Skip to navigation – Site map

Real estate market and urban transformations: spatio-temporal analysis of house price increase in the centre of Marseille (1996-2010)

Guilhem Boulay


Over more than ten years, France has experienced a twofold increase of residential housing prices. This was largely fuelled by the credit conditions and a general fear of the future. A strong negative correlation between initial price level and its increase leads to a massive trend of spatial homogenization of the prices. Such a tendency is fairly vigorous in the metropolitan area of Marseille, and particularly in its extremely poor and long devaluated centre, where prices have risen by over 200% since the mid-1990s. This real estate market evolution is emblematic of Neil Smith’s rent gap hypothesis. Yet, a detailed survey of spatial differentials of price increase shows that contiguous central neighbourhoods with similar price levels before the recent tremendous upsurge are affected by appreciably different price rises. Therefore, this unequal price evolution cannot be understood without a clearer analysis on the ‘potential land value’ than provided by Smith’s model. In the context of such housing price increase and uncertainty, prices transmit noisy signals to investors. Thus, State involvement, as in one of the most important operations of urban redevelopment nationwide (Euroméditerranée), appears as a guarantee for investors and also accounts for the spatial differentials of price growth rates.

Top of page

Full text


  • 1  Thanks are due to Claude Napoléone and Ghislain Géniaux who supported collaboration with the INRA- (...)

1Though countless articles have been addressing and evaluating the relationship between space, the value of land and the value of real estate, a fundamental gap still remains in this literature. Despite the recent and significant fluctuations in real estate markets, very little attention has been devoted to unpacking specific empirical cases in a medium-term perspective. Even if economists have renewed and widened their traditional approaches to the access-space-trade-off model by the use of hedonic regression techniques, economic methodology remains characterized by the neo-classical paradigm: “the dominant theoretical framework for microeconomic analysis of housing markets is the standard, contemporary Walrasian synthesis of the neo-classical framework” (MacLennan and Tu 1996: 388). This approach relies upon the concept of equilibrium (Derycke 1996) and therefore pays little attention to prices variations. However neglected these variations were, we consider they should be taken as a heuristic opportunity. As O’Sullivan and Gibb (2003: 1) argued, “it is a field of inquiry where, in Joan Robinson’s terms, thinking in ‘theory time’ is a poor substitute for recognising the role of real, non abstract, historical time”. In this context, the methods of statistical cartography can be used as part of a spatio-temporal approach of such a market phenomenon: this is not a solely positivist way to capture this phenomenon but rather an effective way to question the theoretical frameworks we usually use.

2One major conclusion stemming from spatial analysis of housing price increases is that there is a remarkable catching-up process in the evolution of prices between boroughs. By catching-up process, we mean the convergence of prices in a given spatial area (Hamnett 2009). This process strongly supports the rent gap hypothesis first proposed by Neil Smith in the 1970s and 1980s discussions concerning the gentrification of American inner cities. The American geographer Smith (1979: 545) defined this gap as “the disparity between the potential ground rent level and the actual ground rent capitalised under the present land use”. According to Smith’s hypothesis, this gap creates an opportunity for investors hoping to earn great returns. Thus, gentrification would be “a back to the city movement of capital, not people”, to quote the subtitle of his seminal paper (Smith 1979).

3Though often used (Badcock 1989, Clark 1988, 1995, Shin 2009) and sometimes criticized or critically analyzed (Bourassa 1993, Millard-Ball 2000), this model marked a significant step in the spatio-temporal analysis of property markets. This hypothesis thus constitutes an ambitious and serious attempt to link the spatial variability of house prices with medium- or long-term market changes. However, this article will not try to assess the rent gap hypothesis, for our interests are not primarily in land market or in gentrification processes. Instead, we explore the implications of this hypothesis in the context of the spatial analysis of housing price increases at a fine scale throughout an entire metropolitan area over a period of ten years, which is a relatively novel use of the model. Though the rent gap hypothesis has been conceptualized on a broad scale, in practice assessments of Smith’s hypothesis suffers from an important methodological drawback: detailed analysis of property market is usually limited to isolated areas such as inner city neighbourhoods. This neglects other phenomena that relativise not the bridging of the rent gaps in themselves but the conditions that make them possible. As a result, statistical cartography reveals some implicit components of the rent gap hypothesis and clarifies some links between the development of the prices and the changes occurring in urban spaces.

4The metropolitan area of Marseille (see Maps 1 and 2), one of the major cities in France, is a particularly relevant and interesting field to demonstrate this approach: long very devaluated, Marseille experienced an exceptional increase in real estate prices, in particular in the most dilapidated segments of the market. The central and inner city have also recently seen the development of one of the most significant urban renewal operations in France – the Euroméditerranée Operation of National Interest (Opération d’Intérêt National, OIN).

Map 1. The Marseille Metropolitan Area

Map 1. The Marseille Metropolitan Area

Source: National Institute of Statistics and Economic Studies (INSEE, census). Cartography by the author.

Map 2. The Municipality of Marseille showing the location of the pictures taken

Map 2. The Municipality of Marseille showing the location of the pictures taken

Source: Euroméditerranée. Cartography by the author.

5In this paper, the first section examines the achievements and the limitations of the spatial approach of real estate markets. The second section presents the data and the methods necessary for establishing the results outlined in the third section. Evidence demonstrates that rent gaps are not filled by processes of real estate price increases alone, but that prospects for gap reduction are also defined and limited by the existing, and historical, sociospatial order. The fourth section shows that this subversion depends highly upon the involvement of public authorities, which makes such reinvestment possible assuming the role of guarantor and building trust.

