Abstract
To understand the relation between artistic reputation and sales prices from popular online auction sites, we analyzed over 3,000 online auction sales of 87 American artists who had expertise in watercolor using the websites eBay, LiveAuctioneers, and EBTH. Traditional indexes of reputation were measured by museum placements, as well as archival holdings, and book references and we also included the number of Instagram mentions. We used hierarchical linear modeling to investigate the relation of reputation with average prices as well as price trends over time, controlling for individual painting characteristics such as size and medium. All of our artistic reputation indicators, including number of Instagram hashtag references, were positively related to the average sales price per artist. Our results extend the literature on the role of artistic reputation to account for social media via Instagram, as well as provide one of the first empirical investigations of popular online art auctions.
The Internet has influenced the art world in numerous ways, making it easier for artists to publicize and sell their art, as well as making it easier for consumers to purchase art without setting foot in intimidating auction houses and galleries (see Fernandes & Afonso, 2020; Vickers, 2005). Although the Internet has influenced the buying habits of art consumers as well as the way we conceptualize artistic reputation (e.g., the number of times an artist is mentioned on a social media platform), there have been few studies that have analyzed how traditional and modern reputation indicators are related with sales price of art over time. In our study, we investigate the role of artistic reputation as measured by some traditional cultural indicators (e.g., museum placements) as well as a modern one, Instagram hashtag mentions, and relate these indicators to auction sales prices from popular online sites. Specifically, we track online prices of a group of twentieth Century American artists who had some specialization in the watercolor medium and we link buying behavior to artistic reputation.
In their landmark work of the sociological aspects of artistic reputation, Lang and Lang (1988, 1990) conducted an analysis of early twentieth Century printmakers in the United States and Great Britain to identify factors that led to posthumous neglect or fame. They defined reputation as “an objective social fact, a prevailing collective definition based on what the relevant public ‘knows’ about the artist” (Lang & Lang, 1988, p. 84) and they argued that reputation-making activities such as preserving biographical information for later audiences were key, as was visibility of works in important cultural institutions. Their theoretical framework (Lang & Lang, 1988), as well as their initial analysis of twentieth Century American and British printmakers (Lang & Lang, 1990) was important in that it helped lay a foundation for trying to understand why some artwork maintains popularity or increases over time, whereas other work fades into obscurity. The wide availability of data due to the Internet has made it possible to test their hypotheses in an even more rigorous manner than was possible in their original work. We do so in this research using a select group of American artists, known for their watercolor expertise (though most painted in other mediums as well), who are tracked over time using data from online art auction sites. We relate online auction prices to a variety of indicators of reputation.
For much of its history, art collecting was viewed as solely an activity for the rich who had large amounts of disposable income and could gain entry to high-powered auction houses such as Sotheby's and Christie's (Gersh, 2019). In fact, this is still the perception of the industry for many, fueled by news of record multi-million dollar sales of popular artists such as Warhol, Picasso, and Monet. Much of the literature on artist sales prices has focused on determining the viability of art as a long-term investment, as opposed to investigating auction sales to better understand the nature of artistic reputation (e.g., Agnello, 2002; Mei & Moses, 2002; Reeneboog & Spaenjers, 2013; Reitlinger, 1961). With widespread access to the Internet, however, art auction houses and dealers have turned to the Internet to sell to new markets, often selling art that would be lower in price than that typically carried at Christie's or Sotheby's (see Gersh, 2019). There are online auction sites that cater to more modest-priced budgets, such as LiveAuctioneers.com, that allow auction houses to run auctions simultaneously with in-person or online-exclusive auctions. In addition, general auction sites such as eBay.com and Everything But the House (EBTH.com) allow, under different business models, for users to either directly upload items for auction or direct sale to consumers (eBay) or take items for consignment from owners and list on their curated site (EBTH). These websites have allowed consumers with smaller amounts of money to participate in art collecting and they have allowed dealers to sell a wider range of artists to that broader audience. In our analysis, we focus on these newer auction sites and track a group of American artists as a way to explore price trends and artistic reputations over time.
We chose American artists who were known to have at least some expertise in the watercolor medium given that watercolor artwork tends to be more affordable than oil-based painting and, hence, might be more likely to be sold in online auctions where the target audience might have more financial constraints than the “high-rollers” who attend auctions from the big houses. In a discussion of the history of watercolor painting within the United States, Foster noted that “watercolors [were] much easier to buy than oil paintings, especially for those unused to purchasing original art” (Foster, 2017, p. 219). Watercolor paintings tend to be smaller in size, are perceived to have less durability, and are viewed as taking less time to complete than oil paintings, and hence tend to be priced lower than oil paintings (see Foster, 2017; O’Hara, 1946; Reeneboog & Spaenjers, 2013). Given the greater accessibility and lower costs of watercolor paintings compared to other forms of art, we thought that there would be a wider range of paintings made available through more accessible online auction sites compared to higher-end types of art. In addition, given the lower prices, forgeries are less likely to be encountered, a particular problem in today's art world (see Walton, 2006). We should note also that the artists in our sample, although known as watercolorists, often worked in other mediums as well. For completeness, we included works in other medium though controlled statistically for differences in prices across mediums.
