SlideShare a Scribd company logo
1 of 25
Elizabeth Thurman ECN 405 2L
1
Model 1 Paper
Cross Sectional
Introduction:
The cost of housing is an important factor when choosing where to live and also
in considering the economic well-being of a state. High housing costs can be an
indication of higher paying jobs and higher prices in general in a particular area. When
housing prices are increased, it indicates an increased demand for housing in that area.
Often, when prices are high, one must achieve a certain level of education in order to
obtain a high paying job in a high-priced area.
The question I would like to analyze is how personal income affects housing
prices in a state, and also how level of education ties in with level of income to affect
said housing prices. Therefore, in this paper, I will be using personal income and level
of education as my testing variables in a regression analysis to see if the amount of
money a consumer makes, and their level of education will have an effect on the
dependent variable, housing prices.
Also, I will be exploring the impact of income and education on home mortgages
throughout this paper by using five related studies in the form of journal articles. I will
be exploring how each article relates and shapes my model and question. I will then
pose my economic model that I will use to answer my question along with its estimation
and corresponding graphs.
Elizabeth Thurman ECN 405 2L
2
Further, I will test to see if I have evidence of heteroscedasticity which could
imply that I have a violation of constant variance in my error terms. I will then test for
endogeneity which will allow me to assess whether there is a correlation between an
independent variable and an error term. If the test results show that I do have an
endogenous model, my regression coefficient would cause the model to be biased, and
therefore violating the OLS rules for being the best estimator.
After testing for homoscedasticity and endogeneity, I will do a Ramsey reset test
for zero mean to see if my model is specified correctly, test for normality, and preform a
Wald test. Then, I will perform a final weighted estimation on my variables and then test
a binary model for states west versus east of the Mississippi River.
Review of the literature:
The article by Stephen Malpezzi, "Housing prices, externalities, and regulation in
US metropolitan areas" analyzes the determinants of housing prices which vary widely
across the United States. The study focuses mainly on city and metropolitan areas. It
uses a simple supply and demand framework to assess how regulatory actions effect
housing prices and uses factors such as income and population changes.
This article has helped me to shape my model because of the use of population
and income in relation to housing prices. I have included both as variables in my model
to study their effect. In Malpezzi’s article, I like that he related the housing prices to
metropolitan areas and chose to do my binary testing based on states east and west of
the Mississippi. This is because although both coasts tend to have a relatively large
number of metropolitan areas compared to the “inner states”, the east coast states
Elizabeth Thurman ECN 405 2L
3
seem to have more populated areas throughout. Highly populated areas tend to have
high demand for housing, thus increasing the prices of homes (or other housing
amenities). This is why I especially wanted to include population; however, I used
population for an entire state which will not give me the more specific relation to housing
prices that I would have liked.
The article "Education and income" by Hendrik S. Houthakker, analyzes income
levels in relation to a person’s level of education. It is performed as a cross sectional
analysis of different age groups across a single year. The article concludes that “capital
values increase uniformly as education increases”. However, it finds that those with
college levels 1-3 do not fare as well as those with only a high school diploma.
This concept led me to use the testing variable ‘level of education’ because those
who afford higher priced homes may be correlated with higher paying jobs which would
most likely require more than a high school diploma. Because the study found a
positive correlation between level of education and level of income, I decided it would
be a reasonable variable to use along with the amount of personal income.
In the article "Valuation of education and crime neighborhood characteristics
through hedonic housing prices" by Robin A. Dubin and Allen C. Goodman, housing
prices are studied as a bundle of neighborhood characteristics. Among the
characteristics in the bundle are those of crime and neighborhood schools. The study
found that housing values were influenced in the Baltimore metropolitan area
significantly when the two variables were studied together.
Elizabeth Thurman ECN 405 2L
4
This article inspired me to use state crime rates in 2010 to see how it affects the
cost of housing. Because one would intuitively guess that housing prices are cheaper
where there is more crime, I would like to test for a correlation. Many homeowners want
to live in an area where there is low crime not only for the safety of their family and safe
schools, but also for the safety of their belongings. Areas with high breaking and
entering rates may also drive the nearby cost of housing down. However, sometimes it
costs more to live in a city where the crime rates tend to be higher. This is another
reason I would also like to add crime rates as a variable. However, I will be using crime
as the amount of burglaries and will use it as a state data and not specifying it to
suburban, city, or rural housing, although this would be an interesting study.
The article "Housing and income" by Alan R. Winger finds that income and
housing do have a relationship and should involve other factors such as the permanent
income of an individual. In the past, studies have concluded that there is a relationship;
however, the specific factors involved where not clear. The article notes that you must
take into account that the decision to purchase a home was made in the past relative to
when each study is performed. Therefore it is harder to conclude what factors are
considered when a consumer decides to buy their home. The article mentions a paper
by Margaret Reid who studied the effects and also found that there is a strong
correlation between income and the housing market.
Because a strong correlation was found, and the two variables of housing related
to income are widely studied, I was inspired to also test for a correlation between the
two. If I were to analyze a second model, I would include a test for income over a
certain time period to see if it fluctuates with housing prices because income can also
Elizabeth Thurman ECN 405 2L
5
fluctuate. Thus, what a consumer may be able to afford in one year, may not be true in
the next.
Similarly, in the article, "Housing and permanent income: Tests based on a
three-year re-interview survey” by Lee, Tong Hun, the author points out that as
consumption is tested against income, income is usually tested at a point in time.
Instead, he notes, particularly with the consumption and purchase of mortgage loans
(housing), it is better to test the relationship with a fixed income over time. This is
because purchasing power may fluctuate and although a consumer may be making a lot
in a particular year, they may not make as much in following years and therefore not
affect the mortgage market as greatly.
This article gave some great insight into the notion that it must be kept in mind
we are testing income and housing at one point in time. Although results were found,
indication that income over time is a more accurate measure when related to housing
consumption, I decided to keep my variable of income because I am interested to see if
there is any correlation with housing prices combined with other factors such as the
population’s level of education, level of crime, GDP per capita, etc.
The Model:
Y(House Prices) = B0 + B1(Personal Income) + B2(Level of Education) + B3(State
Taxes) + B4(Unemployment Rate) + B5(GDP) + B6(Population) + B7(Burglary) + U
This is a cross sectional model measuring 2010 data on average housing costs
in relation to income and level of education in each of the 50 states. (Data used can be
found in Appendix A)
Elizabeth Thurman ECN 405 2L
6
Assumptions about disturbance term:
1. Linearity – The expected value of Y is linearly associated given the X’s. Also, the
average value of the error term is equal to zero given X.
2. Unbiased – Error terms are independent of one another.
3. Homoscedasticity - Constant error variance. This means that the variance of the
errors is the same regardless of X. V(ε|xi)=σε2. The degree of random noise is
the same regardless of the value of the X’s.
4. Mean Independence – The error term is independent of the X variables.
5. Normality – The error terms are normally distributed.
Variables:
House Price Index: This is the dependent variable (Y).
Personal Income per capita: This is a testing independent variable. This should have a
positive effect on housing prices because the greater purchasing power consumers
have, the more housing they will buy, thus increasing demand and therefore pricing of
houses.
Level of Education: This is another testing independent variable. It should have a
positive effect on housing prices. Those who obtain a higher level of education tend to
earn higher income, thus have higher purchasing power driving up demand and prices
of housing.
Total State Taxes: Independent variable. This should have a positive impact on
housing prices. As taxes increase, so do property taxes.
Elizabeth Thurman ECN 405 2L
7
Unemployment Rate: This is another independent variable. If the unemployment rate is
