Urban areas are in a constant state of change. This evolution of the built environment is a complex, multifaceted phenomenon driven by both demand and supply forces. On the demand side, shifting demographics (Haase, Haase, Kabisch, & Bischoff, 2008; Nelson, 2006; Pitkin & Myers, 2008), economic factors (Glaeser & Gottlieb, 2006; Glaeser & Shapiro, 2003) and consumer preferences (Myers & Gearin, 2001; Nelson et al., 2004) act simultaneously to shape the demand for urban locations. In terms of supply, the durability of the existing built environment (Evans, 1985) together with land use regulations actively limit the ability of the property market to directly respond to demand changes (Cheshire & Sheppard, 2005). Speciﬁcally from the policy perspective, two decades of growth management legislation in the United States and abroad have sought to replace the low-density, peripheral growth common during the 1970s and 1980s with more centralised and higher density development (Evans, 1999; Weitz, 2012); a concerted effort that has enjoyed mixed success (Boarnet, McLaughlin, & Carruthers, 2011; Dempsey & Plantinga, 2013).
In order to realise the urban intensiﬁcation desired by policy and demanded (at varying degrees) by the market, the redevelopment of existing properties is required. Urban land use change, though heavily regulated by the public sector, is accomplished in large part by the private property market. In other words, the likelihood of redevelopment at any given site at any given time is driven primarily by economic concerns regarding the proﬁtability of investment (Rosenthal & Helsley, 1994). As a result, investors and homeowners seeking to make the most proﬁtable investment as well as policy-makers aiming to create successful urban intensiﬁcation plans require a greater understanding of the property-speciﬁc characteristics that inﬂuence redevelopment.
The broader study of land use change has garnered considerable attention in the planning (Iacono, Levinson, & El-Geneidy, 2008; Wilson & Song, 2011), economics (Bell & Irwin, 2002; Irwin, 2010) and geography (Parker, Manson, Janssen, Hoffmann, & Deadman, 2003; Verburg, Schot, Martin, Dijst, & Veldkamp, 2004) literature. The vast majority of this work, however, has been focused on the conversion of vacant, rural or exurban land into urban uses. Empirical enquiries into the small-scale and often piecemeal succession of land uses within dense urban areas are far less common. To address this deﬁciency, this research examines the inﬂuence of four types of factors – policy, physical, neighbourhood and market – on the probability of low-rise residential redevelopment in the City of Seattle.
The primary goal of this work is to merge the micro-economic theory and research of the real estate discipline with the longitudinal methods frequently used in the planning and applied geography literature (An & Brown, 2008). To do so, this study estimates a duration model with parcel-level land use change data from the City of Seattle over the time period of 2003–2012. A beneﬁt of using a duration model is that the values of the independent variables for each longitudinal observation may vary over time. This is in contrast to the static predictor variables required by the probit models used in previous cross-sectional research on urban residential redevelopment (Dye & McMillen, 2007; Rosenthal & Helsley, 1994). The ability to account for the impact of changes to the surrounding neighbourhood is critical in a dynamic urban environment.
In general, the ﬁndings reached validate the existing theoretical and empirical work, suggesting that factors that directly increase (decrease) the potential profitability or size of the redevelopment also increase (decrease) the redevelopment probability of existing single-family homes. As an addition to the literature, the role of policy and neighbourhood factors are also illustrated. Finally, this research ﬁnds that the effect of option values on low-rise urban redevelopment decisions conﬂicts with the effects shown in previous work focused on new, suburban development.
This paper proceeds by introducing the relevant theory and empirical work on redevelopment. The study design, data and methodology are then summarised. Results from the models are presented followed by a discussion of the ﬁndings and recommendations for market participants and policy-makers.
2. Literature review
Demolition and subsequent redevelopment of an existing property becomes economically viable when the value of the redeveloped property, minus demolition and construction costs (the net land value), is greater than the value of the property in its current use (Dye & McMillen, 2007; Muth, 1969; Wheaton, 1982). The theoretical framework surrounding redevelopment as an economic decision and as part of the broader urban development and monocentric city models were developed two to three decades ago (Brueckner, 1977; Capozza & Helsley, 1990; Capozza & Sick, 1994; Clapp, 1977; Wheaton, 1982). Much of the immediate work that followed sought to solidify the theoretical expectations of urban growth with redevelopment under varying assumptions such as open vs. closed cities (Brueckner, 1980) or perfect foresight vs. uncertainty (Braid, 2001).
An early empirical test of the decision to demolish single-family homes validated the general theory using data from Vancouver, British Columbia (Rosenthal & Helsley, 1994). In this work, demolition probability increased when land values eclipsed the value of the property in its current use. A number of follow-up studies have utilised data from Chicago to test the impact of other factors on the probability of demolition. Smaller home sizes, low lot coverage and high land leverage (ratio of land value to total value) are all shown to increase the probability of demolition for single-family homes (Dye & McMillen, 2007; Weber, Doussard, Bhatta, & Mcgrath, 2006). Neighbourhood socio-economic factors showed little impact (Weber et al., 2006), but proximity to rail transit and historic village centres did increase the likelihood of a single-family home being torn down (Dye & McMillen, 2007).
Analyses of other land uses show similar market dynamics in the commercial/ industrial areas of Chicago (Munneke, 1996) and the apartment market in Seoul, South Korea (Lee, Chung, & Kim, 2005). Related descriptive work on knock down and rebuild activity in Sydney, Australia ﬁnds that the underlying motivations for redevelopment can vary widely depending on the owner’s ﬁnancial situation or type of ownership (Wiesel, Freestone, & Randolph, 2013). Finally, at a coarser scale, neighbourhood-level land use transitions are shown to undergo cyclical trends in redevelopment and growth due to ﬁltering processes and externalities (Rosenthal, 2008).
