.2170 0.8284-0.0004 SSreHUSRankBV -0.0605 0.1040-0.5820 0.5608-0.0542 SSreHALTRank -0.0044 0.0013-3.5080 0.0005-0.0040 SSreHALTRankBV 0.4615 0.0515 8.9600 0.0000 0.4130 SOClass 0.0016 0.0005 3.3220 0.0009 0.0015 SOClass2-0.000004 0.0000-2.3020 0.0214 0.0000 Year1995-0.0487 0.0492-0.9920 0.3214-0.0436 Year1996-0.0380 0.0439-0.8660 0.3866-0.0340 Year1997 0.0680 0.0435 1.5650 0.1177 0.0609 Year1998 0.0494 0.0437 1.1300 0.2586 0.0442 Year1999 0.1248 0.0438 2.8490 0.0044 0.1117 Year2000 0.0283 0.0433 0.6530 0.5140 0.0253 Year2002-0.1058 0.0432-2.4480 0.0144-0.0947 23
Table 4. Summary Statstcs of Predcted Market Prces Standard Devaton Mnmum Value Maxmum Value Average Sale Horses lnp - l nˆ P -0.073 0.65-2.12 2.75 P - Pˆ -$478.57 $7,846.22 -$46,876.28 $155,214.37 RMSE 7859.54 % predcted prces above sale prce 69.87% No-Sale Horses lnp - l nˆ P -0.0069 0.60-2.02 2.30 P - Pˆ -$552.49 $7,014.18 -$54,431.83 $60,892.98 RMSE 7031.55 % predcted prces above fnal bd 67.04% 24
10000 8000 Geldng Prce, $/head 6000 4000 2000 Stallon Mare 0 0 5 10 15 20 Age, years Fgure 1. Model Predcted Effect of Age on Market Prce by Sex (all other characterstcs evaluated at the mean of the seres for geldng, stallon, and mare). 2500 Premum, $/head 2000 1500 1000 500 Halter Western Pleasure 0 0 20 40 60 80 100 Rank Fgure 2. Model Predcted Premum for a Ranked Sre by Class (all other characterstcs evaluated at the mean of the seres for western pleasure and halter). 25
9000 Prce, $/head 8000 7000 6000 5000 0 50 100 150 200 250 300 Sale Order Fgure 3. Effect of Sale Order on Sale Prce. 26