Historic Preservation and Neighbourhood Change

Document Sample
Historic Preservation and Neighbourhood Change Powered By Docstoc
					Urban Studies, Vol. 41, No. 8, 1587–1600, July 2004




Historic Preservation and Neighbourhood Change

N. Edward Coulson and Robin M. Leichenko
[Paper first received, August 2003; in final form, November 2003]



Summary. Historical designation has become an important tool in efforts to revitalise central-
city neighbourhoods. Yet designation has also come under scrutiny because of its presumed
association with gentrification and displacement of lower-income residents. Using Fort Worth,
Texas, as a case study, the paper asks whether historical designation is associated with demo-
graphic change in neighbourhoods. It is found that historically designated areas started out with
slightly worse neighbourhood indicators than those without designation—a finding that is
consistent with the idea that preservation efforts are targeted to areas in ‘need’ of revitalisation.
However, we find no evidence that preservation efforts altered the demographic composition of
neighbourhoods. This finding runs counter to the notion that historic preservation is a precursor
to gentrification.


1. Introduction
Historical designation is a device that be-                     often thought to be one of the most draconian
stows recognition on particular properties be-                  of land use policies.
cause of their importance, in some great or                        Since by its very nature historical desig-
small way, to the history of the city or region                 nation focuses upon old buildings, its impli-
in which they are located. While historical                     cations for land use are in turn primarily
designation takes place at the local, state and                 directed towards old neighbourhoods. Since
national levels, the putative goal in all cases                 the experience of many urban areas is that
is the preservation of properties with histori-                 their oldest neighbourhoods are also those
cal and/or aesthetic appeal that would other-                   most in need of exogenous stimulus of some
wise be neglected or even demolished.                           sort, designation and preservation of historic
Designation has both positive and negative                      properties and historic districts has become
impacts on the owner of the building. While                     an important tool in efforts to preserve cen-
it confers a certain prestige on the property,                  tral-city neighbourhoods and to promote
the certification can be costly to attain and                    economic development in blighted urban ar-
often imposes substantial restrictions on how                   eas (Newman, 2001; Listokin et al., 1998;
the property may be maintained and altered.                     Slaughter, 1997; Rypkema, 1995; Wonjo,
It also—and usually typically—prevents                          1991). A number of recent studies have
demolition of the unit. The effect of desig-                    found that, despite the costs and restrictions
nation can therefore be dramatic and it is                      that are associated with historical desig-

N. Edward Coulson is in the Department of Economics, Penn State University, University Park, PA 16802, USA. Fax: 814 863
4775. E-mail: fyj@psu.edu. Robin M. Leichenko is in the Department of Geography, Rutgers University, New Brunswick, NJ
08901, USA. Fax: 732 445 0006. E-mail: rleichen@rci.rutgers.edu. The paper was presented at the conference on ‘Analysis of
Land Markets and the Impact of Land Market Regulation’, Lincoln Institute of Land Policy, July 2002. The authors thank Austin
Troy and the other conference participants for insightful comments, an anonymous referee for helpful comments and Sachiyo
Takata for research assistance.
0042-0980 Print/1360-063X On-line/04/081587–14  2004 The Editors of Urban Studies
DOI: 10.1080/0042098042000227028
1588                     N. EDWARD COULSON AND ROBIN M. LEICHENKO


nation, it is in fact associated with increases    examined: diversity of population as mea-
in housing values within designated districts      sured by the Simpson index of diversity,1
(Leichenko et al., 2001; Clark and Herrin,         growth rate of population, change in the res-
1997; Schaeffer and Millerick, 1991) and           idential vacancy rate, percentage change in
that designation has a positive external effect    median income and change in the owner-oc-
on the prices of properties which are in the       cupancy rate.
same neighbourhood as designated units but            Our primary results concern the changes
which themselves are not so identified (Coul-       that take place in the housing and demo-
son and Leichenko, 2001). Thus historical          graphic characteristics of tracts between
designation may have important effects on          1990 and 2000. Our central result is that
the broader neighbourhood.                         nothing happens. The changes which occur
   Much of the interest in the broader, exter-     in the tracts between the two Censuses are
nal effects of designation is related not to the   not correlated, either conditionally or uncon-
price implications of designation, per se, but     ditionally, with the existence or extent of
to questions concerning the impacts of desig-      historical designation.
nation on the demographic composition of              The remainder of the paper is organised as
urban neighbourhoods (Schneider, 2001;             follows. In section 2, we review the econ-
Smith, 1998). Does historical designation          omic literature on neighbourhood transition
slow or halt income and racial ‘tipping’ of        and the role that historical designation might
neighbourhoods, a process which entails a          play within these models that describe tran-
transition in the demographic composition of       sition. Section 3 describes our data and Sec-
a neighbourhood from higher-income and             tion 4 presents the econometric analysis of
predominantly White households to lower-           that data. Section 5 concludes.
income and predominantly minority ones?
By the same token, does designation promote
                                                   2. Neighbourhood Transition
gentrification, whereby a neighbourhood oc-
cupied by lower-income, often minority resi-       The filtering model—an early reference is
dents, undergoes an upward shift in average        Muth (1972)—begins with the assumption
income and a change in demographic compo-          that high-income people have a higher ‘inter-
sition towards predominantly upper-middle-         nal’ depreciation rate for housing. That is,
class, White urban professionals? Both of          while high-income households offer bid rents
these questions have important implications        for new property that are higher than bid
for the use of historical designation as a local   rents from households with lower income,
development strategy, yet there has been lit-      the high-income bid rents for any given unit
tle systematic evidence addressing them.           of property decline faster over time, because
This paper represents a first effort to fill this    the decline in quality that occurs as units
gap.                                               depreciate has higher (in absolute value)
   In this paper, we examine the effects of        value to high-income households. In the
historical designation on the demographic          model of Bond and Coulson (1989), a two-
and housing characteristics of urban neigh-        income neighbourhood is considered. In such
bourhoods, using Fort Worth, Texas, as a test      a case, the bids are as in Figure 1, where H
case. Fort Worth is ideal for such a purpose       and L denote the bids of households of high
because of the extent to which historical          and low income. Note that the slope of the
designation has been implemented, either ex-       former is steeper than that of the latter, indi-
plicitly for aesthetic purposes, or implicitly     cating a greater decline in value. The bids of
as a development tool. Using census tracts as      low-income households exceed those of
the unit of observation, we examine the im-        high-income households at turnover age, T,
pact of the existence and extent of historical     and so the housing changes hands at the
preservation on tract demographic and hous-        point in time when the unit reaches that age.
ing characteristics between 1990 and 2000.         With more than two income groups, a unit
Five demographic and housing indicators are        turns over a number of times, until such point
                     HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE                       1589