2.1 Option values
The impact of option values on the development (or redevelopment) decision has also gathered interest among urban scholars. Option values exist in situations where an investment (development or redevelopment in the case of real estate) is irreversible and the future is unknown (Titman, 1985). The basic theory suggests that higher levels of uncertainty can result in higher potential value from waiting to invest which, in turn, decrease the likelihood of (re)development in the present. A number of studies on real estate development support this theory (Bulan, Mayer, & Somerville, 2009; Clapp, Bardos, & Wong, 2012; Cunningham, 2006, 2007; Towe, Nickerson, & Bockstael, 2008). In short, the additional value provided by the option to wait until the future to (re)develop can have the effect of slowing the overall development rate or increasing the time to (re)develop as waiting for future gains may allow a developer/owner to invest at a higher density/intensity and therefore increase the present value of future rent streams.
Early empirical attempts to examine option values in real estate development (real options) have focused on the conversion of vacant land to urban uses, most often single-family housing. A pair of studies on the efﬁcacy of the urban growth area (UGA) boundary in suburban King County, WA have shown that the decreased uncertainty (reduced option values) caused by the UGA led to increased development within the boundary (Cunningham, 2006, 2007). The effect of option values on single-family home development, however, has been shown to be conditional on the size of the individually permitted subdivisions, where larger subdivisions are less impacted by the speculative effect of market volatility (Towe et al., 2008).
More recent work has looked at the relationship between (re)development and option value in urban areas. Examining condominium development in Vancouver, BC, Bulan et al. (2009) ﬁnd that competition by other developers can moderate the impacts of risk (and therefore option value) on development, essentially showing that the beneﬁts to waiting may be decreased with competitive pressure. Grovenstein, Kau, and Munneke’s (2011) work on redevelopment in Chicago highlights the variation in option values across land uses. By modelling the differences between maximum potential developments and actual prices, they ﬁnd that high density residential and warehouse properties may exhibit option values of up to 10 to 12% of the total property value, whereas similar ﬁgures for small retail and industrial sites are in the range of 1 to 2%. Finally, Clapp and Salavei (2010) show that option values can be a signiﬁcant contributor to value for older single-family homes and that failure to account for these inﬂuences in hedonic price studies can result in biased estimates of independent variables, especially those associated with redevelopment probability such as structure age, condition and lot size.
In sum, empirical research on the determinants of redevelopment is a small, but growing set of literature. The ﬁndings to date validate the basic theory of redevelopment in which future proﬁtability is balanced against the value of the current use. The existing studies examining demolition probability exclusively estimate probit models to evaluate the cross-sectional probability of teardown/demolition. The consideration of option value, or the value of waiting to development, has also been added to the discussion. Option values have been shown to impact development levels and timing as well as impart measurable value premiums. A few recent studies have focused on option values in urban areas (Clapp & Salavei, 2010; Grovenstein et al., 2011); however, none explicitly consider redevelopment in the low-rise, detached single-family and small multi-family residential market. Additionally, very little work considers both the demolition of existing properties as well as the construction of a new use in its place or examines the issue of redevelopment probability using longitudinal data and methods.
3. Research design and data
Redevelopment is the primary process by which dense, urban environments evolve. This paper builds on the existing empirical research by examining the entire redevelopment process, from demolition to the ensuing new construction. Whereas previous work has measured redevelopment cross-sectionally, this research uses a longitudinal database to estimate the drivers of redevelopment probability over 10 years of land use change within the City of Seattle; before, during and after the housing market bubble of the mid-2000s.
The study area is limited to the City of Seattle, located at along the eastern shore of the Puget Sound and at the heart of the Seattle–Tacoma metropolitan statistical area. The entire metropolitan region stretches across four counties, King, Kitsap, Pierce and Snohomish and is home to approximately 3.5 million residents. The City of Seattle claims a population of approximately 630,000 as of the 2010 census. The most densely developed municipality in the region, Seattle is also the centre of redevelopment activity.
Seattle is a particularly interesting case for the study of low-rise urban redevelopment. Job growth in the central area of the city has been very high over the past decade, with ﬁrms like Amazon and, recently, Expedia, moving their headquarters into or near the CBD. Other large software companies such as Microsoft, Adobe and Google have recently opened large ofﬁces within the city limits as well. The demand for city living has increased markedly. On the supply side, the city’s overarching Urban Villages plan – active since 1994 – has created policies that encourage the preservation of most existing single-family neighbourhoods while focusing densiﬁcation into well-deﬁned growth areas and corridors. The combination of high demand and structured policy aiming to cluster redevelopment render the Seattle experience an interesting study in low-rise redevelopment.
3.1 The Seattle context
Redevelopment occurs across a number of land uses and at a variety of densities. The factors that drive redevelopment likely differ signiﬁcantly by use and density. To avoid confusing these various use-speciﬁc processes, this study focuses on a single type of redevelopment; the change from detached single-family (including single-family structures currently converted to multi-unit dwellings) to low-rise or medium density residential uses – deﬁned here as townhomes, rowhouses (terraced housing), small apartments and small condominium structures. Certainly, large redevelopment projects such as high-rise residential towers, new business parks and full neighbourhood redevelopment schemes (e.g. South Lake Union in Seattle) are very important contributors to the densiﬁcation of a city. This type of redevelopment gets nearly all of the press and most of the focus in the academic literature. Very little attention, however, is paid to the small scale, piece-by-piece redevelopment of single-family homes to slightly denser multiple-family dwellings. Yet, this piecemeal process can slowly increase local density to a level where public transit is more efﬁcient and neighbourhood retail centres are economically viable. It also has the obvious impact of providing additional housing supply in desirable areas which can release pressure on rapidly increasing housing prices. For these reasons, this study focuses on the transition of single-family homes to small, multi-family structures.