low, the housing rate may be higher because more people are able to earn a living, thus
driving demand for housing.
GDP per capita: Independent variable. This should have a positive effect on housing
prices. Economically healthy states who have higher GDP may have higher costs of
living due to its many consumers earning a living.
Population: This is an independent variable. The population in a state may cause
higher costs for housing because with more people comes a higher demand for
housing. When supply is limited, and housing demand increases, so must the price.
State Total Burglary: Independent variable. This may have a negative impact on home
prices because housing tends to be cheaper if there is a higher rate of crime in the area.
The demand for housing in these types of areas is lower than safer areas.
I will use this model to answer my question by testing the independent variables,
Personal Income and Level of Education against Housing Prices. I will test to see how
correlated the variables are and whether they are significant.
I will then do a test on a binary variable for which I chose to use states west and
states east of the Mississippi river. I chose this criteria for my binary because I wanted
to compare if there was a difference between the west coast and east coast. The
United States tends to be split with higher populations toward each coast, having lots of
land with lower populations throughout its middle section. This is why I chose to
compare the large populations of California and Texas (these are the states with the
Elizabeth Thurman ECN 405 2L
8
higher population data on the west coast) to the states Florida, New York, Ohio,
Pennsylvania and Illinois (higher population data on the east coast).
Initial estimation and hypothesis testing:
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:16:48
Sample:1 50
Included observations:50
Variable Coefficient Std. Error t-Statistic Prob.
C -115.4761 88.06088 -1.311322 0.1967
INCOME 0.003878 0.003856 1.005642 0.3202
EDUCATION 10.95013 2.965625 3.692351 0.0006
TAXES 2.28E-06 8.96E-07 2.543248 0.0147
BURGLARY -0.000868 0.000305 -2.840839 0.0068
CAPITAGDP -0.001513 0.001888 -0.801578 0.4272
UNEMPLOYMENT 7.641442 4.836332 1.580008 0.1214
R-squared 0.654494 Mean dependentvar 329.1984
Adjusted R-squared 0.606284 S.D. dependentvar 94.81226
S.E. of regression 59.49163 Akaike info criterion 11.13873
Sum squared resid 152187.9 Schwarz criterion 11.40641
Log likelihood -271.4681 Hannan-Quinn criter. 11.24066
F-statistic 13.57588 Durbin-Watson stat 1.928779
Prob(F-statistic) 0.000000
Elizabeth Thurman ECN 405 2L
9
When testing the dependent variable (housing prices) against all other variables,
initially I find that burglary has a negative correlation with housing prices. This was as I
expected. All variables seem to have a correlation with my dependent variable;
however, GDP per capita seems to have the least correlation when compared with other
variables. Also, this variable has a negative beta, which I expected the correlation to be
positive because the income of a state increases as taxes increase which would mean
the housing costs more.
The following are correlation graphs. This gives a visual display the relationship
between each variable against the dependent variable.
100
200
300
400
500
600
700
30,000 35,000 40,000 45,000 50,000 55,000 60,000
Personal Income per capita ($)
HousePriceIndex(inhundreds)
Personal income appears to have a generally positive correlation, as expected.
Elizabeth Thurman ECN 405 2L
10
100
200
300
400
500
600
700
0 20,000,000 60,000,000 100,000,000
Total State Taxes (in thousands)
HousePriceIndex(inhundreds)
It appears as though total state taxes remain around one area with varying house
prices. This makes sense since state income may be in the same general area, the
prices of its houses vary much more greatly.
100
200
300
400
500
600
700
16 20 24 28 32 36 40
level of Education
HousePriceIndex(inhundreds)
Level of education appears to have a positive correlation with housing prices, as
expected.
Elizabeth Thurman ECN 405 2L
11
100
200
300
400
500
600
700
0 40,000 80,000 120,000 200,000
State Total Burglary
HousePriceIndex(inhundreds)
State burglary appears to vary. As burglary levels remain around a certain number,
house prices seem to vary. However, it appears as burglary rates increase, housing
prices don’t stay at high levels.
100
200
300
400
500
600
700
30,000 40,000 50,000 60,000 70,000
GDP per capita
HousePriceIndex(inhundreds)
GDP per capita appears to have a roughly positive correlation, however not very
strongly. I expected a stronger correlation.
Elizabeth Thurman ECN 405 2L
12
100
200
300
400
500
600
700
2 4 6 8 10 12 14
Unemployment Rate
HousePriceIndex(inhundreds)
Unemployment rate does not appear to have a very strong correlation; however
unemployment tends to stay around a certain level while house prices vary.
100
200
300
400
500
600
700
0 10,000,000 20,000,000 30,000,000 40,000,000
POPULATION
HousePriceIndex(inhundreds)
The population variable does not appear to have a high correlation with the housing
price variable.
Elizabeth Thurman ECN 405 2L
13
Conclusion/Test Results:
In conclusion, I have found that housing prices are more correlated with my
testing variable income, than with the testing variable level of education. In my final
estimation I found that my variables are significant, and that burglary is negatively
correlated as I expected.
Due to the high F-stat in my Breusch-Pagan-Godfrey test for heteroscedasticity
(Appendix B), I find that I reject the null concluding that my model has implication of
heteroscedasticity. Therefore, my error terms may be correlated to some degree and
also have varying distributions and variances.
In my test for endogeniety found in (Appendix B) I found that because my t-stat of
-0.586935 was less than my t-crit of 2.02, I fail to reject the null, concluding that I do not
have endogeneity. Therefore, my variables are not correlated with the error term.
In the Ramsey reset test for zero mean I find that my model is specified correctly.
The normality test shows that my model is not a normal distribution. The Wald test
shows that my testing variables are not jointly significant at a value of 0. In the final
weighted estimation I found that most all of my variables tend to be significantly
correlated with the exception of GDP per capita, as it was in my initial estimation.
Therefore, I would conclude that while I would want to include more data
variables, housing prices are related to the level of one’s income. There is also a higher
correlation between housing prices and burglary rates. It has a negative correlation,
meaning as burglary rates are higher, housing prices tend to be lower. If I created a
Elizabeth Thurman ECN 405 2L
14
new model, or added onto this model, I would do more testing for crime rates and use it
as a testing variable rather than a control variable.
References:
Article 1
Dubin, Robin A., and Allen C. Goodman. "Valuation of education and crime
neighborhood characteristics through hedonic housing prices." Population and
environment 5.3 (1982): 166-181.
Article 2
Houthakker, Hendrik S. "Education and income." The Review of Economics and
Statistics (1959): 24-28.
Article 3
Lee, Tong Hun. "Housing and permanent income: Tests based on a three-year
reinterview survey." The Review of Economics and Statistics (1968): 480-490.
Article 4
Malpezzi, Stephen. "Housing prices, externalities, and regulation in US metropolitan
areas." Journal of Housing Research 7 (1996): 209-242.
Article 5
Winger, Alan R. "Housing and income." Economic Inquiry 6.3 (1968): 226-252.
Elizabeth Thurman ECN 405 2L
15
Appendix A:
State
House
Price
Index (in
hundreds
)
Personal
Income
per
capita ($)
level of
Educatio
n
Total State
Taxes (in
thousands)
Unemplo
yment
Rate
GDP per
capita
Populatio
n
State
Total
Burglary
Binary:
West=1,
East=0
Alabama 291.11 33,894 21.9 8,419,911 9.3 36156 4,785,570 42,034 0
Alaska 280.23 45,565 27.9 4,522,927 7.9 68656 713,868 3,105 1
Arizona 263.87 33,993 25.9 10,719,958 10.4 38222 6,408,790 50,771 1
Arkansas 245.27 32,017 19.5 7,559,898 7.9 37658 2,922,280 32,511 1
California 411.31 42,282 30.1 107,195,465 12.3 51546 37,333,601 228,857 1
Colorado 342.25 41,689 36.4 8,575,262 9 49923 5,048,196 26,153 1
Connecticut 409.47 55,216 35.5 12,344,106 9.3 64766 3,579,210 15,172 0
Delaware 437.58 40,969 27.8 2,763,032 8 62994 899,711 7,515 0
Florida 296.59 38,478 25.8 30,484,883 11.3 38258 18,846,054 169,119 0
Georgia 284.15 34,341 27.3 14,782,779 10.2 41894 9,713,248 96,723 0
Hawaii 448.63 41,668 29.5 4,837,862 6.8 48694 1,363,731 8,663 1
Idaho 278.66 32,100 24.4 2,951,703 8.7 34825 1,570,718 6,502 1
Illinois 318.26 42,033 30.8 27,795,759 10.4 50296 12,839,695 75,399 0
Indiana 245.34 34,344 22.7 13,795,221 10.1 43207 6,489,965 47,115 0
Iowa 246.02 39,033 24.9 6,809,344 6.3 46052 3,050,314 16,656 1
Kansas 234.66 38,811 29.8 6,492,996 7.1 43556 2,858,910 19,404 1
Kentucky 287.18 32,929 20.5 9,531,404 10.2 37746 4,347,698 30,311 0
Louisiana 241.53 37,199 21.4 8,758,633 7.4 48594 4,545,392 45,435 1
Maine 462.13 37,213 26.8 3,489,953 8.2 38374 1,327,366 7,359 0
Maryland 423.05 50,035 36.1 15,237,748 7.8 54080 5,787,193 36,542 0
Massachusetts 617.39 51,487 39 20,090,563 8.3 60354 6,563,263 37,767 0
Michigan 241.09 35,082 25.2 22,208,870 12.7 39056 9,876,149 73,868 0
Minnesota 312.11 42,572 31.8 17,208,877 7.4 50641 5,310,337 24,415 1
Mississippi 243.14 30,834 19.5 6,268,823 10.5 31331 2,970,047 30,444 0
Missouri 278.25 36,606 25.6 9,707,053 9.3 42610 5,996,063 44,043 1
Montana 353.21 34,612 28.8 2,142,809 6.7 36918 990,527 3,654 1
Nebraska 252.66 39,926 28.6 3,864,897 4.7 49119 1,829,838 8,326 1
Nevada 216.76 36,657 21.7 5,835,963 13.8 44102 2,703,230 22,226 1
