Why Local Governments Impose Land-Use Restrictions Bengte Evenson Department of Economics Illinois State University William C. Wheaton Department of Economics and Center for Real Estate MIT August 2002 ABSTRACT We formalize a general equilibrium theory of a metropolitan area housing market whose local governments influence house price by regulating their supplies of land. The regulatory decisions are made by current residents who already own housing, but the impact of these decisions on prices is determined by new entrants who must purchase housing. The choice of how much to regulate is shown to vary by town size, existing town density, and the amount of open land potentially available for development. These results are broadly consistent with recent empirical research as well as with the findings of a direct empirical test which we perform using a unique new dataset from Massachusetts. *We wish to thank Emek Basker, Tom Davidoff, Mark Lewis, Paras Mehta, Sendhil Mullainathan, Mike Noel, and lunch and seminar participants at MIT for helpful advice and comments. Any errors or omissions are our own.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Why Local Governments Impose Land-Use Restrictions
Bengte EvensonDepartment of Economics
Illinois State University
William C. WheatonDepartment of Economicsand Center for Real Estate
MIT
August 2002
ABSTRACT
We formalize a general equilibrium theory of a metropolitan area housing market whose localgovernments influence house price by regulating their supplies of land. The regulatory decisionsare made by current residents who already own housing, but the impact of these decisions onprices is determined by new entrants who must purchase housing. The choice of how much toregulate is shown to vary by town size, existing town density, and the amount of open landpotentially available for development. These results are broadly consistent with recent empiricalresearch as well as with the findings of a direct empirical test which we perform using a uniquenew dataset from Massachusetts.
*We wish to thank Emek Basker, Tom Davidoff, Mark Lewis, Paras Mehta, SendhilMullainathan, Mike Noel, and lunch and seminar participants at MIT for helpful advice andcomments. Any errors or omissions are our own.
1
1. Introduction
Locally imposed land-use regulations are thought to be an important supply-side
determinant of house prices. In a recent cross-section study of this relationship, Glaeser and
Gyourko (2002) find evidence that relatively high house prices are almost entirely due to zoning
restrictions. They conclude, “if policy advocates are interested in reducing the cost of housing,
they would do well to start with zoning reform.”
The extent to which zoning and other land-use regulations increase housing prices is only
beginning to be understood. In part, this is because existing research has able to examine the
empirical relationship only on the most limited of terms. Since land-use regulation is the domain
of more than 18,000 local jurisdictions in the U.S., collecting data on and constructing measures
of regulation is exceedingly difficult. Thus, the number of “stylized facts” about land-use
regulations and their impacts is quite limited. This makes the development of more formal models
somewhat difficult, as there is little to guide the modeling effort.
Most land use regulations generally do one of two things: either they limit the acreage
within a town that can be built upon, or they limit the amount of development that can occur on a
given acreage. Examples of the former include local designations of conservation land, permanent
agricultural use, or public open space. In the latter category, the most common regulations require
minimum lot sizes or else limit the ratio of Floor-to-Land Area (FAR). Cities and towns can also
lengthen the permitting process, but the impact of this on development inherently is a more
complicated time-dependent process and cannot be modeled statically. In this paper we take land
use regulations to be of the first type, limiting developable land, which we assume generates
some form of perceived or real public good for the residents of the town enacting the regulations.
If setting aside land from development creates a local public good then with Tiebout-type
competition, this should result in higher house prices. In this formulation, house prices have to be
a “cost” which mobile residents are willing to pay for whatever public benefit results. Towns with
the most open land become highly desirable enclaves of expensive housing. To complete this
“model” however, a decision rule is needed that local residents use for determining the set aside
amount of land.
The most obvious decision rule is that residents simply wish to maximize the value of
their house - as a form of wealth (Bruce Hamilton (1978)). This is consistent with some type of
broader inter-temporal model in which existing residents are able to either eventually cash out of
their housing market or leave bequests to their offspring. Higher house prices then increase the
real wealth of these homeowners, which eventually increases either their own or their offspring’s
consumption. Although this formulation is intuitively appealing, its fundamental assumption is at
2
odds with how a Tiebout-type model determines the impact of those land use regulations on
house prices. In a static model, how can house prices be both a “cost” as well as a wealth
creating benefit?
The only author to try and reconcile this dilemma, is Brueckner (1996), who does so by
creating two categories of residents who view house prices differently. He assumes that owners
regard house prices positively since they generate capital gains, but renters view house prices as
nothing but a cost. He then goes on to postulate a political process where the interests of each
group are resolved. In our model, all residents are owners; rather we distinguish between existing
and future residents. The decision to regulate leads to capital gains for existing residents by
altering the price that future residents must pay to acquire housing. In contrast to overlapping
generation models, we assume that current residents do not care about the utility of future
residents. We justify this assumption with some data showing that most adults do not live in the
same housing market they were raised in.
With two cohorts or generations of consumers, we create the following simple, two-
period game to be played between the two.
? In the first period, existing residents set aside some fraction of their potentially
developable land as “open space”. This decision is presumed to be irrevocable. This set
aside has some “cost” (a fraction of the land’s value, for example, to be paid annually in
taxes), which is offset to some degree by the gains they receive from higher house prices.
? The impact on house prices is determined in the second period, when future residents
develop the land that has not been irrevocably set aside. Towns that have set aside a
significant portion of their land not only constrict supply, but also affect the utility that
future residents receive when they develop in that town. These two effects jointly
determine the pattern of future house prices across towns
? The model is fully rational in that the period two price impacts are fully incorporated by
period one existing residents when making their decisions.
This seemingly simple model structure turns out to be quite complicated analytically, but
we are able to derive a series of comparative static results. In small towns (as measured by market
share), the land set aside by the existing residents generally leads to a larger increase in period
two house prices than does land set aside by the existing residents of a larger town. Trading this
off against the cost of land set-aside, the existing residents of smaller towns will opt to set aside
(or regulate) more. In a symmetric Nash equilibrium, then, an otherwise identical metropolitan
3
area with many smaller jurisdictions would be expected to have more overall land regulation.
This seems consistent with some recent empirical literature.
Two other hypotheses emerge from the model as well. First, towns with more open land
available for future development per current resident should choose to restrict a smaller fraction
of this land from development. The intuition behind this is that with fewer residents, the cost each
resident faces when determining the optimal amount of land-use regulation is greater. Secondly,
towns in which existing residential density is high should also choose to restrict less land from
development. The intuition here is that with lower land (and housing) consumption, the dollar
capital gains induced by regulating land-use will be smaller.
To test these predictions, we use a unique new data set on land-use and its regulation in
Massachusetts, collected between 1999-2001. The data allow us to determine the number of new
housing units that can be built in each town given current land-use laws as well as the total
amount of land on which these units can be built. The data also has extensive information on
existing density as well as all land that is potentially buildable. The empirics confirm the
predictions of the model.
The paper is organized as follows. Section 2 reviews the literature on local regulations
and their impact on housing market outcomes. Section 3 develops the first part of our model,
about how house prices are determined by future residents, conditional on the amount of land set
aside today. Section 4 then introduces the 2-period game by describing how existing residents’
utility is determined by future market prices. This section also develops the comparative static
predictions. Section 5 discusses the theoretical predictions, how we test them empirically, and the
results and implications of these tests. Section 6 concludes.
2. Literature Review
2.1 Theoretical Literature
In a 1978 paper, Bruce Hamilton proposes in an informal model in which zoning restricts
the supply of housing. Local jurisdictions then determine their optimal level of regulation by
maximizing residents’ housing wealth, which amounts to maximizing home values in the
constituency. 1
Home values between competing local governments are determined in a housing market
by “new entrants”. This creates a fundamental tension; each government must trade-off the desire
of existing residents for high house prices with the desire of these “entrants” for low house prices.
4
Hamilton argues that the more fiercely a government had to compete for new entrant demand,
current demand already being locked in, the more land it must devote to housing in order to keep
house prices low and attract housing demand. Competition constrains the decision over the
amount of land to devote to uses other than housing.2
A metropolitan area that consists of many small towns will face more internal
competition for new housing demand. Conversely, a metropolitan area that consists of a single
large town will face no competition whatsoever. If an area consists of only one local government,
incoming households cannot choose to live in another town in response to relatively high house
prices unless they decide to leave the housing market altogether. Hamilton posits that relatively
large local governments will take advantage of their monopoly power over the supply of land to
restrict land more and hence increase home values for their existing residents. But the question of
who “entrants” are and why existing residents in one town are not potential entrants for another is
left hanging.
