I. Introduction
The purpose of this study is to examine the impact of a railroad station and railroad tracks on residential property values within one borough. Existing research has separately examined the effects of railroad stations and railroad tracks on residential property value but has not examined them together. Since a railroad station necessitates railroad tracks it is important to study these variables together.

II. Literature Review
Several studies were performed in the first half of the 1990’s that showed the impact of railroad stations on nearby property values (Diaz). A study that analyzed several different California transit lines found the impact of the railroad stations was generally mixed. Housing values near the Bay Area Rapid Transit stations in San Francisco were positively impacted by the railroad stations. In Alameda County, residential properties had a $2.29 increase in value for every meter closer to BART stations, while residential properties had a $1.96 increase in value for every meter closer to the Bart stations in Contra Costa County. Rail stations on the San Diego light rail trolley line also had a positive impact on nearby housing values, which had a $2.72 increase in value for every meter closer to the trolley stations (Landis, 1994).

Some California railroad lines’ stations had a negative impact on property value, including San Mateo County’s CalTrain commuter line and the San Jose light rail. Residential property values decreased $1.97 per meter closer to the stations; however this negative impact might have been due to the interaction between the railroad stations and their proximity to industrial and commercial areas. Similarly, the CalTrain stations had a negative impact that may have resulted from rail cars that were not designed to be quiet and noise levels higher than most other rail lines (Brinckerhoff, 2001). The study incorrectly implies that rail cars noises are associated with rail stations, when they are actually a problem associated with railroad tracks.

In most other studies, railroad stations had a positive impact on residential property value. In Portland, Oregon homes within 500 meters of the MAX Eastside line, enjoyed a 10.6% premium. A further conclusion of this study showed that the positive effects of accessibility negated the negative effects of nuisance, but only in areas where transit plays a major role in people’s lives (Musaad, 1993). A study in Dade County, Florida similarly showed a 5% premium for real estate near the Miami Metrorail compared to all other real estate in Miami (Gatzlaff, 1993).

A study in suburban Philadelphia encountered similar problems. To determine the net change in real estate values associated with the Southeastern Pennsylvania commuter rail system (SEPTA), house values for census tracts with SEPTA stations were compared to census tracts that did not include SEPTA stations. It was determined that a 3.8% premium existed on median housing values for real estate in Philadelphia suburban census tracts with SEPTA railroad stations (Voith, 1991). Furthermore, “A 1993 study of Montgomery County, the most populated county in the Philadelphia metropolitan area, looked at annual variations in property values by census tracts. Those census tracts with train stations enjoyed a 7.8% premium over similar homes in tracts without a rail station” (Monmouth County Planning Board, 1997).

One problem with these studies was that they only measured the benefits of property values in towns with rail stations compared to towns without rail stations. It is clear that areas with railroad stations typically enjoy a premium on property values over areas without railroad stations; however the effects of immediate proximity to railroad stations on property values within towns that have these stations is undeveloped.

There are fewer studies pertaining to the effect of railroad tracks on property values. In Oslo, Norway residential properties within 100 meters of railroad tracks had their values reduced by an average of 23%, which might be due to aesthetics, vibrations, and noise. Houses outside of 100 meters from the railroad tracks did not have a statistically significant reduction in property value (Strand, 2000).

Other studies show the negative impact of transportation noise on residential property value. It was determined that for road noise louder than 50 decibels, properties were discounted 0.88% per decibel. Additionally for road noise louder than 39 decibels, there was a discount of 0.6% per decibel for each property (Nelson, 1982). Air travel noise is a similar nuisance as railroad noise because it is of a permanent nature and imposed on the property owners. Single-family homes under or near heavy flight corridors suffer a universal negative impact of property value (Bell, 2001).

The current study combines effects of railroad stations and railroad tracks on property values. It studies the impact of residential property values due to proximity of railroad station and railroad tracks within a town, rather than comparing property values of towns with stations and tracks to property values in towns without stations and tracks.