                        Figure 1. Filtering and neighbourhood transition.



as the lowest-income group’s bid for the unit     of the same race, then an exogenous influx of
becomes zero, at which point it is demol-         new Black residents will lower Whites’ and/
ished or otherwise removed from the housing       or raise Blacks’ willingness-to-pay for adjac-
stock.                                            ent units, and cause a turnover of units from
   New construction can occur at that, or         White to Black households. The increased
some earlier, point (see Bond and Coulson,        proportion of Blacks in the area may increase
1989), but in the absence of this, the income     the willingness of White households to move
of a unit’s occupants will inexorably decline     out, or Blacks to move into the neighbour-
over time and so neighbourhood income will        hood, and the changing racial composition of
decline just as inexorably as more and more       the neighborhood encourages yet further in-
units reach the turnover age. Studies of          and out- migration. Quite rapid shifts in the
filtering (Brueckner, 1977 and Phillips, 1981)     demographic composition of the neighbour-
have relied on this relationship between the      hood are therefore possible; the neighbour-
age distribution of the housing stock and the     hood ‘tips’. The turnover can also be
income composition of the neighbourhood to        income-oriented, rather than race-oriented, as
test the filtering model.                          found in Coulson and Bond (1990). While it
   As a theory of neighbourhood change, the       might be the case that all income-groups like
filtering model is somewhat limited, relying       having high-income neighbours, as long as
exclusively as it does on the housing unit as     the bids for high-income neighbourhoods are
the object of interest. There is no role for      higher among high-income households, an
neighbourhoods in this framework except as        exogenous increase in low-income house-
a collection of individually filtering resi-       holds could trigger the same sort of ‘tip’
dences.                                           from high- to low-income residents in a
   On the other hand, the tipping model           given neighbourhood.
(Schelling, 1978; other references include           A lacuna in the traditional tipping litera-
Miyao, 1978) relies on neighbourhood exter-       ture is the source of the influx of residents
nalities to generate neighbourhood transition,    who alter the demographic composition of
without reference to the condition of housing     the neighbourhood and begins the tipping
units. For example, if Whites dislike living in   process. Bond and Coulson (1989) combine
the same neighbourhood as Blacks and/or if        the filtering and tipping models, so that the
all households prefer to live with neighbours     filtering of the oldest housing in the area
1590                       N. EDWARD COULSON AND ROBIN M. LEICHENKO


provides the source of the ‘exogenous’ shock          deterministic trend that favours (say) Blacks,
to the neighbourhood composition. They                in the sense that it increases the desirability
identify ‘tipping’ with the outcome of a              of the neighbour to one group over another.
shock to an unstable mixed-neighborhood               In such a case, the neighbourhood can reach
equilibrium and ‘filtering’ with the case of           a point of inevitability, where everybody
stable     mixed-neighborhood           equilibria.   knows that the neighbourhood is going to
Whether or not the mixed equilbrium is sta-           turn over at some point, and a kind of back-
ble depends on the number of houses at the            ward induction leads all (say) Whites to
turning-point, T, and the relative dislike of         move out immediately. Thus turnover can
low-income neighbourhoods by low- and                 happen very rapidly.
high-income households. In cases with un-                So, what is the role of historical desig-
stable mixed neighbourhoods, there are sta-           nation in neighbourhood turnover in light of
ble, but segregated, equilibria involving all         these models? In the view of those who
high- or all low-income residents in the area.        would see preservation efforts as a develop-
   Thus in Figure 1 the tipping age, T, has           ment tool, the starting-point for such analysis
been lowered from T to T* because both                must be as configured in Figure 2. In Figure
types of household lower their bids over time         2, let age H be the age at which designation
as some housing has filtered from high- to             occurs. Figure 2 assumes a change in taste
low-income residents. However, the effect on          that manifests itself as a discrete jump in bids
high-income households is greater and so              by high-income people, leading to an in-
this generates the decline in tipping age. The        crease in the price of the unit, as demon-
process continues, until all residents are low-       strated in Leichenko et al. (2001). (Although
income. Such an outcome is stable. In the             there is little published evidence on the slope
absence of further exogenous shocks, the              of the price path after designation, we have
neighbourhood remains in this state. How-             drawn it as positive.)2
ever, if demolition and redevelopment occur,             In the ‘pure filtering’ version of residential
the neighbourhood could either ‘tip’ upwards          succession and neighbourhood turnover, this
or simply remain in an all-high-income state.         is the end of the story. By bestowing desig-
   Another extension of the tipping model             nation on a property, the property turns over
has been proposed recently by Frankel and             from low-income households to high-income
Pauzner (2002). In Bond and Coulson                   households and a limited form of gen-
(1989), the neighbourhood transition is gov-          trification occurs. Thus we would expect
erned by history—more specifically, by the
initial conditions which are captured by the
distribution of vintages of the housing stock.
The more heterogeneous the housing stock,
the less likely it is that tipping will occur. On
the other hand, it is possible that neighbour-
hood transitions are more often motivated by
expectations of future events. One then ob-
tains the usual multiple equilibria: if (say)
Whites expect neighbourhoods to turn over
to minority households, they sell—and the
expectation of turnover creates the turnover
itself. If Whites do not hold such expecta-
tions, then turnover need not occur.
   The innovation of Frankel and Pauzner is
to provide conditions under which a unique
equilibrium could arise. One of the ways in           Figure 2. Filtering and neighbourhood transition
which this could happen is if there is a                        with historical designation.
                      HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE                         1591

