Regional Science and Urban Economics 101 (2023) 103914 Contents lists available at ScienceDirect Regional Science and Urban Economics journal homepage: www.elsevier.com/locate/regec Real estate prices and land use regulations: Evidence from the Law of Heights in Bogotá Diego Buitrago-Mora a ,1 , Miquel-Àngel Garcia-López a,b ,∗,1 a Department of Applied Economics, Universitat Autònoma de Barcelona, Edifici B, Facultat d’Economia i Empresa, 08193 Cerdanyola del Vallès, Spain b Institut d’Economia de Barcelona, Spain ARTICLE INFO ABSTRACT JEL classification: Between 2015 and 2017, the Law of Heights (Policy-562) regulated areas of urban renewal in specific locations R14 of Bogotá (Colombia). Using a novel dataset based on detailed information at the block level between 2008 R31 and 2017, we study whether this policy affected real estate prices. Our empirical strategy compares the price R58 per square meter before and after Policy-562 in treated blocks and in control blocks with similar pre-treatment Keywords: traits. Results show that prices increased more in treated blocks than in the rest of the city. We also provide Real estate prices evidence that results are heterogeneous from a temporal, land use and strata point of view. Land regulations 1. Introduction rules on urban land use, but with a low level of enforcement, and finds that the impact of a greater degree of regulation on formal market The relationship between land use regulations and real estate prices prices is unclear. Monkkonen and Ronconi (2013) finds a negative is well documented in developed countries (Quigley and Rosenthal, relationship between regulation and land prices in the three major 2005; Turner et al., 2014; Freemark, 2020; Greenaway-McGrevy et al., Argentinian metropolitan areas with higher levels of regulation and 2021). In general, empirical evidence centered on housing markets lower levels of compliance. For the case of Beijing (China), Ling et al. finds that a greater degree of regulation not only increases housing (2013) find that land control policies accelerated housing prices when prices (Ihlanfeldt, 2007), but also accelerates their reduction in an they were implemented. Finally, Brueckner et al. (2017) find that economic recession (Huang and Tang, 2012), and the effects vary building height restrictions in terms of floor area ratio increases land considerably at the intra-city level (Kok et al., 2014). prices in Chinese cities. On the other hand, little is known about this relationship in devel- This paper aims to contribute to this literature by studying the oping countries. Mayo and Sheppard (1996) compare housing supply regulations in Malaysia, Thailand, and South Korea. Brueckner and impact of a particular regulation, the so-called Law of Heights (Policy- Sridhar (2012) find that building height limits caused spatial expansion 562), on real estate prices in a city of a developing country, Bogotá of Indian cities. Monkkonen (2013) focuses on Indonesia, a country (Colombia), between 2008 and 2017. Using annual data for 837,505 with an important informal housing market, with particularly stringent registered lots.2 grouped in 42,993 blocks, we rely on an empirical ∗ Corresponding author at: Department of Applied Economics, Universitat Autònoma de Barcelona, Edifici B, Facultat d’Economia i Empresa, 08193 Cerdanyola del Vallès, Spain. E-mail addresses: diego.buitrago@uab.cat (D. Buitrago-Mora), miquelangel.garcia@uab.cat (M.-À. Garcia-López). URLs: https://diegob901.wixsite.com/diegobuitragomora (D. Buitrago-Mora), https://miquelangelgarcialopez.wordpress.com (M.-À. Garcia-López). 1 We are grateful for comments from the Editor, two anonymous referees, Antonio Avendaño, Jan Brueckner, Alonso Cardenas, David Castells-Quintana, Paula Herrera, Hans Koster, Marta MJ, Joel and Maria MG, Anna Matas, Rosella Nicolini, Luis Quintero, Daniel Rey, Guillermo Salas, Guillermo Sinisterra, Adam Storeygard, Josh van Ommerem, and Elisabet Viladecans, as well as from participants at the European and American Meetings of the Urban Economics Association, the Urban and Regional Economics Workshop, the Simposium of the Spanish Economics Association, and seminars. Financial support from the Ministerio de Ciencia e Innovación of Spain (research projects RTI2018-097401-B-I00, RTI2018-097434-B-I00, PID2021-128227OB-I00 and TED2021-132184B-I00), Ministerio de Ciencia, Tecnología e Innovación of Colombia (convocatoria doctorado en el exterior 860-2019), and COLFUTURO of Colombia (programa credito-beca 2017) is gratefully acknowledged. 2 In this research, we interpret a lot as an area of land with one or more owners that may contain one or several properties 3 The strata system in Colombia is a system of subsidizing public services by regulating their prices. Every block in Colombia’s urban areas, including Bogotá, are assigned to an specific strata level, from 1 to 6, depending on its physical characteristics and surrounding conditions. Strata 1 to 3 receive a subsidy from the higher strata 4 to 6. See Appendix A for a further definition of the strata system in Bogotá and its relationship with Policy-562. https://doi.org/10.1016/j.regsciurbeco.2023.103914 Received 27 November 2020; Received in revised form 18 April 2023; Accepted 24 May 2023 Available online 12 June 2023 0166-0462/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 strategy based on Difference-in-Differences (DiD) techniques to com- exemplifies this. This Law enshrines how to use urban land and grants pare real estate prices before and after the implementation of Policy- cities with more than 100,000 inhabitants the freedom to draft their 562 in treated blocks and in control blocks with similar pre-treatment master zoning plan or Plan de Ordenamiento Territorial (POT). Ac- traits. Besides the average effects, we also explore the heterogeneity of cording to