Mapping policy pathways: Urban referencing networks in public art policies

This article examines the dynamics of inter-referencing between cities and develops the concept of the ‘Urban Referencing Network’ as a representation of the references made by cities to one another in policy documents. The study employs public art policies, specifically the Percent for Art policy, to investigate the structure of inter-referencing within the urban referencing network. Using a corpus of policy documents from 26 Anglophone cities with over one million residents, we analyse 150 documents containing 2178 inter-references. Combining network measurements and regressions, we explore the emergence of central nodes and the mechanisms influencing their formation. The broader field of arts and cultural policies, with its extensive inter-urban connections and professional networks, provides fertile ground for studying urban referencing networks. By integrating literature on policy mobility and urban networks, this study contributes to a deeper understanding of the circulation of urban ideas and the interplay between cities in policy-making processes. The results demonstrate that only a few cities, including New York, Chicago, London, Seattle, Los Angeles, and Montreal, emerge as central nodes, attracting the other cities’ attention. Attributes of the referenced cities, like economic importance, iconicity and early adoption, determine to a great extent who are the most central nodes.


Introduction
Colloquially known as 'the Bean', Anish Kapoor's sculpture Cloud Gate is a large mirror-like elliptical shape located in Chicago's Millennium Park that reflects the city's skyline and the nearby visitors.Despite its close identification with Chicago, similar 'beans' have mushroomed elsewhere, such as Cleveland, New York, Jerusalem and Karamay (China) -most, but not all, were created by Kapoor.In early March 2018, a heated debate sparked between Chicagoans and Houstonians after the instalment of a new 'bean' in Houston's Glassall School of Art.For example, after Chicago Tribune columnists suggested that the Houston 'bean' will 'act as an additional reminder [for Houstonians] of their poor life choices', the Houston Chronicle fired back, asking whether 'Chicago (is) feeling defensive?' (Capps, 2018).
The inter-referencing between Chicago and Houston and the comparison of the 'beans' exposes one edge of a wider network, in which cities take part when initiating a policy intervention, or a particular public art installation.The comparison between the two cities invites questions like 'who was first?', 'whose ''bean'' is more instragramable?',and also 'which city is better to live in?' (Capps, 2018).Similar comparison and inter-referencing emerge occasionally around various urban models, such as abstract ideas like the Creative City (Oakley, 2012), a particular policy programme like the participatory budget (Peck and Theodore, 2015), or iconic built forms (Patterson, 2012).
However, this comparative mode wherein cities construct 'mental maps' (Bunnell, 2015;Ong, 2011;Patterson, 2016), as suggested is not just an internal, 'mental' or 'cognitive' process.How cities direct their attention to one another has become an integral part of municipal policy making (Peck and Theodore, 2015;Wood, 2022).As the bean instance reveals, cities may be either competing or emulating one another, learning about a particular policy programme, or ranking themselves next to others (Bok, 2021;Gonza´lez, 2011;Lecavalier et al., 2023;McCann and Ward, 2010;Patterson, 2012;Silvestre and Jajamovich, 2023).Interreferencing may include performative elements, such as visualising these mental maps and including them in presentations and documents (Prince, 2014a).It may involve external actors, like professional networks and global consultants (Acuto and Leffel, 2021;Keidar, 2023;Lecavalier et al., 2023), or may be influenced by the broader structural forces of economic and cultural globalisation.
Building both on the inter-referencing literature within the policy mobility scholarship (Bunnell, 2015;McCann and Ward, 2010;Roy, 2011) and on the literature about urban networks (Acuto and Leffel, 2021;Neal, 2012;Neal et al., 2021), in this paper, we show the potential of a systematic investigation of inter-referencing dynamics of a particular policy-making space, namely public art policies.To make the concept of 'mental maps' more empirically grounded, we offer a more articulated concept -that of the 'urban referencing network'.We operationalise this network as a representation of all the references cities make to one another in policy documents within a certain policy model.With a combination of methods including network measurements and multilevel regressions, we ask: what are the contours and attributes of the referencing network?Who are its central nodes?And what are the mechanisms that affect their emergence?
