University of Toronto



Category expectations, category spanning and market outcomesAnne BowersRotman School of ManagementUniversity of Toronto105 St. George StreetToronto, Ontario, Canada M5S 3E6Anne.Bowers@rotman.utoronto.caForthcoming, Advances in Strategic ManagementAbstract: The growth of research on the cognitive origins of market performance has focused on the impact of categories as a primary cognitive mechanism by which exchange occurs. In this research, performance outcomes are typically reduced when firms and products fail to meet audiences’ expectations about membership into categories. The ensuing literature has focused on spanning categories as evidence of not meeting audience expectations while largely ignoring the specific study of expectations themselves. This paper argues that expectations for market behavior are important in their own right, and can impact market outcomes even when categorical boundaries are respected. Using the market for engagement rings as a setting, I show how lack of adherence to expectations can both increase and decrease market value even as the engagement rings adhere to categorical boundaries. Rather than simply focusing on category spanning as evidence that audience expectations have not been met, the findings suggest that expectations should be considered explicitly, with implications for competitive strategy. A fundamental way of processing information by individuals is via the use of categories. At their simplest, categories are socially determined groups of attributes that bind some objects together while separating them from other, unrelated objects (Zerubavel 1991). Categories help answer the question, “What is this?” through cognitive simplification: once an audience knows what category an object belongs to, they can then have expectations for what it does. Categories occur in all facets of human cognition, but have gained traction in recent research as a way of explaining market behavior and performance outcomes. Firms use categories to develop strategies for competitive behavior in markets (Porac, Thomas and Baden Fuller 1989, Zuckerman 2004), and potential buyers and intermediaries also use categories to help them evaluate market products (e.g. Zuckerman 1999). A chief tenet of the categories approach in markets is that a product or firm’s market identity is based in part by the understanding of which category it is associated with, and, importantly, the underlying expectations for market behavior (Hsu 2006). Category membership is not binary, however, as objects can have partial membership in multiple categories, such that within a given category, members fit into the category to a greater or lesser degree (Hannan 2010). For example, Hannan (2010) discusses how audiences expect anything called a circus to be a traveling act with tents, animals and clowns, but notes that traveling acts with clowns but no animals are like circuses, but also like theatre. Since audience members use categories in order to evaluate and understand objects, knowledge of which category an object belongs to is meaningful. To continue the example, a family might have different expectations (what kinds of spectacle, availability of food, suitability for children) based on whether they believed they were going to a circus or the theater, and these in turn influence the value they assign to particular events. This insight has led to a focus on category membership as a critical dimension along which performance should be examined, and, for firms, an area in which they should focus strategic efforts (Porac, Wade, and Pollock 1999, Kennedy 2008, Fleischer 2009).The literature on category membership has established that those that span categories often suffer performance penalties. This finding is extremely robust, having been reaffirmed in wine, the stock market, movies, employment histories, software, the arms industry, credit ratings and eBay auctions. For example, multiple category membership (and thus atypical partial membership in a specific category) leads to lower evaluations for movies, and eBay sellers that sell in multiple categories receive less for their items (Hsu 2006, Hsu, Hannan and Kocak 2009). Conversely, focused employment histories and single varietal wines are associated with better outcomes (Leung 2014, Negro and Leung 2013). The penalties occur because “products that incorporate features from multiple categories are perceived to be poor fits with category expectations and less appealing than category specialists” (Hsu, Hannan and Kocak 2009, p. 150). That is, the spanning of categories is taken as evidence of violation of category expectations.Although such findings use the violation of expectations as the mechanism leading to reduced performance, they do so without examining which specific expectations are not being met. To demonstrate empirical evidence of the multiple category discount, researchers use mathematical methods of measuring category overlap for a given firm. For example, Leung and Sharkey (2013) use a simple count of the number of categories to determine whether a loan belongs to multiple categories. Zuckerman (2004) uses the average proportional overlap of analyst coverage to show categorical membership among publicly traded firms (also used by Bowers 2014 and Tan and Roberts 2010). Those with low scores are covered by analysts whose portfolios contain few similar stocks, and thus span categories. Hsu (2006) uses the Jaccard coefficient of similarity as a basis for measurement of genre overlap in movies, which is defined as the intersection divided by the size of the union of pairs of movies. In each case, the measure captures the degree to which an object fits into a category based on its similarity to other objects. All of these measures are then used to investigate the impact on some measure of market value or performance, such as appeal, price, revenues, or survival, with the underlying mechanism being the violation of category expectations.The value of the mathematical approach is that it creates a precise measure that allows for partiality of membership to be determined for a large number of categories simultaneously, as in the stock market or software development. With a single number generated for each firm or product, membership in any category is commensurate with membership in any other category. For example, Zuckerman’s (2004) measure captures the extent to which a stock is followed by analysts who all cover the same stocks, but equates a firm that is covered by media as well as tobacco analysts similarly to a stock that is covered by technology and utility analysts. Since none of these measures address the specific category combinations that an object belongs to, the measures do not capture which violations of expectations, if any, are occurring. That is, the measures do not capture whether it is the lack of suitability for children that makes audiences dislike circuses that are also like theatre, or if it is the disappointment of expecting an animal act and not seeing one, or if it is something else entirely. Instead, the assumption is that being a member of more than one a category, no matter what category, implicitly violates expectations of audience members, although which expectations are not defined. Logically, then, the converse should also be true: a full category member should violate none of the expectations for category membership, and presumably, audiences find it agreeable because it acts as they expect. If both of these conditions are true, then simply measuring the spanning of categories is sufficient to establish evidence of the relationship between the failure to meet category expectations and the resultant decline in market value, and the direct study of expectations is unnecessary. Firm strategies, in turn, should focus on understanding the nature of category spanning and the positioning of competitors across categories within a market. On the other hand, should objects that span multiple categories adhere to the underlying expectations, or if the violation of expectations is possible without spanning categories, then the practice of focusing only on category spanning as evidence of violation of expectations yields incomplete insight into the cognitive aspects of market behavior and the strategies available to firms as a result.This paper addresses the latter case, that violation of expectations can occur without spanning categories, to argue that expectations for market behavior are important in their own right, and can impact market outcomes even when categorical boundaries are respected. To make an argument for studying expectations explicitly, in the sections that follow I develop an empirical case study of the market for engagement rings. I dispute the notion that full membership into a category automatically means no violation of category expectations by demonstrating that some full members of the engagement ring category can have differing market value based on whether or not they meet audience expectations. By restricting my analysis solely to those members of a single category, I effectively hold the category constant while examining the difference in expectations, allowing for a clean test of the expectations themselves on market value. Using archival data from eBay as well as a controlled laboratory experiment, I then show how violating expectations can both increase and decrease market value depending on the measure of market value used. Empirically, the findings provide insight into the market for diamond rings. Although well studied as an example of cartels or discontinuous pricing (e.g. Scott and Yelowitz 2010), the expectations underlying