Cámara de Comercio de Bogotá (2018), a POT comprises the effects by year of the treatment subperiod, the main land uses of a set of goals, guidelines, policies, strategies, programs, actions, and the blocks, and the strata where they are located.3 norms aimed at directing and managing the physical development and There are various reasons why Bogotá and its Law of Heights land use in the territory. Thus, the zoning plan constitutes a road map (Policy-562) provide an excellent testing ground of the relationship for the long-term (12-year) development of urban and rural areas to between regulation and real estate prices. First, the new policy aimed consolidate a ‘coherent’ city model. to regulate the conditions for urban renewal not in the whole city, The first POT of Bogotá was approved in 2004 and it classified the but only in some specific areas. As a result, it is possible to identify territory according to three structures: (i) a main ecological structure, treated blocks. Second, the treatment period of this policy is also easy (ii) a functional and services structure, and (iii) a socioeconomic-spatial to identify: It was in force between 2015 and February 2016, but new structure. More specifically, the 2004 POT regulated height limits, floor projects were still approved and executed between March 2016 and area ratio and developer payments that affected all areas of the city December 2017. Third, the Law of Heights increased the degree of land indistinctly. The norm remained in force until December 30th 2021, use regulation in Bogotá because, despite relaxing the height limits for when a new POT (Law 555 of 2021) was approved for the 2022–2035 the new buildings (which required a monetary compensation), the new period. set of rules clearly increased construction costs. Finally, detailed data In December 2014, Bogotá implemented a new policy (562 of 2014) at the lot level is available for the 2008 to 2017 period. regulating the conditions for urban renewal in defined city areas. The In general, this paper furthers our understanding of the effects of policy aimed to promote the improvement, beautification, develop- land use regulations. The related empirical literature shows that they ment and, in particular, densification of some specific parts of the limit city size (Hannah et al., 1993), increase real estate prices (Quigley city with public and private interventions. Unfortunately, there is no and Rosenthal, 2005; Ihlanfeldt, 2007; Huang and Tang, 2012), fol- technical document justifying the selection of the areas (see Fig. 2(a)). low the market (Wallace, 1988; Garcia-López et al., 2015), and, in It seems that they were close to public transportation (Transmilenio) general, affect many other aspects of development (Cheshire and Shep- and main roads, to metropolitan and zonal parks, to facilities (public pard, 2004). Furthermore, regulations seems to negatively affect wel- safety, defense and justice, food supply and consumption, hospitals, fare (Turner et al., 2014). As above mentioned, most of the literature fairgrounds, cemeteries and public administration services), and they has focused on developed countries, and only few recent works has analyzed other countries with inconclusive and, sometimes, opposite were not protected (not developable land). However, it is also true that results. This paper contributes to this literature by providing empiri- other areas satisfied the above mentioned characteristics and were not cal evidence for a particular regulation in a city of a middle-income selected (for example, areas in the south of the city with many illegal developing country. settlements). Our results show that, on average, Policy-562 positively affected To achieve these goals, Policy-562 first removed height limits on real estate prices. In particular, our pure DiD approach reports an new buildings conditional on some payments from the developers. In estimated effect of 33.5% in treated blocks. This result holds when general, these payments in Colombia refer to the amount of area (𝐴) we consider more balanced samples of treated and untreated blocks that developers must give to the city. This land comes from the lots in terms of observables by combining DiD with Propensity Score and to be developed and it is used to satisfy the ‘needs’ of the surrounding Nearest Neighbor Matching techniques. When we follow Brueckner area in terms of public space, road infrastructure, parking lots, front et al. (2017) matched-pair approach to consider balanced samples in gardens, or public services, among others. It is calculated as follow: terms of unobservables, we estimate a Policy-562 effect of 16.4%. 𝐴=𝑃 ×𝐾 Finally, in Appendix D we show that the effect of Policy-562 is het- erogeneous in three dimensions. By year, the effect decreased during where 𝑃 is the total lot area, and 𝐾 is the payment factor. the treatment subperiod. By main land uses of blocks, Policy-562 only Secondly, Policy-562 modified developer payments (𝐴) by updating affected Residential and Services prices. By strata, while Policy-562 the value of 𝐾. Under the 2004 POT, 𝐾 had a unique value of 0.20. increased prices in low strata 1 and 3 and high strata 6 treated blocks, Under Policy-562, the value of 𝐾 depended on the floor area ratio it decreased prices in high strata 4 and 5. (𝐹 𝐴𝑅, the ratio between a building’s total floor area and the total lot The rest of the paper is structured as follows. In Section 2, we area). briefly describe land use regulation in Colombia and in Bogotá, with an Table 1 reports 𝐾 values for different floor area ratio intervals: especial attention to the Law of Heights. In Section 3, we present the The higher the 𝐹 𝐴𝑅, the higher the 𝐾. It also shows that developer city of Bogotá, the dataset to study real estate prices at the block level, payments were lower under Policy-562 when the floor