Faster, taller, better: Transit improvements and land use policies

Chen, Liming; Hasan, Rana; Jiang, Yi; Parkhomenko, Andrii Working Paper Faster, taller, better: Transit improvements and land use policies ADB Economics Working Paper Series, No. 702 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Chen, Liming; Hasan, Rana; Jiang, Yi; Parkhomenko, Andrii (2023) : Faster, taller, better: Transit improvements and land use policies, ADB Economics Working Paper Series, No. 702, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS230480-2 This Version is available at: https://hdl.handle.net/10419/298148 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org ADB ECONOMICS WORKING PAPER SERIES NO. 702 November 2023 Faster, Taller, Better Transit Improvements and Land Use Policies This paper examines how the effects of transit improvements vary with land use policies using the case of Bengaluru, one of India’s largest cities with a metro system and subject to low floor-area ratios. We build a quantitative spatial model and show that the metro system increases citywide output and welfare, even net of costs. However, the net gains are several times larger when floor-area ratio limits are relaxed near metro stations (transit-oriented development) or in the city center. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. FASTER, TALLER, BETTER TRANSIT IMPROVEMENTS AND LAND USE POLICIES Liming Chen, Rana Hasan, Yi Jiang, and Andrii Parkhomenko ASIAN DEVELOPMENT BANK The ADB Economics Working Paper Series presents research in progress to elicit comments and encourage debate on development issues in Asia and the Pacific. The views expressed are those of the authors and do not necessarily reflect the views and policies of ADB or its Board of Governors or the governments they represent. ADB Economics Working Paper Series Liming Chen, Rana Hasan, Yi Jiang, and Andrii Parkhomenko No. 702 | November 2023 Liming Chen ([email protected]) is an urban economist in the Sectors Group. Rana Hasan ([email protected]) is a regional lead economist and Yi Jiang ([email protected]g) is a principal economist at the Economic Research and Development Impact Department, Asian Development Bank. Andrii Parkhomenko (parkhomenko[email protected]) is an assistant professor at the Department of Finance and Business Economics at the University of Southern California. Faster, Taller, Better: Transit Improvements and Land Use Policies Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2023 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 8632 4444; Fax +63 2 8636 2444 www.adb.org Some rights reserved. Published in 2023. ISSN 2313-6537 (print), 2313-6545 (electronic) Publication Stock No. WPS230480-2 DOI: http://dx.doi.org/10.22617/WPS230480-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” inthis publication, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This publication is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. Notes: In this publication, “$” refers to United States dollars. ADB recognizes “China” as the People’s Republic of China, “U.S.” as the United States, and “Bangalore” as Bengaluru. ABSTRACT We study the interaction between transit improvements and land use policies in the context of Bengaluru, one of India’s largest cities. The city inaugurated a metro system in 2011. Yet it has low building heights even near metro stations, reflecting low floor-area ratio limits. We construct a rich dataset that includes information on travel times between 198 wards, parcel-level land use, and building heights from satellite images. We then build a quantitative spatial model where heterogeneous workers choose among different commuting modes. The simulations show that the metro system increases citywide output and welfare, even net of costs. However, the net gains are several times larger when floor-area ratio limits are relaxed near metro stations (transitoriented development) or in the city center. Moreover, the metro and transit-oriented development are complementary—their joint effect on incomes, prices, and welfare is greater than the combined effect of the two policies implemented separately. Keywords: urban, transit, land use, building heights, transportation, transit-oriented development, spatial equilibrium, development, India JEL codes: R31, R33, R41, R42, R52 ________________________ We thank Glenita Amoranto and Marjorie Villanueva Remolador for excellent research assistance, Xin Huang and his team from Wuhan University for providing building height data, and Ming Luo from Sun Yatsen University for digitizing land use information. We are also grateful to Gabriel Ahlfeldt, Radha Chanchani, Lucas Conwell, Saugata Dasgupta, Gilles Duranton, Joseph Mariasingham, Partha Mukhopadhyay, Sharad Saxena, as well as seminar and conference participants at the ADB Economists’ Forum, USC, ADB, UEA European Meeting in Milan, and Conference on Infrastructure and Urban Development in the Developing World in Tokyo for helpful comments and suggestions. This study was partly funded by the Japan Fund for Prosperous and Resilient Asia and the Pacific financed by the Government of Japan through the Asian Development Bank. Corresponding author: Andrii Parkhomenko ([email protected]) 1 Introduction In recent decades, rapid urbanization and accompanying congestion have increased the demand for more efficient mass transit in developing countries. The introduction of metro rail systems is often seen as a key part of the solution. In India, about 15 cities currently have active metro rail systems and many more are under construction or in planning. Given the high cost of metro systems, it is important to understand how effectively they improve within-city mobility and provide wider economic benefits. At the same time, Indian cities impose strict controls on building heights through limits on floor-to-area ratios (FARs). A comparison of FARs in central business districts of major cities around the world shows that some of the lowest FARs are in India—between 1 and 1.5 for Chennai, Delhi and Mumbai compared to over 10 for cities such as Chicago, Los Angeles, New York, Tokyo, and Singapore (World Bank 2013). Analysis of building heights around the world by Barr and Jedwab (2023) indicates that large Indian cities have very few tall buildings. While the effects of transit improvements and land use policies have been studied extensively, little is known about their interaction. The goal of this paper is to examine how the benefits of transit depend on existing land use regulations, as well as how these regulations can be redesigned in order to maximize the benefits of transit. In this paper, we study the interaction of transit improvements and land use policies in the context of Bengaluru, one of India’s largest cities.1We build a quantitative spatial model, where lowand high-skilled workers commute between residence and workplace and choose between different modes of transportation. The Bengaluru metro was inaugurated in late 2011. It currently has 52 stations and is being expanded. We first calibrate the model to pre-metro Bengaluru at the ward level using a rich dataset we constructed. Our data includes unique building height estimates from multi-view remote-sensing images, land use information at parcel level from digitized land use maps for existing and future land use plans, commuting time data for multiple modes of transportation from Google Maps, and travel behavior data from a travel survey. We complement these data with ward-level data on employment, residents, and wages. We then simulate the model by examining the introduction of a metro network in Bengaluru and land use reforms that would allow for higher density in central areas of the city or near metro stations. 1The city of Bengaluru is also known as Bangalore. The city’s official name was changed from Bangalore to Bengaluru in 2006. 2 We find that the introduction of the metro increases welfare by about 1.3% and that these gains do not differ between lowand high-skilled workers. At the same time, our empirical findings suggest that both residential and commercial development is severely restricted in the central areas of the city: the height gradient decreases gradually as it moves away from the city center, with a sharp decline at the boundary of central wards. To understand how the welfare effects of the metro depend on these density constraints, we run counterfactual experiments in which the introduction of the metro is accompanied by a relaxation of FAR limits in wards with metro stations (“transit-oriented development” or TOD) or in central wards of the city (“central upzoning”). Even a modest relaxation of FAR limits on residential and commercial development boosts welfare gains from the metro to 3.4% in the case of TOD and to 3.2% in the case of central upzoning. We also evaluate the proposed 2031 Land Use Plan that called for restrictions on commercial development in the city center. We show that the lack of coordination between transportation improvements and land use changes can be detrimental to the city’s economy. Our back-of-the-envelope calculations suggest that, while the metro increases Bengaluru’s output by 6% or $1.7 billion per year, these gains are mostly due to endogenous migration into the city. On a per-worker basis, the gains are almost entirely offset by the costs of building and maintaining the network. Only when metro construction is complemented by a relaxation of FAR limits, do the gains significantly exceed the costs. What makes TOD materially different from central upzoning is that the former policy is complementary with the metro: their joint effect on incomes, prices, and welfare is larger than the combined effect of these two policies when implemented separately. This is because increases in productivity and floorspace supply are positively correlated with improvements in transport connectivity. The complementarity between TOD and the metro unlocks additional gains equivalent to about $64 million or one-half of annual operating costs of the metro system. Furthermore, we demonstrate that modeling FAR limits is quantitatively important. Our findings