such a market have yet to be analyzed. The findings suggest that competition in a market dominated by social beliefs is not straightforward. Theoretically, the findings highlight that forgoing explicit studies of category expectations in favor of category spanning studies ignores how cognitive perceptions shape market outcomes. This in turn limits understandings of potential market strategies and competitive actions available to firms in a given category. Because categories are particularly meaningful in facilitating exchange, understanding how expectations shape markets should lead to opportunities for both buyers and sellers.THE MARKET FOR ENGAGEMENT RINGSThe first step in establishing that audience expectations matter independently of category membership requires identifying a category and its membership requirements and then demonstrating that full members of the category can violate audience expectations. In this section, I establish diamond solitaire rings as full members of the engagement ring category and provide evidence that some diamond solitaire rings can violate audience expectations for engagement rings while still being clearly identified as members of the engagement ring category.At its simplest, the category of engagement rings comprises any ring that is given to a future bride to mark her commitment to be married. Engagement rings were originally the privilege of the wealthy, but when changes in manufacturing techniques and the discovery of additional diamond mines occurred during the late part of the 19th century, diamonds become more accessible, prompting a push toward matrimonial jewelry (Brining 1990). In fact, engagement rings were originally touted for both men and women (Howard 2008), although the male tradition did not catch on. Many observers of the modern day engagement ring market note its start coincident with DeBeers’ 1947 advertising campaign “A diamond is forever,” although research suggests that the market had begun to expand well before then (Brinig 1990, Koskoff 1981, Trex 2010). The original DeBeers advertisting images, echoing advertisements since the 1930s, touted a single diamond as an expression of undying love. As a result, while any ring could be an engagement ring, the diamond solitaire ring became synonymous with the notion of an engagement ring. By the 1950s, diamond engagement rings were the standard for brides (Brinig 1990, Epstein 1982) and by 1965, 80% of brides had a diamond engagement ring, a trend which continues today (O’Rourke 2007). Such rings consistently occupy prominent places in engagement ring buying guides and message boards (Trex 2010). In fact, the diamond solitaire ring is so ubiquitous that if asked what an engagement ring is, most would recall the characteristics of a diamond solitaire ring, as, aside from the white dress worn by a bride, it is one of the most enduring symbols of marriage (Howard 2008). When a particular object is so identified with a category that it becomes synonymous with that category, that object is considered a prototype (Waguespack and Sorenson 2012, Durand and Paolella 2013, Rosch 1975). Prototypes contain all of the attributes of a single category, so much so that their attributes set the standard for understanding the category. In the case of diamond solitaire rings, the critical attributes are a metal band, typically gold or platinum, and a single diamond. All rings that contain these physical features are distinctly identified with the engagement ring category and no other category (such as wedding rings). Within a category, the value of a prototype can be determined not by the presence of its attributes, since it is synonymous with those attributes, but rather the levels of them. For diamond solitaire rings, the criteria for value are well known, developed and standardized in the 1950s by the Gemological Institute of America, and echoed in advertising for jewelry stores and buying guides that are published for consumers. Diamonds are differentiated based on cut, color, clarity, and weight (or carats)—commonly known as the “4 Cs” (Rapaport 2012). Highly valued diamonds are colorless and have few inclusions (impurities from other substances such as minerals or scratches) and a balanced cut that allows maximum light reflection (Scott & Yelowitz 2010). Although the presence of such features match particular rings to the engagement ring category, a focus solely on the physical attributes of the ring does not address the audience expectations of the category of engagement ring. Category expectations are socially derived beliefs about how category members should behave, or how a particular category member should be used (Hsu Hannan & Kocak 2009, Hsu 2006, Zerubavel 1991). Although often described as synonymous with the attributes that determine membership, expectations are in fact separate from membership, because the link between particular expectations and a given category is not automatic. For example, the traditional Japanese art of Mukimono creates intricate carvings out of fruits and vegetables. Yet, the expectation is that such carvings are appreciated as works of art, and not as food, even though they are made entirely of edible materials and may appear on a plate next to actual food to be consumed (Haydock and Haydock 1989). Other garnishes, such as maraschino cherries on a sundae, are also placed decoratively, but are expected to be consumed. Thus, expectations are not automatically linked to category membership.The existence of engagement rings in the first place is a function of social beliefs and tradition about how marriages should be marked, and thus audience expectations for engagement rings focus on both the ring’s history as well as its usage, rather than solely its physical attributes. Historically, diamonds were popular for engagement rings because they were considered “pure” gemstones, whose clear color and strength matched the purity of the bride and the impending marriage (O’Rourke 2007, Trex 2010). Over time, an engagement ring became symbolic of the impending marriage itself, such that a pure diamond was a harbinger for a happy marriage, and anything else dangerously risked an unhappy union (Bernard 2014, Opperman 2012). The expectation that engagement rings should be pure is reinforced by a strong aversion to those rings that are tainted. Such tainting can happen for a number of reasons, including sourcing them from conflict or theft, but taintedness is particularly noticeable when rings are connected to a failed relationship. “I could never wear a ring that was from a divorce. Yeah, I consider it "tainted" or bad luck. My mom still has her rings from my dad and...just no.” (female, age unknown, 2012)“I'm a little superstitious about rings so I would want to know the history of the ring. If its (sic) a used ring from a happy marriage then I would feel like its (sic) a blessing but a used ring from an unhappy marriage would feel tainted. If you don't know the history then give me tin foil over the unknown any day.” (female, age 30, 2012)“I wouldn't want it, however I am superstitious and believe that if you wear an engagement ring from a failed relationship that your relationship is doomed to suffer the same fate.” (female, age unknown, 2012)The quotations highlight that tainted rings may fulfill all the requirements for membership in the engagement ring category via their physical attributes—stone, shape, design, but may nonetheless violate category expectations about purity. Thus, a diamond solitaire ring, despite being the prototypical engagement ring, could still be tainted if associated with a failed relationship. Therefore membership in the engagement ring category is no guarantee that a ring conforms to audience expectations. However, merely establishing that an object can be a full member of a category and nonetheless violate audience expectations does not establish that expectations themselves can impact performance. To examine this link, in the next section, I explore two outcomes capturing an object’s market value, price and authenticity. Both are particularly important to the setting of diamond engagement rings and strategy more generally. Authenticity captures the degree to which a ring possesses the attributes the sellers describe. Price is represents the agreed upon exchange value of the ring, and is negotiated between buyer and seller. Tainted diamond rings and price:In general, tainted objects are perceived as socially dangerous (even though they are not technically dangerous), and as a result individuals avoid them out of fear of contamination (Douglas 1966, Ashforth, Kreiner, Clark and Fugate 2007). Such fears manifest as general sense of uncomfortableness and extend to superstitious beliefs (Rozin, Milman and Nemeroff 1986). For example, life insurance policies, considered dangerous as they challenged religious beliefs about the equation of life and money, faced hurdles in market acceptance because some individuals feared that owning such tainted policies would hasten their death (Zelizer 1978). Additionally, laws were debated in some states in the 1980-90s that required disclosure for real estate that had been occupied by individuals living with HIV out of fear of contamination, despite the fact that infection with the virus could not occur simply through existing in the same space (Hartog 1994, Brown and Turlow 1996). Importantly, the fear connected to tainted objects is socially derived--that is, it is not necessary that the objects themselves must truly be contaminated or dangerous. Such fears translate directly into the desirability of owning or paying for tainted objects. For example, individuals asked to examine