Modeling the Commercial Property Value Using Ordinary Least Squared (OLS): A Case Study of Putatan, Sabah and Limbang, Sarawak

_____________________________________________________________________________________________________ Real Estate is an asset that provides profitable investment in return. Commercial property constitutes an important part of the real estate sector. In valuing commercial property, rental value is an essential component for valuers in applying valuation methods. Determining the rental value usually a difficult process as it involves a lot of influence factors. There are various factors that can be used but not the same for every commercial property. Therefore, this paper shows the modeling valuation comparison between two commercial property areas of Putatan and Limbang that represent the outskirts of the city in Sabah and Sarawak respectively. The purpose of this study is to find an effective approach to develop a suitable model for commercial property valuation using OLS and subsequently intends to identify factors that influence the commercial properties for both study areas. The OLS technique was used for this study to develop the property valuation model in Putatan and Limbang. The outcome shows that both study areas can be modeled using OLS for property valuation using similar factors but the Limbang area produced higher accuracy than Putatan based on the adjusted R2 value. However, in terms of the significant of the property value influence factors, both Limbang and Putatan produced different significant factors. Thus, it shows that most of the outskirt city commercial property valuation must be modeled using different influence factors. The model will benefit the local authorities, especially for commercial property valuation. Ultimately, revaluation also can be done easily with low cost, less time and few people needed for this approach.

www.msocialsciences.com Millington (2001) defines property valuation as the art, or science, of estimating the value for a specific purpose of a particular interest in property at a particular moment in time, considering all the features of the property and also considering all the underlying economic factors of the market, including the range of alternative investments. Property valuation is also used to estimate property values for various uses such as sales, purchases, mortgages, rentals, insurance, estate taxes, stamp duty, and valuation of real estate taxes (Shapiro et al., 2009). Other than that, property valuation not only helps local authorities to calculate the right property tax to generate income in a developing area, but it also led investors to find suitable property areas for marketing, helping developers to track potential areas for the construction of real estate and propose refinancing or sale of real estate to the owner (Oliver, 2008).
Study regarding property valuation modeling on commercial property especially shop houses were never conducted in Limbang and Putatan area compared to other major cities in Sabah or Sarawak (Selamat Jati Yanjah, 2017;Christy Onyo, 2017;Watson Jeffrey Kaling, 2018). However, Limbang and Putatan were developing districts with various development activities being carried out such as housing development in selected areas construction of new industrial areas. Thus, this research is important to determine and assess to which extent the development has brought changes to the property sector in Putatan and Limbang.
Regression analysis is an analysis that is widely used in spatial statistics (Alexander and Schuster, 1998) and it can be used to model, examine, and explore spatial relationships, and that is to explain the factors behind the observed spatial patterns and predict the results (ESRI, 2017). Ordinary Least Square (OLS), which is one of the regression techniques, often used in estimating the rental value of the property based on location (Scott, 1988;Wyatt, 1997;So et al, 1997;Oliver and Ibrahim, 2007).
According to Oliver (2008), property valuation using spatial statistics within Geographic Information System (GIS) integration is very important as it is able to accelerate, facilitate and contribute to assessment activities. However, to measure or identifying the factors of influence towards the commercial property value, the property areas have to be segmented into separate models instead of an overall model (McCluskey et al., 2011). Therefore, this study was conducted to determine whether the two areas of Limbang and Putatan share the same factor of influence or must be modeled with different factors of influence.

Location factor
Location is a key factor in determining a property's value (Oliver and Ibrahim Sipan, 2008). A good location will influence the degree of visitors' accessibility to a business location (Netzell, 2013). Therefore, investors or businessmen are trying to find the best and will directly increase demand for the business lot. According to Gallimore et al. (1996), location will increase the value with multiple location influence. In other words, the influence of location is not seen in terms of the location of the business lot but rather the influence of the relative location of the business lot with the surrounding amenities such as the distance of business lot from the mall's main entrance and the distance from the amenity.
According to Wyatt and Ralphs (2003), location factor influencing commercial properties include access to markets, suplliers, open spaces, parking lots and train stations. Nevertheless, the influence of location factor on retail commercial real estate differs from business complex. This is due to the location factors affecting the business complex is more internal.
In identifying the factors that affect the rental value, there are several techniques that can be used such as Ordinary Least Squares (OLS). In this study, the OLS technique was used to see the influence of location factors in general in affecting the rental value.