suggest that most previous work that evaluated the benefits of transit improvements may overstate their welfare gains by not taking into account density limits that are commonplace in many cities around the world. We also show that our results are robust to various modifications in parameters and model assumptions. The results of this paper highlight the importance of jointly designing transportation and land use policies. The fact that important elements of India’s urban planning norms can limit the economic gains from investments in transport infrastructure has not been 3 lost on reform-oriented policymakers. Indeed, India’s central government has taken several initiatives to induce state governments to improve urban planning and align it more closely with transport infrastructure investments. Among the most recent initiatives is a National Transit Oriented Development Policy, formulated in 2017, that provides states with guidelines and incentives to encourage densification along urban mass transit corridors. The extent to which India’s cities will apply these recommendations remains to be seen, and in this context, research that evaluates alternative urban planning norms is highly relevant for policymakers. The main contribution of our paper is to bridge the gap between two distinct strands of literature. First, it relates to the literature that evaluates the welfare impact of transit infrastructure (Heblich et al. 2020;Balboni et al. 2020;Severen 2021;Warnes 2021;Zárate 2022;Tsivanidis 2023;Velásquez 2023;Conwell 2023;Kreindler et al. 2023). Many of these papers focus on developing countries. For instance, Tsivanidis (2023) examines how the bus rapid transit (BRT) system in Bogotá affects welfare; Zárate (2022) looks at how a new subway line in Mexico City lowers barriers to accessing formal employment; while Balboni et al. (2020) study the differential effects of the BRT system in Dar es Salaam on lowand high-income residents. Our paper contributes to this literature by focusing on how the interplay of land use policies and metro investments can amplify or dampen the welfare benefits of new transit infrastructure. Second, it relates to the literature that evaluates the costs of restrictive land use regulations (Allen et al. 2016;Herkenhoff et al. 2018;Hsieh and Moretti 2019;Anagol et al. 2021;Song 2022;Acosta 2022;Yu 2022;Duranton and Puga 2022;Parkhomenko 2023). Most of this literature studies developed countries and does not explicitly take into account transportation networks. Moreover, it primarily focuses on the effects of land use restrictions on housing supply. Our paper adds to this literature by examining how land use regulations may not only constrain housing supply but also undermine the effectiveness of transportation improvements. This paper also contributes to the literature on the effect of land use management on city form and labor market outcomes (Bertaud and Brueckner 2005;Geshkov and DeSalvo 2012;Brueckner and Sridhar 2012). For example, Bertaud and Brueckner (2005) use a monocentric city model to show that height limits cause spatial expansion of cities, which in turn results in welfare losses due to higher commuting costs. Brueckner and Sridhar (2012) provide supporting empirical evidence by examining cities in India. The remainder of the paper is organized as follows. Section 2describes the city of Bengaluru and the data. Section 3examines the key patterns in the building heights and land use use data, and how they conform with the relationships highlighted by the theory. 4 Section 4describes the theoretical framework. Section 5describes the methodology used to quantify the model. Section 6analyzes the effects of introducing the metro and changing land use policies. Section 7concludes. 2 Setting and Data In this section, we describe the city of Bengaluru, its metro network, and FAR policies. We also discuss the data used to document various facts about building heights and land use, and to construct the quantitative model. 2.1 The City of Bengaluru Bengaluru, the capital of the Indian state of Karnataka, has experienced decades of rapid economic growth as a leading center of information technology and high-tech industries. The city has expanded dramatically, growing from 160 square kilometers in 1991 to 741 square kilometers in 2011. Its population was 8.4 million in 2011, more than double the population twenty years earlier, and it is one of the densest cities in the world, with 11,876 people per square kilometer. Private vehicle ownership increased from 58 to 503 per 1,000 inhabitants between 1981 and 2013. However, public transportation has not kept up with the pace of urbanization, and restrictive zoning regulations have contributed to urban sprawl, making Bengaluru one of the most congested cities in the world (ADB 2019; Akbar et al. 2023a,b). Bengaluru Metro. The Bengaluru Metro is an important initiative to address increasing traffic congestion. It is being built and operated by Bangalore Metro Rail Corporation Limited (BMRCL), a joint venture between the Government of India and the Government of Karnataka. The Bengaluru Metro consists of three phases with a total length of 170.4 km and 125 stations. Figure 1shows the map of the existing and planned lines of the metro. Phase 1 (43.3 km) was completed and fully operational in 2017. It comprises the East–West and the North–South lines. Phase 2 (74.1 km) extends the existing lines and adds two new lines, and is expected to be fully operational by 2024. Phases 2A and 2B (51.5 km) are planned and will add two more lines to the network. Construction costs for all phases are estimated at $7.2 billion in 2020 prices, with annual operating costs estimated at $0.7 million per km. Appendix Figure C.1 11 Table 2: Statistics for Land Use, Building Heights, and FARs (1) (2) (3) (4) (5) (6) Variable N Mean S.D. Min Median Max Panel A: Land use Total ward area (hectares) 198 359 486 32 167 2,926 2015 land use Total land use area (hectares) 198 222 252 27 127 1,851 Share by land use type Residential 198 0.53 0.19 0.05 0.56 0.86 Commercial 198 0.19 0.14 0.01 0.14 0.78 Share by land use type for wards with metro station(s) Residential 70 0.47 0.18 0.11 0.47 0.83 Commercial 70 0.19 0.14 0.02 0.15 0.64 Share by land use type among parcels within 500 meters of the metro station Residential 112 0.41 0.22 0.00 0.42 0.79 Commercial 112 0.29 0.20 0.02 0.22 0.87 2031 plan Total land use area (hectares) 198 320 429 33 145 2,627 Share by land use type Residential 198 0.55 0.19 0.00 0.59 0.93 Commercial 198 0.10 0.14 0.00 0.04 0.94 Panel B: Area, height, and distance to city center of parcel Residential Area (square meters) 115,650 1,602 5,075 0 377 429,423 Mean building height (meters) 115,650 2.6 1.3 0.0 2.6 43.6 Max building height (meters) 115,650 4.3 2.6 0.0 4.1 82.2 Distance to city center (meters) 115,650 7,576 4,454 226 7,281 21,929 Commercial Area (square meters) 43,720 1,521 9,531 0 399 1,083,295 Mean building height (meters) 43,720 2.6 1.1 0.0 2.6 11.7 Max building height (meters) 43,720 4.4 2.3 0.0 4.2 69.1 Distance to city center (meters) 43,720 7,589 3,875 203 7,221 21,858 Panel C: Ward-level FAR Residential 198 1.05 0.16 0.52 1.04 1.54 Commercial 198 0.94 0.23 0.22 0.97 1.86 FAR = floor-to-area ratio, N = number of observations, S.D. = standard deviation. Note: This table presents summary statistics for variables that describe land use, building heights, and FARs. Source: Authors’ estimates. minimum height required by the Indian National Building Code (2005). Panel C shows that the ward-level FAR ranges from 0.52 to 1.54 for residential land and from 0.22 to 1.86 for commercial land, lower than the FAR limits in the 2015 plan. This indicates that 12 although the city is one of the densest in the world, it is dominated by low-rise development. In the counterfactual exercises in Section 6, we consider the implications of increasing ward-level FAR in the vicinity of the metro network or in the city center. 3 Building Heights and Land Use In this section, we discuss empirical evidence on the building height gradient, transitoriented development, and compare the 2015 and the 2031 land use plans. 3.1 Building Height Gradient Bertaud and Brueckner (2005) built a monocentric city model in which the floor area per unit of land is subject to an upper limit. In equilibrium, the restriction is binding in the central part of the city and nonbinding elsewhere. Figure 3illustrates the building height patterns of the model with unrestricted and restricted FAR, respectively. In the absence of FAR limits, building heights decline smoothly from the center to the edge of the city. In contrast, with FAR limits, building heights are not only considerably lower in the city center, but their height gradient is also relatively flat as one moves away from the center. The height starts to decline at a certain distance to the center. We examine the actual building height patterns in Bengaluru in two ways, using the data at land parcel level. First, we run an OLS regression of the log of the parcel’s mean and maximum heights on the parcel’s distance to the city center, the squared distance, and the interactions of the two distance variables with an indicator of whether the parcel is outside the ORR, while controlling for land area. The results (Panel A in Table 3) show that for the different types of land use, both the parcel’s mean and maximum heights increase with parcel area, with the coefficients significantly higher for the maximum height. Conditional on land area, the mean and maximum heights decrease linearly for commercial land and in a convex way for residential land as the distance to the city center increases. Our results show that the commercial height gradient is lower than the residential gradient, which is consistent with the prediction of Ahlfeldt and Barr (2022).12 Moreover, the coefficients of the interaction between distance to the center and outside-ORR indicator are negative and statistically significant. This indicates that mean 12Building a general equilibrium monocentric city model with endogenous land use, they show that in the absence of height limits commercial development has a steeper density gradient than residential development. 