The Influence Of Transportation Accessibility On Property Value

Contributors to this project include Viggy Kumaresan, Azucena Morales, Yifei Wang, and Sicong Zhao.

What Does Transportation Have To Do With Housing Prices?

For the first time in human history, more humans live in cities than in rural areas. This urbanization movement has had many dramatic effects on city life, including a rise in housing demand as well as an increasing reliance on public transportation [1]. Today, in dense cities like New York City, transportation availability is considered an ever-important factor in selecting a place to live. However, many others features also come into consideration such as unit size, amenities, proximity to landmarks, and safety.

From a quantitative standpoint, exactly how valuable are different transportation methods in an urban property? Using residential property sale price as a proxy for value, we can attempt to answer this question. By integrating features of housing proximity to the Metropolitan Transit Authority (MTA) and bike share stops as well as Taxi and Uber data, we are able to predict NYC housing prices and ultimately better understand the impact of each transportation channel on property value. The results of this analysis have implications for the urban planners and real estate developers by offering the quantification of the perceived value of transportation services.

Industry Disruption

Previously, researchers have developed machine learning models to predict housing prices using various neighborhood features, such as crime rate, local business presence, and access to highways [2]. In recent years, as more urban data has become available, researchers have improved these models by integrating additional data sources such as local census data and education profiles [3]. In regards to transportation specifically, increased accessibility has been shown to increase housing prices, due to the decrease in commuting costs for both tenants as well as landlords. On the other hand, there are also negative effects of transportation such as noise, traffic, pollution, and crime that can have potentially negative effects on property value [4].

Nowadays, there are more options for mobility in cities apart from subway transportation. Services such as Uber, Lyft, and Citi Bike are changing the way people transport daily. Because New Yorkers can now carpool into Manhattan for just a few dollars, many of residents are rethinking their priorities as homebuyers [5]. Buyers who do not live close to a subway station, have found alternatives in these services. This industry disruption does not mean that traditional public transportation is being replaced, but instead that a wider, holistic transportation system may begin to affect real estate property value.

Data Sources

In order to assess the impact of transportation accessibility on housing prices in New York City, New York, we integrated several data sources:

NYC Property Prices And Transportation Stations

Taking a look at housing prices in NYC, the highest property prices cluster around lower Manhattan and West Brooklyn with property sales decreasing with distance from this hub. Subway and bike stations also cluster around this area. Citi Bike System stations in particular are exclusively located in lower Manhattan. This posed a challenge during analysis as distance to the nearest bike station can also be considered a proxy for being in this high priced area.

Data Preprocessing

Linking transportation information to each property was integral to this research. Each property was geolocated using Google Maps API. Features based on property proximity to transportation methods were calculated using these coordinates.

Results

In a comparison between linear regression, k-nearest neighbors (KNN), classification and regression tree (CART), neural network, and random forest approaches, a random forest approach produced the most accurate predictions, explaining 61.2% of variance in property prices. Random Forest models are especially useful for interpretation because they produce ranked feature importance scores to indicate which variables were most useful in property price predictions.

Random Forest Feature Importance

Importantly, for each modeling approach, performance was compared to an identical model without the inclusion of transportation accessibility. By including transportation data from subways, bike shares, taxis and Uber to our random forest approach, we explain an additional 5% of the variance in housing prices. In terms of specific features, we still see that square footage and year built are the most predictive features in our baseline model. However, we also see that the nearest distance to a bike and subway stations are also important predictors.

Our random forest model produced a variety of predictions, some accurate, some inaccurate, and some mediocre. The housing sale price in NYC Property Sales dataset has a long tail on both sides; some houses are extremely expensive while others are extremely cheap. There are undoubtedly other features not included in our datasets contributing to the price, such as luxury amenities or views. Since the transportation features are determined mainly by property location and not property features itself, our model tends to average the house price in certain areas, and thus fail to predict the outliers.

Final Thoughts

So maybe location isn’t the most important feature in a property after all, at least as it pertains to proximity to transportation options. Even after integrating transportation accessibility into the analysis, the most influential property features were still square footage and age. That said, transportation accessibility does appear to be predictive of housing prices in some capacity. The addition of transportation features did explain an additional 5% of variance in housing sale prices. Moreover, transportation features such as distance to bike and subway stations were also highly important features in our random forest model.

It should be noted that this analysis does not determine a causal relationship between transportation accessibility and price. Often transportation infrastructure is created in accordance with population density and demographics. For example, although a properties distance to the nearest bike station was the third most important feature in our model, all bike stations are located in Manhattan and Brooklyn. This finding most likely only indicates that these areas are already popular and desirable. Future work can hopefully continue this research through a causal frameworks. Ultimately, urban planners, real estate developers and home buyers can all benefit from this work through better understanding of the value of transportation accessibility in property value.

References

[1] Strekas, T. (2005). "New York as a Model for the Study of Urbanization." CUNY Institute to Nurture New York’s Nature.

[2] Jingyi Mu, Fang Wu, and Aihua Zhang. (2014). “Housing Value Forecasting Based on Machine Learning Methods.” Abstract and Applied Analysis, vol. 2014, Article ID 648047, 7 pages.

[3] Gao, G., Bao, Z., Cao, J., Qin, A. K., Sellis, T., & Wu, Z. (2019). Location-Centered House Price Prediction: A Multi-Task Learning Approach. arXiv preprint arXiv:1901.01774.

[4] Kilpatrick, Throupe, Carruthers, & Krause (2007). The Impact of Transit Corridors on Residential Property Values. Journal of Real Estate Research, 29(3), 303-320.

[5] Small, E. (2014). Sayonara, subway: How ridesharing apps are changing the real estate calculus for brokers and developers. The Real Deal.

An ABM Approach to Addressing the Opioid Crisis: How can we predict the effects of public health intervention on a unique community of pain patients?

• More

• Shoot me an email
• Connect on LinkedIn
• Check out my Github
• Download my Resume