a shirt which they saw had recently been tried on by another person valued the shirt at a lower price than those who simply saw it hanging on a rack (Argo, Dahl and Morales 2007), reflecting a belief that clothes tried on by others are dirty. The opposite effect, that those objects endowed with luck or good fortune should have higher prices is also true: housing prices in China for apartments with lucky numbers sell more quickly and for greater prices than those that lack such numbers (Shum, Sun and Ye 2012).Individuals recognize that an engagement ring from a failed relationship still appears visually as an engagement ring, yet they still feel uncomfortable about the idea of purchasing it. “I wouldn’t want to do it. Sure, from a practical point of view I guess it makes sense, but I’d always know where the diamonds came from etc. and that wouldn’t sit well with me.” If buyers feel emotional unease or worry about contamination, they are unlikely to pay high prices for a tainted ring. Comparatively, pure rings have neither social nor personal risk and should therefore sell for higher prices. Thus, tainted engagement rings should sell for lower prices than those that are pure.Tainted rings and authenticityThe issue of the desirability of any tainted ring for a particular wearer, as determined in its price, is separate from any perception that that ring is authentic. Authenticity refers to both literal claims (such that an authentic version of Microsoft Office is one that is not counterfeit) and cultural values (such as an authentic tourist experience), and has been seen as important for attracting consumer attention on success (Beverland, Lindgreen and Vint 2008). This study focuses on literal authenticity, which is critical to product markets, particularly when provenance impacts value, as in antiques and art markets. In this study, authenticity captures the degree to which a ring legitimately embodies the claims made about it, such as its cut, color, carats and clarity. Diamond engagement rings are a product with tremendous information asymmetry and significant cost. Although the criteria on which a diamond’s value is determined is well known, the actual appraisal and grading of the ring is all but impossible to determine for the average buyer, who can use subjective visual cues (“big” or “sparkly”) to assess the ring but lacks skill in noticing flaws in the gemstone or even whether or not it is a diamond. Buyers worry that they may accidentally purchase a diamond that is not real, or one that is worth significantly less than what they paid for (Thau 2013). Thus, authenticity is an important dimension on which sellers in the diamond industry compete.Most engagement rings are worn for the entirety of marriage (as immortalized in the “A diamond is forever” slogan), and are seen as symbolic of a particular relationship at a point in time rather than as mere pieces of jewelry, which an owner might buy or sell like any other worn object. Therefore, the sale of such a used ring is an unusual event outside of its original purchase. When such a ring arrives on the secondary market, buyers are likely concerned about its authenticity. Typically such concerns can be ameliorated through a variety of methods that each represent a signal to prospective buyers because they are costly to sellers: warranties, certification programs, and seller reputation all assist buyers in feeling confident that they understand the complete history of the object, allowing them to make accurate assessments about the appropriate price (Melnick and Alm 2002, Jin and Kato 2006, Cabral and Hortacsu 2010). However, social means can also serve as a way of increasing confidence in transactions: both familiarity between buyer and seller as well as trust between the two of them reduce information asymmetry, particularly on large purchases, such as cars or houses (DiMaggio and Louch 1998). More generally, social disclosure has been shown to benefit relationships, including increasing trust, as revealing information about oneself brings others closer to them (Collins and Miller 1994, Dindia 2009).Since sellers face a significant amount of uncertainty in selling a used diamond engagement ring, disclosing its history may increase the ring’s perceived authenticity. A failed relationship is a legitimate reason for wishing to sell a ring, but disclosure is required, since taintedness is not a physical defect. Such a disclosure suggests that the ring itself is not stolen, but rather is the product of unfortunate circumstances. Revealing the failed relationship also may suggest that the ring would not have been sold had the relationship endured, which appeals to authenticity by implying the seller would prefer it for him or herself save for the circumstances. Without such disclosure, buyers may instead be unsure of the ring’s origins, and therefore doubt the authenticity of the ring. This is particularly true because other reasons, like needing money for other bills, or receiving a larger ring, challenge the symbolic value of the ring and instead focus on it as a physical object. A tainted diamond engagement ring, with its legitimate reason for being offered for sale, may be seen as more authentic (that is, more likely to match the claims made by sellers) than its non-tainted counterparts. DATA AND METHODSI examine the effect of taintedness on price using archival data from the online auction web site eBay. I examine the effect of taintedness on authenticity with a controlled experiment taken online by a wide sample of adults. The experiment also replicates the archival eBay study. Taken together, the two studies allow for the establishment of the effect and claims about causality to be made.ARCHIVAL DATAMy data come from eBay auctions between January 1, 2011 and February 15, 2012 in category 152899, diamond solitaire rings, and are provided by Terapeak, eBay’s licensing arm. eBay provides several advantages as a research setting, among them that the listings are designed to present all information necessary for a buyer, and the multiple listings observable in a day allow a comparison of alternatives available for the buyer. As a result, approximating the search process requires fewer assumptions about market behavior. Listing a diamond solitaire ring on eBay is a relatively straightforward process. After choosing the diamond solitaire category, sellers are prompted to answer questions about the ring’s physical attributes using a combination of drop down menus and free text. Sellers can choose from a variety of paid advertising features, such as an enhanced photo gallery with larger pictures or using a reserve price (a non-revealed price which must be met for the item to be sold). Sellers also provide an item description—an unlimited text field which they can use to share information about the ring. Sellers disclose a great deal of information in their item descriptions. Some of this information is clearly designed to enhance the trustworthiness of their listings. For example, sellers highlight their years of experience or their quality ratings and satisfied customers. Others included photographs of certification by gemological associations or insurance appraisals. Additionally, many sellers described specifically the origins of the ring for sale, positive or negative. For example, one seller wrote, “Wore the Rings for ten months (divorced). Has lifetime insurance through Zales” and another noted, “I need a car more than I need a ring at this point in my life.” Such disclosures are in line with research suggesting that giving a reason of any kind—even non-sensical ones—improves outcomes for individuals (Langer, Blank and Chanowitz 1978), and providing no explanation generally leads to reduced information outcomes. I excluded listings with multiple rings, listings of “moissanite” rings (a diamond substitute), and listings with starting prices that were greater than $20,000. I also removed auctions that were longer than three weeks, since prior literature and the eBay web site suggest most individual sellers use shorter time frames. I also exclude unusual outliers, for example, rings listed as 172 carats. Importantly, I did not include rings with buy-it-now or make-an-offer options, so all sold prices are the results of completed open ascending price auctions. This yielded a set of 1,560,998 listings and 14,710 sales. Only 3870 rings (.2 percent) contain reference to a failed relationship. Because of the relatively low number of failed rings, I use a matched sample without replacement based on the 3870 rings (Elfenbein & McManus 2007). For each ring, I searched for up to five rings from listings that closed within three days of the focal ring. I attempted to match these rings as much as possible to the failed relationship rings. I chose the variables for matching based on regressions of purchase on the 1,560,998 observation sample, weighting more heavily those variables that were of greater significance in the regression. Because there were many thousands of possible matches for each failed ring, I used a two-step process. First, a few hundred rings were chosen at random that matched the carats of the failed ring. Then, within that reduced sample of rings, close matches were attempted based on a weighting of carats first, followed by cut, color, clarity, metal type, as well as the start price and number of listings by the seller. The latter criterion is important because it allows me to examine sellers of equal experience. I chose the five rings closest to the failed ring in each of these characteristics. Additionally, I tried several different matching scenarios; varying the weighting (removing