13 Figure 3: Height Gradient with Unrestricted and Restricted FAR in a Monocentric City FAR = floor-to-area ratio. Note: This figure is adapted Figure 1 from Bertaud and Brueckner (2005). and maximum heights decrease considerably faster outside the ORR than within the ORR. Second, we estimate a locally weighted regression of the log of the mean or maximum height of a parcel on its distance to the city center. To control for land parcel area, we first residualize the two variables by regressing them on the log of parcel area. The results are plotted in Figure 4. For residential and commercial land, the heights gradually decrease as one moves away from the city center, and then sharply after hitting a certain distance from the center. Simple calculations suggest that the turning points fall between 5 and 7 kilometers from the city center, which is in the vicinity of the ORR alignment. These findings are consistent with the theoretical predictions in Bertaud and Brueckner (2005). They imply that FAR restrictions exist, and are likely to be binding within the ORR and nonbinding outside. 3.2 Transit-Oriented Development Our data could also be used to examine another land use policy, transit-oriented development (TOD). TOD has emerged as a paradigm for more sustainable, livable urban planning by integrating land use and public transport. The Government of India has advocated TOD in the National Urban Transport Policy (2014), the National TOD Policy (2017), and the Metro Policy (2017), and called for cities to raise maximum permitted FARs around transit stations. However, progress on TOD projects has been slow due to various constraints, such as legacy urban development issues, policy and 14 Table 3: Regressions of Building Heights on Parcel Characteristics by Land Use Residential Commercial Mean log Max log Mean log Max log height height height height (1) (2) (3) (4) Panel A: Parcel-level height on parcel area and location variables Area (log) 0.042*** 0.121*** 0.015*** 0.143*** (0.001) (0.001) (0.002) (0.002) Distance to city center -0.012*** 0.003 -0.002 -0.001 (0.004) (0.003) (0.005) (0.005) Squared distance to city center -0.002*** -0.002*** -0.001** -0.001** (0.000) (0.000) (0.000) (0.000) Distance to center x Outside ORR -0.039*** -0.040*** -0.021*** -0.024*** (0.003) (0.002) (0.004) (0.003) Squared distance to center x Outside ORR 0.002*** 0.002*** -0.000 0.001* (0.000) (0.000) (0.000) (0.000) Constant 0.876*** 0.822*** 0.928*** 0.641*** (0.009) (0.008) (0.018) (0.015) Observations 115,370 115,504 43,668 43,700 R-squared 0.114 0.195 0.093 0.191 Panel B: Parcel-level height on proximity to metro station Area (log) 0.044*** 0.123*** 0.016*** 0.143*** (0.001) (0.001) (0.002) (0.002) Distance to city center -0.049*** -0.036*** -0.018*** -0.022*** (0.002) (0.002) (0.003) (0.002) Squared distance to city center -0.000*** -0.000*** -0.002*** -0.001*** (0.000) (0.000) (0.000) (0.000) Within 500 meters from metro 0.060*** 0.052*** -0.029*** -0.003 (0.005) (0.005) (0.008) (0.006) Constant 0.942*** 0.894*** 0.995*** 0.712*** (0.007) (0.006) (0.015) (0.013) Observations 115,370 115,504 43,668 43,700 R-squared 0.109 0.189 0.084 0.182 Panel C: Parcel-level height on the host ward with metro station Area (log) 0.044*** 0.123*** 0.016*** 0.143*** (0.001) (0.001) (0.002) (0.002) Distance to city center -0.052*** -0.040*** -0.015*** -0.021*** (0.002) (0.002) (0.003) (0.002) Squared distance to city center -0.000*** -0.000** -0.002*** -0.001*** (0.000) (0.000) (0.000) (0.000) In a ward with a metro station 0.003 0.039*** -0.060*** -0.001 (0.004) (0.004) (0.006) (0.005) Constant 0.970*** 0.901*** 0.994*** 0.711*** (0.006) (0.006) (0.015) (0.013) Observations 115,370 115,504 43,668 43,700 R-squared 0.107 0.189 0.086 0.182 ORR = outer ring road. Note: Standard errors are in parentheses; *** p<0.01, ** p<0.05, * p<0.1. Source: Authors’ estimates. 15 Figure 4: Height Gradients of Bengaluru (Residualized) Panel (a): Residential −1.5 −1 −.5 0 .5 Log Mean Height (Residualized) −4 −2 0 2 Log Distance to City Center (Residualized) Residential, Mean Height, Residual −1 −.5 0 .5 Log Max Height (Residualized) −4 −2 0 2 Log Distance to City Center (Residualized) Residential, Max Height, Residual Panel (b): Commercial −1 −.5 0 .5 Log Mean Height (Residualized) −4 −2 0 2 Log Distance to City Center (Residualized) Commercial, Mean Height, Residual −.8 −.6 −.4 −.2 0 .2 Log Max Height (Residualized) −4 −2 0 2 Log Distance to City Center (Residualized) Commercial, Max Height, Residual Note: The figures show lowess regression results using residualized log of height on residualized log of distance to city center after their respective regressions on log area. Source: Authors’ estimates. regulation, involvement of multiple agencies in planning and development of urban infrastructure, land acquisition issues, and insufficient financing (Ramulu et al. 2021). Among the indicators reflecting the application of the TOD principles, functional diversification and densification of areas surrounding the transport nodes are considered among the most important. Our data on land use and building heights allow us to undertake a quantitative assessment of the status quo of the two indicators in Bengaluru. In panel A of Table 2, we present the shares of land area for residential, commercial, and industrial use for 70 wards with existing or planned metro station(s), and for areas within 500 meters around the 112 metro stations, respectively. Compared to citywide 16 land use allocations, the wards with metro stations have a smaller land share devoted to residential purposes (47.2% vs. 53.4%) and slightly more for commercial purposes (19.3% vs. 18.5%). When we look at areas within 500 meters of the stations, the share of land devoted to commercial use increases to over 29%, while the share of residential land is only 41%. While it is not clear what the land use mix is for TOD, the higher relative concentration of commercial land use next to metro stations suggests that jobs are more likely to locate in places that are relatively easy to access by metro. However, the number of jobs that locations next to metro stations can support depends not only on land use, but also on density. To check how much density these locations can sustain, we estimate two sets of regressions. First, we regress the log mean or maximum height at the parcel level on whether the parcel is within 500 meters of a metro station, while controlling for the area, distance, and squared distance to the city center. As our general spatial equilibrium exercise takes the ward as the unit of analysis for policy simulations, we also examine how height varies between wards hosting metro stations and those that do not. As shown in panel B of Table 3, residential buildings in the 500-meter neighborhood of metro stations are 6% taller than those outside, conditional on land area and location relative to the city center. However, commercial buildings near metro stations are 3% shorter. Moreover, the difference for residential buildings is economically small considering that the permissible FAR for TOD in Bengaluru is set at 4 (Jain and Singh 2019), while the actual average FAR in residential parcels is 1.05. The results in panel C of Table 3suggest that our conclusions do not change if instead of using a 500-meter radius around stations, we simply use an indicator of whether the parcel is in a ward with at least one metro station. Overall, we find no evidence of TOD in Bengaluru. This suggests that there is considerable room to increase density in areas near existing and planned metro stations. Using the counterfactual analysis in Section 6, we will quantitatively assess how relaxing FAR limits in parcels surrounding metro stations affects welfare. 3.3 Land Use Allocations across the 2015 and the 2031 Master Plans In 2017, the Bengaluru Development Authority (BDA) prepared the draft 2031 Land Use Master Plan, which contained significant proposed changes to land use, particularly the addition of land for commercial and residential land use outside the center and the reduction of commercial use in the center. Although the plan has since been retracted, it offers a valuable opportunity to assess land use planning in Bengaluru. 17 Table 4: Comparison of Land Use Allocations between the 2015 and the 2031 Master Plans (1) (2) (3) (4) (5) Number Resid. Comm. Other Total of wards 2015 allocation 198 18,530 6,649 18,731 43,910 2031 allocation 198 31,895 4,823 26,566 63,284 Change from 2015 to 2031 All wards 198 13,365 -1,826 7,835 19,374 Inside ORR 132 1,177 -1,059 1,982 2,100 Outside ORR 66 12,188 -767 5,853 17,274 With metro station 70 7,070 -667 3,527 9,930 Without metro station 128 6,295 -1,159 4,308 9,444 Inside ORR, with station 40 373 -320 788 842 Inside ORR, without station 92 804 -740 1,194 1,259 Outside ORR, with station 30 6,697 -348 2,739 9,089 Outside ORR, without station 36 5,491 -419 3,114 8,185 ORR = outer ring road. Note: Land area is in hectares. “Other” land use type includes agriculture, defense, forest, open space, public utilities, public and semi-public, quarry, transport and communication, water bodies, and national green tribunal’s buffer. Source: Authors’ estimates. Table 4shows a comparison between the 2015 and 2031 land use allocations and reveals the following patterns. First, there is a significant increase in residential land use (13,655 hectares), but a decrease in commercial land use (1,826 hectares). As a result, the supply of residential land increases by 72%, while commercial land shrinks by 27%.13 Appendix Figure C.6 shows the map of the changes in land use stipulated by the plan. Second, the increase in residential land comes from different sources depending on whether the ward is inside or outside the ORR. For wards inside the ORR, the increase in residential land is mainly at the cost of land allocated to commercial use (i.e., rezoning). Outside the ORR, most of the increase in residential land is due to drawing down of land from the “unallocated” category.14 Third, we find no obvious consideration of TOD in the 2031 Land Use Plan. Whether or not a ward contains a metro station does not seem to matter for changes in 13Total land increases by 19,374 hectares or 44%. The additional land comes from the “unallocated” land category, which is land inside the study area but previously not identified for any land use. 14For example, see Appendix Figure C.2 for a ward, whose residential land area has increased from 381 hectares in 2015 to 1,019 hectares in 2031. Only a small amount of the increase is from rezoning commercial land, while most of the new residential land comes from the previously unallocated land. 18 land allocation patterns. With or without a metro station, wards tend to have more land for residential use and less for commercial use. In section 6.2, we will examine the welfare implications of the 2031 Land Use Plan if it were to be implemented. 