area ratio was and the procedure to identify the blocks (un)affected by Policy-562. bellow 4. On the contrary, Policy-562 payments were higher than 2004 The empirical strategy based on Difference-in-Differences techniques POT ones for higher floor area ratios. is discussed in Section 4. Section 5 presents the main results and Using an example discussed by Ruiz and Moncada (2017), in Table 2 robustness checks, and Section 6 concludes. we compute developer payments under the 2004 POT and Policy-562 for a residential project with 100 m2 apartments in a lot of 8694 2. The law of heights (Policy-562) m2 (138 m × 63 m). First, to build 100 apartments (Columns 1 and 2) a developer would have to give to the city 1739 m2 of the lot Colombia has a national land use regulatory framework that can area under the 2004 POT, but only 52 m2 under Policy-562. Second, be considered strong in the Latin American region.4 Law 388 of 1997 developer payments would be roughly the same with the two policies when building 310 apartments (Columns 3 and 4). Third, to build 4 According to Cabeza (2006), Latin American countries can be classified 433 apartments5 (Columns 5 and 6), developer payments under Policy- according to their level of land use regulation. First, countries with specific 562 would be 120% higher (3817 m2 vs. 1739 m2 ). Finally, if we (centralized) national laws on land use planning (Uruguay, Colombia, Sal- consider the maximum number of floors that could be built according vador, Honduras, and Cuba). Second, countries with several (decentralized and non-coordinated) regional laws (Argentina, Bolivia, Ecuador, Venezuela, and 5 Mexico). Finally, countries without land use regulation laws (Chile, Paraguay, Because of the 2004 POT height limits (10-story buildings), computations Brazil, Panama, Costa Rica, and the Dominican Republic). in Column 5 are hypothetical. 2 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Table 1 3. Data Policy-562 𝐾 values to compute developer payments in Bogotá. Floor area ratio 𝐾 3.1. Bogotá (Colombia) 2.0 < 𝐹 𝐴𝑅 ≤ 2.4 0.006 2.4 < 𝐹 𝐴𝑅 ≤ 2.8 0.035 2.8 < 𝐹 𝐴𝑅 ≤ 3.3 0.092 We study the metropolitan area of Bogotá, with 10,121,956 inhab- 3.3 < 𝐹 𝐴𝑅 ≤ 4.0 0.197 itants in 2021 according to Departamento Administrativo Nacional de 4.0 < 𝐹 𝐴𝑅 ≤ 4.4 0.322 Estadística (DANE) living in 4000 km2 , that is, with roughly 2530 4.4 < 𝐹 𝐴𝑅 ≤ 5.0 0.439 inhabitants per km2 . 5.0 < 𝐹 𝐴𝑅 ≤ 6.5 0.553 Fig. 1 shows the urban and rural areas of metropolitan Bogotá. As 6.5 < 𝐹 𝐴𝑅 ≤ 9.0 0.655 9.0 < 𝐹 𝐴𝑅 ≤ 14 0.757 can be noticed, two-thirds of the city is rural (in green). We focus the 𝐹 𝐴𝑅 > 14 0.833 analysis on the urban areas, which includes 19 municipalities (black lines). After the city, the municipality is the largest level of zoning. For planning purposes, the city is also divided into 108 zonal planning units (ZPU) (red lines) and their 1090 neighborhoods. the 2004 POT (10) and the maximum number of apartments per floor6 according to each policy (31 and 43), developer payments by apartment 3.2. Real estate prices would increase by 58% (from 5.61 (=1739/310) (Column 3) to 8.88 (=3817/430) m2 per apartment (Column 6)). It is important to clarify that, under this policy, developer payments To measure real estate prices, we use the dataset developed by could be also monetary. That is, if the amount of land (𝐴) that was to Secretaría Distrital de Planeamiento (SDP). It is based on annual studies be given to the city was not available in the area (or was less than of the real estate market monitoring the trends in the commercial value 2000 m2 ), the developer could make a monetary payment (based on of properties. Opposed to the traditional cadastral values, these SDP cadastral values) that the city would use for infrastructures and urban values contain real estate market elements such as sales offers, leases amenities in other areas. and financial transactions, and appraisals.8 SPD prices represent the In February 2016, Policy-562 was repealed, among other reasons, commercial reference values (per m2 ) and reflect the dynamics of the because its approval was considered illegal. By that date, 901 projects real estate market.9 were approved, and 2362 applications had been filled while the new The SPD dataset also includes information about the floor area policy was in force. Between March 2016 and December 2017, most (m2 ) and the predominant land use of the lots (residential, manufac- applications were approved and executed. The 2016 Resolution 079 re- turing and services). Unfortunately, no other property characteristic voked Policy-562. The cancellation of the decree meant that Policy-562 (e.g., height) is included in the dataset. had no effect on newly issued construction licenses as of February 22, Our initial sample includes data for 837,505 registered lots in 2017. 2016. However, any license requested prior to February 21, 2016, if au- They represent 88% of registered lots.10 To avoid inconsistencies due to thorized, was governed by Policy-562. Similarly, all projects approved missing values in previous years,11 we fix these 2017 lots for the whole and under construction with Policy-562 continued to adhere to this studied period. By doing so, we avoid inconsistencies due to missing policy even after the repeal declaration and until project completion.7 values in the previous years. Then, we group lot data into blocks and Policy-562 was also important for the city budget. According Sec- we end up with 42,993 blocks. The real estate price at the block level retaria de Hacienda de Bogotá (2015), 200,000 million COP (US$ 50 is then computed as the average price (per m2 ) of the lots that make million) in developer payments were raised in 2015, representing 20% up each block. and 2% of non-tax revenues and total revenues, respectively. Compared to 2004 POT payments between 2005 and 