Brueckner (1996 and 1998) takes a different tact to examining how it is that housing
prices can be both a wealth generating benefit and a cost – by examining growth controls when
there are owners and renters. Brueckner’s model has a system of open communities and fixed
land consumption for each household. Towns that impose growth controls reduce their boundary,
population and house prices. Towns trade this off against an assumed negative externality from
greater size. With regulation creating both prices appreciation and a direct amenity, he finds that,
as the number of cities gets large, the incentives to impose land-use restrictions in the form of
growth controls decreases. 3 His conclusion that larger jurisdiction take advantage of monopoly
power is similar to that of Hamilton
In a related paper that does not deal directly with land use regulations, Hoyt (1999)
studies the effect of town size on government expenditure decisions. In his model, small
jurisdictions are able to capitalize the effects of expenditures much more than large jurisdictions.
This induces residents to monitor governmental expenditures more carefully. 4 This is similar to
our result; we find that smaller towns are better able to capitalize the effects of regulation into
house prices, which leads to the prediction that smaller towns should regulate more.
1 Hamilton allows for renters, but they are taken to be indifferent to the regulatory policy because anincrease in wages is assumed to exactly offset any increase in rents.2 Hamilton does not allow the land-use regulation to also create an amenity for residents, and there are nohousing or other features to distinguish towns. Thus the exact nature of the downward sloping housingdemand function by new residents is never specified.3 All towns are assumed identical, so all towns choose the same size in a Nash symmetric equilibrium. Anasymmetric Nash model is not investigated.4 Hoyt assumes that politicians want to overspend, and his model predicts that they will be allowed to do somore in larger towns.
5
2.2 Empirical Literature
There are only two articles that attempt to directly test the predictions of Hamilton and
Brueckner, that in metropolitan areas with only a few large jurisdictions, land use regulations
should be more prevalent and hence house prices higher. Ozanne and Thibodeau (1983) and
Evenson (2002) do find a significant relationship between home prices and governmental
fragmentation. However, the sign of the relationship in each case is exactly opposite of that
predicted by Hamilton’s model. They find that areas with more small governments tend to have
higher house prices, and more inelastic housing supply, suggesting that smaller governments
regulate more.
There is a longer literature, which attempts to more directly link some measure of land-
use restrictions to housing market outcomes (Malpezzi, 1996). Glaeser and Gyourko (2002)
perform a cross-market analysis on individual house prices, concluding that house prices
generally are not high relative to the replacement costs of construction in most parts of the U.S.
In the areas that are exceptions, they find that land-use regulation is a very important in
determinant of house price levels. This conclusion however is derived with a simple cross-section
equation using a metropolitan wide measure of land use regulatory delay. This measure of
regulation comes from a largely undocumented and limited survey of local governments. Other
similar studies, such as those of Black and Hoben (1985), Segal and Srinivasan (1989) and Mayer
and Somerville (2002), also suffer from using measures of local regulations that are not clearly
defined and carefully calculated.
A final strand in the empirical literature deals with trying to explain which governments
are more likely to enact such regulations. Bates and Santerre (2001) use data from the state of
Connecticut to estimate the demand for public open space. They find that private and public open
space are not close substitutes. Hsieh, Irwin and Libby (2001) estimate the effects of local rural
zoning regulation on land use and urban land conversion across Ohio. They find little effect, and
conclude, “rural zoning in Ohio is not designed to prevent conversion [of rural land to urban
land]; it is more a matter of guiding development, and most local zoning assumes that when the
demand is there, variance will be granted.”5 This implies that zoning regulations and land-use
could be jointly and endogenously determined, casting doubt about simple cross section OLS
results.
5 Hsieh, Irwin and Libby (2001), pg. 13.
6
3. The Impact of Land-Use Regulation on House prices
To examine how land use regulations impact house prices, we develop a static model that
is similar to those in the spatial economics literature. Jurisdictions compete in a Tiebout (1956)
type of environment with perfect mobility between towns. House prices are a “cost” and equalize
the utility of future residents across towns. The jurisdictions form a closed housing market so that
the (common) level of utility is endogenous. Rather than use Brueckner’s assumptions, and have
household land consumption be fixed while town land area is variable, we do the reverse.
Following Hoyt (1999), town land area is fixed and household land consumption is variable. The
future population of the market as a whole is given and the distribution of this population across
towns determines the density of development in each town (private land consumption). Land can
be restricted from development into what we call public land. Public land increases utility
directly, but also lowers private land consumption and hence raises land prices in equilibrium. A
simplification we use throughout is that private land consumption and housing consumption are
the same. The notation we use is as follows:6
xi, hi, ai : per-capita consumption of other goods, private land (housing), and public land
by residents of town i.
pi , ? : price of private land (housing) and the fraction of that price that it costs town i to
set aside public land.
y, N : (common) income level and MSA population.
ni, Li : population and total available land area of town i.
Individuals have preferences over the consumption of private land as well as public land:
),,( iiii ahxUU ? . (1)
For expositional purposes and ease of computation, we assume that the specific functional form
of the utility function is
? ?? ?? ?22 )( iiiiiiii ahahpyaxhU ?????? ? (2)
because the individual’s budget constraint is
)( iiii ahpxy ????? ? . (3)
Public land consumption achieved through regulation, and as such costs some fraction ? of its true
opportunity value (the price of private land consumption). The price of x is normalized to 1.
6 For the remainder of the paper we use lower-case letter to signify individual-level variables and upper-case letters to signify town-level variables. Despite calling it public land, we treat regulation as a privategood in this paper.
7
For efficient resource allocation, the derivative of Ui in (2) with respect to hi must equal
zero, which sets the equilibrium price at
? ?iii ah
yp
????
?2. (4)
Without loss of generality, it is useful to assume that each town sets ai by choosing a
share si, 10 ?? is , of its open land to restrict. Then
iiii Lsna ??? (5)
Land that is not regulated to be public is devoted to private consumption (housing) so that:
iiii Lahn ?? )( (6)
Combining (2), (4), (5) and (6) we can determine a town’s utility as a function of the town’s
population, as well as its decision over the public-private share of land,3222
)1(2)1(
)1()1(2
)1(???
????
????
????
?
???????
????
????
? ????
????
?????
?
????
?????
???
i
i
ii
ii
i
ii
i
ii
ii
ii n
Lss
ssyn
LsnL
sss
syU
??. (7)
In (7), utility is single peaked in the share of land devoted to public use and strictly
decreasing in the number of residents that develop the remaining private land. The total
population of the metropolitan area is fixed at N. Each town realizes an equilibrium population
nj, which is the result of individual location decisions. The equilibrium number of residents in
each town will be realized when residents are indifferent across locating in any town. Therefore,
each town must satisfy an equal-utility constraint *UU i ? , where U* is the market-wide level of
utility. As a town increases si within the range where si improves utility, more residents will move
in, prices will rise and private land consumption will fall so as to equalize utility with that of the
other towns in the market. The converse holds if increases in si reduce town utility. Considering
two towns, i and j, and incorporating (7), we get:
322322
)1(2
)1(
)1(2)1(
???
????
????
?
???
?
????
?????
?
????
????
????
?
??????
j
j
jj
jj
i
i
ii
ii
n
L
ss
ss
nL
ssss
??. (8)
Solving (8) for ni yields
???
????
????
????
?
????
?????
????
?
??????
?
?
j
i
jj
jj
ii
iiji L
Lss
ss
ssss
nn3
1223
122
)1(2
)1(
)1(2)1(
??. (9)
8
If the market consists of only two towns with a fixed aggregate population, it must be the
case that ij nNn ?? . If the total land area in the housing market is fixed at L, we can also
say ij LLL ?? Substituting in and solving (9) for ni yields
31
22
)1(2)1(
,)( ?
??
????
?
??????
?????
???
ii
iii
jjii
iii ss
sszwhere
zLzzLzLN
n?
. (10)
Substituting (10) into (4)-(6) determines the system wide level of utility as a function of
the regulatory choice, si, and sj.
? ?33
* )( jjii zLzzLNy
U ???????
???? (11)
Figure 1 shows the relationship between system wide U* and each town’s open space
decision si, plotted for towns of various sizes.7 Interestingly, system wide utility is maximized at
a share of public land that is common to all towns, regardless of size. This also is evident from
(11), wherein utility is maximized with respect to si when iz is so maximized and from (10), iz is
both independent of town size and varies only with si.
Using (10), we solve for each town’s equilibrium land price, pi*, as a function of the share
of open space in both towns.
? ?? ?jjiiii
ii zLzzLss
zNyp
??????????
?)()1(2
*
?(12)
Figure 2 depicts how town land prices vary with si when the other town’s share is held at
the common utility maximizing solution. The most important observation is that when open space
is at the system wide utility maximizing solution, land prices are not maximized. Since the
market-wide level of utility is endogenous in this model, there is no reason why consumers would
select si to maximize land prices as Hamilton (1978) postulated. This is because the price of land
represents both the cost of private housing (land) as well as the cost of providing public open
space. The ratio of these two costs is given by the parameter ?. As ? increases, the utility-
maximizing share of land devoted to open space declines. Prices are actually well behaved (i.e.