Factors That Influence Artistic Reputation
Lang and Lang (1988) highlight two key factors in the long-term survival of an artist's reputation and hence the value of his or her art: (1) visibility of the art and (2) documentation of the person behind the art. The former is most often measured as the placement of works in key museums or collections, whereas the latter can be documented in the maintenance of an artist's archival material and monographs highlighting the artist's work and biography. Lang and Lang (1990) linked reputation of twentieth Century British and American etchers to sales of prints, though their empirical analysis was limited given large datasets of sold prices were largely unavailable. Subsequent researchers have linked artistic reputation to auction prices empirically. Braden and Teekens (2019) found that an artist's highest auction price was related to status connections (e.g., placement in exhibitions with high status artists) as well as with number of references in books. Fraiberger, Sinatra, Resch, Riedl, and Barabási (2018) also found that artists who had access to institutions via exhibitions and museum collections were likely to have higher prices. Reeneboog and Spaenjers (2013) found that artists mentioned in art history textbooks had higher prices than those not mentioned.
There has been a fair amount of research that has aimed to predict auction prices for artists though the stability of artists’ prices have rarely been empirically studied. Many experts, however, have speculated on the value of art over time. For example, one dealer wrote on social media “It is inescapable that original fine art will almost never decline in value. Instead, most original fine art increases in value over time. And once in a while, an artist becomes the next ‘cause celebre’ and the originals by that artist take off in value” (Wesley Barrett Fine Arts, Instagram Post, 1/16/2021). Contrasted with that, another expert wrote “Only a small percentage of art increases in value over time” (Bamberger, 2019, location 560). In our analyses, we investigate whether reputation is linked to the average sales price as well as to changes in prices over time.
Hypotheses
Lang and Lang (1988) hypothesized that artists whose work is stored in cultural institutions are related to higher sales and more likely to be related to prices sustained over time. The mechanisms for this are several-fold. First, potential consumers are more likely to encounter artists’ work when it is stored in museums and hence likely to seek out artists’ work that is in museums. Second, the knowledge that an artist has work in a particular museum is viewed as a key indicator of an artist's prestige; in fact, museum placements are often mentioned in the advertisement of particular items. Presence in a museum's permanent collection suggests that experts (e.g., museum curators) believe that the artist's work deserves preservation and storage. Bamberger noted “The fact that museums or corporations own works by an artist is an indication that art-world experts and authorities have recognized the art as significant” (2019, location 2333). Gersh noted “Presence in a museum's permanent collection tend to validate the importance of an artist. The more museums, the greater the validation.” (Gersh, 2019, location 501). Thus, we hypothesize:
Artists with more museum placements will tend to have higher mean prices.
Artists who have biographical information stored and accessible to the public are more prone to discovery by the public as well as resurgence due to access by scholars and aficionados. For artists with little biographical information available, there is little beyond the art itself that can help potential buyers understand the value of the art. The traditional way of measuring accessibility of cultural information is through the placement of material in archives.
Artists with material available in archives will tend to have higher mean prices.
In addition to archival information, which would usually only be perused by scholars and true aficionados as opposed to casual potential buyers who are unlikely to visit archives, information via books should be available to a wider audience and so we predict a similar positive relationship, and perhaps stronger. Bamberger wrote “The more references that include an artist and the more sources of information you can find online, the more collectible his or her art is (2019, location 2283). Book references have been used as an index of reputation in other artistic realms as well, such as Baroque composers (Ceulemans, 2010). In addition, some artists also published their own writings providing other opportunities to promote their work.
Artists with more mentions in books will tend to have higher mean prices.
In addition to traditional cultural institutions that hold a particular artists’ work, social media is another index of cultural prestige, albeit one that does not connote the cultural prestige by curators given that anyone can post, comment, and share art. Social media mentions, however, may connote a more democratic measure of prestige that serves as an indicator of prestige and hence should be related to pricing. To measure social media mentions, we summed the number of Instagram hashtag mentions for each artist. Content posted on Instagram is unique from other social media platforms such as Facebook in that people post on Instagram to get “likes” whereas people may post on Facebook to spark discussion, provoking negative (angry faces) and positive (thumbs up and heart) reactions. Although people can comment on Instagram posts, the platform is not well-suited for discussion compared to Facebook where discussions are quite common. In our tabulating of Instagram mentions, museum visitors used the forum to highlight pieces they had admired, collectors bragged about pieces they now owned, and businesses posted art that they wished to sell. We reason that hashtag mentions of artists on Instagram are mainly positive and should relate to the cost of an artist's work.
