Differentiated Urban Effects Around a Large-Scale Entertainment Arena: Evidence from the O₂ and Greenwich Peninsula, London

Abstract

Large-scale entertainment venues are often positioned as anchors of mixed-use redevelopment, yet their surrounding effects may not unfold uniformly across indicators. This study examines differentiated urban effects around The O₂ and Greenwich Peninsula in London by comparing treatment and control Middle Layer Super Output Areas (MSOAs) within the Borough of Greenwich. Using Office for National Statistics housing data for 1995–2025 and NOMIS business-count data for 2016–2024, the analysis combines descriptive comparison, pre-2007 comparability assessment, static difference-in-differences, annual event-study estimation, and total and sector-level business analysis. The housing results show a persistent locational premium in the treatment areas. Static DID estimates were not statistically significant, but annual event-study estimates suggest relative strengthening in selected post-opening years. The business results show a substantially larger business base in the treatment areas but only a limited aggregate relative growth advantage. Sectoral patterns are uneven: retail and arts-related activities perform more strongly, accommodation and food services show a modest advantage, and office-oriented services show little relative difference. Overall, the findings indicate differentiated urban effects rather than uniform neighborhood transformation and suggest that venue-centered redevelopment should be evaluated through multiple indicators and with attention to sectoral composition.

1. Introduction

Large-scale entertainment venues are increasingly positioned within broader processes of urban redevelopment rather than treated as stand-alone leisure facilities [1,2,3]. Recent studies further connect professional sports and entertainment venues to wider revitalization agendas [4], visitor-oriented placemaking, and complementary amenities [5]. More recent review work also links sport-led projects to place-branding strategies [6], while emerging built-environment research suggests that stadium development can reshape surrounding urban form and functional composition over time [7]. In this broader view, the significance of a major venue lies not only in its event-generating capacity, but also in its potential to shape the direction and intensity of nearby urban change.

Yet the urban effects of such projects remain strongly contested. Longstanding critiques have questioned whether stadiums and arenas generate broad-based local economic development, arguing instead that many reported gains are limited, redistributed, spatially uneven, or difficult to detect in aggregate indicators [8,9]. More recent survey and policy-oriented reviews reach a similar conclusion, emphasizing that professional sports venues rarely function as reliable engines of generalized urban growth and that public claims about jobs, income, and wider economic uplift are often overstated [10,11]. Even where localized gains are observed, they may remain modest in scale and partly offset by displacement or crowding out of other spending and activity [12]. The key issue, therefore, is not simply whether a major entertainment venue matters, but how its surrounding effects are distributed across different dimensions of urban change.

Existing research has often examined venue-centered change through one outcome domain at a time. In London, sports stadia and mega-event announcements have been linked to nearby residential price premiums and locational revaluation [13,14]. Evidence from other settings likewise suggests that sports facilities can affect surrounding housing markets under particular proximity and amenity conditions [15,16,17]. A separate body of work examines localized commercial activity and adjacent business concentration [18]. What remains less clear is whether housing-value change and local business change move together within the same redevelopment setting or diverge in substantively important ways. This study addresses that issue through the concept of differentiated urban effects, used here to denote the possibility that these dimensions follow different directions, magnitudes, and temporal paths. A venue-centered project may therefore reinforce locational value without generating an equally broad-based pattern of local commercial expansion.

The O₂ and Greenwich Peninsula provide a particularly suitable case through which to examine this problem. The O₂ is one of London’s most prominent entertainment destinations, but it is also embedded within the longer-term redevelopment of Greenwich Peninsula as a mixed-use urban district rather than functioning as an isolated venue [19]. Official planning documents consistently frame the peninsula as a major redevelopment area combining residential growth, commercial activity, leisure uses, and broader place-making objectives [20,21]. This strategic role is reinforced by its position within London’s Opportunity Area framework, which identifies Greenwich Peninsula as a significant site of planned growth and transformation [22]. The case is therefore analytically valuable because it allows the study to examine whether housing-value change and local business change evolve in a consistent way within a venue-centered redevelopment setting, or whether they diverge across indicators.

Against this background, the study addresses three questions:

(1)

Do the areas surrounding The O₂ exhibit a persistent locational-value premium relative to internal Greenwich controls, and how should that premium be interpreted over time?

(2)

Do the O₂-adjacent areas display stronger local business performance, and does that pattern differ between aggregate and sector-specific business measures?

(3)

What does the divergence between housing-value change and business change reveal about differentiated urban effects in a venue-centered redevelopment setting?

To address these questions, the paper adopts a within-borough comparative design that combines long-run housing analysis with later-period total and sectoral business analysis. Its contribution lies in evaluating multiple indicators within a single venue-centered redevelopment setting, distinguishing aggregate from sector-specific business outcomes, and treating divergence across indicators as a substantive analytical finding rather than as a residual inconsistency. In doing so, the study argues that the surrounding effects of a major entertainment anchor are better understood as differentiated across dimensions of urban change than as a single and uniform process of neighborhood transformation.

