PDF Version

Files

Does online bidding imply a more dispersed auction programme? This paper compares 1,937 art-associated sale listings at Christie’s, Sotheby’s and Phillips from 2021 to 2025. The outcome is the distribution of published sale locations. Three findings qualify a single account of geographical decentralization. First, New York, London, Paris and Hong Kong account for 95.5% of listings, with different city hierarchies across firms. Core-city dominance persists under shared print/edition title screens, while cross-house concentration rankings are sensitive to product scope and naming rules. Second, location coverage, distributional balance and core-city shares produce different comparisons between online-only and auctioneer-led sales. No common direction describes all houses and measures. Third, incomplete bidding-format information leaves some conclusions better established than others. For Sotheby’s, the main evidence specification bounds the online-only core share at 90.9%–95.5%, while allowing either sign for its difference from auctioneer-led sales. A stratified audit finds sale-specific physical-preview evidence for thirteen of thirty online-only events, giving some location labels a documented service interpretation. Across these checks, core-city dominance is more stable than precise firm rankings or format comparisons. The study describes auction programmes rather than identifying the causal effect of online bidding.

1 Introduction

International auction houses associate many online-only sales with familiar cities. This observation raises a question more precise than whether the internet eliminates distance: how are their sale programmes allocated across places, and does an online bidding format imply a more geographically dispersed allocation? An auction can accept bids remotely while remaining assigned to a particular city. Understanding its geography therefore requires distinguishing the bidding mechanism from the distribution of the listed programme.

This paper compares Christie’s, Sotheby’s and Phillips from 2021 to 2025. Its main outcome is the city distribution of published sale listings: the locations to which each house assigns its recorded auctions. These labels do not map every specialist, warehouse, buyer or object. A separate physical-preview audit checks whether selected online-sale locations correspond to a documented local service. The three houses are major international intermediaries selected for comparison, rather than a claim about the first three positions in every global ranking. All observations come from public auction records and event information.

The closest previous discussion already connects online auctions to cities and physical previews (Van Miegroet, Alexander and Leunissen, 2019). The contribution here is to show what a common five-year comparison changes about that account. It establishes a shared pattern of concentration while identifying three distinctions that a single-house example or a simple online-versus-offline comparison can obscure.

First, a common urban core accommodates different firm configurations. New York, London, Paris and Hong Kong account for 1,849 of 1,937 retained listings, or 95.5%. Yet Paris has a substantial place in Christie’s and Sotheby’s programmes and only one observation at Phillips. Phillips’ activity is overwhelmingly assigned to New York, London and Hong Kong. Concentration is shared; the role of each city is not.

Second, online-only bidding has no single observed geographical ordering relative to auctioneer-led sales. Coverage counts the places reached, core shares measure dependence on a specified group, and concentration indices capture balance across the whole distribution. These dimensions need not move together. In Sotheby’s classified records, online-only auctions have a slightly lower core share but a higher HHI. The comparison across houses also differs, and some pooled orderings reverse within individual years. The evidence replaces an undifferentiated question about decentralization with separate, observable comparisons.

Third, classification gaps limit some conclusions more than others. Sotheby’s unresolved formats prevent a definitive ranking of the two formats’ core shares. Nevertheless, even unfavourable allocations preserve a high online-only core share. The bounds make the distinction substantive: incomplete format information can leave the direction of a difference unknown while still establishing a strongly concentrated programme.

The preview audit adds a limited interpretation of the location measure. Thirteen of thirty sampled online-only sales have event-specific physical-viewing evidence; all twelve that permit an exact city comparison match their published sale city. This supplementary result connects some labels to an advertised service. The main advance is the comparative account: city concentration is common across the three houses, while their hierarchies, format comparisons and evidential limits differ.

2 Related Literature and Research Questions

Van Miegroet, Alexander and Leunissen (2019) provide the closest precedent. They examine online and offline auction activity across cities and discuss physical previews associated with online-only auctions. Their distinction between online-only mechanisms and auctions that merely accept internet bids is essential here. The present study builds on this account by comparing three houses over the same five years and asking which geographical patterns are shared, which depend on the firm or measure, and which survive incomplete format classification.