Land transitioning from single-family, detached uses to low-rise residential uses represents the most common form of redevelopment within the City of Seattle. Limiting the set of observations to low-rise residential redevelopment allows this analysis to focus on a single set of zoning categories within the City of Seattle; the Lowrise (LR) zones. By analysing only properties within these zones, the maximum potential building densities for redevelopment are relatively comparable, potential new structures similar in use and entitlement variations between the three zones can be clearly accounted for in the statistical models. Additionally, the vast majority of redevelopment in LR zones is conﬁned to a single parcel, thereby avoiding the necessity of land assemblage and holdout issues that may result in such situations (Miceli & Sirmans, 2007).
The LR zones – LR1, LR2 and LR3 – are the basic transition zone for low-rise residential uses and are widespread throughout the City of Seattle. These areas are primarily clustered in the city’s Urban Villages, the areas in which redevelopment is speciﬁcally being targeted. The LR1 zone possesses the most restrictive density regulations and the LR3 the least. Allowed uses are limited to existing single-family structures and new townhomes, rowhouses, small apartments and small condominium buildings. LR zones allow overall ﬂoor-to-area ratios (FARs) ranging from .9 to a maximum of 2.0. The exact allowed densities vary incrementally within each zone depending on the type of construction, Urban Village location, proximity to transit and conformance to City’s green building code (City of Seattle, 2011). In the majority of situations, the allowable density limits range from FARs of 1.0 (LR1) to 1.4 (LR3). Most of the properties within these zones not having already redeveloped are composed of older detached single-family residential structures built prior to 1950 and, therefore, are particularly susceptible to redevelopment.
This study is focused on the transition of land containing existing single-family dwelling to new, low-rise residential uses within LR zones. As of 2003, 83671 individual tax parcels met these criteria: (1) a single-family detached structure and (2) location in an LR zone. Those parcels that have redeveloped to a higher density residential use over the study period are identiﬁed in the binary REDEV variable, the primary dependent variable. The time of redevelopment or the time at which the observation was censored (end of study period) is recorded in the REDEVTIME variable. Over the 10-year period, 1122, or 13.41% of the universe of parcels redeveloped into a higher density residential use.2 The location of the study observations are shown in Figure 1, colour-coded by zoning designation.
An urban process like redevelopment is inherently spatial. As a result, the spatial context of the data is important in developing models and explaining the phenomenon of interest. Figure 2 shows the probability of redevelopment in LR zones as a surface ﬁtted over the extent of the City of Seattle.3 As shown, conversion rates are much higher in north Seattle as well as in the southwest part of the city (in the area known as West Seattle).
3.3 Independent variables
Independent variables that explain the risk or probability of redevelopment fall into one of four categories; (1) policy; (2) physical; (3) location/neighbourhood and (4) market. Variables are introduced by category below. A detailed discussion of the development of those requiring additional calculations can be found in the online appendices to this paper.4 Table 1 contains basic information on each variable, while summary statistics are shown in Table 2.
The primary governing forms of legal entitlements within the city are zoning regulations. To measure zoning designation, three variables indicate location within the three LR zones (LR1, LR2 and LR3). The FARs in these zones generally range from 1.0 to 1.4.
As previous research suggests, both the existing structure and the characteristics of the lot can inﬂuence the probability of teardown, a precursor to redevelopment (Dye & McMillen, 2007; Weber et al., 2006). Data collected on the structure include the size of the main structure, STRSIZE, the year the main structure was built, YEARBUILT, the condition of the property, COND, the quality of construction, BLDGGRADE,5 an indicator variable denoting that the home has been divided up into multiple living units, XUNITS, and a variable indicating that a separate, accessory dwelling unit is present on the property, ADU.
Characteristics of the lot are described by variables representing the lot size, LOTSIZE, the shape or square-ness of the lot, SHPFCTR, the King County Assessor’s assessment of view quality, VIEW and challenging site topography, TOPO. The use of accessory dwelling units, lot shape, views and topography represent additions to the set of physical variables commonly considered in the literature.
Existing empirical work suggests that the difference between existing density and potential density under existing zoning regulations may prove illustrative in accounting for redevelopment probability (Grovenstein et al., 2011). As a result, a combined physical and policy variable, FARDIF, is developed which compares the density of the existing property (represented as the FAR or the total structural square footage/total lot square footage) to the potential density given the property’s zoning.6
Next, a set of location or neighbourhood variables are developed. These include measures of accessibility, socio-economic and neighbourhood design. Accessibility is measured as the straight-line distance to the centre of downtown Seattle, DISTCBD.
Census 2010 median income in $10,000, MEDINC, and percentage of renter households, RENTPER, at the block group level are included as indicators of socio-economic status. Neighbourhood design is composed of features that are relatively stable, such as the street network, as well as those that are not, such as overall neighbourhood density and the mixture of land uses in the area. As a result, two different types of variables are developed here, those that are static (time-invariant) and those that change over the 10-year period of the study (time-dependent). The time-invariant design variable is a measure of intersection density, INTDENS. Time-dependent measures of neighbourhood design include land use mix, LUMIX, overall neighbourhood FAR, NBHFAR, and adjacency to a commercial, industrial or large multiple-family use, ADJCIMU. Time-dependent variables are collected once annually to form a longitudinal data-set. The term ‘neighbourhood’ is also ambiguous. This study uses GIS analysis to create a 400 m (1/4 mile) radius around each observation from which to calculate neighbourhood design variables.
Finally, four market-based variables are developed. First, measures of market change and uncertainty are created to examine the impact of option values on low-rise urban redevelopment. The ﬁrst is a metric of overall neighbourhood price appreciation over the prior three years, NBHTREND, and second, the level of uncertainty surrounding this trend, NBHUNCERT.
These two variables loosely follow the trend and uncertainty variables used by Cunningham (2007) and Towe et al. (2008). Finally, two indicators of local competition (Bulan et al., 2009) or priming (Wilson & Song, 2011) are developed. These variables, NBHREDEV and ADJREDEV, measure the number of low-rise residential redevelopments having occurred over the past three years within the neighbourhood (400 m radius) and the adjacent parcels, respectively.