9
single peaked) over the full range of si for every value of ? save zero. If it costs literally nothing to
set land aside, prices will go to infinity as all land is devoted to public use. Devoting all land to
open space is certainly not utility maximizing. In the case of ? = 0 the system utility is highest
when the share of public land is 2/3, which is not that different from its value in Figure 1 (0.58
when ? = 0.50).
This static, one-period model exhibits little explanatory power with reference to the
enactment of land use restrictions. While its assumptions seem quite reasonable, towns of
different size, and wealth all select the same share of open space. The perfect mobility of
residents, which allows the equal utility condition to be satisfied, causes each town to choose an
identical optimum. Prices do respond differently in larger as opposed to small towns, but
residents are, as they should be, maximizing utility and not prices. In order to make the model
generate different behavior, we introduce a second type of consumer – existing residents - who
unambiguously benefit as land prices increase in the manner described above.
4. The Regulation Decision by Existing Residents
The existing residents of a town already own some amount of land (interchangeably
housing) in the town. This entitled group, lives only in a first period, and makes an irrevocable
decision about how much of the remaining land will be preserved forever – as opposed to being
developed in period 2. Note that this decision has no direct impact on them – since the land is
undeveloped in period 1, regardless of how they zone it. The decision does impact future prices,
which when discounted turn into today’s prices. This is the only economic link between the
residents of each period. Were it the case, as in overlapping generations models, that current
residents care about the utility of future residents, then the model would be considerably more
complicated. Here they are concerned only about how their decisions impact their own wealth.8
Although we allow existing residents to have the same utility function as future residents,
their consumption levels will be quite different. The private land (or housing) consumption of
existing residents is fixed at hio; as is current town population at nio. Town land area, Li, now
represents total undeveloped land, or land that existing residents can decide to allocate to either
public or private use. For simplicity, we assume that the first period is relatively short, so the
regulatory decisions made by existing residents do not take effect until the second period.
7 The assumptions used to make Figures 1 – 7 are given in Table 2.8 In the 1990 census slightly less than half of urban adult residents report living in the same state they wereraised in. Presumably this fraction would be much lower if the question were phrased as the samemetropolitan area or housing market. Thus for the majority of adult voters, their land use regulatorydecisions will not directly impact the welfare of their offspring.
10
Alternatively, the per-capita public land consumption of existing residents (aio) does not depend
on their choice of si since during the first period it will be vacant regardless of how it is zoned.
Thus for current residents aio = Li / nio.
Finally, we allocate some of the costs of land-use regulation to existing residents as well
as future residents. It is reasonable to assume that the initial costs of public land acquisition are
bonded and then tax financed. Using the same notation as in the previous section, future residents
pay share ? and current residents pay share 1-? of the costs of open space acquisition, pi si Li ( = pi
ai ni). The assumption that the cost shares sum to one is for convenience only – the results hold if
they sum to less than one as well.
While current residents have the same utility function as future residents, their budget
constraint is very different. Existing residents have already purchased land (housing), so the
future price of private land (multiplied by the units of land, hio, they own) enters positively into
their utility function. Their share of the total cost of open space acquisition is converted into a
per-capita amount by dividing by nio. Denoting current resident variables with an o subscript,
utility is:
? ?? ?220 )
)1(( ioio
io
iiioiiioioio ah
nna
hpyahxU ???
????
? ????????
?(13)
from the budget constraint:
ioio
iiioi x
nna
hpy ??
????? ))1(( ? . (14)
Substituting equation (12), in which land prices were determined by future residents, into
(13) and at the same time allowing aio = Lio / nio yields existing resident utility solely as a function
of the land-use decisions made by each town, si and sj.
? ?? ? ? ?2
)1()()1(2 ??
?
????
????
????
????
????
? ?????
??
????
?
??????????
??io
iio
io
iiio
jjiiii
iio n
Lh
nLs
hzLzzLss
zNyyU
??
.
(15)
Several features of this expression warrant discussion. First, if current residents bear none
of the costs of public land acquisition, i.e. ? = 1, the utility of existing residents is maximized by
maximizing future land prices. While this is consistent with Hamilton (1978), it is difficult to
imagine a realistic situation in which land set-aside is truly free to those undertaking it. Second,
existing residents in towns of different sizes (in this case, different amounts of open space, Li)
11
will attain their maximum utility at different levels of si. Finally, as reflected in the land-price
equation, town i’s choice of si will depend on town j’s choice of sj. This implies that the Nash
equilibrium will be generated from reaction functions, and that the solution is likely to be
asymmetric if towns are different.
Figure 3 depicts existing residents’ utility as a function of si assuming ? > 0. Initially,
utility rises since future prices will equal zero if the town sets aside no open space. As si
increases, land prices increase. However, the per-capita cost of open space acquisition will also
increase. At some point, the costs of acquisitions will dominate the benefits from capital gains,
and utility will begin to decrease in si. Utility continues to fall up to the point at which prices also
begin to decrease in si (Figure 4). At this point, utility begins to rise again as the falling prices
ease the costs of acquiring additional open space. At these relatively high levels of regulation, the
gains from cost mitigation dominate the loss of capital gains. Utility at si = 1 must equal that at si
= 0 because zi again equals zero; hence, there must be a single interior maximum that solves the
following first order condition.
???
????
? ?????
??
????
????
? ???
io
iiio
i
i
io
ii n
Lsh
sp
nL
p)1()1( ??
. (16)
Since utility is uniquely maximized at a positive value, equation (15) implies that at that
value ???
????
? ????
io
iiio n
Lsh
)1( ?must also be positive. Therefore, from (16) land prices are still
increasing in si at the utility maximizing value. At the optimum, an increase in land-use regulation
will always raises land values.
The utility maximizing solution to (15) gives a value of si that depends on the value of sj
and the other model parameters. The most important parameters define existing town
characteristics: land area per current resident (Li / nio), town size (Li relative to L, or nio relative to
N), and existing residential density (hio). Each of these influences the reaction functions, defined
as si as a function of sj. We next present a series of figures that illustrate how the reaction
functions, and hence the Nash equilibrium, shift in response to a change in these town-level
parameters.
To determine the symmetric Nash Equilibrium, we treat the metropolitan area as
composed of L / Li (or N / nio) identical towns. The equilibrium is defined by the combination of
equations (8)-(9), the identity ?? ji n -N n , and the assumption that all other towns select a
12
common sj = si. Note that equations (10)-(12) remain exactly the same under this equilibrium as
under the asymmetric equilibrium. This implies that we can treat all other towns save the one
under consideration as a composite. Each town individually selects si assuming that all other
towns choose sj (which is not necessarily the same as si). The Nash equilibrium then involves
solving one town’s reaction function subject to the added constraint that sj = si. . Graphically, we
look for the intersection of the reaction function with the 45o line.
To determine the asymmetric Nash Equilibrium we examine the intersection of two
reaction functions; the second is simply the inverse of the first except that it is calculated
assuming different town characteristics (for example, Lj = L - Li in the case where the town sizes
differ). As shown under the symmetric equilibrium, the second town may be either an individual
town or a composite (but different) town.
Town Land Area Per Capita: Li / nio
Figure 5 9 depicts the reaction functions showing the choice of open space by town i as a
function of the share chosen by town j. These reaction functions are quite flat over most of the
possible values of sj. Each schedule reflects a different assumed value of the town’s land area per
capita, Li / nio. A larger Li / nio implies that the cost of regulation is greater. When a town has
more open land per capita, each existing resident must purchase more actual acreage for a given
si. This leads them to set aside less open land for public use. Thus, the reaction functions for
towns with increasing open land per capita are progressively lower in Figure 5.
The symmetric Nash Equilibrium involves the intersection of one town’s schedule with
the 45o line. The relevant schedule is lower when the town has more open land per capita. This is
tantamount to having either fewer people or more open land in the metropolitan area as a whole,
and will result in less land being optimally regulated throughout the market. In an asymmetric
metropolitan area composed of one town with a lot of land per capita and one town with only a
little land per capita, the former town will set aside less land and the latter will set aside more.
Existing Residential Density: hio
The reaction schedules in Figure 6 reflect different assumed values of existing residential
density, hio. When existing residents consume more land, the capital gains from open space
creation become more important relative to the costs of purchasing that space. This shifts the
reaction schedule upward and leads them to set aside more open land for public use.
In a metropolitan area where current residents consume large amounts of land or housing
(lower density), the symmetric Nash Equilibrium involves the intersection of a higher reaction
9 As noted above, the assumptions used to make Figures 1 – 7 are given in Table 2 and representativesolution values in Table 1.
13
schedule with the 45o line. Hence, more open space is set aside as public land. In an asymmetric
metropolitan area composed of one town with low density and one town with high density, the
former will set aside more land.