III. Overview
To measure the effects of the railroad station and railroad tracks on residential property values a hedonic pricing model, a form of regression analysis, was used. Such a pricing method is common among real estate studies because housing prices are related to its multiple measurable characteristics. Thus, it was possible to isolate the effects of both railroad tracks and a railroad station on the values of residential property, while controlling for other variables such as physical characteristics of the houses.

For the study, 777 residential units were selected from North Wales, Pennsylvania in Montgomery County. A SEPTA railroad station is located in the approximate center of the .72 square-mile borough, which also serves as a census tract. A commuter rail line crosses through the town, traveling west to Doylestown and east to Center City, Philadelphia.

North Wales was selected for several reasons. First, there is little industry and commercial properties in the borough. As explained in past studies, such properties may significantly interact with the effects of the railroad, making the railroad falsely appear to have a negative impact on residential property value. Additionally, there are no parks, landfills, nor highways nearby, which have all been proven to have a significant impact on property values (Crompton, 2001 and Recihert, 1997).

All houses sold after 1970 were included in this study. Descriptions of houses and the most recent sale price for each property were obtained through the County Board of Assessment office. Using tax maps scaled in feet, the closest walking distance to the railroad station was measured from all property fronts, as well as the shortest straight-line distance from the railroad tracks to the nearest point of the property. Straight-line distance was deemed a sufficient measure for proximity to the railroad tracks, because nuisance factors such as vibrations need not travel on roadways and sidewalks to approximate location, as in the case of people walking to railroad stations.

IV. Functionality
Typically two types of functions are used in the hedonic pricing model: the linear and log-linear models. The linear model assumes that the dependent variable, selling price, is normally distributed. The Shapiro-Wilk tests the hypothesis that the selling prices are normally distributed. Since p<.001, that hypothesis is rejected. Therefore the housing prices are not normally distributed, so the log-linear model is used, which allows for nonlinear price effects.

This log-linear model enables continuous variables to represent price elasticities, while categorical variables represent percentage impacts on selling price. The price elasticities are percentage changes in the independent continuous variables that affect the percentage change in housing values. Similarly the percentage impacts, obtained from the anti-log of the parameter estimates of the categorical variables, represent the premiums if the indicated variables are included for a residential property (Kang, 1987).

The log-linear equation used for this hedonic pricing model is: ln(price)=a1[ln(Con1)]+aN[ln(ConN)]+b1(Cat1)+bN(CatN)+c1(Int1)+cN(intN)+d+e where,
a = estimated regression coefficients on continuous variables
b = estimated regression coefficients on categorical variables
c = estimated regression coefficients on discrete variables
Con = continuous variables
Cat = categorical (dummy) variables
Int = discrete variables
d = intercept (constant)
e = a random error term

V. Variables and Interaction effects
Housing prices differ based on various physical characteristics. As in most hedonic housing price models, age and size are important continuous variables. Lot size in square feet, living area in square feet, and front footage of the property are all included size measurements. Total rooms, bedrooms, and stories highly interact with the size of living area and each other. As seen in Graph A, living area is positively linearly correlated with the total number of rooms in a house. Number of bedroom and stories are similarly correlated with living area as well; therefore it is only necessary to include size of living area in the model. Studies show that type of house and type of house exterior do not have significant impacts on property values and are typically not included in hedonic pricing models.

Amenities such as living on Main Street, having a full basement, and usage of oil energy were also included dummy variables. Since studies show that commercial and industrial zones influence property values, these two dummy variables were included, as well as a residential zone dummy variable. Fortunately, the correlations between an industrial neighborhood and distance to the railroad station and tracks were significantly low. As seen in Graph B, properties located in industrial zones are scattered throughout the borough. Some industrial zoned residential properties are 2500 feet from the station and almost 2000 feet from the tracks, while others were closer.

Discrete variables were included for number of fireplaces, number of half bathrooms, number of full bathrooms, and number of years prior to 2004 that the house was sold. This variable was necessary to reflect the constant changes in interest rates and inflation, which severely impact residential property values.