The existing literature on redevelopment and teardown probability exclusively examines cross-sectional data using probit model speciﬁcations (Dye & McMillen, 2007; Rosenthal & Helsley, 1994; Weber et al., 2006). Recent studies looking at suburban land use change have adapted duration models for this task (Towe et al., 2008; Wilson & Song, 2011) A simulated comparison of the two model types has shown that when time-varying independent variables are present and when the study involves an extended time period, duration models outperform probit speciﬁcations (Wang, Brown, An, Yang, & Ligmann-Zielinska, 2013). As the data for this study are longitudinal and contain time-varying independent variables, a duration model speciﬁcation known as the Cox Proportional Hazards model is used here to estimate low-rise residential redevelopment probability in the City of Seattle.
A Cox Proportional Hazards model directly estimates the hazard rate (the instantaneous probability of an event) with a log-linear model speciﬁed as follows (Allison, 2010; Vermunt, 1997):
where h0(t) is the baseline hazard (variable over time), βj is the coefﬁcient vector and xj is the matrix of independent variables. A key component of the Cox model is that the underlying base hazard, h0, does not need to be known or speciﬁed a priori since this constant cancels out during the partial likelihood calculation used to estimate the model. This model has been estimated in a multiplicative form, meaning that an estimate of .2 indicates a 20% change in redevelopment risk due to a one unit increase in the independent variable. In interpreting the raw coefﬁcient estimates, values less (greater) than 0 equate to a reduction (increase) in the probability of the event.7
To examine the factors correlated with low-rise residential redevelopment in the City of Seattle, a number of sequentially more complex Cox Proportional Hazards models have been estimated using the data discussed above. Summary model results, including coefﬁcient estimates and indicators of statistical signiﬁcance α = .05) are shown in Table 3.8 Model 1 examines only physical and policy variables. In Model 2, the combined FARDIF variable is added. Model 3 sees the addition of neighbourhood variables, while the market variables are inserted into the speciﬁcation in Model 4. The Results section that immediately follows covers the basic output of the models in sequential order. A more thorough examination of the resulting coefﬁcient estimates and their importance in explaining low-rise residential redevelopment in Seattle is addressed in the subsequent Discussion section.
The impact of basic policy and physical variables on redevelopment probability are estimated in Model 1.9 All of the variables show signiﬁcant impacts on the risk of redevelopment, with the exception of structure condition, COND, and quality, BLDGGRADE. All variables show the expected signs with the exception of the BLDGGRADE and VIEW variables. In Model 2, FARDIF is added to the speciﬁcation. The FARDIF variable contains a linear combination of the STRSIZE, the LOTSIZE and one of the zoning designation variables (LR1, LR2 or LR3) and, as a result, two of these three variables are removed when FARDIF is entered into the model due to near perfect multicollinearity that results. FARDIF shows a signiﬁcant positive association with increased redevelopment probability and accounts for the considerable change in the lot size (logLOTSIZE) chefﬁcient.
Location and time-varying neighbourhood variables are added to the speciﬁcation in Model 3. In doing so, the model now uses the panel data with the total number of observations increasing from 8367 (Models 1 and 2) to 77,764. Four of the seven location variables – RENTPER, DISTCBD, NBHFAR and ADJCIMU – are signiﬁcant at the α = .05 level. As expected, overall neighbourhood density, NBHFAR and adjacency to a commercial, industrial or multiple-family use, ADJCIMU, have a positive relationship with redevelopment. Per cent of renter-occupied units, RENTPER, on the other hand, shows a negative relationship with redevelopment probability.10
Finally, the market variables are added in Model 4. Localised price trends, NBHTREND, show no signiﬁcant impact, but price deviations, NBHUNCERT, that are higher than expected do exhibit a positive relationship with redevelopment.11 Both measures of competition or priming effects are positively related to redevelopment probability. By adding these four market variables, all other co-variates have maintained their sign though coefﬁcient magnitudes have shifted slightly. This relative consistency over the four sequential speciﬁcations speaks to the robustness of the model.
In addition to the coefﬁcients, the bottom of Table 3 contains a number of model diagnostics and measures of ﬁt. The ﬁrst, concordance, is a measure of the discriminatory power of the model.12 The discriminatory ability of the four models range from a low of .757 for Model 1 to a high of .781 for Model 4, suggesting that this set of models correctly discriminates between a redeveloped and non-redeveloped property (by measure of risk score) 76 to 78% of the time. Across all speciﬁcation types, the concordance measure can be used for comparison. As the models become more complex left to right in Table 3, the concordance measure increases, indicating that each change to the model helps discriminate to a great degree than the one before. Measures of the log-likelihood (logLik) and Akaike information criteria (AIC) are also shown and also suggest improvement with each model addition.13
Finally, to aid in visualising and comparing these results across variables, Figure 3 illustrates the impact on redevelopment probability associated with a one standard deviation increase (or 1 unit increase for binary variables) for all variables in the preferred model, Model 4.
The impacts of legal development entitlements are highly inﬂuential in the redevelopment process. Location within an LR2 or LR3 zone provides moderate increases in the allowable FAR of about .2–.3 for LR2 and .2–.5 for LR3 compared to the LR1 zone. Yet, location within an LR2 or LR3 zone doubles the odds of redevelopment (Model 1). These ﬁndings suggest that there is a threshold point in terms of density at which redevelopment to low-rise, medium density housing is proﬁtable and that a small increase in FAR from 1.0 to 1.3 can cross this threshold leading to greatly increased rates of redevelopment. A mix of policy and physical characteristics, the FARDIF variable (Models 2 through 4) shows a strong and consistently positive impact on redevelopment risk across all models in which it is used. An increase in this difference of .1 FAR relates to an increase in the probability of redevelopment of around 35 to 39%.