Artists with more Instagram hashtag mentions will tend to have higher mean prices.
In addition to investigating the relation between reputation and average prices across artists, we speculated that reputation would be related to change in prices over time. Artists with strong reputations would be more likely to have prices increase over time, whereas artists with poor reputations would be likely to have prices decrease or remain stable over time.
Artists with more reputation (i.e., more museum placements, Instagram mentions, book references, and archival material) will have more positive trajectories of prices over time than artists with less reputation.
In addition to variables related to reputation, we were also interested in understanding the relation between reputation of auction house with prices in this online auction format. Previous research has shown that artwork sold by more prestigious auction houses is related to higher prices (Reeneboog & Spaenjers, 2013; Teti, Sacco, & Galli, 2014). We thought it would be important to replicate that research in this sample. It may be that the differences between more prestigious auction houses and less prestigious auction houses is less important in an online format where there is little direct contact with the auction house. Also, if buyers in general are less knowledgeable about art, they may care less about whether a piece of art is sold by an established auction house or an individual with no history of selling art (in the case of eBay). We hypothesize that prices in general would be higher for art sold from LiveAuctioneers and EBTH because those websites sell items curated at least to some extent, compared to eBay, for which any individual can upload a work of art for sale.
After controlling for painting characteristics, art sold on LiveAuctioneers and EBTH would tend to be higher in price compared to art sold on eBay.
Method
the Artists
We considered all artists in Norman Kent's 100 Watercolor Techniques (Kent, 1968), a book that featured 100 artists who had at least some degree of prominence within the watercolor community. Nearly all artists in the book were active members of the American Watercolor Society, the preeminent organization at the time for watercolor artists, and Kent chose artists to represent an aesthetically diverse range, including artists who were primarily realist in orientation as well as some who were primarily abstract. Kent, himself, was the editor at the time of American Artist, a leading magazine for artists. We excluded six artists who were still alive in 2020 as we were interested in date of death as a predictor of auction prices (see Itaya & Ursprung, 2016). We wanted to capture artists for which the large majority of their work would be sold via non-traditional electronic auction houses and so we excluded artists who had 10 or more sales of over $5,000 USD through the auction houses of Sotheby's and Christie's. Artists that were excluded for this reason were Charles Burchfield, Henry Gasser, Dong Kingman, Ogden Pleissner, Andrew Wyeth, and William Zorach. We excluded these artists for several reasons. First, the focus of Lang and Lang's (1988) analysis was on artists who had not yet made it to the upper echelon in artist ranks. Those two auction houses represent the upper echelon of artistry. Works by those artists do occasionally get sold in the online platforms studied in this analysis, though given the notoriety of the artists, their work is much more susceptible to forgery than the artists in this analysis. Additionally, by focusing on artists who have yet to hit the upper echelon of sales and auction houses, we were more likely to capture the full range of an artist's auction sales record via the studied platforms. Finally, we excluded one artist for which we could find zero online sales (Flora Smith). Based on the inclusion criteria, we had a final sample of 87 watercolor artists (80 males and 7 females). A full list of artists in the study can be found in Appendix A.
the Paintings Analyzed
We analyzed past auction records using several online tools. LiveAuctioneers.com is an auction platform used by numerous auction houses to include an online component to an auction that may also include in-person bidding. We also included records from Everything But the House (EBTH.com), an auction website that includes fine art. Finally, we analyzed eBay sales results using Worthpoint, a site that compiles sales records..
We excluded items that were listed as “attributed to” given that those items have uncertain provenance and are often viewed as untrustworthy (see Gersh, 2019) and previous researchers have found to have prices drop 50% or more (Reeneboog & Spaenjers, 2013). We also excluded prints, drawings, and sculptures. With these criteria, we found a total of 3,772 individual item sales across the 87 artists. 71.7% of those items were sold on LiveAuctioneers, 26.8% were sold on eBay, and 1.5% were sold on EBTH.