Town size (Land Area): Li
To examine a change in Li holding the ratio Li / nio fixed, we must also allow a change in
current town population. Thus, the reaction schedules in Figure 7 reflect different town sizes in
terms of both open land area and population. In widespread simulations larger towns always have
lower reaction schedules, although an exact explanation for this is hard to come by. All we know
from (15) is that it has something to do with the observation in Figure 2, that prices react
differently to land use decisions in small as opposed large towns. Thus a metropolitan area
composed of many small towns will have more open space set aside for public use, while an area
composed of only a few larger towns will have less. In an asymmetric metropolitan area
composed of one larger town and one smaller town, the larger town will set aside less land and
the smaller town more.
The conclusions from these simulations seem to hold over a wide range of parameter
values. The one exception involves the impact of town size when current residents pay little or
nothing for open land. When ? is very close or equal to one, current residents will choose a value
of si that maximizes future land values. While the land values given in (12) are still well behaved,
larger towns seem to have monopoly power over smaller ones, inducing them to select greater si
than smaller towns. For values of ? < 0.8572, the reaction schedules in Figure 7 hold.
5. Empirical Tests
5.1 Data 10
To see if the predictions of the model hold, we were able to acquire a unique new dataset
on land-use and its regulation in Massachusetts. These data were created by the Massachusetts
Executive Office of Environmental Affairs in conjunction with the state’s 13 Regional Planning
Boards. Each of the 351 towns in Massachusetts worked with consultants in one of these Planning
10 “These digital data represent the efforts of the Massachusetts Executive Office of Environmental Affairsand its agencies to compile or record information from the cited source materials. EOEA maintains anongoing program to record and correct errors in these data that are brought to its attention. EOEA makes noclaims as to the absolute validity or reliability of these data or their fitness for any particular use. EOEAmaintains records regarding all methods used to collect and process these digital data and will disclose thisinformation upon request.” (MAEOEA)
14
Boards to create a standardized set of three Geographic Information Systems (GIS) maps and
supporting data describing current and potential land use.11
Current land-use data were created by interpreting aerial photographs taken of the state in
1999. First, all absolute constraints to development, such as land that is already developed, land
that is permanently protected by federal, state and local legislation, or land that is devoted to
cemeteries, transportation facilities or power lines, were identified by town. Second, partial
constraints to development, such as wetlands or land requiring special permits, were identified.
Third, each town’s zoning codes were added to the dataset so that the number of additional
residents, students, and housing units that can be added to the town (given current land-use) could
be determined. These are referred to as the town’s buildout measures.12 We use housing-unit
buildout to estimate land-use regulation.
The dataset used in this study is restricted to Massachusetts cities and towns included in
the Boston CMSA, with the exception of Boston itself for which data is not available. Dropping
Boston proper should not affect our analysis of land-use decisions since Boston has virtually no
open land which is subject to the land-use decision. The aerial photographs show that Worcester,
the second largest town in the state, also has no land on which new housing can be built given
current land-use laws. These two cities make virtually no decision about how much new building
to allow, and so are dropped from the analysis.13
Another 18 towns were also dropped from the analysis because their data were not
consistent.14 Aside from the fact that many are south of Boston, the characteristics of these towns
are fairly similar to the characteristics of the sample in general.15 An additional 21 towns analyze
not only new building, but proposals for redevelopment resulting in new residential building as
well. On average, these towns are closer to Boston (14 miles out as opposed to 27), smaller (10.5
square miles as opposed to 19) and more densely populated (approximately 6400 people per
square mile as opposed to 1850) relative to the sample in general. Since these towns are a priori
different we allow them to have a unique effect on regulation in the analysis.
11 A map of the 351 towns in Massachusetts is given is Appendix 1. The EOEA’s detailed description ofeach map type and a sample map is reproduced in Appendix 2.12 Under the EOEA definition, buildout is defined as the number of units that can be built on open landunder current land-use laws equaling zero. This definition does not allow for demolition and densityincreases.13 It is possible that a very small decision may be made, such as not to build housing on current golfcourses. However, parks and other public open space are not included in the regulatory decision, makingthe assumption of no decision probably fairly accurate.14 Abington, Amesbury, Berkley, Carver, Chelsea, Dartmouth, Dighton, Franklin, Grafton, Halifax,Hanson, Kingston, Mansfield, Marion, Mattapoisett, Norton, Plympton, and Winthrop.15 Including a dummy variable indicating towns in Plymouth County, which contains most of the droppedobservations, does not affect the results significantly.
15
The final dataset includes all towns in Eastern Massachusetts with the exception of Cape
Cod and the islands, several southern towns (included in the Providence, Rhode Island MSA
rather than the Boston CMSA) and the exceptions noted above. The geographic area of the
sample relative to the rest of the state is shown in gray in Appendix 1, with towns that were
dropped from the sample blacked out. Holland is the furthest west at 61 miles out from Boston.
Add in protected land data.
Share of Land Regulated: si
The total amount of open land on which housing could theoretically be built, Li, is
estimated for each community using actual land-use data as provided by MassGIS (the
Massachusetts Office of Geographic and Environmental Information).16 Li includes land used for
intensive or extensive agriculture, forests, nurseries or orchards; land devoted to participation
recreation such as golf and tennis courses; and open land with the exception of urban public land,
power lines, and cemeteries.17 In removing urban public land, we implicitly assume that areas
such as public parks will never be opened to development, although this assumption is not critical
to the results.
The EOEA data provides an estimate of the amount of land open to future residential
development under current land-use laws, Hi, exclusive of road frontages and other development
requirements.18 Subtracting this from the total amount of open land that could be built on gives an
estimate of total restricted open land in the town,.19 The share of open land restricted, si, is created
as defined by equation (5), so. As the share of land restricted to development increases, land-use
regulation increases.
This construction implies that the final measure of share regulation includes two types of
restricted land. First there is land that is technically buildable, but upon which residential use is
prohibited. This includes golf courses, playgrounds, orchards, agriculture or conservation land.
Second, there is land that is both technically buildable and available for development, but which
is restricted to account for development requirements (roads or other rights of way) or
geographical features which restrict building (steeply sloped land or wetlands).
Model Parameters: Li / nio, hio, and Li
16 This data is interpreted from the same aerial photographs as the buildout data, but actual land-use iscategorized more coarsely.17 A complete list of land-use categories is given in detail in Appendix 3.18 Hi, a town-level variable, equals hI ? ni. Each town creates its own estimate of HI to ensure accuracy,guided by EOEA employees to ensure consistency and comparability.
16
The measure of open land per capita, Li / nio, is created as open land available for
residential use divided by the number of existing housing units in the town, taken from 2000
Census data.20 The measure of existing residential density, hio = Hio / nio, is created as the
amount of land currently devoted to residential development, estimated using the MassGIS land-
use data, divided by the number of existing housing units in the town in 2000. The variable
measuring town size, Li holding Li / nio constant, may be measured either in terms of open land
area (Li) or population (nio).
It is important to note that the open land area variable is, by construction, positively
related to the share of open land regulated. As such the estimated relationship between the two
will reflect both a causal and a mechanical relationship. The effect of the mechanical relationship
will be to bias against finding that smaller towns regulate more (as Li increases, si decreases) as
predicted above, and bias toward finding Hamilton's result that larger towns regulate more. In
order to isolate the causal relationship, we use nio to measure town size and instrument for total
open land in the measure of open land per capita with the town’s distance from Boston as the
crow flies.21 These two variables are over 60% correlated, but distance from Boston has been
established since 1920. In fact, more than 90% of the towns in Massachusetts were established in
their current form by 1900, so distance from Boston has almost certainly been unaffected by
current land-use decisions.22
5.2 Summary Statistics
Summary statistics for each variable are presented in Table 3. An average town controls
11 square miles of open space, of which approximately 65% is regulated. This latter may be
slightly high since we are not able to distinguish forested land that is protected from forested land
that is considered open, leading us to overestimate si.23 Plymouth has the most open land, nearly
65 square miles, while Nahant has the least after Boston and Worcester have been dropped.
Quincy regulates nearly all of its land, Wareham almost none.24 Figure 8 shows the distributions
19 Ai, also a town-level variable, equals aI ? ni.20 We use population and housing units interchangeably. Results using population available upon request.21 These results do not change significantly if the total land area of the town is used as an instrument.22 Of these, 18.2% were established in their current form by 1800. By comparison, the first comprehensivezoning regulations were established in New York City in 1916, and recognized as lawful by the SupremeCourt 10 years later. (Wheaton (1993)).23 Taking all land held in perpetuity out of the open space dataset does not, for the most part, significantlychange the results. Taking out all forestland leads to a result of Hi > Li in a majority of the towns, which isclearly incorrect. These results are available upon request.24 The high regulation in Quincy stems from the fact that much of the western part of the city is forestedland that cannot be built on.