Other work connects art markets with particular urban settings. Schuetz and Green (2014) examine gallery clustering and affluent demand in New York, while Molho (2021) compares the regional-hub roles of Singapore and Hong Kong. Ortiz Arellano (2021) considers auction houses as global firms, and Moine (2022) compares the geographic markets of Christie’s, Sotheby’s and Phillips. These studies motivate attention to differentiated city roles, while using geographical scales and market outcomes distinct from the present sale-count comparison.

The intermediary provides an economic reason to examine the programme rather than only the bidding channel. Auction houses supply expertise, estimates, cataloguing, consignment services, buyer relationships and transaction administration. Platform economics describes how attracting buyers can make an intermediary more valuable to sellers and vice versa (Rochet and Tirole, 2003). An established city programme may support these relationships even when bids arrive remotely. Agglomeration mechanisms—sharing facilities, matching works with expertise, and learning—offer related reasons for concentrating complementary activities (Duranton and Puga, 2004). Different category specializations and client networks can also make the same city more important to one house than another.

Research on auction pricing and intermediation clarifies why these functions matter. Beggs and Graddy (2009) document anchoring in auction prices and estimates; their earlier study examines lower subsequent returns following failure to meet a reserve price (2008). Thompson (2010) emphasizes branding and major intermediaries in the contemporary art market. These contributions concern information, expectations and reputation within exchange. They provide background for asking how auction programmes are arranged, rather than direct explanations of the observed city shares. The proposed mechanisms are an interpretive framework; the present data do not distinguish their causal importance.

Three exploratory research questions organize the evidence. RQ1: How far do the houses share concentration in a small set of cities, and where do their city configurations differ? RQ2: How do online-only and auctioneer-led programmes compare in location coverage, distributional balance and dependence on core cities across houses and years? RQ3: Which conclusions remain established when bidding formats cannot be classified completely? The preview audit is a supplementary measurement check: it asks whether some published online-sale locations correspond to documented physical viewing. These questions developed during the initial Christie’s investigation and subsequent comparative extension; they were not preregistered.

3 Data & Measurement

The geographic dataset combines public listing archives for Christie’s, Sotheby’s and Phillips. Each house-specific sale identifier counts once, irrespective of revenue, the number of lots or the number of bidding sessions it contains. The outcome therefore describes programme allocation rather than turnover, successful transactions, employment or artistic quality. It also differs from the geography of buyers and consignors, whose residences are not observed.

The harmonized time window is 2021–2025, while category selection follows documented house-specific art-associated departments or query groups. The sample includes some mixed collections and art/design events; it is not a perfectly matched fine-art basket. A geographically identified source frame contains 1,940 listings. Three exclusions improve alignment with the research unit: Phillips’ natural-history sale Out of This World, Sotheby’s Impact Gala wrapper that combines separately listed subevents, and the museum-managed Art Bash fundraiser for which Sotheby’s supplied an auctioneer. The cleaned sample contains 669 Christie’s, 1,003 Sotheby’s and 265 Phillips observations.

Table 2 makes the category asymmetry explicit. Christie’s retains department labels, Phillips supplies department combinations, and Sotheby’s frozen query set identifies the categories searched without preserving each record’s memberships. A fully harmonized three-house department comparison cannot be recovered from those fields. An additional shared title screen selects advertised prints and editions using the same expressions and exclusions across firms, with a documented photographic-print exception. It retains related works on paper only alongside an eligible print/edition expression; a narrower version removes those mixed titles. Appendix F records the rules and all decisions. This is a comparable-title sensitivity, not a census of identical products.

Years follow the published final or closing date, subject to a documented Phillips date correction checked in sensitivity analysis. Cities remain published sale-associated labels. In particular, United Arab Emirates is retained as a regional label rather than silently recoded to Dubai. The four comparison cities—New York, London, Paris and Hong Kong—were inherited from the initial Christie’s case and are not reselected to maximize the three-house result. The set therefore predates the cross-house extension but was informed by observed Christie’s records. Ranked-city and leave-one-city-out summaries test how much the descriptive conclusion depends on that choice.

The main format distinction is between auctioneer-led live bidding and online-only mechanisms. The former may accept remote bids and does not guarantee permission to attend a room. Sotheby’s September 2024 Modern and Contemporary African Art sale, for example, provided live bidding without room participation (Sotheby’s, 2024). Online-only means that the evidence identifies a wholly online bidding mechanism. Most such events are timed auctions, but the category also includes an explicitly documented all-online sealed system. A catalogue on the internet, a multi-day date range or a preview address alone cannot identify the bidding format.