We included several other variables that were chosen as control variables that were likely to influence the price of a piece. We coded the size of the piece measured in square inches (M = 422.1, SD = 374.4), with the idea that larger pieces would on average be related to higher prices than smaller pieces (see David, 2016; Reeneboog & Spaenjers, 2013; Towse, 2011). Although all artists in our study had expertise in watercolor, there were a variety of painting mediums in our sales records, and so we also included the medium of the piece with the expectation that oil paintings would be related to higher prices than watercolor pieces and other mediums such as gouache, acrylic, and casein (Reeneboog & Spaenjers, 2013; Teti et al., 2014). The vast majority of our paintings sold were watercolors (71.4%) with oil paintings being the second most frequent medium (21.6%), with gouache (2.4%), pastel (1.4%), and acrylic (1.1%) being the other mediums represented with at least 1% of all paintings sold. Additionally, we included whether the painting was framed or not, expecting that framed pieces would be related to higher prices than unframed pieces (Loria, 1965). 76.1% of the pieces in our sample were framed, 18.7% unframed, and for 5.1% we were unable to determine. Finally, we included whether an individual piece was included in a lot sold with other pieces by the same artist. When paintings were sold in a lot of other paintings by the same artist, the final price was divided by the number of pieces and allocated equally. We expected that pieces combined in a lot would sell at a lower individual price than if sold separately. 9.1% of the paintings in our sample were sold in lots with other paintings; the vast majority of paintings (90.9%) were sold individually.
Reputation Indicators
Museum Placement
One factor in the preservation of an artist's work is the placement of his or her artwork in top cultural institutions, and in this case, holdings of work in art museums. We included 100 art museums throughout the United States. We consulted lists of top American museums (e.g., David, 2016) to ensure that we included prestigious art museums and then chose additional museums to ensure that we had geographic dispersion. Museums were included only if they had online, searchable databases of their collection. We tabulated the number of works that were included in each museum as well as counted the number of museums who held an artist's work. In our subsequent analyses, we focus on counts of museums who held particular work as opposed to number of total works across all museums. In a few cases, museums affiliated with the hometown of a particular artist had a large number of works held by that artist, even though very few other museums held work by that artist. For example, the Rochester-based artist Ralph Avery had over 60 paintings held by the Memorial Art Gallery in Rochester, though in our sample of museums studied, only one other museum held work by him. By focusing on number of museums holding work (as opposed to total number of works held), we felt like we had a more accurate measure of an artist's reputation. We included separate counts for painted work, along with total work given that some artists who were also printmakers may have much larger numbers of work given the ability to distribute multiple copies of an individual work to multiple institutions. We found similar results for both indicators (i.e., r = .72 between the two indicators and none of the subsequent analyses differed across the two indicators) and so report the total number of museums holding work throughout the results. The three museums that held work by the most artists in our sample were the New Britain Museum of American Art (24 artists’ work in their collection), the Smithsonian American Art Museum (23) and the Philadelphia Museum of Art (15). Twelve of the museums we considered did not hold works by any of the 87 artists we studied.
Publications
The website AskArt.com compiles information about artists and tabulates references to artists in published books and reference sources. We used their tabulations as an index of artistic reputation. Reeneboog and Spaenjers (2013) found that works by artists are priced 13.5% higher after an artist has been included in an important art history reference book. Note, we excluded works written by the painters themselves when compiling numbers with the idea that references by others would be more valuable as indicators of reputation than self-mentions.
Archival Material
We searched ArchivGrid, a meta-search engine that searches over 2,000 research archives throughout the United States and tabulated the number of archives that held material by each particular artist. To qualify, the archive had to have a distinct archival entry for the artist. Artists who were mentioned as part of another artist's entry were not counted.
Instagram Mentions
Another index of reputation was the number of times an artist was indexed on Instagram using a hashtag, which is a way for users to index posts by using the artist's name within a hashtag, such as #earlhorter. 1 As Gersh noted, Instagram has become an essential tool for living artists to market their work, as well as galleries and museums to promote their holdings: “Instagram has changed the face of art marketing. It has become the most powerful marketing instrument of art sales” (Gersh, 2019, location 436).
