Hedonic modeling of residential rents in Helsinki

Abstract

This thesis investigates what are the residential rent determinants in Helsinki and how the impact of the housing attributes differs between districts, questions that interest many parties, such as investors, of the real estate market. The study focuses on new rental agreements of rental apartments for which there are no restrictions with regard to their rent levels. These kinds of rental dwellings are most strongly linked to the owner-occupied housing market as well as the financial market.

The rentable area, age and location of a flat have the greatest effects on the rents of market based rental apartments in Helsinki, according to the estimated models. Rent levels decrease exponentially as the rentable area increases, while the flat's age decreases the rent level most when considering flats of the age of 30 and substantially over 90 years olds. In addition, a sauna increases rent level by approximately 5 percent, and rental flats located in non-subsidized buildings also have circa 4-5 percent higher rents compared to flats located in interest subsidized buildings in Helsinki. Furthermore, new apartments have a premium in the rent level. The floor level also increases rent level.

However, the neighbourhood and locational characteristics also play important roles in determining the rent level of rental apartments. Distance to the CBD decreases housing rents as the distance increases, in ceteris paribus. Apartments located in the proximity of a metro or train station have lower rent levels than dwellings located within a circa 1-2 kilometre distance from the nearest station - thus the impact of the station on the rent level is rather quadratic than linear. Distance to a highway, main road and seashore also impact the residential rents. Furthermore, the image of the district reflects either positively or negatively on rent levels in Helsinki.

Nevertheless, the effects of the attributes vary across Helsinki, indicating spatial heterogeneity of the parameters. For example, a flat's age has a different impact on the flat's rent level between the major districts of Helsinki. The effect in the Southeastern Major District of Helsinki is not as large compared to the Northeastern Major District. Moreover, the rent premium of new flats is strongest in the southeastern areas of Helsinki.

In addition to spatial heterogeneity, spatial autocorrelation is also strongly involved in housing rents due to clustered apartments in same buildings and same areas. In consequence, the traditional OLS models produce biased estimates due to the spatial effects while the spatial models outperform the OLS models in terms of AIC and Likelihood ratio test statistics as well as when the goodness-of-fit between the models is compared. However, the spatial models do not describe the rent level variation correctly either since there are several unobserved housing characteristics that the models do not take into account but which still have an effect on the rent levels. Moreover, one has to take into account the representativeness of the data used in estimation when inferring and applying the results. Therefore, the estimated models have some shortcomings when they are applied, for example, in rent level evaluation across Helsinki.