Document View

We analyze appreciation rates across comparable designated and undesignated neighborhoods in Memphis, Tennessee. Using appreciation rates potentially nullifies the objection to using assessed values in such an analysis while also mitigating some of the bias inherent in the differences between otherwise similar designated and undesignated neighborhoods. Nonetheless, in accord with the previous studies, after controlling for numerous housing characteristics, we find that properties in neighborhoods designated historical by the Memphis Landmarks Commission had appreciation rates above those in other similar neighborhoods. We also find that new properties benefit as much, perhaps even more, than older properties from being in a historic district. [PUBLICATION ABSTRACT]

The designation and preservation of historical sites is an increasingly widespread tool in urban design and planning. In part, its application is intensifying because there are few vocal opponents to preserving our cultural heritage. It is implicitly an urban-oriented policy because nearly all of the U.S.'s historic building stock is located in its cities, and, moreover, older neighborhoods often are the parts of cities in greatest need of external stimuli. Thus, public incentives encouraging private investment in historic properties have attributes that give them the appearance of being a partial remedy for a myriad of urban problems.

To date, a national tax credit up to 20% of the rehabilitation costs is availabl^ to owners of historic property. Additionally, 15 states offer state income tax credits and 25 states permit some form of property tax abatement that benefits older, if not strictly historic, stock (Beaumont and Pianca 2001). To qualify for many of these incentives, properties often must be officially designated as being "historic," be eligible for such designation or be a "contributing" building in an officially designated historic district. Thus, historic designation can be bestowed either on individual sites or on entire neighborhoods. In addition to the explicit tax benefits, designation often can add a certain cachet to a property.

But designation is more than just an information transmission device or a means of catching a tax break. Typically, it also restricts use of the property, restricts the types of refurbishment and rehabilitation that can be undertaken and burdens property owners with responsibilities for upkeep and maintenance that go beyond those found in usual zoning and building code regulations. Indeed, in the case of neighborhood designation, regulations that often accompany historic preservation appear to provide a mechanism ensuring neighborhood upkeep (Leichenko, Coulson and Listokin 2001, Coulson and Leichenko 2004). Inefficient levels of maintenance are a result of a prisoner's dilemma-like interaction in which property owners have an incentive to invest only in low levels of maintenance, regardless of their neighbors' maintenance behavior. The natural upshot is that all neighbors can wind up employing this strategy, resulting in an overall downward spiral in the quality of housing stock in the neighborhood. In such a situation, everybody is made worse off than if they all had agreed to provide high levels of maintenance. The restrictions embodied in the designation of a historical neighborhood may have the potential to induce owners to internalize this neighborhood externality that comes about when maintenance drops below efficient levels. Thus, at least from a theoretical perspective, compliance with preservation restrictions could overcome the momentum of low levels of neighborhood-wide investment in properties.

But the restrictions that come with designation are a double-edged sword. Stringent building codes also can discourage the restoration of older properties (Saltzman 1995). In particular, by their very nature they restrict the way in which property may be altered or refurbished and may require large maintenance expenditures to preserve or restore the historical character of the building or neighborhood or may significantly delay revenue generation on the property. Perhaps the most common theoretical argument is that designation can prohibit a property from attaining its highest value and best use. For example, it could detract from a property's value by prohibiting its conversion to another land use, such as from a current single-family property to a multistory office building. Thus, the ultimate effect of designation on property values is theoretically ambiguous.

The empirical literature on the topic reveals mixed results, though it is fair to say that studies of historic designation have discovered that it leads to higher property values. Some of the earliest studies of the price effects of designation involve simple comparisons of neighborhood averages. Examples include Scribner's (1976) study of Alexandria, Virginia, and Rackham's (1977) study of the Georgetown area of Washington, DC. Both found that their focal historic districts had higher property values than those of comparison neighborhoods. Heudorfer (1975), on the other hand, found that historic areas in New York City had lower prices than the putatively comparable undesignated neighborhoods. In the largest "comparison" study we know of, Shipley (2000) undertook a comprehensive examination of designated property in Canada and, by comparing average appreciation rates of designated property and nonhistorical property on a city-by-city basis, found that in most cities designated properties had greater appreciation rates than other properties.