HLM Analyses
We used hierarchical linear modeling (HLM) to analyze and test our hypotheses (see Raudenbush & Bryk, 2002) given the nestedness of our data (i.e., painting sales are nested within individual artists and hence do not satisfy independence requirements of traditional multiple regression), using HLM 8.0 (Raudenbush, Cheong, & Congdon, 2019; see Raudenbush & Bryk, 2002 for an excellent reference on HLM). We had the outcome of an individual auction sale, which we predicted was a function of several painting variables such as size and medium, as well as auction year. We used the final price, including the auctioneer fee which is paid for by the LiveAuctioneer auction winners. Painting sales were nested within individual artists and we would expect that different artists would be related to different prices, and that those differences would be themselves related to the reputation indictors mentioned previously. HLM equations are like a series of regression equations. In our case, Level 1 was:
There are two equations for Level 2. In Equation (2), the Level 1 intercept is regressed onto the Level 2 predictor (Reputation). In Equation (3), the Level 1 slope is regressed onto the Level 2 predictor (Reputation):
Results
Table 1 presents descriptive statistics and intercorrelations between artist-level variables. The number of paintings sold in our database ranged from 1 (Robert Conlan, Hertha Furth, and Boris Leven) to 303 (Milford Zornes) with the average number of paintings sold being 44.49 across the 87 artists (SD = 57.73). The average price ranged from $67.00 for Carl Molno to $7,449.00 for Walter Biggs with the mean price across all 87 artists to be $659.67 (SD = 908.04). In terms of museum placements, there were 17 artists who had zero works in any of the 100 museums, with 5.74 being the mean number of museum placements across the 87 artists (SD = 10.04). In addition, the most prominently placed artists in our sample were John Costigan (61%) and Earl Horter (39%). The number of Instagram hashtag references ranged from 0 (23 artists) to 167 for Clarence Holbrook Carter, with a mean of slightly over 12 per artist (M = 12.23, SD = 25.49). The number of archival sites ranged from 0 (21 artists) to 11 for Clarence Holbrook Carter, with a mean of nearly 3 archival sites (M = 2.90, SD = 2.99). Finally, the number of book references tabulated from AskArt ranged from 1 (5 artists) to 85 for Milford Zornes with a mean of 15.70 (SD = 13.42).
Descriptive Statistics and Intercorrelations for Artist-Level Variables.
Note. N = 87. All correlations are statistically significant (p < .05).
As seen in Table 1, all pairwise correlations between artist-level variables were statistically significant. Overall, these correlations provide evidence that the reputation indicators were related to a general reputation factor as well as indicating significant collinearity between these predictors. In addition, all reputation indicators were significantly positively correlated to average price, showing that on average, artists with higher scores on the reputation indicators commanded higher prices. In addition, there was a positive correlation between the reputation indicators and the number of paintings sold, ranging from .27 to.50, indicating the items that scored higher on the reputation indicators also had more frequent sales via the online auction platforms.
Hypothesis Testing
Given the collinearity of the reputation indicators, we ran separate HLM analyses for each indicator, using the reputation indicator as the predictor of Level 1 intercepts and slopes. In each HLM analysis, all other variables were kept identical. The results of the HLM analyses used to test Hypotheses 1–6 are found in Tables 2–5. Level 1 variables related to painting characteristics were all found to be significant in the expected directions. We reference the results from Table 2 using Museum Placements as the reputation indicator. Painting size was positively related to auction prices (β2 = 1.48, p < .05), framed paintings were more expensive than unframed paintings (β3 = 154.27, p < .05), items sold in lots were less expensive than items sold individually (β4 = −146.53, p < .05), and oil paintings were related to higher prices (β5 = 534.69, p < .05). It should be noted that the coefficients reported are just like traditional regression weights, which are related to the relationship between the feature and the outcome, controlling for all other terms in the Level 1 equation. This is important to acknowledge because painting characteristics tend to be correlated. For example, oil paintings tend to be larger than paintings in other mediums. These regression terms reported here control for these colinear relationships. Results were similar in terms of magnitude and significance of each of these painting characteristics, except that the multiple versus single lot terms were not significant in two of the four HLM analyses run (using archives as reputation indicator in Table 3 and using book references in Table 4).
Slopes-as-Outcomes Model in Hierarchical Linear Modeling (Museum Total).
Note. Level 1 n = 3,772; Level 2 n = 87. Framed is coded 0 = Unframed, 1 = Framed. Multiple is coded 0 = Items sold by themselves, 1 = Items sold with other items. Painting medium is coded 0 = Other, 1 = Oil. Auction source is coded 0 = LiveAuctioneers/EBTH, 1 = eBay. Museum total = Total number of museums an artist has paintings in; Year of auction = Year that a painting was sold; Size of painting = Size of the painting in square inches; Framed = Whether the painting is framed; Multiple = Whether the painting was sold with other items or by itself; Painting medium = Whether the painting was oil-based or another medium; Auction source = Website a painting was sold on. Statistically significant values are in
Slopes-as-Outcomes Model in Hierarchical Linear Modeling (Archives).
