Essays on location choice: agglomeration, amenities and housing

(1)

Thesis

Reference

Essays on location choice: agglomeration, amenities and housing

BOUALAM, Brahim

Abstract

Why do people live where they live and what are the consequences of these decisions?

People and firms allocate themselves unevenly across space. In spite of large costs incurred when living in metropolitan areas, economic agents remain extremely concentrated in a finite number of cities. This thesis explores several factors that may influence these location decisions and their effect on economic agents. The first essay questions the role of cultural amenities in explaining the location of firms and residents across American cities. Next, the second essay examines a consequence of the geographic concentration of workers. It evaluates the influence of urban density on the quality of the match between workers' education and their occupation. Finally, the last essay documents the importance of housing vacancies in France. It examines how distance and access to the city center affect the housing vacancy rate in suburban municipalities.

BOUALAM, Brahim. Essays on location choice: agglomeration, amenities and housing. Thèse de doctorat : Univ. Genève, 2015, no. GSEM 12

URN : urn:nbn:ch:unige-753990

DOI : 10.13097/archive-ouverte/unige:75399

Available at:

http://archive-ouverte.unige.ch/unige:75399

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

agglomeration, amenities and housing

by

Brahim Boualam

A thesis submitted to the

Geneva School of Economics and Management, University of Geneva, Switzerland,

in fulfillment of the requirements for the degree of PhD in Economics

Members of the thesis committee:

Prof. Frédéric Robert-Nicoud, Supervisor, University of Geneva Prof. Céline Carrère, Chair, University of Geneva

Prof. Marcelo Olarreaga, University of Geneva

Prof. Gilles Duranton, Wharton School, University of Pennsylvania

Thesis No. 12 August 2015

(3)

(4)

J’adresse mes premiers remerciements à Frédéric Robert-Nicoud, mon directeur de thèse.

Sa disponibilité constante, ses relectures minutieuses, et ses nombreux conseils ont été précieux lors de l’écriture de ce travail. J’ai eu une très grande chance de bénéficier d’un encadrement aussi remarquable et de réaliser ainsi cette thèse dans les meilleures conditions. Je te remercie fortement pour ton investissement dans mon travail et pour tout ce que tu as su m’apprendre sur le plan de la recherche.

Je tiens à remercier l’ensemble des membres de mon jury de thèse, à commencer par le Professeur Gilles Duranton qui a accepté de lire et de commenter ce travail. La qualité des commentaires et suggestions re¸cus lors de mon colloque privé m’ont permis (et me permettront) d’améliorer chacun de ces chapitres.

Je remercie également Céline Carrère – présidente de mon jury – et Marcelo Olarreaga pour avoir suivi l’élaboration de cette thèse durant plusieurs années. Ce suivi actif, de même que vos nombreux conseils et encouragements m’ont été très bénéfiques. Ce fut un véritable plaisir de travailler avec vous, tant sur le plan de la recherche que de l’enseignement.

Je tiens également à remercier toutes les personnes qui m’ont accompagné pendant mon doctorat. Je remercie tous mes collègues de l’Université de Genève : professeurs, as- sistants et amis. J’ai pris un immense plaisir à travailler à vos côtés. Un remerciement particulier à Grégoire – qui au quotidien a fait d’un jeune Parisien survolté un Genevois beau- coup plus mesuré – et Pramila pour les nombreuses expériences d’enseignement partagées.

J’adresse un immense merci à Vanessa, Virginie, Joëlle W. et Joëlle D.M. pour avoir partagé mon quotidien pendant plus de cinq ans. Votre optimisme, votre bonne humeur, votre légèreté, votre sens de l’humour et de la dérision, votre capacité à vivre des situations dont nous seuls avons le secret et votre fidélité sans faille... bref vos innombrables qualités m’ont changé ! Un immense merci pour ne jamais m’avoir laissé m’apitoyer sur mon travail et la difficulté qu’il représentait par moment. Et surtout merci pour le nombre infini de journées, soirées, week-ends et voyages passés ensemble : je n’oublierai jamais ces moments.

Un remerciement également au reste de la folle équipe qui m’a motivé, entouré et

´

epaul´e, et qui inclut notamment Suzanne, Jenn, Alej, Maria, Anja, Tristan, Rose, Marco, Dany... et tellement d’autres que je ne peux citer mais que je remercie.

Enfin, je remercie les personnes qui m’ont accompagné bien avant mes années de thèse. Ma famille, et en particulier ma mère pour l’exemple de courage et de patience qu’elle est pour moi. J’adresse un remerciement particulier à ma sœur aˆınée, Nadia, pour ses nombreux encouragements en dépit de sa perplexité quant à mes centres d’intérêt académiques ! Merci à Caroline : il me faudrait 100 pages supplémentaires pour dresser la liste des choses pour lesquelles je te suis reconnaissant ; Flora, pour m’avoir offert le monde ; Jérôme, qui a toujours eu davantage confiance en moi et en mon travail que je n’en ai jamais eu : ton soutien et ton aide ont été précieux. Un merci à mes amis de Paris

(5)

et en particulier `a Fany et Sofia : les changements de villes ou de continents n’ont pas eu raison de notre amiti´e !

Enfin, un immense merci à l’ensemble des personnes qui, à un moment ou à un autre, ont croisé mon chemin et ont contribué, par leur bienveillance et leur présence, à faire de mes années d’études des années heureuses et incroyablement enrichissantes.

(6)

People and firms allocate themselves unevenly across space. In spite of large costs incurred when living in metropolitan areas – such as congestion, pollution or high land prices – households and firms remain extremely concentrated in a few number of cities. In the U.S., the three biggest urban areas (New York, Los Angeles and Chicago) host 13% of American residents. This share even exceeds 25% in the Greater Tokyo area, where 35 millions of Japanese inhabitants are concentrated. The location of residents and firms is not an exogenous phenomenon. It results from the aggregation of individual decisions, referred to as a location choice. This thesis aims at exploring several factors that may influence these microeconomic decisions, and their effect on economic agents.

The first essay questions the role of cultural amenities in explaining the location of firms and residents across American cities. As many cultural policies have been justified by the positive effect of cultural vitality on the attractiveness of cities, this chapter asks if a better cultural milieu can actually make a city more livable for residents and firms.

This empirical work shows that, even if the arts and culture might be appealing to some agents, their impact on location decisions is rather moderate.

The second essay examines a consequence of the geographic concentration of workers. It evaluates the influence of urban density on the quality of the match between a worker’s education and her occupation. It focuses on a new dimension of labor matching by measuring the link between an individual’s field of study and the tasks she will be performing in her work. Acknowledging the fact that the distribution of workers across space is not random, this work shows that the quality of skill match improves with the density of the labor market. It also proves that both a better skill match and a higher city size positively affect individual earnings.

