1. INTRODUCTION
Residential verticalisation has become a central dimension of contemporary urban transformation, reshaping city form and infrastructure worldwide. Between 1990 and 2014, building density increased in 72 of 200 cities analysed by Angel et al. (2021), reflecting a growing reliance on vertical growth. This shift can expand housing supply, mixed uses and accessibility, but it also risks producing monotonous, overcrowded and segregated environments with limited long-term adaptability (Clark & Moir 2015).
The urban development of Santiago de Chile provides a valuable case study. The city’s growth has been directly shaped by market dominance and lifestyle structures organised around purchasing power (Vergara-Perucich & Boano 2021). Santiago has experienced accelerated urban verticalisation, concentrating large-scale projects in strategic areas from a financial and profitability perspective, producing a markedly different urban landscape, especially in central and pericentral areas since the late 20th century.
Despite the growing literature on urban verticalisation (White & Serin 2021), there is still a gap in accurately identifying the levels or degrees of verticalisation and understanding the urban attributes that correlate with these levels in Latin American cities. In this article, levels of verticalisation are understood as the intensity and morphological configuration of high-rise residential development, combining building height, built density, number and size of dwellings, and the degree of plot utilisation. Within this spectrum, the terms ‘extreme’ or ‘larger scale verticalisation’ are used to describe projects that exceed conventional thresholds of density and scale, concentrating a large number of relatively small units within a single building or parcel. Although measured quantitatively, these dimensions also capture qualitative differences in the real-estate product (amenities, construction standards and, ultimately, intended target groups).
This work seeks to fill this gap through two main objectives: to develop an empirical methodology to identify different levels of verticalisation beyond simple floor counts; and to analyse the urban attributes associated with these levels in Santiago.
These objectives are articulated through the following research questions:
How can distinct levels of residential verticalisation be identified in a Latin American metropolis using building- and plot-level information?
Which urban, socio-economic and regulatory factors are associated with the emergence of extreme or larger scale verticalisation?
Extreme verticalisation can be expected to concentrate in pericentral areas with lower land values, weaker planning instruments and high accessibility to mass transit, rather than in high-income municipalities. The resulting typology captures morphological contrasts between projects and serves as a proxy for differentiating housing products and their likely target groups according to dwelling size, density, quality and location.
The paper is structured as follows. Section 2 presents a critical review of the literature on urban verticalisation, highlighting the gaps this study aims to address. Section 3 details the case study, data sources and methodological procedures. Section 4 analyses the results and their implications for the city of Santiago. Finally, Section 5 discusses the findings and outlines the main conclusions.
2. LITERATURE REVIEW
2.1 VERTICALISATION OVERVIEW
Verticalisation refers to the production and transformation of urban space through residential high-rise developments that reshape cityscapes and economic and social relations (Altrock 2022; Rojas Symmes et al. 2023). Although it is a global phenomenon, local circumstances and institutional complexities make verticalisation strongly context dependent (Glauser 2022), producing distinct spatial configurations, from skyscraper clusters to transformed public spaces and revitalised neighbourhoods (March & Lehrer 2019). Cities such as New York, Hong Kong, Singapore and Tokyo have embraced vertical urbanism by stacking functions, increasing density and concentrating investment in high-rise districts (Bruyns et al. 2020). In these contexts, high-rise neighbourhoods attract capital through high-end housing, while everyday services and public facilities are increasingly accommodated within mixed-use towers (Hirai 2022).
Similar processes are observed in Europe, where densification policies in London and other cities affect residential values and urban form, and residential skyscrapers in Germany blend modern high-rise typologies with traditional fabrics as tools for both luxury housing and urban revitalisation (Livingstone et al. 2021; Altrock 2022). Verticalisation has been associated with self-sufficient, mixed-use environments in Asia and the Middle East (Akristiniy & Boriskina 2018), and with ‘condoisation’ in Canada, where condominium forms reorganise governance and everyday life in mid- and high-rise housing (Grisdale & Walks 2022). Although operating under different legal frameworks, similar dynamics can be observed in Santiago’s condominium-style high-rise projects, where collective ownership and governance arrangements structure everyday life in vertical enclaves. In London, Cerrada Morato (2022) shows how planning systems and political contexts can compromise building quality and public space under intense land-market pressure.
The effects of high-rise housing remain contested in planning research, and there is growing recognition that local political contexts and regulatory systems critically shape how such buildings are produced (Glauser 2022). In this debate, studies foreground questions of social inclusion and exclusion, framing urban renewal and redevelopment as key expressions of densification (Lai et al. 2018). Higher densities can enhance infrastructure efficiency and economic vitality, but also generate environmental, social and health challenges (Berghauser Pont et al. 2021). This tension underscores the difficulty of balancing the benefits of densification with its unintended consequences, and has led scholars to emphasise the need for planning and design approaches that address not only building footprints but also the full three-dimensional space and its wider urban effects (Borsdorf & Hidalgo 2010; Nethercote 2018; Webb & White 2022).
