House Value Estimation - Research Paper Example

 HOUSE VALUE ESTIMATION CONSIDERING LOCATION AND OTHER FACTORS THROUGH REGRESSION MODEL ABSTRACT The value of a house is considered to be subjected to many factors like the manufacturing quality of the house, its architecture, the design of the house and the locality in which it is situated etc. This paper attempts to estimate the relationship between the house price/value and factors that influence it. This is done by introducing a model for predicting this relationship using regression. The influencing factors can be termed as the characteristics for regression. These characteristics, gathered for regression, are of two types namely the home characteristics and the neighborhood characteristics. The home characteristics can further be classified into ‘the number of rooms in a house’, ‘the number of full and half bath rooms’, ‘availability of garage’, ‘cellar’, ‘balcony’ etc. Similarly, the characteristics related to the vicinity or the neighborhood of a house can be termed as the neighborhood characteristics. These characteristics can further be classified into ‘roadside’, ‘traffic intensity’, ‘bothersome location’, ‘crime prone vicinity’ etc. The scope of this paper is to consider all the above mentioned factors for the assessment of house value by testing the level of their significance. The expected outcome is termed as a prediction model. INTRODUCTION The assessment of the price of a house is generally assessed through the few apparent factors like the construction value, design and location of the house. The rough estimation usually does not encompass the related factors that are important to decide the value of a house and have a significant impact on the assessment process. The study of the factors other than the common factors is important to identify their role in the determination of the price or value of a home. Generally a realtor’s claim would always be that the location is the most important factor when it comes to determining the value of a house or home. The formulation of a model in this regard is attempted through ‘regression’. For concrete assessment of this claim, some elaborated factors are included in this exercise. A home or a house can generally be viewed in terms of the properties it holds internally and the characteristics of the environment outside that home (collectively referred to as ‘location’). The null hypothesis here would be the realtor’s claim i.e. the location is the most important factor in assessing the house value whereas the alternate hypothesis refutes by stating that this is not the only factor but there are other significantly effective factors that are needed to be taken into account. LITERATURE REVIEW There are a number of processes that can be used to predict house values. These include various probabilistic methods. Multivariate Spatial Method, Time Series analysis, FootyForecast Forecasting methods( a method that is similar to simple sequence method), ordinary least square regression and logistic regression, ‘hedonic model and artificial neural network model’, ‘non-parametric latent manifold model’. The procedure that we would incorporate in order to evaluate the house values from our system is regression analysis. Regression analysis incorporates the effects that a number of independent variables have on a dependent variable. In the current study it is attempted to evaluate that what do independent variables such as the number of half and full baths present in a house, the vicinity of the house and other factors etc have of the value of a house. Home characteristics The properties related to the home structure, architecture, design and construction quality and the provisions it extends are considered to come under home characteristics. The main characteristics included in this study are the number of rooms, number of bath rooms, provision of central air conditioning etc. Neighborhood characteristics The outside environment of a home generally referred to as location comprises of the neighborhood characteristics. The study includes few like crime effected vicinity, High or traffic volume around etc. DATA AND METHODOLOGY Variables and Model Construction The model for home value estimation employed below analyzes and test the role of different home characteristics and neighborhood characteristics in determining the values of respective homes/houses. The dependent variable here is value. This variable contains the values of houses measured in $1000. The regression model is adapted to run the estimation task of these values. The variables are categorized on the basis of their nature and their relationship to the home internals or location outside home. The respective categories are variables related to home characteristics and variables related to the neighborhood characteristics. Table 1: Sample Statistics for House value regression model variables Dependent Variable N (number of houses) Mean House Value St Dev Value 1654 186921 81999 Home characteristics Baths(Full) (+) 0 1 115000 - 1 1157 162006 62209 2 421 230538 79813 3 61 321536 77906 4 11 367298 52950 5 1 400040 - 6 2 250000 212132 Halfb(Half) (+) 0 974 164253 74984 1 631 217862 80586 2 46 237677 79403 3 3 2600013 121678 Rooms (+) 2 4 107750 58254 3 34 121882 64590 4 141 133107 61599 5 309 150972 66346 6 354 166354 65406 7 368 192259 68043 8 225 228801 75696 9 123 249214 83335 10 48 260733 81830 11 24 298507 108617 12 15 327352 76501 13 6 285007 95246 14 3 366693 57758 Condod (This unit is Condominium?) (-) 0 1498 191573 82233 1 156 142244 64629 Cellard (has a full cellar?) (+) 0 631 161793 81234 1 1023 202420 78594 Garaged (Has a garage?) (+) 0 718 157700 70021 1 936 209336 83479 Porchd (Has a porch?) (+) 0 314 170116 77666 1 1340 190859 82515 Newkitd (New kitchen?) (+) 0 1469 185715 81807 1 185 196494 83110 Septic (Septic System) (+) 0 1129 179079 81088 1 525 203784 81479 Airsysd (Has central a/c ?) (+) 0 1455 182795 79252 1 199 217088 94716 Neighborhood