17
of these two variables across the sample. Most towns have a relatively smaller amount of open
land and a relatively larger share of open land restricted.25
The average town currently supports 1,839 houses per square mile of residentially
developed land. The town with the fewest housing units is Oakham, a secluded town in central
Massachusetts. Cambridge has the most housing units of the cities included in the dataset, nearly
45,000. 26 The number of housing units is 55% correlated with open land per capita, of which
Oakham again has the most and Cambridge has the least after only Everett and Somerville.
5.3 Results
The model generates three specific predictions about the relationship between town
characteristics and land-use regulation: (1) towns with more open land available for future
development per current resident should choose to restrict less land from development, (2) towns
in which existing residential density is high should choose to restrict less land from development,
and (3) towns that have a larger share of regional population or land should restrict less land.
Table 4 shows the results of a log-log regression with the share of open land restricted as
the dependent variable.27 All the coefficients have their predicted sign and only the coefficient on
larger towns – those with more housing units – is not significant statistically. It must be
remembered that by construction, our measure of the share of open land restricted is positively
correlated to the amount of open land per capita, so our analysis could be biased against finding
the expected result. The significant coefficients are approximately half again larger when the
mechanical relationship between the share of open land restricted and open land per capita is
instrumented for with distance from Boston.28 However, their statistical significance is not
different from the simple regression results.
To interpret the magnitude of the coefficients, the log-log formulation allows us to work
with elasticities remembering that the average town has 65% of its land restricted. Average open
land per capita is .003 square miles. Slightly more than doubling this is represents a 1 standard
deviation increase and would lower the restricted share of this land from 65% to 45%. Similarly,
a standard deviation increase in residential density, from the mean of 1839 units per square mile
25 The distribution of the share of open land per capita is even more highly skewed to the left than is totalopen land.26 Boston and Worcester have 251,935 and 70,723 housing units, respectively.27 The log-log specification was chosen because letting a percentage increase in land lead to a percentagedecrease in the share of land regulated seems more reasonable a priori than letting a unit increase in landlead to a unit decrease in the share of land regulated. All the regressions in this paper use the log-logspecification. Results from levels specifications are available upon request.28 If total land area is used as an instrument rather than distance to Boston, the coefficients fall slightly.
18
would reduce the restricted share from 65% to 39%. Finally, if the average town size of 8303 is
doubled (again about a standard deviation) the restricted share drops from 65% to 62%.
If towns are operating anywhere near the equilibrium levels of open space share, equation
(15) implied that increasing that share should yield higher house prices. Furthermore any
difference between towns that generates a greater set-aside share (e.g. lower residential density)
should also yield higher house prices. Table 5 presents the result of two regressions between
house prices and the land set aside share, when the other determinants of the share choice are
variously controlled for. House prices are measured as the average selling price of a single-family
home in 2000.29 The results further support the theory as increasing the restricted share of open
land always increases prices, regardless of other controls. Similarly, the three variables that
determine this share also have the expected effects. More total open land per capita and higher
residential density both reduce house prices. Finally, greater total open land (alternatively
population) also decreases house prices. However, this result goes away once the other if open
land per capita and current observed density are included in the specification. The weakness of
this result matches that in the share regression (Table 4. 30
6. Implications and Conclusions
We have developed a model that produces a number of predictions about what we should
observe in local land-use regulation.
1). Symmetric Nash solutions strongly suggest that more fragmented metropolitan areas
(areas with many small jurisdictions) should impose more land-use regulation (public land set
aside from development). This in turn will lead to higher land prices.
2). The symmetric solutions also suggest that metropolitan areas with low residential
density or high total open land per capita should impose less land-use regulation.
3). In an asymmetric solution, larger towns, towns with a lower residential density and/or
towns with more overall open land per capita within a metropolitan area should impose less land-
use regulation.
Further, we have argued that any model supporting these conclusions must play off
differences in the incentives faced by two types of agents: those who own existing land and are
concerned with capital gains, and those who do not and so regard capital gains only as an increase
in the cost of housing. The latter agents must be part of the model in order for prices to be well
29 Data from http://www.dls.state.ma.us/allfiles.htm30 This effect remains even controlling for distance to Boston. In this case, the coefficient is –0.22 with astandard error of 0.05.
19
determined. Without the former agents, it is difficult to obtain a very rich model in terms of how
towns select the amount of regulation.
We then use a unique dataset on residential zoning and development in the state of
Massachusetts to test the predictions of the model. These data give direct measures of regulation
that are much more precise than any found in the previous literature. In examining the restriction
of land available for housing development, we found strong evidence for two of the three
predictions. First, if the cost of a given share of regulation increases through an increase in the
open land per capita, ceteris paribus, a town will choose to regulate less. Second, if the benefit of
regulation increases through an increase in current land consumption per capita, ceteris paribus, a
town will choose to regulate more. Finally, we found only insignificant evidence to suggest that
small towns are able to capitalize the effects of land-use regulation more fully than large towns
(implying that small towns should restrict the share of open land more). Thus, we cannot say with
any confidence that the data support either our theory or Hamilton’s.
20
ReferencesBates, Laurie J. and Rexford E. Santerre. (1994). “The Determinants of Restrictive Residential
Zoning: Some Empirical Findings,” Journal of Regional Science 34(2), 253-63.
Bates, Laurie J. and Rexford E. Santerre. (2001). “The Public Demand for Open Space: The Caseof Connecticut Communities,” Journal of Urban Economics 50, 97-111.
Black, J. Thomas and James Hoben. (1985). “Land Price Inflation,” Urban Geography 6(1), 27-49.
Brueckner, Jan K. (1996). “Modeling Urban Growth Controls,” University of Illinois at Urbana-Champaign College of Commerce and Business Administration Working Paper.
Brueckner, Jan K. (1998). “Testing for Strategic Interaction Among Local Governments: TheCase of Growth Controls,” Journal of Urban Economics 44, 438-467.
Chambers, Daniel N. and Douglas B. Diamond, Jr. (1988). “Regulation and Land Prices,” Paperread at the American Real Estate and Urban Economics Association Midyear Meeting,Washington D.C., June.
Epple, Dennis and Thomas Romer. (1991). “Mobility and Redistribution,” Journal of PoliticalEconomy 99(4), 828-58.
Epple, Dennis and Glenn J. Platt. (1998). “Equilibrium and Local Redistribution in an UrbanEconomy when Households Differ in Both Preferences and Incomes,” Journal of UrbanEconomics 43(1), 23-51.
Evenson, Bengte (2002). “Understanding House Price Volatility: Measuring and Explaining theSupply Side of Metropolitan Area Housing Markets,” Unpublished Paper.
Fair, Ray C. (1972). “Disequilibrium in Housing Markets,” Journal of Finance 27(2), 207-221.
Glaeser, Edward L. and Joseph Gyourko (2002). “The Impact of Zoning on HousingAffordability,” Harvard Institute of Economic Research Discussion Paper Number 1948.
Greene, William H. (1997). Econometric Analysis, 3rd Ed. Upper Saddle river, NJ: Prentice-Hall.
Hamilton, Bruce W. (1978). “Zoning and the Exercise of Monopoly Power,” Journal of UrbanEconomics 5, 116-130.
Hendershott, Patric H. and Thomas G. Thibodeau. (1990). “The Relationship Between Medianand Constant Quality House Prices: Implications for Setting FHA Loan Limits,”AREUEA Journal 18(3), 323-34.
Hoyt, William H. (1999). “Leviathan, Local Government Expenditures, and Capitalization,”Regional Science and Urban Economics 29(2), 155-71.
Hsieh, Weh-hua, Eland G. Irwin and Lawrence W. Libby. (2001). “The Effect of Rural Zoning onthe Allocation of Land Use in Ohio,” Unpublished Paper.
21
Maisel, Sherman J. (1963). “A Theory of Fluctuations in Residential Construction Starts,” TheAmerican Economic Review 53(3), 359-383.
Malpezzi, Stephen. (1996). “Housing Prices, Externalities, and Regulation in U.S. MetropolitanAreas,” Journal of Housing Research 7(2), 209-241.
Malpezzi, Stephen (1999). “A Simple Error Correction Model of House Prices,” Journal ofHousing Economics 8, 27-62.
Malpezzi, Stephen (2000). “Tales From the Real Side: The Implications of Urban Research forReal Estate Finance in Developing and Transition Economies,” Unpublished Paper.
Mayer, Christopher J., and C. Tsuriel Somerville. (1996). “Land Development and the Supply ofNew Housing: Unifying Empirical and Theoretical Models of Housing Supply,” UBCCentre for Real Estate and Urban Economics Working Paper.
Mayer, Christopher J. and C. Tsuriel Somerville (2000). “Land Use Regulation and NewConstruction,” Regional Science and Urban Economics 30(6), 639-62.
Murray, Michael P. (1999). “Subsidized and Unsubsidized Housing Stocks 1935 to 1987:Crowding out and Cointegration,” Journal of Real Estate Finance and Economics 18(1),107-124.