Ordinary Least Squared (OLS)
OLS is the most well-known analysis of all regression techniques (Scott and Janikas, 2010) as it has a function to predict a dependent variable from a set of independent variables (Tabachnick and Fidell, 1989). Shaw and Wheeler (1985) also argue that OLS is used to produce a relationship model between dependent variables with one or more independent variables, measuring error by using relationships to make predictions against dependent variables and measuring the relationship strength (correlation) between variables dependent and independent variables.
OLS is a global model that forms the single equation to represent the relationship between the object to be modeled with each variable description (Oliver and Edmund;2010) It is quite easy to examine the model's assumptions such as linear, variance constant and the effect of outliers using a simple graphical method (Hutcheson and Sofronio, 1999). The formula can be stated as follows: Where, V is the estimated rental value for each shophouse which is calculated as the sum of B0 (constant) and location influence variables (B1X1…… BnXn).
With the integration of GIS, OLS now becomes part of spatial statistics that could be utilized to perform spatial regression in global environment. Such an approach potentially could help local authority to speed up revaluation process and reduce cost.

Methodology
In this study, for data collection, the non-random selection sampling techniques was used in this research. The questionnaires were distributed by the researchers to obtain data on the rent to owners of shops in the study area. Whereas, for the existing data such as maps of shophouse lots were obtained from Limbang Division Land and Survey Department, Putatan District Council and Property Services and Valuation Department. In this study, the data of 232 ground floor shop houses rental have been collected in the area of Limbang and 100 ground floor rental for Putatan area as shown in Figure 1   Meanwhile, 20 location factors such as renovation, main roads, secondary roads, educational centers, leisure centers, banks , police stations, post offices, bus stops or taxis, ferry terminals, public toilets, places of worship, petrol stations, parking lots, shopping complexes, office complexes, sports complexes, farmer 's markets, convenience and position have been used for the OLS modeling purpose for both study areas.
Attribute data for the location factors in this research using the "Dummy" code or indicator variables (1 or 0) that describes the effect of the location of the property. For example, the code "1" indicates that the property was influenced by the location factor, behind the code "0" describes the area that was not influenced by the location factor. Generally, the more the area influenced by the location factor, the higher the influence received. In addition, this research does not use measuring tools such as GIS buffer analysis or network analysis to measure the distance between the location factor and the area. Based on the results above, it can be said that the OLS model for Limbang is better than Putatan, in which the adjusted R2 of 0.70 for Limbang is higher than Putatan (0.59). Theoretically, the higher the R2 the better the model to predict the value of the property (Hishamuddin M. Ali, 1998). OLS model is also capable to identify the significant and the strength of the influence of the location factor to the shophouses' rental value (Oliver, 2006).

www.msocialsciences.com
To assess the goodness of fit and residual of the model, the comparison between different regression models were examined with the Akaike Information Criterion (AIC) measurement. Comparison between the spatial model would be examined with the model having the lowest AIC will be the better model provided that the differences in value are more than 3 (Fotheringham et al., 2002;ESRI, 2013). AIC for Limbang proves to be the better model as the value of 2892.2994 was smaller than Putatan with 5993.4010.
Based on the OLS analysis, the OLS model for Limbang district shows that there are only 6 significant factors that influence the shophouse property value at 95% accuracy level which consists of Market, Petrol Station, Public Toilet, Recreational Park, Position and Office. While the Putatan district only has 5 significant factors that comprise of renovation, road, main rubbish area, shopping complex, and Hotel.
To determine the strength and type of relationship the independent variable has to property value, the coefficient for each of the independent variable were measured. Table 2 and Table 3 show the coefficient value of each significant independent variable for the Limbang and Putatan area respectively. For Limbang in Table2

Discussion
Ultimately, based on the two models of OLS Limbang and OLS Putatan, it can be said that the significant influence factors for both models are different even though there are using a similar list of variables. This shows that each area need to be developed with different model to accurately estimate the property value of the area. The Limbang model, however, provides the highest accuracy and suitable to be used for the authorities compare with the Putatan model. This could be due to the factors involved in the model really reflect the commercial property value in the Limbang area (Christy Onyo, 2017). A different set of influence factor might be needed to obtain an accurate OLS model for Putatan.

Conclusion
Identifying a significant property valuation factor is a difficult task especially when conducting manual valuation. By using OLS modeling, this activity can be done easily. Other than that, OLS modeling can help determine the accuracy of the property value prediction. Although both study area of Limbang and Putatan are within the outskirt city, the modeling shows that the significant influence factors for the commercial property valuation are different. Therefore, different models must be developed for different commercial property areas to accommodate the significant influence factors that exist in the said area.
Even though the OLS model can produce good property value prediction but this can only be done on a global scale. Another regression modeling through Geographical Weighted Regression (GWR), also capable of conducting property value estimation but on a local scale. This is suitable for non-stationarity property value in the area, which means there are high differences of value in certain areas that only GWR can detect. Therefore, GWR modeling would be another method that should be explored for further studies in Putatan and Limbang. Finally, this study will benefit the local authorities, the investors and the business people to identify the property value influence and the estimated value with low cost, less time and few people needed using this approach.