2014, Policy-562 raised 50% 3.3. Areas (un)affected by Policy-562 of them in just 15 months (Cámara de Comercio, 2015). Finally, it is important to mention that Colombia and, in particular, As we explain in more detail in the next sections, we study the Bogotá have an active law enforcement system with a low percentage impact of Policy-562 on real estate prices with a before–after analysis of informality and a reasonable time to approve building permits. On that compares the evolution of prices in treated areas (affected by average, 12,000 building permits are issued every year in Bogotá. Each Policy-562) and untreated areas (unaffected by Policy-562). permit is issued in an average of 50 calendar days. Secretaría Distri- tal de Gobierno (SDG) is responsible of the related law enforcement according to article 135 of the National Police Code. On average, 900 8 The annual appraisal process is carried out by the cadastral unit (Unidad stop-workers orders are issued every year: 62% of them for not having Administrativa Especial de Catastro Distrital, UAECD), an autonomous entity any type of building permit, 30% due to breach obligations related to belonging to the Bogotá finance office and independent from SPD. Appraisals the construction process itself, 7% for allocating a property to a use are processes that reflect the characteristics of homogeneous geographical and other than that authorized in the building permit, and 1% to protect economic zones to determine the current value of properties. New projects and properties of cultural, historical and architectural interest. This scenario development plans only affect these values once the properties are physically differs from other developing countries like Indonesia, with restrictive changed. In other words, SDP prices do not respond to regulatory changes via land registration and building permits (160 days), and inefficient law appraisals that happened at the same time that the norm changed. On the enforcement (Monkkonen, 2013). contrary, SPD prices adjust in the medium and long term. 9 As a robustness check, we compared the SPD dataset with the best available alternative dataset (Coordenada Urbana developed by Cámara 6 This number depends on other requirements of the policies (e.g., the land Colombiana de Construcción CAMACOL), which includes average transaction use index) and explains why the number of floors is different for the two prices at the neighborhood level. Both datasets are highly correlated and a policies in the three studied scenarios in Table 2. simple test for difference of the means shows that they are not statistically 7 The repeal decree literally says: ‘‘... If, during the term that elapses different. Unfortunately, we did not have access to individual transaction between the application for a license or its modification and the issuance of prices. the administrative act that grants the license or authorizes the modification, 10 According to the 2017 cadastral census, there were 2,543,290 properties there is a change in the urban regulations that affect the project submitted in 951,749 registered lots. 11 the applicant will have the right to have the license or modification granted For example, when new lots are added to the city boundaries, or when based on the urban planning regulations in force at the time of the filing of lots are excluded because they are merged due to the construction of new the application, provided that it has been submitted legally and duly ...’’. buildings. 3 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Table 2 Developer payments in a residential project: 2004 POT vs. Policy-562. 200 Apartments 310 Apartments 433 Apartments 2004 POT Policy-562 2004 POT Policy-562 2004 POT Policy-562 [1] [2] [3] [4] [5] [6] Total lot area (m2 ) 8694 8694 8694 8694 8694 8694 Number of floors 6 5 10 7 14 10 Total floor area (𝑃 ) (m2 ) 20,000 31,000 43,000 Floor area ratio (𝐹 𝐴𝑅) 2.30 3.57 4.98 Payment factor (𝐾) 0.20 0.006 0.20 0.197 0.20 0.439 Developer payment (𝐴) (m2 ) 1738.80 52.16 1738.80 1712.72 1738.80 3816.67 Fig. 1. Urban and rural areas in Bogotá. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) The identification of the affected areas of the city is challenging We begin by identifying with zero the blocks located in ZPUs of because, first, this information is not at the same spatial level of municipalities without areas designated under Policy-562. Then, we use aggregation as that of real estate prices (block level), and, second, we a lower level of aggregation, the ZPUs, and assign a value of 1 to blocks do not have a map of the blocks (we only know their municipalities, located in ZPUs with more than 75% of their total area affected by Policy-562. For ZPUs with less than 75% of affected area, we use an ZPUs and neighborhoods). In fact, all we can resort to are documents smaller spatial unit, the neighborhood, and repeated the exercise: We and paper maps of the city in which the areas affected by Policy-562 assign a value of 1 to blocks located in neighborhoods with more than are presented schematically and without precise geographic detail. For 75% of their total area affected by Policy-562. example, Fig. 2(a) is a paper map published by the planning authority At the end of this procedure, we identify 7700 blocks affected by identifying the ZPUs of the city affected by Policy-562 (in yellow), non- Policy-562 (18% of blocks) (the blue areas in Fig. 2(b)) and 35,293 affected (in white), and under special protection (in red). It is important unaffected blocks (the yellow areas in Fig. 2(c)). The former are our to notice that not all blocks that make up each ZPU were affected by (initial) treatment group and the latter our (initial) control group. Policy-562. 