Ozanne, Larry and Thomas Thibodeau. (1983). “Explaining Metropolitan Housing PriceDifferences,” Journal of Urban Economics 39, 434-41.
Rose, Louis A. (1989). “Urban Land Supply: Natural and Contrived Restrictions,” Journal ofUrban Economics 24(3), 325-45.
Segal, David and Philip Srinivasan. (1989). “The Impact of Suburban Growth Restrictions onU.S. Housing Price Inflation, 1975-78,” Urban Geography, 6(1) 14-26.
Shen, Q. (1996). “Spatial Impacts of Locally Enacted Growth Controls: The San Francisco BayRegion in the 1980s,” Environment and Planning B: Planning and Design 23, 61-91.
Thorson, James A. (1996). “An Examination of the Monopoly Zoning Hypothesis,” LandEconomics 72(1), 43-55.
Tiebout, Charles M. (1956). “A Pure Theory of Local Expenditures,” Journal of PoliticalEconomy 64, 416-26.
Topel, Robert, and Sherwin Rosen. (1988). “Housing Investment in the United States,” Journal ofPolitical Economy 96(4), 718-740.
White, Michelle J. (1975). “Fiscal Zoning in Fragmented Metropolitan Areas,” In Edwin S. Millsand Wallace E. Oates, eds., Fiscal Zoning and Land Use Controls. Lexington, MA:Lexington Books.
Wheaton, William C. (1993). “Land Capitalization, Tiebout Mobility, and the Role of ZoningRegulations,” Journal of Urban Economics 34(2), 102-17.
22
PLYMOUTH
OTI S
WARE
BARRE
BECKET
PETERSHAM
SAVOY
BOSTON
ATH OL
TAUNT ON
MONSON
COLRAIN
DARTMOUTH
NEW SALEM
CARVER
PERU
BARNSTABLE
WESTPORT
HOLDEN
BLANDF ORD
REHOBOT H
FALM OUTH
CONWAY
ORANGE
SUT TON
ASHFIELD
WESTF IELD
DISFIELD
MIDDL EBOROUGH
CHARLTON
DOUGL AS
PALMER
SANDWICH
CHEST ER
GROTON
GRANVILLE
RUTL AND
NANTUCKET
ROWE
BOURNE
WARWIC K
HARDWICK
HEAT H
HAWL EY
WINDSOR
IPSWICHASHBY
TOL LAND
EAST ON
WAREHAM
BELCHERTOWN
SPENCER
ROYALSTON
LAKEVILLE
GILL
FAL L RIVER
NORT ON
GRANBY
BRIM FIELD
WENDELL
LUDLOW
ANDOVER
PELHAM
ADAMS
HAVERHILL
OXFORD
STERLI NG
WINCHENDON
UPT ON
HADLEY
FREETOWN
WARREN
TRURO
PRINCETON
DEER FIELD
STURBRI DGE
STOW
WORCEST ER
ACT ON
ASHBURNHAM
ROCHEST ER
UXBRIDGE
SHARON
AMHERST
FL ORI DA
DENNIS
TOWNSEND
GTON
HARVARD
NORT HFIEL D
HUBBARDST ON
FRANKLIN
MONT AGUE
WESTF ORD
DUDLEY
TON
SHIRETEMPLETON
AGAWAM
MASHPEE
BILLERICA
DRACUT
SOUTHWI CK
SUDBURY
WEST MINSTER
CONCORD
SPRINGFI ELD
BOXFORD
OAKHAM
BOL TON
DIGHT ON
FIT CHBURG
HINGHAMGRAF TONHOLYOKE
NEWBURY
DUXBURY
HOPKINTON
LANCASTER
METHUEN
HARWICH
CANTON
GARDNER
ATTL EBORO
SWANSEA
LEOMINSTER
CHICOPEE
NORTHAMPT ON
LEYDEN
WALES
MARSHF IELD
LEVERET T
BREWSTER
LEICESTER
YARMOUTH
LUNENBURG
NATI CK
DOVER
NSDALE
QUINCY
WALPOLE
EDG ARTOWN
LYNN
NORWELL
WHAT EL Y
ESSEX
RAYNHAM
PEMBROKE
HAL IF AX
WORTHINGTONCHEST ERFIELD
GOSHEN
MENDON
ROWLEY
NEWTON
HUNT INGT ON
PEPPERELL
SHUTESBUR Y
HAMPDEN
RUSSELL
AUBURN
WESTON
FRAMINGHAM
MILLI S
SEEKONK
SHEL BURNE
WRENTHAM
BROCKTON
PHI LL IP STON
CHIL MARK
BRIDGEWATER
CHARL EMONT
PAXTON
ERVING
SOUTHAMPTON
SHIRLEY
PLAI NF IEL D
HANSON
MIDDLEFIELD
WILBRAHAM
BERLIN
BUCKLAND
BOYLSTON
KINGSTON
BERKLEY
MANSF IELD
MARION
LOWEL L
AYER
MILTON
WEST TISBURY
PEABODY
HATF IELD
LINCOLN
WESTHAMPTON
GREENFIEL D
ACUSHNET
SCIT UATE
WELL FLEET
BEVERLY
HOLLISTON
MILF ORD
WILLIAMSBURG
TEWKSBURY
MILLBURY
CUMM INGT ON
LI TTLETONCARLISLE
NO RFOLK
WAYLAND
HANOVER
NORTH ANDOVER
BERNARDST ON
SHREWSBURY
BELLINGHAM
WEYMOUTH
WEBST ER
ORLEANS
MEDF IELD
HAMI LT ON
FOXBOROUGH
LEXINGTON
SALISBURY
SHERBORN
DANVERS
BEDF ORD
EAST HAM
WOBUR N
DUNSTABLE
SOUT HBRI DGE
RTH ADAMS
MARLBOROUGH
HOLL AND
PLYMPTON
HUDSON
ASHL AND
WALTHAM
SAUGUS
NEW BEDF ORD
WESTBOROUGH
WILM INGT ON
STOUGHTON
BROOKFIELD
MEDWAY
NEEDHAM
BRAINT REE
NEW BRAINT REE
CHATHAM
DEDHAM
MIDDLETON
AMESBURY
NORT HBRIDGE
SALEM
MONROE
SOUT H HADLEY
TOPSFI ELD
MATTAPOISETT
READING
FAI RHAVEN
WEST BROOKF IELD
GL OUCESTER
NORT HBOROUGH
SUNDERLAND
NORT H BROOKFIELD
TYNG SBOROUGH
PLAINVILLE
MONT GOMERY
ABING TON
LYNNF IELD
NORW OOD
RANDOLPH
LARKSBURG
ROCKLAND
COHASSET
BURLINGTON
SOUTHBOROUGH
EASTHAMPTON
WEST WOOD
WEST SPRI NGFI ELD
WENHAM
WEST NEWBURY
WELL ESLEY
GEORGETOWN
BLACKST ONE EAST BRIDGEWATER
MEDFORDCLINTON
MERRIMAC
NORTH READING
NORT H ATT LEBORO UGH
AVON
WEST BOYLST ON
TISBURY
GOSNOLD
GROVELAND
SOM ERSET
WEST BRI DGEW ATER
REVERE
WHIT MAN
BOXBOROUGH
WAKEF IELD
HOLBROOK
LAWRENCE
LONGMEADO W
EAST LONGMEADOW
ROCKPORT
PROVINCETOWN
CAMBRI DGE
OAK BLUFF S
STONEHAM
GAY HEAD
EAST B ROOKF IELD
M ANCHESTER
MAYNARD
NEWBURYPORT
MIL LVIL LE
HULL
BELMONT
HOPEDALE
MELROSE
EVERETT
MARBL EHEAD
WATERT OWN
CHELSEA
SWAMPSCOTT
WINT HROP
NAHANT
Appendix 1: MassGIS Data Map
23
Appendix 2: GIS Map Details
Map 1: Absolute Development Constraints
Map 1 allows communities to see how they have used the land within their municipal
boundaries to date. It depicts, in varying colors, lands that have already been developed or
protected as well as lands tat are absolutely constrained to development. The colored data-layers
correspond to various reasons that land in the community is unavailable for future growth:
permanently protected open space, existing residential, commercial or industrial development and
state or federal regulations such as the Rivers Protection Act. The map also allows communities
to see the amount and location of remaining land that may be available for future development.
This available land for development appears in white on the map.
The first major category of land that is unavailable for future growth is land that is
permanently protected. Map 1 indicates, in varying shades of green, lands that cannot be
developed because they are either under a conservation restriction or are owned outright.
Protection could be attributed to a variety of entities including: an EOEA agency such as the
Department of Environmental Management or the Department of Food and Agriculture; a non-
profit conservation organization such as Massachusetts Audubon Society or the Trustees of
Reservations a local land trust; the community's Conservation Commission; or individuals. For
the purpose of the buildout analysis, lands held in Chapter 61 are not considered permanently
protected because they could be developed in the future. Golf courses were not considered
permanently protected unless future growth is prohibited by a conservation restriction.