Finally, the last essay documents the importance of housing vacancy rates in France.

It examines how access to the city center affects the housing vacancy rates in the suburbs.

After describing the evolution, characteristics and spatial distribution of vacant residential units in France, this essay emphasizes – both theoretically and empirically – the link between access to the city center and housing vacancies. The analysis shows that distance to the core city – and thus to the main job center – is associated with more vacancies. This result is consistent with a negative impact of distance on the matching process between landlords and renters.

(7)

(8)

Comment expliquer la localisation géographique des populations et quelles en sont les principales conséquences ? Les individus et les entreprises ne sont pas répartis de fa¸con uniforme sur le territoire. En dépit de larges coûts supportés par les agents économiques dans les villes – tels que les problèmes de congestion, de pollution ou de loyers élevés – les ménages et les firmes restent très largement concentrés dans un nombre restreint de métropoles. Aux Etats-Unis, les trois aires urbaines les plus peuplées (New York, Los Angeles et Chicago) accueillent 13% des résidents américains. Cette proportion atteint jusqu’à 25% dans l’agglomération de Tokyo, où 35 millions d’habitants sont concentrés.

La localisation des résidents et des entreprises n’est pas un phénomène exogène. Elle résulte de l’agrégation de décisions individuelles, désignées sous le terme de choix de localisation. Cette th`ese a pour objectif d’explorer plusieurs facteurs qui influencent ces décisions microéconomiques, et leur effet sur les agents économiques.

Le premier essai questionne le rôle des aménités culturelles dans la localisation des entreprises et des résidents entre villes américaines. Un grand nombre de politiques culturelles ont été justifiées par l’effet positif du dynamisme culturel sur l’attractivité des villes. Aussi, ce chapitre tâche de savoir si un meilleur environnement culturel permet en effet de rendre une ville plus attractive pour les habitants et les entreprises. Ce travail empirique montre que même si les arts et la culture peuvent être considérés comme attrayants pour un certain nombre d’agents économiques, leur impact sur les choix de localisation est plutôt modéré.

Le second essai étudie un des effets de ces phénomènes de concentration spatiale. Il

´

evalue l’influence de la densité urbaine sur la qualité de l’appariement entre le cursus scolaire des travailleurs et leur emploi. Il s’intéresse à une dimension originale de cette question en mesurant le lien entre le domaine d’étude et le type de tâches réalisées par les travailleurs dans leur emploi. Tenant compte du fait que la distribution des travailleurs dans l’espace n’est pas un phénomène aléatoire, ce travail montre que la qualité de l’appariement s’accroˆıt avec la taille du marché du travail. Il montre par ailleurs qu’un meilleur appariement entre éducation et emploi, de même qu’une plus grande taille de la ville, affectent tous deux positivement le salaire per¸cu par les travailleurs.

Enfin, le dernier chapitre documente l’importance des logements vacants sur le marché immobilier. Il questionne ainsi l’effet de l’accès au centre-ville sur les taux de logements vacants en périphérie. Après avoir décrit l’évolution, les caractéristiques et la distribution spatiale des logements vacants en France, cet essai souligne – à la fois théoriquement et empiriquement – le lien entre distance et accès au centre-ville d’une part, et l’ampleur de la vacance immobilière d’autre part. L’analyse montre que la distance au centre – et donc au principal bassin d’emploi – est associée à davantage de logements vacants. Ce résultat peut s’expliquer par un effet négatif de la distance géographique sur le processus d’appariement entre propriétaires et locataires.

(9)

(10)

Acknowledgements i

Abstract iii

R´esum´e v

Introduction 1

1 Does culture affect local productivity and urban amenities? 9

1.1 Introduction . . . 9

1.2 Related literature . . . 11

1.3 Identification strategy . . . 13

1.4 Data description . . . 15

1.4.1 Measuring culture . . . 15

1.4.2 Dependant variables: wages and housing rents . . . 16

1.4.3 Control variables . . . 17

1.4.4 Stylized facts on culture, wages and rents . . . 18

1.5 Empirical results . . . 19

1.5.1 Baseline regressions . . . 19

1.5.2 Instrumental variables . . . 24

1.5.3 Robustness checks . . . 25

1.6 Concluding remarks . . . 36

2 Getting a first job: quality of the labor matching in French cities 43 2.1 Introduction . . . 43

2.2 Data . . . 47

2.3 Measuring matching . . . 48

2.4 Empirical strategy . . . 51

2.5 Empirical results . . . 52

2.5.1 Baseline regressions . . . 52

2.5.3 Instrumental variables . . . 59

2.5.4 Differences in ability and sorting . . . 66

2.6 Skill match and urban wage premia . . . 72

2.7 Conclusion . . . 80

3 Access to the city center and housing vacancies in the suburbs 87 3.1 Introduction . . . 87

3.2 Stylized facts on housing vacancies . . . 90

(11)

3.2.1 Size and evolution of vacancies . . . 91

3.2.2 Distribution of vacant dwellings in the Paris area . . . 93

3.2.3 Characteristics of vacant dwellings . . . 94

3.3 Theoretical model . . . 96

3.3.1 Model setup . . . 96

3.3.2 Renters’ utility . . . 97

3.3.3 Landlords . . . 98

3.3.4 Rent determination . . . 99

3.3.5 Equilibrium and comparative statics . . . 100

3.4 Commuting patterns in the Paris area . . . 104

3.5 Empirical analysis . . . 105

3.5.1 Baseline results . . . 106

3.5.3 The effect of commuting time . . . 114

3.6 Conclusion . . . 118

Conclusion 125

(12)

(13)

(14)

“220 millions Americans crowd together in the four percent of the country that is urban. 35 million people live in the vast metropolitan area of Tokyo, the most productive urban area in the world. [...] We choose to live cheek by jowl, in a planet with vast amounts of space.” (Glaeser,2008)

Glaeser (2008) shows that if the entire world population was randomly distributed within the State of Texas, each four persons household would be surrounded by 400 square meters of free area with no one else around. We could enjoy a life without congestion, with free space in a less polluted and noisy environment. By contrast, people choose to live close to each other and firms decide to locate in clustered areas where many of their competitors are also concentrated. In the U.S., the top-100 economically active counties account for more than 40% of total employment while representing only 1.5% of total U.S land area (Hanson, 1999). Similarly, Fujita and Thisse (2002) emphasize that one core region in France, “Ile de France (the metropolitan area of Paris) which accounts for 2.2% of the area of the country and 18.9% of its population, produces 30% of its GDP”.