only minor changes to the neighbourhood             effects are strong enough, any neighbour-
and only to the extent that designation occurs      hood with designated properties would con-
within its borders.                                 tribute to such correlations, regardless of the
   However, advocates of preservation be-           extent to which designation occurs, although
lieve in the ability of historical designation to   the definition of neighbourhoods would have
effect broader changes in neighbourhoods.           to be fine enough not to blunt the measure-
Wagner (1993) and Rypkema (1994) de-                ment of the spatial impact of this catalyst
scribe the external effects that can take place     effect.
after preservation efforts occur. Wagner               An intriguing possibility is that, following
(1993) describes preservation as a ‘catalyst’       the model of Frankel and Pauzner (2002),
for investment in other properties in blighted      historical designation acts as the determinis-
or depressed areas. Rypkema notes that              tic force which engenders the ‘zone of in-
                                                    evitability’ which in turn generates
  one renovation supports another … as
                                                    immediate neighbourhood transitions. That
  more properties are rehabilitated, lenders
                                                    is, the designation of historical properties
  are more interested in making loans … and
                                                    within a neighbourhood generates a deter-
  terms become more attractive (Rypkema,
                                                    ministic outcome of all White or high-in-
  1994, p. 68).
                                                    come residents which is observed by all
However, Listokin et al. (1998) note that the       residents, creating the immediate turnover of
evidence for these external effects is anecdo-      the neighbourhood. In this case, we should
tal rather than systematic.                         observe an immediate impact; neighbour-
   The model of Bond and Coulson (1989)             hoods with designated properties should im-
can be used to describe how such ‘reverse           mediately have different demographic and
filtering’ or broader neighbourhood change           housing characteristics at the point in time of
can occur. At H, the designated housing turns       designation, or soon after.
over from low to high income; this increases           To summarise, if historical designation has
the desirability of the neighbourhood to both       impacts, then
rich and poor households but, under the as-
sumptions of Bond and Coulson (1989), the
                                                    1. the ‘pure’ filtering model predicts that
effect is greater on the bids of rich house-
                                                       only minor neighbourhood effects will be
holds and so (as in the case with no desig-
                                                       observed as a result of designation; such
nation) the age at which the original turnover
                                                       changes should be soon after designation
(from rich to poor) occurs increases and the
                                                       occurs, but then only to the extent that
average income of the neighbourhood in-
                                                       designation permeates the neighbourhood;
creases as well. Something like the catalyst
                                                    2. the Bond–Coulson filtering/tipping model
effect then occurs, because the neighbour-
                                                       suggests that major neighbourhood effects
hood becomes even more desirable to high-
                                                       will happen in any neighbourhood where
income individuals, increasing T, and so on
                                                       designation occurs, but will perhaps occur
(as if redevelopment were occurring).
                                                       only over some longer time-period;
Whether this is a ‘stable’ neighbourhood
                                                    3. the Frankel–Pauzner model suggests that
transition, or an example of ‘reverse tipping’
                                                       the upward mobility of whole neighbour-
(i.e. immediate transition) depends on the
                                                       hoods with designation will occur very
precise age distribution of the housing stock
                                                       soon after designation.
and the desirability of having high-income
neighbours.
   If this model is correct, we would expect        We turn now to the empirical evidence about
to see neighbourhood demographics and               the relationship between historical desig-
housing characteristics conditionally corre-        nation and neighbourhoods, and consider
lated in a positive manner with the existence       whether the actual Fort Worth experience
of historical designation. If the externality       coincides with any of these models.
1592                     N. EDWARD COULSON AND ROBIN M. LEICHENKO




                  Figure 3. Concentration of designated property in Fort Worth.



3. Data                                           ignated as historic either by placement on the
We use the city of Fort Worth, Texas, as our      National Register of Historic Places or by
laboratory to test assertions about the effect    recognition from the Texas Historical Com-
of historical preservation on neighbourhoods      mission or by local historical commissions.
and their transition. Our original database,      So far as we have been able to determine,
which came from the Tarrant County Ap-            most of these properties received their desig-
praisal District as part of a broader study of    nation in the 1980s through to 1990, al-
the economic impact of historical preser-         though the earliest recognition date we can
vation in Texas (see Leichenko et al., 2001,      find is 1979. The substantial majority of
for further information), consists of data on     these properties (1245) are in one of three
over 100 000 residential properties in the city   large historic districts: Elizabeth Avenue,
of Fort Worth. The database contains infor-       Grand Avenue and Fairmount–Southside.
mation on the historical status of each of the    The remaining 93 historic properties are des-
properties as well as a number of housing         ignated individually as historic, but are not
characteristics such as would typically be        located in one of these three historical dis-
available on a property assessment form.          tricts.
These data on individual housing units were          This concentration of historically desig-
geocoded and linked to the census tract in        nated properties within historical districts
which the housing unit is located, of which       does not imply that Fort Worth historical
236 are in the final data-set. The tract-aver-     properties are concentrated in just a few par-
age data are then linked to demographic in-       ticular census tracts. As illustrated in Figure
formation on the tract from the 1990 and          3, which identifies census tracts containing
2000 Censuses.3                                   historic properties, many census tracts lo-
   Our database includes 1338 residential         cated throughout the city contain either his-
properties in Fort Worth that have been des-      toric districts or individual historically
                     HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE                                  1593