The second major color grouping on Map 1 is pink and red. Lands depicted in pink on
Map 1 are lands that were developed in the community as of the time of the buildout analysis to
the best knowledge of the town and RPA planners. The amount of pink land on the map relative
to the overall area of the map is a rough indication of the extent of development in the
community.
Lands shown in dark red indicate the community's residential subdivisions approved
since 1990 (or in some cases 1985). Each dark red subdivision is numbered. These numbers
correspond to a Subdivisions Table on the map containing the year of the subdivision, its number
of units and its total acreage. Many tables may also include the number of units in the subdivision
that have been built as of the date of the buildout analysis.
In addition to the permanently protected and developed lands, other land areas that are
deemed not available for future growth are shown in different colors on Map 1. The gold
indicates miscellaneous undevelopable lands such as gas line easements, power transmission
24
lines, cemeteries, sewer pumping stations and corrections to the land use data. Gray indicates
lands that are unavailable for future growth due to the Rivers Protection Act. Since the Rivers
Protection Act does not allow development within the first one hundred feet from named
perennial streams (25 feet in some communities and Designated Densely Developed areas), this
area is shown as unavailable for future growth.
Depending on local regulations, land in flood zones or wetlands could also be depicted on
Map 1 as undevelopable by nature of being an absolute constraint. Some communities, however,
allow non-upland areas to contribute to an individual building lot thus allowing structures to be
built in flood zones as long as the structures comply with the Massachusetts Building Code and
do not alter flood levels.
Map 2: Developable Lands and Partial Constraints
This map shows the amount and location of land within a community that is available for
future development. It is the inverse of Map 1 -- showing, in color, the area that appeared on Map
1 as white. The varying colors are used to identify the particular zone in which land is classified.
Although color classifications vary, a common color palette used statewide generally depicts
residential zones in shades from yellow to orange to brown (to generally reflect density),
industrial zones in purple, and commercial zones in red. Hatched patterns indicate partial
constraints to development, such as wetlands, the second 100-foot buffer zone under the Rivers
Protection Act, or overlay districts, such as a water protection overlay district which limits
impervious surface, bans underground storage tanks and requires that a special permit be issued
for the storage of hazardous materials.
Map 3: Composite Map
The third map in the buildout analysis series is intended to simplify the information seen
on the first two maps by illustrating in purple all land that is available for future development, in
yellow all land which is unavailable for future growth by reason of current development or
permanent land protection, and in a transparent stipple dot pattern all land that is partially
constrained from development.
A Community Snapshot: The Orthophotograph Aerial Map
Each buildout map series includes an aerial photo of the community taken from
approximately 15,000 feet above the ground. This map, a picture of the community at the same
scale as the buildout maps, is useful for determining the relationship between current landmarks
25
and the land depicted on the other maps of the buildout series. It is also useful to examine the
patterns of subdivisions, buildings, power lines, roads and other geographic features, such as
recreational fields, water bodies, forests and farmlands, which are also readily apparent. The
regional planning agencies (RPA) and consultants who completed the buildout analyses also used
this map as a type of base map from which to derive new GIS layers and to check the accuracy of
other maps produced for the buildout series.
The Blueprint for Development: The Zoning Map
This colorful map shows a community's current zoning. Each zoning district in the
community is designated by a separate color that corresponds to a key on the map. Essentially, it
acts as a blueprint for a community's future development, showing how it has divided its land to
accommodate varied development interests. For example, the Zoning Map shows where the
community wants to place houses (i.e. residential) and businesses (i.e. retail,
commercial/industrial) and their relationship to each other. A number of communities have a
single zoning district, allowing many forms of land development to occur throughout the
community, while some have as many as 50 districts, separating certain uses.
Most of the buildouts include a GIS zoning map of the community. However, the map
was added to the buildout series after the first group of buildouts had been completed. The
Zoning Map was added because many communities lacked an accurate GIS based zoning map
and much of the buildout analysis is based upon zoning.
26
Example of Map 1: Zoning and Absolute Development Constraints, Cambridge, MA
FreshPond
A-2
C-1
B
B
OS
C
B
B C-3A-1
B
C-3
IB
B
B
PD6
OS
O-2
C-3
BC
PD2
B
OS
PD5
C-1
IB- 2
C-1
B
A-2
A-2
A-2
O-2
C3-B
C-1
B
BA
BA
IA
SD- 8
B
PD4
A-2
BA-2
BC
PD3
SD- 3
MXD
C-1
A-2
C-1
C-2
IB- 1
A-1
A-2 B
B
IA- 1
BB- CS O
CRDD
O-3A
B
SD- 6
B
C-2
C-2
OS
C-2
BA
C-2
C-1
OS
IA- 1
A-2
C-2
C-3
BB- HS H
O-2 -H SH
BA -1
C-1
C-1
C-2
BA
C-1
OS
BA -2
A-1
C-1A
BA
B
BB
OS
OS
SD- 2
IA- 1
C-2A
C-1
C-1
BA
C-1
SD-5
C-1
C-2
C-1
B
C-1
C-2B
SD- 9
C-1
OS
OS
SD- 1
BA-1
OS
C-2
C-2
SD-10
C-2
C3-B
IA- 2
BA
C-1
C-1
C-1
C-3
C-3
BB O-1
OSC-1
OS
IA
PD1
C-1
B
C-1
C-1
BB
C-1
OS
IB- 1
SD- 7
BB-1
C-1
C-1
C-3
C-2B
C-1A
C-1
OS
C-1
C-2B
BA
C-2
OS
C-2A
OSC-1A
OS
O-3-H SH
OS
O-2 -H SH
SD- 10
O-1
IA- 1
C-3
C-1A
SD-11
IA-2
PD1BA-2
OS
BA
BA
C-2
IA- 1
C-3
C-2A
O-3 -H SH
BA
OS
OS
C-2
BA
OS
C-2
BA -1
C-3
BA
O-1
C-1A
C-2
OS
OS
BA-1
O-2 -H SH
C-2B
BA -1
C-2
BA
IA- 1
OS
C-2
OS
C-2B
O-1
C-2
C-2B
OS
BA -1
C-1
OS
BB- 2
C-2A
SD- 11
C-2B
O-1
IA
BA-1
C-2
C-1
BA-1
BA -1
IA-1
BC -1
OS
BA
O-2 -HSHC-1
C-1
C-3
IB-1 BC
A-2
27
Appendix 3: Land-Use Categories and Open Space Measure31
LAND USE CODE DEFINITIONS
The two land use code items (LU21_CODE and LU37_CODE) in the polygon attributeand history tables represent two classifications of land use. The 21 category classificationaggregates the categories in the 37 category classification as follows:
CODE ABBREV CATEGORY DEFINITION1 AC Cropland Intensive agriculture2 AP Pasture Extensive agriculture3 F Forest Forest4 FW Wetland Nonforested freshwater wetland5 M Mining Sand; gravel & rock
6 O Open Land Abandoned agriculture; power lines;areas of no vegetation
7 RP ParticipationRecreation Golf; tennis; Playgrounds; skiing
8 RS SpectatorRecreation
Stadiums; racetracks; Fairgrounds;drive-ins
9 RW Water BasedRecreation Beaches; marinas; Swimming pools
10 R0 Residential Multi-family11 R1 Residential Smaller than 1/4 acre lots12 R2 Residential 1/4 - 1/2 acre lots13 R3 Residential Larger than 1/2 acre lots14 SW Salt Wetland Salt marsh15 UC Commercial General urban; shopping center16 UI Industrial Light & heavy industry
17 UO Urban OpenParks; cemeteries; public &institutional greenspace; also vacantundeveloped land
18 UT Transportation Airports; docks; divided highway;freight; storage; railroads
19 UW Waste Disposal Landfills; sewage lagoons20 W Water Fresh water; coastal embayment21 WP Woody Perennial Orchard; nursery; cranberry bog
22 - No Change Code used by MassGIS during qualitychecking
The additional categories in LU37_CODE are:CODE ABBREV CATEGORY23 CB Cranberry bog (part of #21)24 PL Powerlines (part of #6)25 RSB Salwater sandy beach (part of #9)26 RG Golf (part of #7)27 TSM Tidal salt marshes (part of #14)28 ISM Irregularly flooded salt marshes (part of #14)29 RM Marina (part of #9)30 - New ocean (areas of accretion)
31 http://www.state.ma.us/mgis/lus.htm
28
31 UP Urban public (part of #17)32 TF Transportation facilities (part of #18)33 H Heath (part of #17)34 CM Cemeteries (part of #17)35 OR Orchard (part of 21)36 N Nursery (part of #21)37 - Forested wetland (part of #3)
All land use categories were aggregated from 104 categories originallydefined in 1971. Further information on them can be obtained from ProfessorWilliam MacConnell at the Dept. of Forestry, University of Massachusetts,Amherst.