Obviously, by comparing average property values without controlling for other differences between designated and undesignated lots, structures or neighborhoods, the aforementioned analyses are neglecting other possible explanations for the observed differences in historical and undesignated property values. Regression models using individual properties as the unit of observation can overcome this problem. As it happens, regression-based studies also have generally confirmed a positive effect of designation on property values (Ford 1989, Asabere and Huffman 1994a, Clark and Herrin 1997, Leichenko, Coulson and Listokin 2001). Alternatively, however, others (Asabere, Huffman and Mehdian 1994, Asabere and Huffman 1994b) find that the special restrictions on designated multiunit properties had adverse effects on their values in Philadelphia.

Because, from a theoretical perspective, the extent of preservation that arises from historical designation depends primarily on the trade-off between the internal and external impacts of the regulations related specifically to designation, one might expect that associated price impacts might depend on the stringency of those regulations (only in the case of local regulations-national designation has the same basic influence everywhere, although there are cases where a property has both). Schaeffer and Millerick (1991), in a study of Chicago, find that national designation has a positive impact but local designation has a negative impact, which they attribute to its greater regulatory stringency. Coulson and Leichenko (2001) also found national designation of individual properties to be more value enhancing in their study of Abilene, Texas.

Finally, whatever the effect of designation and preservation regulations on the historic property, undesignated property that shares a border with historic ones should unambiguously experience positive impacts because they bear none of the cost of the regulation and experience a presumably positive externality-that of basking in the glow of an enhanced and maintained property. While Clark and Herrin (1997) did not find any such increases, Coulson and Leichenko (2001) found strong positive price effects from having a number of historical properties in the same census tract in their study of Abilene, Texas.1

Much of the literature focusing on historic designation's effect on property values has done so by analyzing the differences across neighborhoods that are subjectively deemed to be similar. Unfortunately, it undoubtedly is quite difficult to select undesignated neighborhoods that have properties that are sufficiently close in age, style and size to those in the designated neighborhoods to facilitate an unbiased statistical comparison. After all, some underlying set of characteristics of the designated neighborhoods has suggested to policy makers that the subject neighborhoods should be allotted an official historic status, while the selected comparison neighborhoods should not. For example, it may be that the officially designated historic neighborhoods were selected because they embraced architecturally unique structures or a better maintained stock or simply from a planning perspective could serve as a sort of buffer zone for a neighboring commercial district if it was improved. Almost any rationale used to select a comparison neighborhood can also help to explain relatively higher property prices in the designated neighborhood. Hence, identifying higher property values or appraisals in historically designated versus undesignated neighborhoods, even using regression techniques, is at best weak proof that designation yields higher property values.

Somewhat stronger proof of designation's effect on property values can result if one can demonstrate that historic property values proportionally appreciate at a significantly different rate from that of undesignated ones during the same period and in the same city. That is, while similar arguments can be made with regard to price changes as for those in the preceding paragraph on price levels, the arguments are mitigated somewhat because the effect of unobserved timeinvariant characteristics, including those associated with the selection process described above, can be eliminated.

In this article, we undertake a study of a number of the issues regarding the price effects of designation using Memphis, Tennessee as our laboratory. The study data set has a number of features that recommend it for this kind of study. The data characterize the traits of properties in 11 well-defined neighborhoods (as opposed to census tracts).2 While all of these neighborhoods are "old," only about half of them are designated. Of these, all have national designation but some carry in addition a local designation bestowed by the Memphis Landmarks Commission. Some of the designated neighborhoods have relatively large shares of new buildings in them.

The richness and size of the database allow for the explanation of several interesting research questions. We put some extra emphasis on the pricing patterns for properties of different ages. Essentially, we set out to answer the following question: What is the differential effect of age between properties in neighborhoods that are historically designated and those that are not officially designated as historic? Not all buildings in designated areas are old, but in this case they are subject to similarly stringent regulations. Do such properties "bask in the glow" of the historical cachet even though they are not themselves historical? We address these questions through a standard hedonic model that allows for a quite flexible response by the dependent variable to the "age" characteristic.