This pattern of spatial concentration is particularly strong in urbanized areas. Few cities indeed manage to concentrate a large share of economic activities and populations. In the U.S. for instance, more than 80 per cent of total population lives in an urbanized area, among which 13% lives in the three biggest cities (New York, Los Angeles and Chicago)¹. Besides, the attractiveness of cities even strengthens over time. By the end of the 20^th century, 75% of European population was living in urbanized areas against 38% in 1900 and 12% in 1800 (Bairoch, 1985).

The location of residents and firms results from the aggregation of individual decisions, referred to aslocation choice. The standard economic approach for understanding such individual choices starts with a comparison between the costs and benefits induced by this decision. Therefore, what are the gains and losses associated with spatial agglomeration for inhabitants and firms?

Several costs induced by agglomeration are well-known. For instance, housing is more expensive in highly populated cities. In Paris for example, the average rent reached 21.6 euros per square meters in January 2013 against 9.6 in the second biggest French metropolitan area². This fact is illustrated in Figure 1. The graph shows the correlation between housing rents and city population in American cities. It reveals that rents are positively correlated with city size.

Other costs are also well established: time for commuting, congestion, pollution or insecurity are also positively correlated with urbanization. As an illustration, Figure 2

1Author’s calculation based on data from the U.S. Bureau of Labor Statistics

2Data from the Rent Observatory for the Paris region (OLAP).

(15)

Figure 1: Rents and population

Data on rents: U.S. Department of Housing and Urban Development (HUD) Data on population: U.S. Census

displays the relationship between crime rates and population in U.S. cities. Crime rates are indeed positively and statistically correlated with urban population.

Figure 2: Crime rates and population

Data on crime: Federal Bureau of Investigation (FBI) Data on population: U.S. Census

Urban costs are many. So, why do people and firms still want to locate in big cities?

The answer relies on the gains acquired by residents and firms when located in crowded areas. From the viewpoint of inhabitants, the simplest conjecture is that monetary costs induced by urban concentration – such as high rents or high prices – are compensated by higher wages. In such a case, purchasing powers are equalized across cities such that workers face a simple trade-off between high wages and high prices. Available data support this idea, as shown in Figure 3. As population increases, the average wage of workers rises. This phenomenon is the so-called urban wage premium. Estimating this elasticity of wages with respect to city size is a topic of extensive research in the field of urban economics. Wage elasticities with respect to urban density traditionally range between 0.04 and 0.1 (Combes et al., 2010). A general discussion on the urban wage premium and the associated literature is provided in the second essay of this thesis.

(16)

Figure 3: Wages and population

Data on wages: Occupation Employment and Wage Estimates, Bureau of Labor Statistics Data on population: U.S. Census

However, estimates of wage and rent elasticities with respect to city size do not necessarily provide similar results. This is the case when comparing estimates provided by Figures 1 and 3. Similarly, Albouy and Stuart (2013) underline that “in the United States, population densities vary across space far more than the prices of labor and housing”. Such a discrepancy implies that part of the attractiveness of cities is explained by other factors.

One can easily argue that residents can find in big cities a large variety of goods and services that cannot be found in the countryside or in small towns. Examples include the presence of many coffee shops, restaurants and bars; good cultural entertainments such as museums, theaters, art galleries; and many places for shopping. Such place-specific attributes are usually referred to as consumption amenities³. A positive correlation between consumption amenities and city size is also supported in the data. For American cities, Figure 4 depicts the simple relationships between city population and the number of restaurants and drinking places, museums and theaters, and clothing stores⁴. These factors contribute to explain why cities can be appealing to residents. Exploring the influence of such urban amenities on location choices is the topic of the first chapter of this thesis.

For firms, agglomeration benefits are less straightforward. First, firms also face the additional costs induced by congestion: workers may be late every morning and the delivery of intermediate goods may be delayed if commuters and trucks are stuck in traffic jams. Second, high wages and high rents imply that a firm’s total cost is higher in big cities. Such costs make dense cities less appealing for firms. Third, it is hard to believe that profit maximizing firms deeply care about museums, clothing stores or drinking places. Consequently, why do firms still want to locate in dense areas?

The existence of agglomeration economies is the major explanation for this fact.

Already in 19^th century, A. Marshall (1890) established a typology of the main sources of agglomeration economies arising from geographic clustering. Positive externalities arise

3Amenities are defined as “all characteristics of a city which could influence the desirability of a city beyond local wages and prices” (Diamond,2013)

4The three variables are expressed in absolute terms rather than per capita as varieties of these amenities are substitute to each other.

(17)

(a) Museums (b) Restaurants

(c) Clothing stores

Figure 4: Amenities and population

Data on amenities: County Business Patterns Data on population: U.S. Census

from strong linkages between intermediate and final producers, knowledge spillovers, and the pooling of workers and firms in the labor market⁵. Duranton and Puga (2004) propose a different but related classification of the sources of agglomeration economies based on three micro-founded mechanisms: sharing, learning and matching. Accordingly, agglomeration economies arise from the sharing of a large variety of intermediate inputs and indivisible investments (in the spirit of Hirschmann’s linkages), a greater ability to create and spread knowledge and ideas (in line with Jacobs (1969) or Glaeser (1999)) and a better matching between economic agents. Chapter 2 explores the relevance of this last source of agglomeration economies by evaluating how urban density affects the matching between employers and employees in the labor market.

Finally, when a worker (or a firm) decides to locate in New York or any other city, she has to reach a decision on her exact location. Will she live in Manhattan, South Bronx or Brooklyn? Indeed, metropolitan areas do not constitute perfectly homogeneous areas.

They are usually made of multiple jurisdictions, including one or several core centers where

5Empirical evidence of the sources of agglomeration economies are provided in Rosenthal and Strange (2004).

(18)

people work, and many suburban municipalities where they live. The features of these municipalities can strongly vary: local taxes might be different as well as the availability and affordability of housing goods, the presence of natural and consumption amenities or the distance to job centers. These variables are potential determinants of a worker’s willingness to live in a specific location⁶. Therefore, the internal structure of a city can strongly influence the location patterns. Understanding how people allocate themselves among jurisdictions within a given metropolitan area is the topic of the last chapter of this thesis.

To summarize, this collection of essays aims at explaining the patterns of spatial concentration and the uneven distribution of agents across space. To this end, it examines the determinants and effects of location choices faced by workers and firms. It answers the three following questions, by considering three dimensions of location choices:

What are thedeterminantsof location choiceacrosscities? The case of cultural amenities The first essay questions the role of a specific type of amenities in explaining the location of firms and residents across American cities. It focuses on the impact of cultural amenities, broadly defined as the presence of cultural and artistic equipment, infrastructures and workers at the local scale.