seller characteristics, for example, or weighting all variables equally) did not significantly impact the results. The matching yielded a final sample of 14,532 rings (including the 3870 failed relationship rings), of which 291 sold.Although my inquiry focuses on the price of the ring, my results may be biased if I do not address the fact that not all rings sell, and thus all rings do not have a sales price. I use a Heckman selection model that adjusts for the sample selection by first estimating the likelihood of sale using a probit model, and then estimating the price using ordinary least squares with a variable to account for selection (Mannin, Duan and Rogers 1987, Heckman 1976). The unit of analysis is the individual auction listing.Dependent variable. The dependent variable measures the final sales price obtained through eBay’s open ascending price auction. The sales price was highly skewed, and so I took the natural log. Independent variable. My independent variable is an indicator variable coded “1” if an item description discusses a failed relationship and 0 otherwise. Failed relationships were determined from a computerized search of the item description for words indicating a divorce (thus rings that had been worn by married individuals), or other failure, such as a failed engagement. The search terms for each type were based on textual analysis of approximately 4000 item descriptions. Divorces were determined by “divorce” and its roots, “marriage ended,” “ex- husband/wife,” “not/no longer married.” Other failures were determined based on the words “ex-fiancé(e),” “previous girlfriend/boyfriend,” “called off,” “fiancé/e cheated,” ”not going to happen,” “ended abruptly,” “right ring wrong guy” “my loss is your gain,” “want to move on with my life,” “I don’t love him anymore,” “things didn’t work out,” “did not last,” “going to get engaged,” “split up,” and “broken off.” The resultant item descriptions were then examined to ensure that the ring was in fact from a failed relationship—for example, a ring that was described with the phrase, “This is not a ring from a divorce” would be flagged as a possible failed relationship by the computerized search, but would be excluded afterward.Control Variables. I needed to control for a number of characteristics of both the rings and the sellers that may influence the price and likelihood of sale for each listing. All variables, unless otherwise indicated, are measured contemporaneously on the day of the auction closing.The market price for rings is based in part on the cut, color, clarity and size of the diamond. The most popular cut of diamond is the round or brilliant cut, but diamonds can be cut into many shapes. Sellers on eBay could independently type the style of cut of the ring into a specific menu; as a result, sellers listed many types of diamond cuts, including Asscher, triangle, heart and cushion—as well as, in some instances, inappropriate descriptors like “sparkly.” Because of this, I created a set of categorical variables for princess/square, oval, marquise, or pear shaped diamonds, a residual category for other cuts, and used round as the reference category, since round diamonds are the most popular. The color of a diamond is a letter that varies from D (colorless) to Z (light yellow) but most diamonds in the sample fell in the D to I range. For ease of interpretation in the regression, and because there are only 291 observations in the second stage, I converted this into a numerical scale ranging between 0 and 7. Clarity is a subjective judgment by a jeweler based on the level of inclusions, or imperfections, in an individual diamond. For ease of interpretation as well as sample size concerns, I again converted this into a numerical scale. Carats, the weight or size of a diamond, is perhaps the most important dimension for buyers. The data on size was highly skewed, with very few rings listed as greater than 3 carats. To avoid losing observations, since some of the divorced rings did not include a value for the carats of the ring, I created a categorical variable for whether or not the seller specifically listed the carats as well as including the actual carats listed with zero as the reference. (Dropping the observations without carats does not change results.) I also added a squared term for carats to account for nonlinearity in demand. I included categorical variables for the metal of the setting, indicating platinum or gold with all others as the reference category. Additionally, I created an indicator variable that accounted for a national brand name being included, such as Kay Jewelers or Tiffany. Since rings certified by a grading agency may attract more attention, I added a variable coded 1 if the listing title included that the ring was certified by any grading agency and 0 otherwise. Measures of the presence of a reserve price or additional pictures were coded as 1 when present and 0 otherwise. Both of these options require the payment of additional fees, and so may also suggest the seriousness of the seller to make sure the ring is sold. I included the start price, which, given its highly skewed nature, was log transformed. I also controlled for the number of words in the item description, the duration of the auction in days, and the number of solitaire rings available on eBay that day. Engagement ring specialists were indicated if the seller focused solely on listing engagement rings (that is, all of his/her prior listings were in category 152899 on eBay). I controlled for the natural log of the number of listings the seller had on eBay, as well as the number of categories in which he or she had listed items, since prior research suggests that the items of non-focused sellers may suffer a penalty (Hsu, Hannan and Kocak 2009). In the first stage regression predicting likelihood of sale, I used the seller’s success rate, or percentage of successful sales in the prior month, as a variable that would impact likelihood of sale but be less likely to impact the price.In the regression of sales price, I included the count of the number of bids that an item received as well as the total number of engagement rings the buyer had purchased in the past year. I also included the seller’s feedback score, a measure of the reputation of the individual seller. Unfortunately, this variable was available only for completed sales. Descriptive Statistics.I report the means, standard deviations and correlations for all rings in Table 1. Although sixty-six percent of listed rings are round in cut, fifty-six percent of sold rings are round in cut, and 19% are square. Most rings are set in gold, and the average carats listed is 1.26, although the average carats of sold rings is 1.03. The average clarity is 5.3, which approximates SI1 clarity, which is in the midrange of all possible clarities and the most common clarity grade. Three percent of the listings had a jeweler’s name. Approximately 21% of the rings come from a failed relationship, which reflects the successful matching process. The correlations are largely unremarkable and in expected directions.RESULTSI present only full models of the regression equation in Table 2. I include both the selection equation and the final corrected model. In the first stage, the variable for failed relationship is significant and negative, suggesting that, controlling for other characteristics, rings from failed relationships are less likely to be sold.The regression on price is shown in model 2 of Table 2. If the variable for failed relationship is significant and negative, this suggests that not meeting audience expectations can impact market value. A few control variables are worth noting. The variables for marquise, oval, and princess shaped diamonds are negative, suggesting these cuts sell for less than round cuts. Carats and carats squared are significant in an inverted-U shape. Platinum rings sell for more than gold or other metals, as do rings with a higher starting price and those with more bids. These variables are in the expected direction. The variables for certification and jeweler’s name are not significant. One reason these results might occur is that the measures are rather broad; if only certain jewelers or certain certifications are valuable, the broader measures would not capture an effect. The variable for a failed relationship is significant and negative, suggesting that rings from failed relationships are associated with lower prices than those that are not from such relationships.Several concerns exist with the eBay data. One critical issue involves the price setting aspect of the auction. If sellers are na?ve about the appropriate price of their rings, they may set the price too high or too low. If buyers are informed about prices, the former case should decrease the likelihood of sale, while the latter should be taken care of by the auction mechanism. The concern for the regression is whether those who have failed relationships systematically over- or underprice in a way that is different from other sellers. For example, sellers with failed rings might be more desperate, and so might underprice their ring in hopes of a quick sale. Although a close reading of many item descriptions suggests that mispricing occurs with all sellers (the most common issue being to overprice due to using the appraisal value of the ring as an anchor for its actual value rather than its sales price), I ran a regression predicting start price using all variables in the first stage to see if failed rings systematically priced higher or lower. The variable was positive but not significant (p<.34). Additionally, in the experiment below start price is the same in