Structural characteristics such as the shape of the lot, SHPFCTR, and multiple units in the structure (a proxy for prior conversion to duplex/triplex/quadplex), XUNITS, show a positive relationship with redevelopment risk, as expected. More square lots allow for greater building footprints in most cases due to setback restrictions in the LR zones. Prior conversion of an existing single-family home to a multi-family use, XUNITS, signals that the property is being used as an investment as opposed to an owner occupied home, a fact that is likely to increase the probability of the owner’s willingness to sell and/or teardown the structure if ﬁnancial calculations suggest that such an action would be proﬁtable. Structure condition, COND, and quality, BLDGGRADE, show no signiﬁcant relationship to redevelopment. Better condition and quality do, ceteris paribus, increase the value that must be paid for a to-be-torn down home; however, land values throughout most of the City of Seattle make up at least 50% of the total value of the property – often much more – and, as a result, the additional premium paid for a slightly better condition or higher quality home is a very small percentage of the overall sales price of the property. Newer homes, YEARBUILT, additional dwelling structures, ADU, and difﬁcult topography, TOPO, decrease redevelopment probability, all of which are in line with the basic redevelopment theory.
View quality, VIEW, shows a signiﬁcant negative relationship with redevelopment probability. An explanation for this is that properties with a view have a moderate level of topographical challenge, not enough to warrant a classiﬁcation by the county assessor, but enough to limit potential redevelopment via higher construction costs and decreased density. Additionally, properties with good views are more likely to have high land values due to the view amenity, a cost which may not be able to be fully passed on to future buyers or renters since only some portion of the potential multi-unit redevelopment may be able to capitalise on the view.
Within the location/neighbourhood variables, moderately strong relationships to redevelopment are found. Overall neighbourhood density, NBHFAR, and adjacency to a commercial/industrial/multi-family use, ADJCIMU, both show signiﬁcantly positive relationships with redevelopment probability. These results are expected, as both overall density and adjacency to non-single-family uses are indicators of an area with demand for redevelopment. At the neighbourhood scale, however, both land use mix, LUMIX, and intersection density, INTDENS, show no signiﬁcant impact on the likelihood of redevelopment. At ﬁrst these results may appear counterintuitive since the walkability provided by dense street networks and mixed uses are often cited as amenities (Matthews & Turnbull, 2007; Pivo & Fisher, 2011). Most low-rise residential redevelopment, however, is composed of owner-occupied housing units (townhomes and small condominium projects) the buyers of which may value the quiet streets and easy parking that are more common in single use, residential neighbourhoods (Song & Knaap, 2003). The neighbourhood boundaries used in the GIS calculation here are small, 400m, and these ﬁndings signal that the cost and beneﬁts of close proximity to mixed use areas are offsetting at the 400 m scale. Finally, the ratio of nearby dwellings occupied by renters, RENTPER, exerts a negative impact on redevelopment risk, while median income, MEDINC, shows no signiﬁcant effect. Again, due to the prevalence of owner-occupied units in low-rise redevelopment, proximity to areas largely composed of renters is considered a nuisance, resulting in decreased demand. The ambiguous effect of median income could be due to issues of causality, i.e. it is difﬁcult to disentangle whether income drives redevelopment or if the redevelopment is signiﬁcantly changing the demographics of the neighbourhood and, therefore, observed income levels.
The ﬁnal variables added to the models are those dealing with market effects. First are measures related to market price trends and the uncertainty in these trends, grouped as option value effects. NBHTREND measures the impact that localised home price trends have on the risk of redevelopment. As Table 3 shows, localised home price trends have an ambivalent effect on the probability of redevelopment; a ﬁnding that warrants discussion. Option value theory suggests that in the face of rising prices (potential proﬁt), an owner may decide to withhold investment in the hopes that in a future time period the property may be developed to a more intensive highest and best use, therefore resulting in a higher present value to the owner than a redevelopment today. In this situation, there is value in the option to wait and increasing prices will have a negative inﬂuence on redevelopment probability.
However, two problems exist with this line of thinking in the case of low-rise residential redevelopment in the City of Seattle. The ﬁrst is that most properties that are redeveloped must be purchased by a developer. As a result, increasing prices raise the cost of purchasing the property which may not mean an increase in overall proﬁtability from the redevelopment, despite the rising market. Existing option value theory (Titman, 1985) is generally concerned with an owner who is also the developer, a situation which is not very common in the case of redevelopment from a detached single-family home to a low-rise residential use. Simply put, the current owners are not the redevelopers. The second issue with option value theory in this context is that zoning and other land use regulations limit the potential highest and best use that may be achieved in the future. Therefore, while increased prices (market demand) may increase the ultimate future price that may be charged for units, they usually cannot directly increase the total intensity (number of units) that can be built, barring a change in legal entitlements. In such a situation, option value is likely to be greatly diminished by existing land use regulations that limit the highest and best use of the property.14 The data in this study deal solely with properties that have not changed zoning over the 10-year study period. Given these two issues, the ﬁndings of these models, though contrasted with much of the existing work on option values (Bulan et al., 2009; Cunningham, 2006, 2007; Grovenstein et al., 2011) are reasonable. One rationale for this contradiction is the fact that most of the previous studies deal with suburban vacant land. In other words, due to the greater inﬂexibility in land use regulations in an urban setting, option values play less of a role in urban redevelopment decisions than in subdivision development at the periphery.
A related issue to be considered here is that within a Cox Proportional Hazards model the impact of time is controlled by the partial likelihood calculation. In other words, the coefﬁcient estimates on NBHTREND represent the change in redevelopment probability due to differences in local house price trends over space not over time. As a result, the coefﬁcient estimates here are interpreted as indicating the effect that greater price appreciation in one area has on redevelopment, ceteris paribus, to the same property in an area with a lower price appreciation trend at the same point in time. Given this, the insigniﬁcant effect of localised price trend simply suggests that spatial differences in price trends have a de minimis impact on likelihood of redevelopment. In sum, the two theoretical issues noted above combined with the methodological constraints noted here render the deviation from earlier published ﬁndings on option values credible.