We develop the concept of the urban referencing network via the case of public art policy, and in particular the Percent for Art policy -a popular funding mechanism for public art.To do so, we examine a corpus of policy documents published between 1959 and 2020 in 26 cities with more than one million residents in the Anglophone world.This corpus contains 150 documents with 2178 inter-references made by cities.The broader arts and cultural policy field in which public art takes place is a vibrant and wellestablished domain that fosters meaningful connections between cities.Since its inception in the mid-20th century, arts and culture policies have formed an extensive network of references (Alasuutari and Kangas, 2020).Initially rooted in nation states and national initiatives, this space has evolved into an inter-urban realm as culture gained prominence in municipal policymaking (Oakley, 2012).Furthermore, the establishment of professional networks for urban leaders, such as Placemaking-X (Acuto and Leffel, 2021), has further enhanced this inter-urban space.Notably, cities like Bilbao, Barcelona (Gonza´lez, 2011), Glasgow (Garcia, 2005) and Austin (Grodach, 2012) have emerged as exemplars of cultural-based regeneration within this dynamic environment.As such, it is a fertile case to illustrate the potential and options of investigating Urban Referencing Networks.
The article is structured as follows: We first articulate the core discursive dynamics of inter-referencing processes between cities, following the literature on policy mobility.We integrate existing research that focuses on inter-referencing as 'mental maps' with the scholarship on urban networks, thereby introducing the concept of the 'Urban Referencing Network'.To illustrate this concept, we employ the specific context of public art policies.Subsequently, in the methodology section, we outline the process of constructing the Urban Referencing Network using tools from natural language processing.In the analysis section, we employ network measurements to delineate the contours of the public art referencing network.Lastly, we scrutinise the mechanisms that establish certain cities as exemplary models with a set of regression analyses.Through this comprehensive exploration, our article presents novel avenues for investigating inter-referencing processes between cities and for exploring the circulation of urban ideas.

Inter-referencing cities: From mental maps to urban referencing networks
Policy mobility research investigates the various dynamics by which cities learn from one another, with special attention to how policy models and best practices are generated, diffused, and adapted in circulation across locations (Peck and Theodore, 2015;Keidar and Silver, 2023).Key to policy mobility is the notion of inter-referencing, which refers to the myriad ways in which cities direct their attention to one another, such as by comparing themselves to other cities on various metrics or rankings, looking to one another as examples of approaches to emulate or avoid, or defining themselves as sharing a set of policy goals and interests (Bunnell, 2015).Inter-referencing situates cities in relation to the experience and reputation of other cities (McCann and Ward, 2010;Ong, 2011), and at the same time constructs a policy field by demonstrating various policy applications (Peck and Theodore, 2015).Inter-referencing is often a performative act within local political-economic dynamics of cities (Bok, 2021;Peck and Theodore, 2015), which operates as a coalition magnet (Silvestre and Jajamovich, 2020).For instance, studies reveal that planners and developers in Asian cities employ a common comparative mode, by which cities situate themselves to one another vis-a´-vis rankings on various developmental criteria through featured terminology such as 'need to catch-up'.This sort of referencing serves the purpose of persuading residents to accept controversial projects.The referential mode also reshapes the subjective knowledge and experience of residents towards their cities, offering them, for instance, comparative resources to conceptualise themselves as the city with the biggest shopping mall or the highest tower in the continent (Ong, 2011).Bunnell (2015) describes three main images used in inter-referencing processes: cities as paradigms, cities as pathways, and cities as policy description.First is the paradigmatic city, which depicts a prototype of what all cities will become in the future, more as a prophecy rather than a strategic vision.An example is portraying Los Angeles as the harbinger of planetary urbanisation (Bunnell, 2015), or Jerusalem as representing the complex demographic future of Israel (Keidar, 2018).Second is cities as pathways.Pathway cities are revolving around those positioned at the top of the global hierarchy of urbanity, illustrating the developmental pathway for other cities to follow.New York, London and Tokyo became archetypes for their global status (Sassen, 1991), but also less central cities like Copenhagen (Adelfio et al., 2022) are acknowledged for the 'pathway' they present for good urbanity.Third is cities as a policy description, which presents a model for a policy strategy that can urge an urban transformation.In this category, cities like Barcelona, Bangalore, Bilbao, Porto Alegre, Singapore and Vancouver have transformed into a policy description city because of a particular intervention (Kennedy, 2016).According to Bunnell (2015), each of these images has a referential effect on other cities.For example, the pathway cities offer an open-ended passage to 'world cityness', while policy description cities propose a particular recipe.