Note. Level 1 n = 3,772; Level 2 n = 87. Framed is coded 0 = Unframed, 1 = Framed. Multiple is coded 0 = Items sold by themselves, 1 = Items sold with other items. Painting medium is coded 0 = Other, 1 = Oil. Auction source is coded 0 = LiveAuctioneers/EBTH, 1 = eBay. Museum total = Total number of museums an artist has paintings in; Year of auction = Year that a painting was sold; Size of painting = Size of the painting in square inches; Framed = Whether the painting is framed; Multiple = Whether the painting was sold with other items or by itself; Painting medium = Whether the painting was oil-based or another medium; Auction source = Website a painting was sold on. Statistically significant values are in
Slopes-as-Outcomes Model in Hierarchical Linear Modeling (Book References).
Note. Level 1 n = 3,772; Level 2 n = 87. Framed is coded 0 = Unframed, 1 = Framed. Multiple is coded 0 = Items sold by themselves, 1 = Items sold with other items. Painting medium is coded 0 = Other, 1 = Oil. Auction source is coded 0 = LiveAuctioneers/EBTH, 1 = eBay. Book ref. = Number of book references; Museum total = Total number of museums an artist has paintings in; Year of auction = Year that a painting was sold; Size of painting = Size of the painting in square inches; Framed = Whether the painting is framed; Multiple = Whether the painting was sold with other items or by itself; Painting medium = Whether the painting was oil-based or another medium; Auction source = Website a painting was sold on. Statistically significant values are in
Slopes-as-Outcomes Model in Hierarchical Linear Modeling (Instagram).
Note. Level 1 n = 3,772; Level 2 n = 87. Framed is coded 0 = Unframed, 1 = Framed. Multiple is coded 0 = Items sold by themselves, 1 = Items sold with other items. Painting medium is coded 0 = Other, 1 = Oil. Auction source is coded 0 = LiveAuctioneers/EBTH, 1 = eBay. Museum total = Total number of museums an artist has paintings in; Year of auction = Year that a painting was sold; Size of painting = Size of the painting in square inches; Framed = Whether the painting is framed; Multiple = Whether the painting was sold with other items or by itself; Painting medium = Whether the painting was oil-based or another medium; Auction source = Website a painting was sold on. Statistically significant values are in
In testing the effects of reputation on mean artist prices, we expected that the Level 2 slope terms would be significant, γ01 Hypotheses 1, 2, 3, and 4 predicted that artists with more museum placements, material available in archives, mentions in books, and social media mentions, respectively, would have higher mean prices. As can be seen in Tables 2–5, in support of each of these hypotheses, museum placements (γ01 = 26.99, p < .05), archives (γ01 = 84.32, p < .05), book references (γ01 = 22.56, p < .05), and Instagram mentions (γ01 = 11.81, p < .05) each were statistically significantly related to final price.
Hypothesis 5 predicted that artists with higher reputation indicators (i.e., more museum placements, material available in archives, book references, and social media mentions), would have more positive trajectories of prices over time. In all four HLM analyses, there was no significant intercept term relating year of auction to overall price trends, γ10. This suggests that across all artists there was no significant trend for prices to increase or decrease over time. The absence of an overall trend, however, does not preclude the possibility that for some artists, the individual trend would be for increasing prices or for others it could be decreasing prices. To test Hypothesis 5, which predicted trends over time would be related to reputation, we investigated the Level 2 slope terms γ11. Across all four HLM analyses, none of these slope terms were significant, demonstrating that none of the Level 2 reputation variables statistically significantly predicted the Level 1 slope for year of auction.
Hypothesis 6 predicted that final price would be higher for paintings sold on LiveAuctioneers and EBTH (both sites that involve some curation) compared to eBay, after controlling for the size of the painting, whether the painting was framed, whether the painting was sold in a lot, and the painting medium. As shown in Tables 2–5 the final price of paintings sold on eBay and EBTH was lower than paintings sold on LiveAuctioneers in each of the HLM models, but this relation was not statistically significant. Therefore, Hypothesis 6 was not supported.
Discussion
The rise in internet auctions and sales has provided a way for artists, galleries and auction houses to reach wider and more diverse audiences. This online vista has the potential for opening up art collecting to a broader economic market, allowing for the collecting of artists who are not just the elite few prized in top tier art museums and praised in art history textbooks. In this analysis, we focused on a group of 87 artists who all started in at least a similar space over 50 years ago, being mentioned in a book designed to promote watercolor artistry. In 1968, Norman Kent compiled a collection of watercolorists to help provide tips for up-and-coming artists, pairing a reproduction of a watercolor painting along with a page of text providing background, advice and tips. He arranged a diverse group of painters, some who were already well-established (Andrew Wyeth), a few who were already dead (Earl Horter), and some who were up-and-coming. Since the publication of that book, the reputations of these 100 artists have changed, with many of the artists fading into obscurity and others being held by museums across the country. Nearly 50% of our sample had at most one painting held in the 100 museums in the United States that we investigated, with nearly 25% having zero works held. Some of the artists’ work is bought and sold on a frequent basis long after their death, whereas for several of the artists, there were few sales. With the rise of internet auctions and the easy-to-access nature of these auctions, we expect there will continue to be a significant market for artists who have been generally forgotten by the top-tier cultural institutions. In addition, new ways of measuring cultural reputation now exist that do not require cultural gatekeepers, such as museum curators, famous galleries, and well-published art critics, to determine which artists are worth preserving. In our analysis, the number of Instagram hashtag mentions that an artist had was more highly correlated with mean sale price of an artist's work compared to the number of museum placements.