all conditions, which eliminates the concern.A greater concern is that the failed relationship variable may be capturing other issues about sellers that lead to lower prices for such rings. This is concern is reinforced by a casual examination of individual listings. Even in listings that use approximately the same number of words, the content of the listings does seem to differ between failed relationships and those without, with non-failed relationship listings tending to include far more information about selling practices, such as very detailed information about shipping methods and returns, while failed relationship listing tend to discuss more about the rings themselves. These listings occasionally contain lengthy personal narratives that would likely preclude the sale because they suggest that the seller is unreliable, such as the seller who whose listing included the following narrative (sic), “My fiance cheated on me recently and decided to leave me and our daughter high and dry for another man,Also leaving me with 20,000.00 in credit card debt.So she took her ring off to shower and forgot it as she ran off with another man…..” Such deeply concerning narratives should reduce the likelihood of any bids at all, which would be captured in the first stage regression. Indeed, the mean number of words for failed rings is 2026.72 and only 1194.82 for the non-failed rings in the full sample. In the sample of sold rings, however, the mean number of words is 455.21 for failed rings and 616.65 for non-failed rings. In the experiment below, the narratives are identical, which eliminates this confounding issue. EXPERIMENTIn order to explore the effect of taintedness on authenticity, I used an experiment that examined buyers’ perceptions of authenticity and also, for external validity, replicated the archival eBay findings. I chose to focus on buyer perceptions because testing seller motivations would require subjects to assume the role of a seller (and thus adopt, in some instances, a persona of divorce), while acting as a buyer allows them to respond with fewer assumptions. Additionally, measuring buyer perception is consistent with measuring price, as price in an auction is ultimately set by buyers (even though sellers set the starting price). I used three conditions. In the first condition, subjects read an item description that included the information suggesting a happy marriage; in the second condition, subjects read an item description that included information suggesting the ring was from a divorce; and in the control condition, subjects read an item description from a store. A total of 598 subjects participated, but data from 46 of them were unusable because they did not actually answer the survey or failed the manipulation checks. Procedure. Subjects were recruited online via ’s Mechanical Turk (MTurk) worker interface and invited to participate in a short experiment about online buying behavior for $3.00. MTurk subjects have been found to be more diverse than student experimental populations and slightly older (33 years versus 24 years) than adult populations (Buhrmester, Kwang, and Gosling 2011). However, results of experiments using traditional experimental populations have been replicated using MTurk subjects in fields ranging from psychology to political science (Buhrmester, Kwang, and Gosling 2011; Berinsky, Huber and Lenz 2013, Rand 2012). Thus, MTurk subjects are considered to be equal to other experimental populations in the United States. Subjects who consented to the study were randomly shown one of the three ads. Each ad showed several photographs of the ring as well as a GIA authenticity and grading certificate, which the seller promised to include with the ring. The seller (“Pamela_el” in the divorce and happy marriage conditions; “Ring Depot” in the store condition) had 100% positive feedback on 4893 sales, was top rated, and accepted returns. In each case the seller offered a .70 carat round brilliant cut 18kt white gold diamond engagement ring with a starting bid of $300.00 and original purchase price of $3500. The only difference across the three ads was the item description. In the happy marriage condition, the item description included the following information: “I am still happily married--I am selling the ring because I prefer to wear only a wedding band because I work with my hands.” Those in the divorce condition saw the exact same ad, except that the seller noted, “Due to a divorce, I am auctioning this gorgeous .70 carat diamond ring.” The item description ended with the comment, “Since my ex and I split up I don’t wear it anymore, but someone else should!” Those in the store condition saw an advertisement that noted “This ring would cost $3500 in a retail store. Due to excess inventory, you can have it for much less than that!” Full item descriptions are available in Appendix A.Approximately half of the participants were asked, as part of assessing their likelihood of purchase: “How sure are you that this ring is authentic? (That is, how sure are you that you would receive the item described if you were the winning bidder?)” They then indicated their likelihood on a scale of 1 to 7. The rest of the participants were asked, “If you were planning to buy an engagement ring on eBay, how much money would you pay for this ring?” They were able to choose a value on a sliding scale from $0 to $4,000. These subjects were also asked how authentic they thought the ring was as above (after they chose a value) and were also asked “How desperate do you think the seller is to sell this ring? (That is, how important is it for the seller to receive cash quickly?)” Both sets of subjects also answered a series of questions designed to capture their beliefs about engagement rings, purchasing behavior, and provided demographic information. Several of these demographic variables were used as controls in the ensuing regressions. Gender was coded as 1 if female, zero otherwise. Subjects indicated their marital status, although I converted this into an indicator variable coded 1 for married and zero otherwise. Age was coded in five year increments starting at age 20. Income was measured in $30,000 increments, starting at $0. I measured uncertainty by timing the length of time each respondent spent before answering the main question of interest. Because this was skewed, I took the natural log. Table 3 presents combined descriptive statistics and correlations of the online subjects for both the authenticity and price dependent variables. In the entire subject pool, 50% were female and 34% were married. The average age of subjects was between 25-29, and the average income was between $30-60,000. Subjects were largely evenly split between groups, with 35% in the divorce condition, 34% in the happy marriage condition, and the rest in the store condition. RESULTSAuthenticity Experiment: In Figure 1 I present the mean price of the three conditions. The bar charts suggest that respondents found authenticity highest for the divorce condition, followed by the happy marriage condition and lastly the jewelry store condition. This suggests differences between the three groups, but not definitively. In Table 4 I present results from an ordered probit regression on authenticity. I used an ordered probit because the dependent variable is an ordered scale; however, results are robust when using ordinary least squares. The coefficient for the divorce variable is significant and positive, while the coefficient for the happy marriage condition is not significant. Thus, the divorce disclosure is seen as significantly more authentic than the jewelry store or the happy marriage conditions.Price Experiment: One concern with running an experiment is external validity—the population of subjects may be significantly different than the general population (Brewer and Hunter 1989). If the results for the price regression are similar to the results from the eBay data, then I can be more confident that the significant authenticity finding is generalizable, and that attributing the results in the eBay archival data to the failure of the relationship is valid. In Figure 2 I present the mean price of the three conditions. The bar charts suggest differences across prices in the three conditions, with the engagement ring from a divorce selling for the lowest price, the ring from the online jewelry store selling for the highest price, and the ring from the happy marriage selling for somewhere in between. Such a table is indicative of differences in the three groups, but not statistically definitive. In Table 5, I show results from regressions on price with the same control variables as before as well as categorical variables for the divorce condition and the happy marriage condition, with the jewelry store condition as the reference category. Interestingly, those who rated the seller as desperate chose lower prices. The other control variables were not significant. The coefficient for the divorce variable is significant and negative, while the coefficient for the happy marriage condition is negative but not significant. This suggests that the prices in the divorce ring condition are significantly lower than the jewelry store or the happy marriage conditions, consistent with the archival eBay data.Subjects were given an opportunity to explain their answers. Several subjects noted that the explanation for selling was important to them, for example, “The seller has sold thousands of items, has a perfect feedback score, is an eBay top-rated seller, and provides a satisfactory explanation for why the item is being sold. The certification provided with the ring also appears to be