Christie’s and Phillips have format labels throughout their selected samples. Sotheby’s requires a separate evidence treatment. Three nested scopes distinguish auction-house or identified event-partner evidence; additional original reports and previews; and additional attributed archives or listings. The middle scope is the main specification. It supports 351 auctioneer-led and 417 online-only classifications, leaving 235 records unresolved. The narrower and broader scopes are sensitivities, not equally authoritative sources. An original preview documents an announced mechanism rather than independently proving its realization.

All geographic records remain in the concentration analysis regardless of format uncertainty. Classification evidence may be easier to recover for prominent collections or earlier years, so supported formats are not assumed representative of unresolved ones. Record-level URLs, matching rationales, exclusions and alternative classifications remain in the accompanying data.

Completeness refers to the enumerated public archive frame; deleted, unpublished or out-of-scope sales are not observed.

The physical-preview audit uses a separate sampling step within the 774 known online-only events. Two records are randomly selected without replacement from each of the fifteen house-year cells, producing 30 observations. The draw is fixed before retrieving preview evidence, and missing information never triggers replacement. The audit preserves the geographic and format analysis as frozen, even where subsequent inspection provides additional corroboration. It tests the interpretation of selected city labels, not the completeness of the entire online-only population.

4 Empirical Strategy

The main outcome is the share of a house’s listings assigned to the four comparison cities. This fixed-core measure addresses whether a recognizable urban centre persists. Location counts measure coverage. The Herfindahl–Hirschman index measures balance across all recorded labels, including differences within the core. Shannon entropy provides a complementary dispersion measure, while the reciprocal of HHI expresses concentration as a number of equally active locations. For location shares pᵢ, the definitions are:

These measures answer different questions. Adding a location with one sale expands coverage immediately but may have little effect on balance. Moving activity from Paris to New York leaves the fixed-core share unchanged but can increase HHI. A fall in the core share therefore need not coincide with lower overall concentration. The analysis reports annual observations before pooling, and pooled indices are recalculated from combined counts rather than averaged across years. Both sale-weighted and equal-house pooling are checked.

Observed location counts depend on programme size. An exact rarefaction calculation compares expected distinct labels when both formats contribute the smaller of their two listing counts, adapting the without-replacement method of Hurlbert (1971). A second calculation matches counts separately within each year and measures the expected five-year union of labels. These standardize exposure within the recorded frame; they do not recover unclassified formats. Appendix E gives definitions and results.

Association diagnostics ask whether houses distribute events similarly across locations and whether format is associated with location within each fully classified house. Pearson table statistics and Cramér’s V summarize these differences. Because several expected cells are small, conditional Monte Carlo permutations provide reference distributions. House labels are shuffled within year. Formats are shuffled within each house’s year-by-category-family cells, preserving these aspects of programme composition. The procedure uses 9,999 draws and the add-one p-value calculation recommended by Phipson and Smyth (2010).

These are exploratory conditional diagnostics, not causal tests. Exchangeability is a reference assumption rather than random assignment, and stratification cannot remove all product differences or dependence among related collections. Effect sizes and observed shares remain more informative than a binary significance label.

For Sotheby’s, bounds address the unresolved formats without assuming random missingness. Let n be the known number of events in a target format and k its known core count. Let uC and uP denote unresolved core and peripheral records. Assigning all unresolved peripheral records, and no unresolved core records, to the target gives its lowest feasible core share. Reversing this assignment gives its highest:

Records not assigned to the target can belong to another mechanism or remain outside the binary comparison. The construction consequently does not require every unknown event to be either live or online-only. Bounds are conditional on accepted classifications and geography; they do not cover source errors or absent events. They are feasible-value ranges, not confidence intervals. Jointly feasible assignments also bound the difference between the online-only and auctioneer-led core shares.

The preview audit records affirmative physical-viewing evidence, an explicit statement excluding physical viewing, or insufficient information. A general contact address, blank viewing heading or virtual appointment is insufficient. Selected highlights qualify, but do not establish that every lot was exhibited. Conservative coding requires a viewing arrangement expressly linked to the selected sale. Two cases with named membership in a shared exhibition programme are retained separately in an extended coding. This precise programme distinction was clarified during adjudication after retrieval, rather than independently preregistered.