Over the last decade, cultural planning has been increasingly developed to foster economic and employment growth at the urban scale and to stimulate urban revitalization. A major justification for these policies relies on the externalities associated with the arts and culture: through its effects on creativity, a better cultural environment might improve both business and living conditions and thus attracts people and firms (Center for Urban Future, 2002; Department for Culture, Me- dia and Sport, 2006; European Commission, 2010; OECD, 2006). Besides, famous examples such as Glasgow or Bilbao proved that cultural vitality can favor urban recovery and improve the image of a city.

I then ask if a better cultural milieu can make a city more livable for inhabitants or improve its business environment for firms. I investigate how intercity differences in cultural environment affect the location choices faced by firms and households.

Following an identification procedure derived from a theoretical model proposed by Roback (1982), I estimate a wage and a rent equation on a panel of 346 U.S.

metropolitan areas between 2005 and 2011. Such strategy allows distinguishing the contribution of culture on production and consumption amenities.

The essay provides evidence that correlations that exista priori between cultural specialization and factor prices are positive. However, these relationships may easily vanish when other city-specific characteristics are accurately controlled for, and when problems of endogeneity are properly taken into account. Therefore, the arts and culture seem to play a moderate role in affecting the location of economic agents.

What are the consequences of location choice across cities? The effect on skill match The second essay examines a potential effect of urban density. Namely, it focuses on a positive externality that arises when workers decide to locate in dense and

6See for instance the impact of location taxation on residents in Charlot et al. (2013), and on firms in Duranton et al. (2011).

(19)

populated areas. It evaluates how urban agglomeration favors the matching between occupations and skills in the labor market.

Empirically, the extent of skill mismatch is important. Using several indicators for evaluating the quality of skill match, existing studies show that around 60% of total working population is occupied in a job for which another field of degree would be more appropriate (Abel and Deitz, 2012; Allen and Velden, 2001). Besides, the urban economic literature states that the positive effect of density on the matching between economic agents (including the match between employers and employees) is one of the main sources of agglomeration economies (Duranton and Puga, 2004;

A. Marshall, 1890). The latter being a major explanation for the existence of cities, it is quantitatively important to interact these two concepts.

This chapter evaluates the effect of urban density on the quality of the match between workers’ field of education and their occupation. Firstly, I propose a new measure to evaluate the quality of this skill match, based on observed patterns pertaining to the French labor market. I then apply this measure to a sample made of students who entered the French labor market for the first time in 2004. This work explores how employment density of a local labor market affects the matching between young workers’ field of study and the first position they found in this market.

Empirical findings show that urban density favors the quality of skill match for new entrants. Econometric results take into account problems of spatial sorting and reverse causality between density and matching. Finally, additional wage regressions are consistent with both the existence of a wage premium in big cities and a positive effect of matching on wages.

What are the determinants of location choice within cities? The distance to the center The last essay explores how geographical distance to the center of a city affects the extent of housing vacancies in suburban jurisdictions. In many large cities, problems of access to housing are considered as critical. However, a closer look at the data shows that the number of housing vacancies is very high in most urbanized areas, and little is known about these empty dwellings. The existing literature on this topic is very scarce and mostly based on American or English data aggregated at the metropolitan level. Nevertheless, it is valuable to explore housing vacancies and occupancy for more disaggregated geographical units of analysis.

Using detailed time-series data on housing for French municipalities, I document the evolution, characteristics and spatial distribution of vacant residential units in France. A particular attention is paid on the Paris metropolitan area, where housing vacancies are surprisingly large.

Then, I extend a standard model of job search to analyzing the housing market.

I develop this theoretical model to describe how distance to the central business dis- trict of a city affects the distribution of vacant dwellings in that metropolitan area.

This model provides interesting predictions regarding the effect of distance and commuting costs on the occupancy rate that prevails in a jurisdiction. I empirically test these predictions using geographic information on housing and commuting for the Paris metropolitan area. The econometric analysis shows that distance to the core city is associated with more vacancies, and that this impact has to be distinguished from the effect of commuting costs.

(20)

References

Abel, Jaison R. and Richard Deitz (2012). “Agglomeration and job matching among college graduates”.

Albouy, David and Bryan Stuart (2013). “Urban Population and Amenities”.

Allen, Jim and Rolf van der Velden (2001). “Educational mismatches versus skill mismatches: effects on wages, job satisfaction, and on-the-job search”. In: Oxford Eco- nomic Papers 53.3, pp. 434–452.

Bairoch, Paul (1985). De J´ericho `a Mexico. Villes et Economie dans l’Histoire. Ed. by Paris. Gallimard.

Center for Urban Future (2002). “The Creative Engine: How Arts and Culture is Fueling Economic Growth in New York City Neighborhoods”. In: A New York City Policy Research Report.

Charlot, Sylvie, Sonia Paty, and Michel Visalli (2013). “Assessing the impact of local taxation on property prices: a spatial matching contribution”. In: Applied Economics 45.9, pp. 1151–1166.

Combes, Pierre-Philippe, Gilles Duranton, Laurent Gobillon, and S´ebastien Roux (2010).

“Estimating agglomeration economies with history, geology, and worker effects”. In:

Agglomeration Economics. Ed. by Edward L. Glaeser. University of Chicago Press, pp. 15–66.

Department for Culture, Media and Sport (2006). “Creative Industries Economic Esti- mates Statistical Bulletin”. In: UK.

Diamond, Rebecca (2013). “The determinants and welfare implications of us workers’ di- verging location choices by skill: 1980-2000”. In:Job Market Paper, Harvard University December.

Duranton, Gilles, Laurent Gobillon, and Henry G. Overman (2011). “Assessing the Effects of Local Taxation using Microgeographic Data”. In: The Economic Journal 121.555, pp. 1017–1046.

Duranton, Gilles and Diego Puga (2004). “Micro-Foundations of Urban Agglomeration Economies”. In:Handbook of Regional and Urban Economics. Ed. by J. V. Henderson and J. F. Thisse. Vol. 4. Elsevier. Chap. 48.

European Commission (2010). “Green Paper: Unlocking the Potential of Cultural and Creative Industries”. In: Brussels.

Fujita, Masahisa and Jacques-Francois Thisse (2002). “Agglomeration and market inter- action”. In: 2002011. CORE Discussion Papers.

Glaeser, Edward L. (1999). “Learning in Cities”. In: Journal of Urban Economics 46.2, pp. 254–277.

— (2008). Cities, Agglomeration, and Spatial Equilibrium. Oxford University Press.

Hanson, G.H. (1999). “Market Potential, Increasing Returns, and Geographic Concentra- tion”. In: 439.

Jacobs, J. (1969). The Economy of Cities. Random House, New York.

Marshall, A. (1890). Principle of Economics. Macmillan, London.

OECD (2006). “Competitive Cities in the Global Economy”. In: Territorial Reviews.