2. Literature Review

2.1. Venue-Centered Redevelopment and Selective Local Effects

Within the redevelopment literature, sports and entertainment venues are typically interpreted as anchors that help organize broader spatial strategies rather than as isolated buildings [1,2,3,4]. This perspective also links major venues to visitor-oriented placemaking and related amenity clusters [5]. Recent work on place branding through sport events broadens this view by showing how venue-centered projects can be mobilized as place-branding devices within wider urban strategies [6]. Emerging built-environment evidence likewise suggests that stadium projects may reshape surrounding urban form and functional structure over time [7]. In this sense, venue-centered projects are best understood as parts of wider redevelopment processes rather than as self-contained facilities whose effects can be read only at the site level.

At the same time, claims that such projects generate broad-based local development remain heavily qualified. Early critiques argued that stadium-led development often produces redistribution rather than robust new growth [8,9]. Recent review studies retain that skepticism and show that generalized claims about jobs, income, and metropolitan uplift are frequently overstated [10,11]. More focused evidence from stadium-linked redevelopment settings also suggests caution, indicating that even where localized gains occur, net increases may be modest and partly offset by displacement or crowding out of other activity [12]. Critical work on stadium-led regeneration likewise shows that visible redevelopment can coexist with uneven local outcomes and exclusionary pressures [23]. Recent culture-led regeneration research further suggests that local effects are filtered through contested meanings [24], governance arrangements and redevelopment logics [25], and locally negotiated priorities rather than unfolding as a single uniform process [26]. The literature therefore supports viewing venue-centered redevelopment as a selective and context-dependent form of urban change, not as an automatic engine of broad-based local uplift.

2.2. Property-Market Capitalization and Locational Value

One of the clearest themes in the venue literature is that effects may be more visible in property markets than in aggregate economic indicators. In London, venue-related redevelopment and mega-event anticipation have been associated with nearby residential price premiums and locational revaluation [13,14]. Research from other settings reports similar housing-market responses around sports facilities, especially where proximity and amenity value shape local demand [15,16,17]. More recent evidence from large sports-park development likewise points to dynamic housing-price effects around event-related landscapes [27]. Property-market evidence therefore offers an important lens through which the spatial implications of venue-centered redevelopment can be observed.

However, capitalization should not be treated as synonymous with broadly shared neighborhood benefit. Property-based evidence does not necessarily support strong claims about community-wide gains [28]. More recent urban revitalization research also suggests that land-value effects vary by project type and local context [29], while sport-led redevelopment can produce housing and redevelopment outcomes that are unevenly distributed rather than uniformly shared [30]. Accordingly, the present study interprets housing-value change as evidence of locational revaluation and capitalization rather than as a direct measure of inclusive neighborhood improvement.

2.3. Local Business Dynamics, Visitor Economy, and Sectoral Unevenness

Compared with the property literature, evidence on local business effects is less consistent and more spatially specific. Some studies identify localized commercial gains in immediately adjacent business districts rather than broadly distributed economic expansion [18]. Other work emphasizes visitor spending, amenity bundling, and planned placemaking in shaping how arena-anchored areas attract consumption-oriented activity [5]. Agglomeration-based accounts similarly suggest that sports facilities may intensify clustered service activity in locations most directly exposed to venue-related flows [31]. Business effects, when they occur, therefore appear more likely to be localized and mechanism-specific than automatically generalized across the surrounding urban economy.

Business effects are also unlikely to be uniform across sectors. Uses more directly tied to footfall, leisure consumption, and destination visibility may respond differently from office-oriented or less customer-facing activities [5]. Recent work on neighborhood commercial change shows that regeneration pressures can produce selective restructuring, uneven survival, and differentiated vulnerability among local establishments [32,33]. Aggregate business counts may therefore conceal important internal variation, making sector-specific change analytically necessary rather than optional.

2.4. Methodological Approaches and the Unresolved Gap

The methodological literature reflects the same fragmentation visible in the outcome literature. One stream evaluates broad local economic claims through city-, county-, or metropolitan-level indicators and commonly finds limited generalized growth effects [10,11,12]. A second stream focuses on property-market responses and is better suited to identifying shifts in locational value and capitalization around major facilities [27,28]. A third stream examines localized commercial activity, visitor-oriented concentration, or adjacent-district business performance, thereby capturing more spatially focused patterns of venue-related change [5,18,31]. Each approach has produced important insights, but each captures only part of the broader urban effects associated with venue-centered redevelopment.

The unresolved gap lies here. Existing studies more commonly isolate housing-market responses [27,28] or localized commercial effects [5,18,31] than examine both within the same venue-centered redevelopment setting. Research that also distinguishes aggregate business concentration from sector-specific business change while considering temporal variation in housing outcomes remains especially limited. Recent redevelopment research beyond the venue literature likewise suggests that employment responses can be project-specific and spatially uneven [34], reinforcing the need for outcome-sensitive comparative frameworks. This study addresses that gap by combining a within-borough comparative design with a multi-indicator framework that distinguishes among locational-value premium, aggregate business concentration, and sector-differentiated local business change. In doing so, it treats divergence across indicators not as residual inconsistency, but as the central analytical problem through which the surrounding effects of a major entertainment anchor should be evaluated.

The following section translates this literature-based framework into the empirical design used for the O₂ case.

