Mortgage Refinancing Activity: An Explanation [1990 ... - Springer Link

J Real Estate Finan Econ (2006) 33: 75–86 DOI 10.1007/s11146-006-8275-4

Mortgage Refinancing Activity: An Explanation [1990–2001] Jill L. Wetmore & Chiaku Ndu

# Springer Science + Business Media, LLC 2006

Abstract Mortgage refinancing activity reached unprecedented high levels during 1990–2001. Using GARCH to control for heteroskedasticity and separating the data into regimes to control for potential structural changes over time, we estimate a model explaining changes in mortgage refinancing activity over the period studied. We find changes in refinancing activity to be negatively related to current as well as past changes in the 30-year mortgage rate with a declining significant lag over time. Similarly, there is a significant lagged dependent variable with a declining lag. Moreover, mortgage refinancing activity is a substitute for other investments during certain regimes. These results offer evidence that home owners cash out the mortgage for other investments. The lags suggest that the process is delayed for a variety of reasons. The declining lag signals a faster response by consumers. The reasons for a faster response include a consumer perception that interest rates have Bbottomed out,’’ the presence of an increase in consumer sophistication, and improvements in technology and market coordination that facilitate and reduce the cost of the refinancing process. Keywords Banking . Interest rates . Mortgages . Mortgage prepayment . Refinancing

Introduction Home mortgage refinancing activity has grown to unprecedented levels during the years 1990–2001 and is an important economic activity. Individuals have been known to refinance their mortgages repeatedly over a short time! Since this activity is so important to the economy, the ability to forecast it is of interest to managers of lending institutions and the literature. J. L. Wetmore (*) Saginaw Valley State University, 317 Curtiss Hall, University Center, MI 48710, USA e-mail: [email protected] C. Ndu Eastern Connecticut State University, 83 Windham Street, Wilimantic, CT 06226, USA e-mail: [email protected]

76

J Real Estate Finan Econ (2006) 33: 75–86

We develop and estimate a model explaining changes in the mortgage refinancing index published by the Mortgage Bankers Association of America.1 The results of this paper will assist managers of lending institutions to forecast the direction of refinancing activity to be expected thus facilitating the planning of lending activity and other features of asset/liability management. Important information on expected changes in income streams to companies servicing mortgages is implied. Expected volatility and direction of the bond market may be predicted as well (Zuckerman, 2002).

Literature Search Mortgage refinancing activity is an important topic in the literature. A clear negative relationship exists between the level of refinancing activity and the level of interest rates.2 A refinement of this relationship is that mortgage refinancing activity increases when the yield curve is steep (Kau and Keenan, 1995 and Abrahams, 1997). From the view of the individual, refinancing the mortgage is appropriate when the prevailing rate is lower than the contract rate.3 Similarly, using the analogy of a mortgage as an option, several authors posit that a mortgage is an option and the mortgagor will refinance to reduce the market value of the loan below the call price. That is, as interest rates decline below the contract rate, the cash flows to pay off the mortgage are discounted at a lower rate and the value of the mortgage increases to the financial institution or investor owning the mortgage. At this time, mortgagors will find it financially feasible to pay off the mortgage and refinance. See, for example, Chen and Ling (1989), Schwartz and Torous (1989) and Yang and Maris (1993). Refinancing behavior is strongly influenced by individual borrower and property characteristics.4 For example, a decline in housing prices or individual credit ratings will preclude refinancing activity. Refinancing activity levels depend on the sophistication and credit rating of the mortgagor and changes in housing values in the area.5 Individuals who understand the level of cash savings resulting from mortgage refinancing and who are familiar with and unafraid of the process are more likely to engage in mortgage refinancing. In support of the concept of individual and property differences affecting mortgage refinancing activity, some authors argue that a Bburn out’’ rate exists. That is, mortgages are refinanced early (and often) in their lives and as mortgages become Bolder’’ are less likely to be refinanced (Stanton, 1995 and Bennett et al., 1999). Reasons for this Bburn out’’ may include the following: the mortgagor plans 1

This index will serve as a proxy for refinancing activity.

2

See, for example, Hendershott and Van Order (1988); G-Yohannes (1988); Chen and Ling (1989); Stone and Zissu (1990); McConnell and Singh (1994); Abrahams (1997); Archer et al. (1997b); Bennett et al. (1999); Brady et al. (2000); and Harding (2000). 3

See Chen and Ling (1989); Stone and Zissu (1990); Bennett et al. (1999); and Brady et al. (2000).