Much of the foundational writing about inter-referential practices among cities suggests that inter-referencing creates 'mental maps' by which cities understand themselves (Bunnell, 2015;McCann and Ward, 2010;Ong, 2011).These maps tend to overrepresent political, economic, and cultural centres, while leaving more marginal cities 'off the map' (Roy, 2011).These 'cognitive' relations are not only constructed by the domination of key leading cities.They are increasingly bound up with inter-urban connections defined by financial, professional, informational, and migration networks among cities (Acuto and Leffel, 2021).For example, as consultants become an integral part of municipal policy-making (Hurl and Vogelpohl, 2021), inter-referencing is more than central cities dominating the cognitive maps of others.It is also an institutional process supported by professional urban networks and as part of interactions between municipal agencies with various experts (Keidar, 2023).As the policy mobility literature has recently shown, cities on the 'demand' side like Buenos-Aires and Rio De Janeiro are those which construct model cities, like Barcelona, as reference points (Silvestre and Jajamovich, 2020).Cognitive maps are not only structured by the power of central nodes, but are rooted in a wider array of factors, through which knowledge circulates.
The study of inter-referencing can benefit from being joined with the rich research literature on urban networks (Neal, 2012;Neal et al., 2021).Initially inspired by observations about how globalisation transformed cities into spaces of flows, and is increasingly binding them together in a global and hierarchical space of financial and cultural flows (Castells, 1996;Friedmann, 1986;Sassen, 1991), this research has expanded to study diverse forms of urban networks.Examples include urban networks based on air travel (Smith and Timberlake, 2001), transportation (Derrible and Kennedy, 2009), telecommunication (Choi et al., 2006), scientific knowledge (Matthiessen et al., 2010) and urban governance networks (Acuto and Leffel, 2021).
Urban network studies highlight overlapping hierarchies that vary depending on the type of relationships involved, and often aim to understand the processes that generate particular network configurations among cities.A key task is to identify the most central nodes that dominate the flow of content within the network (Li and Neal, 2022).The centrality of such nodes can be measured in several ways, each highlighting different qualities: Degree counts the number of connections with other cities; Closeness measures the proximity to other cities; Betweenness evaluates the amount of times in which cities operate as links to other cities; and Authority assesses the connections to other important nodes.A complementary element to the nodes' centrality is the edges, which represent the content that flows from one node to the other, whether these are air flights connecting cities (Smith and Timberlake, 2001), or research exchanges linking academic institutions (Matthiessen et al., 2010).
In the context of research on urban interreferencing in policy mobility processes, the urban network approach suggests that the mental maps envisioned by scholars of interreferencing are embodied in the concrete practices by which some cities orient their policy documents towards others.When the author of a policy report refers to some cities versus others, their texts construct a map in which some cities are more central than others.These 'urban referencing networks' extend the more established research on inter-referencing dynamics among cities, joining it with the urban networks literature.

The public art referencing network
Arts and culture policies in general and public art policy in particular are part of a policy domain of extensive inter-referencing.
Studies on municipal and regional interreferencing illustrate how cultural organisations situate themselves within their broader cultural ecologies (Burnill-Maier, 2023;Comunian, 2011;Gilmore, 2013;Kaddar et al., 2022;Patterson, 2016;Yavo-Ayalon et al., 2019).However, the study of cultural production networks of a global scale is still relatively new (Kloosterman et al., 2019;Zhang and Chen, 2021).Even so, from its emergence in the middle of the 20th century, formal cultural policy developed in an international referential space.In the late 1960s, for example, UNESCO invited member states of multiple regions to prepare national cultural policy reports, leading many of them to open ministries of culture for the first time.UNESCO utilised peer pressure as a main circulation mechanism of the cultural policy concept, encouraging member states to include international comparisons and inter-references to other cities in their reports (Alasuutari and Kangas, 2020).
