SPACE AND TIME: A COMPETING RISKS ANALYSIS OF THE IMPACT OF PROPERTY TAXES AND ZONING RESTRICTIONS ON RESIDENTIAL DEVELOPMENT

• Author: H. Allen Klaiber,Douglas H. Wrenn
• Published date: 01 August 2019
• Date: 01 August 2019
• DOI: http://doi.org/10.1111/iere.12380

INTERNATIONAL ECONOMIC REVIEW

Vol. 60, No. 3, August 2019 DOI: 10.1111/iere.12380

SPACE AND TIME: A COMPETING RISKS ANALYSIS OF THE IMPACT OF

PROPERTY TAXES AND ZONING RESTRICTIONS ON RESIDENTIAL

DEVELOPMENT∗

BYDOUGLAS H. WRENN AND H. ALLEN KLAIBER1

Pennsylvania State University, U.S.A.; The Ohio State University,U.S.A.

Property taxes and zoning restrictions are prevalent tools for managing land use. We combine microlevel

data on residential subdivision development from the Baltimore, Maryland, metro area from 1994 to 2007 with a

competing risks model to examine how both policies influence the density and timing of residential development.

Consistent with theory, we demonstrate that the impact of both policies on optimal density and timing depends

on whether density and time are substitutes or complements in the profit function. Our results have important

implications as they provide key empirical insights into how property taxes and zoning interact to influence

development patterns.

1. INTRODUCTION

Understanding residential housing supply is important for explaining urban land use. It is also

critical for designing effective policy (Nechyba and Walsh, 2004; Gyourko and Molloy, 2015;

Glaeser and Gyourko, 2017). Property taxes and zoned-density restrictions, or minimum-lot

zoning, are common tools for managing land use; their effect on urban land use has received

considerable theoretical attention (Turnbull, 1988, 1991; Brueckner, 2000). Although the impact

of taxes and zoning on the optimal timing of development is clear, their impact on the joint

density–timing decision is ambiguous. In this article, we combine parcel-level data on residential

development outcomes with a multinomial duration model to provide empirical clarity to this

theoretical ambiguity. Specifically, we are interested in (i) how property taxes and zoned-density

restrictions affect the optimal conversion time of undeveloped land to residential development

and (ii) how those changes impact the optimal lot size, or density class, chosen by developers.

It is common to model the conversion of undeveloped land to residential use as an optimal

stopping problem. Holding the choice over density constant, a policy-driven decrease in housing

prices should delay optimal development time, all else equal. Extending this logic to include a

choice over time and density creates ambiguity in how changes to prices and zoning restrictions

influence development outcomes. We can use theory to describe how the optimal development

decision, and its impact on the timing of development across different density classes, depends

on whether density and time are substitutes or complements in the developers’ profit function.

When density and time are complements in the developer’s profit function, it implies that

demand for density, or capital per unit of land, is rising faster than the demand for land area—

that is, consumers demand denser development outcomes. Given this, an increase in taxes and/or

a reduction in zoned density will delay development given that both policies promote less dense

∗Manuscript received April 2018; revised July 2018.

1We are grateful to conference participants at the Agriculture and Applied Economics Association and Association

of Environmental and Resources Economist summer meetings and seminar participants at the SUNY Binghamton,

Virginia Tech, and the University of Illinois for helpful comments. We are also indebted to two anonymous referees for

valuable suggestions for improvement. This research was supported by a cooperative agreement with the U.S. Forest

Service Northern Research Station and National Science Foundation grants CBET 1160961, DEB 0410336, and GSS

1127044. Please address correspondence to: Douglas H. Wrenn, Department of Agricultural Economics, Sociology,

and Education, The Pennsylvania State University, 112A Armsby Building, University Park, PA 16802. Phone: 814 865

9216. Fax: 814 865 3746. E-mail: dhw121@psu.edu.

1097

C

(2019) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

1098 WRENN AND KLAIBER

development. Conversely, when density and time are substitutes, it implies that demand for

land is rising faster than demand for capital per unit of land—that is, consumers demand less

dense development. Here, an increase in taxes and/or reduction in zoned density will delay

development. Consequently, the final impact that property taxes (price policy) and zoning

(quantity target) have on when and how densely land is developed is an empirical question that

depends on the fundamentals of the underlying market.

To test the preceding hypotheses and provide clarity to the debate concerning the impact

of taxes and zoning on development, we estimate an instrumental variable (IV) competing

risk (CR) model using parcel-level data on historical subdivision development. In the CR

model, developers choose the optimal time (year) to convert a parcel of land to a residential

development and the optimal density, where density is defined as a multinomial choice over

low-, medium-, or high-density development. Both decisions are impacted by housing prices

and zoning restrictions. To generate consistent estimates of the impact of prices and zoning

on conversion, we control for price and zoning endogeneity using a control function (CF)

method suitable for instrumentation in a nonlinear model (Heckman and Robb, 1985; Rivers

and Vuong, 1988; Papke and Wooldridge, 2008; Petrin and Train, 2010). We instrument for

housing price using spatial equilibrium theory and empirical insights provided by the urban

location choice literature (Bartik, 1987; Bayer and Timmins, 2007; Bayer et al., 2007; Wrenn

et al., 2017). Specifically, we use spatially lagged, exogenous factors from distant neighborhoods

as instruments for housing prices in focal neighborhoods. Our instruments for zoning are based

on distance measures to a series of drinking-water reservoirs and the legacy effects that those

reservoirs have on contemporary density restrictions in our study region.