Theoretical frameworks for the analysis of spatio-temporal dynamics of housing markets

The limitations of econometric approaches to real estate markets

6In classical land rent theory and amongst neoclassical economists, land and real estate markets have historically been discussed from a purely theoretical point of view (Guigou 1982). This attitude stems first and foremost from their commitment to building the pure, analytical frameworks of the New Urban Economics, which have been described as “urban economic theories based upon searching for market equilibrium grounded in the principle of utility maximization” (Derycke 1996: 64, our translation). During the 1980s, an increasing number of economists advocated the use of econometrics to break with these excessively static models: “Remarkably there has been virtually no econometric estimation of the competitive housing models developed over the last 15 years” (Arnott 1987: 981). These accurate econometric works can be divided in two main approaches.

7The first one is the Hedonic Pricing Method (HPM), which has been used and adapted by many scholars (Sheppard 1999, Malpezzi 2003). However detailed and seemingly completed this method may be from a technical point of view, we too often forget that HPM is subject to serious limitations. Notably, HPM provides us with numerous snapshots of the housing markets does have the capacity to analyze evolution over time. HPM provides accurate information about spatial differentials at any given moment but appears to be less efficient to explain local variations of real estate prices. Moreover, HPM continues to be characterized by the prescriptive paradigm of equilibrium, imposing relatively restrictive assumptions and conditions of validity (Beckerich 2001).

8The second econometric approach, which has recently and massively entered the works of econometrics, deals with the evolution of the market over several years but is confined to very broad scales such as the national and sometimes metropolitan scales. These contributions are limited almost exclusively to testing and measuring housing bubbles (Diewertet al. 2009, Duca et al. 2010, Glaeser et al. 2008, Goodman and Thibodeau 2008, Hwang Smith and Smith 2006). As a result, they usually understand housing prices according to their capacity to reflect the ‘fundamentals’ of the market, and not according to their links with the production of urban space.

The rent gap hypothesis: a multi-scale modelling of housing markets

9The rent gap hypothesis breaks with the abovementioned approaches, which keep the spatial and the temporal aspects of the housing markets separate. In so doing, this hypothesis fills the vacuum created by the relative disinterest of many geographers for questions of land and real estate markets. Smith’s pattern refers to central and inner city districts where the process of devalorisation of the fixed capital creates a gap between the Actual Land Rent (ALR) and the Potential Land Rent (PLR). The transition between the ‘downward sequence’ and the ‘back to the city of capital’ is famous: “As well as creating barriers to the further valorisation of capital in the built environment, however, the steady devalorisation of capital creates longer term possibilities for a new phase of valorisation, and this is exactly what happened in the inner city” (Smith 1982: 147).

10Beyond this widely accepted pattern, one should bear in mind that rent gaps are unthinkable outside a theoretical framework based upon many scales and many temporalities. Following Harvey (1985), one can say that the back-to-the-city movement of capital is conceptualized in conjunction with the process of capital switching, which has been recorded several times (Davis 1997, Nappi-Choulet 1999, Smith and Defilippis 1999). The very fact that the filling of a rent gap corresponds to a spatial fix à la Harvey implies inclusion of both temporal and spatial fluctuations of prices: “That the intensity of the process came to parallel the wider economic cycle of accumulation and crisis since the 1980s indicates that gentrification is more, not less, integrated into the larger movement of economy” (Smith and Defilippis 1999: 639). This is why the rent gap hypothesis may be a missing link in the already depicted classical econometric approach between space and time. The rent gap is a true historical and geographical model of urban property markets, and Smith (1982: 150) strongly emphasizes the point that the kind of rent gaps we are examining nowadays are a contingent species of a more general genus: “Suburbanisation was a concrete spatial response to the depression of the 1890s and 1920s in the sense that suburban development opened up a whole series of investment possibilities which could help to revive the profit rate. […] Albeit a reversal in geographic terms, the gentrification and redevelopment of the inner city represents a linear continuation of the forces and relations that led to suburbanisation”.

Housing price dynamics in France and in Marseille since the mid-1990s

11As in most of emerging and wealthy countries, housing prices have dramatically risen from the end of the 1990s onwards. Unlike what was observed in many others countries, the housing market did not collapse after the subprime mortgage crisis occurred: in 2010, prices had regained their historical levels of 2007. According to the National Institute of Statistics and Economic Studies (INSEE), French housing prices have more than doubled over the past 15 years (index base = 100 in 1996 was 269 in 2011). In the Marseille area, prices increased even more: they tripled over the same period. The city experienced one of the most impressive housing price increases amongst major French cities, placing the city at the forefront of the housing price spike (Boulay 2011a). These details explain why Marseille is an ideal experimental case study to assess the spatial distribution of housing investment, estimated to have increased in the metropolitan area from €800 million in 1996 to €2.5 billion ten years later (Boulay 2011b), even though this figure was calculated for old property market and thus underestimates the total amount invested in property market (since old houses represent only 85% of the market, and are less expensive than the newly constructed ones).

A geographical focus on housing markets: data and methods

Disaggregated and georeferenced data on real estate sales

12The French property market is particularly opaque. Traditionally closely connected with a very strong conception of private property and the cult of fiscal secrecy, this opacity is nowadays partly due to the emergence of a market of statistical data – there are nearly 100 producers of free or commercial data (Anah 2005). Each of these databases includes a flaw: lack of metadata, relatively high levels of aggregation, vague georeferencing, limited information or variables limited to a specific market segment, etc. (Buhot 2006, Schmitt 2009).