Across the Global South, contemporary verticalisation takes diverse forms that combine global capital, symbolic projects and uneven local benefits. In Sub-Saharan Africa and Middle Eastern metropolises, isolated ‘skyscraper hubs’ and iconic mega-towers are deployed as tools for city branding and speculative investment (Bini & d’Alessandro 2017). In Jakarta, Mumbai and Phnom Penh, financialised housing regimes and uneven planning generate ‘differential verticalisation’ that mixes elite projects with dense resettlement high-rises for low-income groups (Rahmawati & Rukmana 2022; Burte 2024; Nam 2017; Fauveaud 2020), while in Chongqing topography and vertical circulation embed verticality in everyday governance and urban life (Jin 2022).
Overall, this body of work underscores the complex interplay of regulatory frameworks, market dynamics, social relations, and architectural possibilities that shape verticalisation and its spatial selectivity. Taken together, these international experiences show that verticalisation is not a uniform or purely luxury trend, but a context-dependent process whose drivers and social effects vary across governance and market regimes, an insight that informs the analysis of Santiago’s extreme verticalisation patterns.
2.2 VERTICALISATION IN LATIN AMERICA
Urban verticalisation has also become a central phenomenon in Latin America, driven by intertwined processes of urban accumulation, reinvestment and financialisation. High-rise housing is used to maximise land use within regulatory parameters and to channel capital into real estate, in line with broader debates on the financialisation of urban space (Shin 2011; De Mattos 2016; Aalbers 2017; Rolnik 2017). A large body of research has examined these dynamics and their spatial impacts in cities such as Buenos Aires (Tella et al. 2011; Romano et al. 2024), Santa Fe (Fedele & Martínez 2015), Santiago (Vicuña 2020; Rojas Symmes 2017; Orellana et al. 2022), Concepción (Pérez et al. 2019; Vicuña et al. 2024) and several Brazilian cities (De Oliveira et al. 2015; Melo & Matana Júnior 2020; Almeida et al. 2025).
Urban renewal policies have frequently underpinned these processes, seeking to repopulate and revitalise deteriorated centres, address building obsolescence and transform pericentral industrial areas (Paquette Vassalli 2020). Local governments often act as market facilitators, adapting planning regulations to attract vertical investment and its associated tax and land-value gains, as in Santiago de Chile (Vergara-Perucich 2018). Major cities such as Buenos Aires, São Paulo and Lima have experienced waves of luxury towers and compact units (including micro-apartments) fuelled by global capital flows and public–private renewal schemes that reorder land-use patterns (Klaufus et al. 2017; Yunda & Sletto 2020; Arroyo 2024).
Verticalisation has profoundly reshaped housing supply in the region. Large towers have introduced very small apartments (sometimes as small as 17 m2 in Chile), generating overcrowding and the stigmatisation of vulnerable groups, especially migrants (Orellana et al. 2022). Public and academic debate has labelled some of these complexes ‘vertical ghettos’ due to poor living conditions and social isolation (Rojas Symmes et al. 2023). In Concepción, recent research documents a dispersed, market-led pattern of high-rise housing that increases built residential density without equivalent population growth, reflecting weak planning guidance and deep morphological change in central and pericentral neighbourhood (Vicuña et al. 2024). Similar dynamics are observed in Bogotá, where public and private investments reshape socio-spatial orders and reinforce selective forms of renewal (Yunda & Sletto 2020).
In Brazil, verticalisation illustrates this dual nature: it can support more compact urban forms and more efficient land use, but also produces car dependency, congestion and environmental degradation, particularly in vulnerable areas (Melo & Matana Júnior 2020). In cities such as Porto Alegre and Córdoba (Argentina), compact studio flats and redevelopment schemes conceived as financial assets strengthen central land valorisation, displacement and socio-spatial segregation (Almeida et al. 2025; Michelazzo & Salguero 2017).
Overall, Latin American verticalisation displays a clear ambivalence: it can modernise and compact urban form, yet it often reproduces or intensifies urban inequalities. When carefully regulated, it can support social inclusion and more efficient infrastructure, but its impacts vary sharply across social groups and city areas (Borsdorf & Hidalgo 2010; Aquino & Gainza 2014; Silveira & Silveira 2014). Nonetheless, few studies in this regional literature have systematically typified extreme forms of verticalisation using multidimensional quantitative indicators at the building scale, particularly in contexts such as Santiago, where vertical growth concentrates in low land-value, weakly regulated pericentral areas. This study addresses that gap.
3. MATERIALS, METHODS AND CASE STUDY
The study area corresponds to the consolidated urban area of the Santiago Metropolitan Area (Gran Santiago), understood as the continuous built-up fabric of the city. According to the National Institute of Statistics of Chile (INE), this urban continuum covered 83,789 ha and housed approximately 6,797,632 inhabitants in 2019, representing 35.6% of the national population and spanning 34 municipalities. Of these, only 29 enter the empirical analysis, because the remaining five did not register any eligible multifamily residential building permits during the period 2010–19 and therefore could not contribute building-level observations or municipality dummy variables to the regression models.