3. Materials and Methods

3.1. Study Area and Research Design

This study examines The O₂ and the wider Greenwich Peninsula as a case of venue-centered urban change in Greater London. Rather than treating The O₂ as an isolated entertainment facility, the analysis situates it within a broader mixed-use redevelopment setting that combines residential development, leisure uses, commercial activity, and longer-term place-making processes [20,21]. The aim is to assess whether areas most directly associated with this venue-centered redevelopment setting display different patterns from other areas in the same borough across multiple indicators of urban and market change.

To do so, the study adopts a within-borough comparative design in which treatment areas adjacent to The O₂ are compared with non-treatment Middle Layer Super Output Areas (MSOAs) within the Borough of Greenwich. This borough-internal strategy is intended to improve comparability by limiting the broader locational, institutional, and market heterogeneity that would arise in cross-borough comparisons within London. It therefore provides a structured way to examine whether venue-adjacent areas followed different trajectories from other areas in the same borough under a shared local administrative and planning context.

At the same time, this design does not remove all possible sources of confounding. Spatial dependence, neighborhood-level differentiation, and the wider redevelopment processes unfolding across Greenwich Peninsula may still shape the observed outcomes. The estimates should therefore be interpreted as evidence of differential outcomes associated with a venue-centered redevelopment setting rather than as proof of a single-facility effect.

The spatial logic of treatment is based on proximate exposure rather than on a simple geometric buffer. In this case, the relevant redevelopment influence is assumed to operate through venue-adjacent MSOAs that form part of the connected Greenwich Peninsula urban field. For this reason, treatment is defined through contiguous and proximate Greenwich MSOAs rather than through a mechanically uniform radial threshold. This approach is intended to capture the redevelopment zone most directly associated with The O₂ and its surrounding transformation while remaining consistent with the administrative-statistical geography used in the source datasets. The resulting study-area configuration is shown in Figure 1.

3.2. Data Sources, Spatial Frameworks, and Variable Construction

The analysis is organized around two complementary empirical streams: housing prices and business counts. These streams address related but distinct dimensions of urban change and differ in temporal coverage, spatial framework, and inferential scope. The housing stream is used to examine long-run capitalization patterns and relative price dynamics, whereas the business stream is used to examine later-period commercial concentration and sector-differentiated business trajectories.

The housing-price stream uses Office for National Statistics (ONS) small-area median house-price data at the MSOA level for 1995–2025 [35]. The principal dependent variable is median house price, modeled in logarithmic form in the regression analysis to reduce skewness and facilitate relative interpretation. The study also uses ONS residential sales counts for the same geography and period. These sales counts are included as activity-related sensitivity variables to assess whether the main housing results remain similar when local transaction activity is taken into account. Because transaction activity may itself respond to neighborhood change, redevelopment expectations, and market conditions, these variables are treated as supplementary sensitivity measures rather than as fully exogenous controls. Median house price is interpreted here as a proxy for locational value and capitalization within the local housing market, rather than as a direct measure of welfare gain or broadly shared neighborhood benefit.

Because the ONS price and sales files contain multiple year-ending observations within calendar years, the housing analysis uses two related data structures. First, pooled MSOA-period observations are retained for descriptive summaries that reflect the original reporting structure of the ONS release. Second, an annualized MSOA-year panel is constructed by averaging within-year observations for each MSOA. This annualized panel is used for dynamic estimation, pre-period comparability assessment, and annual sensitivity checks. This distinction is important because descriptive counts based on pooled observations should not be interpreted as annual counts in the same way as the annualized trend figures and panel regressions.

The business-count stream uses NOMIS UK Business Counts data at the 2011 MSOA level for 2016–2024 [36]. Two forms of business evidence are employed. The first is total business count, which is used to compare absolute business concentration, treatment–control gaps, and normalized later-period trajectories. The second is sector-level business count, which is used to assess whether aggregate business patterns conceal sector-differentiated dynamics. For this purpose, the sector analysis groups the NOMIS categories into four analytically interpretable clusters: Retail; Accommodation and Food Services; Arts, Entertainment, Recreation, and Related Other Services; and Office-Oriented Services. The final cluster combines information and communication, financial and insurance activities, property, professional and scientific activities, and business administration and support services.

The housing and business streams are not forced into a single spatial framework. Housing prices and residential sales are analyzed under the current MSOA framework used in the ONS release, whereas business counts are analyzed under the 2011 MSOA framework used in the NOMIS series. This distinction reflects the structure of the underlying data. Imposing a mechanically identical spatial frame across incompatible boundary systems would create more distortion than allowing each empirical stream to remain internally consistent within its own official geography. The two streams are therefore analytically aligned through a shared treatment logic, but not through an identical unit structure.

No physical instruments, reagents, or proprietary agents were used in this study. Data processing, statistical analysis, and visualization were conducted using Python 3.13 (Python Software Foundation, Beaverton, OR, USA), including the pandas, NumPy, statsmodels, GeoPandas, and Matplotlib libraries.

Table 1. Data sources, variables, spatial units, temporal coverage, and analytical roles.