N Hampshire 404.59 44,963 32.8 2,271,936 6.1 47224 1,316,614 5,441 0
New Jersey 485.57 50,941 35.4 25,927,891 9.6 56025 8,802,707 38,732 0
New Mexico 299.21 33,175 25 4,295,237 7.9 39316 2,064,982 21,014 1
New York 576.76 49,582 32.5 63,807,610 8.6 60974 19,398,228 64,973 0
N Carolina 314.54 35,435 26.5 21,514,930 10.8 43778 9,559,533 102,690 0
N Dakota 254.68 43,275 27.6 2,645,695 3.8 51254 674,344 1,966 1
Ohio 245.55 36,199 24.6 23,583,596 10 42342 11,545,435 106,521 0
Oklahoma 203.06 35,912 22.9 7,082,161 6.9 39377 3,759,263 37,476 1
Oregon 373.79 35,898 28.8 7,475,135 10.7 49538 3,837,208 19,637 1
Pennsylvania 373.96 41,635 27.1 30,169,122 8.4 45976 12,710,472 55,171 1
Rhode Island 474.92 43,013 30.2 2,568,759 11.7 46277 1,052,669 6,121 0
S Carolina 317.6 32,669 24.5 7,242,724 11.2 35078 4,636,361 46,156 0
S Dakota 290.44 40,613 26.3 1,321,228 5.1 46507 816,211 3,181 1
Tennessee 288.17 35,426 23.1 10,513,788 9.8 39649 6,356,683 64,235 0
Texas 223.8 38,065 25.9 39,516,186 8.2 47617 25,245,178 228,597 1
Utah 320.5 32,447 29.3 5,237,427 8.1 42075 2,774,424 15,017 1
Vermont 436.9 40,134 33.6 2,511,387 6.4 42097 625,793 3,366 0
Virginia 403.22 44,836 34.2 16,411,055 7.1 52084 8,024,417 30,629 0
Washington 414.56 42,547 31.1 16,106,154 9.9 52850 6,742,256 55,164 1
W Virginia 217.41 31,798 17.5 4,803,704 8.4 34818 1,854,146 10,756 0
Wisconsin 303.86 38,728 26.3 14,368,569 8.5 44431 5,689,060 26,566 0
Wyoming 274.93 45,025 24.1 2,158,260 7 66256 564,222 2,149 1
Elizabeth Thurman ECN 405 2L
16
Appendix B:
Test for Heteroscedasticity:
HeteroscedasticityTest:Breusch-Pagan-Godfrey
F-statistic 6.258169 Prob. F(1,42) 0.0163
Obs*R-squared 5.705965 Prob. Chi-Square(1) 0.0169
Scaled explained SS 3.957595 Prob. Chi-Square(1) 0.0467
Test Equation:
DependentVariable:RESID^2
Method: LeastSquares
Date: 12/11/14 Time:14:53
Sample:1 50
Included observations:44
Variable Coefficient Std. Error t-Statistic Prob.
C -6821.033 4419.425 -1.543421 0.1302
CAPITAGDP 0.236573 0.094568 2.501633 0.0163
R-squared 0.129681 Mean dependentvar 4054.004
Adjusted R-squared 0.108959 S.D. dependentvar 5592.571
S.E. of regression 5279.105 Akaike info criterion 20.02529
Sum squared resid 1.17E+09 Schwarz criterion 20.10639
Log likelihood -438.5564 Hannan-Quinn criter. 20.05537
F-statistic 6.258169 Durbin-Watson stat 2.313727
Prob(F-statistic) 0.016345
Elizabeth Thurman ECN 405 2L
17
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:14:55
Sample:1 50
Included observations:44
Weighting series:ABS(VRES)
Weight type: Variance (average scaling)
Variable Coefficient Std. Error t-Statistic Prob.
C -108.2291 52.54895 -2.059586 0.0463
INCOME 0.012422 0.002053 6.050098 0.0000
TAXES 2.13E-06 5.89E-07 3.615020 0.0009
BURGLARY -0.000847 0.000182 -4.642958 0.0000
CAPITAGDP -0.002872 0.001753 -1.638338 0.1096
UNEMPLOYMENT 9.806262 3.717914 2.637571 0.0120
Weighted Statistics
R-squared 0.872892 Mean dependentvar 247.1635
Adjusted R-squared 0.856168 S.D. dependentvar 207.7145
S.E. of regression 27.66106 Akaike info criterion 9.604052
Sum squared resid 29075.11 Schwarz criterion 9.847351
Log likelihood -205.2892 Hannan-Quinn criter. 9.694279
F-statistic 52.19178 Durbin-Watson stat 1.870739
Prob(F-statistic) 0.000000 Weighted mean dep. 313.1386
Unweighted Statistics
R-squared 0.547538 Mean dependentvar 327.4782
Adjusted R-squared 0.488004 S.D. dependentvar 96.43644
S.E. of regression 69.00406 Sum squared resid 180939.3
Elizabeth Thurman ECN 405 2L
18
Durbin-Watson stat 2.070976
Test for Endogeniety:
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:14:08
Sample:1 50
Included observations:44
Variable Coefficient Std. Error t-Statistic Prob.
C -122.7353 100.7347 -1.218402 0.2310
INCOME -0.008659 0.022321 -0.387937 0.7003
TAXES 2.01E-06 1.15E-06 1.752352 0.0882
EDUCATION 11.58633 4.461864 2.596747 0.0135
Elizabeth Thurman ECN 405 2L
19
CAPITAGDP 0.008608 0.017224 0.499797 0.6203
BURGLARY -0.000785 0.000376 -2.088543 0.0439
UNEMPLOYMENT 9.239402 5.959531 1.550357 0.1298
IVRES -0.010193 0.017367 -0.586935 0.5609
R-squared 0.634308 Mean dependentvar 327.4782
Adjusted R-squared 0.563202 S.D. dependentvar 96.43644
S.E. of regression 63.73555 Akaike info criterion 11.31033
Sum squared resid 146239.9 Schwarz criterion 11.63473
Log likelihood -240.8272 Hannan-Quinn criter. 11.43063
F-statistic 8.920510 Durbin-Watson stat 1.908336
Prob(F-statistic) 0.000002
Test for Zero Mean:
RamseyRESET Test
Equation:UNTITLED
Specification:CPIHOUSING C INCOME TAXES EDUCATION CAPITAGDP
BURGLARY UNEMPLOYMENT
Instrumentspecification:C INCOME TAXES EDUCATION CAPITAGDP
BURGLARY POPULATION
Omitted Variables:Powers of fitted values from 2 to 4
Value df Probability
F-statistic 2.214646 (3, 34) 0.1043
Likelihood ratio 7.853523 3 0.0491
F-test summary:
Sum of Sq. df
Mean
Squares
Elizabeth Thurman ECN 405 2L
20
Test SSR 24134.14 3 8044.714
Restricted SSR 147639.3 37 3990.252
Unrestricted SSR 123505.2 34 3632.506
LR testsummary:
Value df
Restricted LogL -241.0367 37
Unrestricted LogL -237.1100 34
Unrestricted TestEquation:
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:14:20
Sample:1 50
Included observations:44
Variable Coefficient Std. Error t-Statistic Prob.
C 11126.00 7566.153 1.470497 0.1506
INCOME -0.245853 0.165883 -1.482087 0.1475
EDUCATION -585.4722 400.5812 -1.461557 0.1530
TAXES -0.000137 9.33E-05 -1.469711 0.1508
CAPITAGDP 0.081795 0.055551 1.472444 0.1501
BURGLARY 0.051451 0.035064 1.467355 0.1515
UNEMPLOYMENT -441.7197 300.9174 -1.467910 0.1513
FITTED^2 0.275000 0.179630 1.530927 0.1350
FITTED^3 -0.000558 0.000353 -1.579705 0.1234
FITTED^4 4.15E-07 2.54E-07 1.635528 0.1112
R-squared 0.691159 Mean dependentvar 327.4782
Elizabeth Thurman ECN 405 2L
21
Adjusted R-squared 0.609407 S.D. dependentvar 96.43644
S.E. of regression 60.27027 Akaike info criterion 11.23227
Sum squared resid 123505.2 Schwarz criterion 11.63777
Log likelihood -237.1100 Hannan-Quinn criter. 11.38265
F-statistic 8.454350 Durbin-Watson stat 1.869290
Prob(F-statistic) 0.000002
Normality:
0
1
2
3
4
5
6
-40 -30 -20 -10 0 10 20 30 40 50
Series: Standardized Residuals
Sample 1 50
Observations 44
Mean 0.151793
Median -3.511059
Maximum 48.16692
Minimum -43.24634
Std. Dev. 26.00272
Skewness 0.266158
Kurtosis 1.794042
Jarque-Bera 3.185773
Probability 0.203338
Wald Test:
Equation:Untitled
Test Statistic Value df Probability
F-statistic 28.84905 (2, 38) 0.0000
Chi-square 57.69810 2 0.0000
Null Hypothesis:C(2)=C(3)=0
Null Hypothesis Summary:
Elizabeth Thurman ECN 405 2L
22
Normalized Restriction (= 0) Value Std. Err.
C(2) 0.012422 0.002053
C(3) 2.13E-06 5.89E-07
Restrictions are linear in coefficients.
Final Weighted Estimation:
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:14:55
Sample:1 50
Included observations:44
Weighting series:ABS(VRES)
Weight type: Variance (average scaling)
Variable Coefficient Std. Error t-Statistic Prob.
C -108.2291 52.54895 **-2.059586 0.0463
INCOME 0.012422 0.002053 ***6.050098 0.0000
TAXES 2.13E-06 5.89E-07 ***3.615020 0.0009
BURGLARY -0.000847 0.000182 ***-4.642958 0.0000
CAPITAGDP -0.002872 0.001753 *-1.638338 0.1096
UNEMPLOYMENT 9.806262 3.717914 ***2.637571 0.0120
Weighted Statistics
R-squared 0.872892 Mean dependentvar 247.1635
Adjusted R-squared 0.856168 S.D. dependentvar 207.7145
S.E. of regression 27.66106 Akaike info criterion 9.604052
Sum squared resid 29075.11 Schwarz criterion 9.847351
Log likelihood -205.2892 Hannan-Quinn criter. 9.694279
Elizabeth Thurman ECN 405 2L
23
F-statistic 52.19178 Durbin-Watson stat 1.870739
Prob(F-statistic) 0.000000 Weighted mean dep. 313.1386
Unweighted Statistics
R-squared 0.547538 Mean dependentvar 327.4782
Adjusted R-squared 0.488004 S.D. dependentvar 96.43644
S.E. of regression 69.00406 Sum squared resid 180939.3
Durbin-Watson stat 2.070976
*Reject at alpha = 0.10
**Reject at alpha = 0.05
***Reject at alpha = 0.01
Binary Estimation:
DependentVariable:CPIHOUSING
Method: LeastSquares
Date: 12/11/14 Time:16:30
Sample:1 50
Included observations:50
Variable Coefficient Std. Error t-Statistic Prob.
C 59.01909 118.0173 0.500089 0.6198
INCOME 0.008922 0.004040 2.208542 0.0332
INCOME*BINARY -0.001853 0.005813 -0.318785 0.7516
TAXES 1.84E-06 1.47E-06 1.249160 0.2191
TAXES*BINARY 5.24E-07 1.89E-06 0.277112 0.7832
BURGLARY -0.000798 0.000511 -1.562475 0.1263
BURGLARY*BINARY 4.04E-05 0.000684 0.059132 0.9531
Elizabeth Thurman ECN 405 2L
24
CAPITAGDP 0.000345 0.003074 0.112353 0.9111
CAPITAGDP*BINARY -0.002311 0.004385 -0.527089 0.6011
UNEMPLOYMENT -6.536023 8.989954 -0.727036 0.4715
UNEMPLOYMENT*BINAR
Y 13.60267 9.166913 1.483888 0.1459
R-squared 0.633596 Mean dependentvar 329.1984
Adjusted R-squared 0.539646 S.D. dependentvar 94.81226
S.E. of regression 64.32953 Akaike info criterion 11.35745
Sum squared resid 161393.2 Schwarz criterion 11.77810
Log likelihood -272.9363 Hannan-Quinn criter. 11.51764
F-statistic 6.743987 Durbin-Watson stat 1.987158
Prob(F-statistic) 0.000006
Appendix C:
Test for heteroscedasticity:
H0: The model contains homoscedasticity
H1: The model contains heteroscedasticity
Critical value F(1,42): 4.07
F Statistic: 6.258169
Therefore reject null. Conclude the model is heteroscedastic
Test for endogeniety:
H0: The model is not endogenous
H1: The model is endogenous
DF: 36
Tcritical: 0.10, 0.05, 0.01 = 1.302, 1.683, 2.422 respectively.
Elizabeth Thurman ECN 405 2L
25
Tstat: |-0.586935|
Therefore, conclude that the model is not endogenous.
Test for zero mean:
H0: Γ = 0 (model is specified correctly)
H1: Γ ≠ 0 (model is not specified correctly)
Fstatistic = 2.214646 < Tcritical = 2.89
Therefore, fail to reject H0.
Normality:
H0: skewness & kurtosis = 0
H1: otherwise
Tcritical: 1.696 < Jarque-Bera: 3.1858
Therefore, reject H0, conclude my model is not normal.
Wald test:
H0: B1=B2=0
H1: Otherwise
Fcritical: 3.24 < Fstatistic: 28.84905
Therefore reject Ho. My testing variables are not jointly significant at a value of 0.