4. Empirical strategy To identify whether or not each of the 42,993 defined blocks are affected by the Policy-562, we follow a top-down analysis, i.e. from 4.1. Timing of the analysis the largest level of aggregation to the smallest one, in order to obtain a dummy variable that takes a value of 1 for areas included under Using the SPD dataset, we have information on real estate prices Policy-562 and 0 otherwise. from 2008 to 2017. We split this period into two subperiods. First, 4 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Fig. 2. From a paper map to GIS maps of (un)affected blocks by Policy-562 Notes: In Fig. 2(a), yellow and white zones are ZPUs affected and unaffected by Policy-562, respectively. Red zones are ZPUs under special protection. In Figs. 2(b) and 2(c), blue and yellow zones are blocks in areas affected and unaffected by Policy-562, respectively. In both figures, gray lines are municipality boundaries. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) the treatment subperiod (2015–2017) considers the years in which Policy-562 was in effect (2015 and February 2016) and the years in ln(Price𝑖𝑡 ) = 𝛽0 + 𝛽1 × Policy-562𝑖 × After-562𝑡 which the last projects approved by Policy-562 were developed (March + 𝛽2 × Time-variant controls𝑖𝑡 (1) 2016 and 2017). Second, the subperiod 2008–2014 is the period before treatment. + 𝛽3 × Time-invariant controls𝑖 + 𝜐𝑡 + 𝜖𝑖𝑡 where ln(Price𝑖𝑡 ) is the log of the average property price in block 𝑖 in year 𝑡. 4.2. Estimated equation Policy-562𝑖 is a dummy equal to one if block 𝑖 is affected by the new policy, and zero otherwise. After-562𝑡 is a dummy equal to one We estimate the effect of the Law of Heights (Policy-562) on real if year 𝑡 corresponds to the period of implementation of the Law of estate prices using a Difference-in-Differences (DiD) strategy. In par- Heights (2015–2017), and zero otherwise. We are interested on the DiD ticular, with our 10 year dataset, we estimate the following equation: estimator, that is, on the estimated value of 𝛽1 , the coefficient of the 5 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 interaction between Policy-562𝑖 and After-562𝑡 . It measures the effect based on the average price of the lots that make up each block, we of the new policy in treated vs. untreated (control) areas. weight block-year observations by the number of lots-year. Time-variant controls𝑖𝑡 is a vector of time-variant block and ZPU The estimated coefficient of interest is positive and statistically characteristics. First, we control for the log of the average floor area significant in all columns and decreases when we add control variables (m2 ) in the block. Second, to control for socioeconomic characteristics, and, in particular, when we control for block fixed-effects. Our pre- we add the log of the number of inhabitants per hectare (density) and ferred result is in Column 4, it reports an estimated coefficient of 0.289 the log of population per household. Summary statistics are reported indicating that blocks affected by Policy-562 experienced an increase in in Table B.2 of Appendix B. real estate prices around 33.5% higher than untreated blocks. Time-invariant controls𝑖 is a vector of time-invariant ZPU character- Table B.1 in Appendix B shows that average prices of treated blocks istics. First, we control for time-invariant socioeconomic characteristics increased from 592,000 to 1,942,000 COP/sq.m. between 2014 and with dummy variables for each of the five strata. Second, we add 2017, which represents a total growth of 228.1% in the treatment controls for the accessibility to the city’s main services such as the period. As a result, the Law of Heights explains roughly 15% of this log of km2 of metropolitan parks, the log of km2 of zonal parks, growth. Similarly, if we consider that average prices of untreated blocks the number of health-related private institutions (small and medium), increased by 120.7% (from 241,000 to 533,000 COP/sq.m.), Policy-562 and the number of facilities (public safety, defense and justice, food would explain a third of the difference in growth rates between treated supply and consumption, hospitals, fairgrounds, cemeteries and public and untreated blocks. administration services). These variables are from 2017. In the same Since Bogotá’s real estate market is not perfectly segmented by group, we added the number of Transmilenio stations, the system of block, in Appendix C we consider potential spillover effects when prices Bus Rapid Transit (BRT) responsible for the majority of public transport in one block are affected by prices (or their determinants) in nearby trips in the city.12 blocks. In particular, in Table C.1 we add controls for the log of average Finally, 𝜐𝑡 are year fixed-effects, and 𝜖𝑖𝑡 is an error term with the price per square meter in neighboring blocks located at different dis- usual properties. tances and ZPUs. The estimated coefficient of interest remains positive In our preferred specification we replace the time-invariant controls and statistically significant in all specifications. Furthermore, these with block fixed-effects (𝛼𝑖 ) that fully control for all time invariant results are not statistically different from our preferred specification in differences between blocks: Column 4 of Table 3. ln(Price𝑖𝑡 ) = 𝛽1 × Policy-562𝑖 × After-562𝑡 In Appendix D we investigate the heterogeneity of the above re- (2) sults. First, we study whether the effect of Policy-562 changed over + 𝛽2 × Time-variant controls𝑖𝑡 + 𝜐𝑡 + 𝛼𝑖 + 𝑢𝑖𝑡 time during the treatment period. Results in Column 1 of Table D.1 4.3. On the parallel trends assumption shows that the positive effect of this policy on prices decreased every year (from 2015 to 2017). We relate this decreasing effect with the To use the DiD strategy, we assume parallel trends, which implies political context of Bogotá during these years and, in particular, the that the time effects (𝜐𝑡 ) take account of any time trend in the data that announcement and effective