Theoretically speaking, uncertainty of any kind can make waiting to invest more proﬁtable; however, it is possible that positive deviations from expected trends may have different effects than negative ones on redevelopment decisions. The result from Model 4 (Table 3) suggests that, indeed, local house price trends that deviated from expectations, NBHUNCERT in positive (negative) manner have a strong positive (negative) relationship with an increased odds of redevelopment even when controlling for other co-variates as well as the localised home price trends for three years prior to the redevelopment. Due to the strength of this relationship, there may be some concern that this metric is simply accounting for the housing market bubble, a phenomenon that has a correlation with higher rates of redevelopment in Seattle. Limiting this concern is the fact that the Cox Proportion Hazards model speciﬁcation allows for the base hazard of the event (redevelopment in this case) to vary over time, therefore eliminating the biasing effects any correlation between uncertainty measures and the increased overall likelihood of redevelopment during the housing boom.
Unlike the NBHTREND variable, spatial variations in price trend uncertainty, NBHUNCERT, do have a positive inﬂuence on the decision to redevelop. This ﬁnding indicates that rapid, positive and somewhat unexpected, changes in neighbourhood price dynamics are a key factor attracting redevelopment. In such a situation, developers who may be searching for the best value or speculating on the next hot neighbourhood – buying land cheap and beneﬁting from neighbourhood improvement – when choosing redevelopment sites are quick to invest in the new, trendy neighbourhood of the moment.
Finally, the impacts of previous neighbourhood, NBHREDEV, and adjacent, ADJREDEV, redevelopment are tested. In other literature, similar variables are either considered to measure competition in the market (Bulan et al., 2009) or a priming effect indicating that the market is ready for a particular type of development (Wilson & Song, 2011). In both cases, the impact of these variables is hypothesised to increase redevelopment. The results here show similar relationships as overall neighbourhood redevelopment counts and adjacency to previous low-rise redevelopment are both associated with an increase in the probability of redevelopment. Conceptually, these ﬁndings make sense as competition can decrease option value and encourage development while priming effects reduce the ﬁnancial risk that a new development type in a given area may have. In terms of priming effects, the evidence of existing, successful developments in an area will help convince lenders of the feasibility of a project and therefore increase the chance of a developer ﬁnding sufﬁcient ﬁnancing for a redevelopment project.
Overall, the strongest relationships, both negative and positive, are those involving the size of the current structure, size of the lot, the zoning designation and the difference in densities between existing and potential developments. These ﬁndings, much like previous work (Dye & McMillen, 2007; Rosenthal & Helsley, 1994; Weber et al., 2006), validate the primacy of proﬁtability in explaining redevelopment likelihood. Adjacency to a recent low-rise redevelopment also exerts a large, positive impact on redevelopment probability, while view amenities and challenging topography dampen the likelihood of a redevelopment event.
This research extends the literature on low-rise redevelopment by measuring impacts from four sets of factors; policy, structural, neighbourhood and market. The ﬁndings validate previous work that identiﬁed basic structural and land factors such as home size, home age and lot size as the key indicators of redevelopment probability. The results of this study add further understanding as to the importance of legal entitlements (zoning) as well as insight into which neighbourhood factors do and do not contribute to the likelihood of low-rise redevelopment. Overall, factors that increase the potential income from a redevelopment or lower the costs of such development will have a positive impact on redevelopment probability. In sum, investors and developers looking to redevelop land containing single-family homes and/or policies seeking to encourage urban intensiﬁcation cannot ignore the physical constraints and the existing built environment, speciﬁcally in regards to lot and home sizes. The interaction between these physical variables and legal entitlements is a critical factor in determining low-rise residential redevelopment probability.
In terms of neighbourhood effects, overall neighbourhood density and adjacency to non-single-family uses show a positive relationship with redevelopment probability. Similarly, the presence of nearby low-rise redevelopment also increases redevelopment probability. These relationships highlight the priming effects in real estate development as both developers and their ﬁnanciers use nearby market activity to gauge the potential success of a proposed project. Finally, an examination of option value factors does not ﬁnd that option values inhibit low-rise redevelopment. Much of the existing empirical work and basic theoretical underpinnings of option values do not measure urban land use changes, strict density constraints or a preponderance of owner-occupiers; conditions found in LR zones in the City of Seattle. More work on the impact of option values on land use change under inﬂexible zoning regimes and/or within dense urban areas is needed.
Finally, this study provides a methodological contribution to the urban redevelopment literature. Previous analyses of dense, urban redevelopment have extensively performed cross-sectional analyses with logit or probit model speciﬁcations. To date, prior work using duration models has focused on analysing the factors contributing to peripheral, low-density suburban development. This study contributes by being the ﬁrst to apply a duration model to urban redevelopment decisions. Use of a duration model permits extension of traditional cross-sectional studies to a more robust longitudinal research design. It also allows for the effects of time-varying independent variables such as neighbourhood density and changing market trends to be considered in modelling redevelopment probability.
No potential conﬂict of interest was reported by the author.
Supplemental data for this article can be accessed here [http://dx.doi.org/10.1080/09599916. 2015.1048705].
1. The following ﬁltering criteria were applied to the original data-set of 8715 properties in order to eliminate observations with uncommon data or characteristics not representative of a typical single-family home at risk for redevelopment. Removed were observations with: (1) lot size less than 2000 or more than 20,000 square feet; (2) structural size less than 400 or greater than 8000 square feet; (3) a structure of more than four stories; and (4) a 2003 assessed value per square foot of land greater than
$100 or per square foot of improvement greater than $300. In sum, 348 total observations were removed from the original data-set of 8715 (3.99%).
2. Of these redevelopments, 98% involved townhome or rowhouse construction. The remaining 2% are composed of small apartments and condominiums.