The methods comprise two quantitative analysis stages (Figure 1). The first identifies different levels of residential verticalisation; the second analyses the urban factors associated with them. The period of analysis covers 2010–19, a decade characterised by intense construction activity and a marked expansion of high-rise residential projects in Santiago (Moreno-Alba 2025). During these years, Chile’s construction and real-estate markets were driven by strong private-sector dynamism, increasing financialisation of housing supply and a relatively permissive regulatory environment, with heterogeneous municipal enforcement capacities and densification incentives often concentrated in pericentral areas with lower land values. These conditions created a development context particularly conducive to large-scale verticalisation, shaping both the volume and the spatial patterns of the building permits analysed in this study.
Figure 1
Methodological process.
Source: Author.
The empirical dataset consists of 892 residential building permits drawn from the 2144 permits for buildings with five or more floors issued between 2010 and 2019 within the consolidated urban area. These 892 cases correspond to those permits for which a precise spatial match could be established between the permit database and the georeferenced cadastral shapefile; permits that could not be reliably geocoded or linked to a specific parcel were excluded, as the spatial match was required to compute block- and buffer-based contextual indicators. Although the resulting dataset is not a probabilistic sample, it is highly representative: the municipal distribution of the 892 permits closely mirrors that of the full 2144-permit universe, preserving the metropolitan structure of high-rise residential construction.
3.1 IDENTIFYING VERTICALISATION
Different levels of residential verticalisation at the building scale were identified using information from formal residential building permits. For each of the 892 permits, a set of indicators was constructed to describe the intensity and morphology of the project: net residential density (dwellings/ha), built density (built m2 per plot m2), average apartment size, building height (floors), total number of dwellings, construction quality (ordinal scale), number of building installations, rooms per unit, built m2 per room, and dummies for three-or-more-room units and mixed use within the building. Variable definitions and descriptive statistics are reported in Table A1 in Appendix A in the supplemental data online.
To reduce dimensionality and capture the main latent dimensions of verticalisation, principal component analysis (PCA) was applied to the standardised building-level variables (Peres-Neto et al. 2005). The retained components summarise joint variation in density, height, unit size and plot utilisation; eigenvalues, explained variance and component loadings are presented in Tables B1 and B2 in Appendix B in the supplemental data online. A cluster analysis was performed on the component scores (Everitt et al. 2011) to group permits into internally homogeneous and mutually distinct types of high-rise development. The number of clusters is determined by standard stopping rules (Calinski & Harabasz 1974; Duda et al. 2001) combined with substantive interpretability.
The resulting clusters represent empirical typologies of verticalisation, ranging from low-intensity projects with larger units and lower densities to extreme or larger scale verticalisation, characterised by higher buildings, very high residential and built densities, smaller dwellings and lower construction quality. In subsequent analyses, cluster membership was used in two ways: as a binary outcome distinguishing the extreme verticalisation cluster from all others; and as an ordinal variable that orders clusters from lower to higher verticalisation intensity.
This classification is data driven and depends on the availability and quality of formal building permit records. It does not capture informal or unpermitted construction, and the specific indicators and thresholds used here reflect the Chilean regulatory and data context. However, the procedure (combining permit-level morphological indicators, PCA and clustering) is replicable in other cities where similar administrative information is available, and it can be adapted by adjusting the input variables to local market and regulatory conditions.
3.2 TERRITORIAL CONTEXT ANALYSIS
To characterise the urban environment into which each building is inserted, contextual indicators were created to describe pre-existing density, land values, accessibility, land-use composition and socio-economic conditions around each permit. These indicators are derived from census and cadastral information observed before the issuing of the building permit, so they represent the receiving context rather than the changes produced by the project. Detailed definitions and descriptive statistics are reported in Table C1 in Appendix C in the supplemental data online.
For every permit, a 300-m buffer around the parcel was delineated, a radius commonly used in urban research as an approximation of a three- to five-min walking distance and a compromise between very local conditions and statistically stable neighbourhood measures. All census blocks whose centroid falls within the buffer are included, and block-level indicators are averaged to obtain one value per permit. Residential density is measured as net housing density (dwellings/ha), where the denominator is the sum of parcel areas within each block, excluding public rights-of-way such as streets and pavements. Built density is defined as built floor area (m2) over parcel area, equivalent to a floor space index. Both measures therefore refer to net, rather than gross, densities.
The contextual indicators also include measures of accessibility and infrastructure (such as proximity to public transport and major roads), land-use mix, land prices and a socio-material index capturing socio-economic conditions. In addition, municipality-level dummy variables are incorporated as fixed effects to capture differences in local planning instruments and political–administrative contexts for the 29 municipalities that registered eligible multifamily residential permits.
These contextual indicators and the typology derived in Section 3.1 were then used in two types of regression models. A binary logistic regression (Uberti 2022) models the probability that a permit belongs to the extreme verticalisation cluster, while an ordered logistic regression (McCullagh 1980) uses the ordinal cluster variable to capture gradations in verticalisation intensity. In both cases, the dependent variable is explained by the set of contextual indicators and the set of municipality dummies. Coefficients for the final specifications are presented in the Results section; intermediate models and robustness checks are reported in Appendix E in the supplemental data online.