4

See Giliberto and Thibodeau (1989); Stanton (1995); Archer et al. (1997a,b); Caplin et al. (1997); Peristiani et al. (1997); and Harding (2000). 5 See Caplin et al. (1997); Follain and Ondrich (1997); Green and LaCour-Little (1999); Bennett et al. (2001); Stein (1995); and Chan (2001).


77

to move soon or the mortgage is about to be paid off thus transaction costs will not be amortized, the mortgagor"s credit rating is downgraded precluding refinancing, or housing values may have dropped in the area (creating an inverted situation where the mortgage is larger than the value of the house) thus the individual_s application to refinance is declined. Despite excellent reasons to refinance, many individuals delay or deny refinancing. Using option terminology, several authors find that although the mortgage option is Bin the money,’’ it may not be refinanced as indicated earlier.6 Mortgagors may choose to delay refinancing if they perceive that rates will drop lower. Discount rate and federal funds rate target changes are often used by stock market participants as a Bsignal’’ of the future direction of interest rates and expected inflation (Thorbecke and Alami, 1994). Mortgagors may use the signal to decide whether to refinance now or wait for rates to drop further. Refinancing may be delayed because the transaction costs are too high.7 However, securitization of loans reduces loan origination costs and encourages refinancing (Todd, 2001). Points represent an additional financing cost and may also delay the refinancing decision (Stone and Zissu, 1990). Points tend to be competitive and are often attached to below market rates so they are not expected to be a significant factor when developing the model. A desire to Bcash out’’ the mortgage is an important reason to refinance the loan (Giliberto and Thibodeau, 1989 and Brady et al., 2000). The house is perceived to be a part of one_s investment portfolio and the owner(mortgagor) wishes to free up the equity for other uses such as alternative investments, debt consolidation, and other purchases. If it appears that alternative investments are poised to return more than the rate on the mortgage, then it would be feasible to take the low price funds offered by the refinancing process and invest them elsewhere. Similarly, if the consumer wants to purchase an item, refinancing offers low-cost funds for this purpose. Debt consolidation is possible reason for refinancing and may facilitate income tax deductions. The Tax Reform Act of 1986 removed interest expense tax deductions on all loans but mortgage loans on primary residences. This meant that individuals have an incentive to refinance the mortgage with its deductible interest and use the funds to pay off loans with interest in nondeductible categories such as credit-card loans. As a result, tax law changes have been an important factor in the refinancing decision.8 Finally, there may be a structural change in the relationship between the change in mortgage refinancing activity and changes in other economic variables during the period studied in this paper [1990–2001]. According to Bennett et al. (2001), prepayment speeds increased in the 1990s compared to the 1980s. They posit that this increase is the result of increased consumer sophistication and knowledge of the refinancing process combined with changes in the refinancing process through

6

See, for example, Hendershott and Van Order (1988), Stone and Zissu (1990), and Stanton (1995) among others.

7 See, for example, Hendershott and Van Order (1988); G-Yohannes (1988); McConnell and Singh (1994); Abrahams (1997); Archer et al. (1997b); Tai and Przasnyski (1999), and Harding (2000). 8 See, for example, G-Yohannes (1988), Followill and Johnson (1989) Stone and Zissu (1990), and Archer et al. (1997b).

78


technological improvements and other process changes that facilitate and reduce the cost of refinancing. Since mortgage lending institutions expect to earn fees and would like to forecast the rate of refinancing, knowledge of factors explaining refinancing activity is of importance to managers of these firms. The creation and estimation of a comprehensive model assist the determination of which factors affect refinancing activity as well as the level of consistency of this relationship. The next section models refinancing behavior and discusses the methodology to measure the relationships between changes in refinancing activity and changes in other economic variables.

Model Creation, Data Collection, and Methodology Discussion of the Model We assume a refinancing mortgagor wants to reduce loan payments and free up cash from home equity for other uses. Lower interest rates could reduce payments if the maturity of the loan remains unchanged. The cash obtained from the refinancing can be used for alternative investments or purchases or to consolidate other debt under the umbrella of a mortgage on a primary residence for which interest payments remain deductible. The availability and cost of the funds should affect the level of refinancing activity. External events such as level of unemployment and current housing prices drive credit ratings and individual access to refinancing. The stock market serves as an alternative investment to a house. Other types of housing loans such as home equity loans substitute for refinancing the mortgage. Finally, refinancing activity takes place in an environment of increasing consumer sophistication and technological advances. That is, as time passes, the activity level is expected to increase in volume and speed in response to changes in economic variables. The model to be estimated and expected signs of the coefficients are shown below: Refinit ¼ t þ f ðXit Þ

ð1Þ

Where Refin = Change in log of refinancing index; t = time; and f (Xi):

Variable Xi

Definition

Expected sign

Refin(j1, j2)

Change in log of refinancing index over a one or two period lag. Change in 30-year mortgage rate current, one, or two period lag. Change in difference between 30-year T-Bond rate and 3-month T-Bill rate or term structure.