3. Rates are calculated as the redevelopment rate of the nearest 100 observations to each grid point on the map.
5. Condition and quality based on determinations by that King County Assessor.
6. The resulting calculation is scaled by a factor of 100 for interpretation purposes.
7. The models used in this research have been estimated in the R software package (R Core Team, 2013) using the coxph() function from the survival package (Therneau, 2013). The data is singly, right-censored (Type I).
8. Full model results are available in the online appendices, hosted at www.andykrause. com/publications.
9. The hedonic price literature has shown that non-linear (Clapp & Salavei, 2010) or vintage effects (Bitter, 2014; Coulson & McMillen, 2008) often exist when modelling the relationship between home age and home prices. Tests of non-linear effects on the YEARBUILT variable added no additional explanatory power to the model. Additionally, tests of vintage effects suggest that grouping homes by vintage decreases the explanatory power of the model and does not impact other coefﬁcient estimates. As a result, neither non-linear nor vintage effects were considered in the ﬁnal model speciﬁcations. In short, non-linear and vintage effects found to impact home prices may be less pronounced in determining redevelopment probability.
10. Sensitivity tests using alternate neighbourhood deﬁnitions of 200 and 800 m produced very similar parameter estimates for both the neighbourhood/location variables and other co-variates.
11. In this context, localised is deﬁned as the nearest 500 home sales within the past three years.
12. From Therneau (2013, p. 79): ‘Concordance is deﬁned as Pr(agreement) for any two randomly chosen observations, where in this case agreement means that the observation with the shorter survival time of the two also has the larger risk score. The predictor (or risk score) will often be the result of a Cox model or other regression. For continuous covariates concordance is equivalent to Kendall’s tau, and for logistic regression it is equivalent to the area under the ROC curve. A value of 1 signiﬁes perfect agreement, .6–.7 is a common result for survival data, .5 is an agreement that is no better than chance and .3–.4 is the performance of some stock market analysts’.
13. Note that AIC and logLik values cannot be compared between Models 1/2 and Models 3/4 due to differences in the underlying data.
14. The ‘binding-ness’ of the zoning regulations in these zones is evidenced by the threshold effect on redevelopment probability observed when moving from LR1 to LR2 or LR3 zones.
Notes on contributor
Andy Krause joined the University of Melbourne as a lecturer in Property in 2015. He teaches courses in property research methods and property development. He earned a doctorate from the Interdisciplinary Urban Design and Planning program at the University of Washington in Seattle, WA, USA. Prior to and during his doctoral studies, he maintained a position as a senior analyst at Greenﬁeld Advisors – a Seattle-based valuation consulting ﬁrm – where he focused on building automated valuation models for use in litigation. He has also served as a consulting data scientist for Zillow, a web-based real estate information ﬁrm. His research focuses in two main directions: (1) development of automated valuation models, speciﬁcally in regards to providing land valuations; and (2) understanding the role of data systems in identifying, analysing, visualizing and, ultimately, encouraging urban redevelopment. He is a member of the American Real Estate Society and the Western Regional Science Association and regularly presents research at related conferences. His work has been published in the Journal of Real Estate Research, the Journal of Property Investment and Finance, Urban Geography, Cities and the Journal of Property Tax Assessment and Administration.
Andy Krause http://orcid.org/0000-0002-4771-5623
Allison, P. (2010). Survival analysis using SAS: A practical guide (2nd ed.). Cary, NC: SAS Institute.
An, L., & Brown, D. (2008). Survival analysis in land change science: Integrating with GIScience to address temporal complexities. Annals of the Association of American Geographers, 98, 323–344.
Bell, K., & Irwin, E. (2002). Spatially explicit micro-level modelling of land use change at the rural–urban interface. Agricultural Economics, 27, 217–232.
Bitter, C. (2014). Subdivision vintage and housing prices: Do home buyers value traditional development? Urban Studies, 51, 1038–1056.
Boarnet, M., McLaughlin, R., & Carruthers, J. (2011). Does state growth management change the pattern of urban growth? Evidence from Florida. Regional Science and Urban Economics, 41, 236–252.
Braid, R. (2001). Spatial growth and redevelopment with perfect foresight and durable housing. Journal of Urban Economics, 49, 425–452.
Brueckner, J. (1977). The determinants of residential succession. Journal of Urban Economics, 4, 455–459.
Brueckner, J. (1980). Residential succession and land-use dynamics in a vintage model of urban housing.
Regional Science and Urban Economics, 10, 225–240.
Bulan, L., Mayer, C., & Somerville, C. (2009). Irreversible investment, real options, and competition: Evidence from real estate development. Journal of Urban Economics, 65, 237–251.
Capozza, D., & Helsley, R. (1990). The stochastic city. Journal of Urban Economics, 28, 187–203.
Capozza, D., & Sick, G. (1994). The risk structure of land markets. Journal of Urban Economics, 35, 297–319.
Cheshire, P., & Sheppard, S. (2005). The introduction of price signals into land use planning decision-making: A proposal. Urban Studies, 42, 647–663.
City of Seattle. (2011). Seattle’s low rise zones. Retrieved August 22, 2013, from http:// www.seattle.gov/dpd/cms/groups/pan/pan/documents/web\_informational/dpds021571.pdf
Clapp, J. (1977). Urban land use succession under risk. Urban Studies, 14, 73–77.
Clapp, J., Bardos, K., & Wong, S. (2012). Empirical estimation of the option premium for residential redevelopment. Regional Science and Urban Economics, 42, 240–256.
Clapp, J., & Salavei, K. (2010). Hedonic pricing with redevelopment options: A new approach to estimating depreciation effects. Journal of Urban Economics, 67, 362–377.
Coulson, N., & McMillen, D. (2008). Estimating time, age and vintage effects in housing prices. Journal of Housing Economics, 17, 138–151.
Cunningham, C. (2006). House price uncertainty, timing of development and vacant land prices: Evidence for real options in Seattle. Journal of Urban Economics, 59, 1–31.