Preview counts describe documentation within the fixed sample. Known selection probabilities would permit design weighting to the accepted online-only frame, despite equal allocation across house-years. Such weights cannot determine whether unresolved events actually offered previews, or represent events excluded because their bidding formats remain unknown. The analysis therefore reports documented confirmations and unresolved cases rather than estimating actual preview prevalence. A GDP regression addresses a different question and would require a comparable universe of non-host cities; it is outside this design.

5

A Common Urban Core with Different Hierarchies

The main sample contains 1,937 listings across thirteen published location labels. New York, London, Paris and Hong Kong contain 1,849, or 95.5%. The result appears at every house: 642 of 669 Christie’s events, 943 of 1,003 Sotheby’s events and 264 of 265 Phillips events. High pooled concentration therefore does not arise solely from combining one centralized firm with two dispersed ones. RQ1 therefore yields a shared pattern of concentration in the full selected programmes.

The hierarchy within the core differs. New York is the largest location at each house, with 270 Christie’s, 374 Sotheby’s and 115 Phillips listings. London contributes 176, 292 and 98, respectively. Paris is substantial at Christie’s and Sotheby’s, with 96 and 151, but appears only once at Phillips. Hong Kong contributes 100, 126 and 50. Phillips’ New York, London and Hong Kong programmes together account for 263 of its 265 events. Its Paris and Singapore observations tie below these three centres.

Thus, the extension beyond Christie’s supports a common pattern of concentrated organization, not an identical four-city model. Describing Paris as equally central to all three firms would conceal an important organizational difference. The data also do not rank the cities’ artistic quality: an auction listing reflects the house’s programme and category selection, rather than the creative output of the host city.

HHI is 0.2756 at Christie’s, 0.2634 at Sotheby’s and 0.3607 at Phillips. Their effective location counts are 3.63, 3.80 and 2.77, compared with observed footprints of eight, ten and five labels. Entropy is correspondingly lowest at Phillips: 1.0869, against 1.4569 and 1.5201. These measures show why a list of international locations can overstate the balance of an organizational network. A peripheral label can be present while accounting for very little activity.

The core’s prominence also persists annually. Its share ranges from 93.7% to 98.3% at Christie’s, 92.8% to 96.0% at Sotheby’s and 98.0% to 100% at Phillips. The upper panels of Figure 2 show that pooled dominance is repeated in every observed house-year. The lower panels separate this persistent core from the more variable comparison of format-specific concentration, discussed next.

Alternative cores clarify the result’s scope. The three largest locations account for 81.6% of Christie’s, 81.5% of Sotheby’s and 99.2% of Phillips listings. Sotheby’s top three include Paris rather than Hong Kong. Removing New York from the fixed core reduces the pooled share to 56.3%, so 95.5% is specific to the four-city definition. Table A8 reports all deletions. HHI and effective counts describe the whole geographic distribution without selecting a core boundary.

The house-by-location table gives Cramér’s V of 0.180 and a conditional permutation p-value of 0.0001. Under the year-preserving reference scheme, the observed distributions differ. This is compatible with a shared core, since firms can concentrate in similar places with different weights. Equal-house pooling produces a core share of 96.5%, compared with the sale-weighted 95.5%, so the common result survives giving the smaller Phillips programme equal weight. Neither calculation represents the global market share of the three businesses.

6 Online Formats and the Persistence of Core Cities

Christie’s and Phillips permit complete within-frame comparisons of the two formats. Christie’s has 401 auctioneer-led and 268 online-only sales. Online-only sales involve seven location labels, compared with six for auctioneer-led sales, and have a lower pooled HHI: 0.2650 against 0.2855. Yet 244 of the 268 online-only events, or 91.0%, remain in the core. Additional peripheral activity and a dominant metropolitan centre coexist.

Phillips has 176 auctioneer-led and 89 online-only sales. Each format involves four locations, while online-only HHI is lower at 0.3392 against 0.3754. Their core shares are 98.9% and 100%. Here a more even online distribution does not require more locations. Raw footprints show wider online-only coverage at Christie’s and a tie at Phillips. Both retain a dominant core, despite different coverage comparisons.