The final variables were those being analyzed in this study: feet from the closest point of each property to the railroad tracks and feet from the railroad station’s platform on sidewalks and roadways to the front of each property, assuming an estimation of shortcuts through the parking lot of the railroad station. These two variables show no interaction with a low correlation coefficient. Therefore each distance can be analyzed separately to seek its impact on residential property value.

VI. Regression Results
A stepwise regression process, which adds significant variables and removes insignificant variables while constantly recalculating the model, was used to achieve an adjusted R-square value of approximately 80% with an F-value of 252. Several variables were removed from the model due to their insignificance. Out of 16 variables, 12 remained in the model and significant at the .05 level. Residual tests and Cook-D statistics indicate the model and data were adequate and an accurate analysis may be extracted.

A list of the variables, estimated parameters, standard errors, and p-values are presented in Table 1. The number of years prior to 2004 that the house was sold was highly significant and the most important variable in the model.

Usage of oil and number of fireplaces were removed from the model. Factors such as living on Main St., living in an industrial zoned neighborhood, and the existence of a full basement are related to residential property value and with estimated parameters of .164, -.123, and .124 are significant at 5%. As explained in the log-linear model these parameters represent premiums or discounts. Thus a property with a full basement compared to a property without a full basement had a 13.2% premium added to the house. This is determined by taking the antilog of the parameter, which is .124 (Kang, 1987). The negative parameter of the industrial zoned neighborhoods is consistent with past studies that show a negative effect from proximity to industrial areas (Diaz).

Several other discrete and continuous variables impacted the value of a residential property. Number of full bathrooms and number of half bathrooms, with estimated parameters of .077 and .057, tested significant at 5%. The natural logs of the front footage, lot size, living space, and age all proved to impact property value, with estimated parameters of .089, .073, .175, and -.047, were significant at 5%.

Finally, both walking distance to the railroad station and straight-line distance to the railroad tracks tested significant.

VII. Railroad calculations
Both the distance from the railroad station and the railroad tracks, with estimated parameters of .041 and .028 tested significant with p-values of .024 and .0047. Therefore, within the borough of North Wales, properties far away from the railroad station and railroad tracks enjoy a premium on their residential property values.

Because of the log-linear regression model, the distance parameters serve as price elasticities. Every 100% increase in feet away from the railroad station generates a 2.88% increase in residential property value on average, while every 100% increase in square feet away from the railroad tracks generates a 1.96% increase in residential property value on average.

Calculations can be determined by adding the antilog of the distance to the base price and then comparing. For example a house worth $200,000 at 500 feet away from the railroad station, would be worth $205,765 at 1000 feet away and $194,396 at 250 feet away. Similarly a house worth $200,000 at 500 feet away from the railroad tracks would be worth $203,919 at 1000 feet away and $196,155 at 250 feet away. Furthermore, a house valued at $200,000 at 500 feet away from railroad tracks would be worth approximately $179,000 for properties closest to the railroad at about 10 feet away.

Some residential housing units are located as close as 200 feet from the railroad station, while the residential housing units on the edge of the borough are located as far as approximately 4000 feet. Thus, residential property values on the edge of North Wales borough, farthest from the railroad station have about a 13% premium compared to residential housing units closest to the railroad station.

VIII. Conclusion
In North Wales borough, from 1970-2003 sales prices were influenced by the walking distance to the railroad station and straight-line distance to the railroad tracks. Proximity to the railroad tracks had a negative impact on property value because of nuisance factors such as noise and vibrations.

Proximity to the railroad station also had a negative impact on property value within the borough of North Wales. Although property values are higher in a town or census tract with a railroad station than in a town or census tract without such stations, property values are higher on the edge of such towns with train stations, than immediately near the train station. These results suggest that railroad stations provide convenience to the entire town or borough, but the nuisance factors such as aesthetics, noise, and traffic only occurs in the immediate vicinity of the railroad station.

Return Home
Return to Academics
Copyright 2004, Kevin Semanick