4 Spatial Model of Bengaluru Next, we build a quantitative spatial model of Bengaluru to assess the general equilibrium effects of the introduction of the metro in the context of land use restrictions. The model belongs to the class of quantitative spatial equilibrium models with commuting.15 Consider an urban area that consists of a finite set Iof discrete locations (wards), each populated by workers, firms, and floorspace developers. Workers differ by skill: low and high. They supply their labor to firms and consume residential floor space and a numeraire consumption good. Workers suffer disutility from time spent commuting between home and work. They commute by either public or private transport, and the choice of transportation affects commute time. The choice of residence and employment locations depends on commute time, wages at the place of employment, housing costs and amenities at the place of residence, and idiosyncratic location and transport mode preferences. Firms use labor and commercial floorspace to produce the numeraire consumption good, which is traded within the urban area at no cost. Firms’ total factor productivity depends on agglomeration spillovers that are increasing in local density of employment. Developers use land and the numeraire to produce floorspace, which can be used for residential or commercial purposes. The supply of floorspace in each location is restricted by zoning regulations and FAR limits. Total employment in the city is endogenous and responds to changes in expected utility. More details of the model can be found below. 4.1 Workers A worker has skill s∈ {L,H}. The total number of workers with each skill is Ns. Workers make choices in the following sequence. First, they choose the location of residence i∈ I and the location of work j∈ I. Second, they choose transportation mode m∈ M ≡ {T,P}to go from ito j, where Tstands for transit and Pfor private. Third, they choose the quantity of goods and housing to consume. Workers also experience two preference shocks: (1) shock for the residence-workplace pair, υij, drawn from the Fréchet 15A canonical example of this model is Ahlfeldt et al. (2015). Sturm et al. (2023) show how such a model can be built for a city in a developing country with minimal data requirements. 19 distribution with cdf F(υ)=exp(−υ−ϵ); (2) shock for the transportation mode, ˜ υm, drawn from the Fréchet distribution with cdf ˜ F(˜ υ)=exp(−˜ υ−σ). Utility and demand. The utility function of individual nis given by us mij(n)=υij(n)Xs iEs j(cs G,ij)1−γ(cs S,ij)γ˜ υm(n)Bm dmij ,(4.1) where cGand cHdenote the consumption of goods and housing (shelter), and dmij is the utility cost of commuting from location ito location jusing mode m. Parameters Xs i,Es j, and Bmare amenity terms associated with residence location, workplace location, and transportation mode. In the baseline model, we assume that residential amenities Xs iare exogenous. Later, in Section 6.6, we study how sensitive our results are to this assumption. The budget constraint is ws j=cs G,ij +qRics H,ij,(4.2) where qRi is the cost of renting one unit of residential floorspace in location i. The commuting cost dmij is modeled as the “iceberg” cost: dmij =eκtmij ,(4.3) where tmij is the time in minutes it takes to travel from ito jusing mode m. Parameter κ measures the strength of the relationship between the cost of commuting and the commuting time.16 The demand functions for housing and the consumption good are cs G,ij =(1 −γ)ws jand cs H,ij =γws j qRi .(4.4) The indirect utility function, conditional on idiosyncratic shocks, is vs mij =Xs iEs j ws j qγ Ri Bm dmij .(4.5) 16The exponential functional form is standard in the literature. See Ahlfeldt et al. (2015), Tsivanidis (2023), Zárate (2022), and many others. 20 Choice probabilities. The probability that an s-skilled worker chooses location pair (i,j)is given by πs ij =(vs ij)ϵ ∑ i′∈I ∑ j′∈I (vs i′j′)ϵ,(4.6) where vs ij ≡∑ m∈M (vs mij)σ 1 σ (4.7) is the expected value of choosing pair (i,j)before knowing the realization of the transportation mode shock. The probability that this worker chooses mode m, conditional on having chosen residence-workplace pair (i,j), is πs m|ij =(vs mij)σ ∑ m′∈M (vs m′ij)σ.(4.8) Finally, the probability of choosing residence i, workplace j, and commuting mode mis πs mij =πs m|ijπs ij.(4.9) Welfare. The expected utility of a worker with skill s(and our measure of worker’s welfare) is Vs= Γ (σ−1 σ)Γ(ϵ−1 ϵ)∑ i∈I ∑ j∈I ∑ m∈M (vs mij)σ ϵ σ 1 ϵ ,(4.10) where Γ(·)denotes the Gamma function. Labor supply. The choice probabilities specified above determine local residential population by skill, Ns Ri =∑ j∈I ∑ m∈M πs mijNs,(4.11) 27 estimated by Saiz (2010) for the US. He estimates the elasticity of 1.75. Since in our model the elasticity is (1 −η)/η, we obtain η=0.3636. The migration elasticity is set to ζ=3, following Bryan and Morten (2019) and Tsivanidis (2023). In Section 6.6, we evaluate the sensitivity of counterfactual results to the values of several of these parameters. Commuting parameters. To estimate the Fréchet elasticity of the preference shock for residence-workplace pairs ϵ, we use the pseudo-Poisson maximum likelihood (PPML) estimator and maximize the following log-likelihood function, ln L ≡ ∑ m∈M ∑ i∈I ∑ j∈I Nmij ln (φMmφRiφWje−κϵtmij ),(5.1) where Nmij is the number of commuters from ito jthat use mode m;φMm,φRi, and φWj are mode, origin, and destination fixed effects that subsume all relevant local variables that appear in the expressions for location choice probabilities; and tmij is the commuting time from ito jwith mode m. Table 6reports estimation results. Because we cannot separately identify the commute cost elasticity κand the Fréchet elasticity ϵ, we first estimate the product κϵ and obtain 0.0902. Then, to recover ϵwe set κ=0.013, as estimated by Tsivanidis (2023) for the city of Bogota, Colombia.21 Thus, our estimate of ϵis equal to 6.9359 =0.0902/0.013. In Section 6.6, we study how sensitive our results are to this value. Transport mode choice. The parameter that reflects preferences for transport modes Bmis normalized to 1 for transit and calibrated for private vehicles during model inversion to match the fraction of commuters who use the private mode, as described in Appendix Section B.1. The calibrated value of BPis 1.19, which reflects the relative preference for using a private vehicle. Table 6: Estimation of ϵ tmij –0.0902 (0.0020) Residence f.e. yes Workplace f.e. yes Observations 78,012 Pseudo R20.2101 Note: This table reports estimated values of −κϵ from equation (5.1). Standard errors are in parentheses. Source: Authors’ estimates. 21Other studies that estimate the commute cost elasticity arrive at similar estimates. For instance, Ahlfeldt et al. (2015) find a value of 0.011 and Zárate (2022) finds 0.009. 28 We calibrate the Fréchet elasticity of the transportation mode preference shock σas follows. The relative probability of choosing transit versus private mode on a given route is πs T|ij πs P|ij =      vs Tij vs Pij       σ .(5.2) The elasticity of this probability ratio with respect to commuting time is ∂ln (πs T|ij/πs P|ij) ∂ln tTij =−σκtTij.(5.3) Note that the left-hand side can be decomposed into the difference between the own-elasticity, ∂ln πs T|ij/∂ ln tTij, and the cross-elasticity, ∂ln πs P|ij/∂ ln tTij. We use available estimates of ownand cross-elasticities of transit ridership with respect to commuting times on transit and by car from Frank et al. (2008): −0.39 and 0.02. This implies that −σκtTij =−0.41. Finally, using the average commuting time by transit in Bengaluru, 26.55 minutes, and the calibrated value of κ, we obtain σ=1.1878. In Section 6.6, we investigate how our results depend on this value. Local parameters. Local parameters Xs i,ωs j,ψj,ϕfi,aj, and Es jare computed by inverting the model, i.e., finding parameters that are consistent with the data being an equilibrium of the model. A detailed description of the model inversion is contained in Appendix Section B.1. Appendix Figure C.5 shows the distribution of residential (Xs i) and employment (Es j) amenities, as well as total factor productivity Ai, while panels (a)–(d) in Appendix Figure C.4 show skill-specific residential population and employment. Appendix Table C.2 summarizes ward-level population and job counts, population and job density, wages, floorspace prices, and other variables used in the quantitative model. Land use parameters. The value of νfi, the share of land zoned for use f, is obtained from the data on the fraction of land zoned for commercial or residential use in each ward. Land area of a ward, Λi, is equal to the sum of land used for commercial and residential purposes. As can be seen from equation (4.34), the expression for floorspace prices depends on whether the FAR limits are binding (i.e., the development has reached maximum height and no further development is possible) or not binding in a given ward. Moreover, the equation shows that we can only identify the FAR limit ¯ hfi when the limit is binding and construction productivity ϕfi when the limit is not binding. 