Annual and conditional comparisons further qualify the pooled results. Figure 2 shows that Christie’s online-only HHI is higher in 2021 and 2025, despite being lower over the full period. Phillips has a near-tie with a small reversal in 2023 (online minus auctioneer-led HHI = 0.0012). Sotheby’s online-only HHI is higher in every classified annual subset; its later comparisons must be read alongside the coverage row. Christie’s format-by-city association has Cramér’s V of 0.238 and a conditional p-value of 0.0001. Phillips’ corresponding values are 0.136 and 0.3997. The latter is not evidence that the distributions are identical. These diagnostics and HHI orderings describe different aspects of the recorded programmes; neither estimates the effect of switching an otherwise identical auction online.

Sotheby’s known-format subset contains 351 auctioneer-led and 417 online-only events. Within this classified subset, online-only sales cover eight location labels compared with nine for auctioneer-led sales. Their observed HHIs are 0.2870 and 0.2634, respectively, while their core shares are 93.0% and 94.0%. The higher online-only HHI despite a slightly lower core share illustrates the analytical distinction between core dominance and balance among locations.

These Sotheby’s figures require particular care. Main-scope format coverage is 99.6% in 2021 and 100% in 2022, but only 59.3%, 52.8% and 58.4% in 2023–2025. Supported records therefore disproportionately represent earlier years. Reports may also favour more prominent sales. The observed format rows are valid descriptions of supported events, but cannot be treated as complete full-period estimates. Appendix tables retain the annual coverage and the two alternative source scopes.

Matching listing counts gives expected live/online-only location counts of 5.56/7.00 at Christie’s, 9.00/7.84 within Sotheby’s known subset, and 3.51/4.00 at Phillips. Matching counts within years preserves these directions (Table A7). Phillips’ raw four-location tie becomes a modest expected-coverage difference, partly because subsampling may omit its singleton Paris live sale. Online-only coverage is therefore wider at Christie’s and Phillips after count matching, but the ordering differs within Sotheby’s classified subset.

Bounds permit a stronger statement about core dominance than about the ordering of formats. Of the 235 unresolved Sotheby’s records, 225 are in the core and ten elsewhere. Under the main evidence scope, the auctioneer-led core share can range from 91.4% to 96.4%, and the online-only share from 90.9% to 95.5%. Even assigning every unresolved peripheral record to online-only, with none of the unresolved core records, leaves approximately nine in ten online-only events in the core.

The online-only-minus-auctioneer-led core-share difference nevertheless ranges from −5.49 to +4.07 percentage points. Both endpoints are attainable through opposing allocations of unresolved core and peripheral records. Narrower primary-only evidence produces a wider interval, −6.70 to +6.36 points; adding attributed sources narrows it to −5.05 to +2.48. All three scopes permit both signs. The primary-only online lower bound is 89.90%, so a literal above-90% assertion is specific to the main scope. Across the evidence standards, the more defensible general statement is that roughly nine in ten remain in the core at the lower bound.

The nearly complete Sotheby’s 2021–2022 comparison gives an additional check: 457 of 458 events have supported formats, and online-only HHI remains higher than auctioneer-led HHI, at 0.2913 against 0.2542. Christie’s and Phillips show the opposite pooled ordering in those years. RQ2 and RQ3 thus yield different kinds of answers: format comparisons vary across measures, firms and years, whereas high core shares survive much of the uncertainty about Sotheby’s formats.

7 Physical Previews and the Meaning of City Labels

The preview audit asks whether online-sale cities can be connected to a tangible service. The sample contains ten events per house, with two from every year. Every selected event has a recoverable official body, but recoverability of a sale page does not guarantee information about physical viewing. Sources include official sale descriptions and releases, as well as dated exhibition records embedded in public catalogue pages. Where available, exhibition location is read separately from the sale-city field.

The thirty events fall into three mutually exclusive documentation categories: thirteen with event-specific physical-preview evidence, two with evidence linking the sale to a shared exhibition programme, and fifteen with neither form of sufficient confirmation. The direct confirmations comprise four Christie’s, eight Sotheby’s and one Phillips event; both programme-only cases are Christie’s sales. No selected record explicitly rules out physical viewing. The conservative confirmation count is thirteen, rising to fifteen when shared-programme evidence is accepted.