               Table 1. Means and standard deviations for Fort Worth census tracts

                                                          Tracts without               Tracts with
                                All tracts                 designation                 designation
                            Mean      Standard        Mean         Standard        Mean         Standard
 Variable                             deviation                    deviation                    deviation

Vacancy rate, 1990             0.11         0.06            0.11         0.06            0.13        0.06
Ownership rate, 1990           0.60         0.22            0.61         0.23            0.58        0.19
Population. 1990           4 312.98     2 284.30      4   313.70     2 196.14     4    310.52    2 581.97
Percentage Black, 1990         0.15         0.25            0.16         0.25            0.13        0.22
Percentage Hispanic, 1990      0.13         0.17            0.11         0.13            0.19        0.24
Simpson index, 1990            0.69         0.17            0.69         0.17            0.68        0.17
Median income, 1989       32 905.48    16 281.15    33    520.79    16 424.05     30   831.69   15 761.81
Average year built         1 952.43        15.98     1    954.03        15.56      1   947.05       16.38
Average living area        1 505.56       537.00     1    477.31       537.67      1   600.76      528.58
Number of designated           5.57        30.43            0.00         0.00           24.33       60.33
  homes
Historical designation         0.23          0.42           0.00           0.00          1.00        0.00
  dummy
Change in vacancy rate         0.05          0.05           0.05           0.05          0.04        0.06
Change in ownership rate       0.02          0.08           0.02           0.08          0.01        0.09
Change in Hispanic             0.09          0.08           0.09           0.08          0.09        0.09
  percentage
Change in Black                0.01          0.07           0.01           0.07          0.00        0.07
  percentage
Change in Simpson              0.12          0.11     0.12665              0.11          0.10        0.12
  index
Population growth rate         0.33          1.86           0.35           2.08          0.26        0.71
Percentage change              0.41          0.36           0.41           0.29          0.43        0.29
  median income




designated properties. In total, 54 of the 236      use in our analysis and then presents
tracts included in our database contain his-        stratified means and standard deviations for
torically designated properties. As Figure 3        tracts without and with any historical desig-
also indicates, there is gratifying variation in    nation. The row labelled ‘Historical Desig-
the geographical scope and intensity of his-        nated Dummy’ provides information about
torical designation.                                the indicator variable which equals one if any
   Designation in Fort Worth is not a               such property exists in the tract. As can be
phenomenon only of the inner city, but also         seen from Table 1, 23 per cent of the tracts in
has presence in more suburbanised tracts.           the sample have at least one historically des-
Moreover, tracts that overlap with historic         ignated property. From the following row,
districts, and therefore have a more substan-       we find that across all tracts there is an
tial connection to historic properties, are also    average of 5.57 properties per tract, but in
not confined to the central city—although, as        those tracts with designated property the av-
can be seen on Figure 3, the most concen-           erage is 24.33.
trated placement of designated properties is           The columns in Table 1 provide compari-
indeed in the centre of the city. This set of       son between tracts that have designation and
tracts is the centre of the Elizabeth Avenue        those that do not. As can be observed, the
and Fairmount–Southside Historic Districts.         two types of tract differ slightly, but only
   Table 1 presents the means and standard          slightly, in their housing and demographic
deviations of the tract-level variables we will     characteristics. While the average popula-
1594                       N. EDWARD COULSON AND ROBIN M. LEICHENKO

              Table 2. Determinants of the existence and extent of historic designation

  Independent variable                    Dependent            Dependent              Dependent
                                      variable HDUM           variable ND             variable ND
                                           (probit)             (Poisson)             100 (Poisson)
                                                        4                       4                     4
  Population, 1990                           5.82 10               2.59 10               1.55 10
                                            (1.30)               (18.16)                (2.47)
  Ownership rate, 1990                       1.27                  3.32                  2.67
                                            (1.95)               (12.88)                (3.07)
  Vacancy rate, 1990                         4.09                 18.65                  8.48
                                            (1.92)               (26.58)                (3.26)
  Percentage Black, 1990                     0.554                 7.97                  1.050
                                         ( 1.06)               ( 16.79)               ( 1.54)
  Percentage Hispanic, 1990                  1.20                  1.35                  0.207
                                            (1.84)               (10.16)                (0.30)
                                                        4                       5                     5
  Median income, 1989                        1.13 10               2.26 10               1.91 10
                                         ( 1.14)                ( 6.27)               ( 1.47)
  Average year built                         0.0245                0.0645                0.0575
                                         ( 3.18)               ( 22.62)               ( 5.73)
  Average living area                    9.09 10 4                 0.00211               0.0014
                                            (3.91)               (28.34)                (5.78)
  Pseudo R2                                  0.0044                0.4755                0.1572

  Note: The above table provides the coefficients and t-statistics (in parentheses) for regressions of the
indicated dependent variable on the independent variables in the first column.