29 To obtain the values of ¯ hfi, we rely on the evidence for density gradients presented in Section 3.1 and assume that the wards inside the ORR have binding FAR limits. Thus, the calibrated FAR limits ¯ hfi in wards within the ORR are equal to the observed FAR. For wards outside the ORR, we assume that their FAR limit is equal to either the average FAR limit of within-ORR wards or the observed FAR, whichever is greater. This implies that some wards outside the ORR where the density of development is higher than the average within the ORR are also assumed to have binding FAR limits.22 In Section 6.5 below, we study how our results would change if the FAR limits were nonbinding everywhere. To obtain the values of ϕfi, we proceed as follows. In wards where the FAR limit is nonbinding, ϕfi can be identified from equation (4.34) using observed heights and the predicted value of qfi (see Appendix Section B.1 for details). In wards where the limit is binding, we assign the average value of ϕfi in nonbinding wards. In some wards, this means that the value of ϕfi is such that the equation (4.34) takes the value of the second argument of the max operator, which is a contradiction because the ward was assumed to have a binding FAR limit. For these wards, we increase ϕfi to the level that makes both arguments of the max operator equal to each other. 5.2 Model Validation While we have the data on commuting flows, we do not feed them directly into the model.23 Yet, the correlation between the flows in the model and in the data is high: 0.63 for transit flows and 0.48 for private vehicle flows. Panels (e) and (f) of Appendix Figure C.4 show the model-predicted residential and commercial floorspace prices. Although we do not have the data to compare with our model-predicted prices, the prices tend to be higher in the central areas of the city, as is the case in most cities around the world. We can also examine the model’s prediction for the value of time (VOT). An increase in commute time by one hour implies that indirect utility falls to e−60κ≈46% of the original level. Since indirect utility is linear in wages, it means that a worker is willing to give up 54% of wages to reduce commute by one hour. Assuming an eight-hour workday, this implies that the VOT is 54% ×8=432% of hourly wage, which is remarkably close to Kreindler (2023)’s VOT estimate of 370% of hourly wage for the city of Bengaluru. 22Twelve out of 66 outside-ORR wards have binding residential FAR limits, and 7 out of 66 wards have binding commercial FAR limits. 23If we did, we would have to calibrate pair-specific shifters for each location pair, and given that many pairs have zero commuters, many shifters would have a value of zero. These pairs would inevitably have zero commuters in any counterfactual. See Dingel and Tintelnot (2021) for a discussion of this issue. 30 6 Effects of the Metro and Land Use Policies In this section, we examine several counterfactual scenarios in which the metro is introduced and land use policies are changed. 6.1 Introduction of the Metro In our first counterfactual, we introduce the metro without making any other changes. We compare an equilibrium without the metro (the “benchmark economy”) with an equilibrium where the full metro network with 125 stations (see Figure 1) is constructed. Preferences between private and public transportation are the same as before and are determined by the combination of BTand σ. Now, however, workers can choose their preferred mode of transit: bus or metro. We assume that they always choose the faster option, i.e., tTij =min {tbus,ij,tmetro,ij}. In Section 6.6, we also consider a scenario in which workers have an explicit preference to ride the metro. In the data, the metro is a little faster on average: the weighted average one-way commute time (weighted by transit commuting flows) is 28.6 minutes by bus and 26.6 minutes by metro. As a result, commuters on routes well served by the metro prefer it to the bus. Panels (a) and (b) of Figure 5show the counterfactual changes in residents and jobs. There is mild decentralization of both: as shown in column (1) of panel A in Table 7, the share of residents in central wards declines by 0.7 percentage points, while the share of jobs declines by 1 percentage point. There is particularly strong growth in residents in selected wards to the southeast and southwest of the city center, and strong job growth in the southeast. These places are seeing a large increase in their residential and firm CMA thanks to the construction of metro stations. Relocations of jobs and residents feed into floorspace prices: panels (c) and (d) show that changes in residential and commercial prices largely mimic the patterns in resident and job movement. Column (1) of panel A in Table 7summarizes the aggregate results of this counterfactual experiment. Wages barely change because the metro has a negligible effect on labor productivity. The population increases by 4%, causing an increase in residential floorspace prices of nearly 2% and commercial prices of nearly 3%. Welfare gains are similar for highand low-skilled workers, amounting to about 1.3% for an average resident. What drives these welfare gains? Note that wages do not change much and housing costs go up, which reduces utility from consumption for an average worker. The key source of welfare gains is the increase in the residents’ CMA.24 24CMA is a sufficient statistic for welfare gains, as shown in Tsivanidis (2023), and it also incorporates 31 Figure 5: Local Effects of Introducing the Metro Panel (a): Residents Panel (b): Jobs Panel (c): Residential floorspace prices Panel (d): Commercial floorspace prices Panel (e): Residential CMA Panel (f): Firm CMA CMA = commuter market access. Note: The maps show results for the counterfactual where the metro is introduced. Panels (a) and (b) show changes in residents and jobs. Panels (c) and (d) show changes in residential and commercial floorspace prices. Panels (e) and (f) show changes in weighted-average (by skill and transport mode) residential and firm commuter market access. Red lines represent the metro network and red crosses are metro stations. Source: Authors’ estimates. changes in wages and housing costs, as can be seen from equations (4.17) and (4.19). 32 Table 7: Aggregate Effects of Metro and Land Use Policies Introduce metro: – ✓–✓–✓–✓ Increase FAR next to stations: – – ✓ ✓ – – – – Increase FAR within the ORR: – – – – ✓ ✓ – – 2031 land use plan: – – – – – – ✓ ✓ Panel A: Employment, wages, prices, and welfare (1) (2) (3) (4) (5) (6) (7) Residents, % chg 4.0 5.8 10.5 5.5 9.7 -18.5 -14.6 low-skilled 3.9 4.8 9.5 4.8 9.0 -18.3 -14.5 high-skilled 4.2 6.9 11.8 6.3 10.7 -18.7 -14.8 Shares, p.p. chg residents within ORR -0.7 2.0 1.0 2.7 1.9 -4.5 -5.5 jobs within ORR -1.0 4.0 2.4 4.7 3.7 -8.9 -10.4 residents next to stations 0.5 1.7 2.5 -0.5 0.0 0.4 0.9 jobs next to stations 0.4 3.7 4.6 -0.4 0.0 6.2 6.6 Wages, % chg 0.1 2.9 3.0 2.5 2.6 -7.7 -7.6 low-skilled 0.1 2.4 2.6 2.2 2.3 -7.9 -7.7 high-skilled 0.0 2.8 2.8 2.4 2.3 -7.4 -7.4 Residential floorspace prices, % chg 1.9 0.5 2.0 0.6 2.5 -12.5 -11.9 Commercial floorspace prices, % chg 2.8 -2.7 -1.4 -2.5 0.1 29.9 32.7 Land prices, % chg 4.1 8.8 13.8 8.1 12.5 7.9 13.2 Welfare, % chg 1.3 1.9 3.4 1.8 3.2 -6.6 -5.1 low-skilled 1.3 1.6 3.1 1.6 2.9 -6.5 -5.1 high-skilled 1.4 2.2 3.8 2.1 3.4 -6.7 -5.2 Panel B: Commuting patterns (0) (1) (2) (3) (4) (5) (6) (7) Mean time to work, min. 22.7 23.4 21.8 23.4 21.8 23.5 22.0 23.7 Private mode use, % 60.4 59.3 60.3 59.2 60.4 59.3 60.3 59.2 Transit use, % 39.6 40.7 39.7 40.8 39.6 40.7 39.7 40.8 bus 39.6 25.0 39.7 24.7 39.6 25.0 39.7 24.7 metro 0.0 15.7 0.0 16.1 0.0 15.8 0.0 16.1 low-skilled 39.8 40.9 39.9 40.9 39.8 40.9 39.9 40.9 high-skilled 39.4 40.6 39.5 40.6 39.4 40.6 39.4 40.6 FAR = floor-to-area ratio, ORR = outer ring road. Note: Panel A shows the counterfactual changes in employment, wages, prices, and welfare. Panel B shows commuting patterns. Column (0) shows results for the benchmark economy. Other columns show counterfactual results. The header indicates which of the four adjustments are considered in a given counterfactual. “% chg” refers to percentage changes, and “p.p. chg” refers to percentage point changes. Source: Authors’ estimates. As panel (e) of Figure 5demonstrates, nearly all residents benefit from better connectivity and access to jobs brought by the metro. Similarly, as panel (f) shows, firms in most locations enjoy better access to workers. 33 Column (1) of panel B in Table 7summarizes the counterfactual commuting patterns. The share of residents using transit increases by 1.1 percentage points, from 39.6% to 40.7%. The majority of those who use transit still take the bus, but for 15.7% of Bengalureans, the metro becomes the primary mode of transportation to work.25 The uptake does not differ much across skill groups. Interestingly, despite the fact that metro makes commuting faster on many routes, workers’ commutes are getting slightly longer: the average travel time to work increases from 22.7 to 23.4 minutes. There are two important reasons for this result. First, some workers are switching to transit thanks to the shorter travel times. However, transit is still slower than private vehicles and the commuting time of an average worker increases. Second, workers have idiosyncratic location preferences. A more efficient transportation network creates more options as workers are freer to choose their preferred locations and can take combinations of residences and jobs that are farther apart. 