As mentioned in the introduction, there have already been several analyses of auction house prices for a large number of artists and art works. Most of that work, though, has focused on the prices of famous artists whose work is seen in museums across the world and who have already enshrined themselves into the canons of art history. One of the main contributions of this research study is broadening the analysis to include artists whose reputations are not set in stone, as well as broadening the sources of auction data to focus on online auction sites that do not have the perceived barriers of entry that top auction houses like Christie's and Sotheby's have.
In terms of the overall findings, most of our results parallel findings from other analyses of auction house sales data, with painting characteristics playing a predictable role in final prices. Larger paintings, on average, were associated with higher prices, as were oil paintings compared to other mediums. Items that were combined with other items and sold in one lot tended to be priced lower than items sold separately, 2 and items that were framed had higher prices than works that were unframed. The source of the auction, whether on LiveAuctioneers or EBTH, where work tends to be run through professional auction houses with employees who have some kind of expertise in the art market or eBay, where items can be uploaded by someone without any kind of expertise, did not make a difference. This was surprising to us as we hypothesized that buyers may value the expertise of professional auction houses more than that of eBay sellers who often cannot judge the quality of an artwork, occasionally mistaking a print for a watercolor and often unable to provide condition reports for artwork. Perhaps buyers of artwork on these auction sites can get the information needed to make decisions, or buyers tend not to care about the shape of the work. Future research should investigate the role of auction house reputation in pricing and analysis of larger data sets should be conducted to further investigate this.
Our research also provided empirical evidence for the value of reputation indicators in determining sales prices. Artists whose work was placed in more museums, cited in books, and whose biographical information is held in archives were related to higher prices than work by artists who have relatively little information out there. Our work confirms Lang and Lang's (1988) hypotheses of the importance of reputation and is consistent with other subsequent research (Braden & Teekens, 2019; Fraiberger et al., 2018; Reeneboog & Spaenjers, 2013). In fact, an exploratory analysis was run as suggested by a reviewer, and all reputation indicators loaded on a single factor. 3
A further extension of our research is that the frequency of social media mentions via Instagram was also correlated with sales price data, suggesting that cultural indicators of reputation should be broadened beyond traditional sources. This research suggests that artists who wish to preserve their reputation should now consider digital methods as well as the traditional sources. It should be noted, however, that our results may not generalize to living artists who are actively using Instagram as a tool to promote their own work. The Instagram references in our study were all done by third parties done to promote paintings for sale, to brag about artwork that the poster owned, or to educate social media users about work in a particular collection. More research should be done to further investigate the role of social media in establishing and maintaining an artist's reputation, including the influence of other types of social media. Platforms such as Facebook that are well-suited for discussion of posted content may solicit more negative reactions to artwork (Instagram only allows for likes, not dislikes) and thus may provide additional insight regarding the positive or negative reputation of an artist. Future research could examine how the popularity of an artist (regardless of positive/negative reputation) influences auction sales prices over time. In addition, social media reputation indicators should be related to other reputation indicators such as work being owned by important collectors (Braden, 2016) and artists taking part in important museum exhibitions (Braden, 2021; Fraiberger et al., 2018).
Our hypotheses related to price changes over time were not supported. In general, prices were fairly stable over the roughly 20 years of pricing data that we were able to obtain. Reputation indicators did not predict whether prices would increase or decrease over time, as we hypothesized. We explored this relationship, examining whether an artist's death date might explain increases or decreases as well as reputation, but could not find any statistical results for death date. Further work should consider investigating the role of reputation and prices over time. It may be that the roughly twenty years of pricing data did not allow us to accurately model changes over timeWith more time and auction sales, researchers will be able to model changes in reputation via social media. Finally, there were a large number of artists who had relatively few sales in the twenty-year period, making it difficult to untangle statistical effects with such few sales. As the window of auction sales data continues to increase, we hope to investigate more long-term trends.