Of the thirteen main confirmations, twelve permit exact city comparison, and every one agrees with the original sale-associated city. The remaining case advertises a Dubai preview, while the frozen source label is United Arab Emirates. It is included in the preview count but treated only as region-consistent. This preserves the original geographic definition rather than creating a perfect match by changing it after inspection.

The sources identify specific arrangements. Christie’s July 2022 Contemporary Edition sale advertised in-person viewing at its Rockefeller Center galleries. Phillips’ 2022 Whitechapel Gallery Art Icon Charity Auction gave a London address under its viewing information. Sotheby’s May 2022 American Scene sale advertised a York Avenue preview and in-person appointments (Christie’s, 2022; Phillips, 2022; Sotheby’s, 2022a). Such records connect remote bidding with a physical point of contact. They do not imply that bidders had to attend it.

The link is not exclusive to the four-city core. Sotheby’s Marisa Perret-Rezzonico sale advertised selected works in Zurich in March 2023 (Sotheby’s, 2023), while Christie’s Middle Eastern sale advertised selected highlights in Dubai (Christie’s, 2023). A peripheral programme can therefore also combine online bidding and local viewing. The economic distinction is between bidding format and the spatial organization of services, rather than between core cities with physical services and peripheral places without them.

A further official example, outside the fixed pilot, is Phillips’ SPOTLIGHT online sale, UK040223. Its announcement links bidding on 7–14 July 2023 to viewing at the London Berkeley Square galleries on 7–13 July (Phillips, 2023b). The sale already belongs to the main geographic dataset but was not drawn for the preview audit. It corroborates the coexistence of online bidding and announced physical viewing without changing the pilot’s selection or confirmation counts.

These examples connect particular sale labels to announced physical viewing. The audit does not establish that undocumented events lacked previews. Unequal sampling probabilities could be adjusted using the known house-year frame counts, but weighting would still leave their actual service status unresolved and would not recover Sotheby’s unclassified online-only events. The reported counts therefore concern documentary confirmation, while city agreement describes the confirmed cases with comparable labels. Neither provides a ranking of the houses’ actual preview provision.

8 Robustness and Economic Interpretation

Several alternative definitions preserve the core result. Retaining the three excluded records leaves each house’s five-year core share essentially unchanged. A broader Christie’s art-tag rule increases its sample to 769 and gives a 95.8% core share. Excluding sixteen identified mixed Phillips sales leaves 249 events and a 99.6% share, although HHI rises to 0.3818. Product scope therefore matters for exact concentration comparisons even when the broader urban-core conclusion is stable.

The shared print/edition title screen adds a product-scope comparison across all three houses (Table A9). It retains 60 Christie’s, 39 Sotheby’s and 41 Phillips listings. Core shares remain 100%, 94.9% and 100%, while HHIs of 0.4594, 0.4556 and 0.4539 are closely clustered around 0.46. The smaller cross-house gap is more informative than their slight numerical ranking. The narrower exclusion of titles explicitly naming related works on paper leaves 50, 39 and 21 observations and HHIs of 0.4856, 0.4556 and 0.5510. The company ranking therefore depends on the scope even within these advertised print/edition programmes. Core dominance survives both definitions more clearly than an exact cross-firm HHI ordering.

This comparison strengthens one finding while limiting another. The whole-programme HHI ordering does not survive every restriction of the listing frame, but the screen does not estimate how much of that difference is caused by composition. Different naming practices, lots and prices remain. The narrower rule removes twenty Phillips New York listings explicitly naming works on paper; the change reflects title conventions as well as possible product differences. Format cells are also uneven: the main screen includes only three classified Sotheby’s auctioneer-led sales and four Phillips online-only sales; eleven Sotheby’s formats remain unknown. These cells are reported for transparency, without a general within-product format ranking or a new significance test.

Collapsing multiple listings within the same house, location and final date reduces counts to 567 Christie’s, 802 Sotheby’s and 249 Phillips units. Core shares remain 95.2%, 92.5% and 99.6%. This check reduces the influence of dividing one day’s programme across several catalogues. It does not turn the observations into staff hours or physical occupancy, but shows that catalogue granularity alone does not account for core dominance.