tions of designated and undesignated tracts           1000 square feet bigger than buildings in
are nearly identical, the designated tracts           undesignated areas.
have rather greater percentages of their popu-
lation claiming Hispanic heritage, while the
                                                      4. Model and Results
percentage of the population that is Black is
lower. Tracts with historical designation also        Our first regressions are those where the
seem to have slightly more depressed econ-            dependent variable is an indicator of the
omic indicators; they have a slightly higher          extent to which historical designation exists
vacancy rate, lower home-ownership rate and           in a census tract in 1990. Table 2 presents
lower median income. The variables above              these results. In the first column, the depen-
are all derived from the 1990 Census data             dent variable is the dummy variable for the
but, as noted, we have fairly complete data           existence of any historical property and we
on single-family properties in these tracts           run probit regressions to model this. As can
from our appraisal data. Although we con-             be seen there, several of the indicators are
structed a number of indicators of ‘tract             correlated with the existence of designation.
housing quality’, we confine ourselves to two          Obviously the average age of the tract is
in this analysis: average building age and            important, but so is the size of units. This is
interior square feet. These two seem ad-              of interest if only because age and size are
equate for instrumenting building quality and         negatively correlated. Given two tracts with
the difference between tracts with and with-          the same average age (and other characteris-
out designation. Buildings in historic tracts         tics), the one with the larger housing units is
are (naturally enough) somewhat older, but            more likely to have some historical desig-
the difference in average ages is only about          nation. This is presumably the result of his-
seven years. A larger and more interesting            torical designation being more likely to occur
difference is that of building size; designated       in places that have been historically wealth-
tracts have buildings that are on average over        ier than others of the same vintage. However,
                     HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE                           1595

while this may have been the case in the past,     lation of the tract is now significant, which is
it is not so today; the income levels of the       sensible given that we are trying to explain
designated tracts do not have conditionally        counts of homes in a tract. But the rest of the
higher income. The coefficient on 1989 in-          coefficients reflect the results in column (1):
come is negative and insignificant.                 tracts with greater vacancy rates, proportions
   However, there are some demographic dif-        of owner-occupied houses and buildings with
ferences. The coefficients indicate that, at        older vintages or more space have more des-
standard levels of significance, tracts with        ignated properties. However, income differ-
designated properties in 1990 were, on aver-       ences are still not significant.
age, slightly more Hispanic than others.              How do we interpret these results? The
Significantly, the vacancy rate is higher in        results do not support the theories of neigh-
those neighbourhoods as well, which is con-        bourhood change described above; if such
sistent with the idea that designation is di-      ‘upward tipping’ theories were true, we
rected at those areas with more depressed          would expect that in 1990 the indicators
housing markets. However, it is also the case      would be significantly better for tracts with
that neighbourhoods with designation have          designation than those without. One would
higher home-ownership rates (and recall that       have expected indicators such as (and per-
we have controlled for building size); thus, if    haps especially) vacancy rates to have been
it is the case that home-owners are more           lower in those areas in which designation has
politically aware (as suggested by                 occurred. It would seem rather that we are
DiPasquale and Glaeser, 1999), this                modelling the choices made by historical
coefficient is consistent with a picture of         commissions (or neighbourhood activists) to
historical designation as the result of neigh-     designate properties in the tract. There is
bourhood activism.                                 clearly an element of choice involved. Lis-
   We explore this topic in a slightly different   tokin et al. (1998, p. 460) quote Rose (1981):
way in column 2, which uses a Poisson count        “The phrase ‘historical preservation’ is so
regression model to examine the determi-           elastic that any sort of project can be
nants of the number of designated properties       justified”, although Listokin et al. (p. 461)
in a tract. The Poisson was used because of        say that most preservation efforts are
the integer nature of the variable and because     “judicious, so that historic designations
the unconditional distribution actually some-      reflect legitimate concerns to protect a com-
what resembles a Poisson.                          munity’s historical resources”. However, if
   The results of this regression are, in a        historical designation is indeed a develop-
word, incredible. The goodness-of-fit mea-          ment tool, then it is presumably aimed at
sure is 45 per cent, and all of the coefficients    locations that would receive more benefit
are highly significant, and of the signs that       from place-based development. These regres-
are to be expected (and the same as in the         sions are therefore not assessing the causal
previous probit). One of the problems with         effects of historical designation, but the re-
this regression is that it may be influenced        verse.4
unduly by the few tract observations with             Therefore, we turn now to the second set
large numbers of historical properties. As a       of regressions that attempt to estimate the
check on these results, we run the same            changes between 1990 and 2000 that might
regression, but eliminating from the sample        be ascribed to historical designation. These
any tract with more than 100 designated            regressions are now meant to be causal since
properties, and these results are rather more      all of the historical designation is prior to the
humbling. The goodness-of-fit measure is re-        year 1990. They are of the form
duced to a bit under 16 per cent and all of the
                                                       90–00   yi   f(x90)
t-statistics are reduced in magnitude. The
major difference between these results and         where, the left-hand side is the change from
the probit of column (1) is that the popu-         1990 to 2000 of some tract characteristic of
1596                      N. EDWARD COULSON AND ROBIN M. LEICHENKO




Figure 4. Change in ownership rate (orc) vs 1990 ownership rate (or90). Note: 1 indicates a tract with
                                historically designated properties.