Description of the Data

Our sample consists of appraisal data from several thousand properties in 11 different neighborhoods in Memphis. Six of these neighborhoods are historic, by which we mean they are included in the National Register of Historic Places. Of these six, three, Shadowlawn, Hein Park and East Buntyn, have no further designation from the local historic preservation agency, the Memphis Landmarks Commission. Annesdale-Snowden is designated by the Commission as a Historic Preservation District, while Central Gardens and Evergreen are named Historical Conservation Zones. A Preservation District is distinguished from a Conservation Zone by the greater number and severity of restrictions with which renovation of its properties must comply. The nonhistoric neighborhoods, the benchmark against which we measure the impact of historical preservation (in its various forms), were chosen by the Memphis Landmarks Commission to match each of the historic neighborhoods in their broad neighborhood characteristics (except for Evergreen, which was deemed too unique to find a good match). For example, Central Gardens is a high-income neighborhood, with house prices to match, so Chickasaw Gardens, another high-priced area of similar (though slightly younger) age and demographic characteristics, was also selected for the database.

To be sure, none of these neighborhoods is entirely made up of aged buildings. Evergreen, for example, has a large number of recently built properties because it was originally in the pathway of a proposed interstate highway. Demolition of many older structures occurred there as a result. In the end, the highway was never built, and the newly vacant properties, over the course of time, were developed.

The Memphis Landmarks Commission provided us data from their appraisal information system that included value and housing characteristics for properties located in each of the historical and nonhistorical neighborhoods. From the data set that was extracted for our use, we assembled 5,889 usable observations on properties, for both the years 1998 and 2002; these years immediately follow fresh reappraisals conducted by the Shelby County Property Assessors office in 1997 and 2001. From these, we calculated the log differences of the appraised values, that is, appreciation rates, for use as dependent variables in our analysis.3 We use a large number of characteristics in our regression models including a number of interaction terms; for convenience, we present in Table 1 the means of housing characteristics customarily thought to be important in the valuation process, stratified by historical status. In addition, we provide the standard error of the difference. It should be noted that a hypothesis test on the difference of means test would reject the equality of the characteristic means; this is in part due to the large sample sizes, but it also indicates the importance of conditioning on housing characteristics and not simply comparing the neighborhoods.

Enlarge 200% Enlarge 400%

The comparison neighborhoods are nevertheless seen to be the appropriate comparisons; while not quite as old as the designated neighborhoods, they are old by any usual standard in U.S. metropolitan areas. And despite the result of any difference in means test, they are similar in quality, as measured by the Table 1 characteristics. The undesignated properties are somewhat smaller in terms of living area, but they have larger lots and a somewhat similar number of bathrooms. Despite the relative similarity in physical characteristics, the historically designated neighborhoods have an appreciation rate that is more than twice that of the undesignated neighborhoods. This may be due to the historic status or due to the differences in characteristics. To know more, we turn to regression analysis.

The Models

Enlarge 200% Enlarge 400%

The coefficients of the housing attributes in the appreciation rate specification are, as can be seen, changes in the coefficients over time. We are, in essence, asking which attributes are associated with high appreciation rates, and in particular we are asking if historical designation is one of those. The coefficients in the changes in housing attributes reflect the value of the hedonic coefficient as of 2002. Such changes are exceedingly rare, and our specification includes only changes in the number of rooms and number of bathrooms.

Discussion of Findings

In this section, we develop a set of models by progressively enhancing the specification in Equation (2). We do so to demonstrate how the apparent effects of historic designation are influenced by full articulation of the array of factors that can enhance a property's value.

Model 1 : The Gross Effects of Designation

The first model merely regresses the 1998-2002 nominal appreciation rate on the binary variable National Designation, which takes on the value of 1 for a neighborhood in the National Registry. As noted earlier, for the neighborhoods in our database, all areas designated as historic by any institution have been placed on the National Registry. Hence, the implications of National Designation can be interpreted more generally here simply as "historic designation." For any given α coefficient, the associated impact of designation (or any other binary characteristic) is measured by the value of ea - 1 (Halvorsen and Palmquist 1980) and they are, in this instance, identical to what one can discern from Table 1. National Designation adds 13.0% to the appreciation rate, a fact that agrees with the indications of the bulk of the literature reviewed in the introduction-that historic designation has a positive effect on housing prices and appreciation rates. This coefficient is statistically significant at all conventional levels of type I error (its t ratio is 16.85).