13Against this background, the Perval data we use in this paper have several advantages. These data, produced and controlled by the notaries, are disaggregated, georeferenced in the cadastre, they cover all market segments throughout France, and they encompass about 100 variables addressing the date and the legal conditions of the transaction, the home seller and the home purchaser, the real property itself. However, the data compilation depends on personal, voluntary initiative – the notary is not required to send the data to the Perval Company – and most of the variables collected are declarative – the notary can make mistakes in filling the fields or neglect it. Perval data are consequently samples, and this implies a necessary assessment of their quality, to evaluate whether their representativeness and significance comply with the requirements of a meaningful statistical treatment. This sometimes leads to use error correction strategies (Boulay 2011b). The sample we are working on is made up of about 120,000 real estate sales, mainly old properties’ sales – about 85% of the total sample. Figures are available for 1996, 1998, 2000, 2002, 2004 and 2006, that is to say for periods marked by the sharpest house price increase. Given the fact that we are primarily concerned with the house price increase, we will rely on 1996 and 2006 transactions.

Selection of the observations

14About 2000 sales have been removed from this sample for several reasons – sales by public auction instead of private sales and statistical aberrations corresponding to errors or donations. We also applied a second selection criterion to neutralize structural effects in this 118,000 sales sample. Indeed, some properties are structurally more expensive than others, even when the property price is brought back to the m2 pre-tax price: a m2 of floor area in old properties is cheaper than one in newly built properties; a m2 in a house is more expensive than in an apartment; a m2 in an average size housing unit is cheaper than in a small or a large one, etc. Therefore, to capture only price differentials that are due to location, we need to eliminate the influence due to the heterogeneity of the sample. This is why we will only work on a well-defined market segment, which corresponds to the modal segment throughout the metropolitan area of Marseille: old three-room apartments. This helps reducing the risk of biased results due to local variations in the composition of the samples.

15For those reasons, we will place special emphasis on this particular segment of the Marseille metropolitan area property market, but we will also set out results for the totality of the market to show that this kind of sample construction does not arise as an artefact for which the results may differ from those obtained for the whole market.

Mapping the real estate market

16A cartographic approach should maximize the opportunities offered by the Perval data. Due to privacy reasons, real estate sales are georeferenced at the scale of the cadastral section, that is to say an intermediary scale between the cadastral parcel and the municipality. The sizes of these administrative spatial units vary between a few hectares and a few square kilometres, depending on the human density of the municipality. The boundaries of the sections shift with demographic growth and urban sprawl but do not follow any precise rule.

17As a result, the geographer faces a twofold problem in mapping real estate prices: shape and size effects on the one hand, the non-comparability and the incompatibility of the cadastral sections over time on the other. The first problem stems from the non-regularity of the cadastral sections’ spatial pattern, the second problem arises because the cadastral sections change over time (Lloyd 2011). The best solution is to redistribute the real estate sales into a regular grid in order to limit the impact of the Modifiable Areal Unit Problem (MAUP) on the results (Pumain and Saint Julien 1997). Choosing a grid does not constitute a sufficient and entirely satisfactory solution per se, as the problem of optimizing the size of the grid cells remain (that is to say the problem of the best level of aggregation) (Wong and Amrhein 1996). Admittedly, the quadrat analysis provided valuable guidance on the optimal size of grid cells (Greig-Smith 1964), but while useful these results are primarily used for studying the geometric attributes of a points pattern, and not its semantic attributes. Thus Greig-Smith formula is particularly suited for the study of densities, and not for the study of central values. However, it can be noted for information purposes that Greig-Smith’s formula would result in 2196 m by 2196 m square grid cells for 1996, and 1878 m by 1878 m square grid cells for 2006.

18Taking prices into account entails a trade-off between the ‘focal length’ – the size of the grid cells – and the ‘statistical significance’ – the number of sales per cell. At the same time, we also need to consider that any change in these two parameters leads to change in the total number of cells considered since we need to determine a minimal threshold of sales per cell – we decided to keep only the cells with at least five sales every year. On the other hand, even if the choice of a particular market segment has sharply reduced the number of sales on which this paper is based, this choice results in quite homogenous samples and thus reduces the risks related to the sometimes low number of sales per cell. Table 1 shows the values for each of the parameters of this trade-off.

Table 1. Values for parameters

Size of the grid cells

Number of cells with at least five old three room apartment sales every year

Average number of old three room apartment sales per cell (1996)

Average number of old three room apartment sales per cell (2006)

Total area taken into account (km²)

750 m by 750 m





1000 m by 1000 m





1500 m by 1 500 m





2000 m by 2000 m





Source: author.

19Ultimately, we have chosen 1500 m by 1500 m square grid cells, which is a quite detailed scale for a time series study. Note that this size is smaller than the one resulting from Greig-Smith’s formula. This choice results from an arbitrage between all the parameters and therefore may be called into question. To prove that this choice is not ad hoc, we will also set out results for all sizes of grid cells. Obviously, this relatively small number of grid cells corresponds in substance to the main municipalities of the Marseille metropolitan area (see Map 2).

Can all price increases be explained by the rent gap hypothesis?