interest, yi, and x90 is a set of regressors dated   is straightforward to summarise the lesson of
1990 (or perhaps earlier) among which will           that table. We obtain no significant outcomes
always be an indicator for the extent of his-        for neighbourhoods that are the result of
torical designation in the tract. The depen-         economic development that arises from his-
dent variable will alternately be: population        torical    designation.     Both     HD     and
growth rate; the change in the owner-occu-           NUMDESG are insignificant in both the con-
pation rate; the change in the vacancy rate;         ditional and unconditional regressions. The
the percentage change in median tract in-            sole exception is in the multivariate regres-
come; and the change in the Simpson index            sions with the vacancy rate as the dependent
of (ethnic) diversity. The Simpson index of          variable and HD as the designation measure.
diversity is measured as the sum of squared          But even then the coefficient on HD is posi-
shares of each population group within the           tive, indicating that, after controlling for the
tract and is used as a measure of ethnic             other attributes of a neighbourhood, those
change in the tract. We use this, as opposed         tracts with historical designations had in-
to changes in population shares of various           creases in residential building vacancy rates
ethnic groups, because it was felt that this is      higher than those without. This case aside,
more what planners, and others, wish to see          the t-statistics for the preservation variables
emerge from a process of neighbourhood               are almost always less than one. Thus, our
revitalisation.5                                     interpretation is straightforward. Historical
   For each dependent variable, four regres-         designation and preservation have no impact
sions were run. Bivariate regressions with           on neighbourhood composition.
HD and NUMDESG as the sole independent                  What one does observe from this set of
variable were calculated, as were multiple           regressions is convergence. In several of the
regressions with 1990 measures of popu-              regressions, the only significant variable is
lation, vacancy rates, ownership rates, Black        the ‘own’ variable—the 1990 ‘starting-point’
and Hispanic percentages and size and age of         for that particular dependent variable. This is
the housing stock, along with 1989 median            an indication of convergence of tracts to-
income.                                              wards the mean. For example, tracts with
   The results are presented in Table 3 and it       high ownership rates experience a relative
                                                           Table 3. Impacts of historic preservation on demographic and housing characteristics
                                                             ownership ownership ownership ownership                                                                                                                   Percentage Percentage Percentage Percentage
            Population Population Population Population        rate      rate      rate      rate                    vacancy      vacancy      vacancy      vacancy      Simpson      Simpson      Simpson      Simpson median      median      median     median
           growth rate growth rate growth rate growth rate 1990–2000 1990–2000 1990–2000 1990–2000                    rate         rate         rate         rate        index        index        index        index   income     income     income     income

Intercept 0.348         0.333        0.224       –0.0110      0.020        0.0171       1.321                                  –0.0445      0.3550       0.4653       –0.1266      –0.3570      –0.1223      –0.3397     0.413     1.25           0.414     1.22
          (2.52)        (2.70)       (0.02)      (0.00)       (3.48)       (3.28)       (1.73)                                 (–13.51)     (1.10)       (1.44)       (–14.81)     (–0.41)      (–15.57)     (–0.039)    (14.48)   (0.37)         (16.39)   (0.37)
HDUM –0.0866                         –0.093                   –0.0098                   –0.011                   –0.0024                    0.0122                    0.250        –0.010                                0.0162    0.0213
          (–0.30)                    (–0.31)                  (–0.81)                   (–0.86)                  (–0.31)                    (2.24)                    (1.40)       (–0.71)                               (0.28)    (0.32)
NUMDESG                 –8.34 10–4             –0.149 10–3                 1.83 10–4               1.82 10–4                   –1.46 10–4              6.93 10–5                              2.54 10–4 2.16 10–4                                 0.0005    4.29 10–4
                        (–0.021)               (–0.037)                    (1.09)                  (1.07)                      (–1.36)                 (0.93)                                 (1.07)    (–1.14)                                   (0.78)    (0.59)
                                                                                                                                                                                                                                             –5
Population,                          0.0002443 .0002447                                 –2.26 10–6 –2.57 10–6                               –3.91 10–6 –3.77 10–6                  –1.97 10–6           –2.05 10–6                 –3.2 10                  –3.2 10 –5
 1990                                (–4.39)   (–4.41)                                  (–0.94)    (–1.07)                                  (–3.84)    (–3.68)                     (–0.61)              (–0.64)                    (–2.59)                  (–2.60)
Ownership                            –0.8300   –0.841                                   –0.0715    –0.0771                                  –0.0254    –0.0225                     0.0304               0.0289                     –0.166                   –0.166
 rate, 1990                          (–1.12)   (–1.14)                                  (–2.28)    (–2.42)                                  (–1.87)    (–1.65)                     (0.81)               (0.77)                     (–1.17)                  (–1.17)
Vacancy                              4.082     4.10                                     –0.0671    –0.0958                                  –0.7272    –0.7194                     –0.1278              –0.1229                    –1.33                    –1.33
 rate, 1990                          (1.54)    (1.55)                                   (–0.59)    (–0.84)                                  (–15.01)   (–14.70)                    (–0.99)              (–0.95)                    (–2.72)                  (–2.75)
Percentage                           –1.04     –1.05                                    –0.0071    –0.0030                                  0.0435     0.0426                      0.0785               0.0771                     0.195                    0.194
 Black,                              (–1.70)   (–1.70)                                  (–0.27)    (–0.11)                                  (3.86)     (3.75)                      (2.66)               (2.62)                     (–1.53)                  (–1.52)
 1990
Percentage                           –0.793      –0.781                                 –0.0243     –0.0351                                 –0.0183      –0.0157                   0.4911                    0.4939                0.053 (0.33)             0.046
 Hispanic,                           (–0.95)     (–0.93)                                (–0.67)     (–0.97)                                 (–1.20)      (–1.01)                   (12.24)                   (12.33)                                        (0.29)
 1990
Median                               1.38 10–5 1.40 10–5                                –1.27 10–7 –9.31 10–8                               –1.88 10–7 –2.15 10–7                  1.21 10–6                 1.23 10–6             –1.23 10–5               –1.24 10–5
 income, 1989                        (1.17)     (1.19)                                  (–0.25)    (–0.18)                                  (–0.87)    (–0.99)                     (2.07)                    (2.13)                (–4.61)                  (–4.64)
Average                              6.12 10–4 7.31 10–4                                –0.00062 –0.00048                                   –1.56 10–4 –2.15 10–4                  5.20 10–5                 4.27 10–5             1.900 10–7               1.44 10–5
 year built                          (0.07)     (0.08)                                  (–1.57)    (–1.24)                                  (–0.94)    (–1.29)                     (0.11)                    (0.10)                (0.00)                   (0.01)
Average                              –1.12 10–4 –1.19 10–4                              –1.62 10–5 –2.11 10–5                               1.12 10–5 1.35 10–5                    1.57 10–5                 1.51 10–5             –3.12 10–4               –2.99 10–4
                                                                                                                                                                                                                                                                         HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE




 living area                         (–0.38)    (–0.41)                                 (–1.27)    (–1.69)                                  (2.07)     (2.54)                      (1.06)                    (1.06)                (–0.57)                  (–0.56)
R2          0.0004      0.0002       0.1230     0.1231     0.0028                       0.0862     0.0876                      0.0076       0.5971     0.5897     0.0095           0.4560    0.0056          0.4581                0.244                    0.245

Note: This table provides the coefficients and t–statistics (in parentheses) for regressions of the indicated dependent variable on the independent variables in the first column.
                                                                                                                                                                                                                                                                         1597
1598                     N. EDWARD COULSON AND ROBIN M. LEICHENKO


decline in ownership rates during the 1990s       associated with neighbourhood turnover of
and those with low ownership rates see an         any type.
increase. Observe Figure 4, which gives the          We experimented with other measures, in-
graph of ownership rate change (orc) against      cluding growth in the stock of housing, other
initial ownership rate (or90). The symbol ‘1’     diversity indexes and simple growth in the
indicates a tract with designated properties.     minority populations, and used other mea-
The ticks fan inwards and are roughly nega-       sures of the extent of historical designation.
tively sloped as would be expected given the      (Moreover, in view of the absence of
convergence of ownership rates, but no his-       spillovers within tracts, we eschew examin-
torical tract is associated with any large in-    ation of intertract spillovers and other spatial
crease in home-ownership.                         consideration.) The overall results remain
   There are a few other significant t-statis-     that no significant change in tract demo-
tics scattered through Table 3. Tracts with       graphic composition is associated with any
large Black populations are associated with       measure of historical designation.
increases in the vacancy rates and changes in
the diversity index are associated with large
                                                  5. Conclusion
initial Black and Hispanic populations. This
is rather like a convergence result in itself.    Designation of historic properties and his-
Also, tracts with larger populations and          toric districts is a popular, yet somewhat
lower vacancy rates are associated with           controversial, tool for revitalisation of older
smaller growth rates in median tract income.      central-city areas. In this paper, we explore
   However, the effect of historical desig-       the effects of historical preservation on the
nation is not entirely neutral. As noted in the   characteristics of neighbourhoods in Fort
introduction, our previous work has found         Worth, Texas, during the period from 1990
that historical designation increased the level   to 2000. In particular, we examined the im-
of housing prices for individual homes (Le-       pact of the existence and extent of historical
ichenko et al., 2001) and the level of prices     preservation on tract demographic and hous-
in census tracts that contained historical        ing characteristics between 1990 and 2000.
homes or districts (Coulson and Leichenko,            Our overriding conclusion is that historical
2001). Although housing prices are some-          designation does not lead to gentrification, or
what outside the scope of the present paper,      any other kind of neighbourhood turnover.
we performed similar regressions (i.e. with       We found that, while there is some evidence
identical conditioning variables) to those re-    that areas are chosen for preservation efforts
ported in Table 3. Using the percentage in-       with neighbourhood revitalisation in mind,
crease (over the period 1990–2000) of             the decade or so following designation pro-
median housing price in the tract as the          duced no significant change in neighbour-
dependent variable, we found that those           hood demographic composition. Those who
tracts with historically designated homes had     fear that designation will lead to displace-
a significantly higher rate of increase in prop-   ment of lower-income residents may be
erty values.6 Thus it would seem that the         mollified on that account. Concerning the
effects of historical designation are not too     effects of designation on local economic con-
noisy to blunt totally the measurement of any     ditions, we found evidence that historic pres-
effects. However, the only observable effect      ervation increases property values, but has
is that of the increased desirability of the      little effect on other measures such as va-
neighbourhood; whether that is due to the         cancy rates and rates of owner-occupancy.
increased cachet of historical designation, or    Those who hope that designation will lead to
due to the strictures placed on the owners of     dramatic economic development benefits
such property, remains an area of research.       may be disappointed on that account.
What seems clear from the evidence pre-               Our findings have several implications for
sented in Table 3 is that designation is not      urban development policy. Our results sug-
                       HISTORIC PRESERVATION AND NEIGHBOURHOOD CHANGE                               1599