Empirical Stream	Dataset	Main Variable	Spatial Unit	Time Span	Analytical Role
Housing prices	ONS	Median house price; residential sales count	Current MSOA	1995–2025	Descriptive comparison; pre-2007 comparability assessment; static DID; annual event-study; sales-based sensitivity checks
Business counts	NOMIS	Total business count; sector-level business counts	2011 MSOA	2016–2024	Level comparison; gap comparison; normalized index comparison; sectoral descriptive-comparative analysis
Boundary support	ONS Geography	MSOA polygons and codes	Current/2011 MSOA	Cross-sectional	Spatial delineation of treatment and control areas
Site reference	The O2 location	Venue point location	Point	Cross-sectional	Identification of the venue-centered treatment zone

summarizes the data sources, variables, spatial units, temporal coverage, and analytical roles used in the study.

3.3. Treatment and Control Definition Across Empirical Streams

Treatment and control groups are defined separately for the housing and business streams, but both follow the same basic logic: treatment captures the venue-adjacent Greenwich Peninsula exposure zone, whereas control captures the remaining Greenwich MSOAs outside that zone.

For the housing stream, treatment is defined under the current MSOA system as the MSOA containing The O₂ (Greenwich 040) and its three adjacent Greenwich MSOAs (Greenwich 037, Greenwich 038, and Greenwich 041). The housing treatment group therefore comprises four current MSOAs. The control group consists of the remaining Greenwich MSOAs not classified as treatment.

For the business stream, treatment is defined under the 2011 MSOA system as six proximate Greenwich units: Greenwich 004, Greenwich 012, Greenwich 014, Greenwich 036, Greenwich 037, and Greenwich 038. The control group again consists of the remaining Greenwich MSOAs outside this treatment zone. The business stream therefore uses six treatment units rather than four. This difference reflects the application of the same proximate-exposure logic under a different official boundary system.

The key point is not that the housing and business streams use the same number of units, but that they apply a consistent spatial principle to different statistical geographies. In both cases, treatment identifies the area most directly associated with the venue-adjacent redevelopment setting, whereas control identifies the rest of the borough. This common logic permits cross-stream comparison at the level of substantive interpretation even though the underlying unit structures are not identical.

Table 2. Treatment and control definitions across empirical streams.

Empirical Stream	Boundary System	Treatment Definition	Treatment Units	Number of Treatment Units	Control Definition
Housing prices	Current MSOA	The MSOA containing The O2 and its contiguous Greenwich neighbors	Greenwich 040; Greenwich 037; Greenwich 038; Greenwich 041	4	Remaining Greenwich MSOAs
Business counts	2011 MSOA	Proximate 2011-MSOA treatment zone around The O2 and Greenwich Peninsula	Greenwich 004; Greenwich 012; Greenwich 014; Greenwich 036; Greenwich 037; Greenwich 038	6	Remaining Greenwich MSOAs

Note: The housing and business streams follow the same proximate-exposure logic but not the same unit count, because they are implemented under different official MSOA boundary systems.

reports the treatment and control definitions used in each empirical stream.

3.4. Analytical Strategy and Interpretation Scope

The empirical strategy is deliberately layered. Rather than relying on a single model or a single outcome, the study combines descriptive comparison, baseline comparability assessment, static panel estimation, dynamic housing analysis, and sector-differentiated business analysis. This strategy is intended to distinguish among level differences, average post-opening differences, dynamic relative patterns, and sector-specific later-period variation.

For the housing stream, the analysis proceeds in four stages.

First, treatment and control areas are compared descriptively through summary statistics, long-run trend figures, and annual treatment–control gaps.

Second, pre-2007 comparability is assessed using both log median house prices and log-transformed residential sales counts. Specifically, the study compares treatment and control areas in terms of pre-period average levels and pre-period linear slopes. This step distinguishes pre-existing level differences from pre-period trend comparability.

Third, the study estimates a set of static difference-in-differences (DID) models to evaluate the average post-opening treatment differential. The baseline specification is written as follows:

$$\log \left(P_{it}\right)=\alpha +{\beta }_{1}Trea{t}_i+{\beta }_{2}Pos{t}_t+{\beta }_3\left(Trea{t}_i\times Pos{t}_t\right)+{\epsilon }_{it}$$

(1)

where $\mathrm{l}\mathrm{o}\mathrm{g}\left(P_{it}\right)$ denotes the logarithm of median house price for MSOA $i$ in year $t$, $Trea{t}_i$ identifies treatment MSOAs, $Pos{t}_t$ identifies the post-opening period, and $Trea{t}_i\times Pos{t}_t$ is the DID interaction term of interest. To strengthen comparability, the study also estimates a specification with year fixed effects and a two-way fixed-effects specification with both year and MSOA effects. Static DID is retained because it provides a transparent estimate of the average post-opening treatment differential across the full post period. In this study, it is used as a summary comparison of average post-opening difference rather than as a stand-alone basis for strong causal inference.

Fourth, the study estimates an annual event-study specification on the annualized MSOA-year panel. Relative-year indicators are defined around the 2007 opening, with year −1 as the omitted reference category. This dynamic specification is used to examine whether relative housing-price effects were concentrated in particular post-opening years rather than being uniformly distributed across the entire post period. Appendix A reports the individual pre-treatment coefficients and a joint pre-trend test. These diagnostics are used to assess how the dynamic estimates should be interpreted in light of pre-period behavior.