Before estimating the models, the correlation structure of the contextual indicators was explored by using a Pearson pairwise correlation matrix and computed variance inflation factors (VIFs) to diagnose multicollinearity. Most variables are continuous and approximately symmetric, and the logistic and ordered logistic models are specified as linear combinations of covariates; linear correlations are therefore informative for identifying highly collinear predictors. Variables with very high VIFs were iteratively excluded, retaining those that remained substantively relevant. The full correlation matrix and VIF statistics are presented in Tables C2 and C3 in Appendix C in the supplemental data online. Some relationships between urban attributes may be non-linear or only monotonic; rank-based measures such as Spearman’s rho could be incorporated in future applications of this methodology as part of extended robustness checks.
4. RESULTS
4.1 VERTICALISATION IDENTIFIED
The 11 project-level variables described in Section 3.1 were used to estimate a PCA. The first three components have eigenvalues greater than 1 and jointly explain 68% of the total variance (see Table B1 in Appendix B in the supplemental data online). Based on the highest loadings (see Table B2 online), component 1 is interpreted as Building intensity, combining high net residential density, built density, building height and number of units with smaller unit and room sizes. Component 2 is labelled Apartment compactness and configuration because it contrasts projects with more generous internal space per room and more traditional layouts against more compact subdivisions. Component 3 is characterised as High-end apartments and amenities, with high loadings on construction quality, installations and number of rooms per unit, capturing a dimension of product standard and internal space.
The scores of these three components were then used as inputs for a k-means cluster analysis. The Calinski–Harabasz and Duda–Hart indices indicated that a six-cluster solution offers a good balance between parsimony and internal separation. Each cluster was characterised by the means of the 11 indicators (Table 1), which allowed for the interpretation of distinct types of vertical housing.
Table 1
Means of different variables according to grouping for development permits.
| VARIABLE | CLUSTER 1 | CLUSTER 2 | CLUSTER 3 | CLUSTER 4 | CLUSTER 5 | CLUSTER 6 |
|---|---|---|---|---|---|---|
| Net residential density (apartments/ha) | 182.05 | 252.85 | 359.84 | 774.40 | 907.56 | 2023.83 |
| Built density (built m2/plot area m2) | 3.50 | 3.69 | 2.62 | 5.61 | 5.20 | 11.68 |
| Gross apartment area (apartments/built m2) | 242.52 | 163.41 | 76.72 | 76.71 | 66.31 | 61.35 |
| Height (floors) | 6.83 | 8.30 | 7.60 | 15.25 | 11.2 | 24.23 |
| Units (apartments) | 57.61 | 62.24 | 116.75 | 209.79 | 175.20 | 492.10 |
| Construction quality (1 = lowest to 5 = highest) | 3.72 | 3.87 | 2.87 | 3.32 | 3.40 | 3.13 |
| Installations (0 to 3) | 2.28 | 2.67 | 1.09 | 2.56 | 2.48 | 2.55 |
| Rooms (rooms/units) | 1.89 | 4.52 | 3.87 | 3.64 | 2.12 | 3.25 |
| Gross rooms area (built m2/rooms) | 157.46 | 36.74 | 20.18 | 21.52 | 30.16 | 19.32 |
| Rooms dummy (1 = permits with ≥ 3 rooms) per unit | 0.17 | 1.0 | 1.0 | 1.0 | 0.08 | 0.96 |
| Mixed-use dummy (1 = permits with mixed-use) | 1.0 | 0.02 | 0 | 0 | 0.12 | 0 |
| Observations | 18 | 282 | 173 | 357 | 25 | 97 |
[i] Source: Author based on National Institute of Statistics of Chile (INE) (n.d.) data.
Representative examples of projects belonging to each cluster are shown in Figure 2 (Google 2025), illustrating differences in height, footprint, dwelling size and urban setting:
Cluster 1: Mixed-use, high-standard mid-rises, with moderate densities, relatively large apartments, higher construction quality and a higher incidence of non-residential floor area.
Cluster 2: Family-oriented mid-rises, with similar densities and larger units but fewer mixed uses.
Cluster 3: Compact, lower quality mid-rises, with smaller dwellings and lower construction quality scores.
Cluster 4: Standard high-rise towers, with tall buildings and high densities but medium-sized units and intermediate quality.
Cluster 5: High-rise family towers, combining high densities with somewhat larger apartments and better internal space configurations than the most extreme cases.
Cluster 6: This represents the most extreme form of verticalisation in Santiago and is labelled extreme micro-apartment towers. It concentrates the highest net residential density, built density and height, together with the smallest apartments and room sizes, the largest number of units per project and comparatively lower construction quality and fewer rooms per unit. This cluster includes 97 development permits and is used in the subsequent analysis as the benchmark for extreme or larger scale verticalisation (Figure 2). The next section moves from this typology to its spatial distribution across the city.
Figure 2
Examples of buildings for the identified clusters.
Source: Google (2025).