Model 2: Gross Effects of Different Designations

In column 2, we separately estimate the impact of the various types of designation available in Memphis. Local Designation indicates the location in a locally designated historic zone or district. Local District Designation denotes the more strictly regulated type of local preservation area in Memphis. Among the neighborhoods in the study data set, this status was uniquely bestowed upon the Annesdale-Snowden neighborhood. As can be seen from the table, the coefficient on National Designation is statistically insignificant but, even so, negative and very small in absolute value. From this it would appear that, in Memphis at least, it is not enough to be listed on the National Register of Historic Places. That is, this particular brand of prestige does not help property values to appreciate faster than usual. On the other hand, the coefficient for Local Designation is significant (the associated t = 16.7), positive and large, indicating that on average such designation adds 18.6 percentage points to property appreciation in Memphis over the course of 4 years. Evidently, the ability of historic designation to add to property values comes only when the restrictions are in place that are beyond those associated with national designation of historic districts. Interestingly, the additional restrictions created under the aegis of a Local District Designation do not add very much to the appreciation rate, for its coefficient is statistically significant only under the most generous of criteria (t = 1.41). And, while its effect is in net positive, the Local District Designation merely added on an average 2.9 percentage points to the rate of property appreciation over the study period.

Model 3: The Effects of Physical and Neighborhood Variables

It is of course possible that these results are generated from other differences in housing attributes other than historic designation alone. Therefore, as shown in column 3, we added a large number of covariates to the regression model. These covariates increase the fit of the model, as measured by R2. We briefly discuss these covariates and the size and significance of their coefficients here.

The most relevant covariate is Age, the number of years (as measured from the year 1998) since the construction of the unit.4 The coefficient is insignificant, and its size quite small (0.0002), although it has the expected negative sign. The size of the coefficient is not a surprise because previous research indicates that the age coefficient in hedonic price equations typically is in the 0.0020.01 range (see, e.g., Rubin 1993). On the other hand, in other hedonic studies the coefficient on the age variable is typically significant; this certainly is not the case here. We surmise that in the context of historical neighborhoods, age has two main confounding effects: a pure aging effect-where greater age is associated with higher maintenance costs-and a cachet effect-wherein older properties become more valuable. (For a sophisticated take on this confounding effect, see Clapp and Giacotto (1998).) We describe our attempt to articulate the age effects in more detail later in this article.

The next group of variables measures the physical attributes of the property. Among them we first elaborate the basic parameters of property assessmentsize. The evidence reveals that properties with a \argerPerimeter, broader building Footprint, and more sizeable Living Area naturally appreciated more rapidly (i.e., were precisely estimated with statistically significant t ratios and had expected positive signs).5

Interestingly, structures with more stories to them also appreciated more rapidly. The continuous variable Stories denotes the number of stories, including fractions of them, in the primary structure. The series of nested binary variables Stories ≥ 1.5, Stories ≥ 2 and Stories ≥ 2.5 take on the value of unity, when the building has more than the indicated number of stories.6 All the variables are significant, and only that Stories ≥ 2.5 was negative. Note that the negative effect of this binary variable is overwhelmed by the corresponding positive effect of the continuous variable Stories.

Because the footprint of the building helps to define or is defined by some of the other measures of building size, we interacted it, assuming that we would discount some of the benefits to property appreciation of the size of the Living Area and the number of Stories by doing so. Indeed, the pair of interactions was significant with negative signs. In fact, the interaction term Footprint × Stories is so large that we felt compelled to calculate whether it washed out that of Stories. We found that the coefficients cancel each other, when a footprint is about 30,000 square feet, a mansion of large proportion. The largest building footprint in our population, however, was 16,731 square feet. Thus, the value of the term implies that all properties in our database received strong appreciative benefits from having more than a single story.