In addition to the main housing models, the study performs annual sensitivity checks that include contemporaneous and lagged log-transformed residential sales counts. These sensitivity models are included because transaction activity may indicate whether the static findings remain similar under variation in local market activity. Since transaction intensity may itself respond to neighborhood change, redevelopment expectations, and market conditions, these models are interpreted as sensitivity checks rather than as definitive solutions to omitted-variable concerns.

For the business stream, the analysis is descriptive–comparative rather than quasi-experimental in the same sense as the housing models. Because the consistent NOMIS series begins only in 2016, the business analysis cannot sustain a full pre/post design tied directly to the 2007 opening. The study therefore evaluates business change through three descriptive devices: absolute level comparison, treatment–control gap comparison, and normalized index comparison with 2016 set equal to 100. Here, 2016 functions as the first year of the consistent business series rather than as an intervention year.

To avoid relying exclusively on aggregate business counts, the study also examines sector-level business patterns. For each of the four sector groups, average counts per MSOA are computed separately for treatment and control areas and converted into 2016-based indices. This sector analysis is used to determine whether aggregate business trajectories conceal uneven sectoral dynamics, particularly between visitor-facing activities and more office-oriented services. The sector figures are interpreted as descriptive evidence of later-period business differentiation rather than as formal causal estimates.

Taken together, the analytical framework implies two different levels of interpretation. The housing stream provides stronger comparative panel evidence through descriptive comparisons, static DID, annual event-study estimates, and sales-based sensitivity checks. The business stream provides later-period descriptive-comparative evidence on total and sector-level business patterns. The estimates are used to evaluate how a venue-centered redevelopment setting is associated with differentiated outcomes across housing values, total business concentration, and sector-specific local business dynamics.

Table 3. Analytical approaches and interpretation scope across housing and business streams.

Outcome/Stream	Main Comparison	Main Methods	Core Output	Interpretation Scope
Housing prices	Treatment vs. Control within Greenwich	Descriptive statistics; long-run trend; annual gap; pre-2007 comparability assessment; static DID; annual event-study; sales-based sensitivity checks	Persistent price premium; average post-2007 treatment differential; dynamic relative effects; robustness to transaction activity	Stronger comparative panel evidence
Business counts	Treatment vs. Control within Greenwich	Descriptive statistics; total level trend; treatment–control gap; 2016 = 100 normalized index; sectoral descriptive-comparative analysis	Absolute business concentration; later-period relative trajectory; sector-specific business patterns	Descriptive-comparative evidence

summarizes the main analytical approaches and the interpretation scope adopted in the study.

The following section reports the empirical results in the same sequence, beginning with baseline comparability and housing dynamics before turning to business concentration and sectoral differentiation.

4. Results

4.1. Baseline Comparability and Descriptive Context

Before turning to post-opening comparisons, the study first examines pre-2007 comparability between treatment and control areas.

Table 4. Pre-2007 comparability between treatment and control areas.

(a). Pre-2007 level comparison
Variable	Control Mean	Treatment Mean	Difference (Treatment–Control)	p-Value
Log median house price	11.585	11.935	0.350	<0.001
Log(1 + residential sales count)	4.584	4.596	0.012	0.920
(b). Pre-2007 linear trend comparison
Variable	Control Slope/Year	Treatment Slope/Year	Difference (Treatment–Control)	p-Value
Log median house price	0.128	0.131	0.003	0.682
Log(1 + residential sales count)	0.080	0.090	0.009	0.749

Note: Pre-2007 refers to 1995–2006. (a) reports differences in pre-period average levels. (b) reports differences in linear annual trends over the pre-period.

reports differences in pre-period average levels and linear slopes for both log median house prices and log-transformed residential sales counts. The treatment areas exhibited a higher pre-period level of log median house price than the control areas, with a mean difference of 0.350 that is statistically significant. By contrast, the difference in pre-2007 price slope is small and statistically insignificant. Residential sales also show no statistically significant pre-period differences, either in average level or in annual slope. Taken together, these results indicate that the treatment areas entered the post-2007 period with a pre-existing housing-value premium, while pre-period trend differences were much less pronounced.

The descriptive housing statistics reinforce this pattern. As shown in

Table 5. Descriptive statistics for housing prices in treatment and control areas before and after 2007.

Group	Period	Mean	Median	Std. dev.	Min.	Max.	Count
Treatment	Pre-2007	176,272	169,225	76,480	65,000	355,000	180
	Post-2007	458,121	466,000	120,692	247,000	715,000	292
Control	Pre-2007	125,169	116,875	58,391	30,500	299,975	1372
	Post-2007	326,351	320,200	114,658	125,000	700,000	2263

Note: Counts refer to pooled MSOA-period observations in the underlying housing-price panel rather than to the annual observations displayed in the trend figures.

, the treatment areas recorded higher housing values than the control areas both before and after 2007. In the pre-2007 period, the mean median house price was 176,272 in the treatment areas, compared with 125,169 in the control areas. In the post-2007 period, the corresponding means increased to 458,121 and 326,351, respectively. Median values follow the same pattern. These descriptive statistics therefore confirm that the treatment areas formed part of a systematically higher-value segment of the Greenwich housing market across both periods.

