Reminders

The final project deadline is Wednesday at 5:30 PM (paper and electronic submission).

Final blog post were due yesterday...

Please return all borrowed books on Wednesday (especially Dr. Hayunga's books!).

Big Blue is buying SPSS Inc. this morning...

Tuesday, July 28, 2009: IBM Corp. (NYSE:IBM) announced Tuesday morning that it is buying predictive analytics software maker SPSS Inc. (NASDAQ:SPSS) for $50 a share in cash, or about $1.2 billion.

Pools add value to homes

The goal of this study is to build a statistical model to capitalize swimming pools into housing prices and find the relationship between swimming pools and the values of residential properties. The data for this study comes from house transactions of Arlington area-83-1 near UTA.

All of the past studies found in the Literature Review section of this paper have found positive correlations with swimming pools. Pools do tend to add value to a home — about 7.7%, according to National Association of Realtors statistics. Methodologies and results from these past studies supported that the effect of swimming pools has a significant importance on property values. Therefore, the hypothesis of this paper is pools add value to homes.

The regression model has R square equal to 0.734 which means that 73.4% of the data is explained by this model. This number is confirmed by the Significance F change and Durbin Watson which is equal to 1.963. (very closed to 2) The model select seven independent variables. In the T statistic analysis, acres, Square ft, ages, and pools have significant level under 0.05. In 95% confidence Interval for B, the value of pools is between 2897.636 and 25447.888. In addition, In the ANOVA analysis, the F-statistic is equal to 131.347 with significant level. When checking multicollinearity, the VIF falls into reasonable range and collinearity diagnostics points out there is no multicollinearity problem. Generally, the Histogram graph looks fine. However, the scatterplot graph has homoskedasticity problem. Therefore, the natural log regression may help correct the flaw of the model. This approach considers a linear relationship by log-transformed dependent variable.

The purpose of this paper is to find the correlation among variables. The author of the paper followed the 5 steps of regression analysis beginning with looking at the data and ending with analyzing. If everything is followed correctly the ending analysis should produce valuable confidence and prediction intervals that can be used in structural modeling. Upon using the SPSS, the outcome of the analysis has shown that there is a positive correlation between the pools and sales prices.

Conserving History

The focus of my analysis was to determine if houses located within a conservation or historic district, in Oak Cliff, command a higher selling price than homes outside of a designated area. To complete this analysis 459 sales were analyzed with a hedonic model over a one year time period. Using the reported sales price based on MLS data, the hedonic model was able to explain approximately 70% of the final selling price. However, there was potentially an issue with heteroskedasticty based on the residual plot. Seeking to mitigate this issue, I used the natural log of the sales price, effectively converting the quantities to percentages. This increased the explanatory power of the model by 6%. Using the natural log method, houses within a conservation or historic district command roughly a 40% premium to similar houses outside of an officially designated area.

New Urbanism: Does it add value?

Hypothesis for this paper is that new urbanist features has a positive impact on property values. A detailed regression model is discussed in this research paper.

Summary

Some of the characteristics of new urbanist communities are small lots containing houses with front porches set close to the street; neighborhoods with plenty of public space; controlled landscape; and pedestrian friendly streets. It stands out from competing conventional housing stock. New urbanist communities have gained popularity either due to the changing demographics or change in taste. Despite this growing popularity, not much research is done to find out the market acceptance of such communities.
For the purpose of this research, a local housing community with the above mentioned characters is identified. Hedonics analysis method is used with sales price as the dependent variable and “new urban” (proxy for new urbanist characters) as one of the independent variables along with other independent control variables.

Results

The stepwise-stepwise regression model has adjusted R square value increased with addition of every variable and regression model explained 88% variation in sale price (adjusted R square = 0.880). The model selects other six independent variables which are also statistically significant. The coefficients for these variables show strong relationship to the sales price as expected at a significance level (min. 0.000**, and max 0.003**), indicating very less chances of accepting null hypothesis. The variable of interest ‘new urban’ show strong positive relation to the sales price ($47,000) with t static (8.7) at significance level (0.000**).