legitimate;” or, “The ring appears to be high quality at a reasonable price. Also the tone of the listing makes the seller appear to be legitimate (not a scammer).” While many of the responses simply stated it was the highest price a subject would be willing to pay or responded to the aesthetics of the ring, several subjects noted that the divorce played a role in their pricing, for example noting, “I feel like it may bring bad luck since she got divorced;” “I wouldn’t pay much for the ring since it comes with the bad history of a divorce, but this is a fair cheap price slightly above what the seller is asking;” “I would be willing to buy this quality diamond ring a good price! Since the person is selling it due to a divorce, I can get it for a good deal, as most people wouldnt (sic) buy it because of this fact;” and “I would not want or buy a ring that was used in a failed relationship/marriage. I am not overly superstitious but I would feel weird and uncomfortable owning a ring from a person whose relationship soured.” I conducted several robustness checks on both regressions, adding variables for frequency of online and eBay purchases as well as choosing different combinations of marital status, such as separating out those who were engaged or divorced. Additionally, I restricted the sample only to those who were eligible to be married, since it could be argued that married individuals are not the appropriate sample audience. In each case, results remained the same. Finally, I also added an interaction between the divorce condition and the authenticity variable. My theoretical interest is in establishing that taintedness can have a positive and negative relationship to market value depending on the measure used. Because of the positive and significant main effect of authenticity, it is important to examine if divorced rings perceived as authentic have a higher purchase price, as such a result would undermine the separate results for authenticity in the first experiment. In Model 2 of Table 5, I collapsed the happy marriage and store conditions for ease of interpretation. The interaction for divorce and authenticity is not statistically significant, while the main effects remain so, lending support for the main effects separately but not together. That is, both authenticity and taintedness impact the purchase price of a ring, but do not do so jointly.It is worth recalling that authenticity here measures only that buyers expect the ring to match the claims made about it, and does not represent a value-based judgment as in other tests of authenticity (e.g. Kovacs, Carroll and Lehman 2014). As a result, the finding that a tainted ring is perceived to be genuine, yet that a tainted ring may not be valued or desired by particular individuals suggests that claim-based authenticity may not be enough to overcome expectations of what an appropriate engagement ring should be. Although most studies of authenticity show that increased authenticity is associated with increased price, in those cases, the experience or object that is authentic is one that is positive, such as a tourist or dining experience or particular craft methods of production (Carroll and Swaminathan 2000, Newman and Dhar 2014). However, there is no reason that an authentic object should necessarily be a positive valence, especially when authenticity captures only the genuineness of an object, and not a particular cultural connection. In this setting, prospective buyers do not doubt the veracity of the ring but nonetheless prefer those that are not tainted.Both the archival data and the experimental data allow for a complementary approach for understanding the effect of taintedness in product markets, and how expectations are linked to market outcomes more generally. They should be considered together since each study has particular drawbacks that the other attends to (Small 2011, Brewer & Hunter 1989). The archival data contains real world sellers using a variety of item descriptions, including differing levels of personal information, differing ring qualities, and different starting prices. However, there are several limitations, among them the lack of seller reputation in the first stage of the purchase equation, difficulty assessing the legitimacy of start prices as well as buyer skill in determining value, and my use of matched data rather than the full sample. Nonetheless, the archival data shows that rings from a failed relationship are associated with lower prices. The experiment replicated the eBay results using random assignment to deal with the heterogeneity of buyer experience with diamonds and therefore ability to assess pricing. The experimental design also addresses the concerns about reputation, concerns about the matching process I use, and issues with sellers’ ability to accurately determine a starting price for their ring by providing advertisements where the only difference is the reason for the sale of the ring. Additionally, buyer perceptions about issues such as desperation can be assessed via the experiment, as can demographic characteristics not available in the archival data. The results also allow for the testing of buyer beliefs, in particular their views about authenticity and about appropriate expectations for engagement rings more generally. The ability to test both an actual market value (price) and a perceived market value (authenticity) would not be possible without the use of both designs.DiscussionThis study began by arguing for a closer look at the expectations that underlie category membership. When objects span categories, one reason for the reduced performance that often ensues is that such objects do not meet audience expectations; implied by this is that full members of the category must by definition meet audience expectations. However, research making this claim does not explicitly measure expectations. The design of this study allowed me to hold category membership constant and look solely at expectations in order to assess whether the explicit study of expectations yielded insight into market outcomes. Diamond solitaire rings are the prototypical engagement ring, and so meet the criteria for full membership into the engagement ring category. However, such rings can differ in the circumstances that lead to their sale. Some rings have no negative associations to prior owners, and others are tainted, associated with a failed relationship. Audiences expect that engagement rings should be pure, and not associated with a failed relationship. Using archival data from eBay as well as a controlled experiment, I showed that tainted rings sell for a lower price than those that are pure. However, tainted diamond rings are seen as more authentic—that is, more likely to match the claims made by the seller, than other diamond rings. These results have implications for both studies of categories as well as the cognitive underpinnings of strategy.For studies of categorical approaches to markets, the findings here suggest that audience expectations deserve specific attention in categories research. Prior literature has focused on category spanning as evidence of the violation of audience expectations, but this study suggests that full members of a category as determined by their attributes can also violate expectations. Here, even a prototypical member of the engagement ring category, a diamond solitaire ring, nonetheless violated audience expectations when it was associated with a failed relationship. Since the violation of expectations can occur for both clearly classified objects and category spanning ones, measuring category spanning alone may not be adequately capturing the underlying expectations that are inherent in category membership. If audiences do not like category spanning objects, the reasons for this may not lie solely with the mis-match for expectations, as is often stated. Similarly, audience appreciation for full members of a category may not be solely because of the match between expectations and category membership. The findings also highlight an important distinction between attributes and expectations themselves. Membership in a category is a function of an object’s possession of particular attributes, but audience expectations lie on top of these attributes and give meaning to them. A single diamond is an attribute that places a ring in the engagement ring category, but purity is an expectation that is placed upon that diamond by the distinctly socially determined nature of that category (Douglas 1966 (2002)). A long standing history suggests that taintedness creates boundaries that help individuals navigate social space and create order (Douglas 1966 (2002)), but taintedness is only one kind of category expectation, and may not be relevant for all categories. The point of the paper is not to establish locked-in-stone predictions for violations of category expectations, but rather to argue that paying attention to exactly what expectations are violated within a category, whether by spanning categories or not, should lead to greater insight about both which measures of market value are particularly relevant to an industry as well as governing predictions about whether value increases or decreases. Both authenticity and price are important indicators of the worth of a diamond engagement ring. By showing that violating the same audience expectation can both increase and decrease market value, depending on the measure chosen, I further highlight the need to consider the relationship between the expectation and the outcome being chosen, rather than assuming a positive relationship between adherence to expectations and all measures of performance