A generalized increase hiding strong catch-up phenomena

20Housing price increase is impressive in all the parts of the metropolitan area of Marseille. Prices of old three-room apartments grew by 230% on average between 1996 and 2006. Amongst the 60 cells investigated, housing price increase assumes a statistical normal distribution. In some parts of the metropolitan area, prices increased tremendously, such as the industrial cities on the shores of Berre’s Pond or the northern part of inner Marseille. In some others, price increase was less pronounced, as in Aix-en-Provence (see Map 3). Since housing price increase statistical distribution is normal, most of the cells have witnessed increases in line with the metropolitan average.

Map 3. Housing price increase in the Marseille metropolitan area, 1996-2006

Map 3. Housing price increase in the Marseille metropolitan area, 1996-2006

Sources: DGFIP and Perval. Data provided by INRA-Avignon. Cartography by the author. Note: mean prices are given for an old three bedroom apartment at 1500 m by 1500 m square cells.

21Spatial differentials in housing price increases all relate to a clearly identifiable fundamental phenomenon: at a detailed scale, there is a strong negative correlation between the prices before the house price spike and the growth rates of prices. Between 1996 and 2006, prices rose in inverse proportion to their 1996 level, regardless if one observes mean prices or median prices (see Figure 1 and 2).

Figure 1. Median housing prices in 1996 and house price increase, 1996-2006

Figure 1. Median housing prices in 1996 and house price increase, 1996-2006

Source: author. Note: pre-tax prices, 1500 m by 1500 m cells, old three bedroom apartments.

Figure 2. Mean housing prices in 1996 and house price increase, 1996-2006

Figure 2. Mean housing prices in 1996 and house price increase, 1996-2006

Source: author. Note: pre-tax prices, 1500 m by 1500 m cells, old three bedroom apartments.

Table 2. Correlations between the 1996 housing prices and the housing price increase between 1996 and 2006

Mean prices

Median prices

R² calculated for all grid cells



R² calculated for all grid cells without outliers



Source: author.

22As shown in Table 2, correlations are fairly high and significant: they explain a good share of the spatial variation of housing price increase. Nevertheless, the quality of the fit suffers from the presence of very few aberrations. So, it is possible to greatly improve the quality of the correlation by eliminating three cells (see the red plots on Figure 2). Such ‘Barro regressions’ (Barro 1991) are typical of β-convergence processes, well known in growth literature: “Numerous researchers have recently examined this question by calculating cross-section regressions of measured growth rates on initial level [...]. Such regressions have, in fact, become known as ‘Barro regressions’, and are widely used to analyze empirical growth dynamics. Evidently, in a Barro regression, a negative coefficient on initial levels is taken to indicate convergence” (Quah 1993). This kind of price convergence is not due to the decision to focus on a particular segment of the market or at a particular scale: we noted such significant correlations (for α = 0.01) for the old three room apartments segment at any grid scale, and likewise for the totality of real estate transactions at any grid cells, regardless if we worked on mean prices or median prices.

23As Quah (1993: 428) demonstrated: “a negative cross-section regression coefficient on initial levels is, in fact, perfectly consistent with the absence of convergence in the sense of (d)” – i.e., a diminution of the cross-section overt time, a process called σ-convergence. It can therefore be useful to check whether the dispersion between the mean prices and median prices per cell declines over time, reporting a σ-convergence. The coefficient of variation appears to be the most suitable indicator to assess the decline of dispersion, and illustrates clearly this phenomenon. Between 1996 and 2006, the coefficient of variation was reduced by 18% if we consider mean prices of old three room apartments, or 21% if we consider median prices. Once again, this σ-convergence is not due to the decision to work on this particular segment or at the 1500 m by 1500 m cells scale: such convergences have been reported at any grid cells, either for the old three room apartments or for all the property transactions.

24It can therefore be concluded that the main spatial dimension of housing price increase is a generalized catching-up phenomenon. Such phenomenon has already been demonstrated in London (Hamnett 2009) and Paris (Guérois and Le Goix 2009). In both papers, the role of low price submarkets is highlighted: “the end of the period, from 2001, marks a new stage in the process of reduction of price differences. This reflects catching-up processes in certain sectors: between 2001 and 2003, some peripheral neighbourhoods have risen steeply in price, experiencing price increases far stronger than the average increase” (Guérois and Le Goix 2009: 5, our translation).

The housing price increase in devaluated segments as a driving force

25Marseille is a historically poor and dilapidated city, however these catching-up phenomena mean that house price increase is particularly strong in the lower housing submarkets, be it defined either by spatial, structural or nested segmentation (Watkins 2001).

26The decline in house price dispersion is a permanent feature in the increasing housing market of Marseille. Such a phenomenon can be demonstrated by various means for many parameters. For example, if we keep focusing on the three room apartments segment, one should note that a square meter of living space in the old property market in the mid-1990s was half the price of its equivalent in the new property market. Ten years later, the former costs roughly the same as the latter (90%). At the same time, the ratio between the 9th and the 1st deciles for the prices of an old three room apartment square metre of living space drops from more than 3 to less than 2.5. This decrease is even more obvious for the ratio between quartile 3 and quartile 1, for decile 9 is the only segment of the market in which prices react differently from the catching-up rule. Indeed, there are a few fairly luxurious segments that are not sensitive to solvency limits. Their prices increased tremendously, as happened for example in the suburban tourist town of Cassis, which was in 1996 the most expensive part of the metropolitan area, and remains the most luxurious submarket in 2006. It corresponds to the outliers shown in red plots in the right parts of the scatter plots (see Figures 1 and 2).