characteristics Metro (-) Suburb 0 1519 189696 80553 Central City 1 135 155698 91466 Crimed (Crime bothersome) (-) 0 (No) 1595 189034 81751 1 (Yes) 59 129797 67180 Noised (Noise bothersome) (-) 0 (No) 1514 188742 82408 1 (Yes) 140 167223 74926 Trafficd (Traffic bothersome) (-) 0 1416 188380 82333 1 238 178237 79597 Inadeqd (Inadequate housing?) (-) 0 1618 187043 81717 1 36 181448 94957 Eroadd (Road in need of repair?) (-) 0 1234 190424 82134 1 420 176628 80823 Ejunkd (Junk in neighborhood?) (-) 0 1437 191829 82444 1 217 154415 71041 (+) indicates positive correlation whereas (-) indicates negative correlation with the dependent variable. For all binary variables, 0 means No and 1 means Yes. The variables that are related to home characteristics are as follows, The variable baths indicates the number of full bathrooms in a house, similarly the halfb is there to indicate the number of half bathrooms in a house. The rooms variable as the name indicates represents the number of rooms. These three variables are continuous in contrast to all other variables which are binary in nature. The other variables that represent the home characteristics include, condod stands for a question that is answered in yes or no, i.e. Is this unit a condominium? Similarly cellard means whether the house has a full cellar? Variable garaged raises a question regarding the availability of garage in the house. Porchd, newkitd, septic and airsysd explore the provisions of balcony, remolded kitchen, septic system and central air conditioning respectively. All of these variables are binary as they are answered as yes or no. The total number of factors that are selected to represent home characteristics amounted to ten this way. The contribution of these factors in determining the value of the house is tested through the respective p-values. The variables that are related to neighborhood characteristics are all binary in nature. The description is as follows, The variable inadeqd represents inadequate housing, eroadd indicates the road in need of repair, ejunkd explores whether there is a Trash/Litter/Junk in neighborhood? The variable metro indicates whether the house is located in the central city or at the suburbs? Variables crimed, noised and traffic indicate the problems that are caused due to criminal activities in the area, noise like industry or other noises and the rate of traffic passing by near to that house respectively. The contributions of these factors in predicting the house value is also tested through the respective p-values. The main aspect of location is highlighted through these variables and the hypotheses presented above is very much related to the significance of these factors. Table 1 clearly shows that the differences of means of houses values grouped through contributing home variables and neighborhood variables are significant. For example the mean value of houses having garage is fairly higher than the mean value of houses with no garage. Whereas in the cases of rooms and bathrooms an increasing trend in the mean value of houses is evident with a disturbance at end due to the suspected outliers. The value of house is regressed for an individual house on the contributing variables from Table 1. These variables are categorized in two types that are home related and location or neighborhood related, therefore specifically, a set of parameter coefficients for each home variable i.e. for full bathrooms (βfb), for half bathrooms (βhb), for rooms (βr), cellar (βcel), garage (βg) septic (βs) condominium (βc), neighborhood variables ejunkd (βej) and crimed (βcr) is estimated. The model includes an intercept (α) and stochastic error term (ε). HWijn = α+ βfbXij + βhbXij + βrXij + βcelXij + βgXij + βsXij + βcXij + βejXij + βcrXij + εijn. The main contributing variables in the model are eight out of seventeen. These predictors include bathrooms, half bathrooms, rooms, cellar, garage, septic, ejunkd and crimed. The increase in number of bath rooms, half bathrooms and rooms shows a proportional increase in the mean house values respectively (outliers are considered as exceptions for this observation). The presence of cellar, garage and provision of septic system accordingly exhibit a fair rise in the value of the house. The area free of crime and litter or junk happens to increase the mean value of the house as compared to the mean values of affected area houses. The condominiums similarly are of low mean values than regular houses. REGRESSION RESULTS Table 2 provides the coefficient estimates and robust standard errors for our house value model. The adjusted R2 is around 42% which seems to be sufficient for a multivariate regression model. The standard errors for all the contributing factors and the respective p-values are also shown in Table 2. The p < 0.05 in all the cases, except the case of condominium yet it is considered due to a very minute difference. Annexure 1 contains the regression models for house values that are formulated considering the home characteristics and neighborhood characteristics individually. The process assisted the identification of the contributing factors also. The p-value in this regard is focused as the main criteria to discriminate both the factors. Table 2: The regression model for house value estimates with contributing home and neighborhood characteristics, adjusted R2, respective standard errors and p-values. House Value Predictors Parameter Estimate (Robust SE) Home characteristics baths 47491.17443 * (3064.19388) halfb 30764.07418 * (3030.070854) rooms 8527.93162 * (1150.75003) cellard 11383.66199 * (3378.920393) garaged 15914.34095 * (3335.279378) septic 6851.8332 * (3370.463478) condod -11144.81745 ** (5912.519284) Neighborhood characteristics ejunkd -19400.67626 * (4609.504325) crimed -23883.49575 * (8422.867105) intercept 39084.32855 * (6697.431044) Adjusted R Square 0.421788854 Observations 1654 Note: Levels of statistical significance are indicated as p < 0.05 (*), p Read More