More Related Content

Similar to Thurman Model 1 Cross Sectional ECN 405

MMSS Senior Thesis 2000
MMSS Senior Thesis 2000MMSS Senior Thesis 2000
MMSS Senior Thesis 2000Eric Morel
 
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final Draft
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final DraftAn Analysis on the Influence of Mortgage Rates on Housing Prices - Final Draft
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final DraftCaleb Goettl
 
Position Essay Examples.pdf
Position Essay Examples.pdfPosition Essay Examples.pdf
Position Essay Examples.pdfJessica Spyrakis
 
Portion Four Housing Affordability: Lecture Note.ppt
Portion Four Housing Affordability: Lecture Note.pptPortion Four Housing Affordability: Lecture Note.ppt
Portion Four Housing Affordability: Lecture Note.pptGoitom Abraha Baraki
 
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...ParanLesterDocot
 
UNIT 1 - WHAT IS ECONOMICS LESSON...pptx
UNIT 1 - WHAT IS ECONOMICS LESSON...pptxUNIT 1 - WHAT IS ECONOMICS LESSON...pptx
UNIT 1 - WHAT IS ECONOMICS LESSON...pptxPreciousChanaiwa
 
In Text Citations Using Apa Endnote Example - Acadial
In Text Citations Using Apa Endnote Example - AcadialIn Text Citations Using Apa Endnote Example - Acadial
In Text Citations Using Apa Endnote Example - AcadialJessica Henderson
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxcurwenmichaela
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxjasoninnes20
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxrichardnorman90310
 
« PreviousHomeNext »Home » Measurement » Levels of Measure.docx
« PreviousHomeNext »Home » Measurement » Levels of Measure.docx« PreviousHomeNext »Home » Measurement » Levels of Measure.docx
« PreviousHomeNext »Home » Measurement » Levels of Measure.docxodiliagilby
 
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...Daniel Katz
 
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docxlorainedeserre
 
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docxnovabroom
 
Topics Essay Writing Competition College Students
Topics Essay Writing Competition College StudentsTopics Essay Writing Competition College Students
Topics Essay Writing Competition College StudentsCheryl Thompson
 
Compare And Contrast Sample Essay College.pdf
Compare And Contrast Sample Essay College.pdfCompare And Contrast Sample Essay College.pdf
Compare And Contrast Sample Essay College.pdfBrenda Cooper
 

Similar to Thurman Model 1 Cross Sectional ECN 405 (19)

MMSS Senior Thesis 2000
MMSS Senior Thesis 2000MMSS Senior Thesis 2000
MMSS Senior Thesis 2000
 
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final Draft
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final DraftAn Analysis on the Influence of Mortgage Rates on Housing Prices - Final Draft
An Analysis on the Influence of Mortgage Rates on Housing Prices - Final Draft
 
Position Essay Examples.pdf
Position Essay Examples.pdfPosition Essay Examples.pdf
Position Essay Examples.pdf
 
Portion Four Housing Affordability: Lecture Note.ppt
Portion Four Housing Affordability: Lecture Note.pptPortion Four Housing Affordability: Lecture Note.ppt
Portion Four Housing Affordability: Lecture Note.ppt
 
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...
AEC 12_Q1_0304_PS_The Impact of Increase of Prices of Basic Commodities to Ho...
 