OECD Publishing.

Roback, Jennifer (1982). “Wages, Rents, and the Quality of Life”. In: Journal of Political Economy 90.6, pp. 1257–1278.

Rosenthal, Stuart S. and William C. Strange (2004). “Evidence on the Nature and Sources of Agglomeration Economies”. In:Handbook of Regional and Urban Economics. Ed. by

(21)

J. V. Henderson and J. F. Thisse. Vol. 4. Handbook of Regional and Urban Economics.

Elsevier. Chap. 49, pp. 2119–2171.

(22)

Does culture affect local

productivity and urban amenities?

Abstract: Does a better cultural milieu make a city more livable for residents and improve its business environment for firms? I compute a measure of cultural specialization for 346 U.S. metropolitan areas and ask if differences in cultural environment capitalize into housing price and wage differentials. Simple correlations replicate standard results from the literature: cities that are more specialized in cultural occupations enjoy higher factor prices. Estimations using time-series data, controlling for city characteristics and correcting for endogeneity weaken the magnitude of this effect. Even though the arts and culture might be appealing to some people and firms, such determinants are not strong enough to affect factor prices at the city level.

Editorial note: A shortened version of this chapter has been published on the May 2014 issue of Regional Science and Urban Economics.

Keywords: Location choice, local amenities, culture, hedonic regressions.

JEL classification: R3, R23, Z10, O18

1.1 Introduction

This paper asks if a better cultural milieu can improve the attractiveness of a city. Cities like Paris, London or New York tend to be more attractive partially because of their remarkable cultural milieus. But are these differences strong enough to be considered as relevant determinants of the location of firms and residents? Are people and firms really willing to pay more to live in the so-called creative cities? To answer these questions, this paper evaluates how culture shapes the relative demand for a city by estimating how differences in cultural specialization across cities capitalize into housing price and wage differentials. Such strategy allows distinguishing the relative contribution of culture on productive amenities – the ability of culture to stimulate productivity throughout sectors – from its effect on consumption amenities – its propensity to offer valuable attributes and services to consumers and to improve quality of life. Using a large sample of U.S.

(23)

metropolitan areas between 2005 and 2011, the empirical analysis shows that cultural determinants are not strong enough to affect factor prices at the city level.

Addressing these questions is important for policy and academic reasons. Cultural policies are increasingly considered as drivers of economic growth and urban recovery. A large share of these policies is justified by the effect of culture on the attractiveness of cities: through its effects on creativity and livability, a better cultural environment may improve the attractiveness of places for firms and households. For instance, the Orga- nization for Economic Co-operation and Development considers that developing cultural equipment and infrastructures favors urban regeneration by helping “to attract creative and innovative populations” and highlights the ability of culture to foster a “city’s livability and attractiveness” (OECD, 2006). Similarly, the European Commission (2010) argues that cultural and creative sectors generate economic growth through enhanced creativity and innovation. In the United States, the Center for Urban Future (2002) states that creative economy’s “greatest strength is the ability to attract other businesses and jump-start neighborhood development. Arts and culture do this by giving local economies their soul. And this is everything, given that “knowledge workers” demand vibrant and dynamic settings in which they can work, live and create.”

Empirical evidence on the contribution of the arts on the attractiveness of places is surprisingly more mitigated. According to Eurobarometer (2006), 77% of European citizens state that they care about culture while data on cultural participation show that very few Europeans really enjoy cultural amenities. For instance, 55% of Europeans never attended a live performance or visited a cultural site over the year (Eurostat, 2011).

Similarly, evidence on firms’ location remains ambiguous: while Kotkin (2000) finds that cultural amenities can have a crucial role in attracting high-tech firms, Bille and Schulze (2006) stress that studies based on interviews usually emphasize that cultural variables are not important determinants in their location decisions.

In spite of a growing interest from policy makers and urban planners, research from scholars in this field remains sparse. An emerging literature offers valuable insights on this topic and some developments received extraordinary attention from policy makers and the media as illustrated by the famous contributions from Richard Florida (2002a,b).

Nevertheless, some of these studies face important limitations. First, they often fail to provide well-defined and exhaustive measures of the cultural intensity of places. Second, they rarely take account of the impact of other city-specific characteristics that might also influence the location choice of economic agents. Third, they do not consider potential problems of reverse causality between cultural variables and economic outcomes. This paper seeks to address these three limitations.

In order to achieve this aim, I propose a measure of cultural employment based on the type of tasks performed by employees to evaluate how cities differ in their specialization in cultural occupations. In that respect, I rely on occupational employment data from the U.S. Bureau of Labor Statistics (BLS) and compute the relative size of cultural employment for 346 U.S. metropolitan areas covering 82% of total U.S. population between 2005 and 2011. To my knowledge, this is the largest sample used to study this topic so far. I focus on occupations that are intrinsically oriented towards the production of non-tradable cultural goods and services since only these may potentially affect the utility of economic agents at the local scale. Similarly, I propose an alternative output-based measure of culture, which describes the accessibility of cultural goods and services for local inhabitants in each city.

Next, I rely on an identification procedure derived from Roback (1982) and estimate

(24)

hedonic wage and rent equations to evaluate the impact of culture on urban production and consumption amenities. I include in the specifications a full set of factors that could potentially influence the location of economic agents by controlling for natural, political and other location-specific variables. I also correct for the endogenous determination of cultural supply in the city by implementing an instrumental variable strategy. I use the amount of federal grants received by individual artists and art organizations every year in each city as an instrument for culture. Federal grants may be quite substantial and therefore constitute a significant shock to the supply of culture at the local scale. Besides, they are not otherwise determined by the current values of wages and rents and instead, mirror the average quality of art work in each city.

The simplest specifications recover findings from the extant literature (Clark and J. R.

Kahn, 1988; Florida and Mellander, 2010; Sheppard et al., 2006). In the cross-section of American cities, my estimates report a positive effect of culture on a city’s attractiveness:

everything else equal, specialization in cultural occupations is positively correlated with median housing rents, consistent with the existence of a higher demand for these locations. When trying to disentangle the relative contribution of culture on households and firms, I find that culture might be considered as a production amenity. However, further empirical investigations using time-series data and controlling for city-specific features reduce the estimated coefficient of the cultural environment on factor prices. Lastly, estimations that correct for endogeneity reveal that the effect of culture becomes negligible.

I conclude that the positive effect associated with culture captures the impact of omitted variables and results from the simultaneous determination of culture, wages and rents.

This interpretation is supported by various additional checks. In line with previous critics, this finding helps to question the existing empirical literature on this topic.