The documented Phillips date alternative moves one event between 2023 and 2024, with annual HHI changes below 0.0003 and no five-year effect. Restricting the period to 2023–2025 gives core shares of 97.4%, 93.4% and 100%. Excluding the two Sotheby’s Sealed-branded listings also preserves the qualitative core result and permits either sign of the format-share gap. These checks address specific measurement risks rather than selecting definitions for a preferred direction.

The economic interpretation is compatible with auction houses concentrating complementary activities while accepting bids remotely. Specialist capacity, facilities and repeated client relationships provide plausible reasons for an established city programme. The geographical comparison shows that such concentration can take different forms across firms; the preview audit identifies one local activity associated with selected online sales. Neither result determines which proposed mechanism matters most. The main descriptive contribution is the combination of a common metropolitan pattern with different company allocations and different format comparisons.

The scope of the evidence is correspondingly specific. Sale counts weight programmes rather than turnover or employment, and the selected category frames do not represent every national or regional auction market. Published locations are not a complete map of operating functions. Historical source recovery also limits format comparisons and the interpretation of preview documentation. Finally, no pre-digital counterfactual or exogenous format change is observed: pandemic disruption, demand, collections and strategy can affect both cities and mechanisms. These limitations concern the interpretation and generalization of the results, not the arithmetic of the recorded distributions.

A next study could follow preview venues, specialist departments and sale locations alongside remote-bidding availability. Lot counts or transaction values could test whether the count-based pattern extends to economic volume. Digital participation and local organization should be measured separately before inferring geographical dispersion.

9 Conclusion

The 2021–2025 programmes of Christie’s, Sotheby’s and Phillips share a concentrated geography, with different city configurations. Four cities account for 95.5% of the retained listings, while Paris plays a markedly different role at Phillips from its role at the other two houses. A shared print/edition title screen preserves high core shares and brings the three HHIs close together; a narrower screen changes their ranking again. Core-city dominance is therefore more stable than the precise ordering of firms. These screens compare advertised programmes, without fully controlling product composition.

Online-only bidding has no single geographical ordering. Coverage, balance and core shares measure distinct features, and their comparisons vary across houses and years. Incomplete Sotheby’s classification supports a high online-only core share even when it cannot determine which format has the higher share. Classification uncertainty limits format comparisons; product and title scope limit firm rankings. A dominant city core is the more stable finding across these checks.

Documented previews give a subset of published online-sale locations a concrete service meaning. The contribution extends earlier accounts of online auctions and physical arrangements by separating shared city concentration from firm configurations and comparisons that the available records cannot settle. Bidders need not assemble in the city for the recorded programme to retain an urban core.

References

Beggs, A., & Graddy, K. (2008). Failure to meet the reserve price: The impact on returns to art. Journal of Cultural Economics, 32(4), 301–320. https://doi.org/10.1007/s10824-008-9077-8

Beggs, A., & Graddy, K. (2009). Anchoring effects: Evidence from art auctions. American Economic Review, 99(3), 1027–1039. https://doi.org/10.1257/aer.99.3.1027

Christie’s. (2021–2025). Auction results, annual archive pages. Snapshots retrieved 5 October 2026. https://www.christies.com/en/results/2021/01 ; corresponding 2022–2025 pages and individual sale URLs are preserved in the source package.

Christie’s. (2022). Contemporary Edition, New York, 8–19 July 2022. Official online auction overview. https://onlineonly.christies.com/s/contemporary-edition/overview/3263?sc_lang=en

Christie’s. (2023). Modern and Contemporary Middle Eastern Art, 1–16 May 2023. Official online auction overview. https://onlineonly.christies.com/s/modern-contemporary-middle-eastern-art/overview/3416?sc_lang=en

Duranton, G., & Puga, D. (2004). Micro-foundations of urban agglomeration economies. In J. V. Henderson & J.-F. Thisse (Eds.), Handbook of Regional and Urban Economics (Vol. 4, pp. 2063–2117). North-Holland. https://doi.org/10.1016/S1574-0080(04)80005-1

Hurlbert, S. H. (1971). The nonconcept of species diversity: A critique and alternative parameters. Ecology, 52(4), 577–586. https://doi.org/10.2307/1934145

Kong, Y. T. (2024, June 25). A unique Kusama statue estimated at US$1.5m heads to Sotheby’s “Sealed” auction without reserve. The Value. https://en.thevalue.com/articles/sothebys-auction-house-sealed-london-yayoi-kusama-statue-polka-dots-mirror-room