4.2. House-Price Trajectories and Post-Opening Dynamics

Figure 2 plots the long-run housing-price trend in logarithmic form. The treatment series remains above the control series throughout the observation period, indicating that the areas surrounding The O₂ consistently occupied a higher-value position within the Greenwich housing market. This pattern is consistent with the descriptive comparisons reported in Table 5 and with the pre-period level difference shown in Table 4.

Figure 3 presents the annual treatment–control gap in mean median house prices. The gap is positive throughout the observed period and generally widens over time. The key point is not that the post-2007 years mark the first appearance of a difference between treatment and control, but that the positive gap persisted and became larger over the longer run. The figure therefore shows the durability of the locational premium associated with the treatment areas, rather than a simple before-and-after break.

Table 6. Estimated treatment-by-post interaction coefficients for log median house prices under three model specifications.

Model	Treatment$\times$ Post Coefficient	Std. Error	t-Value	p-Value
DID	0.0117	0.0418	0.2797	0.7797
DID + year fixed effects	0.0073	0.0173	0.4230	0.6723
DID + two-way fixed effects	0.0106	0.0125	0.8436	0.3989

reports the static DID estimates. Across the three principal specifications, the treatment-by-post interaction remains positive but statistically insignificant: 0.0117 in the baseline DID model, 0.0073 in the model with year fixed effects, and 0.0106 in the two-way fixed-effects model. In substantive terms, these results indicate that when the post-2007 period is summarized as an average treatment differential, the housing-price premium associated with the treatment areas does not reach conventional levels of statistical significance. The annualized two-way fixed-effects estimate is likewise statistically insignificant, and this overall conclusion remains similar when contemporaneous and lagged residential sales counts are included as sensitivity variables.

Figure 4 presents the annual event-study estimates. The dynamic pattern differs from the static DID summary. Most pre-opening coefficients are not statistically distinguishable from zero, although the coefficient at relative year −6 is positive and statistically significant, and the joint pre-trend test is marginal at conventional levels. In the post-opening period, positive relative effects are concentrated in selected medium-run years rather than being distributed uniformly across the full post period. In particular, the coefficients for relative years +3, +4, and +6 are positive and statistically significant, while the coefficient at +2 is positive but only marginally significant.

Taken together, the housing results indicate a persistent locational premium rather than a uniformly generated post-opening break. The treatment areas were already more valuable before the opening of The O₂, and the static DID estimates do not support a strong average post-opening treatment effect. At the same time, the dynamic estimates suggest that additional relative strengthening emerged in selected post-opening years. The housing evidence is therefore more consistent with the reinforcement of an already differentiated setting than with a simple and uniform post-opening transformation.

4.3. Business-Count Trajectories and Sectoral Differentiation

The business-count results show a clear distinction between absolute concentration and relative growth.

Table 7. Descriptive statistics for total business counts in treatment and control areas, 2016–2024.

Group	Mean	Median	SD	Min.	Max.	MSOA- Year Observations	Unique MSOAs
Treatment	635.28	595.00	342.22	225.00	1240.00	54	6
Control	280.84	240.00	126.95	100.00	655.00	243	27

Note: Values are based on the 2016–2024 business-count panel under the 2011 MSOA framework.

reports the descriptive statistics for total business counts over 2016–2024. The treatment areas display a substantially higher business base than the control areas, with a mean count of 635.28 compared with 280.84 and a median count of 595.00 compared with 240.00. This difference indicates that the treatment areas maintained a much denser concentration of business activity over the observed period.

Figure 5 illustrates this absolute level difference over time. The treatment series remains above the control series in every observed year, indicating that the business base surrounding The O₂ was consistently larger than that of the non-treatment Greenwich areas. In absolute terms, therefore, the business stream points to a strong and persistent level advantage.

Figure 6 presents the same total business series in normalized form, with 2016 set equal to 100. Once the level difference is removed, the relative trajectories appear much closer. The treatment areas display some stronger growth in the middle of the period, but the gap narrows substantially in the later years, and by 2024 the normalized series are close. This indicates that the treatment areas did not translate their much larger business base into a comparably strong and sustained aggregate growth advantage over the full 2016–2024 period.

The sector summary reported in

Table 8. Sector-specific business-count growth summary, 2016–2024.

(a) Treatment group growth summary
Sector Group	Mean Count (2016)	Mean Count (2024)	Absolute Change	2024 Index	2024 Index Gap (Treatment–Control)
Retail	59.17	83.33	24.17	140.85	15.63
Accommodation and food services	42.50	54.17	11.67	127.45	2.45
Arts, entertainment, recreation, and related other services	36.67	51.67	15.00	140.91	16.42
Office-oriented services	275.83	247.50	−28.33	89.73	0.13
(b) Control group growth summary
Sector Group		Mean Count (2016)	Mean Count (2024)	Absolute Change	2024 Index
Retail		22.04	27.59	5.56	125.21
Accommodation and food services		14.81	18.52	3.70	125.00
Arts, entertainment, recreation, and related other services		18.15	22.59	4.44	124.49
Office-oriented services		115.74	103.70	−12.04	89.60

Note: “Office-oriented services” is a constructed grouping combining information and communication, financial and insurance activities, property, professional and scientific activities, and business administration and support services. The arts-related category also includes other related services in the NOMIS classification.

clarifies why aggregate business growth appears more muted than the absolute level difference might suggest. Sectoral trajectories are not uniform. Retail shows a stronger treatment trajectory, with a 2024 index of 140.85 compared with 125.21 in the control areas. Arts, entertainment, recreation, and related other services show a similar pattern, with a treatment index of 140.91 compared with 124.49 in the control areas. Accommodation and food services display a more modest treatment advantage, with the two groups remaining relatively close. By contrast, office-oriented services show almost no difference in normalized growth by 2024, with both treatment and control remaining below their 2016 baseline.