In recent decades, cultural policies also became more central in municipal policymaking.As a result of post-industrialisation and rising quality of life concerns, cities increasingly see the provision of cultural services and the construction and maintenance of cultural amenities as a core responsibility (Grodach and Silver, 2012;Silver and Clark, 2016).Cities now share their knowledge in various referential spaces, like in UNESCO's creative networks, and PlacemakingX networks.At the same time, global consulting firms advise cities on how to advance their cultural policy strategies, while extensively using comparative measurements and indexes (Bok, 2021;Prince, 2014b).In addition, within these inter-referential spaces, particular cities became models of cultural regeneration: Bilbao is known for its transformation by its starchitect designed iconic museum (Gonza´lez, 2011), Glasgow for regeneration based on cultural industries (Garcia, 2005), and Austin for its creative class magnetism (Grodach, 2012).Cultural policy thus represents a well-documented and established policy field, with fertile ground for studying referential effects and discursive relations among cities.
Public art policies are a subset of cultural policies, especially apt for examining urban referencing networks.Studies show how public art discourses express various policy dimensions, including the city's cultural identity, its socio-economic order, and the spatial experience of the urban fabric (Keidar and Silver, 2022;Mendelson-Shwartz and Mualam, 2021).In all dimensions, public art policy deals with the aesthetic experience of the city, highlighting both local identity as well as global iconicism.In a globalised world that emphasises the close relationship between culture, economy and space (Lash and Urry, 1994), examining the inter-referencing of public art allows us to emphasise the spread of cultural meanings and symbols.
To explore the urban referencing network of public art, we ask: 1. What are the contours and attributes of the referencing network? 2. Who are the cities that emerge as prominent reference points, and why?
While our past research had described the emergence of public art discourses within a policy space of a global scale (Keidar andSilver, 2022, 2023), this current approach examines whether the mobility of these discourses is driven by a network, and how important peer cities are for the circulation of ideas.

Methodology
To investigate our questions, we pursued a three-stage methodology.First, we constructed our reference network using natural language processing tools and described it with standard network statistics.Second, we gathered metadata to understand its structure, overall, and in terms of out and inreferencing.Third, we analysed the resulting dataset (reference network matched with metadata) using a combination of regression analysis and descriptive visualisations.
The public art referencing network is based on a corpus of 150 public art related documents.We found these documents online for our sample of cities, which includes those with more than one million residents in the Anglosphere (e.g. the United States (N = 13), Canada (N = 6), Australia (N = 5), and, the United Kingdom (N = 2)) -a total of 26 cities. 1 The corpus includes different types of documents, some more bureaucratic, while others are investigative and open-ended.Although different in content and length, all include inter-references to other cities.The oldest document is Philadelphia's 1959 guidelines for the 'Adoption of Percent for Art', which initiated the policy field.The most recent documents are from 2020. Figure 1 plots the number of documents by decade. 2

Building the reference network
We built the reference network by the following steps.Starting with a global list of city names from the R package 'maps', we identified all instances of each city name in each document in our corpus.We manually examined the results to remove references that were not to cities but to the names of individuals, streets, plazas etc. 3 From this, we generate an edge list, in which two cities are connected if one city mentions the other -these edges are our basic unit of analysis (N = 2178).The edge weight is determined by the number of times they mention one another.This is a directed graph, based on who mentions whom: for example, if Toronto mentions Montreal, but Montreal does not mention Toronto, this is a different relationship than the other way around.We also calculate standard network centrality measures, focusing on degree centrality (in and out) and authority.

Understanding the reference network
To pursue questions about the basis of the reference network, we connect the network measures to other metadata.A first set of metadata covers the cities that generate policy documents, which we call 'referencing cities' (N = 26).For each city, we generate variables describing the time and the location of publication.We also added variables describing the context in which each document was produced: first, the administrative body that manages public art in the city, distinguishing between 'Arts and Culture Departments' (N = 15), 'Planning/Design Departments' (N = 3), 'Arms-Length Organisations or Redevelopment Authorities' (N = 5), and 'Mixed Forms' of the above (N = 3).Second, we categorised each document according to its content, style and function, distinguishing between 'Guidelines' (N = 34), 'Public Art Plans', (N = 49), 'Culture Plans' (N = 54) and 'Research Reports' (N = 12).