Moine, C. (2022, February 1). Christie’s, Sotheby’s, Phillips: A look at their relative markets. ArtMarket Insight, Artprice. https://www.artprice.com/artmarketinsight/christies-sothebys-phillips-a-look-at-their-relative-markets

Molho, J. (2021). Becoming Asia’s art market hub: Comparing Singapore and Hong Kong. Arts, 10(2), 28. https://doi.org/10.3390/arts10020028

Ortiz Arellano, E. (2021). Geografía del mercado internacional de obras de arte: Las casas de subastas como empresas globales. Proceedings of the 25th Congreso Internacional de Ciencias Administrativas, FCA–UNAM. https://repositorios.fca.unam.mx/investigacion/memorias/2021/14.01.pdf

Phillips. (2021–2025). Past auctions and results calendar. Snapshots retrieved 5 October 2026. https://www.phillips.com/auctions/past ; https://www.phillips.com/calendar/results

Phillips. (2022). The Whitechapel Gallery Art Icon Charity Auction, UK090122. Official auction overview. https://www.phillips.com/auction/UK090122/overview

Phillips. (2023a). Editions and works on paper, UK030623. Official auction results. https://www.dist.phillips.com/content/web/auction-results/auctionResultsFile_UK030623.pdf

Phillips. (2023b, June 29). Phillips presents SPOTLIGHT: Photographs from a private London collection online auction from 7 to 14 July 2023. Official press release. Retrieved 6 October 2026. https://www.phillips.com/press/release/phillips-presents-spotlight-photographs-from-a-private-london-collection-online-auction-from-7-to-14-july-2023

Phillips. (2025). Out of this world, NY011725. Official auction overview. https://www.phillips.com/auction/NY011725/overview

Phipson, B., & Smyth, G. K. (2010). Permutation P-values should never be zero: Calculating exact P-values when permutations are randomly drawn. Statistical Applications in Genetics and Molecular Biology, 9(1), Article 39. https://doi.org/10.2202/1544-6115.1585

Rochet, J.-C., & Tirole, J. (2003). Platform competition in two-sided markets. Journal of the European Economic Association, 1(4), 990–1029. https://doi.org/10.1162/154247603322493212

Schuetz, J., & Green, R. K. (2014). Is the art market more bourgeois than bohemian? Journal of Regional Science, 54(2), 273–303. https://doi.org/10.1111/jors.12068

SFMOMA. (2025, March 13, updated April 10). Zack Fox to headline SFMOMA’s Art Bash on April 23. https://www.sfmoma.org/press-release/celebrate-creativity-community-at-sfmomas-art-bash-on-april-23/

Sotheby’s. (2021–2025). Auction results index, category filters and individual sale pages. Snapshots retrieved 5 October 2026. https://www.sothebys.com/en/results

Sotheby’s. (2022a). The American Scene including important photographs from the Helene Wurlitzer Foundation. Official auction overview. https://www.sothebys.com/en/buy/auction/2022/the-american-scene-including-important-photographs-from-the-helene-wurlitzer-foundation

Sotheby’s. (2022b). Sotheby’s Impact Gala 2022: Auctions in support of Instituto Terra. Official digital catalogue. https://www.sothebys.com/en/digital-catalogues/sothebys-impact-gala-2022-auctions-in-support-of-instituto-terra

Sotheby’s. (2023). A Friend to Artists: The Marisa Perret-Rezzonico Collection. Official auction overview. https://www.sothebys.com/en/buy/auction/2023/a-friend-to-artists-the-marisa-perret-rezzonico-collection

Sotheby’s. (2024). Modern & Contemporary African Art, London, 27 September 2024. Official auction details. https://www.sothebys.com/en/buy/auction/2024/modern-contemporary-african-art-2

Thompson, D. (2010). The $12 million stuffed shark: The curious economics of contemporary art. St. Martin’s Griffin. (Original work published 2008.) https://us.macmillan.com/books/9780230620599/the12millionstuffedshark/

Van Miegroet, H. J., Alexander, K. P., & Leunissen, F. (2019). Imperfect data, art markets and Internet research. Arts, 8(3), 76. https://doi.org/10.3390/arts8030076

©2026 Ryan Financial Daily