These sector-level results show that the aggregate business trajectory conceals important internal variation. The treatment areas were associated with stronger relative growth in visitor-facing sectors, particularly retail and arts-related activities, while office-oriented services did not exhibit a comparable advantage. The business evidence therefore points not to a uniform pattern of local commercial expansion, but to a differentiated structure in which strong absolute concentration coexists with uneven sectoral change.

5. Discussion

5.1. Persistent Locational Premium and Temporally Uneven Housing Effects

The housing results indicate that the areas surrounding The O₂ were characterized by a persistent locational premium rather than by a simple and uniform post-opening break. The pre-2007 comparability results show that the treatment areas already exhibited a significantly higher level of housing value before the opening, while the difference in pre-period price slope was not statistically significant. The long-run trend and gap patterns likewise show that the treatment series remained above the control series throughout the observation period. Taken together, these findings suggest that the treatment areas entered the post-opening period from an already advantaged position within the Greenwich housing market.

This pattern is more consistent with capitalization and locational advantage than with a narrow before-and-after interpretation. In this case, the housing market appears to have sustained a premium associated with the venue-adjacent redevelopment setting, rather than responding through a singular discontinuity at one moment in time. The O₂ should therefore be understood as operating within a favorable locational context that was already differentiated before opening. In this sense, the housing evidence points more strongly to reinforcement than to creation.

At the same time, the housing results should not be reduced to persistence alone. The static DID estimates do not identify a statistically significant average post-opening treatment differential, which indicates that the post-2007 period does not yield a strong and uniform average effect when summarized through a single interaction term. However, the dynamic estimates suggest that relative strengthening emerged in selected medium-run post-opening years. These two results are not inconsistent. Rather, they suggest that any additional housing response was temporally uneven, becoming more visible in specific post-opening years than in a stable and uninterrupted form across the full post period.

This interpretation is also consistent with the idea that areas with an initial locational advantage may consolidate and extend an already favorable position under particular market conditions. The housing evidence is therefore most plausibly understood as showing that The O₂ became embedded in a broader process of locational reinforcement, within which relative gains became more visible at selected moments after opening.

5.2. Differentiated Urban Effects Across Housing and Business Outcomes

A central implication of the analysis lies in the divergence between the housing and business results. The housing stream indicates a persistent locational premium, whereas the business stream points to a more selective and uneven pattern of change. The treatment areas maintained a substantially larger business base than the control areas over 2016–2024, but the normalized business index shows only a limited and weakening relative growth advantage over time. This distinction between absolute concentration and relative growth is analytically important because a denser business base does not necessarily imply a correspondingly stronger subsequent growth trajectory.

This divergence supports the interpretation of differentiated urban effects. Prior research on venue-centered redevelopment has long distinguished between broad local economic claims and more localized or selective effects [10,11,12]. Work on nearby property markets shows that venue-related development may be capitalized into residential values through locational revaluation, event-related expectations, and amenity effects [17,28]. By contrast, business-side studies more often emphasize localized commercial activity, visitor-oriented concentration, and agglomeration in immediately adjacent areas [5,18,31]. These strands suggest different mechanisms rather than a single, uniform process of change.

The sector analysis clarifies this distinction further. Retail and arts-related activities display stronger treatment trajectories than the control areas, whereas accommodation and food services show only a modest treatment advantage and office-oriented services show almost no relative difference by the end of the period. This pattern is more consistent with selective intensification in visitor-facing activities than with a generalized upgrading of all forms of local business. It is also consistent with recent studies of neighborhood commercial change that emphasize selective restructuring, uneven survival, and differentiated vulnerability among local establishments [32,33].

This distinction matters conceptually because it separates asset-value effects from business-dynamics effects. Housing values may reflect locational revaluation and event-related expectations [13,14], as well as proximity and amenity effects around sports facilities [15,16]. Business outcomes, by contrast, depend more directly on localized commercial attraction, amenity bundling, agglomeration, and sectoral restructuring [31]. The O₂ case therefore suggests strong place-value effects alongside a more uneven pattern of local commercial change. One plausible interpretation is that a large pre-existing commercial base around the venue-centered area limited the extent to which later-period aggregate growth could continue to outpace the rest of the borough, making sector-specific intensification more visible than uniformly stronger overall expansion.

5.3. Methodological Implications and Limitations

The study also has several methodological implications. First, the within-borough comparative design is useful because it reduces the broader heterogeneity that would arise in a wider London comparison. By comparing venue-adjacent and non-treatment MSOAs within Greenwich, the analysis places treatment and control areas within a more comparable local administrative and planning context. This makes the design well suited to identifying differentiated patterns within a shared borough setting.