UNIT 1 - WHAT IS ECONOMICS LESSON...pptx
UNIT 1 - WHAT IS ECONOMICS LESSON...pptxUNIT 1 - WHAT IS ECONOMICS LESSON...pptx
UNIT 1 - WHAT IS ECONOMICS LESSON...pptx
 
In Text Citations Using Apa Endnote Example - Acadial
In Text Citations Using Apa Endnote Example - AcadialIn Text Citations Using Apa Endnote Example - Acadial
In Text Citations Using Apa Endnote Example - Acadial
 
Real Estate Level Forecasting - Review
Real Estate Level Forecasting - ReviewReal Estate Level Forecasting - Review
Real Estate Level Forecasting - Review
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
 
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docxBUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
BUS308 Week 4 Lecture 1 Examining Relationships Expect.docx
 
final is
final isfinal is
final is
 
« PreviousHomeNext »Home » Measurement » Levels of Measure.docx
« PreviousHomeNext »Home » Measurement » Levels of Measure.docx« PreviousHomeNext »Home » Measurement » Levels of Measure.docx
« PreviousHomeNext »Home » Measurement » Levels of Measure.docx
 
community devolepment
community devolepmentcommunity devolepment
community devolepment
 
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...
Quantitative Methods for Lawyers - Class #5 - Research Design Part V - Profes...
 
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
 
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx20   THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
20 THE NEW” HOUSING AND MORTGAGE MARKET SPRING 2016The .docx
 
Topics Essay Writing Competition College Students
Topics Essay Writing Competition College StudentsTopics Essay Writing Competition College Students
Topics Essay Writing Competition College Students
 
Compare And Contrast Sample Essay College.pdf
Compare And Contrast Sample Essay College.pdfCompare And Contrast Sample Essay College.pdf
Compare And Contrast Sample Essay College.pdf
 