27At the urban framework scale, it can be clearly concluded that the progression of Marseille Metropolitan Area within the hierarchy of major French cities is first and foremost due to the increase of the most initially devaluated market segments. This is evident by the fact that that the cheaper the market segment considered (D9, mean, median, D1), the greater the progression of the market segment within the hierarchy of prices between the major cities – the city of Marseille including a large number of these devaluated parts, and the city of Aix-en-Provence including a large number of higher status parts (Boulay 2011a).

How do we define the filling of a rent gap in this context?

28The identification of this generalized catching-up phenomenon challenges the rent gap filling process. Once financial and macro-economic conditions for the filling of a rent gap have been expressed, the process may then be summarized as follows: “a process involving a change in the population of land-users such that the new users are of a higher socio-economic status than the previous users, together with an associated change in the built environment through a reinvestment in fixed capital” (Clark 2005: 258). Consequently, there are three key points in this mechanism: a catching-up process; ‘higher and better uses’ of the ground or building renovation operations; a gentrification induced by the house price increase.

29The first and third points pose a problem: at the metropolitan area scale, the catching-up phenomenon is generalized, and the social structure of home purchasers does not change significantly between 1996 and 2006 – the proportion of the wealthier classes does not progress whereas the part of the lower class slightly does (Boulay 2011b). In these conditions, the sole criterion for determining the specificity of a rent gap filling appears to be the second factor: higher and better uses. Changes in the neighbourhood status can be a proxy index for this criterion.

30When catching-up is generalized, it is necessary to assess whether housing price increase is a mere difference of degree or entails a difference of nature. This phenomenon is very rare as evidenced by the use of Spearman’s rank correlation coefficient (for which 1 means a perfect correlation and 0 the absence of any correlation). This index is a very robust way for judging whether or not a hierarchy is stable. At a detailed scale (grid cells’ one), the coefficient reaches high values (0.87 for mean prices and 0.85 for median prices), reflecting the fact that the cases of subversion of the spatial order inherited from the pre-increase period are very few. Hamnett (2009) came to similar conclusions when studying house price increase in London between 1995 and 2006.It could be argued that such stability in the rankings in prices necessarily results from the short period considered, but the very fact that leapfrogging did occurred for a few number of grid cells weakens this argument.

31Without undermining the rent gap hypothesis, these research findings necessitate a broader consideration of Smith’s model to a better understanding of urban house price increase along with spatial structures.

Real estate market and urban transformations

Rental payments everywhere, fillings of rent gaps in very few places

32Restricting the conditions for defining the fillings of rent gaps does not mean restricting the importance of the rent gap hypothesis or the existence of rental payments at the same time. On the contrary, the past fifteen years bluntly stressed the importance of seriously reconsidering the role of ownership in real estate markets and the importance of countering the isolation of prices from social and political evolutions. Many authors showed that house price increase could not be reduced to a simplistic view of the market where the mere relationship between supply and demand would govern price changes. Housing price increases do not involve a contraction of demand which could, in turn, lead to a downward adjustment of prices (Goodman and Thibodeau 2008), nor does the increase in new housing units lead to a housing price decrease (Donzel et al. 2007). Prices depend on broader financial and social factors, which bring back the question of land rent and housing rent. Consequently, due to macro market conditions for which they are absolutely not responsible, homeowners are given a huge bargaining power. The old property market is particularly illustrative of this phenomenon, as prices of properties tripled while properties themselves did not change to the same degree nor the properties’ prices had to include increased costs of production (i.e. land). Topalov (1984) called this phenomenon ‘housing rent’ by analogy with the land rent.

33Yet, the filling of a rent gap cannot be identified with this ubiquitous housing rent unless we also observe a change in status. Housing price increases in the northern part of inner Marseille makes it clear, notably in the 2nd and 3rdarrondissements, i.e., infra-municipal districts specific to the three largest French cities. These two districts are inner, popular parts of Marseille, long dominated by industrial and port facilities (Dell’Umbria 2006). Considering the two grid cells that cover these two sectors, we notice that despite of their very similar characteristics in the mid-1990s, their prices do not evolve similarly. The difference in degree in growth rates becomes a difference in nature in terms of urban structures: while the 3rdarrondissement still remains one of the cheapest parts of the metropolitan area, far below the average price, the prices of the 2ndarrondissement, equally low in 1996, are now superior to the average price. The status of the 3rdarrondissement did not evolve while the 2ndarrondissement rose rapidly within the metropolitan area’s hierarchy. As showed, the value of Spearman’s rank correlation coefficients, this kind of disruption constitutes an exception.

Public investment as a guarantee for private investment

34Paradoxical as it may seem, the future filling of a rent gap cannot be predicted from the sole existence of a rent gap. As a matter of fact, the very notion of a gap between a current rent and a ‘potential’ one requires fixing the level of this PLR. At the same time, the macroeconomic factors governing the general level of prices are not predictable. The very intensity of housing price increases in the Marseille metropolitan area has demonstrated that market actors themselves cannot visualize what could be the “potential” prices in the future (Boulay 2012). This observation challenges us to conceptualize the market from a socio-psychological point of view. The ubiquity of housing price increase and the stability of the neighbourhoods’ hierarchy highlight some implicit components of the idea of PLR and force a reconsideration of the role of public authorities in the very definition of a ‘potential’ rent. The fact that the housing price increase is ubiquitous entails a fundamental point: the catching-up process itself makes the price signal noisier. Such spatial consequences of housing price increase generate doubt in potential investors’ mind about the legitimacy of the ratio between the price and the value. This could lead to an antiselection process (Cahuc 1998) and therefore could block the operation of the market. We have shown how a set of “market devices” (Callon 1998) could allow the exchanges to continue (Boulay 2012). Yet, these market devices account for the global continuance of the market; they do not explain the cases of ‘over-increase’ as this has occurred in the 2ndarrondissement.