4.2 SPATIAL PATTERNS OF VERTICALISATION AND THEIR LIKELIHOOD
The spatial distribution of both components and clusters reinforces this interpretation. A map of component 1 scores (Building intensity) shows the highest values concentrated in the pericentral belt of the city, with a marked peak in Estación Central and adjacent inner-ring areas (Figure 3), while components 2 and 3 display more diffuse patterns (see Figures D1 and D2 in Appendix D in the supplemental data online).
Figure 3
Spatial distribution of the first principal component.
Source: Author based on National Institute of Statistics of Chile (INE) (n.d.) data.
The cluster map (Figure 4) reveals that cluster 6 is predominantly located in Estación Central, especially near the Alameda corridor and the boundary with Santiago’s historical centre, with additional cases in Independencia, Ñuñoa and San Miguel. Clusters 4 and 5 exhibit similar, though slightly less extreme, patterns, extending towards pericentral and southern municipalities such as La Cisterna and La Florida. By contrast, the lower verticalisation clusters (1 and 2) are concentrated in eastern, high-income municipalities such as Providencia, Las Condes, Vitacura and Lo Barnechea. These spatial patterns confirm a strong central and pericentral concentration of the most intensive and lower standard vertical projects, alongside higher standard, lower intensity verticalisation in wealthier communes.
Figure 4
Location of development permits with the grouping of six clusters.
Source: Author based on National Institute of Statistics of Chile (INE) (n.d.) data.
Vertical growth is thus not randomly distributed, but responds to contextual conditions linked to accessibility, land values, land-use structure and socio-economic differentiation. The following subsections use logistic and ordered logistic models to relate the typology derived in Section 4.1 to the contextual indicators described in Section 3.2 and to municipality fixed effects capturing local planning and governance.
4.2.1 Logistic models
The logistic regression models estimate the probability that a development permit belongs to the most extreme verticalisation cluster (cluster 6) versus all other clusters (Table 2). A sequence of nested specifications was estimated and then highly collinear variables were progressively removed; intermediate models and VIF diagnostics are reported in Table E1 in Appendix E in the supplemental data online.
Table 2
Logistic models.
| VARIABLE | EXPLAINED VARIABLE: 1 = CLUSTER 6; 0 = OTHER CLUSTERS MODEL 5 | |||
|---|---|---|---|---|
| ESTIMATED | SE | PR(> |Z|) | VIF | |
| (Intercept) | 4.379 | 1.955 | 0.025 | |
| Distance to major public transport | –0.011 | 0.004 | 0.005 | 1.4 |
| Residential density | –0.016 | 0.009 | 0.065 | 1.8 |
| Built density | 0.162 | 0.092 | 0.078 | 2.0 |
| Land price | –0.073 | 0.058 | 0.209 | 1.4 |
| Socio-economic level | –0.006 | 0.003 | 0.045 | 2.1 |
| Residential Use Coefficient | –1.144 | 0.542 | 0.035 | 2.2 |
| Dummy to public transport | 0.307 | 0.314 | 0.328 | 1.2 |
| Dummy to motorways | 0.183 | 0.383 | 0.634 | 1.4 |
| Dummy of Conchalí | –16.670 | 4,586 | 0.997 | 1.0 |
| Dummy of El Bosque | –17.190 | 5,323 | 0.997 | 1.0 |
| Dummy of Estación Central | 2.809 | 0.590 | 0.000 | 3.0 |
| Dummy of Huechuraba | –16.310 | 4,128 | 0.997 | 1.0 |
| Dummy of Independencia | 2.127 | 0.609 | 0.000 | 2.0 |
| Dummy of La Cisterna | –1.877 | 1.138 | 0.099 | 1.2 |
| Dummy of La Florida | –0.382 | 0.902 | 0.672 | 1.5 |
| Dummy of La Granja | –16.010 | 7,604 | 0.998 | 1.0 |
| Dummy of La Pintana | –17.670 | 10,750 | 0.999 | 1.0 |
| Dummy of La Reina | –16.350 | 4,283 | 0.997 | 1.0 |
| Dummy of Las Condes | –15.060 | 1,240 | 0.990 | 1.0 |
| Dummy of Lo Barnechea | –16.000 | 3,713 | 0.997 | 1.0 |
| Dummy of Lo Prado | –17.640 | 10,750 | 0.999 | 1.0 |
| Dummy of Macul | –1.089 | 1.112 | 0.328 | 1.2 |
| Dummy of Maipú | –15.560 | 7,532 | 0.998 | 1.0 |
| Dummy of Ñuñoa | –0.225 | 0.576 | 0.696 | 1.8 |
| Dummy of Peñalolén | –15.700 | 3,277 | 0.996 | 1.0 |
| Dummy of Providencia | –15.320 | 1,168 | 0.990 | 1.0 |
| Dummy of Pudahuel | –15.980 | 4,134 | 0.997 | 1.0 |
| Dummy of Quilicura | –16.040 | 10,750 | 0.999 | 1.0 |
| Dummy of Quinta Normal | –0.331 | 0.702 | 0.637 | 1.7 |
| Dummy of Recoleta | –16.690 | 4,678 | 0.997 | 1.0 |
| Dummy of Renca | –16.100 | 3,343 | 0.996 | 1.0 |
| Dummy of San Joaquín | –16.710 | 3,481 | 0.996 | 1.0 |
| Dummy of San Miguel | –1.241 | 0.733 | 0.091 | 1.4 |
| Dummy of Vitacura | –14.990 | 1,552 | 0.992 | 1.0 |
| Dummy of Puente Alto | –14.770 | 10,750 | 0.999 | 1.0 |
| Observations | 892 | |||
| Prob > chi2 | 0.000 | |||
| Pseudo R2 (McFadden) | 0.454 | |||
[i] Note: SE = standard error; VIF = variance inflation factor.