The residual plots for standardized residual to adjusted predicted value do not show any violation of linearity and statistical independence assumption. Little flared pattern is observed mainly due to some higher sales prices in newer homes. It can be attributed to the homes built when the real estate markets were unrealistically bullish and certain custom built homes. Since number of such transactions is low, adjusted weight method technique was not applied. Also the outlier identified show relatively low standard of error i.e. less than four.

To make sure the price differential is not due the age of these communities (newly built) and desired school district, three dummy variables to describe age and one for school district are manually entered. Of those, the model only showed one of the variables i.e. homes less than eight years in age, has statistically positive relation to the sales price ($18821) with t statistics of (2.82) at significance level (0.005**).

Conclusion

The results confirmed the rejection of the null hypothesis and that people were willing to pay premium for the homes in new urbanist communities, which is in accordance with the literature reviewed. But the coefficient suggests a higher premium percentage than what was mentioned in the literature. Some of it can be attributed to the age of the homes as they are newly built. In addition, since all the new urbanist homes are from one subdivision, its location characteristics may also explain the differential in the price.

Best Time to Sell a Real Property - Results

The purpose of this study is to examine if the sale price of a real property may be influenced by which part of the year the sale occurs. Generally, it is perceived that spring and summer are the best times to relocate and purchase a house. Hence there is an expected increase in demand during this part of the year that would possibly increase the price. This study looks at the sales data from Arlington Area 83-1 near UTA for the year 2008 to examine this relationship. The study does not find any significant relationship between the property prices and seasonality.

The study further extends the hypothesis by looking at the interaction between seasonality and the property type and their combined effect on property prices. Single family residences are in high demand during late spring and summer as families tend to relocate in tandem with the school year. The study interestingly finds significant relationship between property prices and the interaction between seasonality and property type. The results indicate that for single family residences selling in spring and summer, the property sells for a premium of $6808 after controlling for other hedonic variables like size, age, etc.

School Distance and Home Prices

The rationale for this study is to establish a negative correlation between property values and the distance from public schools. Many determinants are used when evaluating the price of a home. This study uses stepwise linear regression to establish school distance as a determinant of the value of a piece of property. Some questions that can be addressed through this statistical study are: Does distance from a public school affect the value of a home? How much (in dollar terms) is the value affected? Is there strong enough statistical evidence to indicate a negative correlation between house prices and school distance? These questions can, and are, answered through this statistical study.

The results of the stepwise-stepwise linear regression analysis confirm the hypothesis that home values are negatively correlated with distance from schools. The variables that were entered/removed from the model because they met the level of significance threshold were area, number of bathrooms, four miles, and days on market. The fourth model produced included area, number of bathrooms, four miles, and days on market as the predictors of sales price. The r square for the fourth model was .705 and the adjusted R square was .702. This means that the fourth model produced explains 70.5% of the sales price. The adjusted r square decreased by .03, which is a solid indicator of confidence in the numbers. When looking at the ANOVA table, the F statistic for the fourth model is 202.420 with a p-value of .000. The F statistic is very high and has a very low p-value, thus indicating strong statistical significance. The t-statisitc for the independent variable of four miles in the fourth model is 3.175, which indicates a model that possesses considerable strength and aptness. The coefficients of the fourth model show that a home located three to four miles away from the appropriate high school capitalize a $24,281.52 discount into the price. When looking at the collinearity statistics, a variance inflation factor (VIF) for the fourth model’s variable of four miles is 1.040. This is very low and indicates that there is no multi-collinearity occurring between the independent variables. All four statistically significant independent variables have VIF’s of 2.396 or less. The rule of thumb when determining multi-collinearity between independent variables is that VIF’s less than 10 show that multi-collinearity is not an issue in the model. Any VIF of 10 or more indicates multi-collinearity. Therefore, multi-collinearity is not an issue in the model. The variables that were excluded from the fourth model because they did not meet the statistical level of significance were number of bedrooms, name of highschool, one mile, two mile, and three mile.