Negative

Mort(0, j1,j2) Termst

Negative Positive


Unemp House Stkmkt Primed Morts Downp

Change in unemployment rate. Change in median new housing prices. Change in S&P 500 Index. Change in difference between 30-year mortgage rate and the prime rate. Change in level of mortgage sales. Change in average size mortgage.

79

Negative Positive Negative Negative Positive Positive

As mortgage rates decline, it is expected that refinancing activity will increase because this offers the mortgagor an opportunity to reduce payments and lock in a lower rate. This result is suggested by the mirror image relationship between the refinancing index and the mortgage rates shown in Fig. 1. Individuals desiring to refinance their mortgages may not act promptly for a variety of reasons so a lag is expected in the change in refinancing activity.9 They may expect rates to decline further, be planning to move or pay off the mortgage soon, or be confused by the process. The expected sign is negative as shown in the correlation results of Table 1. The literature suggests that the steeper the yield curve (greater the difference between long-term and short-term rates), the more likely individuals will refinance mortgage loans (Kau and Keenan, 1995; and Abrahams, 1997). Therefore, a positive relationship between changes in refinancing activity and changes in the yield curve is expected. The difference between the 30-year T-Bond and the 3-month T-Bill rate is a proxy for the term-structure.10 As the unemployment rate increases, credit ratings decline, and the ability to refinance a mortgage should also decline. Therefore, there should be a negative relationship between changes in the unemployment rate and changes in the refinancing index. The price of new housing determines if individuals will have sufficient equity in their residences in order to refinance, so a positive relationship should exist between the changes in the price of new housing and changes in the level of refinancing activity.11,12 Individuals perceive their house as a part of their investment portfolio. To the extent that there is a substitution effect between investing in a house or other securities, there should be a negative relationship between changes in a market proxy stock index and changes in the refinancing index. This is, in part, consistent with the Bcash out’’ theme. If the cost of funds is an issue, there may be a substitution effect between refinancing the house and taking a home equity loan (Giliberto and Thibodeau, 1989 and Brady et al., 2000). Since the home equity loan is often priced at a rate related to the prime rate, the difference between the long-term mortgage rate and the prime rate is examined and a negative sign is expected.

9

See Chen and Ling (1989); Stone and Zissu (1990); Bennett et al. (1999); and Brady et al. (2000).

10

We thank an anonymous reviewer for this suggestion. The data are national rather than regional so the effects of regional economic changes are not shown.

11

12 See Caplin et al. (1997); Follain and Ondrich (1997); Green and LaCour-Little (1999); Bennett et al. (2001); and Chan (2001).

80


Fig. 1 The thirty-year mortgage rate versus the log of the mortgage refinancing index (1990–2001)

The availability of mortgage funding is an important concern when refinancing mortgages. The level of mortgage sales is a proxy for the liquidity of the mortgage market. It also helps to reduce transaction costs (Todd, 2001). Therefore, there should be a positive relationship between changes in the refinancing index and changes in the level of mortgage sales. The average down-payment proxies availability of funds and credit risk. That is, a large down-payment offers a cushion against declining housing prices and the current mortgage balance is more likely to be less than the current market value of the house. This situation increases the probability that the refinancing application will be approved. Therefore, a positive relationship should exist between changes in the refinancing activity and changes in the average down payment. Finally, there is a trend of increased activity in mortgage refinancing as awareness of the advantage_s increase and technological and other market improvements facilitate the process (Bennett et al., 2001). To the extent that there is a connection between the current refinancing rate and earlier rates or a delay in the process, we include a first and second level dependent variable lag in the regression equation.13 As time passes, we expect the significant lags to decline. Data Collection Weekly mortgage refinancing indices are collected from Mortgage Bankers of America. Data for the 1990–2001 period are collected from the Appraisal Today web site and the Mortgage Bankers of America web site.14 Prime-rate, 3-month T-Bill, and 30-year T-Bond rates are taken from the Federal Reserve Board of Governors web site.15 Data on the unemployment rate are taken from the Bureau of Labor Statistics Data website. Thirty-year mortgage rate data

13

We tested for lags in the change in refinancing and change in mortgage rates and the two-period lag shown in the results is the maximum lag showing significant results.