outcomes. Current approaches to viewing category expectations, which simply mathematically measure category spanning, would miss the violation of expectations itself, but would also miss how a particular expectation may lead to certain types of performance outcomes and not others, because current approaches do not specify which violations occur across a set of categories and rely on generic relationship of value. This is not to say that broad mathematical approaches should not be used, but rather that deeper insight into underlying audience expectations would more carefully establish the link between the cognition of individuals and resulting performance outcomes, ostensibly the very goal of category research in the first place. Determining whether the same types of violations of expectations occur across all category spanning objects in a particular market space, or whether different combinations of categories violate different expectations inherent in each would give insight to the relationship between particular market categories and could potentially explain variance in market outcomes. Such investigation offers the potential to find new connections among different branches of category research. For example, while many studies support the discount in market outcomes from multiple category membership, other studies find multiple category membership to be beneficial (e.g. Smith 2012, Vergne 2012, Pontikes 2012). Thus far, findings of benefits are seen as exceptional and instances of the boundary conditions of theories of multiple category membership. Examining the different kinds of violations of expectations underlying category membership offers one avenue of investigation that may unite these divergent findings into a more cohesive body of research. As demonstrated here, part of the divergence of findings may come solely from the type of market outcome measured and not from the mechanism per se. In this study, revealing the tainted association with a failed marriage establishes legitimacy for selling a ring (leading to increased perceptions of authenticity), but nonetheless does not make it desirable as a ring (leading to lower purchase prices). Mathematical methods of assessing category membership establish that particular objects or firms violate boundaries to varying degrees, but cannot distinguish amongst the different reasons that this occurs, and therefore deny understanding of the deeper cognitive processes likely operating in many markets. To ignore, for example, the role of taintedness in such a profit driven industry as the diamond industry is to ignore a critical fact about the success of both the primary and secondary markets for rings. Taintedness helps quash what could be a robust secondary market for diamond engagement rings, and thereby increases the profits of jewelers and diamond manufacturers alike, and does so through cognitive, and often superstitious, beliefs. A common complaint about the diamond market is that once purchased, diamonds have little resale value (Epstein 1982). Although marketing campaigns such as “A diamond is forever” and strong-arm tactics by DeBeers make this possible, beliefs in taintedness also play a role. Diamonds from failed engagements are identical to any other diamond in physical attributes, but represent a separate type of ring socially, one that is less valued for purely social beliefs (Zelizer 1989). Ignoring the role that expectations such as these play in determining market outcomes misses opportunities to truly understand how categories, which are distinctly cognitive methods of approaching markets, influence markets as a whole. Focusing on expectations also yields particular strategic insights for the market for diamond rings. Taintedness appears to reduce market prices, yet the rings are seen as more authentic, that is, legitimately described. Since rings from failed relationships sell for less than other rings, arbitrage in the secondary market is possible. Professional buyers who can then “de-contaminate” a ring from its history stand to make increased margins. Additionally, strategic positioning suggests that overcoming the taintedness of rings from failed engagements represents a market opportunity as well. Since these rings are seen as more authentic, finding buyers for whom taintedness is not a concern is an additional opportunity. Although the notion of purity is steeped into the traditions surrounding an engagement, non-traditional practices surrounding engagements are increasing, as part of a trend toward non-traditional practices in weddings more generally (Price 2014). Surveys suggest that women are increasingly being involved in the process of purchasing an engagement ring, at times even splitting the cost of the ring with their fiancé (Marcotte 2013). This coincides with a focus on larger stones and rings as a fashion statement rather than a symbol of purity (Opperman 2012, O’Rourke 2012, Price 2014). When part of the allure of a diamond ring is the ability to “wave [it] triumphantly under every nose you encounter” rather than its symbolic purity (Marcotte 2013), used engagement rings allow for the largest diamond given a certain price (Price 2014). These buyers may not be concerned with the failed relationship leading to the sale of the ring, but instead on the ring’s physical attributes, preferring the maximum ring at a bargain price. Such trends are recent changes to the industry that would be obscured in typical ways of assessing the impact of categorization in markets as well as classic approaches to competitive strategy.More generally, these results point to an opportunity for studying competitive implications arising from the expectations that underlie particular market categories rather than simply the category attributes themselves. Prior research has shown, for example, that firms respond to the penalties of multiple category membership either by attempting to properly establish categorical associations (Kennedy 2008) or by de-diversifying to assume a more focused (and presumably more easily understood) market identity (Zuckerman 2004). However, even within a single category, market expectations impact success as shown here. Firms that can enhance particular expectations, as when DeBeers creates advertising that reinforces how engagement rings should be worn, or that can position themselves in response to expectations, either by countering those expectations in a market (for example, working to reduce superstitious beliefs in the diamond market) or by segmenting the population to find a niche willing to ignore expectations in favor of other attributes, may be able to achieve a competitive advantage. The further study of audience expectations, particularly as they link to category membership through different settings will allow a more complete understanding of how markets function—not only the underlying cognitive roots of how objects are valued, but also how such markets impact the strategies of those who involved with them.ReferencesArgo, J., D. Dahl, and A. Morales. 2007. “Consumer contamination: how consumers react to products touched by others.” Journal of Marketing 70: 81-94.Ashforth, B. E., G. E. Kreiner, M. A. Clark, and M. Fugate. 2007. “Normalizing dirty work: Managerial tactics for countering occupational taint. Academy of Management Journal 50: 149-174.Bernard, T. 2014. “With engagement rings, love meets budget.” New York Times accessed online. , M.F. 1990. “Rings and Promises.” Journal of Law, Economics, and Organization 6:203-215.Brown, R. and T. Turlow. 1996. “Buyers Beware: Statutes Shield Real Estate Brokers and Sellers Who Do Not Disclose That Properties Are Psychologically Tainted.” Oklahoma Law Review 49: 625-650.Bottomley, P., and S. Holden. 2001. “Do We Really Know How Consumers Evaluate Brand Extensions? Empirical Generalizations Based on Secondary Analysis of Eight Studies.” Journal of Marketing Research 38: 494-500.Bowers, A. 2014. “Relative comparison and category membership: the case of equity analysts.” Forthcoming, Organization Science published online in Articles in Advance, May 5 2014Buhrmester, M., T. Kwang, and S. Gosling. 2011. “Amazon’s Mechanical Turk: A New Source of Inexpensive, Yet High-Quality, Data?” Perspectives on Psychological Science 6: 3-5.Cabral, L. and A. Hortacsu. 2010. “The Dynamics of Seller Reputation: Theory and Evidence from eBay.” Journal of Industrial Economics 58: 54-78.Carroll, G. and A. Swaminathan. 2000. “Why the Microbrewery Movement? Organizational Dynamics of Resource Partitioning in the U.S. Brewing Industry.” American Journal of Sociology 106: 715-762.Collins, N. and L. Miller. 1994. “The disclosure-liking link: from meta-analysis toward a dynamic reconceptualization.” Psychological Bulletin 116: 457-475.Dindia, K. 2009. “Self-disclosure research: knowledge through meta-analysis.” In Interpersonal Communication Research: Advances Through Meta-analysis. Eds. M. Allen, R. Price, B. M. Gayle and N. Burrell. New Jersey: Lawrence Erlbaum Associates.Douglas, M. 1966 (2002). Purity and Danger. New York: Routledge Classics.Elfenbein, D. and B. McManus. 2010. “A Greater Price for a Greater Good? Evidence that Consumers Pay More for Charity Linked Products.” American Economic Journal: Economic Policy 2: 28-60.Epstein, E. 1982. “Have you ever tried to sell a diamond?” Atlantic Monthly 23.Haidt, J., C. McCauley, and P. Rozin. 