Source: Author based on National Institute of Statistics of Chile (INE) (n.d.) data.
The final, more parsimonious specification (model 5) preserves the core substantive relationships with acceptable multicollinearity levels. In model 5, accessibility to major public transport emerges as a key contextual driver of extreme verticalisation. Distance to metro or bus stations has a negative and statistically significant effect (–0.011, p < 0.01), indicating that projects located further from these nodes have lower odds of belonging to cluster 6. Residential density in the surrounding area is also negatively associated with extreme verticalisation (–0.016, p < 0.10), while built density has a positive, borderline significant coefficient (0.162, p < 0.10), suggesting that extreme towers tend to be inserted in consolidated but not previously densest environments.
Socio-economic and land-use attributes reinforce this pattern. Lower socio-economic levels around the project increase the probability of extreme verticalisation (–0.006, p < 0.05), and a lower residential use coefficient (–1.144, p < 0.05) indicates that these projects are more likely in contexts where residential use does not fully dominate and other uses are present nearby. Land price retains a negative but non-significant coefficient in the final model, consistent with the descriptive evidence that cluster 6 concentrates in relatively lower value areas.
Municipality dummies capture differences in local regulatory and political–institutional contexts. Estación Central shows a strong, highly significant positive effect, confirming that, all else being equal, permits located there have a much higher probability of belonging to cluster 6 than those in Santiago municipality (the reference category). Independencia also exhibits a positive and significant coefficient, whereas several high-income eastern municipalities present negative, sometimes significant coefficients, indicating a lower propensity for extreme verticalisation in those territories. Overall, the model shows that extreme verticalisation in Santiago is most likely where transit accessibility is high, surrounding socio-economic and land-value conditions are weaker, land uses are more mixed, and municipal regulatory frameworks are more permissive toward intensive high-rise development.
4.2.2 Ordered logistic models
Whereas the logistic model focuses on the probability of belonging to the most extreme verticalisation type (cluster 6), the ordered logistic model exploits the full ordinal structure of the six clusters, from lower to higher intensity. The dependent variable is the cluster rank (1–6), and the same set of contextual indicators and municipality dummies were used as in the previous subsection. For brevity, Table 3 reports only the final specification (model 4), while earlier variants are presented in Table E2 in Appendix E in the supplemental data online. The model also estimates a set of threshold (cut-point) parameters (thresholds 1–5 in Table 3) that partition the latent verticalisation scale into ordered categories; these delimit the boundaries between clusters but are not interpreted substantively here, as the focus lies on the contextual covariates.
Table 3
Ordered logistic models.
| VARIABLE | MODEL 4 | ||
|---|---|---|---|
| ESTIMATED | SE | PR(> |Z|) | |
| Distance to major public transport | –0.0005 | 0.0008 | 0.5720 |
| Residential density | –0.0018 | 0.0040 | 0.6490 |
| Built density | 0.1948 | 0.0897 | 0.0080 |
| Land price | –0.0227 | 0.0086 | 0.0100 |
| Socio-economic level | –0.0047 | 0.0014 | 0.0010 |
| Residential Use Coefficient | –1.3964 | 0.0609 | 0.0000 |
| Dummy to public transport | 0.5079 | 0.2792 | 0.0030 |
| Dummy to motorways | –0.4553 | 0.1257 | 0.0220 |
| Dummy of Conchalí | –1.4346 | 0.1912 | 0.0740 |
| Dummy of El Bosque | –1.5341 | 0.1845 | 0.0730 |
| Dummy of Estación Central | 2.4648 | 4.8061 | 0.0000 |
| Dummy of Huechuraba | –1.5383 | 0.3012 | 0.2730 |
| Dummy of Independencia | 1.5015 | 2.1145 | 0.0010 |
| Dummy of La Cisterna | –0.2701 | 0.3046 | 0.4990 |
| Dummy of La Florida | –0.2929 | 0.3016 | 0.4690 |
| Dummy of La Granja | 0.3201 | 1.9454 | 0.8210 |
| Dummy of La Pintana | –1.6635 | 0.2982 | 0.2900 |
| Dummy of La Reina | –2.3484 | 0.1073 | 0.0370 |
| Dummy of Las Condes | –2.8128 | 0.0285 | 0.0000 |
| Dummy of Lo Barnechea | –2.4609 | 0.0824 | 0.0110 |
| Dummy of Lo Prado | –2.6583 | 0.1090 | 0.0880 |
| Dummy of Macul | 0.0492 | 0.4591 | 0.9100 |
| Dummy of Maipú | –0.7057 | 0.7077 | 0.6220 |
| Dummy of Ñuñoa | –1.2648 | 0.0952 | 0.0000 |
| Dummy of Peñalolén | –3.5770 | 0.0402 | 0.0130 |
| Dummy of Providencia | –3.7542 | 0.0111 | 0.0000 |
| Dummy of Pudahuel | –1.5841 | 0.1683 | 0.0540 |
| Dummy of Quilicura | –1.1190 | 0.5066 | 0.4710 |