Conclusion
There has been little research on the role of twenty-first century online auction sites in the art market and we hope that our research spurs additional studies. Traditional top tier auction houses have turned to the Internet as another outlet to market their goods and during the course of the COVID-19 pandemic, online auctions have become a norm. In addition to studying the role of online art auctions, this is one of the first studies to track the long-term reputations of artists who would be considered outside of the top tier of American artists. This group of artists include some who seem to have been left behind in contemporary writing of American art history. Many of the artists in our study are described by Bamberger: “Most artists are not that widely written about and are listed in few places other than in basic standard references, archives and files” (2019, location 2269). With the availability of accessible online auction and sales tools, the possibility of more niche collecting is possible. Although some researchers have referred to purchasers of art via eBay as interested in “sofa art” (art chosen to match the sofa) instead of investment art (Highfill & O’Brien, 2007), our results suggest an in-between category might be more appropriate, in that our online auction results show that prices are related to measures of artistic reputation, though these artists’ prices did not increase over time as would be expected from investment art.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
Author Biographies
Appendix A
List of Artists in our Study
Anderson, Edward (1923–2005)
Anderson, Harry (1906–1996)
Asplund, Tore (1903–1978)
Avery, Ralph (1906–1976)
Baumgartner, Warren (1894–1963)
Beall, C.C. (1892–1967)
Biggs, Walter (1886–1968)
Blower, David (1901–1976)
Bowes, Betty (1911–2007)
Bradshaw, Glen (1922–2013)
Broemel, Carl (1891–1984)
Burchess, Arnold (1912–1992)
Burton, Jack (1917–2006)
Bye, Ranulph (1916–2003)
Cameron, William Ross (1893–1971)
Carlin, James (1906–2005)
Carter, Clarence Holbrook (1904–2000)
Chapin, Francis (1899–1965)
Christ-Janer, Albert (1910–1973)
Cobb, Ruth (1914–2008)
Coes, Kent Day (1910–2000)
Conlan, Robert (dates unknown)
Cooke, Hereward Lester (1916–1973)
Cooper, Mario (1905–1995)
Costigan, John (1888–1972)
Cotsworth, Staats (1908–1979)
Culver, Charles (1908–1967)
Dahlberg, Edwin (1901–1984)
Detore, John (1902–1975)
Fulwider, Edwin (1913–2003)
Furth, Hertha (1907-c. 1975)
Gould, John (1906–1996)
Haines, Richard (1906–1984)
Hartgen, Vincent (1914–2002)
Heitland, W. Emerton (1893–1969)
Higgs, Robert (1916–1967)
Hoftrup, Lars (1874–1954)
Horter, Earl (1881–1940)
Johnson, Avery (1906–1990)
Johnson, Cecile (1916–2010)
Kent, Norman (1903–1972)
Khouri, Alfred (1915–1962)
Laessig, Robert (1913–2010)
Leven, Boris (1908–1986)
Margulies, Joseph (1896–1984)
Mason, Roy (1886–1972)
Maurice, Eleanor Ingersoll (1901–1995)
Messersmith, Fred (1929–2009)
Molno, Carl (1926–2000)
Monaghan, Eileen (1911–2005)
O’Hara, Eliot (1890–1969)
Ohrning, Rudolph (1930–2011)
Olaf Olson, Joseph (1894–1979)
Peirce, Gerry (1900–1969)
Perrin, C. Robert (1915–1999)
Pike, John (1911–1979)
Pitz, Henry (1895–1976)
Pousette-Dart, Nathaniel (1886–1965)
Quinn, Noel (1915–1993)
Reitzel, Marques (1896–1963)
Riley, Arthur (1911–1998)
Roberts, Morton (1927–1964)
Ross, Alexander (1908–1990)
Samerjan, George (1915–2005)
Santoro, Joseph (1908–1996)
Setterberg, Carl (1897–1983)
Sgouros, Thomas (1927–2014)
Shapiro, Irving (1927–1994)
Sisson, Laurence (1928–2015)
Smith, Jacob Getlar (1898–1958)
Stone, Don (1929–2015)
Strosahl, William (1910–2002)
Thompson, Ernest (1897–1992)
Thon, William (1906–2000)
Vallee, Jack (1921–1986)
Wagner, Ferdinand (1894–1982)
Walsh, John (1907–1994)
Wenrich, John (1894–1970)
Whitaker, Frederic (1891–1980)
White, Doris (1924–1995)
Whitney, Edgar (1894–1987)
Wilcox, Frank (1887–1964)
Wilford, Loran (1892–1972)
Winter, Lumen (1908–1982)
Wright, James Couper (1906–1969)
Youngblood, Nat (1916–2009)
Zornes, Milford (1908–2008)