35The filling of this rent gap depends on the long-term, massive State investment and the involvement of local authorities in this neighbourhood. Since the beginning of the 1990s, this district has hosted the major urban redevelopment operation in France, the Euroméditerranée OIN (Peraldi and Samson 2006, Bertoncello and Dubois 2010). The State and the local authorities promote the “urban revenge” depicted by Smith (1996) and their actions serve as guarantees for the modification of the image and the status of this part of the city (see Illustration 1 and Map 2).

Illustration 1. ‘Euroméditerranée. The operation transforming Marseille. By 2010, the motorway is moving backwards, the city is moving forwards’. Picture taken in the Porte d’Aix sector, a popular district of the 2nd arrondissement, famous for being disfigured by a motorway

Illustration 1. ‘Euroméditerranée. The operation transforming Marseille. By 2010, the motorway is moving backwards, the city is moving forwards’. Picture taken in the Porte d’Aix sector, a popular district of the 2nd arrondissement, famous for being disfigured by a motorway

Source: author.

36This active involvement of public authorities coincides with a “back to the city movement of capital”, exactly as described in Smith’s model (see Illustration 2 and Map 2).

Illustration 2. ‘Living or investing in the heart of a capital: Marseille

Illustration 2. ‘Living or investing in the heart of a capital: Marseille’

Source: author.

37But much more than a simple coincidence, other advertising show that there is a causal connection between this state-driven redevelopment and the massive real estate investment. The public authorities’ involvement clearly appears as a guarantee for potential investors. That is what happens, for example, in the case of ‘Marseille-Provence 2013 European Capital of Culture’ whose flagship project is a new museum within the Euroméditerranée perimeter (see Illustration 3 and Map 1b).

Illustration 3. ‘Soon available here. Live the Murano (sea side or city side). In the heart of the new urban projects of Marseille 2013

Illustration 3. ‘Soon available here. Live the Murano (sea side or city side). In the heart of the new urban projects of Marseille 2013’

Source: author.

38The fact that real estate programs located near but out of the Euroméditerranée perimeter use it as an argument, sometimes on the bordering on a lie, reflects the importance of such public interventions as investment insurances. For example, illustration 4 shows that some real estate ads lie in order to reassure potential home purchasers: the ‘Rue Cristofal’ – Cristofol, in fact – is outside the OIN and in the core of the 3rdarrondissement, the poorest of Marseille (see Illustration 4 and Map 2).

Illustration 4. ‘New program. Located in the heart of the economic and cultural centres of the city: Euroméditerranée

Illustration 4. ‘New program. Located in the heart of the economic and cultural centres of the city: Euroméditerranée’

Source: author.

39Thus, a joint study of housing price increase throughout the metropolitan area and in the inner city of Marseille calls for an inclusion of the rent gap hypothesis in a broader understanding of both the geography of the market and the socio-political construction of value. Badcock’s (1989: 142) conclusions in his rent gap filling assessment in Australia are similar to ours: “There is no reason for suspending judgment in the case of public sector investment [...]. This strategy not only fulfilled the social welfare objectives of the Dunstan Government, but also gave a lead to the private sector and helped to’ prime’ less promising parts of the inner city property market”. In the case of Marseille, this kind of rent gap filling is very clear: while the prices in the 2nd and in the 3rdarrondissements were similar in the mid-1990s, i.e., among the lowest in the city, the 2ndarrondissement is in 2012 one of the more expensive zone in the whole city of Marseille for the new apartments according to the website of the notaries. These territorial restructurings are not, thus, short-term disruptions but more fundamental trends.


40A geographical approach to the property markets, both at a detailed scale and at a wider metropolitan area scale, provides accurate outcomes. But such geographic and cartographic analyses are rare. This is unfortunate since spatial analysis of housing price dynamics provides both basic information and results challenging traditional theoretical frameworks. For example, the catching-up processes combined with the permanence of spatial hierarchies helps to redefine the conditions of validity for the rent gap hypothesis. More generally, analyses of empirical market phenomena could help better re-conceptualize economic theories underlying research in social sciences in general, and in geography in particular. There are a few interesting paths for economic, social and urban geographers.

41First and foremost, an increase in research directed toward such objects is imperative, as is a diversification in the research areas – countries, cities of all sizes, different market contexts, etc. This is the only way to assess the relevance of the few findings available at the moment, and to evaluate our conceptual frameworks’ refutability. This also means that geographers have to fight against the restrictions on access to fine scale data.

42Our understanding of value creation and relationships between space and economic value are limited by these restrictions. This is even more regrettable in that recent years stressed the role of the real estate sector in our societies and how crucial it was to rethink these questions. The prime concern and the central issue have to do with how various phenomena of different scales work together to lead to unstable promotion of space. Smith himself progressively took into account an increasing number of parameters in the understanding of neighbourhood’s values construction (Smith and Defilippis 1999), but more remains to be done to better understand the interactions between global macroeconomic trends, socio-psychological values and the local production of urban space.