14 15

The data from both web sites originates from the same source.

At the time of the writing of this paper, the web site addresses are as follows: http:// www.appraisaltoday.com/mbaold.htm (June 11, 2002), http://www.mbaa.org/news/weekly.app.html (June 11, 2002), http://www.federalreserve.gov/releases/h15/data.htm (June 11, 2002), http:// stats.bls.gov (June 11, 2002) http://www.Freddiemac.com (June 11, 2002).

1.00 j0.015 0.0087 j0.034 0.058 0.056 j0.066 0.00017

0.020

0.016

j0.039

0.058

1.00 0.0074 j0.0046 j0.032 j0.080 j0.10 j0.23 j0.33 j0.17

j0.067

j0.052

j0.090

j0.067

Refin Unemp Morts House Stkmkt Termst Primed Mort Mort (j1) Mort (j2) Refin (j1) Refin (j2) Downp

Unemp

Refin

622 Obs.

0.11

j0.014

j0.00049

0.0099

1.00 0.11 j0.0064 0.029 j0.040 j0.010 0.037

Morts

0.41

j0.023

0.0044

0.074

1.00 j0.0026 0.0045 j0.013 j0.00042 0.0093

House

0.0047

0.033

0.0028

j0.019

1.00 0.0048 j0.017 j0.039 j0.0056

Stkmkt

0.054

0.080

j0.084

j0.043

1.00 0.1970 0.14 0.34

Termst

j0.020

0.038

j0.037

0.0012

1.00 0.70 j0.028

Primed

Table 1 Correlations between variables used in mortgage refinancing regression, 1900–2001

j0.0022

0.072

j0.023

0.040

1.00 0.036

Mort

0.063

j0.025

j0.34

0.038

1.00

Mort (j1)

0.080

j0.033

j0.17

1.00

Mort (j2)

j0.052

1.00

1.00

Refin (j1)

j0.0048

Refin (j2)

1.00

Downp

J Real Estate Finan Econ (2006) 33: 75–86 81

82


are taken from the FreddieMac web site. Measures of the S&P 500 index are taken from the Daily Stock Price Guide published by Standard and Poors. Weighted average housing prices of newly built homes compiled by the Federal Housing Finance Board and the Federal Deposit Insurance Corporation, the amounts of the mortgage, and the level of mortgage sales on the secondary market are collected from the Federal Reserve Bulletin, (various issues).16 Methodology As can be seen in Fig. 1, mortgage refinancing activity appears as an increasing cycle over time. This suggests that regression results may have problems with stationarity or heteroskedasticity. To improve the stationarity of the results and mitigate heteroskedasticity, the first differences of the variables are used.17 To test for heteroskedasticity, the White test is used and the results are significant (White, 1980).18 To correct for this problem, a GARCH estimation is used (Bollerslev et al., 1992).19 The literature suggests that structural breaks in regression coefficients may occur and should be examined (Kane and Unal, 1988). The increasing cyclical pattern over time shown earlier may be caused by changes in consumer behavior thus affecting the regression coefficients (Bennett et al., 2001). Moreover, as we enter a new cycle, consumer behavior may change. To address this issue, the data are divided into three regimes. Each regime change is chosen to correspond with a low point in the cycle after a substantial increase. Chow test results show structural breaks in the regression coefficients so regimes exist.20 To locate the precise point of the break, a maximum likelihood test is implemented. The regression results are shown for the entire period of 1990–2001 and by regime. We use a two-period lagged dependent variable and a two-period long-term mortgage rate lag to explain lags in the decision and implementation processes of refinancing. (Flannery and James, 1984).21 Multicollinearity may be a problem. For example, a correlation test of the variables shown in Table 1 indicates that the 30-year mortgage rate is highly correlated with the prime rate less 30-year mortgage rate which may signal a possible problem of multicollinearity. To determine if multicollinearity is a problem, variables generating insignificant coefficients are dropped from the equation and the regression rerun to determine if the variable had significant explanatory power 16

Monthly data are assumed to be the same for each week during the month. Monthly data include the unemployment rate, levels of mortgage sales, and new house prices. The other data are available on a weekly basis.

17

Unit root tests for the log of the refinancing rate and mortgage rates indicate that first the first difference of the variables needs to be used: Refinancing rate ADF = j3.07:criticalj2.57, PP = j17.60. Mortgage ADF = j2.59: j2.57, PP = j9.77. White test # 2 results are as follows: 135.29 for the entire sample, 95.22 for the first regime, 124.24 for the second regime, 0 for the third regime. Results are available from the authors by request.