The next subset of physical property characteristics pertains to land use type and the number of living units in the structure. More living units denote more income-generating power from the structure. Thus, we expect this variable's sign to be positive. But we thought that, for a given physical size, a building would yield net decreasing returns to the Number of Living Units. Hence, we also tried a set of polynomial terms as well; only the squared and linear terms provided any explanatory power, so the others were dropped. Both of the remaining terms were significant, although the linear term was positive and the squared one was confoundingly negative. Figure 1 displays the net effect of the two terms Number of Living Units and Number of Living Units2. Note that the benefits to appreciation of the number of living units do diminish. The buildings in our database began to do so starting with a third unit. Moreover, the average benefit per unit even appears to go negative, when there are five units.7

Enlarge 200% Enlarge 400%

Because nearly all of the properties (95.4%) we studied were residential, we hypothesized that the limited supply of properties reserved for other land uses within the well-defined neighborhoods sampled here would be in high demand. As a consequence, we expected their appreciation rates to be somewhat higher than average. Our expectations were borne out. The variable denoting properties designated for Commercial Land Use purposes was negative, while that for Other Nonresidential Land Uses was significant and positive.

It is common knowledge that certain types of rooms in a home enhance a property's potential to appreciate more than others. The number of bedrooms, bathrooms and even garage spaces and fireplaces are best known for their ability to support home prices.8 On the other hand, for a specific home size, and after accounting for bedrooms and bathrooms, having more other kinds of rooms robs space from what are typically common areas. Therefore, having more such rooms should be detrimental to its broad market appeal, dampening its ability to appreciate. Thus, we specified the variables for the Number of Bedrooms, Full Bathrooms and Half Bathrooms. These variables all have negative coefficients; roughly speaking this leads to the above interpretation. We also added a variable representing the Number of Exterior Fireplace Stacks. In general, most of these variables yielded no significant effect on the property appreciation rates, although there were exceptions. Having more than the average Number of Nonbath Fixtures as well as more than the average Number of Rooms each yielded a significant but small negative effect on property appreciation rates.9

Equation (2) suggests that the change in attribute levels should also be included in the specification as well as their change between 1998 and 2002, and it suggests that the coefficient should represent the level coefficient (in 2002) from Equation (1). Of course most properties do not experience any changes in their attributes (except time), and so our ability to estimate «2 in level form is greatly limited. We included variables for Change in No. of Bedrooms and Change in No. of Bathrooms, these being the two characteristics that had more than a trivial number of changes. Both these variables are positive and significant at the usual levels, as would be expected. Bathrooms are a particularly potent home improvement.

To account for the very real possibility that historical dwellings are more desirable because of stylistic characteristics that are either currently fashionable or in particularly high demand, we also tested binary variables denoting various kinds of exteriors. Those identified as significant were Frame, Stucco, Brick Veneer and Stone.10 Only Stone and Frame were expected to yield superior appreciative value. Each of the included exterior wall types, perhaps surprisingly, received a positive coefficient: Stucco's was the greatest in magnitude.

Properties in the database were also characterized by a set of 16 architectural styles: Colonial American, English, European, Old Style Two Story, Traditional/Conventional, Bungalow, Ranch, Raised Ranch/Split Level, Shotgun, Contemporary, Cottage, Cape Cod, Townhouse, Rowhouse, and Other. Again, we had no expectations and limited the set included in the model to those that were able to enter in with significance (or close to it). Four remained: Colonial American, Old Style Two Story, Traditional/Conventional and Bungalow. Only Colonial American had a negative coefficient.

We were also able to attempt to articulate style differences through seven roof types: Gable, Hip, Gambrel, Mansard, Pitched/Shed, Mixed and Other. As with the other style variables, we opted to include only those that came in with significance (or close to it) and had few expectations. The single notable exception was our expectation for Mansard roofs, which are associated with Victorian homes, that they might yield higher value. Perhaps surprisingly, only Hip roofs yielded a statistically different result, which was positive and not large in magnitude.