1994. “Individual differences in sensitivity to disgust: A scale sampling seven domains of disgust elicitors.” Personality and Individual Differences 16: 701-713.Hartog, R. 1994. “Psychological Impact of AIDS on Real Property and a Real Estate Broker’s Duty to Disclose.” Arizona Law Review 36: 757-784. Heckman, J. 1976. “The common structure of statistical models of truncation, sample selection, limited dependent variables, and sample estimator for such models” Annals of Economic and Social Measurement 5: 475-592.Howard, V. 2008. Brides Inc. American Weddings and the Business of Tradition. Philadelphia: University of Pennsylvania Press.Hsu, G. 2006. “Jacks of all trades and masters of none: Audiences’ reactions to spanning genres in feature film production.” Administrative Science Quarterly 51:420-450.Hsu, G., M. T. Hannan, and ?. Ko?ak. 2009. "Multiple Category Memberships in Markets: An Integrative Theory and Two Empirical Tests." American Sociological Review 74: 150-169.Kovacs, B., G. Carroll, and D. Lehman. 2013. “Authenticity and Consumer Value Ratings: Empirical Tests from the Restaurant Domain.” Organization Science 25: 458-478.Leung, M. 2014. “Dilettante or Renaissance Person? How the Order of Job Experiences Affects Hiring in an External Labor Market.” American Sociological Review 79:136-158.Leung, M. and A. Sharkey. 2013. “Out of Sight, Out of Mind: Evidence of Perceptual Factors in the Multiple-Category Discount.” Organization Science, forthcoming.Mannin, W., N. Duan and W.H. Rogers. 1987. “Monte Carlo evidence on the choice between sample selection and two part models.” Journal of Econometrics: 35: 59-82.Melnik, M. and J. Alm. 2002. “Does a Seller’s eCommerce Reputation Matter? Evidence from eBay Auctions.” The Journal of Industrial Economics 50: 337-349.Mervis, C. and E. Rosch. 1981. “Categorization of natural objects.” Annual Review of Psychology 32: 89-115.Negro, G. and M. Leung. 2012. “’Actual’ and Perceptual Effects of Category Spanning: An Analysis of Critical Ratings of Italian Wines.” Organization Science 23 forthcoming.Newman, G. and R. Dhar. 2014. “Authenticity Is Contagious: Brand Essence and the Original Source of Production.” Journal of Marketing Research 51: 371-386.O’Rourke, M. 2007. “Diamonds are a Girl’s Worst Friend” Accessed online 06/29/2012., M. 2012. “Secondhand Engagement Rings: Second Best or First Rate?” Huffington Post. (February 28, 2012). Accessed online: , E. 2012. “Two Sides of the Same Coin: How Ambiguous Classification Affects Multiple Audiences’ Evaluations.” Administrative Science Quarterly 57: 81-118.Porac, J., H. Thomas, and C. Baden-Fuller. 1989. “Competitive Groups as Cognitive Communities: the Case of Scottish Knitwear Manufacturers.” Journal of Management 26: 397-416.Rapaport Report. 2012. Diamond Prices 2011 Research Report. Available via .Rozin, P., L. Millman, and C. Nemeroff. 1986. “Operation of the laws of sympathetic magic in disgust and other domains.”?Journal of Personality and Social Psychology, 50, 703-712.Ruef, M. and K. Patterson. 2009. “Credit and classification: The impact of industry boundaries in 19th century America.” Administrative Science Quarterly, 54: 486-520.Scott, F. and A. Yelowitz. 2010. “Pricing Anomalies in the Market for Diamonds: Evidence of Conformist Behavior.” Economic Inquiry 48: 353-368.Shum, M., W. Sun, and G. Ye. 2012. “Superstition and “Lucky” Apartments: Evidence from Transaction Level Data.” Working paper, CalTech.Spence, M. 1973. “Job Market Signaling.” Quarterly Journal of Eocnomics 87: 355-374.Tan, D. and P. W. Roberts. 2010. “Categorical coherence, classification volatility, and examiner added citations.” Research Policy, 39: 89-102.Thau, B. 2013. “Tis the season for counterfeit e-Commerce sites: an insider’s take on how to avoid being snookered by fakes.” . Accessed online. Trex, E. 2010. “Why engagement rings are made with diamonds.” Mental_floss, accessed online. , J.P. 2012. “Stigmatized categories and public disapproval of organizations: A mixed methods study of the global arms industry (1996-2007).” Academy of Management Journal 55: 1027-1052.Waguespack, D.M. and O. Sorenson. 2010. “The Ratings Game: Asymmetry in Classification” Organization Science 22: 541-553.Zerubavel, E. 1991. The Fine Line: Making Distinctions in Everyday Life. Chicago: University of Chicago Press.Zelizer, V. 1978. “Human values and the market: the case of life insurance and death in 19th-century America.” American Journal of Sociology 3: 591-610.Zelizer, V. 1989. “The Social Meaning of Money: “Special Monies.” American Journal of Sociology 95: 342-377.Zuckerman, E. W. 1999. “The Categorical Imperative: Securities Analysts and the Illegitimacy Discount.” American Journal of Sociology 104: 1398-1438.Zuckerman, E. W. 2004 “Structural Incoherence and Stock Market Activity.” American Sociological Review 69: 405-432.Table 1. Descriptive Statistics and correlations, eBay archival data (sold and unsold)Table 2. Two-stage regression of failure on price (ln), eBay archival dataModel 1 (Likelihood of sale)Model 2(Purchase price)Cut (marquise)-0.0488-0.3156(0.1503)(0.1776)*Cut (other)-0.1465-0.1335(0.1067)(0.1344)Cut (oval)-0.1155-1.1471(0.3782)(0.4537)**Cut (pear)-0.34070.2471(0.2758)(0.3387)Cut (princess)-0.1950-0.5419(0.0938)**(0.1248)***Color0.0261-0.0267(0.0228)(0.0268)Clarity0.01900.0623(0.0223)(0.0267)**Carats missing (1=yes)0.60251.1262(0.1574)***(0.2260)***Carats0.70981.5661(0.1316)***(0.1988)***Carats squared-0.1283-0.2561(0.0289)***(0.0394)***Metal (gold)-0.17090.0041(0.1506)(0.1480)Metal (platinum)-0.05040.4726(0.1895)(0.2204)**Length of item description-0.0002-0.0005(0.0001)**(0.0001)***Certified (1=yes)-0.13930.1292(0.0842)*(0.1126)Engagement ring specialist (1=yes)0.13410.1341(0.1266)(0.1697)Number of categories sold in0.0276-0.0382(0.0183)(0.0252)Starting price (Ln)-0.38190.2212(0.0225)***(0.0819)***Jeweler’s name (1=yes)-0.05730.0977(0.1164)(0.1419)Number of prior successful sales0.6721(0.1591)***Auction duration (days)-0.0318-0.0022(0.0131)**(0.0166)Extra pictures (1=yes)0.36770.5505(0.1031)***(0.1275)***Reserve price (1=yes)-1.15390.1174(0.1827)***(0.3029)Number of listings (seller) (Ln)-0.16840.1466(0.0246)***(0.0582)**Number of listings (total) (Ln)0.06270.1980(0.0805)(0.0960)**Seller feedback (Ln)-0.0190(0.0293)Number of prior ring purchases0.0070(0.0029)**Number of bids0.0583(0.0065)***Inverse Mills ratio0.3421(0.2769)Failed Relationship (1=yes)-0.1900-0.3389(0.1016)*(0.1274)***Constant0.54160.9082(0.9301)(1.0957)N14,532291R20.67Standard errors in parentheses. * p<.10; **p<.05; ***p<.01; two-tailed tests.Table 3. Descriptive Statistics, ExperimentTable 4. Ordered probit regression of sales condition on authenticity?Model 1Desperation-0.1479(0.0393)**Uncertainty0.1318(0.1552)Female (1=yes)-0.0006(0.1333)Age-0.0101(0.0318)Income-0.0581(0.0839)Married (1=yes)0.1284(0.1461)Divorce condition0.3122(0.1587)*Happy marriage condition0.0593 (0.1553)Log Likelihood-464.8560PR>X20.0218 N272Standard errors in parentheses, cut points omitted.***p<.01; **p<.02; *p<.05; two tailed tests.Table 5. Regression of sales condition on price (ln)Model 1Model 2Authenticity0.17640.1257(0.0535)**(0.0648)+Desperation-0.1077-0.1092(0.0431)*(0.0430)Female (1=yes)-0.0380-0.0544(0.1545)(0.1540)Age-0.0035-0.0027(0.0387)(0.0385)Income0.09640.0930(0.0976)(0.0974)Uncertainty (Ln)0.12710.1123(0.1067)(0.1071)Married (1=yes)-0.1259-0.1110(0.1767)(0.1766)Divorce condition-0.3667-1.0340(0.1824)*(.5821)+Happy marriage condition-0.1226(0.1809)Divorce * Authenticity0.1471(0.1124)Constant5.54575.8114(0.6171)**(0.6584)R20.090.10N280280Standard errors in parentheses. +p<.10;* p<.05; ** p<.02; *** p<.01; two-tailed tests.Figure 1. Authenticity as indicated by online subject poolFigure 2. Mean price as indicated by online subject pool.Appendix A. Description of eBay advertisement text, experimentDivorce condition:? Buy with confidence!? I accept returns and have 100% positive feedback!? Due to a divorce, I am auctioning this gorgeous .70 carat diamond ring.? It is a round brilliant cut and set in 18k white gold.? The clarity is SI1 (the level of most quality diamonds) and the color is D (very rare and the highest color level you can get), and the cut is very good.? It is certified by the GIA, and I will send you the certificate with the ring.? This ring is absolutely beautiful and I always got many compliments whenever I work it. It cost $3500. But you can have it for much less than that! Bidding starts at only $300. Since my ex and I split up I don’t wear it anymore, but someone else should!Happy marriage condition:? Buy with confidence!? I accept returns and have 100% positive feedback!? I am auctioning this gorgeous .70 carat diamond ring.? It is a round brilliant cut and set in 18k white gold.? The clarity is SI1 (the level of most quality diamonds) and the color is D (very rare and the highest color level you can get), and the cut is very good.? It is certified by the GIA, and I will send you the certificate with the ring.? This ring is absolutely beautiful and I always got many compliments whenever I work it. It cost $3500. But you can have it for much less than that! Bidding starts at only $300. I am still happily married—I prefer to wear only my wedding band because I work with my hands.Store condition:Buy with confidence!? Returns accepted and Ring Depot has 100% positive feedback!? Up for auction is this gorgeous .70 carat diamond ring.? It is a round brilliant cut and set in 18k white gold.? The clarity is SI1 (the level of most quality diamonds) and the color is D (very rare and the highest color level you can get), and the cut is very good.? It is certified by the GIA, and the certificate will be sent with the ring.? This ring is absolutely beautiful and you will get many compliments. The ring would cost $3500 in a retail store, but due to excess inventory, Ring Depot is selling it for much less than that! Bidding starts at only $300. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download