«Daniel Domb An honors thesis submitted in partial fulfillment of the requirements for the degree of Bachelor of Science Undergraduate College Leonard ...»
Theory and Evidence..
An honors thesis submitted in partial fulfillment
of the requirements for the degree of
Bachelor of Science
Leonard N. Stern School of Business
New York University
Professor Marti G. Subrahmanyam Professor Crocker Liu
Faculty Adviser Thesis Advisor
“Many homeowners might have saved tens of thousands of dollars had they held adjustable-rate mortgages rather than fixed-rate mortgages during the past decade.”1 This surprising statement was made by Alan Greenspan on February 23, 2004. Alan Greenspan, who rarely gives financial advice made this statement emphasizing the importance of the financial decisions homeowners make when they purchase a home, and the potential savings that can be made from the correct choices. The housing market is a critical component of our economy, and over the past 3 years, the robust housing market has been the silver lining to the overall lackluster economy.
Over the past year, we have seen mortgage rates drop to record lows week after week. At the same time, borrowing activity has increased to record levels. The housing industry has seen many highs and lows in 2003. The median price of all homes sold hit a record high in August 2003 at $253,900. New construction starts hit a 25 year high in 2003 with 1.85 million units.2 The average effective fixed rate mortgage hit a low of 5.51% in July of 2003. The average adjustable rate mortgage hit a low of 4.70% in July of 2003. With the low rates refinancing hit a record low in May of 2003.3 With rates as low as they are, borrowers can afford more house for their money than ever before.
Although the low rates have helped the economy over the past few years, rates are most likely only going to go up from their current levels.
Alan Greenspan, February 23, 2004, MSNBC http://money.cnn.com/2004/01/21/news/economy/housing_starts/index.htm http://money.cnn.com/2003/07/02/commentary/bidask/bidask/index.htm
-1When homeowners purchase a home they make many decisions. The first decisions purchasers make are the location of the home and the size. Beyond the qualitative choices, each individual must decide the price they are willing to spend on their home. Once the buyer decides on a price to pay for a home, they must decide their desired debt to equity ratio. Some people buy the entire home for cash, but the majority of people use some ratio of their personal cash and the rest is made up of debt that will most likely come in the form of a mortgage. Assuming that the buyer is taking out a mortgage, they must decide what type of mortgage to take out, most likely either a fixed rate mortgage or an adjustable rate mortgage. However, within these two categories of mortgages the mortgagee must decide the term to maturity of the loan. The mortgage could be a 1 year adjustable rate, a 3/1 ARM, 5/1 ARM, 7/1 ARM, a 15 year fixed, a 30 year fixed rate, etc. Each decision a consumer makes about the financing of their home is critical to their future monthly payments.
LITERATURE REVIEW:Unfortunately, there has been very little research into how consumers make their decisions when purchasing a home. Referring to Economic Literature I found a lot of material about the housing market, both general research done and research done on individual areas.4 However, to my knowledge, there have not been any researches spending there time and efforts into examining consumers choices when purchasing a home. After my thesis is complete, I hope that future researchers will now have a starting point for their research.
Many economists spend their careers trying to predict the future path of interest rates. They make predictions for every type of rate from the three month treasury rate to the 30 year fixed mortgage rate. For my thesis I will try and develop a new way to predict the future path of interest rates. I propose that borrowers can predict the future movement of interest rates through their financing choices when purchasing a home.
More specifically, I don’t believe that each individual borrower can predict future interest rates, but I am hypothesizing that one can implicitly draw conclusions about future interest rates by examining the median borrower. I have already established that borrowers make many choices when they purchase a home, and I hope to develop a model that combines simple economic variables and the choices of the median borrower to shed light on the path of future interest rates. I hypothesize that a perfect model cannot exist with consumers’ choices alone, but with the addition of economic indicators related to the housing market the model will improve greatly. I believe that by using the information collected from borrowers’ choices one can predict the future path of interest rates 30, 60 and possibly even 90 days in advance.
1990 through August 2003. However, I originally collected data over a 20 year period. I was hoping to encompass many housing booms and recessions, and see interest rates very
not able to expand my data set beyond the 13 ½ year period I am examining. First, the mid 1980’s brought some interesting times to the interest rate market. The savings and loan crises of the mid 1980’s hurt the speculation of the real estate market and meanwhile fixed rates hit all time highs just below 20%, while adjustable rate mortgages had even higher rates than those fixed rate mortgages. The second reason for starting the data set in 1990 is the ability to track down accurate data for my analysis. Many of the different independent variables I plan to use in my regression analysis were not collected on a monthly basis before 1990. I made the decision to sacrifice multiple economic cycles for a shorter, but possibly more accurate time period. I believe that the most recent housing, and interest rate cycle, more closely resembles the cycles the United States will experience in the future.
I collected my data on a monthly basis, giving me 164 data points for my study. I collected data from monthly publications, mortgage websites, and federal statistics websites. The publications I found useful are Housing Market Statistics a monthly publication that has been published since 1991, and U.S. Housing Market Conditions which is published by the Department of Housing and Urban Development. The mortgage websites that have extensive historical statistics that I found useful are www.hsh.com (HSH Associates), www.mbaa.com (Mortgage Bankers Association of America), and www.freddiemac.com (Freddie Mac). There are two federal websites that have data collected by the government monthly over the past 15-20 years. The first is www.fedstats.gov which contains data from many different federal agencies. The second
Housing Finance Board’s website at www.fhfb.gov/MIRS/MIRS_downloads.htm.
