For urban fringe farmland, it is very important to properly evaluate officially announced land price based on the objective facts, since it is, as a main area of land policy or program, the key component to accurately understand land market and to eff...
For urban fringe farmland, it is very important to properly evaluate officially announced land price based on the objective facts, since it is, as a main area of land policy or program, the key component to accurately understand land market and to efficiently perform the land policy. Nevertheless, urban fringe farmland price has a great gap between officially announced land price and real transaction price. It is thought that valuation by traditional land rent theory and weakness of ordinary least square estimation in current officially announced land price model have caused. To solve these problems, this study has started when it was known that it was essential to make a model of location influence or spatial effects for properly evaluating the urban fringe farmland price.
Former studies showed that immoveability and individual characteristics of land cause spatial dependency and heterogeneity. Also, they have been mostly on urban land price and it is few that spatial econometrics approaches are applied to estimate urban fringe farmland price which is future urbanized land and is greatly affected by accessibility and development potentials. For this reason, this study has estimated urban fringe farmland price in Seoul Metropolitan Area with SAR, SEM, GSAM and GWR model using real transaction price in order to overcome the weakness of traditional OLS estimators which didn't consider any spatial effect on urban fringe farmland price as a main area of land policy and program.
The result of making a comparing between current land price model and accessibility variables added model showed that the latter has better model-fitness and lower root mean square error(RMSE). Moreover, the existence of heteroscedasticity and non-normal distribution in OLS residuals were verified. Also, applying the inverse distance and inverse square distance weight matrix respectively, spatial dependence in OLS residuals is proved to exist with Moran's I, Lagrange Multiplier(LM) error and LM lag test.
In this research, it is analysed 732 agricultural land parcels located at non-urbanized zone in Gwangju, Kyounggi province in South Metropolitan Area(SMA). As a result, tricube weighted GWR model is proved to outperform all OLS and Spatial Econometrics Models on the ground of lower AIC, SC, and RMSE which is 30% lower than OLS model. Meanwhile, Gaussian and Exponential weighted GWR models have quite higher RMSE than tricube.
It could be said that this study has a great significance to estimate spatial effect controlled farmland price model, using real transaction price which is more likely real market price and closer to real land market situation than officially announced land price.
With the base of the result of this study, it is possibly suggested the policy implications as below.
First, spatial econometrics models could be applied in urban fringe farmland where spatial dependency exists. And accessibility variables such as distance to road, city center, and residential area, based on urban land rent theory, should be added in current officially announced land price model. Moreover, the variables should be input systematically connecting to the National Spatial Data Infrastructure(NSDI), Korea Land Information System(KLIS), and Urban Planning Information System(UPIS) operated by Ministry of Land, Transportation and Maritime Affairs(MLTM)
Second, it is strongly suggested to apply transaction price based GWR models for public purpose land price estimation such as landbanking and public land disposal, both indirect and long-term land policy, especially when calibrating other price factors on land price appraisal.