repeal of the Law. is common to both the treatment and control groups (Jones, 2009). The Second, we also explore heterogeneous effects related to the main presence of this common trend prior to the implementation of Policy- land use of the blocks. Results in Column 2 indicates that the Law of 562 means that the behavior of the two groups should be homogeneous Heights only affected Residential and Services treated blocks. On the and independent of the future impact that will affect the treated group. contrary, Manufacturing prices were not significantly affected. Several authors stress the importance of studying this assumption by Finally, we consider the strata where blocks are located. Results comparing the observable characteristics of the treated and control in Column 3 confirm heterogeneous effects of Policy-562 at the strata groups (Zhang, 2017; Givord et al., 2018) which, in this case, means level. While prices in low strata 1 and 3 and high strata 6 treated verifying if there is a systematic difference in the behavior of the real blocks were positively affected, prices in high strata 4 and 5 zones were estate prices prior to the introduction of Policy-562. negatively affected by the Law. Fig. 3 shows the evolution of the average prices in treated and control groups between 2008 and 2017. It shows that, before the Law 5.2. Robustness checks of Heights (2008–2014), real estate prices of the two groups evolved in a similar way and, in fact, they were not statistically different. These Despite the parallel pre-trends reported in Fig. 3, we fear that parallel pre-trends are suggestive evidence in support of the parallel treated and control groups might be different in terms of observables. trends assumption. On the other hand, it is clear that the average prices To alleviate this concern, we consider three alternative methods that of the two groups followed different trends when Policy-562 was in aim to redefine our treated and control groups. First, we apply a place (2015–2017). Propensity Score Matching (PSM) to select treated and controls that are similar in terms of explanatory variables.13 We end up with 34,449 5. Results blocks (80% of the initial sample). The treated and controls groups are made up of 6186 and 28,263 blocks, respectively. Alternatively, we 5.1. Main results consider a Nearest Neighbor Matching (NNM) using the 100-nearest neighbors on all explanatory variables.14 With this method, we select Table 3 reports DiD results when we regress the log of price on a total of 6177 blocks, 3818 treated and 2359 untreated. Finally, we the interacted Policy-562 variable. In Column 1, we follow a pooled follow Brueckner et al. (2017) matched-pair approach and consider strategy and estimate Eq. (1) without control variables. Then, we what we name the Geographical Approach (GA): We focus on the gradually add time-variant (Column 2) and time-invariant (Column 3) control group to select those untreated blocks that are adjacent to controls. Column 4 shows results when we follow a block fixed-effects treated blocks. The idea is that, at this spatial level, adjacent blocks panel strategy and estimate Eq. (2). Since our dependent variable is 13 In Appendix E we provide further details on the method and its 12 We include this variable as time-invariant using most recent values implementation. 14 because there was no new construction of lines or stations between 2013 and Unfortunately, smaller ‘neighborhoods’ do not provide enough number 2020. The last one before such a pause was the enlargement to connect Soacha of observations. On the contrary, bigger ‘neighborhoods’ do not significantly (the neighboring municipality in the south of Bogotá) in 2013. change the number of observations and results hold. 6 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Fig. 3. Evolution of real estate prices in treated and control groups: Mean and S.D. Notes: 7700 treated blocks and 35,293 untreated (control) blocks as described in Section 3.3. Table 3 The effect of Policy-562 (Law of Heights) on real estate prices: DiD main results. [1] [2] [3] [4] Policy-562 × After-562 1.130𝑎 1.168𝑎 0.850𝑎 0.289𝑎 (0.032) (0.033) (0.032) (0.037) Time-variant controls ✓ ✓ ✓ Time-invariant controls ✓ Block fixed-effects ✓ Time fixed-effects ✓ ✓ ✓ ✓ Adjusted 𝑅2 0.100 0.133 0.139 0.217 Notes: 429,930 observations (= 42,993 blocks × 10 years) in each regression. Regressions are weighted by the number of lots that make up each block. Robust standard errors are clustered by ZPU and are in parenthesis. The coefficient of interest remains significant when clustering at the neighborhood and block levels. 𝑎 , 𝑏 , and 𝑐 indicates significant at 1, 5, and 10 percent level, respectively. may only differ on the treatment. In this case, we end up with a total Table 4 The effect of Policy-562 (Law of Heights) on real estate prices: Alternative methods. of 13,546 blocks, that is, the original 7700 treated blocks and 5846 PSM + DiD NNM + DiD GA + DiD untreated blocks (16.7% of the initial untreated sample). [1] [2] [3] Table 4 reports results when we combine the DiD approach with Policy-562 × After-562 0.296𝑎 0.324𝑎 0.152𝑎 the PSM (Column 1), the NNM (Column 2) and the GA (Column (0.040) (0.083) (0.050) 3). As previously, the estimated coefficient of interest is positive and Adjusted 𝑅2 0.214 0.207 0.203 statistically significant in all three alternative approaches. Observations 344,490 61,770 135,460 Regarding the magnitude of the estimated coefficients, the PSM and Notes: Regressions include time-variant controls, block fixed-effects, and year fixed- NNM ones (0.296 and 0.324) are statistically similar to their pure DiD effects. They are also weighted by the number of lots that make up each block. Robust counterpart (0.289) in Column 4 of Table 3. They show that Policy-562 standard errors are clustered by ZPU and are in parenthesis. 