Investigating the Relationship between the Quality of Middle Schools and House Values within the Birdville and Keller ISD in Watauga, Texas.

386 residential sales were investigaed in Watauga, Texas between March 1, 2008 and June 5, 2009. Data was analyzed and missing data points were created by associating with nearby addresses. School rating information was not part of the original MLS dataset. These rating categories were obtained via the Texas Education Agency website. The scope of the investigation was narrowed to include only middle schools as this had the most information per rating category. A preliminary model was constructed with property characteristics having a sufficient amount of data (greater than 20). The adjusted R-squared was 84.6% for the preliminary model. Only statistically significant variables were promoted to the final model. This reduced the number of independent variables from 23 to 9.

The final model had an adjusted R-squared value of 84.8% and showed that parents are willing to pay a 2.89% premium for middle schools rated Exemplary. The level of significance was p<0.076 which is in the gray area. Thus, although quality does matter, it is a characteristic of quality which parents consider. More research is necessary to dissect the term general term "quality" into its constituent parts. Such characteristics may be the quality of the facility, the level of parent-teacher interaction, the technologies available to students, the number of qualified teachers, etc.

Does Mineral Rights Affect Propert Prices?

The Arlington 83-I area is used in my analysis. The results show a significant "mineral premium" on sales price. A property transferring the mineral rights to the buyer increases sales price by $17523.317. This is the most persistent factor influencing sale prices.

My model is--

Sale Price = 8909.460 + 59.116 Square Feet -716.588 Age +16435.744

Total Baths + 17523.317 Mineral Rights

There is a 95% confidence that the population mean for mineral rights lies within the interval

[ 8993.477, 26053.158]

Another finding in my study is the bi-directional causality between sale price and days in the market, I accomplish this by a Granger Causality test.

Can Sellers Get a Higher Price For Their Home During the Summer?

An analysis was conducted in order to determine if there was a relationship between the time of year a house goes on the market and the sales price of a home in the east area of Dallas, Texas in 2008. The date of listing is classified into two categories; Summer Break which occurs from April through August, and School Season which is September through the following March. The study found that there was a significant relationship between time of year and sales price. Since there is a higher volume of sales during the summer time, there is a greater market where buyer preferences are able to be met and regardless of property type there was an effect. When entering an interaction variable between season and property type, the study found that single family homes sold during the months of April through August added a premium of a little over $10,000 to the price of the home. This could account for families that are more apt to move during the summer months while kids are between school.

The Effect of Foreclosure on House Prices

Abstract: The purpose of this study is to examine effects of foreclosure on house prices in the area near the Dallas Forth Worth International Airport. This study finds that foreclosure reduces property value significantly. The results indicate 32.44 percent discount on prices for foreclosed properties after controlling for other variables including size, and age of house. This evidence is consistent with earlier empirical findings.

Richardson Heights Residential Subdivision Regression Analysis:
Days on Market vs. Final Sales Price

I. RESULTS
No underlying model assumptions required for this regression analysis were violated. Each of the observations was independent. Scatterplots and correlations coefficients for all multivariate combinations were evaluated and the use of a linear function form of final sales prices was determined to be justified. The final sales price is a normally distributed continuous random variable with randomly distributed residuals.

A complete documentation of the stepwise regression is included in the reseach paper. Model 3 has the highest coefficient of determination, explaining 67.7% of the variation in final selling price. Models 1, and 2 were rejected because they have a significantly lower level of explanatory power, at 66% and 67% respectively.


II. CONCLUSIONS
The additional independent variables are both feasible and sensible to include in the model as the size and number of bedrooms of homes are known by non-statisticians to be sufficiently correlated to its final selling price (see scatterplots below). Statistically speaking, their t-statistics are significant and validate the simpler model results that size alone explains 66 % of the variability in final selling price, and coupled with more than 3 bedrooms adds another 0.01% of variance explanation.

In conclusion, the Richardson research study demonstrates that size, more than 3 bedrooms, and the days on the market does impact the final sales price.