Thurman Model 1 Cross Sectional ECN 405

  • 1. Elizabeth Thurman ECN 405 2L 1 Model 1 Paper Cross Sectional Introduction: The cost of housing is an important factor when choosing where to live and also in considering the economic well-being of a state. High housing costs can be an indication of higher paying jobs and higher prices in general in a particular area. When housing prices are increased, it indicates an increased demand for housing in that area. Often, when prices are high, one must achieve a certain level of education in order to obtain a high paying job in a high-priced area. The question I would like to analyze is how personal income affects housing prices in a state, and also how level of education ties in with level of income to affect said housing prices. Therefore, in this paper, I will be using personal income and level of education as my testing variables in a regression analysis to see if the amount of money a consumer makes, and their level of education will have an effect on the dependent variable, housing prices. Also, I will be exploring the impact of income and education on home mortgages throughout this paper by using five related studies in the form of journal articles. I will be exploring how each article relates and shapes my model and question. I will then pose my economic model that I will use to answer my question along with its estimation and corresponding graphs.
  • 2. Elizabeth Thurman ECN 405 2L 2 Further, I will test to see if I have evidence of heteroscedasticity which could imply that I have a violation of constant variance in my error terms. I will then test for endogeneity which will allow me to assess whether there is a correlation between an independent variable and an error term. If the test results show that I do have an endogenous model, my regression coefficient would cause the model to be biased, and therefore violating the OLS rules for being the best estimator. After testing for homoscedasticity and endogeneity, I will do a Ramsey reset test for zero mean to see if my model is specified correctly, test for normality, and preform a Wald test. Then, I will perform a final weighted estimation on my variables and then test a binary model for states west versus east of the Mississippi River. Review of the literature: The article by Stephen Malpezzi, "Housing prices, externalities, and regulation in US metropolitan areas" analyzes the determinants of housing prices which vary widely across the United States. The study focuses mainly on city and metropolitan areas. It uses a simple supply and demand framework to assess how regulatory actions effect housing prices and uses factors such as income and population changes. This article has helped me to shape my model because of the use of population and income in relation to housing prices. I have included both as variables in my model to study their effect. In Malpezzi’s article, I like that he related the housing prices to metropolitan areas and chose to do my binary testing based on states east and west of the Mississippi. This is because although both coasts tend to have a relatively large number of metropolitan areas compared to the “inner states”, the east coast states
  • 3. Elizabeth Thurman ECN 405 2L 3 seem to have more populated areas throughout. Highly populated areas tend to have high demand for housing, thus increasing the prices of homes (or other housing amenities). This is why I especially wanted to include population; however, I used population for an entire state which will not give me the more specific relation to housing prices that I would have liked. The article "Education and income" by Hendrik S. Houthakker, analyzes income levels in relation to a person’s level of education. It is performed as a cross sectional analysis of different age groups across a single year. The article concludes that “capital values increase uniformly as education increases”. However, it finds that those with college levels 1-3 do not fare as well as those with only a high school diploma. This concept led me to use the testing variable ‘level of education’ because those who afford higher priced homes may be correlated with higher paying jobs which would most likely require more than a high school diploma. Because the study found a positive correlation between level of education and level of income, I decided it would be a reasonable variable to use along with the amount of personal income. In the article "Valuation of education and crime neighborhood characteristics through hedonic housing prices" by Robin A. Dubin and Allen C. Goodman, housing prices are studied as a bundle of neighborhood characteristics. Among the characteristics in the bundle are those of crime and neighborhood schools. The study found that housing values were influenced in the Baltimore metropolitan area significantly when the two variables were studied together.
  • 4. Elizabeth Thurman ECN 405 2L 4 This article inspired me to use state crime rates in 2010 to see how it affects the cost of housing. Because one would intuitively guess that housing prices are cheaper where there is more crime, I would like to test for a correlation. Many homeowners want to live in an area where there is low crime not only for the safety of their family and safe schools, but also for the safety of their belongings. Areas with high breaking and entering rates may also drive the nearby cost of housing down. However, sometimes it costs more to live in a city where the crime rates tend to be higher. This is another reason I would also like to add crime rates as a variable. However, I will be using crime as the amount of burglaries and will use it as a state data and not specifying it to suburban, city, or rural housing, although this would be an interesting study. The article "Housing and income" by Alan R. Winger finds that income and housing do have a relationship and should involve other factors such as the permanent income of an individual. In the past, studies have concluded that there is a relationship; however, the specific factors involved where not clear. The article notes that you must take into account that the decision to purchase a home was made in the past relative to when each study is performed. Therefore it is harder to conclude what factors are considered when a consumer decides to buy their home. The article mentions a paper by Margaret Reid who studied the effects and also found that there is a strong correlation between income and the housing market. Because a strong correlation was found, and the two variables of housing related to income are widely studied, I was inspired to also test for a correlation between the two. If I were to analyze a second model, I would include a test for income over a certain time period to see if it fluctuates with housing prices because income can also
  • 5. Elizabeth Thurman ECN 405 2L 5 fluctuate. Thus, what a consumer may be able to afford in one year, may not be true in the next. Similarly, in the article, "Housing and permanent income: Tests based on a three-year re-interview survey” by Lee, Tong Hun, the author points out that as consumption is tested against income, income is usually tested at a point in time. Instead, he notes, particularly with the consumption and purchase of mortgage loans (housing), it is better to test the relationship with a fixed income over time. This is because purchasing power may fluctuate and although a consumer may be making a lot in a particular year, they may not make as much in following years and therefore not affect the mortgage market as greatly. This article gave some great insight into the notion that it must be kept in mind we are testing income and housing at one point in time. Although results were found, indication that income over time is a more accurate measure when related to housing consumption, I decided to keep my variable of income because I am interested to see if there is any correlation with housing prices combined with other factors such as the population’s level of education, level of crime, GDP per capita, etc. The Model: Y(House Prices) = B0 + B1(Personal Income) + B2(Level of Education) + B3(State Taxes) + B4(Unemployment Rate) + B5(GDP) + B6(Population) + B7(Burglary) + U This is a cross sectional model measuring 2010 data on average housing costs in relation to income and level of education in each of the 50 states. (Data used can be found in Appendix A)
  • 6. Elizabeth Thurman ECN 405 2L 6 Assumptions about disturbance term: 1. Linearity – The expected value of Y is linearly associated given the X’s. Also, the average value of the error term is equal to zero given X. 2. Unbiased – Error terms are independent of one another. 3. Homoscedasticity - Constant error variance. This means that the variance of the errors is the same regardless of X. V(ε|xi)=σε2. The degree of random noise is the same regardless of the value of the X’s. 4. Mean Independence – The error term is independent of the X variables. 5. Normality – The error terms are normally distributed. Variables: House Price Index: This is the dependent variable (Y). Personal Income per capita: This is a testing independent variable. This should have a positive effect on housing prices because the greater purchasing power consumers have, the more housing they will buy, thus increasing demand and therefore pricing of houses. Level of Education: This is another testing independent variable. It should have a positive effect on housing prices. Those who obtain a higher level of education tend to earn higher income, thus have higher purchasing power driving up demand and prices of housing. Total State Taxes: Independent variable. This should have a positive impact on housing prices. As taxes increase, so do property taxes.