The rest of this paper is organized as follows: Section 1.2 selectively reviews recent empirical developments related to this topic. Section 1.3 describes the identification strategy that is applied to assess the effect of the culture on amenities. Section 1.4 describes the data. Section 1.5 presents the main econometric results and address several identification problems that arise when estimating this type of hedonic equations. Finally, Section 1.6 concludes.

1.2 Related literature

Initially considered as minor and valueless economic sectors, culture-related industries received a growing attention from policy makers and scholars. Non-negligible efforts have been made recently to evaluate the economic contribution of the cultural sector on the economy of cities. An important series of developments emphasizes the short-run effects of cultural attributes by looking at the contribution of specific cultural events, industries or infrastructures. Based on study cases and economic impact evaluations, these studies reveal the existence of multiplier effects associated with cultural spending and most of these papers focus on the direct and indirect effects of culture on tourism¹.

In the field of urban economics, several attempts have been made to evaluate the contribution of cultural attributes on the attractiveness of cities. The seminal work of

1See Bille and Schulze (2006) for a critical review of these empirical studies. Such papers include the study of single cultural events such as the economic contribution of the Guggenheim Museum in Bilbao (Plaza, 2006), the creation of a new cultural center in Catalonia (Llop and Arauzo-Carod, 2011) or the global influence of the arts in a specific area (Myerscough (1988) for Glasgow or Scanlon and Longley (1984) for New York).

(25)

Clark and J. R. Kahn (1988) evaluates the returns of six cultural infrastructures² using traditional hedonic approaches and find a positive willingness to pay for five of these amenities. Glaeser et al. (2001) take into account a series of cultural indicators – live performance venues, art museums, movie theaters and bowling alley – to assess the role of such consumption amenities on population growth across cities. Similarly, Glaeser and Saiz (2003) introduce the number of museums as a control variable to explain population growth³. Next, Albouy (2008) uses a raking based on the Places Rated Almanac (Sav- ageau 1999) to take into consideration artistic and cultural amenities in evaluating quality of life indices. He then finds a positive valuation of cultural and recreational amenities by consumers. Carlino and Saiz (2008) use the number of leisure trips as revealed preferences for local recreational amenities and find that beautiful cities enjoy relatively higher population growth. Sheppard et al. (2006) also emphasize the positive influence of cultural vitality – measured as the expenditure of local non-for-profit cultural organizations – on housing values for 11 cities in the U.S. Finally, Ahfeldt and Mastro (2011) find a positive willingness to pay for living close to iconic architecture.

The most striking development arises with the contribution of Richard Florida (2002a,b).

Florida (2002b) examines the geographic distribution of a new socioeconomic class of workers, the creative class, characterized by their specific lifestyle and their strong in- volvement in creative activities. The author intends to show that the presence of this class of people is associated with higher levels of openness and human capital. Similarly, a particular attention is paid on a smaller category of creative people, called bohemians, mainly made of artists and cultural workers. Florida eventually finds a positive correlation between its bohemian index – measured as the relative number of artists in the metropolitan area compared to the national equivalent – and the relative share of edu- cated people as well as the concentration of high-tech industries. He concludes that the presence of this class of people tends to attract more intelligent, talented and tolerant people⁴ which in turns favor the expansion of skill-intensive and innovative activities such as high-technology industries. Furthermore, Florida and Mellander (2010) extend this analysis and show that the prevalence of bohemians and gay people is associated with higher housing values and wages on a cross-section of 331 American cities in 2000.

This theory received extraordinary attention from urban policy makers and media. It strongly contributed to popularize the idea that culture may be an important driver of growth in modern cities. However, this analysis has been sharply criticized by academic researchers. The definition of the creative class as well as the bohemian community and measures of talent or urban openness are subject to strong controversies. Florida (2002b)’s creative class indeed encompasses a very large variety of working occupations which are strongly unrelated to creativity⁵. As a result, some of Florida’s empirical findings capture the impact of human capital and educational attainment rather than the effect of culture or creativity (Glaeser,2005; Markusen, 2006). Similarly, Hoyman and Faricy (2009) show that Florida’s theory is not very helpful in explaining urban growth. By contrast, standard human and social capital models better predict employment and income growth in the U.S. In addition, Montgomery (2005) criticizes the ranking of creative cities proposed by Florida based on his index. As a example, it makes Manchester be the most creative city

2The number of museums, zoos, dance companies, theaters, opera and instrumental music groups.

3The coefficient on this variable appears to be insignificant.

4Thebohemian index being significantly correlated with bothmelting-potandgay indices.

5For instance, occupations such as managers, tax collectors or dental hygienist are included in his definition of the creative class (Markusen,2006).

(26)

in the U.K., before London. The causal relationship described by Florida is also deeply criticized because causality is only inferred from simple correlations (Marcuse,2003; Peck, 2005; Sawicki, 2003). Similarly, we may suspect positive relationships between culture and both housing values and incomes found in Florida and Mellander (2010) to capture the impact of other omitted city-characteristics because econometric specifications do not include any control variables.

Even if such attempts prove that growing efforts are made to evaluate the effect of culture on households and firms’ location, these contributions remain sparse and face important limitations. First, many empirical studies fail in providing a well-defined measure of culture. Actually, studies often account for a restricted number of cultural attributes and then tend to underestimate the size of cultural sector. Inversely, measures based on employment data usually encompass all high-skilled workers or employment in cultural industries even if the tasks performed by workers are not related to culture (Scott, 2000).

Likewise, some measures tend to overlap cultural indicators with recreational variables such as parks or sport infrastructures. One contribution of this paper is to propose a more accurate measure of culture that only includes workers occupied in cultural and artistic activities based on very detailed employment data. Second, many empirical studies do not control for the influence of city specific characteristics that are not otherwise related to culture. This failure will be addressed in Section 1.5. Third, concerns regarding the causal relationship between culture and economic outcomes are rarely mentioned.

I correct for potential endogeneity biases using instrumental variable techniques in the empirical section.

1.3 Identification strategy

To determine how differences in the cultural landscape across cities affect the location of households and firms, I rely on an identification strategy proposed by Roback (1982).

This model extends the standard hedonic approach from S. Rosen (1979) to heterogeneous labor markets⁶.

We consider a finite number of non-overlapping cities in which people chose to live and work. People and firms move freely across space in order to reach the highest level of utility and economic profit. Preferences of the representative household are defined over the consumption of a composite and a housing good. The iso-utility curve of this household is represented by the upward-sloping curveU₁in Figure 1.1. The indirect utility function being strictly increasing in wages and non-increasing in prices, any increase in housing prices must be compensated by a rise in wages in order to keep the level of utility constant at ¯u.

Similarly, firms determine their optimal consumption of labor and land in order to maximize profits. Firms behavior is described by the downward-slopping iso-profit curve Π₁ in Figure 1.1: any increase in wages must be compensated by a fall in land prices in order for total costs and economic profits to remain constant. Wages and rents prevailing in city care determined at equilibrium point a.