6.2 Land Use Policies As we have seen, the metro brings nontrivial welfare gains. However, these gains may not reach their full potential because in many locations where the metro increases demand for housing or commercial real estate, supply is limited by the FAR restrictions. As Figure 6shows, in most wards where the FAR limits were binding in the benchmark economy, the limits remain binding in the counterfactual. In a few wards where real estate demand declined, the limits became non-binding; however, several wards that received stations and where the limits were not binding are now hitting their FAR limits. In what follows, we study how changes in land use regulations interact with the introduction of the metro. First, we simulate a policy of TOD by increasing FAR limits near metro stations. Second, we increase FAR limits in all central wards. Finally, we change zoning as envisaged by the 2031 Land Use Plan. Transit-oriented development. We simulate what would happen to ward-level FAR if we implement TOD by raising the FAR of land parcels within 500 meters of metro stations from their current level to 2.26 This results in varying increases in FARs at the 25The small increase in the transit share but large reallocation from the bus to the metro parallels Tsivanidis (2023)’s findings for the city of Bogotá. He finds that about fifteen years since the introduction of the bus rapid transit it accounted for 21% of commutes, while the share of conventional buses fell from 74% to 48%. However, the share of car commutes changed little, from 17% to 15%. 26The limit of 2 is below the current limit of 4 near metro stations. However, as discussed in Dhindaw et al. (2021), the FAR of 4 may not be achievable because most land parcels near stations are too small to accommodate tall buildings. 34 Figure 6: Binding and Nonbinding FAR Limits (Introduction of the Metro) Panel (a): Residential FAR Panel (b): Commercial FAR FAR = floor-to-area ratio. Note: The figure shows the wards where FAR limits are not binding in the benchmark (BM) and counterfactual (CF) economies (“never bind”), wards where FAR limits bind in BM but do not bind in CF (“bind to non-bind”), wards where FAR limits do not bind in BM but bind in CF (“non-bind to bind”), and wards where FAR limits bind in both BM and CF (“always bind”). Source: Authors’ estimates. ward level. The affected wards include those with metro stations (70 wards) and those with land parcels within 500 meters of a station. Figure 7shows the percentage changes in FAR limits in each ward for both residential and commercial development. Within the ORR, where the FAR limits are binding, the average ward sees a 15% increase in residential FAR limits and a 27% increase in commercial FAR limits. Appendix Figure C.7 demonstrates that almost all wards where the FAR limit was binding expand their floorspace to the new limit, making FAR restrictions binding again. We proceed in two steps. First, we increase the FAR limits without introducing the metro. Then, we also introduce the metro. This allows us to understand the effects of each policy individually and then to study the complementarity between the two types of policies. Column (2) of panel A in Table 7shows that less strict density limits lead to much lower growth in residential prices and a nearly 3% fall in commercial prices. These price changes are a combination of greater supply in response to less tight restrictions and greater demand from nearly 6% additional workers that move into the city. Small increases in housing prices allow workers to maintain their consumption levels, while lower prices of commercial real estate allow employers to raise wages. This policy results in a welfare gain of 1.9% for an average worker. It also leads to a large relocation of jobs and residents to central wards and wards with metro stations. In column (3), we report the results of an experiment in which we introduce the metro in addition to increasing FAR limits. Population growth doubles from the previous experiment and, as a result, residential prices increase more, while the decline in 35 Figure 7: Transit-Oriented Development Panel (a): Residential FAR Panel (b): Commercial FAR FAR = floor-to-area ratio. Note: Panel (a) shows the percentage increase in the FAR limit for residential development in the counterfactual with transit-oriented development, while panel (b) reports the increase for commercial development. Source: Authors’ estimates. commercial prices is much smaller. There is a greater movement of jobs and residents to locations with metro stations and there is greater wage growth. In addition, greater concentration near metro stations leads to a larger increase in metro use, as shown in column (3) of panel B in Table 7. Unlike the scenario with metro only, in which the welfare gains for highand low-skilled workers were similar, the counterfactuals with TOD result in greater gains for the high-skilled and larger migration of high-skilled labor into Bengaluru. This occurs primarily because TOD leads to greater employment within the ORR, where wards have relatively high job amenities Es jfor high-skilled workers.27 Thus, by disproportionately treating central wards that specialize in high-skilled jobs, TOD and the metro yield larger benefits for high-skilled workers. Central upzoning. Next, we increase the FAR limits in all locations within the ORR uniformly by the same total amount as in the previous experiment, i.e., by 15% for residential parcels and 27% for commercial parcels within the ORR. This counterfactual attempts to alleviate the problem of constrained development in the city center without explicitly targeting areas near metro stations. As before, we proceed step by step: first, we increase the FAR limits without introducing the metro, then we increase the FAR limits and introduce the metro. 27This reflects the abundance of high-skilled jobs in the city center in the data. 36 Columns (4) and (5) of Table 7summarize the results. Many of the outcomes are comparable to the counterfactuals with TOD. One difference is that in this scenario, there is much more concentration of jobs and residents in the city center, which is not surprising since the policy treats only central wards. At the same time, the share of residents and jobs near metro stations decreases in this experiment, since many stations are located outside of the center and are not treated by the policy. The welfare gains in this scenario are somewhat smaller than the gains from the TOD policy. As with TOD, high-skilled residents benefit more from upzoning central wards than their low-skilled counterparts because these wards specialize in high-skilled jobs. 2031 Land Use Plan. In the following counterfactual experiment, we examine how the introduction of the metro would affect the local economy if the urban development authority’s proposed land use plan for 2031 were fully implemented. As discussed in detail in Section 3.3, the plan increases the total land available for development. However, while the amount of land available for residential development increases, the amount of land available for commercial development decreases. Moreover, land supply increases predominantly in the periphery of the city where few people live and which are far from most jobs (see Appendix Figure C.6). We first implement the 2031 plan without introducing the metro, and then implement the plan and introduce the metro. Columns (6) and (7) of Table 7show the results. The economy suffers huge welfare losses of 5.1%–6.6%. The main reason is the reduction of land that can be used for commercial development. The prices of commercial floorspace increase by 30%–33%, making it more difficult for employers to hire workers. As a result, wages fall by nearly 8% and city-wide employment declines by 15%–18%. This policy also significantly reduces the share of residents and jobs in the city center. At the same time, the share of jobs next to the metro stations increases by 6 percentage points, primarily because a number of wards in the southeast receive several metro stations and do not see large reductions in commercial land. The 2031 Land Use Plan dramatically reduces the land area zoned for commercial use, especially in central wards. What would be the optimal way to rezone the city in light of the introduction of metro? To tackle this question, we run a series of counterfactuals in which, instead of implementing the 2031 plan, we increase the area of land zoned for commercial and industrial use by 5,000 square meters while keeping the total land area fixed in each ward, one by one. We then look at the city-wide welfare gain resulting from commercial rezoning and building the metro in each ward. Panel (a) of Figure 8shows that commercial rezoning in the vast majority of wards yields city-wide welfare gains. 43 consensus in the literature about the value of this elasticity.34 Then we recalibrate the benchmark economy and rerun all counterfactuals. The results can be found in Appendix Table D.2. Most of the results are more pronounced compared to the main counterfactual. For example, welfare gains are larger and population growth is stronger. The main reason for the larger effects is that the influx of residents into Bengaluru increases the amenities in the city, which encourages even more people to move to the city. Extra utility of riding the metro. In the benchmark model, we assume that commuters have the same utility when they take the metro or the bus, and simply choose the faster mode. However, anecdotal evidence suggests that riding the metro may be a more pleasant experience because metro cars are newer, cleaner, and feel safer than buses. We explore this idea by adding a 5% utility gain for metro rides. In other words, commuters are indifferent between taking a bus or taking the metro and incurring a 5% higher commuting cost. As can be seen in Appendix Table D.3, most of the results are larger than in the main set of counterfactuals. The extra utility of metro rides encourages more people to come to Bengaluru and more transit commuters to choose the metro over the bus. No migration. Our main counterfactuals allow for migration in and out of Bengaluru. In the counterfactual with the metro only, employment in the city goes up by nearly 4%, while in counterfactuals with the metro and the land use policies employment increases by about 10%. To understand how the implications of metro and land use policies depend on migration, we close the city and maintain fixed city-wide employment in all experiments. The results are shown in Appendix Table D.4. The welfare gains without migration are much larger. In the absence of immigrants from other parts of India, floorspace prices fall, benefiting both firms and workers. Fixed productivity. In our main counterfactual experiments, we allow local productivity to change endogenously in response to changes in employment. This channel is standard in spatial equilibrium models and has received plentiful empirical support. To understand its role in our results, we mute this channel by setting λ=0(see equation 4.26) and then recalibrate the model. 