| Dummy of Quinta Normal | –0.0711 | 0.4621 | 0.8860 |
| Dummy of Recoleta | –2.5438 | 0.0660 | 0.0020 |
| Dummy of Renca | –0.9277 | 0.2846 | 0.1970 |
| Dummy of San Joaquín | –0.9427 | 0.2555 | 0.1510 |
| Dummy of San Miguel | 0.1127 | 0.3737 | 0.7360 |
| Dummy of Vitacura | –3.8207 | 0.0136 | 0.0000 |
| Dummy of Puente Alto | –0.8260 | 0.6798 | 0.5950 |
| Threshold 1 | –10.9381 | 0.9818 | |
| Threshold 2 | –5.8216 | 0.9057 | |
| Threshold 3 | –4.4072 | 0.8953 | |
| Threshold 4 | –1.2126 | 0.8778 | |
| Threshold 5 | –0.9052 | 0.8782 | |
| Observations | 892 | ||
| LR chi2(43) | 749.09 | ||
| Prob > chi2 | 0.0000 | ||
| Pseudo R2 | 0.2973 | ||
| Log-likelihood | –885.23 | ||
[i] Note: SE = standard error.
Source: Author based on National Institute of Statistics of Chile (INE) (n.d.) data.
The ordered model largely confirms and generalises the patterns identified in the binary specification. Built density around the project shows a positive and statistically significant association with the verticalisation rank (0.1948, p < 0.01), indicating that more intensive typologies tend to emerge in already consolidated fabrics. By contrast, land price has a negative and significant coefficient (–0.0227, p = 0.01), and socio-economic level also presents a negative, highly significant effect (–0.0047, p < 0.01), suggesting that higher verticalisation clusters are more likely to be located in relatively lower value, lower income areas.
Land-use composition exerts an additional influence. The residential use coefficient is negative and strongly significant (–1.3964, p < 0.01), indicating that strictly residential surroundings reduce the odds of belonging to higher verticalisation clusters; extreme and higher intensity verticalisation tends instead to appear in mixed or transitional contexts. Regarding accessibility, the dummy for proximity to major public transport shows a positive and statistically significant coefficient (0.5079, p < 0.01), confirming that projects within the defined catchment of Metro or bus infrastructure are more likely to occupy the upper part of the verticalisation scale. Conversely, proximity to motorways has a negative and significant effect (–0.4553, p < 0.05), suggesting that adjacency to high-capacity road corridors is less attractive for the most intensive residential towers.
Municipal fixed effects in the ordered model reproduce the territorial selectivity observed in the logistic specification. Estación Central exhibits a large and highly significant positive coefficient (2.4648, p < 0.01), even after controlling for accessibility, densities, land-use structure and socio-economic conditions, confirming its central role in hosting the highest verticalisation clusters. Independencia also presents a positive and significant effect (0.5015, p < 0.01), while several eastern municipalities show negative and statistically significant coefficients, including Las Condes (–2.8128, p < 0.01), Providencia (–3.7542, p < 0.01), Vitacura (–3.8207, p < 0.01) and Lo Barnechea (–2.4609, p < 0.05). These results indicate a systematically lower propensity for high-intensity verticalisation in high-income municipalities compared with the reference category (Santiago municipality). Taken together, the logistic and ordered logistic models show that extreme and higher intensity verticalisation in Santiago de Chile is jointly shaped by transit accessibility, pre-existing built form, socio-economic vulnerability, land-use structure and permissive or flexible local planning frameworks.
5. DISCUSSION AND CONCLUSIONS
The findings of this study on Santiago’s urban verticalisation align with Global South urban development trends, where limited land availability, transport infrastructure and land-use regulation drive vertical growth (Bruyns et al. 2020; White & Serin 2021). In Santiago, the strong relationship between proximity to major public transport stations and the likelihood of high-rise development highlights the centrality of accessibility in shaping densification: transport hubs operate as anchors for high-density projects, mirroring other global cities but with a distinctive socio-spatial configuration.
The contribution to the field of urban planning and design is a quantitative examination of residential verticalisation and its territorial drivers. Findings support the existing literature on urban landscape transformations, aligning with studies such as Rojas Symmes et al. (2023), Cerrada Morato (2022), Vergara-Perucich & Aguirre-Núñez (2020) and Nethercote (2018) that emphasise the complexity and varied implications of high-rise development. What distinguishes this work is the empirical typification of verticalisation and the estimation of how density, land use, accessibility, land values and municipal context shape high-rise development.