At the same time, the design does not eliminate all possible sources of confounding. Internal borough comparisons cannot fully remove spatial dependence, neighborhood heterogeneity, or the influence of wider redevelopment processes unfolding across Greenwich Peninsula. The estimates should therefore be interpreted as evidence on differential outcomes within a common local context, rather than as exhaustive identification of a single-facility effect.

Second, the housing and business streams operate on different inferential levels. The housing analysis provides stronger comparative panel evidence because it combines long-run descriptive comparisons, static DID, dynamic estimates, and activity-based sensitivity checks. The business analysis is more limited because the consistent business series begins only in 2016. For that reason, the business results are best interpreted as later-period descriptive-comparative evidence rather than as a full pre/post design parallel to the housing stream.

Third, the dynamic housing results require careful interpretation in light of pre-period behavior. The annual event-study estimates indicate that relative strengthening was concentrated in selected post-opening years, but the pre-opening pattern is not perfectly flat. The dynamic specification is therefore most useful for identifying temporal unevenness in relative treatment effects, not for establishing a strong claim of fully satisfied parallel trends.

Fourth, the inclusion of residential sales counts as sensitivity variables helps assess whether the housing results remain similar when local market activity is considered, but this does not eliminate broader omitted-variable concerns. Other time-varying influences, such as income change, accessibility shifts, and housing supply conditions, may still matter. In addition, the housing and business streams are implemented under different official MSOA boundary systems. This preserves internal consistency within each empirical stream, but it also limits the extent to which the two streams can be treated as mechanically identical spatial representations of the same urban field.

Taken together, these limitations do not diminish the value of the study’s comparative design, but they do define its proper interpretation. The study is strongest when used to identify differentiated patterns across housing and business outcomes within a shared borough context, and less suited to supporting a narrow claim that a single intervention produced a uniform and fully isolated causal effect.

5.4. Planning Implications

From a planning perspective, the results suggest that major entertainment venues should not be evaluated through a single headline indicator. In the present case, stronger housing-value performance, a much larger business base, and uneven sectoral growth do not point to one uniform pattern of neighborhood change. Instead, they indicate that venue-centered redevelopment can reinforce some dimensions of urban advantage while leaving others more limited, uneven, or sector-specific [5].

One implication is that absolute concentration and relative growth should be evaluated separately. A venue-adjacent area may sustain a large commercial base without necessarily outperforming surrounding areas in normalized growth over time. A second implication is that sectoral composition matters. Retail and arts-related activities may benefit more directly from a major entertainment anchor, whereas office-oriented services may remain comparatively weak. Planning evaluation should therefore distinguish between broad-based local expansion and selective growth concentrated in visitor-facing sectors.

A further implication is that housing-market gains should not be treated as synonymous with broadly shared local economic uplift. The results show that locational value can strengthen even when business growth remains uneven. This means that venue-led projects should be assessed through multiple dimensions of change, including housing, local business structure, and sectoral composition, rather than through a single measure of success.

More broadly, the findings suggest that a flagship venue is unlikely to function as an automatic engine of broad-based local development on its own. The urban significance of such a facility depends on how it interacts with the surrounding redevelopment context, local sectoral structure, and the wider planning strategy within which it is embedded. Land-use coordination, public-realm integration, long-term redevelopment sequencing, and support for local business adaptation are likely to matter if venue-centered investment is expected to generate more widely shared and durable benefits.

6. Conclusions

This study examined the differentiated urban effects associated with The O₂ and Greenwich Peninsula by comparing treatment and control MSOAs within the Borough of Greenwich across housing values and business activity. The study’s central contribution lies in showing that the effects of a major entertainment anchor embedded within a wider mixed-use redevelopment setting are not uniform across indicators, but instead emerge as a divergence between locational-value capitalization and local business dynamics.

The housing results point to a persistent locational premium rather than to a simple post-opening break. The treatment areas already exhibited higher housing values before the opening of The O₂, and this premium remained visible in the long-run trend and treatment–control gap patterns. At the same time, the static DID estimates did not identify a statistically significant average post-opening treatment differential. The dynamic estimates suggest that additional relative strengthening appeared in selected post-opening years rather than across the full post period. Taken together, the housing evidence is most consistent with the reinforcement of an already advantaged locational position within Greenwich.

The business results show a different pattern. The treatment areas maintained a substantially larger business base over 2016–2024, but the normalized business trajectory indicates only a limited aggregate relative growth advantage. Sectoral analysis further shows that this pattern was uneven. Retail and arts-related activities displayed stronger relative performance, accommodation and food services showed only a modest advantage, and office-oriented services showed little relative difference by the end of the period. The business evidence therefore suggests selective sectoral intensification rather than broad-based local commercial expansion.

Overall, the findings support an interpretation of differentiated urban effects. Housing-value capitalization, aggregate business concentration, and sector-specific commercial change did not move together, and they should not be treated as interchangeable indicators of neighborhood transformation. For planning and evaluation, major entertainment venues should therefore be assessed through multiple outcomes and with attention to sectoral structure and the wider redevelopment context in which they are embedded.

This study also has limitations. The housing and business streams operate under different boundary systems and at different inferential levels: the business series begins only in 2016, and the dynamic housing estimates do not warrant a strong claim that parallel trends are fully satisfied. Future research could extend this analysis through longer local economic time series, more detailed consumer-activity indicators, and additional comparative cases.