No abstract

This paper analyzes the impact of the newly operated Dubai Metro on the sale transaction value of dwellings and commercial properties. The effect is estimated for properties within different catchment zones of a metro station using difference-in-differences and hedonic pricing methods on both repeated cross-sectional data and pseudo panel data. Our estimates show a positive effect of the metro on sale values of both residential and commercial properties, although the effect is stronger for commercial properties. The models also reveal that the effect of the metro on the value of dwellings and commercial properties is largest within 701 to 900 meters of a metro station and is about 13 percent and 76 percent, respectively. IntroductionA large number of researchers have examined the effect of rail systems on property values and the range of estimates varies substantially across studies (Mohammad et al. 2013). The majority of studies suggests that proximity to rail stations enhances property values (e.g., Laakso 1992;Pan and Zhang 2008;Voith 1991;Weinberger 2001), some indicate a negative impact at certain locations mainly due to negative environmental externalities (e.g., Bolling, Ihlanfeldt, and Bowes 1998;Cervero 2003;Du and Mulley 2006), and a few show no noticeable effect (Gatzlaff and Smith 1993).The empirical literature has generally adopted hedonic pricing (HP) methods to estimate the relationship between rail and property values. Although HP models can control for unobserved heterogeneity across properties, they cannot identify a causal relationship between rail access and property value, only indicative results. For example, stations may be found at high-valued areas like a commercial hub zone, and HP models typically do not account for the impact of these attributes on the reported values (Billings 2011;Gibbons and Machin 2005).As an improvement to the HP models, a number of studies have recently started using an innovation-based model, the difference-in-differences (DID) estimator (e.g., Agostini and Palmucci 2008; 1995;McMillen and McDonald 2004). Although the application of this method differs across studies, it generally relies on comparing prices before and after the treatment for the properties that experienced the treatment (treated) compared to those that did not (control). Some studies have also compared results obtained from DID to those from conventional HP models. In a study on the effect of new rail stations on dwelling prices in London, Gibbons and Machin (2005) found that HP models produced statistically larger effects than DID models. On the other hand, Agostini and Palmucci (2008) found that the DID models produced higher estimates of the effect of Santiago Metro on property values, compared to HP models. The difference in the magnitude of estimates between the DID and HP models in the two studies may be related to the data structure; while Agostini and Palmucci (2008) used repeated cross-sectional data, Gibbons and Machin (2005) used pseudo panel data.This paper contributes to...

No abstract