The logistic and ordered logistic models highlight the complex and multidimensional nature of vertical development in Santiago. Results show a greater propensity for verticalisation in areas with lower socio-economic levels and land values and fewer residential zones, in contrast to more affluent areas where vertical projects are less common, likely due to higher land prices and stronger community resistance to densification. These findings challenge conventional assumptions that associate high-rise development primarily with wealthy municipalities, showing instead a pattern driven by affordability and market dynamics rather than luxury housing demand.
The spatial configuration of the clusters shows that the most intensive forms of verticalisation tend to accumulate in lower income pericentral zones with lower land values and more limited access to public green spaces and urban amenities. This reveals a structural imbalance in Santiago’s verticalisation: affordability-driven projects concentrate in vulnerable areas, reinforcing pre-existing inequalities in access to green space, services and housing quality. Verticalisation thus operates as a spatially selective process rather than a uniform densification strategy. Pre-existing built density also emerges as a significant factor, with denser areas more predisposed to further vertical development. However, extreme densification raises concerns about liveability, as illustrated by the proliferation of micro-apartments averaging 17 m2, where overcrowding diminishes residents’ quality of life. These patterns emphasise the need for planning approaches that balance densification with liveability.
Municipality-specific patterns underscore this territorial selectivity. Estación Central exhibits a significantly higher probability of high-rise residential development, consistently across models, linked to permissive planning instruments, flexible zoning and a strategic location near major transit corridors such as the Alameda and metro lines, together with relatively low land prices and a predominance of underutilised parcels. Conversely, municipalities such as Las Condes, Providencia and Vitacura exhibit markedly lower probabilities of high-rise development. They are characterised by higher land values, well-established residential fabrics and politically active communities that have historically resisted densification. In these areas, planning frameworks prioritise heritage conservation, low-density zoning and environmental quality, creating strong barriers to vertical expansion.
A one-size-fits-all approach to densification may therefore exacerbate spatial inequalities, reinforcing exclusion in high-income communes while concentrating vertical growth and its associated challenges in vulnerable pericentral areas. Differentiated planning strategies are recommended: tailored to the socio-economic, morphological and political characteristics of each territory, so that verticalisation contributes to urban integration rather than fragmentation.
Socio-economic factors are thus pivotal in shaping these patterns, complementing studies such as Vergara-Perucich & Boano (2021), which examine the market’s influence on urban growth and lifestyle changes. The study confirms that lower income zones are more likely to undergo verticalisation, a pattern consistent with the concentration of development pressure where land values are lower and planning instruments more permissive. Conversely, higher income neighbourhoods tend to restrict densification by mobilising political, legal and social resources to preserve low-density environments, protect property values and maintain residential exclusivity. These dynamics reflect the power that affluent groups exert over urban decision-making, shaping planning and development permissions to maintain exclusivity (Vicuña 2020; Vergara-Perucich & Boano 2021).
Although verticalisation can support urban revitalisation and improved access to services (Morigi & Bovo 2016; Fedele & Martínez 2015), it predominantly targets low-income households, often in rental housing, reinforcing existing patterns of stratification (Vergara-Perucich & Aguirre-Núñez 2020). Qualitative studies highlight the challenges of extreme densification in Santiago, such as reduced functional areas of apartments (Rojas Symmes et al. 2023) and overcrowded shared facilities, underscoring how unregulated housing markets exacerbate social segregation and inequality, perpetuating cycles of urban inequity (Vicuña 2020; Paquette Vassalli 2020).
The dimension of social inclusion and exclusion in the context of verticalisation (Lai et al. 2018) is therefore particularly relevant for Santiago. The city’s experience demonstrates the need for careful management of vertical growth to prevent precarious living conditions and social segregation (Orellana et al. 2022). In this regard, the role of public and private investment in shaping socio-spatial orders (Yunda & Sletto 2020) is crucial, as it can either reinforce existing inequalities or be leveraged to promote more equitable access to housing and urban amenities. Environmental implications, such as potential loss of historical buildings, increased pollution and pressure on infrastructure, further highlight the need for sustainable development strategies that mitigate impacts while improving quality of life, especially in vulnerable communities.
Finally, Santiago’s verticalisation is embedded in broader macro-level dynamics of capital flows, housing financialisation and urban neoliberalism (Cerrada Morato 2022; Vergara-Perucich & Boano 2021). During the period 2010–19, strong private-sector dynamism, the expansion of mortgage credit and investment funds, and a relatively permissive regulatory environment created favourable conditions for large-scale vertical projects in pericentral areas. This phenomenon, mediated by local institutional complexities (Glauser 2022), calls for site-specific approaches to urban planning and governance. These dynamics underline the importance of careful planning and design to ensure sustainable and inclusive urban environments (March & Lehrer 2019; Webb & White 2022).
COMPETING INTERESTS
The authors declare they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
DATA ACCESSIBILITY
The code and aggregated indicators used in the analysis are available from the corresponding author upon reasonable request.
SUPPLEMENTAL DATA
Supplemental data for this article can be accessed at: https://doi.org/10.5334/bc.698.s1