The open space measure used for the denominator of the share of restricted land includes
land used for intensive agriculture (1), extensive agriculture (2), forest (3), open land with the
exception of power lines (6-24), participation recreation (7), urban open land with the exception
of urban public space and cemeteries (17-31-34) and woody perennial land with the exception of
cranberry bogs (21-23).
29
Table 1: Numerical Simulations of The Land-Use Regulation Decision
Variables
Total Open Land in Town 1 (L1,fixed) 0.5 2 3.33 4 5 3.33Total Open Land in Town 2 (L2, fixed) 9.5 8 6.67 6 5 3.33Total Open Land in Town 3 (L3, fixed) 3.33
New Residents to be Housed (N, fixed) 10 10 10 10 10 10Income (y, fixed) 5 5 5 5 5 5Housing Consumption of Current Residents (hio, fixed) 0.25 0.25 0.25 0.25 0.25 0.25Current Population in Town 1 (n1o) 0.5 2 3.33 4 5 3.33Current Population in Town 2 (n2o) 9.5 8 6.67 6 5 3.33Current Population in Town 3 (n3o) 3.33Proportion of Regulation Costs which can be Shifted to New Residents (?, fixed) 0.5 0.5 0.5 0.5 0.5 0.5
Outcomes
Equilibrium Price in Town 1 (p1*) 6.52 3.86 3.22 3.02 2.81 2.88
Equilibrium Price in Town 2 (p2*) 2.43 2.48 2.56 2.66 2.81 2.88
Equilibrium Price in Town 3 (p3*) 2.88
Equilibrium Number of New Residents in Town 1 1.11 2.66 3.75 4.26 5 3.33Equilibrium Number of New Residents in Town 2 8.89 7.34 6.25 5.74 5 3.33Equilibrium Number of New Residents in Town 3 3.33
Utility of New Residents (Ui) 0.007 0.027 0.039 0.043 0.045 0.059360
Utility of Current Residents in Town 1 (U1o) 1.493 1.403 1.383 1.378 1.374 1.367Utility of Current Residents in Town 2 (U2o) 1.387 1.374 1.372 1.372 1.374 1.367Utility of Current Residents in Town 3 (U3o) 1.367
Optimal Share Restricted Land in Town 1 (s1*) 0.20 0.18 0.17 0.16 0.15 0.17
0.66Optimal Share Restricted Land in Town 2 (s2*) 0.05 0.10 0.12 0.13 0.15 0.17
Optimal Share Restricted Land in Town 3 (s3*) 0.17
Total Optimal Restricted Land 0.57 1.15 1.39 1.45 1.48 2.32
30
Table 2: Assumptions Made in Figures
Figure 1,2 3.4 5 6 7
Total Open Land 10 10 10 10 10Open Land in Towns 0.5-9.5 0.5-9.5 2,8 2,8 2-8
New Residents to be Housed (N) 10 10 10 10 10Income (y) 5 5 5 5 5Housing Consumption of Current Residents (hio,) 0.25 0.25 0.25 0.25-0.75 0.25Total Current Population 10 10 10 10 10Current Population in Towns 0.5-9.5 0.5-9.5 0.5-8 2,8 2-8Proportion of Regulation Costs which can be Shifted to New Residents (?, fixed) 0.5 0.5 0.5 0.5 0.5
Outcomes
Equilibrium Prices 4.42 2.43-6.52
Utility of New Residents (Ui) 1.37-1.49
Utility of Current Residents 0.26
Optimal Share Restricted Land 0.58 0.05-0.20 0.02-0.18 0.12-0.81 0.1-0.19
31
Table 3: Summary Statistics
Variable Name N Mean
Standard
Deviation Min Max
Total Open Land (Sq. Miles) 191 11.05 9.22 0.14 64.82
Share Open Land Restricted 191 0.65 0.16 0.07 0.99
Open Land per Capita 191 0.003 0.004 0.000005 0.03Existing Residential Density 191 1,839 2,135 378 16,874Housing Units, 2000 191 8,303 8,505 591 44,725
Towns Allowing Redevelopment 18 0.09 0.29 0 1
Distance From Boston2 191 26.98 13.73 5 61
Notes:1 Data from Massachusetts EOEA and MassGIS2 Distance from Boston is calculated as the crow flies.
32
Table 4: Analysis of Share Open Land Restricted
Dependent Variable: Share Open Land Restricted
OLS IV
Open Land per Capita -0.15 -0.23 (0.04)*** (0.07)***
Existing Residential Density -0.22 -0.34 (0.11)** (0.12)***
Housing Units, 2000 -0.01 -0.04(0.05) (0.06)
Constant 0.19 0.76(0.56) (0.64)
N 191 191R2 0.12 0.34
Notes:1 All variables are measured in logs.2 Data from Massachusetts EOEA and MassGIS3 All regressions allow towns that include redevelopment in their analyses to have a unique effect on
regulation.4 *** denotes significance at the 1% level, ** denotes significance at the 5% level and * denotes significance
at the 10% level.
33
Table 5: Effect of Open Land on Average House Price
Dependent Variable: Average House Price 20004
OLS OLS
Open Land -0.09 -0.0007 (0.03)*** (0.04)
Share Open Land Restricted 0.30 0.22 (0.08)*** (0.07)***
Open Land per Capita -0.36 (0.07)***
Existing Residential Density -0.88 (0.15)***
Constant 12.49 16.12 (0.08)*** (0.68)***
N 185 185R2 0.17 0.56
Notes:1 All variables are measured in logs.2 Data from Massachusetts EOEA and Department of Revenue Division of Local Services3 All regressions allow towns that include redevelopment in their analyses to have a unique effect on
regulation.4 *** denotes significance at the 1% level, ** denotes significance at the 5% level and * denotes significance at the 10% level.5 Three of the 191 towns did not have data for single-family homes, so data for all residential sales were
used.
34
Figure 1: Relationship Between the Share of Land Regulated and Utility of Future Residents
0
0.05
0.1
0.15
0.2
0.25
0.3
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Share of Land Regulated
Util
ity o
f Fut
ure
Res
iden
ts Town Size = 0.5
Town Size = 2
Town Size = 4
Town Size = 5
Town Size = 6
Town Size = 8
Town Size = 9.5
35
Figure 2: Relationship Between Share of Land Regulated and House Prices
0
1
2
3
4
5
6
7
8
9
10
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Share of Land Regulated
Hou
se P
rice
Town Size = 0.5
Town Size = 2
Town Size = 4
Town Size = 5
Town Size = 6
Town Size = 8
Town Size = 9.5
36
Figure 3: Relationship Between Share of Land Regulated and Uitlity of Current Residents
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.2 0.4 0.6 0.8 1 1.2
Share of Land Regulated
Util
tiy o
f Cur
rent
Res
iden
ts
Town Size = 0.5
Town Size = 2
Town Size = 4
Town Size = 5
Town Size = 6
Town Size = 8
Town Size = 9.5
37
Figure 4: Relationship Between Share of Land Regulated and House Prices
0
2
4
6
8
10
12
14
16
18
20
0 0.2 0.4 0.6 0.8 1 1.2
Share of Land Regulated
Hou
se P
rice
Town Size = 0.5
Town Size = 2
Town Size = 4
Town Size = 5
Town Size = 6
Town Size = 8
Town Size = 9.5
38
Note: Assumes both towns are of size 5 with a current housing consumption of 0.25.
Figure 5: Reaction Functions Varying Town Land Area per Current Resident
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Share of Land Regulated in Town of Size 8
Shar
e of
Lan
d R
egul
ated
in T
own
of S
ize
2
Li/nio = 1
Li/nio = 2
Li/nio = 4
Li/nio = 1
Li/nio = 2
Li/nio = 4
39
Note: Assumes both towns are of size 5 with an open space per capita of 1.
Figure 6: Reaction Fuctions Varying the Housing of Current Residents
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Share of Land Regulated in Town of Size 8
Shar
e of
Lan
d R
egul
ated
in T
own
of S
ize
2
hio = 0.25
hio = 0.5
hio = 0.75
hio = 0.25
hio = 0.5
hio = 0.75
40
Note: Assumes both towns have an open space per capita of 1 and a current housing consumption of 0.25.
Figure 7: Reaction Functions Varying Open Land per Town
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Share of Land Regulated in Town of Size 6,7 or 8
Town Size = 1
Town Size = 3
Town Size = 5
Town Size = 5
Town Size = 7
Town Size = 9
41
Figure 8: Distribution of Open Land Available for Development and Share ofOpen Land Regulated
Den
sity
Kernel Density Estimateli
0 20 40 60 80
0
.02
.04
.06
Den
sity
Kernel Density Estimatesi
0 .5 1
0
1
2
3
Den
sity
si
Den
sity
Li
Related Documents