Metadata connected to the 205 referenced cities is in line with the urban networks literature.We particularly focused on variables relevant to the sub-field of public art: Economic Importance -based on the Globalisations and World Cities Research Network (GaWC) index; Iconicity  Note: The plot depicts the public art referencing network for our complete corpus of policy documents over the entire period (1959-2020), with nodes representing cities and edges representing the direction and intensity of referencing.It is a simplified version of the full network, as it includes only nodes with in-degree centrality greater than 5. Node size is based on in-degree centrality scores representing how many mentions each city received, and node colour is based on out-degree centrality, with darker shades indicating cities that refer to other cities at higher rates.New York is the most referenced city, while Toronto is the highest referencing city.Edge thickness is proportional to edge weight, with thicker lines representing more referencing from one city to the other.The thickest edge shows the extensive referencing of Toronto to Montreal.
Finally, we examine this compiled information using standard analytical methods.Multi-level regression models help investigate factors that predict which cities are more or less referenced.We include the referencing city as a level two variable, in order to control for referencing patterns that may be driven by distinctive practices of individual cities.Table 1 lists all the factors included in the regressions.
This three-stage methodology indicates how the attributes of the referencing network and its prominent reference points could be investigated comprehensively.However, as with any attempt to represent the social world, the public art referencing network we empirically generated is skewed by various factors, in this case, especially city size and language, so that it captures the mental map of big and newer cities of the anglosphere.It does not represent, for instance, inter-referencing of Italian cities using public art for culturalhistoric purposes (Pratt and Hutton, 2013), or the more recent extensive inter-referencing of Asian cities using public art for enhancing their cultural industries (Gu et al., 2020;Kong et al., 2006;Tran, 2023), or even the referencing practice of shrinking cities in the Anglosphere (Zebracki et al., 2019), which often use arts and culture for urban revitalisation.The choice of the referencing cities sample clearly shapes which segment of the overall mental map would be observed, and exploring expanded and alternative samples is an important area for further research.

The public art referencing network: Results
What are the contours and attributes of the referencing network?
The public art referencing network represents various kinds of referencing patterns.Figure 2 depicts a simplified version of this full network over the entire period, showing only nodes with degree centrality greater than 5. Node size is relative to a referenced city's in-degree centrality, to highlight cities that attract the most mentions.Node colour is based on outdegree centrality, with darker shades indicating cities that refer to other cities at higher rates.Edge thickness is proportional to the edge weight, with thicker lines representing more referencing from one city to the other.
Cities in the network can be sorted into one of four types, according to their overall degree centrality results, composed of being referenced by others (In-Degree, 'ID') and referencing others (Out-Degree, 'OD').The first is cities that are both highly referenced by others and highly reference others, like New York (ID = 19, OD = 21), Los Angeles (ID = 13, OD = 16), and San Francisco (ID = 12, OD = 14); these cities are depicted in Figure 2 by larger nodes with darker colours.The second is heavily referencing cities that are not referenced by others, like Phoenix (ID = 4, OD = 34), and Calgary (ID4, OD = 32); these cities are depicted by smaller nodes and darker colours.The third is cities that are highly referenced by others, but are not referencing themselves, like Seattle (ID = 15, OD = 0), London (ID = 14, OD = 0) and Portland (ID = 10, OD = 0); these cities are depicted by larger nodes and a white colour.The fourth is the majority of the sample, that includes cities with lower volume of referencing others and being referenced; these cities are not represented in Figure 2. The last two types are shaped by our sample method, which only includes policy documents from 26 cities in English-speaking countries with populations larger than one million residents.Theoretically, in a different sampling of documents, the cities in the third and the fourth group may be included in the first or the second group.
Who are the most central cities of the public art referencing network?The cities with the highest in-degree centrality (the largest nodes in Figure 2) appear to be mainly 'pathway cities', in Bunnell's (2015) terms, like New York (ID = 19), London Note: These figures show the highest ranked cities according to their in-degree centrality and their authority indexes.In-degree counts the number of cities that reference a particular city, and ranges in our sample from 0 to 19.Authority prioritises cities that are referenced by other highly referenced cities, and ranges from 0 to 1 (Kleinberg, 1998).
(ID = 14), and Los Angeles (ID = 13).Other than London, both New York and Los Angeles are part of the first referencing type -cities which are both highly referencing and referenced.