Although we do not possess information on the actual revenues and costs associated with

each development project, our data and empirical strategy allow us to identify the key statis-

tics of interest—price and zoning elasticities. These elasticity values capture exactly what we

want to know about the relationship between policy-driven changes in housing prices and

zoning restrictions. Indeed, our model and data allow us to stratify projects based on density

and empirically assess how changes in taxes and zoning impact the conversion rates of low-,

medium-, and high-density projects—that is, our elasticity values are sufficient to assess the

impact of taxes and zoning (Chetty, 2009).

Our subdivision data as well as our price and zoning data come from a three-county region in

the Baltimore, Maryland metro area.2To create these data, we combined detailed Geographic

Information System (GIS) data with historical subdivision maps from the region and assembled

the complete history of subdivision activity from 1994 to 2007. These data include information

on the original land parcel, including a rich characterization of the attributes on each parcel,

as well as information on the number of building lots created within each subdivision. To

create our variable on zoning, we overlaid GIS data with geo-located images of zoning maps to

create a measure of the restrictiveness of zoning. To create our house-price index, we used data

on housing sales and estimated a series of hedonic models. Finally, to determine the density

outcomes, or choices, in our CR model, we combined the actual density outcomes from our

subdivision data with a clustering algorithm designed for univariate data (Bellman, 1973; Wang

and Song, 2011). We used a density measure based on the number of building lots created

per acre of land for each subdivision and this clustering algorithm and determined the optimal

number of clusters, or density cutoffs (low, medium, and high) by maximizing a set of Bayesian

information criterion (BIC) statistics.

The results from our model provide several important insights. First, we demonstrate that

failure to account for price and zoning endogeneity produces biased and inconsistent estimates

of the responsiveness of land conversion decisions (density and timing) to changes in housing

prices and zoning. Price elasticity values across all densities are more than 3.5 times larger

in the IV model, and the elasticity value for zoning in the high-density equation is 2.7 times

2Our data come from Baltimore, Carroll, and Harford Counties. Our data do not include the City of Baltimore. We

provide some additional explanation for our choice of data in Section 4.

PROPERTY TAXES AND ZONING 1099

higher. We also find that price elasticity values in the low- and medium-density equations and

zoning elasticity in the high-density equation are elastic in the IV model compared to inelastic

in the non-IV setup. These results suggest that failure to account for endogeneity is leading

to incorrect conclusions regarding the impact of property taxes and zoning restrictions on the

density–timing decisions of developers.

Next, we find that the responsiveness of land conversion, across all density classes, is greater

in absolute value for changes in prices than for equivalent changes in zoning. Specifically, we

find that marginal changes in housing prices lead to changes in the probability of land conversion

that are 4.1, 6.5, and 3.2 times higher for low, medium, and high density, respectively, compared

to equivalent changes in zoning. To the extent that changes in property taxes lead to equivalent

changes in housing prices, our results suggest that price shocks produce a greater impact on

conversion decisions than equivalent zoning shocks. One qualification is that these comparisons

are based on marginal changes in both prices and zoning. Given that zoning changes tend to

occur at larger scales, a better comparison, for policy purposes, may be marginal changes in

prices compared with larger scale shifts in zoning. We choose to analyze marginal changes in

this article to provide a direct comparison between our results and theory and provide a clearer

comparison among policies.

Finally, as a direct test of the theory on the impact of taxes and zoning on density and

timing, we find that (i) a decrease in housing prices leads to an increase in optimal development

time for all density classes, (ii) a decrease in housing prices leads to a larger increase in optimal

development time as density increases, and (iii) a reduction in zoned density leads to an increase

in optimal development time for high-density outcomes and a decrease in optimal development

time for low- and medium-density outcomes. These results suggest that density and time are

substitutes in the latent profit function for low- and medium-density outcomes for zoning

(consumers prefer lower density development) and complements for high-density outcomes for

zoning and low-, medium-, and high-density outcomes for property taxes (consumers prefer

higher density development). The substitutability of density and time with respect to zoning

changes in the low- and medium-density models follows from the fact that demand for land is

rising faster than the demand for capital, and a policy-driven decease in density intensifies this

relationship, thus speeding up development for lower density development options.

We also use the results from our IV model in a series of land use simulations to further

investigate the impact of tax policy and zoning on development outcomes. We run a baseline

simulation, with no policy change, and a set of simulations where we impose, uniformly across all

parcels and time periods, a 1% property tax, a 1% decrease in zoned density, and a 1% change

in taxes and zoning occurring simultaneously. For each simulation, we compare our baseline

results—predictions for the number of developments created within each density class—with

predicted developments following each policy change. The results from this exercise provide

additional insights into the impact of property taxes and zoning restrictions on development

outcomes. Our results demonstrate that common land use policies that apply property taxes

and density restrictions simultaneously have the potential to intensify the rate of lower density

development, relative to high-density development. To the extent that this type of development

takes place in more remote areas, this type of policy has the potential to expand the urban

footprint and increase sprawl.

The rest of the article proceeds as follows: In Sections 2 and 3, we present our theoretical and

empirical models. In Section 4, we provide an overview our study region, a description of our

data, and a description of the process used to create the density categories in our CR model.

Section 5 presents our main results, and Section 6 presents our land use simulations and some

additional discussion. Section 7 concludes.

2. THEORECTICAL FRAMEWORK

Before presenting our empirical model and data, we provide a brief overview of the pre-

dictions from the theoretical literature on the property taxes, zoning, and the density–timing