Jerry Burbridge, Barrett Shepherd, Ali Samee

Location (second post -- question)

After taking advice from Dr. Hansz and Ramya I have developed a new model that expands upon the one previously posted. My "Location" paper is focusing on the variable: distance from a franchise restaurant. I split the distance into 4 dummy variables: B(Distance, 1/3, 2/3, 3/3, 4/3). My previously posted model had 3 of 4 distance variables kicked out for different reasons.

Following advice, I added a "Pool" dummy variable. It changed everything. Now 3 of 4 distance variable are inside the model explaining the Sales Price. The problem is that I lost .o1 R^2. The new model has a R^2 value of .777 instead of .787. Both models show strength, aptness, and follow the rules of regression.

I am inclined to choose the new model. Any advice?

The Value of a Cul-de-sac

Cul-de-sacs provide homeowners with several benefits and features. Among them is reduced traffic. This results in increased security. Also, a cul-de-sac promotes a more interactive and close knit environment. This analysis will investigate whether there is a positive correlation between being located in a cul-de-sac and home values. The data collected for this study is from 83-1 Arlington, Texas. From that data, the variables will then be compared to see if they have an effect on home prices. And then finally, houses located on cul-de-sacs will be compared to all other houses to determine if, in fact, there is a premium for the location.

With the data used, the size, age, number of bedrooms, number of bathrooms, lot size and the location in relation to a cul-de-sac were all analyzed. The variable of site location was then broken down into three variables to find significance in the model. Dummy variables were used to analyze the variables of homes located on a cul-de-sac, homes located on a dead end street, and homes located within a grid type street. The model removed the variable of cul-de-sac location due to the lack of significance in the model.

In conclusion, the data determined that there is a price premium for homes located within a dead end street compared to homes located within grid type streets. The results of the regression model explain that cul-de-sacs do not have a big significance on home sale prices. Each area of study can have different results as cul-de-sacs can be more valuable in other areas than Arlington. After conducting this analysis, it is evident that homes on a cul-de-sac do not have any extra value however; homes located on a dead end street do have a small premium over other homes in the Arlington area.

Summary Post for Final Project

The Effect of Parks on Residential Home Values
The proximity of parks, green spaces, and trails can have an impact on residential home values. The analysis that is conducted explores the correlation between the proximity to parks and the sales prices. The data comprises of sales information from area 83-1 in Arlington, Texas. The sales information along with the distances to parks has been studied and evaluated. The main focus is to determine exactly how home values are affected by amenities such as parks, green spaces, and trails.
The descriptive variables of the analysis include property size, number of bedrooms, number of bathrooms, lot size, age, swimming pools, days on the market, and the distance to the nearest park. The variable of interest in this analysis is the distance of the parks and green spaces from the residential properties. 13% of the properties are located within 750 feet of a park; 28% of the properties are located 750 to 1,500 feet from a park, 35% of the properties are located 1,500 to 2,500 feet from a park, and 24% of the properties are located over 2,500 feet from a park.
After evaluating the park distances and the other variables of the properties, there is a strong correlation in park distances to sales prices. The regression analysis explains that there is a premium of approximately $4,763 for residences located within 750 feet of a park. The analysis also states that there is a premium of approximately $1,077 for residences located within 750 to 1,500 feet of a park. For homes located over 2,500 feet from a park, the analysis states there is a discount of $1,304. The model did not analyze the variable of homes located 1,500 to 2,500 feet from parks, due to the lack of significance.

Location (Summary Post)

The paper titled Location builds the case that distance from franchise restraunts can be accounted for in the sales price. Franchise restaurants are representative of major roads, commercial districts, noise, and light pollution. The hypothesis is that there is a "sweet spot" where postive externalities are maximized, and negative exernalities minimized. The distance is between 1 mile and 4/3 miles from franchise restaurants. It is believed that this distance allows for an adaquate buffer from commercial use while still allowing for convenient access.