However, two cities illustrate alternatives to the 'pathway city' portrayal.The first is Chicago, which was found to be the second most referenced city (ID = 16), despite the fact it is sometimes said to suffer from a 'network void' and found to be 'off the map' of some important world city networks (Derudder and Taylor, 2021).In this sample of public-art policy documents, Chicago is mentioned by most of the referencing cities, in some cases with a direct reference to the 'Bean'.The City of Sydney, for instance, speaks about Cloud Gate as 'the poster child for the city it transformed'.The City of San Antonio speaks about Millenium Park in which the 'Bean' is located as a 'premier example of a contemporary art collection with engaging and interactive installations located in an urban park setting'.The second is Seattle, which was not included in the sample of referencing cities, but still stands out as the third most referenced city (ID = 15), despite its relatively smaller population size and lower position in the economic global ranking (7 out of 12 in the GaWC scale).Both Chicago (Flanagan, 2008) and Seattle (Finkelpearl, 2000) are recognised in the literature as a public art 'policy description city' (Bunnell, 2015).This may be a factor that determines their centrality.
However, there is more than one way to measure centrality.The authority score (Figure 3), which captures cities that are referenced by other highly referenced cities, slightly changes the list of most prominent reference points.Montreal, not a central city within the global economic network but rather an early adopter of the Percent for Art policy, is the most central city, with a big advantage over New York and San Fransisco on this metric.This emphasises the importance of being among the first and having a long-term commitment to the policy issue.Other surprising entries are the Canadian cities Mississauga and Winnipeg.The multi-level regression analyses allow us to systematically test the mechanisms that lead particular nodes in the network to become the most referenced cities.
Figure 2 also presents the structure of ties that buttress the rise of these reference points.The strength of these ties is represented by the thickness of the edges.Overall, the connections are often unidirectional.Montreal and Toronto, for instance, are tightly connected by the initiative of one side.Toronto, which is the highest outreferencing city (OD = 50), references Montreal extensively -103 times in four different documents.This likely influences Montreal's high authority score.However, the opposite rarely occurs.Montreal only references Toronto twice.Montreal is also referenced by Calgary, Chicago, Dublin, Ottawa, Sydney and Vancouver, but none references Montreal at the same volume that Toronto does.The multi-level regressions below, which include the referencing cities in the model as random effects, control for these city-specific referencing patterns.
Additional strong unidirectional connections can be seen between Philadelphia.which extensively references Los Angeles, and from San Antonio to Houston.Such unidirectional connections may cross nation states, as seen for instance the Toronto to Chicago relationship.They may also occur between cities that are less central in the global economy, like Calgary, which references San Jose intensively.The strong presence of unidirectional ties emphasises what Silvestre and Jajamovich (2020) calls the 'demand' side of reference points.Toronto, for instance, often uses Montreal in order to make a political statement locally.However, the lack of bidirectional ties also highlights that inter-referencing is often not a dialogue between cities.
Who are the cities that emerge as prominent reference points and why?
Figure 4 shows the results of multi-level regressions, capturing how different attributes of the referenced and referencing cities are associated with the centrality of cities within the public art referencing network.
Among the attributes of the referenced cities, economic importance, iconicism and population size may align with what Bunnell (2015) describes as the 'pathway cities' image, while attributes like early adopters and membership in UNESCO's creative city networks may align with the 'policy description in Note: This plot shows the results from two different multilevel regression models.Dependent variables are in-degree centrality (blue) and authority (red).The nodes represent the coefficients, and the lines delineate confidence intervals.The fixed-effects integrated in the models are divided into two groups, and distinguished by the dotted lines.First, above the dotted line, are attributes of the referenced cities, including economic importance, iconicity, population, membership in UNESCO, early adopter, capital and North American cities.Second, below the dotted line are attributes of the referencing cities, including document type (cultural plan is the reference category), and administrative unit (armslength organisations and redevelopment authorities is the reference category).The random effects capture variance among referencing cities (N = 26), and are not included in this figure for efficiency.cities' image.We review the results along these images.Moreover, we examine cities' centrality both in terms of the number of cities referencing them (in-degree score, in grey), and the extent by which they are referenced by other referenced cities (authority score, in black).We compare the results to distinguish between the different qualities of referential centrality: one based on wider public recognition, and the other on the recognition of connoisseur cities.