gest that designation has direct benefits in                 t-statistic of 2.20 in the binary regression.
terms of property values—a finding that                      Slightly weaker t-statistics were observed
                                                            when the number of designated properties in
reaffirms the conclusions of our prior re-                   the tract was used instead.
search (Leichenko et al., 2001). Higher prop-
erty values, in turn, are likely to have indirect
benefits for cities in the form of higher prop-          References
erty tax revenues. With regard to neighbour-
                                                        BOND, E. W. and COULSON, N. E. (1989) External-
hood revitalisation efforts, however, our                 ities, filtering, and neighborhood change, Jour-
results suggest that historical preservation              nal of Urban Economics, 25, pp. 231–249.
may not be as effective a tool as is some-              BRUECKNER, J. (1977) The determinants of resi-
times thought. If the aim of urban policy is to           dential succession, Journal of Urban Econom-
revitalise deteriorating, older neighbour-                ics, 4, pp. 45–59.
                                                        CLARK, D. E. and HERRIN, W. E. (1997) Historical
hoods, then historic districting represents               preservation and home sale prices: evidence
only a partial solution. More direct measures             from the Sacramento housing market, The Re-
in central-city areas—such as incentive pro-              view of Regional Studies, 27, pp. 29–48.
grammes to promote the purchase of vacant               COULSON, E. and BOND, E. (1990) A 2 Review of
housing by owner-occupants—are also re-                   Economics and Statistics, 72, pp. 433–443
                                                        COULSON, N. E. and LEICHENKO, R. (2001) The
quired.                                                   internal and external impacts of historical des-
                                                          ignation on property values, Journal of Real
Notes                                                     Estate Finance and Economics, 23(1), pp. 113–
                                                          124.
1.   The Simpson index is defined in section 3.          DIPASQUALE, D. and GLAESER, E. (1999) Incen-
2.   The bids by low-income people may rise as            tives and social capital: are homeowners better
     well, although for simplicity we eschew that         citizens?, Journal of Urban Economics, 45,
     possibility in Figure 2.                             pp. 354–384.
3.   As is typical, there were some 1990 tracts         FRANKEL, D. and PAUZNER, A. (2002) Expecta-
     that were split into two tracts for the 2000         tions and the timing of neighborhood change,
     Census. These tracts were recombined into a          Journal of Urban Economics, 51, pp. 295–314.
     single tract to preserve comparability be-         LEICHENKO, R., COULSON, E. and LISTOKIN, D.
     tween the two Censuses. There were no other          (2001) Historic preservation and residential
     border changes or other issues with tract            property values: an analysis of Texas cities,
     comparability. In view of the results we re-         Urban Studies, 38(11), pp. 1973–1987.
     port in the next section, it might be thought      LISTOKIN, D., LISTOKIN, B. and LAHR, M. (1998)
     that census tracts are too large an area to          The contributions of historic preservation to
     observe the spillover results of historical des-     housing and economic development, Housing
     ignation. We believe, however, that tracts are       Policy Debate, 9, pp. 431–478.
     in fact an appropriate level of observation,       MIYAO, T. (1978) Dynamic instability of a mixed
     first because there are some impacts of pres-         city in the presence of neighborhood externali-
     ervation we have observed—particularly,              ties, American Economic Review, 68, pp. 454–
     price changes—in other studies (as described         463.
     above). Furthermore, part of our interest is in    MUTH, R. (1972) A vintage model of the housing
     whether the extent of preservation matters           stock, Regional Science Association Papers
     and it is easier to obtain sample variation of       and Proceedings, 30, pp. 141–56.
     this when tracts are used.                         NEWMAN, H. (2001) Historic preservation policy
4.   We did not find any causal impact of prior            and regime politics in Atlanta, Journal of Ur-
     designation on any of the 1990 indicators.           ban Affairs, 23, pp. 71–86.
     Since these results are so similar to the re-      PHILLIPS, R. (1981) A note on the determinants of
     gressions presented in Table 3, we omit them         residential succession, Journal of Urban Eco-
     for convenience.                                     nomics, 9, pp. 49–55.
5.   However, letting y be population proportion        ROSE, C. (1981) Preservation and community:
     of Whites, Blacks or Hispanics had similar,          new directions in the law of historic preser-
     insubstantial results.                               vation, Stanford Law Review, 33, pp. 473–534.
6.   Using HDUM as the measure of historical            RYKEMA, D. (1994) The Economics of Historic
     designation yielded a coefficient on HDUM             Preservation: A Community Leader’s Guide.
     of 0.088 and a t-statistic of 1.98 in the full       Washinton, DC: National Trust for Historic
     regression and a coefficient of 0.085 and a           Preservation.
1600                      N. EDWARD COULSON AND ROBIN M. LEICHENKO

RYPKEMA, D. (1995) Economics and historic pres-     SLAUGHTER, H. B. (1997) Integrating economic
  ervation, Historic Preservation Forum, 9, 39-       development and historic preservation in Pitts-
  45.                                                 burgh, Pennsylvania, Historic Preservation Fo-
SCHAEFFER, P. V. and MILLERICK, C. A. (1991)          rum, 11, pp. 41–44.
  The impact of historic district designation on    SMITH, N. (1998) Comment: the contributions of
  property values: an empirical study, Economic       historic preservation to housing and economic
  Development Quarterly, 5, pp. 301–331.              development, Housing Policy Debate, 9,
SCHELLING, T. (1978) Micromotives and Mac-            pp. 479–485.
  robehavior. New York City: Norton.                WAGNER, R. (1993) Urban downtown revitaliza-
SCHNEIDER, T. (2001) From monuments to urban          tion, Historic Preservation Forum, 7, pp. 53–
  renewal: how different philosophies of historic     58.
  preservation impact the poor, Georgetown          WOJNO, C. T. (1991) Historic preservation and
  Journal on Poverty Law & Policy, 8, pp. 257–        economic development, Journal of Planning
  281.                                                Literature, 5, pp. 296–307.