  • 7. Elizabeth Thurman ECN 405 2L 7 Unemployment Rate: This is another independent variable. If the unemployment rate is low, the housing rate may be higher because more people are able to earn a living, thus driving demand for housing. GDP per capita: Independent variable. This should have a positive effect on housing prices. Economically healthy states who have higher GDP may have higher costs of living due to its many consumers earning a living. Population: This is an independent variable. The population in a state may cause higher costs for housing because with more people comes a higher demand for housing. When supply is limited, and housing demand increases, so must the price. State Total Burglary: Independent variable. This may have a negative impact on home prices because housing tends to be cheaper if there is a higher rate of crime in the area. The demand for housing in these types of areas is lower than safer areas. I will use this model to answer my question by testing the independent variables, Personal Income and Level of Education against Housing Prices. I will test to see how correlated the variables are and whether they are significant. I will then do a test on a binary variable for which I chose to use states west and states east of the Mississippi river. I chose this criteria for my binary because I wanted to compare if there was a difference between the west coast and east coast. The United States tends to be split with higher populations toward each coast, having lots of land with lower populations throughout its middle section. This is why I chose to compare the large populations of California and Texas (these are the states with the
  • 8. Elizabeth Thurman ECN 405 2L 8 higher population data on the west coast) to the states Florida, New York, Ohio, Pennsylvania and Illinois (higher population data on the east coast). Initial estimation and hypothesis testing: DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:16:48 Sample:1 50 Included observations:50 Variable Coefficient Std. Error t-Statistic Prob. C -115.4761 88.06088 -1.311322 0.1967 INCOME 0.003878 0.003856 1.005642 0.3202 EDUCATION 10.95013 2.965625 3.692351 0.0006 TAXES 2.28E-06 8.96E-07 2.543248 0.0147 BURGLARY -0.000868 0.000305 -2.840839 0.0068 CAPITAGDP -0.001513 0.001888 -0.801578 0.4272 UNEMPLOYMENT 7.641442 4.836332 1.580008 0.1214 R-squared 0.654494 Mean dependentvar 329.1984 Adjusted R-squared 0.606284 S.D. dependentvar 94.81226 S.E. of regression 59.49163 Akaike info criterion 11.13873 Sum squared resid 152187.9 Schwarz criterion 11.40641 Log likelihood -271.4681 Hannan-Quinn criter. 11.24066 F-statistic 13.57588 Durbin-Watson stat 1.928779 Prob(F-statistic) 0.000000
  • 9. Elizabeth Thurman ECN 405 2L 9 When testing the dependent variable (housing prices) against all other variables, initially I find that burglary has a negative correlation with housing prices. This was as I expected. All variables seem to have a correlation with my dependent variable; however, GDP per capita seems to have the least correlation when compared with other variables. Also, this variable has a negative beta, which I expected the correlation to be positive because the income of a state increases as taxes increase which would mean the housing costs more. The following are correlation graphs. This gives a visual display the relationship between each variable against the dependent variable. 100 200 300 400 500 600 700 30,000 35,000 40,000 45,000 50,000 55,000 60,000 Personal Income per capita ($) HousePriceIndex(inhundreds) Personal income appears to have a generally positive correlation, as expected.
  • 10. Elizabeth Thurman ECN 405 2L 10 100 200 300 400 500 600 700 0 20,000,000 60,000,000 100,000,000 Total State Taxes (in thousands) HousePriceIndex(inhundreds) It appears as though total state taxes remain around one area with varying house prices. This makes sense since state income may be in the same general area, the prices of its houses vary much more greatly. 100 200 300 400 500 600 700 16 20 24 28 32 36 40 level of Education HousePriceIndex(inhundreds) Level of education appears to have a positive correlation with housing prices, as expected.
  • 11. Elizabeth Thurman ECN 405 2L 11 100 200 300 400 500 600 700 0 40,000 80,000 120,000 200,000 State Total Burglary HousePriceIndex(inhundreds) State burglary appears to vary. As burglary levels remain around a certain number, house prices seem to vary. However, it appears as burglary rates increase, housing prices don’t stay at high levels. 100 200 300 400 500 600 700 30,000 40,000 50,000 60,000 70,000 GDP per capita HousePriceIndex(inhundreds) GDP per capita appears to have a roughly positive correlation, however not very strongly. I expected a stronger correlation.
  • 12. Elizabeth Thurman ECN 405 2L 12 100 200 300 400 500 600 700 2 4 6 8 10 12 14 Unemployment Rate HousePriceIndex(inhundreds) Unemployment rate does not appear to have a very strong correlation; however unemployment tends to stay around a certain level while house prices vary. 100 200 300 400 500 600 700 0 10,000,000 20,000,000 30,000,000 40,000,000 POPULATION HousePriceIndex(inhundreds) The population variable does not appear to have a high correlation with the housing price variable.
  • 13. Elizabeth Thurman ECN 405 2L 13 Conclusion/Test Results: In conclusion, I have found that housing prices are more correlated with my testing variable income, than with the testing variable level of education. In my final estimation I found that my variables are significant, and that burglary is negatively correlated as I expected. Due to the high F-stat in my Breusch-Pagan-Godfrey test for heteroscedasticity (Appendix B), I find that I reject the null concluding that my model has implication of heteroscedasticity. Therefore, my error terms may be correlated to some degree and also have varying distributions and variances. In my test for endogeniety found in (Appendix B) I found that because my t-stat of -0.586935 was less than my t-crit of 2.02, I fail to reject the null, concluding that I do not have endogeneity. Therefore, my variables are not correlated with the error term. In the Ramsey reset test for zero mean I find that my model is specified correctly. The normality test shows that my model is not a normal distribution. The Wald test shows that my testing variables are not jointly significant at a value of 0. In the final weighted estimation I found that most all of my variables tend to be significantly correlated with the exception of GDP per capita, as it was in my initial estimation. Therefore, I would conclude that while I would want to include more data variables, housing prices are related to the level of one’s income. There is also a higher correlation between housing prices and burglary rates. It has a negative correlation, meaning as burglary rates are higher, housing prices tend to be lower. If I created a
  • 14. Elizabeth Thurman ECN 405 2L 14 new model, or added onto this model, I would do more testing for crime rates and use it as a testing variable rather than a control variable. References: Article 1 Dubin, Robin A., and Allen C. Goodman. "Valuation of education and crime neighborhood characteristics through hedonic housing prices." Population and environment 5.3 (1982): 166-181. Article 2 Houthakker, Hendrik S. "Education and income." The Review of Economics and Statistics (1959): 24-28. Article 3 Lee, Tong Hun. "Housing and permanent income: Tests based on a three-year reinterview survey." The Review of Economics and Statistics (1968): 480-490. Article 4 Malpezzi, Stephen. "Housing prices, externalities, and regulation in US metropolitan areas." Journal of Housing Research 7 (1996): 209-242. Article 5 Winger, Alan R. "Housing and income." Economic Inquiry 6.3 (1968): 226-252.
  • 15. Elizabeth Thurman ECN 405 2L 15 Appendix A: State House Price Index (in hundreds ) Personal Income per capita ($) level of Educatio n Total State Taxes (in thousands) Unemplo yment Rate GDP per capita Populatio n State Total Burglary Binary: West=1, East=0 Alabama 291.11 33,894 21.9 8,419,911 9.3 36156 4,785,570 42,034 0 Alaska 280.23 45,565 27.9 4,522,927 7.9 68656 713,868 3,105 1 Arizona 263.87 33,993 25.9 10,719,958 10.4 38222 6,408,790 50,771 1 Arkansas 245.27 32,017 19.5 7,559,898 7.9 37658 2,922,280 32,511 1 California 411.31 42,282 30.1 107,195,465 12.3 51546 37,333,601 228,857 1 Colorado 342.25 41,689 36.4 8,575,262 9 49923 5,048,196 26,153 1 Connecticut 409.47 55,216 35.5 12,344,106 9.3 64766 3,579,210 15,172 0 Delaware 437.58 40,969 27.8 2,763,032 8 62994 899,711 7,515 0 Florida 296.59 38,478 25.8 30,484,883 11.3 38258 18,846,054 169,119 0 Georgia 284.15 34,341 27.3 14,782,779 10.2 41894 9,713,248 96,723 0 Hawaii 448.63 41,668 29.5 4,837,862 6.8 48694 1,363,731 8,663 1 Idaho 278.66 32,100 24.4 2,951,703 8.7 34825 1,570,718 6,502 1 Illinois 318.26 42,033 30.8 27,795,759 10.4 50296 12,839,695 75,399 0 Indiana 245.34 34,344 22.7 13,795,221 10.1 43207 6,489,965 47,115 0 Iowa 246.02 39,033 24.9 6,809,344 6.3 46052 3,050,314 16,656 1 Kansas 234.66 38,811 29.8 6,492,996 7.1 43556 2,858,910 19,404 1 Kentucky 287.18 32,929 20.5 9,531,404 10.2 37746 4,347,698 30,311 0 Louisiana 241.53 37,199 21.4 8,758,633 7.4 48594 4,545,392 45,435 1 Maine 462.13 37,213 26.8 3,489,953 8.2 38374 1,327,366 7,359 0 Maryland 423.05 50,035 36.1 15,237,748 7.8 54080 5,787,193 36,542 0 Massachusetts 617.39 51,487 39 20,090,563 8.3 60354 6,563,263 37,767 0 Michigan 241.09 35,082 25.2 22,208,870 12.7 39056 9,876,149 73,868 0 Minnesota 312.11 42,572 31.8 17,208,877 7.4 50641 5,310,337 24,415 1 Mississippi 243.14 30,834 19.5 6,268,823 10.5 31331 2,970,047 30,444 0 Missouri 278.25 36,606 25.6 9,707,053 9.3 42610 5,996,063 44,043 1 Montana 353.21 34,612 28.8 2,142,809 6.7 36918 990,527 3,654 1 Nebraska 252.66 39,926 28.6 3,864,897 4.7 49119 1,829,838 8,326 1 Nevada 216.76 36,657 21.7 5,835,963 13.8 44102 2,703,230 22,226 1 N Hampshire 404.59 44,963 32.8 2,271,936 6.1 47224 1,316,614 5,441 0 New Jersey 485.57 50,941 35.4 25,927,891 9.6 56025 8,802,707 38,732 0 New Mexico 299.21 33,175 25 4,295,237 7.9 39316 2,064,982 21,014 1 New York 576.76 49,582 32.5 63,807,610 8.6 60974 19,398,228 64,973 0 N Carolina 314.54 35,435 26.5 21,514,930 10.8 43778 9,559,533 102,690 0 N Dakota 254.68 43,275 27.6 2,645,695 3.8 51254 674,344 1,966 1 Ohio 245.55 36,199 24.6 23,583,596 10 42342 11,545,435 106,521 0 Oklahoma 203.06 35,912 22.9 7,082,161 6.9 39377 3,759,263 37,476 1 Oregon 373.79 35,898 28.8 7,475,135 10.7 49538 3,837,208 19,637 1 Pennsylvania 373.96 41,635 27.1 30,169,122 8.4 45976 12,710,472 55,171 1 Rhode Island 474.92 43,013 30.2 2,568,759 11.7 46277 1,052,669 6,121 0 S Carolina 317.6 32,669 24.5 7,242,724 11.2 35078 4,636,361 46,156 0 S Dakota 290.44 40,613 26.3 1,321,228 5.1 46507 816,211 3,181 1 Tennessee 288.17 35,426 23.1 10,513,788 9.8 39649 6,356,683 64,235 0 Texas 223.8 38,065 25.9 39,516,186 8.2 47617 25,245,178 228,597 1 Utah 320.5 32,447 29.3 5,237,427 8.1 42075 2,774,424 15,017 1 Vermont 436.9 40,134 33.6 2,511,387 6.4 42097 625,793 3,366 0 Virginia 403.22 44,836 34.2 16,411,055 7.1 52084 8,024,417 30,629 0 Washington 414.56 42,547 31.1 16,106,154 9.9 52850 6,742,256 55,164 1 W Virginia 217.41 31,798 17.5 4,803,704 8.4 34818 1,854,146 10,756 0 Wisconsin 303.86 38,728 26.3 14,368,569 8.5 44431 5,689,060 26,566 0 Wyoming 274.93 45,025 24.1 2,158,260 7 66256 564,222 2,149 1