An overall look indicates that the attributes of the referenced cities have a strong and significant effect, while the contextual characteristics of the referencing cities are less salient.Among the attributes of the referenced cities, their economic importance (GaWC ranking) and their iconicism (mentions in experts' lists of iconic artworks) have the strongest positive effect on the degree of references, and a weaker effect on the authority score.Population size has a lower but significant positive effect as well, but its main effect is on the authority index.The significance of these three factors alludes to the gravity of being a 'pathway city' (Bunnell, 2015) -a widely known leader of urbanity, rather than an expert in a particular policy field.
The authority score, however, shows the importance of policy expertise.Other than authority's strong correlation with the population size of referenced cities, authority is found to be highly correlated with early adoption cities, and with members in UNESCO's Creative Networks.Being an early adopter is also positively correlated with in-degree centrality.These associations suggest that cities which engage with the policy area over time and through professional networks tend to become touchstones for other influential cities.
As for other attributes of the referenced cities, a North American domination of the referencing network is also shown with the higher centrality scores of North American cities, in terms of both degree and authority.Somewhat surprisingly, capital cities are not central within the referencing networks, in terms of neither degree nor authority.In the sample these are Ottawa and Washington DC, but also world leading cities like London and Paris.
As noted, the contextual factors in which these references were made -the type of document and the department in which it was produced -are in general less salient.For document type, overall, culture plans (the reference group) are the documents that influence centrality the most.In terms of indegree centrality, all other document types are negative when compared to cultural plans, though the difference with guidelines is not significant.In terms of authority, only research documents, which tend to explore policy models in length, have a stronger effect on the authority of the cities mentioned in them.For the department in which the documents are produced, arms-length institutions and redevelopment authorities (the reference group) are the institutions with the highest in-degree centrality.In other words, these organisations that administer public art outside the official city bureaucracy are the ones with the highest influence on wider public recognition.However, in terms of authority, the other departments, whether planning, arts and culture or a hybrid, have more influence on being referenced by other referenced cities, though the differences are insignificant.

Discussion
We opened by highlighting the playful yet telling exchange between Chicago and Houston regarding their respective 'beans' in local newspapers.This anecdote provides a glimpse into the broader network of intercity interactions and comparisons.The foundational policy mobility literature described such comparative practices as generating 'mental maps' of urban futures and policy pathways.Building on this notion and integrating it with concepts and methods from studies of urban networks, we formulated the concept of an 'urban referencing network' to facilitate the systematic examination of inter-referencing dynamics.
To demonstrate the value of this framework, we conducted an empirical investigation using public art policy documents as a case study.Our primary line of inquiry revolved around delineating the contours and characteristics of the public art referencing network.As an aesthetic intervention intricately linked to the cultural identity, socio-economic order, and spatial fabric of cities, public art encompasses diverse policy dimensions.Thus, analysing the public art referencing network sheds light on the circulation of aesthetic meanings and symbols between cities.Our sample encompassed 150 documents from 26 cities, containing references to over 200 cities and over 2000 distinct inter-city references.To elucidate the network's structure, we employed centrality metrics and visualisations.
The analysis surfaced a select group of pivotal nodes that attract and concentrate other cities' attention.However, referencing often follows a unidirectional pattern, with one city mentioning another without reciprocation.This unidirectionality highlights the 'demand side' of policy ideas (Silvestre and Jajamovich, 2020), suggesting that dialogic exchange within this network may be limited.
We then probed the factors that contribute to a city's prominence within the network by collecting metadata specific to the public art policy field.Using multilevel regressions, we tested correlations between city attributes and two centrality measuresin-degree, capturing overall references, and authority, focused on referencing by influential peers.Attributes linked to Bunnell's 'pathway cities' like economic importance and public art iconicity displayed stronger correlations with in-degree centrality.
Conversely, attributes associated with 'policy description cities', like UNESCO Creative Cities membership and early adoption, exhibited stronger correlations with authority centrality.Based on these mechanisms, two city clusters emerge -global metropoles like New York and cultural pioneers like Montreal.The concentration of influence among so few nodes, given the network's limited dialogic structure, highlights the power of select cities in shaping aesthetic discourses.This echoes the global cities literature' emphasis on the inequality between global cities and the rest, demonstrating that this inequality also exists in the aesthetic realm.When further applying the urban referencing network approach into other policy fields, domain specific data that may distinguish between 'pathway cities' and 'policy description cities', and between overall references and centrality based on policy 'authority', should be used.