H0: β(Distance>1) = 0 ; H1: β(Distance>1) ≠ 0

The model that was built to control for the age, size of the home, and lot is as follows:

YiF= β0 + β1(square feet) +β2(acres) + β3(age) + β4(Distance 4/3) +β5(Distance_between 1/3 and 2/3) + ε

In this model, Distance was turned into a dummy variable dividing 1.3 miles into four portions. Other portions were tested with no success. By splitting Distance into 1/3 portions, the explanatory value of the variable was maximized.

Several variables were thrown out of the model due to statistical irrelevance or issues with correlation. These include the dummy variables for beds and baths, school district, and Distance 1/3, and Distance 3/3. By removing these variable the model was stronger and more apt.

The F-Statistic measures the model's overall strength. The result is: 243.907
The R^2 Statistic is a measurement of confidence. The result is: .787
The Adjusted R^2 controls for the loss of df. The result is: .784
The t-statistics are listed in the order of the model: 2.317; 26.623; 2.951;
-5.453, 2.044; -1.653

After looking at the residuals (Histogram, Normal P-Plot of Standardized Residuals, and Scatter plot) they appear to be normal, linear, homoskedastic, and the kurtosis looked great. Multi-collinearity did not appear to be an issue either.

The end result with regard to the Distance 3/3 variable's B-coefficient is:

P(9,107.512 < B < 26,555.36) =.95


In conclusion, the null hypothesis is rejected.

Heteroskedasticity

Useful Links
White's Test
http://pages.infinit.net/rlevesqu/Syntax/RegressionRepeatedMeasure/WhiteTestStatisticsAndSignificance.txt
http://www.geocities.com/rlevesque/SPSS/HeteroscedasticityTestingAndCorrectingInSPSS1.pdf

Weighted least Squares
http://faculty.chass.ncsu.edu/garson/PA765/wls.htm

Article on New Urbanism and Property Values

I came across the following article on new urbanism and property values.

Song,Yan andGerrit-Jan Knaap, BNewUrbanism and HousingValues:ADisaggregate
Assessment,’’ Journal of Urban Economics, 2003, 54:2, 218–238.

I got reminded about Yogesh Patil's final project. Thought this might help him in his literature review.

Informal UTA-Fort Worth Reception Details...

I had no idea this was going on until Dr. Hansz mentioned it in class. Thanks Dr. Hansz. I'm placing the details here just in case someone else "didn't get the memo". I'm going and I hope to see some familiar faces there...

**********************
MSRE Students:

Just in case you did not receive the e-mail from another source, you are invited to an informal reception for the UTA Real Estate Program in Downtown Fort Worth - Saturday, July 11 from 6:30 pm - 7:30 pm in the "Gallery 76102" at the UTA Fort Worth Center - 1401 Jones Street, Fort Worth, Texas 76102. This informal gathering is open to former, current, and prospective UTA Real Estate Students, Faculty, and Friends. The purpose of this gathering is to promote networking and socializing opportunities among the attendees as well as giving everyone the opportunity to see the new home for the Graduate Programs in Real Estate at the UTA Fort Worth Center in Downtown Fort Worth. At the conclusion of the reception at 7:30 pm, there will be an opportunity for those interested to move to another Downtown Fort Worth venue for dinner, drinks, and further interaction with other attendees. In order to adequately order appetizers and drinks, please send your RSVP to Fred Forgey at forgey@uta.edu before Saturday July 11th. You are welcome to bring your friends, spouses, significant others, etc.
***********************

The Effect of Noise on House Prices

Abstract: The purpose of this study is to examine the effects of noise level on house prices of the area near the Atlanta Airport. We find that noise level and house prices are negatively related. The results indicate that house prices discount $10,505.45 for noise level at 40 after controlling for size of house, age of the house, and number of months old from the first house sale in sample. This evidence is consistent with earlier empirical findings.

SPSS 17.0 Gradpack update

FYI...There is a free update for those who are using SPSS Gradpack version 17.0. The update upgrades version 17.0 to version 17.0.2 and fixes some internal bugs as well as import/export problems, table display problems, among others. On the SPSS toolbar, go to Help - Check for Updates. Then follow the directions.


Eduardo (Ed) Torres-Valdez