Cunningham, C. (2007). Growth controls, real options, and land development. Review of Economics and Statistics, 89, 343–358.
Dempsey, J., & Plantinga, A. (2013). How well do urban growth boundaries contain development? Results for Oregon using a difference-in-difference estimator. Regional Science and Urban Economics, 43, 996–1007.
Dye, R., & McMillen, D. (2007). Teardowns and land values in the Chicago metropolitan area. Journal of Urban Economics, 61, 45–63.
Evans, A. (1985). Urban economics: An introduction. Oxford: Basil Blackwell.
Evans, A. (1999). The land market and government intervention. Handbook of Regional and Urban Economics, 3, 1637–1669.
Glaeser, E., & Gottlieb, J. (2006). Urban resurgence and the consumer city (Discussion Paper No. 2109). Cambridge, MA: Harvard Institute of Economic Research.
Glaeser, E., & Shapiro, J. (2003). Urban growth in the 1990s: Is city living back? Journal of Regional Science, 43, 139–165.
Grovenstein, R., Kau, J., & Munneke, H. (2011). Development value: A real options approach using empirical data. The Journal of Real Estate Finance and Economics, 43, 321–335.
Haase, D., Haase, A., Kabisch, S., & Bischoff, P. (2008). Guidelines for the “Perfect Inner City”. Discussing the appropriateness of monitoring approaches for reurbanization. European Planning Studies, 16, 1075–1100.
Iacono, M., Levinson, D., & El-Geneidy, A. (2008). Models of transportation and land use change: A guide to the territory. Journal of Planning Literature, 22, 323–340.
Irwin, E. (2010). New directions for urban economic models of land use change: Incorporating spatial dynamics and heterogeneity. Journal of Regional Science, 50, 65–91.
Lee, B., Chung, E., & Kim, Y. (2005). Dwelling age, redevelopment, and housing prices: The case of apartment complexes in Seoul. The Journal of Real Estate Finance and Economics, 30, 55–80.
Matthews, J., & Turnbull, G. (2007). Neighborhood street layout and property value: The interaction of accessibility and land use mix. The Journal of Real Estate Finance and Economics, 35, 111–141.
Miceli, T., & Sirmans, C. (2007). The holdout problem, urban sprawl, and eminent domain.
Journal of Housing Economics, 16, 309–319.
Munneke, H. (1996). Redevelopment decisions for commercial and industrial properties.
Journal of Urban Economics, 39, 229–253.
Muth, R. (1969). Cities and housing. Chicago, IL: University of Chicago Press.
Myers, D., & Gearin, E. (2001). Current preferences and future demand for denser residential environments. Housing Policy Debate, 12, 633–659.
Nelson, A. (2006). Leadership in a new era: Comment on “Planning Leadership in a New Era”. Journal of the American Planning Association, 72, 393–409.
Nelson, A., Burby, R., Feser, E., Dawkins, C., Malizia, E., & Quercia, R. (2004). Urban containment and central-city revitalization. Journal of the American Planning Association, 70, 411–425.
Parker, D., Manson, S., Janssen, M., Hoffmann, M., & Deadman, P. (2003). Multi-agent systems for the simulation of land-use and land-cover change: A review. Annals of the Association of American Geographers, 93, 314–337.
Pitkin, J., & Myers, D. (2008). U.S. housing trends: Generational changes and the outlook to 2050. (Special report to the Transportation Research Board Report No. 298). Washington, DC: National Academy of Sciences.
Pivo, G., & Fisher, J. (2011). The walkability premium in commercial real estate investments. Real Estate Economics, 39, 185–219.
R Core Team. (2013). R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. Retrieved from www.r-project.org
Rosenthal, S. (2008). Old homes, externalities, and poor neighborhoods. A model of urban decline and renewal. Journal of Urban Economics, 63, 816–840.
Rosenthal, S., & Helsley, R. (1994). Redevelopment and the urban land price gradient.
Journal of Urban Economics, 35, 182–200.
Song, Y., & Knaap, G. (2003). New urbanism and housing values: A disaggregate assessment. Journal of Urban Economics, 54, 218–238.
Therneau, T. (2013). Survival: Survival analysis. R Package Version 2.37-4.
Titman, S. (1985). Urban land prices under uncertainty. American Economic Review, 75, 505–514.
Towe, C., Nickerson, C., & Bockstael, N. (2008). An empirical examination of the timing of land conversions in the presence of farmland preservation programs. American Journal of Agricultural Economics, 90, 613–626.
Verburg, P., Schot, P., Martin, P., Dijst, J., & Veldkamp, A. (2004). Land use change modelling: Current practice and research priorities. GeoJournal, 61, 309–324.
Vermunt, J. (1997). Log-linear models for event histories. Thousand Oaks, CA: Sage.
Wang, N., Brown, D., An, L., Yang, S., & Ligmann-Zielinska, A. (2013). Comparative performance of logistic regression and survival analysis for detecting spatial predictors of land-use change. International Journal of Geographical Information Science, 27, 1960–1982.
Weber, R., Doussard, M., Bhatta, S., & Mcgrath, D. (2006). Tearing the city down: Understanding demolition activity in gentrifying neighborhoods. Journal of Urban Affairs, 28, 19–41.
Weitz, J. (2012). Growth management in the United States, 2000–2010: A decennial review and synthesis. Journal of Planning Literature, 27, 394–433.
Wheaton, W. (1982). Urban spatial development with durable but replaceable capital.
Journal of Urban Economics, 12, 53–67.
Wiesel, I., Freestone, R., & Randolph, B. (2013). Owner-driven suburban renewal: Motivations, risks and strategies in ‘Knockdown and Rebuild’ processes in Sydney, Australia. Housing Studies, 28, 701–719.
Wilson, B., & Song, Y. (2011). Do large residential subdivisions induce further development? Journal of the American Planning Association, 77, 5–22.