Lastly, the United States Census Bureau provided many statistics that I used in my analysis.
Looking at interest rates from a general overview they are comprised of three factors. First, there is a real rate. The real rate is the amount of return demanded for holding a no risk investment for an instantaneous time period. The best way to find the real rate is to look at the shortest US Treasury rate, like the 3-month Treasury. The next factor is the inflation premium. The inflation premium is very different depending on the time frame of the investment and investors expectations about current and future inflation. Looking at the 1-year Treasury to the 10-year Treasury verses the 30-year Treasury, one can see the impact of different expectations of inflation over the three different time periods. The last factor is the risk premium. The risk premium does not apply to the US Treasury rate as long as we assume that there is no risk that the US government will not be able to pay off their future debt.
The first step in my research is to use the 10-year Treasury as a baseline measure for interest rate movements. Without the 10-year Treasury the final model will not be as statistically significant. For each variable that I suspect as an explanatory factor in the fixed rate mortgage, I will run a simple regression against the difference between the fixed rate mortgage and the 10-year Treasury. The variables I tested include, the percentage of adjustable rate mortgages compared to total mortgage originations, the
the median term to maturity of the loan, the investment to loan ratio of all commercial banks, the number of new single family homes sold, monthly housing starts, an index of the state of the economy going into a recession or coming out of one, the refinance index, and the total volume of mortgage originations. After I determine the relationship of each independent variable to the dependent variable I will run a multi-linear regression showing the statistically significant variables ability to explain the dependent variable, the fixed rate mortgage. This model will consist of both consumers’ choices and economic indicators. Hopefully it will explain away most of the uncertainty in the current fixed rate mortgage.
Once I have model to describe current rates I will run three benchmark regressions. Each regression will have the current fixed rate as the independent variable and the fixed rate plus 1 month, then fixed rate plus 2 months, and the fixed rate plus 3 months as the independent variable. These regressions will be used for comparison to the regressions attempting to predict the fixed rate. First I will develop regressions to predict the fixed rate 1, 2, and 3 months in the future using the variables based on consumers choices and economic indicators. Then, later, I will develop similar regressions again, but without any economic indicators to see if consumers can predict the future path of interest rates without the help of economic indicators. Ultimately, I will compare each of the regressions to the actual rates and see if there is any correlation.
DATA ALAYSIS & RESULTS:
the 10-year treasury rate, because the 10-year rate is most closely aligned with the average length of time of a fixed rate mortgage.5 Most conventional fixed rate mortgages are for either 15 or 30 years. However, in my analysis I took into account all fixed rate mortgages by using an effective time to maturity. Over the 13 ½ years that my study encompassed, the average term to maturity was 27.11 years. This means that more people took out 30 year mortgages than 15 year mortgages, because the average loan is very close to 30 years. The entire data set for term to maturity is between 24.70 years and 28.80 years. These term to maturities do not represent actual lengths of time that mortgages were held, they only measure the average length of mortgages at issuance.
Most mortgages do not make it to maturity. Borrowers either sell their homes or refinance prior to the expiration of the original mortgage. Therefore, the actual observed term to maturity of a fixed rate mortgage is much closer to 10 years. Mortgage institutions understand this phenomenon, and know that individuals will mostly average a 10 year holding period of their fixed rate mortgage. Given this information, the first variable I built into my model to explain the fixed rate mortgage was the government issued 10 year Treasury rate.
Before I ran my first regression I plotted the 10-year Treasury and the fixed rate mortgage on a two line plot to examine any graphical relationships between the two variables. Exhibit 1 displays the graph with the two variables plotted against each other.
The graph shows that the difference between the fixed rate mortgage and the 10-year www.HSH.com, HSH Associates
half of the data set period, the fixed rate mortgage and the 10-year Treasury is fairly close together, and get very close at times. However, after 1997 the fixed rate stayed further above the 10-year Treasury.
My first regression contains only two variables, the 10-year Treasury rate as the independent variable and the fixed rate mortgage as the dependent variable. This simple regression is statistically significant with an f-statistic of about 2210 and yields an R-Sq (adj) of 93.1%. The high R-Sq shows me that a large portion of the fixed rate mortgage can be described by the 10-year Treasury rate. The 93.1% R-Sq seems to make sense for this regression because the 10-year Treasury implicitly has the inflation risk and the real rate built in and both also make up the fixed rate mortgage. Exhibit 2 shows the residuals for this regression and although the R-Sq is very high, the residuals show a distinctive pattern throughout. Comparing those residuals to the graphical display of the two variables in Exhibit 1, one can see that anytime there is a large spike in the rates either up or down, the residuals are at their highest. The early years and the later years of the data set show the most volatility in residuals. Between the dip in 1994 and the spike in residuals in 1998 the simple linear regression is a fairly good model for explaining fixed rate mortgages.
Earlier I claimed that the 10-year Treasury makes a better comparison to the fixed rate mortgage than a shorter time frame risk-free rate, like the 1-year Treasury.
Comparing Exhibit 3, which plots the fixed rate against the 1-year Treasury, to Exhibit 1,