«Daniel Domb An honors thesis submitted in partial fulfillment of the requirements for the degree of Bachelor of Science Undergraduate College Leonard ...»
Although the multi-linear regressions are not accurate in predicting individual month’s mortgage rates, there may be some hidden results that may be helpful. I decided to strip down the models and isolate the variable that I thought told the best story of fixed rate mortgages. The variable I chose to examine is the percentage of adjustable rate mortgages taken out monthly. In Exhibit 39 I ran regressions for the percentage of adjustable rate mortgages and the fixed rate plus 1, 2 and 3 months. First, I plotted the predicted fixed rate 1 month in advance against a 3 month moving average in Exhibit 40.
I chose to plot the predicted rate against a 3 month moving average to eliminate some of the noise that occurs when looking at the results on a month to month basis. The graph in Exhibit 40 shows that based on consumers choices, the trend of fixed rate mortgages can be anticipated through looking at the percentage of adjustable rate mortgages taken out in a given month. Going a step further, I graphed the 1, 2, and 3 month lagged regression in Exhibit 41 against the actual fixed rate. Again, the predicted fixed rate based on each of the three regressions is an accurate prediction for the actual fixed rate.
The question still remains, can consumers predict interest rates? Well consumers can definitely not tell what the effective fixed rate mortgage will be next month or any month in the future. There are way too many variables involved, and there is a certain
an increasingly worse job at predicting rates further in the future than next month. It seems as if the further in the future that consumers try and predict interest rates, the more importance is placed on economic variables to develop a rough estimate of the future fixed mortgage rate.
Originally, I had planned to compare consumers’ predictive abilities to the average economist. Unfortunately, although economists predict many different interest rate measures, they do not predict an effective fixed mortgage rate. The closest predictor is the 30-year mortgage rate, but I cannot compare the economists predictive power compared with consumers, because I would be comparing apples with oranges and would not be able to draw any accurate conclusions.
However, the story does not end there. Although we cannot decipher the exact future mortgage rate from consumers’ choices, we can predict the future path of interest rates through one of those choices consumers make, whether they are taking out a fixed rate mortgage or an adjustable rate mortgage. Inherently, consumers want to save as much money as possible. Therefore, if consumers believe rates are about to rise in the future they will take out more fixed rate mortgages. Conversely, if consumers believe that rates are high and will most likely drop in the near future they will take out an adjustable rate mortgage and refinance later for a fixed rate mortgage when rates drop. I have not come to the conclusion that consumers are optimizing the money they could be paying for a mortgage, I only found that based on the single choice between an adjustable
6.00% 6.00% 4.00% 4.00%
6.00% 6.00% 4.00% 4.00% 2.00% 2.00%
Fixed minus 10 yr treasury = - 0.0026 - 0.0115 Arm Share
- 0.0396 Loan to Value Ratio + 0.00417 price (000,000) + 0.00195 term to maturity - 0.000004 Housing Starts + 0.000001 Refinance Index
- 0.000020 Mortgage originations
The regression equation is Fixed Rate = 0.107 + 0.480 10-yr treasury + 0.0150 Arm Share
- 0.0778 Loan to Value Ratio - 0.00020 price (000,000) - 0.0701 Investment to loan ratio - 0.000013 New sales (an adj) - 0.000851 Peak to trough + 0.00151 term to maturity - 0.000002 Housing Starts + 0.000000 Refinance Index Mortgage originations
- 52 Exhibit 28:
fixed rate + 1month = 0.0888 + 0.705 10-yr treasury + 0.00498 Arm Share
- 0.0994 Loan to Value Ratio - 0.00245 price (000,000) - 0.0350 Investment to loan ratio - 0.000005 New sales (an adj) + 0.000115 Peak to trough + 0.00146 term to maturity + 0.000004 Housing Starts - 0.000000 Refinance Index
- 0.000009 Mortgage originations
The regression equation is fixed rate + 1month = 0.0872 + 0.714 10-yr treasury + 0.00442 Arm Share Loan to Value Ratio - 0.00242 price (000,000) - 0.0352 Investment to loan ratio - 0.000005 New sales (an adj) + 0.00141 term to maturity + 0.000004 Housing Starts - 0.000000 Refinance Index - 0.000008 Mortgage originations
The regression equation is fixed rate + 1month = 0.0217 + 0.984 Fixed Rate + 0.00283 Arm Share - 0.0268 Loan to Value Ratio - 0.00427 price (000,000) + 0.000008 New sales (an adj) Refinance Index + 0.000012 Mortgage originations
The regression equation is Fixed rate + 2month = 0.0997 + 0.800 10-yr treasury - 0.143 Loan to Value Ratio
- 0.00804 price (000,000) - 0.0110 Investment to loan ratio + 0.000005 New sales (an adj) + 0.00144 Peak to trough + 0.00163 term to maturity + 0.000009 Housing Starts - 0.000000 Refinance Index - 0.000014 Mortgage originations
0.004 The regression equation is Fixed rate + 2month = 0.0548 + 0.927 Fixed Rate + 0.00534 Arm Share - 0.0695 Loan to Value Ratio - 0.00981 price (000,000) + 0.000017 New sales (an adj) +
0.000244 term to maturity - 0.000001 Refinance Index + 0.000011 Mortgage originations
10.0% 8.0% 6.0% 4.0% 2.0%
- 64 -