34Meta-analysis in Ahlfeldt and Pietrostefani (2019) finds that the density elasticity of amenities largely depends on the type of amenity. Averaged over 67 studies, the estimates vary from −0.04 to 0.24 (categories 5, 6, 8, 9, and 10 from Table 3). 44 The results are in Appendix Table D.5. Wage growth is slightly lower in all counterfactuals, because the new incoming labor force does not make the city more productive. While welfare gains are somewhat smaller, they are still large enough in counterfactuals with TOD and upzoning in central wards to make the net benefit of constructing the metro positive. Lower skill substitution and labor informality. For our main counterfactuals, we used the elasticity of substitution between lowand high-skilled labor of 2, as estimated for the US by Card (2009). Due to high levels of labor informality, there may be more segmentation of labor markets in a developing country context and, therefore, the two skill groups may be less substitutable. To understand how our results depend on the skill substitution elasticity, we lowered ξfrom 2 to 0.1.35 Then we recalibrate the model and rerun the counterfactuals. Appendix Table D.6 shows that all of the counterfactual results barely change, which suggests that the elasticity of substitution between skills is not a key parameter that determines the effects of the metro and land use policies. Lower ϵ.The estimated value of the Fréchet elasticity of the residence-workplace shock ϵis 6.9359. This is within the range of existing estimates but somewhat above the average. A high value of ϵmeans that workers, when making location choices, put a relatively high value of location fundamentals and a low value on their idiosyncratic preferences. To evaluate how sensitive our results are to this elasticity, we lowered the value of ϵto 4, recalibrated the model, and reran counterfactual experiments. The results can be seen in Appendix Table D.7. Lower value of ϵmeans that workers, when making location choices, put a relatively low value on location fundamentals and a high value on their idiosyncratic preferences. Therefore, workers choose distant location pairs more often which results in longer commutes (the average commute in the benchmark model is nearly 34 minutes, compared to 23 minutes in the main model). However, because marginal commuting costs are increasing in commute time (equation 4.3), the introduction of metro reduces the average commute, while in the main set of counterfactuals the average commute went up. As a result, welfare gains from introducing the metro are significantly larger. These larger gains lead to higher population inflows and, as a consequence, to higher growth in land and floorspace prices. 35With such low substitution, the production function approaches Leontief. 45 Table 9: Complementarity of Transportation and Land Use Policies in Sensititivity Checks Wages Resid. prices Comm. prices Land prices Welfare TOD ORR TOD ORR TOD ORR TOD ORR TOD ORR Main counterfactual 0.05 -0.04 -0.38 -0.09 -1.32 -0.14 0.55 -0.06 0.15 -0.01 Endog. amenities -0.07 -0.16 -1.40 -0.93 -1.74 -0.70 1.20 0.04 0.37 0.06 Extra utility of metro -0.01 -0.07 -0.54 -0.09 -1.27 -0.09 0.64 -0.16 0.19 -0.03 No migration -0.01 -0.01 -0.61 -0.15 -0.86 -0.13 -0.01 -0.01 0.16 0.02 Fixed productivity 0.08 0.00 -0.34 -0.03 -0.83 -0.02 0.57 -0.02 0.15 0.00 Lower skill subst. 0.05 -0.04 -0.38 -0.09 -1.32 -0.14 0.55 -0.06 0.15 -0.01 Lower ϵ0.07 -0.05 -0.20 0.01 -0.74 0.02 0.65 -0.17 0.17 -0.03 Higher σ0.06 -0.04 -0.37 -0.09 -1.29 -0.10 0.56 -0.07 0.15 -0.01 Lower h.s. elasticity 0.18 -0.01 -0.25 0.00 -1.77 -0.12 0.86 -0.07 0.20 -0.02 ORR = outer ring road, TOD = transit-oriented development. Note: The table shows the complementarity between constructing the metro and two land use reforms: TOD and upzoning within the ORR. The first row shows complementarity for the main set of counterfactuals and reproduces the numbers in Figure 9. The following rows show complementarity for sensitivity checks discussed in this section. The complementarity is calculated using formula (6.1). Source: Authors’ estimates. Higher σ.The calibrated value of the Fréchet elasticity of the transportation mode shock σis 1.1878. Very few quantitative spatial models have this elasticity; hence, it is important to understand how much our results depend on its value. We increase σto 2, recalibrate the model, and rerun the counterfactuals. Appendix Table D.8 shows the results. Higher value of σimplies that workers have weaker idiosyncratic preferences for transportation modes. In the counterfactual where the metro is introduced, the share of transit commutes goes up by 1.9 percentage points, compared to 1.1 percentage points in the main counterfactual. However, all other results, such as welfare gains, population growth, and price changes, are strikingly similar to the main set of counterfactuals. Lower housing supply elasticity. In our quantitative model, we used the housing supply elasticity of 1.75, as estimated by Saiz (2010) for the US. However, it is likely that a large city in a developing country has a lower elasticity. For example, Sturm et al. (2023) estimate the elasticity of 1.5 for Dhaka, Bangladesh. To evaluate the sensitivity of our results to the value of housing supply elasticity, we reduce it all the way to 1. According to Saiz (2010), this value corresponds to the elasticities observed in the most regulated and land-constrained cities in the US. Appendix Table D.9 shows that the results barely change. With lower housing supply elasticity, worker inflows are a little bit smaller than in the main set of counterfactuals. Yet, the welfare gains of the metro and land use reforms are about the same. 46 Complementarity of metro and land use reforms. Importantly, as Table 9 demonstrates, none of the modifications that we considered in this section change our conclusion that the construction of the metro is complementary to TOD but not with central upzoning. 6.7 Discussion In this section, we discuss a number of additional factors that may be considered when studying the interaction of transit improvements and land use reforms, and that we leave for future research. Traffic congestion. In our model, commuting speeds do not change in response to changes in traffic volumes. How would our results change if we took into account this margin of adjustment? On the one hand, the introduction of the metro takes 15.7% of commuters off roads although, as shown in column (1) of panel B in Table 7, most of the switchers are previously bus riders and buses presumably contribute less to traffic congestion than private vehicles.36 On the other hand, more workers are moving to the city, adding to the pressure on the existing road network. Finally, any improvement in speed may lead to more trips in the long run (Duranton and Turner 2011). Therefore, whether or not the introduction of metro in Bengaluru leads to less road traffic depends on which of the above-mentioned channels dominate. Environmental effects. The metro certainly produces much less greenhouse gas emissions than private vehicles and buses, and taller buildings tend to be more energy-efficient due to economies of scale. As a result, the introduction of the metro is likely to bring environmental benefits that are likely to be amplified by land use policies such as TOD and central upzoning. On the other hand, these benefits may be mitigated by population inflows. Dynamic effects. The static nature of our model does not allow us to distinguish outcomes for incumbent residents and newcomers. However, increases in land prices in all counterfactuals suggests that welfare gains for the incumbents are larger. The absence of dynamics also prevents us from evaluating potential costs of new real estate development on existing residents. In a comparable context of Mumbai, Gechter and Tsivanidis (2023) find that even though the redevelopment of former textile mills in the 36First, buses allow for a denser layout of passengers inside a vehicle. Second, some streets in Bengaluru have dedicated bus lanes. 47 city center led to significant citywide gains, it imposed large losses on the relocated slum dwellers. Informal housing. We treat all housing in the model as homogeneous. Yet Bengaluru, as many other large Indian cities, has a nontrivial amount of informal housing. An alternative way to view our counterfactuals where the FAR limits are relaxed is an increase in the share of formal housing which tends to be taller, and not an increase in the legal height limit. However, understanding how the introduction of metro interacts with formalizing the housing stock would require a dynamic model with segmented housing markets and property rights, such as Henderson et al. (2020). Population growth. Bengaluru is a rapidly growing city that more than doubled its population from 1991 to 2011. At the same time, in our counterfactual experiments we do not incorporate population growth except via migration from other cities, as is standard in the literature. Thus, our results should be viewed as the effects of building the metro and reforming land use isolated from the effects of population growth. 