I allow both indirect utility and profit functions to be affected by a city specific at- tribute S_c that varies across cities. S_c is defined as a consumption amenity as long as it positively affects consumer’s utility; and as a productive amenity when it enhances firms

6The model is now widely used to estimate the influence of localized amenities. See Ottaviano and Peri (2006) for a full analytical description of the model.

(27)

productivity and therefore economic profits. Sc will represent the state of the cultural environment in city chereafter.

Figure 1.1: Local amenities in a spatial equilibrium

At a spatial equilibrium, workers and firms must be indifferent among locations. Per- fect mobility allows residents and firms to relocate if they can reach a higher level of utility and profit in a different location. Hence at the equilibrium, both utility and economic profits are equalized across cities in order to eliminate further inducements to move. In the presence of a consumption amenity, any increase in S_c must be compensated either by a decrease in wages or by a rise in rents to eliminate residents’ incentives to move to city c. In the case of a production amenity, firms’ cost function being increasing in both factor prices, firms’ incentives to move are arbitraged away either by a rise in wages and in rents.

The impact of an increase in the level of city-amenity is summarized in Figure 1.1.

With a consumption amenity, an increase in S_c shifts the iso-utility curve up to U₂: at equilibrium point b, we observe that residents pay higher rents and receive lower wages but attain the same level of utility thanks to this amenity. With a production amenity, the iso-profit curve shifts from Π₁ to Π₂: at equilibrium pointc, firms endure higher wages and higher rents but reach the same level of economic profits thanks to a higher level of S_c.

We can easily derive from this analysis the hedonic rent and wage equations that describe the relationship between local amenities and factor prices. First, the analysis of the rent equation allows determining the overall effect of culture: the fact that cultural cities experience higher (lower) rents suggests the existence of a higher (lower) demand for these locations, induced either by firms or workers. Next, the wage equation helps us to determine if this positive (negative) effect is dominated by a positive (negative) impact on firms or households. A negative impact on wages mirrors the fact that workers are willing to give up wages to live in high-amenity cities. In contrast, higher wages imply that firms

(28)

are likely to incur higher costs to be located in such high-amenity cities, reflecting the fact that S_c is a productive amenity.

Therefore, the Roback (1982) model provides a very suitable identification procedure for assessing how households and firms are affected by inter-city differences in cultural environment. Table 1.1 summarizes the main results of this model in the case of positive or negative amenities in both consumption or production.

Table 1.1: Analysis of the wage and rent equations Wages ^dw_dS^c

c

>0 <0

Rents _dS^dr^c

c

>0 Production amenity Consumption amenity

<0 Consumption disamenity Production disamenity

1.4 Data description

1.4.1 Measuring culture

I use Occupational Employment Statistics (OES) from the U.S. Bureau of labor statistics (BLS) describing employment for 372 U.S. Metropolitan Statistical Areas between 2005 and 2011. The occupational classification system (SOC) reports employment statistics based on the type of activities and tasks performed by workers independently on the in- dustry in which they take place⁷. Employment data are then disaggregated into 22 major occupations which are in turn broken down into 840 detailed occupations. Markusen (2006) or Glaeser et al. (2001) show that existing measures of cultural employment based on the broadest occupational categories face major limitations. For instance, including the major category “Arts, Design, Entertainment, Sports and Media occupations” tends to overestimate the size of cultural employment. Indeed, this category encompasses occupations such as “Public relations specialists” or “Athletes and sports workers” which can hardly be considered as cultural activities. Conversely, other artistic and culture-related occupations can be found in other major categories such as “Archivists”, “Curators” or

“Museum technicians and conservators”. Therefore, I delineate the sample of cultural occupations by inspecting each detailed occupation title rather than broader occupational categories.

Well-established definitions found in Throsby (2001, 2010), World Intellectual Prop- erty Organisation (2003), KEA European Affairs (2006) and Unesco Institute of Statistics (2009) or UNCTAD (2008, 2010) traditionally define cultural occupations as including:

7See Lin (2011) for additional details on the Standard Occupational Classification System.

(29)

- Core art and heritage occupations, including visual arts (painting, sculpture), performing arts (theater, dance) and heritage (museums, monuments, libraries).

- Cultural occupations, including activities related to films, TV, radio, broadcasting, music recording, press and book publishing.

- Creative occupations, encompassing fashion, design and architecture.

This classification includes all workers occupied in culture related activities. However, it does not reflect the production for or the consumption by local inhabitants. Indeed, several cultural products are easily tradable. Therefore, local production does not necessarily coincide with local consumption. Cultural and creative occupations basically encompass workers occupied in cultural industries which create goods and services that are traded and not locally consumed. Remarkable examples are radio or TV programs: the production of such services in a city is not directly consumed by local residents. Therefore, they are unlikely to enhance the attractiveness of the city in which they are produced. This is why I take benefit of the classification presented above to restrict cultural employment to a few number of occupations that are traditionally consumed where they are produced and which require proximity to final consumers. This includes for instance art teachers, curators, museum technicians, conservators or librarians. A description of these occupations is given in Table 1.A.1 in the Appendix. Using this definition, the share of cultural employment ranges between 0 and 0.8% with a median value of 0.2%. In section 1.5.3, I also use an alternative measure of culture based on the accessibility of cultural goods and services for residents as a robustness check.

The variable is disaggregated at the Metropolitan Statistical Area (MSA) level using the 2003 definition⁸. These geographical units correspond to local labor markets with strong commuting ties between each component. Cultural employment is annually computed for 372 MSAs. I restrict the sample to cities located in contiguous continental US states and exclude New England City and Town Areas (NECTAs) for which a corre- sponding MSA cannot be found. This leaves us with 346 MSAs, covering 82% of total US population.

1.4.2 Dependant variables: wages and housing rents

BLS’s Employment Statistics provide data on wages and income distribution at the MSA level. I use the median wage defined as the median hourly wage for all occupations in each MSA⁹. The subsequent sections intend to study whether variations in the cultural supply are associated with differences in wages. However, if wages in cultural occupations are higher than average, a bigger share of cultural workers in total employment will automatically translate into higher wages in the city. To avoid this problem, I compute the median wage of non-cultural workers only.

Data on rental prices comes from the “50th percentile series” of the U.S. Department of Housing and Urban Development (HUD). HUD annually reports median gross rent estimates for all U.S. metropolitan areas. The latter correspond to gross rent estimates,

8SeeStandards for Defining Metropolitan and Micropolitan Statistical Areas, U.S. Office of Manage- ment and Budget.