While revealing, this analysis has limitations that provide avenues for future research.The sample focuses exclusively on large Anglophone cities and their inter-referencing practices.In revealing their mental map, the urban referencing network we find may potentially overlook alternate central nodes and mechanisms in other regions, and in smaller cities.Expanding the sample, or focusing on a different segment of the mental map, for instance, on inter-references of cities in 'classical Europe', of cities in Asia, or of smaller cities in the Anglosphere, could enrich the observed network dynamics.Moreover, the corpus is based on policy documents, while the textual practice of inter-referencing also appears in the grey literature (Saleem et al., 2023), policy speeches (McCann, 2013), conferences (Silvestre andJajamovich, 2023), and in newspapers and specialised media (Duque Franco and Ortiz, 2020;Whitney, 2022), as the exchange between Chicago and Houston has shown.Expanding the analysis to additional corpuses may highlight additional dimensions of mental mapping, for instance, which type of inter-references are directed to residents, which to stakeholders, and which to other cities.Additionally, we were unable to fully explore the role of document authors due to data constraints.As consultants and specialised agencies are often the actors who actually produce these documents for cities, they have a key role in generating these mental maps.
While examined here through public art policies, the urban referencing network approach could fruitfully be applied to diverse policy domains.Mapping out referencing networks on issues like climate change, affordable housing, or smart cities could reveal distinct network contours and central nodes.Comparing networks across policy areas may uncover certain cities that consistently shape discourses and circulate ideas across multiple domains, as well as specialists in distinct policy areas.Tracking referencing networks longitudinally could also elucidate the co-evolution of network position and policy diffusion.For instance, as Chicago gains prominence in the public art network, we may observe other cities increasingly adopt and reference Chicago's approach, becoming more 'Chicago-like' over time.
Expanding the framework to incorporate each city's discursive tendencies would further enrich the analysis.Cities exhibit distinct styles and logics when discussing public art, rooted in local discourses and meanings (Keidar and Silver, 2023).Integrating measures of cities' rhetorical patterns could elucidate the interplay between discourse and network position.We could examine whether certain rhetorical styles enhance a city's authority within the network.Similarly, examining how each referenced city tends to be discussed may reveal particular attributes or policy elements that garner attention.For example, referencing patterns may highlight innovative funding mechanisms in some cities, striking architecture in others, or economic vitality others.Incorporating these discursive dimensions surrounding both the referencing and referenced cities could unpack the cultural symbols and meanings diffusing through and structuring the network.
Overall, this study demonstrates the potential of the urban referencing network approach for systematically mapping flows of policy ideas between cities.The framework's flexibility enables application across policy domains and integration with discourse analysis.Extending this approach through future research would further advance understanding of how urban policy models emerge, circulate, and reshape the subjective visions of what cities can or should become.The urban referencing network provides a valuable tool for unravelling the discursive construction of urban futures across an increasingly connected but unequal global landscape of cities.

Figure 1 .
Figure 1.Number of policy documents over time.

Figure 2 .
Figure 2. Simplified version of the public art referencing network.

Figure 3 .
Figure 3.The most referenced cities, measured through in-degree and authority.

Figure 4 .
Figure 4. Factors predicting centrality in the public art inter-referencing network.
(Rizzo, 2021) is composed of lists with titles like 'Most Iconic/fascinating/legendary/famous Public Art Around the World' taken from a variety of online sources in English.These sources represent different types of interests in public art: Three lists were composed by the arts, culture, architecture and design magazines, The Artists (2021), Architectural Digest (Mafi and Cherner, 2019) and CONASU ¨R (2019); another two lists were curated by the travel magazines Far and Wide (Lemmin-Woolfrey, 2023) and Departures(Rizzo, 2021); one list was collected by the BBC(The Lonely Planet,  2010); and another one by The Collector (Lesso, 2022) -a magazine of a broader humanistic scope that includes history and philosophy.