Finally, Table 2 reveals coefficients for variables that carry the names of the designated areas. These coefficient values are attached not to binary variables strictly for the neighborhoods identified but to both the historical neighborhood and its matched, undesignated neighborhood. These variables act as controls for unobserved characteristics in these matched neighborhoods. For example, the binary variable Central Gardens Pair indicates location in either that neighborhood or its "companion," nondesignated neighborhood, Chickasaw Gardens. The binary variable Evergreen is an exception, as it has no companion. The neighborhood pair omitted from the list and included in the intercept is that for Hein Park and its companion, Red Acres. As it happens, all the coefficients are positive, indicating (conditionally) higher prices for these neighborhoods (than the Hein Park/Red Acres pair), although Evergreen and Annesdale-Snowden Pair (including its companion Annesdale-Rozelle) yielded no statistically different result.

The most important facts to be gleaned from column 3 of Table 2 are mild changes in the coefficients that describe the historical status of the units. The coefficient on National Designation moves from -0.012 to -0.043 and becomes significant. Hence, accounting for the structural differences exhibited in nationally designated areas indicates that properties in these neighborhoods appreciate more slowly. While the coefficient for Local District Designation increases slightly, it remains imprecisely estimated. The most important finding in this column is that the coefficient on Local Designation remains significant, although its magnitude declines by about 3.8 percentage points to about 14.2%. The ability of local restrictions to significantly raise appreciation rates remains in force even after accounting for differences in the characteristics of properties in the different neighborhoods. Most interestingly, only about 4 percentage points of the difference in appreciation rates in locally designated historic areas is due to these other characteristic differences.

Model 4: A Polynomial Articulation of Age

As noted above, one of the puzzles from this specification is the small and insignificant coefficient on Age. Perhaps this is because part of the age effect is captured by the designation and neighborhood variables, but it would seem more generally that there is great variation in the effects of age when comparing historical and other buildings. While Coulson and Leichenko (2001) used piecewise interaction terms to model this phenomenon, we opt for a more flexible parameterization through the use of polynomials. We add the second through seventh powers of Age to the specification; all were highly significant (except for the cubic, which we subsequently omitted). We interact Age with National Designation to allow the appreciation path of historic and undesignated properties to differ, and the first, second and third powers of Age are interacted with Local Designation to allow the time paths to differ across the different designation modes. We also allow for there to be a shift in the age-appreciation rate relationship for houses built before 1900 (more than 102 years of age) by creating the interaction variable Age χ Built Before 1900, which takes on the value of Age only for those properties built before 1900 and the value zero otherwise.11

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

Enlarge 200% Enlarge 400%

The results of this exercise are presented in the final column of Table 2. The structural variables are robust; their coefficients remain virtually unchanged in terms of significance and magnitude. An exception is a change in the sign of Evergreen, although the coefficient remains insignificant. There is some movement in the variables that indicate historical designation status, but the overall conclusion that the main impact of designation comes about with local designation is unchanged. Indeed, it is even strengthened in that the coefficient on Local Designation increases from 0.133 to 0.211 (23.4%). While Table 2 displays all the coefficients' values for the polynomials, and they are not readily interpreted in that format. Instead, we present Figure 2: a plot of the polynomial effect of Age on the appreciation rate against building age for each of the three types of neighborhoods. (We do not separately measure the Local District Designation here.) There are several interesting things to be gleaned from this figure.

Enlarge 200% Enlarge 400%

First, the impact of neighborhood designation is uniformly greater when that designation is tougher (i.e., emanates from local preservation authority). Moreover, regardless of age, the increment to the appreciation rate is always higher on average for properties in locally designated neighborhoods.

Second, the impact of that local designation tends to be strongest for the newest properties. The difference between the local line and the other two lines is greatest for buildings that are less than a few years old. (Recall that all of the neighborhoods have some newer construction in them, despite the restrictions in these areas.) Moreover, it appears that these buildings reap the most benefit from whatever drives price outcomes in historic districts.

Third, while new buildings had the highest appreciation rates, the apparent appreciative benefit of newness dropped rapidly with age through to those about 25 years old and older. Structures between 24 and 50 years old had similar appreciation rates for undesignated and nationally designated neighborhoods. The appreciation rates of properties in locally designated districts, however, tended to increase with age for those between 28 and 80 years old. It is tempting to interpret this as an example of the conflict of depreciation and style. Buildings in historical areas (at least locally designated areas) that have neither the cachet of being quite new, nor the cachet of being fashionably old will have the lowest appreciation rates. But during the 30-80-year age range buildings acquire whatever historical cachet is available to them, and, within the context of an overall preservation effort, become more desirable.