𝑎 , 𝑏 , and 𝑐 indicates increased prices by 34.5% and 38.3%, respectively. significant at 1, 5, and 10 percent level, respectively. On the other hand, the GA estimated coefficient (0.152) is statisti- cally smaller and differs by a factor of 2 with the pure DiD estimated coefficient (0.289) in Column 4 of Table 3. This GA result indicates that In Table 5 we explore the sensitivity of the results to the chosen Policy-562 (only) caused a 16.4% growth in real estate prices in treated cutoff. First, we consider an smaller cutoff of 25% in Column 1 and a blocks.15 more demanding cutoff of 100% in Column 2. Using these alternative We may also fear that the cutoff used in the definition of blocks thresholds, the number of treated blocks increases from 7700 to 10,488 affected by Policy-562 is somehow arbitrary. As we explain in detail in (25% threshold) and decreases to 3075 (100% threshold). The results of Section 3, treated blocks are those located in ZPUs with more than 75% estimating Eq. (2) confirm the positive and significant effect of Policy- of their total area affected by Policy-562. For ZPUs with less than 75% 562 for the two thresholds. Furthermore, when comparing with the of affected area, we apply this threshold to each of their neighborhoods. result counterpart in Table 3 Column 4 (75% threshold), it is clear that the estimated positive effect increases the higher the threshold. Second, in Column 3 we consider a multilevel treatment by simulta- 15 In some additional robustness checks that are available upon request, we neously using different threshold intervals: Blocks with 25% to less than apply the geographical approach (GA) to the PSM and the NNM samples. In 75% of affected area, blocks with 75% to less than 100% of affected both cases, results hold with significant and smaller estimated coefficients. area, and blocks with 100% of affected area. The omitted category 7 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Table 5 The effect of Policy-562 (Law of Heights) on real estate prices: Alternative measures. Thresholds Intervals Continuous 25%≥ 100% [1] [2] [3] [4] Policy-562 × After-562 0.234𝑎 0.538𝑎 (0.033) (0.050) 25%–75% Policy-562 × After-562 0.060𝑎 (0.010) 75%–100% Policy-562 × After-562 0.119𝑎 (0.008) 100% Policy-562 × After-562 0.560𝑎 (0.009) Continuous Policy-562 × After-562 0.520𝑎 (0.044) Adjusted 𝑅2 0.217 0.217 0.217 0.217 Notes: 429,930 observations (= 42,993 blocks × 10 years) in each regression. Regressions include time-variant controls, block fixed-effects, and year fixed-effects. They are also weighted by the number of lots that make up each block. Robust standard errors are clustered by ZPU and are in parenthesis. 𝑎 , 𝑏 , and 𝑐 indicates significant at 1, 5, and 10 percent level, respectively. refers to blocks with less than 25% of affected area. The estimated A qualifier is important here. As we previously acknowledge, our coefficients confirm the positive effect of Policy-562, which is more research faces an unsolved endogeneity problem related to unobserved important for the most affected blocks (100%). variables and potential measurement errors. As a result, the magnitudes Finally, in Column 4 we consider a continuous treatment variable of the estimated coefficients in our preferred specification should be by directly using the percentage of affected area (instead of a dummy). read with caution. The significant and positive estimated coefficient indicates that each additional 1 p.p. in the percentage of affected area, increased real estate CRediT authorship contribution statement prices by 0.52%. In other words, blocks with a 100% of affected area experienced a 52% increase in their real estate prices. Diego Buitrago-Mora: Concept, Design, Analysis, Writing, Revi- Overall, these alternative threshold results confirm results when sion. Miquel-Àngel Garcia-López: Concept, Design, Analysis, Writing, using the 75% threshold. Revision. In summary, results in Tables 3, 4 and, 5 confirm that Policy-562 affected real estate prices in Bogotá. In particular, while the pure DiD Declaration of competing interest specification in Column 4 of Table 3 shows that prices increased 33.5% more in treated blocks, the GA specification in Column 3 of Table 4 The authors declare that they have no known competing finan- reports an effect of 16.4%. cial interests or personal relationships that could have appeared to A qualifier is important here. There are some identification issues influence the work reported in this paper. that might affect the magnitude of the estimated coefficient. In this sense, our research faces an endogeneity problem. First, we are worried that some unobserved variable determines both real estate prices and Data availability Policy-562. The DiD, the PSM, and the NNM approaches are elaborate ways of comparing blocks that are similar on observable quantities. The data that has been used is confidential. By comparing near neighbors, the GA approach may do better at controlling for unobservables. Second, as shown in Table 5, we also Appendix A. Supplementary data face a measurement error in our measure of Policy-562. Therefore, the magnitude of the positive effect estimated in the DiD, the PSM, the Supplementary material related to this article can be found online NNM and the GA specifications should be read with caution. at https://doi.org/10.1016/j.regsciurbeco.2023.103914. 