9In order to closely estimate the Roback (1982) model, we would like to have information on wages and rental prices for the marginal worker. Because of data unavailability, I use median values instead.

(30)

including utilities (except telephone and other communication medias), at the 50th percentile point of the rent distribution of rental housing units¹⁰. This indicator already con- trols for major differences in housing quality by including only rents of “standard-quality rental housing units (occupied rental units paying cash rent, with full plumbing, with full kitchen, unit more than 2 years old) occupied by recent movers [and excludes] public housing units, newly built units and substandard units”¹¹. To control for an additional major difference in housing characteristics, I use the median gross rent of a 2-bedroom rental unit (Saiz,2007). One minor limitation of these data is that the HUD geographic criterion to compute these estimates may not perfectly match with the definition of metropolitan areas applied by the BLS such that some MSAs might be discomposed into several com- ponents. In such a case, I weight each county component by its population within the MSA to compute the median gross rent of the entire area.

By using median rents and wages, I implicitly assume that changes in the size of the cultural sector uniformly affect all workers and residents. This choice is dictated by two main factors. The first is the availability of data on urban rents, that are exclusively provided at the 50th percentile only. Second, the identification strategy is directly derived from a model of representative consumers and firms.

1.4.3 Control variables

To control for major determinants of rents and wages that are not otherwise related to culture, I include a full set of control variables in the empirical analysis. Differences in the level of human capital across cities may affect both wages and rents through higher productivity of workers or heterogeneous preferences for housing goods. I control for these differences by including the average level of education of each city. I extract data on educational attainment from the annual American Community Survey estimates (ACS) and defineEducation as the share of population of 25 years or more with a bachelor degree (or more).

Using ACS data, I compute the share of foreign born residents in total population to control for the impact of migrants on labor and housing markets. I also control for the racial composition of MSAs by including the share of ‘non-white’ workers in total population, using the Population by Race and Hispanic Origin Table derived from U.S.

Census and the Selected Social Characteristics from the ACS estimates. These variables control for the racial composition of cities and the effect of cultural diversity on prices (Glaeser et al., 1995; Markusen,2006; Ottaviano and Peri, 2006; Saiz, 2007).

Ciccone and Hall (1996) show how population density matters for explaining differences in labor productivity while Duranton and Puga (2004) emphasize the gains from agglomeration experienced in large cities. Rappaport (2008) underlines the relationship between density and urban consumption amenities. I use a measure of population density, defined as the total population per square mile of land area, to control for inter-city differences in urbanization and the role of agglomeration economies.

To control for differences in public goods supplied by local governments, I include spending on public schools approximated by the total per pupil current spending in elementary and secondary schools. Primary data are extracted from the Public Elementary -

10These data are calculated on the same basis as Fair Market Rents but focuses on the 50th percentile point. See Saiz (2006).

11Federal Register Vol. 65, No. 191, Monday October 2, 2000, pages 58870-58875.

(31)

Secondary Education Finance Data from the U.S Census Bureau for the year 2009. Data reported at the school districts level are aggregated to compute an average value at the MSA level.

To account for differences in natural amenities, I follow Saiz (2007) by adding the monthly average temperatures in January and July, the average number of hours of sun- light in January and the average relative humidity in July as control variables. These average data are computed over 30 years (1941-1970) and extracted from the Natural Amenities Scale Dataset from the U.S. Department of Agriculture.

To capture differences in safety conditions across cities and their impact on quality of life, I use data from the Federal Bureau of Investigation (FBI) and compute an indicator ofviolence rate defined as the number of murders, forcible rapes, robberies and aggravated assault per 100’000 inhabitants.

I also include two economic variables: the annual rate of unemployment it the metropolitan area (provided by the Smoothed Seasonally Adjusted Metropolitan Area Estimates series from the BLS) as well as the annual share of employment in the service sector extracted from the U.S. County Business Patterns to account for the shift towards services in the U.S. economy. Including a variable that measures changes in the composition of local production is in line Glaeser and Saiz (2003) or Carlino and Saiz (2008).

Finally, I control for the role of non-tradable consumption goods and their variation over time by including two additional consumption amenities as explanatory variables: the variety of restaurants and drinking places – measured by the number of food and beverage establishments per capita – and the number of amusement, gambling and recreational establishments¹²per capita extracted from the County Business Patterns. These variables mirror the analyses performed by Glaeser et al. (2001), Glaeser and Saiz (2003) or Albouy (2008). As described above, these two variables being excluded from existing definitions of cultural activities, they enter the regressions separately as control variables.

A statistical summary of the dataset is provided in Table 1.A.2 in the Appendix.

1.4.4 Stylized facts on culture, wages and rents

Table 1.2 provides information on the average and median hourly wage as well as median gross rents in U.S. cities. The first column displays general statistics for the whole sample whereas columns (2) to (5) report statistics for cities ranked according to their degree of specialization in cultural occupations. Both wages and rents tend to increase as the level of cultural specialization rises: in average, cities that have a relatively high share of cultural employment face higher wages and higher rents.

Figures 1.2 depicts the relationship between the variable of interest – the cultural employment share – and wages and rents for all U.S. metropolitan areas over the year 2005 and 2011. The graph also reports the fitted line and the estimates of the regression line. Wages and rents are positively correlated with this measure of cultural supply.

The estimate associated with the rent equation shows that a 10% increase in cultural employment share is associated with a 1.32% increase in local rents and a 1.14% increase in wages. The results are similar when using an alternative measure of culture based on access to cultural establishments (described in Section 1.5.3) even if the maginitude of the coefficient is smaller. At first sight, a higher specialiation or accessibility to culture seems to positively capitalize into higher factor prices. Such relationship is in line with Florida

12This variable includes amusement parks, golf courses, fitness and recreational sports centers or bowling alleys.

(32)

Table 1.2: Wages and rents according to the relative specialization of cities in 2010 Cultural employment

All MSAs Lowest Low High Highest

Average wage 19.24 18.34 18.66 19.20 20.83

Median wage 15.10 14.36 14.65 15.06 16.36

Median rent 833.37 785.76 787.37 843.19 914.58

Cultural emp. 888 94 450 908 2093

Observations 346 86 87 86 87

and Mellander (2010) who interpret it as a positive impact of culture on production amenities.

(a) Median wage (b) Median rent

Figure 1.2: Wages, rents and culture (average over 2005-2011)

1.5 Empirical results

The aim of this paper is to determine if the willingness of residents and firms to locate in cities is influenced by the cultural environment of these places. In line with the identification procedure described in Section 1.3, I perform two sets of econometric estimations using alternatively rents and wages as the dependent variable.

1.5.1 Baseline regressions

To determine the effect of culture on both production and consumption amenities, I estimate the following reduced form equation on the panel of U.S. cities over the period