43Finally, another essential point relates to the socio-spatial consequences of rising housing prices. Generally, rising housing prices are studied in the city from the point of view of gentrification as is the case with the rent gap hypothesis does. Much has been learned about gentrification (Lees 2000) but these papers often suffer from a short focal length, and therefore provide little about the dynamics in other areas of the cities, while all the households suffer from higher prices and every part of the cities experienced rising prices. Projects such as Subprime Cities (Aalbers 2012) provide direction and would probably need to be supplemented by other papers.

Top of page


Aalbers MA. (ed.) 2012. Subprime Cities. The Political Economy of Mortgage Markets. London, Wiley Blackwell.

Anah 2005. Les Marchés Locaux de l’Habitat. Paris, Presses de l’Ecole Nationale des Ponts et Chaussées.

Arnott RJ. 1987. Economic theory and housing, in Mills ES. (ed.) Handbook of Regional and Urban Economics. Amsterdam, North Holland: 959-988.

Badcock B. 1989. An Australian view of the rent gap hypothesis. Annals of the Association of American Geographers79(1): 125-145.

Barro RJ. 1991. Economic growth in a cross-section of countries. Quarterly Journal of Economics 106(2): 407-443

Beckerich C. 2001. Biens Publics et Valeurs Immobilières. Paris, ADEF.

Bertoncello B, Dubois J. 2010. Marseille Euroméditerranée, Accélérateur de Métropole. Marseille, Parenthèses.

Boulay G. 2011a. La hausse des prix immobiliers dans les grandes aires urbaines françaises: Anomalie ou normalisation marseillaise? Rives Méditerranéennes 39: 165-181, (Retrieved November 20, 2012).

Boulay G. 2011b. Le prix de la ville. Le marché immobilier à usage résidentiel dans l’aire urbaine de Marseille-Aix-en-Provence (1990-2010). Aix-en-Provence, University of Aix-Marseille 1, unpublished PhD diss.

Boulay G. 2012. Assurer et rassurer le marché: Géographie de l’incertitude économique et dispositifs de marché. Rives Méditerranéennes 42, (Retrieved November 20, 2012).

Bourassa S. 1993. The rent gap debunked. Urban Studies 30(10): 1731-1744.

Buhot C. 2006. Marché du logement et division sociale de l’espace dans les îles du Ponant. Brest, University of Brest, unpublished PhD diss.

Cahuc P. 1998. La Nouvelle Microéconomie. Paris, La Découverte.

Callon M. (ed.). 1998. The Laws of the Market. Oxford, Blackwell.

Clark E. 1988. The rent gap and transformation of the built environment: Case studies in Malmö, 1860-1985. Geografiska Annaler B70: 241-254.

Clark E. 1995. The rent gap re-examined. Urban Studies 32(9): 1489-1503.

Clark E. 2005. The order and simplicity of gentrification – a political challenge, in Atkinson G, Bridge G. (eds) Gentrification in a Global Context: The New Urban Colonialism. London and New York, Routledge: 256-264.

Davis M. 1997. City of Quartz: Los Angeles Capitale du Futur. Paris, La Découverte.

Dell’Umbria A. 2006. Histoire Universelle de Marseille. Marseille, Agone.

Derycke PH. 1996. Equilibre spatial urbain, in Derycke PH, Huriot JM, Pumain D. (eds) Penser la Ville: Théories et Modèles. Paris, Anthropos.

Diewert WE, Nakamura AO, Nakamura LI. 2009. The housing bubble and a new approach to accounting for housing in a CPI. Journal of Housing Economics 18(3): 156-171.

Donzel A, François D, Géniaux G, Napoléone C. 2007. Les Déterminants Socio-Economiques des Marchés Fonciers. Marseille, DRE PACA-PUCA.

Duca JV, Muellbauer J, Murphy A. 2010. Housing markets and the financial crisis of 2007-2009: Lessons for the future. Journal of Financial Stability 6(4): 203-217.

Glaeser EL, Gyourko J, Saiz A. 2008. Housing supply and housing bubbles. Journal of Urban Economics 64(2): 198-217.

Goodman AC, Thibodeau TG. 2008. Where are the speculative bubbles in US housing markets? Journal of Housing Economics 17(2): 117-137.

Greig-Smith P. 1964. Quantitative Plant Ecology. London, Butterworth.

Guérois M, Le Goix R. 2009. La dynamique spatio-temporelle des prix immobiliers à différentes échelles: Le cas des appartements anciens à Paris, 1990-2003. Cybergeo – European Journal of Geography 470, (Retrieved November 20, 2012).

Guigou L. 1982. La Rente Foncière. Paris, Economica.

Hamnett C. 2009. Spatially displaced demand and the changing geography of house prices in London, 1995-2006. Housing Studies 24(3): 301-320.

Harvey D. 1985. The Urbanization of Capital: Studies in the History and Theory of Capitalist Urbanization. Baltimore, Johns Hopkins University Press.

Hwang Smith M, Smith G. 2006. Bubble, bubble, where’s the housing bubble? Brookings Institution Press 1: 1-50.

Lees L. 2000. A reappraisal of gentrification: towards a ‘geography of gentrification’. Progress in Human Geography 24(3): 389-408.

Lloyd CD. 2011. Local Models for Spatial Analysis. Boca Raton, CRC Press.

McLennan D, Tu Y. 1996. Economic perspectives on the structure of local housing markets. Housing Studies 11: 387-406.