6. Conclusions References In this paper, we investigate the effect of the Law of Heights (Policy- Brueckner, Jan K., Fu, Shihe, Gu, Yizhen, Zhang, Junfu, 2017. Measuring the stringency 562) on real estate prices in Bogotá between 2008 and 2017. Our of land use regulation: The case of China’s building height limits. Rev. Econ. Stat. results show that treated blocks experienced an increase in real estate 99 (4), 663–677. http://dx.doi.org/10.1162/REST_a_00650. Brueckner, Jan K., Sridhar, Kala Seetharam, 2012. Measuring welfare gains from prices. On average, the effect of Policy-562 ranges between 16.4% relaxation of land-use restrictions: The case of India’s building-height limits. Reg. (GA approach) and 33.5% (pure DiD approach). This effect is also Sci. Urban Econ. 42 (6), 1061–1067. http://dx.doi.org/10.1016/j.regsciurbeco. heterogeneous from a temporal, land use and strata point of view: It 2012.08.003. decreases in time, it is only related to Residential and Services land Cabeza, Ángel Massiris, 2006. Políticas latinoamericanas de ordenamiento terri- uses, and it is positive in low strata 1 and 3 and high strata 6 and torial. Universidad Nacional de Colombia, Revista Maestría En Ordenamiento Urbano-Regional, p. 19. negative in high strata 4 and 5. Cámara de Comercio de Bogotá, 2018. Plan de ordenamiento territorial. URL http: We think that the contributions made by this paper are relevant. //www.ccb.org.co. First, it provides empirical evidence for a city (Bogotá) in a middle- Cheshire, Paul, Sheppard, Stephen, 2004. Land markets and land market regulation: income developing country (Colombia) and shows that, similar to Progress towards understanding. Reg. Sci. Urban Econ. 34 (6), 619–637. http: developed countries, a greater degree of regulation increases real estate //dx.doi.org/10.1016/j.regsciurbeco.2004.03.001. Cámara de Comercio, Bogotá, 2015. Observatorio de la gestión urbana en Bogotá. p. prices. Second, while most papers focus on the average effects of the 4, URL https://bibliotecadigital.ccb.org.co/handle/11520/14094. regulation, this research also provides empirical evidence on its hetero- Freemark, Yonah, 2020. Upzoning chicago: Impacts of a zoning reform on property geneous effects. In particular, the paper furthers our understanding of values and housing construction. Urban Aff. Rev. 56 (3), 758–789. http://dx.doi. how regulation affects different land uses and income groups. org/10.1177/1078087418824672. 8 D. Buitrago-Mora and M.-À. Garcia-López Regional Science and Urban Economics 101 (2023) 103914 Garcia-López, Miquel-Àngel, Solé-Ollé, Albert, Viladecans-Marsal, Elisabet, 2015. Does Mayo, S.K., Sheppard, Stephen, 1996. Housing supply under rapid economic growth zoning follow highways? Reg. Sci. Urban Econ. 53, 148–155. http://dx.doi.org/10. and varying regulatory stringency: An international comparison. J. Hous. Econ. 5, 1016/j.regsciurbeco.2015.05.008. 274–289. Givord, Pauline, Quantin, Simon, Trevien, Corentin, 2018. A long-term evaluation of Monkkonen, Paavo, 2013. Urban land-use regulations and housing markets in devel- the first generation of French urban enterprise zones. J. Urban Econ. 105, 149–161. oping countries: Evidence from Indonesia on the importance of enforcement. Land http://dx.doi.org/10.1016/j.jue.2017.09.004. Use Policy 34, 255–264. http://dx.doi.org/10.1016/j.landusepol.2013.03.015. Greenaway-McGrevy, Ryan, Pacheco, Gail, Kade, Sorensen, 2021. The effect of upzoning Monkkonen, Paavo, Ronconi, Lucas, 2013. Land use regulations, compliance and land on house prices and redevelopment premiums in Auckland, New Zealand. Urban markets in Argentina. Urban Stud. 50 (10), 1951–1969. http://dx.doi.org/10.1177/ Stud. 58 (5), 959–976. http://dx.doi.org/10.1177/0042098020940602. 0042098012471982. Hannah, Lawrence, Kim, Kyung-Hwan, Mills, Edwin S., 1993. Land use controls and Quigley, John M., Rosenthal, Larry A., 2005. The effects of land use regulation on the housing prices in Korea. Urban Stud. 30 (1), 147–156. price of housing: What do we know? what can we learn? Cityscape 8 (1), 69–137. Huang, Haifang, Tang, Yao, 2012. Residential land use regulation and the US housing Ruiz, Geraldine, Moncada, Ivan, 2017. Análisis multitemporal de la variación del price cycle between 2000 and 2009. J. Urban Econ. 71 (1), 93–99. http://dx.doi. valor del suelo en bogotá durante el periodo 2013–2016, teniendo en cuenta org/10.1016/j.jue.2011.08.001. los cambios de normativa urbana en el plan de ordenamiento territorial de la Ihlanfeldt, Keith R., 2007. The effect of land use regulation on housing and land prices. ciudad (decretos 190 de 2004, 364 de 2013, 562 de 2014, 575 de 2015 y 079 J. Urban Econ. 61 (3), 420–435. http://dx.doi.org/10.1016/j.jue.2006.09.003. de 2016). URL https://repository.udistrital.edu.co/bitstream/handle/11349/5151/ Jones, Andrew M., 2009. Panel data methods and applications to health economics. RuizPalomoGeraldineJohanna2017.pdf?sequence=1. In: Mills, Terence C., Patterson, Kerry (Eds.), Palgrave Handbook of Economet- Secretaria de Hacienda de Bogotá, 2015. Estados financieros. URL https://www.shd. rics. Palgrave Macmillan UK, London, pp. 557–631. http://dx.doi.org/10.1057/ gov.co/shd/estados-contables-basicos. 9780230244405_12. Turner, Matthew A., Haughwout, Andrew, van der Klaauw, Wilbert, 2014. Land use reg- Kok, Nils, Monkkonen, Paavo, Quigley, John M., 2014. Land use regulations and the ulation and welfare. Econometrica 82 (4), 1341–1403. http://dx.doi.org/10.3982/ value of land and housing: An intra-metropolitan analysis. J. Urban Econ. 81, ECTA9823, arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.3982/ECTA9823. 136–148. http://dx.doi.org/10.1016/j.jue.2014.03.004. Wallace, Nancy E., 1988. The market effects of zoning undeveloped land: Does zoning Ling, L., Dao-lin, Z., Ke-lin, H., 2013. The effect of real estate regulation policy on follow the market? J. Urban Econ. 23 (3), 307–326. http://dx.doi.org/10.1016/ house price in Beijing city. In: 2013 International Conference on Management 0094-1190(88)90021-6. Science and Engineering 20th Annual Conference Proceedings. pp. 2078–2086. Zhang, Junsen, 2017. The evolution of China’s one-child policy and its effects on family http://dx.doi.org/10.1109/ICMSE.2013.6586551. outcomes. J. Econ. Perspect. 31 (1), 141–160. http://dx.doi.org/10.1257/jep.31.1. 141. 9