  • 16. Elizabeth Thurman ECN 405 2L 16 Appendix B: Test for Heteroscedasticity: HeteroscedasticityTest:Breusch-Pagan-Godfrey F-statistic 6.258169 Prob. F(1,42) 0.0163 Obs*R-squared 5.705965 Prob. Chi-Square(1) 0.0169 Scaled explained SS 3.957595 Prob. Chi-Square(1) 0.0467 Test Equation: DependentVariable:RESID^2 Method: LeastSquares Date: 12/11/14 Time:14:53 Sample:1 50 Included observations:44 Variable Coefficient Std. Error t-Statistic Prob. C -6821.033 4419.425 -1.543421 0.1302 CAPITAGDP 0.236573 0.094568 2.501633 0.0163 R-squared 0.129681 Mean dependentvar 4054.004 Adjusted R-squared 0.108959 S.D. dependentvar 5592.571 S.E. of regression 5279.105 Akaike info criterion 20.02529 Sum squared resid 1.17E+09 Schwarz criterion 20.10639 Log likelihood -438.5564 Hannan-Quinn criter. 20.05537 F-statistic 6.258169 Durbin-Watson stat 2.313727 Prob(F-statistic) 0.016345
  • 17. Elizabeth Thurman ECN 405 2L 17 DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:14:55 Sample:1 50 Included observations:44 Weighting series:ABS(VRES) Weight type: Variance (average scaling) Variable Coefficient Std. Error t-Statistic Prob. C -108.2291 52.54895 -2.059586 0.0463 INCOME 0.012422 0.002053 6.050098 0.0000 TAXES 2.13E-06 5.89E-07 3.615020 0.0009 BURGLARY -0.000847 0.000182 -4.642958 0.0000 CAPITAGDP -0.002872 0.001753 -1.638338 0.1096 UNEMPLOYMENT 9.806262 3.717914 2.637571 0.0120 Weighted Statistics R-squared 0.872892 Mean dependentvar 247.1635 Adjusted R-squared 0.856168 S.D. dependentvar 207.7145 S.E. of regression 27.66106 Akaike info criterion 9.604052 Sum squared resid 29075.11 Schwarz criterion 9.847351 Log likelihood -205.2892 Hannan-Quinn criter. 9.694279 F-statistic 52.19178 Durbin-Watson stat 1.870739 Prob(F-statistic) 0.000000 Weighted mean dep. 313.1386 Unweighted Statistics R-squared 0.547538 Mean dependentvar 327.4782 Adjusted R-squared 0.488004 S.D. dependentvar 96.43644 S.E. of regression 69.00406 Sum squared resid 180939.3
  • 18. Elizabeth Thurman ECN 405 2L 18 Durbin-Watson stat 2.070976 Test for Endogeniety: DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:14:08 Sample:1 50 Included observations:44 Variable Coefficient Std. Error t-Statistic Prob. C -122.7353 100.7347 -1.218402 0.2310 INCOME -0.008659 0.022321 -0.387937 0.7003 TAXES 2.01E-06 1.15E-06 1.752352 0.0882 EDUCATION 11.58633 4.461864 2.596747 0.0135
  • 19. Elizabeth Thurman ECN 405 2L 19 CAPITAGDP 0.008608 0.017224 0.499797 0.6203 BURGLARY -0.000785 0.000376 -2.088543 0.0439 UNEMPLOYMENT 9.239402 5.959531 1.550357 0.1298 IVRES -0.010193 0.017367 -0.586935 0.5609 R-squared 0.634308 Mean dependentvar 327.4782 Adjusted R-squared 0.563202 S.D. dependentvar 96.43644 S.E. of regression 63.73555 Akaike info criterion 11.31033 Sum squared resid 146239.9 Schwarz criterion 11.63473 Log likelihood -240.8272 Hannan-Quinn criter. 11.43063 F-statistic 8.920510 Durbin-Watson stat 1.908336 Prob(F-statistic) 0.000002 Test for Zero Mean: RamseyRESET Test Equation:UNTITLED Specification:CPIHOUSING C INCOME TAXES EDUCATION CAPITAGDP BURGLARY UNEMPLOYMENT Instrumentspecification:C INCOME TAXES EDUCATION CAPITAGDP BURGLARY POPULATION Omitted Variables:Powers of fitted values from 2 to 4 Value df Probability F-statistic 2.214646 (3, 34) 0.1043 Likelihood ratio 7.853523 3 0.0491 F-test summary: Sum of Sq. df Mean Squares
  • 20. Elizabeth Thurman ECN 405 2L 20 Test SSR 24134.14 3 8044.714 Restricted SSR 147639.3 37 3990.252 Unrestricted SSR 123505.2 34 3632.506 LR testsummary: Value df Restricted LogL -241.0367 37 Unrestricted LogL -237.1100 34 Unrestricted TestEquation: DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:14:20 Sample:1 50 Included observations:44 Variable Coefficient Std. Error t-Statistic Prob. C 11126.00 7566.153 1.470497 0.1506 INCOME -0.245853 0.165883 -1.482087 0.1475 EDUCATION -585.4722 400.5812 -1.461557 0.1530 TAXES -0.000137 9.33E-05 -1.469711 0.1508 CAPITAGDP 0.081795 0.055551 1.472444 0.1501 BURGLARY 0.051451 0.035064 1.467355 0.1515 UNEMPLOYMENT -441.7197 300.9174 -1.467910 0.1513 FITTED^2 0.275000 0.179630 1.530927 0.1350 FITTED^3 -0.000558 0.000353 -1.579705 0.1234 FITTED^4 4.15E-07 2.54E-07 1.635528 0.1112 R-squared 0.691159 Mean dependentvar 327.4782
  • 21. Elizabeth Thurman ECN 405 2L 21 Adjusted R-squared 0.609407 S.D. dependentvar 96.43644 S.E. of regression 60.27027 Akaike info criterion 11.23227 Sum squared resid 123505.2 Schwarz criterion 11.63777 Log likelihood -237.1100 Hannan-Quinn criter. 11.38265 F-statistic 8.454350 Durbin-Watson stat 1.869290 Prob(F-statistic) 0.000002 Normality: 0 1 2 3 4 5 6 -40 -30 -20 -10 0 10 20 30 40 50 Series: Standardized Residuals Sample 1 50 Observations 44 Mean 0.151793 Median -3.511059 Maximum 48.16692 Minimum -43.24634 Std. Dev. 26.00272 Skewness 0.266158 Kurtosis 1.794042 Jarque-Bera 3.185773 Probability 0.203338 Wald Test: Equation:Untitled Test Statistic Value df Probability F-statistic 28.84905 (2, 38) 0.0000 Chi-square 57.69810 2 0.0000 Null Hypothesis:C(2)=C(3)=0 Null Hypothesis Summary:
  • 22. Elizabeth Thurman ECN 405 2L 22 Normalized Restriction (= 0) Value Std. Err. C(2) 0.012422 0.002053 C(3) 2.13E-06 5.89E-07 Restrictions are linear in coefficients. Final Weighted Estimation: DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:14:55 Sample:1 50 Included observations:44 Weighting series:ABS(VRES) Weight type: Variance (average scaling) Variable Coefficient Std. Error t-Statistic Prob. C -108.2291 52.54895 **-2.059586 0.0463 INCOME 0.012422 0.002053 ***6.050098 0.0000 TAXES 2.13E-06 5.89E-07 ***3.615020 0.0009 BURGLARY -0.000847 0.000182 ***-4.642958 0.0000 CAPITAGDP -0.002872 0.001753 *-1.638338 0.1096 UNEMPLOYMENT 9.806262 3.717914 ***2.637571 0.0120 Weighted Statistics R-squared 0.872892 Mean dependentvar 247.1635 Adjusted R-squared 0.856168 S.D. dependentvar 207.7145 S.E. of regression 27.66106 Akaike info criterion 9.604052 Sum squared resid 29075.11 Schwarz criterion 9.847351 Log likelihood -205.2892 Hannan-Quinn criter. 9.694279
  • 23. Elizabeth Thurman ECN 405 2L 23 F-statistic 52.19178 Durbin-Watson stat 1.870739 Prob(F-statistic) 0.000000 Weighted mean dep. 313.1386 Unweighted Statistics R-squared 0.547538 Mean dependentvar 327.4782 Adjusted R-squared 0.488004 S.D. dependentvar 96.43644 S.E. of regression 69.00406 Sum squared resid 180939.3 Durbin-Watson stat 2.070976 *Reject at alpha = 0.10 **Reject at alpha = 0.05 ***Reject at alpha = 0.01 Binary Estimation: DependentVariable:CPIHOUSING Method: LeastSquares Date: 12/11/14 Time:16:30 Sample:1 50 Included observations:50 Variable Coefficient Std. Error t-Statistic Prob. C 59.01909 118.0173 0.500089 0.6198 INCOME 0.008922 0.004040 2.208542 0.0332 INCOME*BINARY -0.001853 0.005813 -0.318785 0.7516 TAXES 1.84E-06 1.47E-06 1.249160 0.2191 TAXES*BINARY 5.24E-07 1.89E-06 0.277112 0.7832 BURGLARY -0.000798 0.000511 -1.562475 0.1263 BURGLARY*BINARY 4.04E-05 0.000684 0.059132 0.9531
  • 24. Elizabeth Thurman ECN 405 2L 24 CAPITAGDP 0.000345 0.003074 0.112353 0.9111 CAPITAGDP*BINARY -0.002311 0.004385 -0.527089 0.6011 UNEMPLOYMENT -6.536023 8.989954 -0.727036 0.4715 UNEMPLOYMENT*BINAR Y 13.60267 9.166913 1.483888 0.1459 R-squared 0.633596 Mean dependentvar 329.1984 Adjusted R-squared 0.539646 S.D. dependentvar 94.81226 S.E. of regression 64.32953 Akaike info criterion 11.35745 Sum squared resid 161393.2 Schwarz criterion 11.77810 Log likelihood -272.9363 Hannan-Quinn criter. 11.51764 F-statistic 6.743987 Durbin-Watson stat 1.987158 Prob(F-statistic) 0.000006 Appendix C: Test for heteroscedasticity: H0: The model contains homoscedasticity H1: The model contains heteroscedasticity Critical value F(1,42): 4.07 F Statistic: 6.258169 Therefore reject null. Conclude the model is heteroscedastic Test for endogeniety: H0: The model is not endogenous H1: The model is endogenous DF: 36 Tcritical: 0.10, 0.05, 0.01 = 1.302, 1.683, 2.422 respectively.
  • 25. Elizabeth Thurman ECN 405 2L 25 Tstat: |-0.586935| Therefore, conclude that the model is not endogenous. Test for zero mean: H0: Γ = 0 (model is specified correctly) H1: Γ ≠ 0 (model is not specified correctly) Fstatistic = 2.214646 < Tcritical = 2.89 Therefore, fail to reject H0. Normality: H0: skewness & kurtosis = 0 H1: otherwise Tcritical: 1.696 < Jarque-Bera: 3.1858 Therefore, reject H0, conclude my model is not normal. Wald test: H0: B1=B2=0 H1: Otherwise Fcritical: 3.24 < Fstatistic: 28.84905 Therefore reject Ho. My testing variables are not jointly significant at a value of 0.