7 Conclusions In this paper, we studied the benefits of improving transportation infrastructure in the presence of land use restrictions. We conducted our study in the context of one of India’s largest cities, Bengaluru, which is building a metro network but where development is severely constrained by zoning and FAR limits. Using detailed data on local labor markets, building heights, land use, and commuting patterns, we built a quantitative spatial equilibrium model of the city and then used it to evaluate the effects of building the metro and reforming land use policies. We found that the construction of the metro led to output and welfare gains, even net of costs. However, these gains could be much larger if the FAR limits were relaxed and more development was allowed in areas near the newly built metro stations or in central parts of the city. We also found that transit-oriented development is complementary to the metro network, which highlights the importance of coordinating land use policies with transportation improvements. Finally, we would like to highlight the policy relevance of our findings. As mentioned earlier, India’s restrictive urban planning norms have come under criticism from reform-oriented policymakers. Our results show how key elements of these norms, namely strict FAR limits and zoning regulations that restrict commercial and residential development, detract from the benefits of modern mass transit systems. Our results also 48 underline the importance of recent efforts by the Indian government to provide incentives for state and city governments to reconsider building regulations and adopt TOD principles that emphasize the need to densify transport corridors. To be sure, implementation of TOD principles is not easy, especially in brownfield locations which are characterized by the presence of large numbers of small residential or commercial plots. Amalgamating these plots into larger land parcels and then availing of much higher FARs and the necessary supporting infrastructure is one way forward. Research examining the conditions under which land amalgamation can be facilitated would be particular useful. In the meantime, as India’s urbanization gathers space, it is critical that TOD principles be adopted in locations where legacy issues are not as constraining. 49 Bibliography Acosta, C. (2022). The incidence of land use regulations. Working Paper. ADB (2019). Fostering Growth and Inclusion in Asia’s Cities. Asian Development Bank. Ahlfeldt, G. and Barr, J. (2022). The economics of skyscrapers: A synthesis. Journal of Urban Economics, 129(C):S0094119021001017. Ahlfeldt, G. M. and Pietrostefani, E. (2019). The economic effects of density: A synthesis. Journal of Urban Economics, 111:93–107. Ahlfeldt, G. M., Redding, S. J., Sturm, D. M., and Wolf, N. (2015). The economics of density: Evidence from the Berlin Wall. Econometrica, 83(6):2127–2189. Akbar, P., Couture, V., Duranton, G., and Storeygard, A. (2023a). The fast, the slow, and the congested: Urban transportation in rich and poor countries. Working Paper. Akbar, P., Couture, V., Duranton, G., and Storeygard, A. (2023b). Mobility and congestion in urban India. American Economic Review, 113(4):1083–1111. Allen, T. and Arkolakis, C. (2014). Trade and the topography of the spatial economy. The Quarterly Journal of Economics, 129(3):1085–1140. Allen, T., Arkolakis, C., and Li, X. (2016). Optimal city structure. Working Paper. Amirkolaee, H. A. and Arefi, H. (2019). Height estimation from single aerial images using a deep convolutional encoder-decoder network. ISPRS journal of photogrammetry and remote sensing, 149:50–66. Anagol, S., Ferreira, F. V., and Rexer, J. M. (2021). Estimating the economic value of zoning reform. Working Paper 29440, National Bureau of Economic Research. Balboni, C., Bryan, G., Morten, M., and Siddiqi, B. (2020). Transportation, gentrification, and urban mobility: The inequality effects of place-based policies. Working Paper. Bank, W. (2013). Urbanization Beyond Municipal Boundaries: Nurturing Metropolitan Economies and Connecting Peri-Urban Areas in India. The World Bank. Barr, J. and Jedwab, R. (2023). Exciting, boring, and nonexistent skylines: Vertical building gaps in global perspective. Real Estate Economics, n/a(n/a). Bertaud, A. and Brueckner, J. K. (2005). Analyzing building-height restrictions: Predicted impacts and welfare costs. Regional Science and Urban Economics, 35(2):109–125. Bharadwaj, K. V. A. (2020). BDA withdraws draft revised master plan - 2031, to review and rewrite. https://www.thehindu.com/news/cities/bangalore/ bda-withdraws-draft-revised-master-plan-2031-to-review-and-rewrite/ article31929610.ece. 50 Brueckner, J. K. and Sridhar, K. S. (2012). Measuring welfare gains from relaxation of land-use restrictions: The case of India’s building-height limits. Regional Science and Urban Economics, 42(6):1061–1067. Bryan, G., Glaeser, E., and Tsivanidis, N. (2020). Cities in the developing world. Annual Review of Economics, 12(1):273–297. Bryan, G. and Morten, M. (2019). The aggregate productivity effects of internal migration: Evidence from Indonesia. Journal of Political Economy, 127(5):2229 – 2268. Cao, Y. and Huang, X. (2021). A deep learning method for building height estimation using high-resolution multi-view imagery over urban areas: A case study of 42 Chinese cities. Remote Sensing of Environment, 264:112590. Card, D. (2009). Immigration and inequality. American Economic Review, 99(2):1–21. Conwell, L. (2023). Subways or minibuses? Privatized provision of public transit. Working Paper. Dhindaw, J., Kumaraswamy, S. K., Prakash, V. S., Chanchani, R., and Deb, A. (2021). Synergizing land value capture and transit-oriented development: A study of Bengaluru metro. Practice Note. World Resources Institute India. Dingel, J. I. and Tintelnot, F. (2021). Spatial economics for granular settings. Working Paper. Duranton, G. and Puga, D. (2022). Urban growth and its aggregate implications. Working Paper. Duranton, G. and Turner, M. A. (2011). The fundamental law of road congestion: Evidence from US cities. American Economic Review, 101(6):2616–52. Frank, L., Bradley, M., Kavage, S., Chapman, J., and Lawton, T. (2008). Urban form, travel time, and cost relationships with tour complexity and mode choice. Transportation, 35(1):37–54. Frantz, D., Schug, F., Okujeni, A., Navacchi, C., Wagner, W., van der Linden, S., and Hostert, P. (2021). National-scale mapping of building height using sentinel-1 and sentinel-2 time series. Remote Sensing of Environment, 252:112128. Gechter, M. and Tsivanidis, N. (2023). Spatial spillovers from high-rise developments: Evidence from the Mumbai mills. Working Paper. Geshkov, M. V. and DeSalvo, J. S. (2012). The effect of land-use controls on the spatial size of US urbanized areas. Journal of Regional Science, 52(4):648–675. Heblich, S., Redding, S. J., and Sturm, D. M. (2020). The making of the modern metropolis: Evidence from London. Quarterly Journal of Economics, 135(4):2059–2133. 51 Henderson, J. V., Regan, T., and Venables, A. J. (2020). Building the city: From slums to a modern metropolis. The Review of Economic Studies, 88(3):1157–1192. Herkenhoff, K. F., Ohanian, L. E., and Prescott, E. C. (2018). Tarnishing the golden and empire states: Land-use restrictions and the U.S. economic slowdown. Journal of Monetary Economics, 93:89 – 109. Carnegie-Rochester-NYU Conference on Public Policy held at the Stern School of Business at New York University. Hsieh, C.-T. and Moretti, E. (2019). Housing constraints and spatial misallocation. American Economic Journal: Macroeconomics, 11(2):1–39. Jain, D. and Singh, E. (2019). Transit oriented development in India: A critical review of policy measures. International Journal of Recent Technology and Engineering, 7(6S2). Kreindler, G. (2023). Peak-hour road congestion pricing: Experimental evidence and equilibrium implications. Working Paper 30903, National Bureau of Economic Research. Kreindler, G., Gaduh, A., Graff, T., Hanna, R., and Olken, B. A. (2023). Optimal public transportation networks: Evidence from the world’s largest bus rapid transit system in Jakarta. Working Paper. Lall, S., Lebrand, M., Park, H., Sturm, D., and Venables, A. (2021). Pancakes to pyramids. World Bank Publications - Reports 35684, The World Bank Group. Li, M., Koks, E., Taubenböck, H., and van Vliet, J. (2020a). Continental-scale mapping and analysis of 3D building structure. Remote Sensing of Environment, 245:111859. Li, X., Zhou, Y., Gong, P., Seto, K. C., and Clinton, N. (2020b). Developing a method to estimate building height from sentinel-1 data. Remote Sensing of Environment, 240:111705. Liu, C., Huang, X., Wen, D., Chen, H., and Gong, J. (2017). Assessing the quality of building height extraction from ziyuan-3 multi-view imagery. Remote Sensing Letters, 8(9):907–916. Liu, C., Huang, X., Zhu, Z., Chen, H., Tang, X., and Gong, J. (2019). Automatic extraction of built-up area from zy3 multi-view satellite imagery: Analysis of 45 global cities. Remote Sensing of Environment, 226:51–73. Parkhomenko, A. (2023). Local causes and aggregate implications of land use regulation. Working Paper. Ramulu, D., Sankar, K., and Randhawa, A. (2021). Challenges of transit oriented development (TOD) in Indian cities. Institute of Town Planners, India Journal, 18(2):37– 55. Roy, S. and Shah, A. (2018). Concerns with Delhi metro. blog.theleapjournal.org/ 2018/12/concerns-with-delhi-metro.html. 52 Saiz, A. (2010). The geographic determinants of housing supply*. The Quarterly Journal of Economics, 125(3):1253–1296. Severen, C. (2021). Commuting, labor, and housing market effects of mass transportation: Welfare and identification. The Review of Economics and Statistics, pages 1–99. Song, J. (2022). The effects of residential zoning in U.S. housing markets. Working Paper. Sood, A. (2022). Land market frictions in developing countries: Evidence from manufacturing firms in India. Working Paper. Sturm, D. M., Takeda, K., and Venables, A. J. (2023). How useful are quantitative urban models for cities in developing countries? Evidence from Dhaka. Working Paper. Tsivanidis, N. (2023). The aggregate and distributional effects of urban transit infrastructure: Evidence from Bogotá’s transmilenio. Working Paper. Valentinyi, A. and Herrendorf, B. (2008). Measuring factor income shares at the sectoral level. Review of Economic Dynamics, 11(4):820–835. Velásquez, D. (2023). Transit infrastructure, couples’ commuting choices, and gender inequality. Working Paper. Warnes, P. (2021). Transport infrastructure improvements and spatial sorting: Evidence from Buenos Aires. Working Paper. Yu, Y. (2022). Land-use regulation and economic development: Evidence from the farmland red line policy in China. Working Paper. Zárate, R. D. (2022). Spatial misallocation, informality, and transit improvements. Working Paper.