Fourth, the appreciation rates of all buildings more than 80 years of age or so in all neighborhoods tended to become successively lower. While we allow for there to be a shift in the function for very old units, and this shift is positive, the relevant coefficient is not particularly large, although it could be deemed statistically significant under a liberal rule. The fact that its significance is in question is not surprising in that only 26 observations in the data set were built before 1900. In this sense, the effect of these very old buildings is included in the drawing of Figure 2 only for interest's sake. We speculate that the overall successive age-associated reduction in the appreciation rate with properties more than 80 years old is the result of a perceived increased need for maintenance and deterioration that occurs with these older properties, which overcomes whatever additional prestige accrues to such units.

There are several obvious caveats when interpreting Figure 2. The first is that the age polynomials are probably fairly collinear; the individual coefficients may be sensitive to small changes in the data or in the specification of the regression, although such small changes that we did implement did not reveal any major alterations to the pattern seen here. A more important caveat is that the higher polynomials can only become less collinear to the extent that there are rather old buildings in these neighborhoods. Such buildings, indeed, do exist in the data set, but as alluded to earlier make up only a small portion of it: 2% of the structures in the sample are over 100 years old. Under normal circumstances, this is a large number of buildings, but in trying to identify statistically significant price patterns for very old units, the evidence we are working with is quite thin. The third caveat is that this is a cross-section and as such is a model of vintages, not of actual aging effects, which would be observed only over time.

Conclusions

The literature on the effect of historic designation of neighborhoods has sharpened over the years. It started in the mid-1970s with comparison of average aggregate neighborhood property values, and since the early 1990s, it has elevated to address individual property values. In this article, we believe we take the analysis one step further by analyzing the change in property values, rather than by simple differences in assessed property values, across comparative designated and undesignated neighborhoods. We believe this nullifies some of the objections of using assessed values in such an analysis, while at the same time mitigating some of the bias that may be inherent in the differences between designated and undesignated neighborhoods that are otherwise deemed to be similar.

Our analysis used a rather unique set of appraisal data for the years 1998 and 2002 obtained from Memphis's Landmarks Commission. As in several prior studies, our data set contained relatively equal numbers of properties in designated and undesignated districts. It also contained a single historic neighborhood with no undesignated companion that had a large swath of historic structures replaced by new construction during the past few decades.

After controlling for numerous variables that mostly pertain to differences in architectural style, functional features and housing quantity, we find across these Memphis neighborhoods that when properties were in neighborhoods zoned historical by the authority of the City of Memphis, it significantly raised property values at rates above those in other similar neighborhoods, that is, 14-23% higher. Given that local designation is a more important determinant than national designation, it is possible to view this result as arising from the stricter guidelines embodied in local designation (which may be manifested in more assiduous upkeep, for example) rather than the cachet effect of designation, although this is a tentative conclusion.

We also found that the relationship between the age of the property and the change in assessed value was quite nonlinear. When plotted, the relationship looks more or less like a bathtub with more rapid rates of appreciation for all young properties (less than 10 years old) and for locally designated, moderately older properties (30-80 years old).

Finally, one of our most interesting findings is that new properties benefit as much, perhaps even more, than older properties from being within a historic district. This is a phenomenon that, as far as we can discern, has not been noticed in the literature to date. Indeed, we intend to pursue this and the appreciation of commercial land use in the near future.

We thank Amy Crews-Cutts, Johan Lundberg, Tom Thibodeau, two anonymous referees and participants at the 2004 AREUEA Midyear and 2004 North American Regional Science Annual Meeting for helpful comments. We also thank Jennifer Tucker and other staff with the City of Memphis for their cooperation and generous assistance. Finally, we acknowledge the assistance of Rodrigo Duran who painstakingly labeled items in the study database, so we could figure out what we were doing.

[Author Affiliation]

* Department of Economics, Penn State University, University Park, PA 16802-3306 or fyj@psu.edu.

** Center for Urban Policy Research, Rutgers University, New Brunswick, NJ 08901-1982 or lahr@rci.rutgers.edu.