Malpezzi S. 2003. Hedonic pricing models: A selective and applied review, in O’Sullivan T, Gibb K. (eds) Housing Economics and Public Policy. Oxford, Blackwell: 67-89.

Millard-Ball A. 2000. Moving beyond the gentrification gaps: Social change, tenure change and gap theories in Stockholm. Urban Studies 37(9): 1673-1693.

Nappi-Choulet I. 1999. Le cycle de l’immobilier de bureaux aux Etats-Unis, in Calcoen F, Cornuel D. (eds) Marchés Immobiliers: Segmentation et Dynamisme. Paris, ADEF: 247-264.

O’Sullivan T, Gibb K. 2003. Housing Economics and Public Policy. London, Blackwell.

Peraldi M, Samson M. 2006. Gouverner Marseille. Enquête sur les Mondes Politiques Marseillais. Paris, La Découverte.

Pumain D, Saint-Julien T. 1997. L’Analyse Spatiale 1. Localisations dans l’Espace. Paris, Armand Colin.

Quah D. 1993. Galton’s fallacy and tests of the convergence hypothesis. The Scandinavian Journal of Economics 95(4): 427-443.

Schmitt G. 2009. Valeurs et usages de l’espace: Approches méthodologiques des dynamiques foncières dans la région Nord-Pas-de-Calais. Lille, University of Lille 1, unpublished PhD diss.

Sheppard S. 1999. Hedonic analysis of housing markets, in Mills ES, Cheshire P. (eds) Handbook of Regional and Urban Economics 3. Amsterdam, Elsevier: 1595-1635.

Shin HB. 2009. Property-based redevelopment and gentrification: The case of Seoul, South Korea. Geoforum 40: 906-917.

Smith N. 1979. Toward a theory of gentrification: A back to the city movement of capital, not people. Journal of the American Planning Association 45(4): 538-548.

Smith N. 1982. Gentrification and uneven development. Economic Geography 58(2): 139-155.

Smith N. 1996. The New Urban Frontier: Gentrification and the Revanchist City. London, Routledge.

Smith N, Defilippis J. 1999. The re-assertion of economics: 1990s gentrification in the Lower East Side. International Journal of Urban and Regional Research 23(4): 638-653.

Topalov C. 1984. Le Profit, la Rente et la Ville. Paris, Economica.

Watkins CA. 2001. The definition and identification of housing submarkets. Environment and Planning A 33(12): 2235-2253.

Wong D, Amrhein CG. 1996. Research on the MAUP: Old wine in a new bottle or real breakthrough? Geographical Systems 3 (2-3): 73-76.

Top of page


1  Thanks are due to Claude Napoléone and Ghislain Géniaux who supported collaboration with the INRA-Avignon and allowed us to use these databases. I have benefited from the enriching comments of the reviewers and Daniel Florentin. The author remains solely responsible for the statements and views expressed herein.

Top of page

List of illustrations

Title Map 1. The Marseille Metropolitan Area
Credits Source: National Institute of Statistics and Economic Studies (INSEE, census). Cartography by the author.
File image/png, 156k
Title Map 2. The Municipality of Marseille showing the location of the pictures taken
Credits Source: Euroméditerranée. Cartography by the author.
File image/png, 240k
Title Map 3. Housing price increase in the Marseille metropolitan area, 1996-2006
Credits Sources: DGFIP and Perval. Data provided by INRA-Avignon. Cartography by the author. Note: mean prices are given for an old three bedroom apartment at 1500 m by 1500 m square cells.
File image/png, 155k
Title Figure 1. Median housing prices in 1996 and house price increase, 1996-2006
Credits Source: author. Note: pre-tax prices, 1500 m by 1500 m cells, old three bedroom apartments.
File image/png, 39k
Title Figure 2. Mean housing prices in 1996 and house price increase, 1996-2006
Credits Source: author. Note: pre-tax prices, 1500 m by 1500 m cells, old three bedroom apartments.
File image/png, 43k
Title Illustration 1. ‘Euroméditerranée. The operation transforming Marseille. By 2010, the motorway is moving backwards, the city is moving forwards’. Picture taken in the Porte d’Aix sector, a popular district of the 2nd arrondissement, famous for being disfigured by a motorway
Credits Source: author.
File image/jpeg, 52k
Title Illustration 2. ‘Living or investing in the heart of a capital: Marseille
Credits Source: author.
File image/jpeg, 80k
Title Illustration 3. ‘Soon available here. Live the Murano (sea side or city side). In the heart of the new urban projects of Marseille 2013
Credits Source: author.
File image/jpeg, 32k
Title Illustration 4. ‘New program. Located in the heart of the economic and cultural centres of the city: Euroméditerranée
Credits Source: author.
File image/png, 1.1M
Top of page


Electronic reference

Guilhem Boulay, « Real estate market and urban transformations: spatio-temporal analysis of house price increase in the centre of Marseille (1996-2010) », Articulo - Journal of Urban Research [Online], 9 | 2012, Online since 20 November 2012, connection on 26 March 2017. URL : ; DOI : 10.4000/articulo.2152

Top of page

About the author

Guilhem Boulay

Guilhem Boulay is a researcher at Ecole normale supérieure de Lyon (ENS), and affiliated to the UMR 7303 TELEMMe Laboratory at the University of Aix-Marseille, France. His research on housing issues aims at better understanding the links between space and the economic theory of value. Email:

Top of page


Creative Commons 3.0 – by-nc-nd, except for those images whose rights are reserved.

Top of page
  • Logo DOAJ - Directory of Open Access Journals