18

The GARCH process is stable. In all regressions, the ! parameters are positive and total less than one.

19

20

The Chow test results are F = 2.90 with a critical F = 1.75.

21

See endnote 4. Also, lags were only significant at two periods.


83

albeit an insignificant coefficient. We find evidence of multicollinearity to appear only during the third regime OLS22 results. As a result, ridge regression results are computed and shown for the third regime OLS results. As a Bpreview’’ of expected results, changes in the refinancing index should be most closely related to changes in the mortgage rate, lags in the mortgage rate, lags in the dependent variable, the stock index, the term-structure, and the prime rate30-year mortgage rate. The results follow in the next section.

Results The results are shown in Table 2 and are computed both for the entire period 1990– 2001 and for each regime. For purposes of comparison, both the ordinary least squares (OLS) and GARCH results are shown together. The results for the final regime June 2000–December 2001 include a Ridge regression as well. The results for the entire period 1990–2001 are similar regardless of whether GARCH or OLS is used. The change in the refinancing index is negatively related to a one and two-period lag in the variable. This suggests a continual trend and volatility in the refinancing index. The results also suggest that individuals do not refinance immediately but are inclined to delay the decision. The change in the refinancing index is negatively related to the change in the stock market index indicating that investing in a house is a substitute for investing in the stock market. This result is consistent with the Bcash out’’ motivation of Giliberto and Thibodeau (1989) and Brady et al. (2000). The change in the refinancing index is negatively related to current, one, and twoperiod lags in changes in the mortgage rates. The negative relationship shows a desire by mortgagors to lock in a lower rate. It is also consistent with the results of Hendershott and Van Order (1988), G-Yohannes (1988), Chen and Ling (1989), Stone and Zissu (1990), McConnell and Singh (1994), Abrahams (1997), Archer et al. (1997b), Bennett et al. (1999), Brady et al. (2000) and Harding (2000). The lag suggests a delay in the decision process or the administration process which is consistent with the results of Hendershott and Van Order (1988), Stone and Zissu (1990), and Stanton (1995). The results for the first regime—January 1, 1990 to January 30, 1995 are similar whether GARCH or OLS are used. The first regime results are somewhat different from those of the entire twelve-year period. The lag of the dependent variable is significant at two-periods but not at one-period. This suggests a longer period of volatility and trend in the index than over the entire sample. The relationship between the refinancing rate and mortgage rates is negative and significant for the current, one, and two-period lags. These results suggest that individuals delay the decision process for a variety of reasons and there may be a lack of customer sophistication and technology advancement to encourage refinancing activity. The results for the second regime—January 31, 1995 to June 25, 2000 again are similar regardless of whether GARCH or OLS is used. Again, there are differences between these results and the results for the entire twelve-year period. The second22

When testing for multicollinearity, F test results are as follows: 0.19 and 0.40 for the entire sample, 2.43 and 1.33 for the first regime, 1.28 and 2.01 for the second regime, and 2.42 and 3.71 for the third regime. Regression results are available from the authors by request.

j0.026 (0.069)

j0.020 (0.056)

j0.62* (0.083)

j0.40* (0.082)

Termst

Prime

Mort

Mort (j1) Mort (j2) Downp

j0.048 (0.078) j0.60* (0.10) j0.42* (0.083) j0.21* (0.079) j1.28 (0.96)

j0.095 (0.13) j0.0066 (0.12) j0.47* (0.17) j0.59* (0.16) j0.32* (0.13) j1.81 (1.99)

j1.16 (0.87)

j0.059 (0.12) 0.067 (0.51)

0.18 j0.11 (0.073) j0.14* (.064) 0.11 (0.22)

GARCH

j2.10**(1.08)

j0.33* (0.13)

j0.58* (0.13)

j0.55 (0.16)

0.079 (0.11)

j0.079 (0.14)

j1.23 (0.76)

0.0053 (0.44)

j0.10 (0.11)

0.15 (0.18)

j0.14* (.062)

0.19 j0.11 (0.064)

OLS

January 1, 1990–January 30, 1995 242 obs.

* is significant at 5% level. ** is significant at 10% level.

Standard error is in parentheses

j1.50 (1.06)

j0.20* (0.071)

j0.79* (0.33)

Morts

Stkmk

j0.12* (0.040) j0.034 (0.11) 0.0051 (0.061) j0.031 (0.37) j0.86* (0.35) 0.028 (0.076)

j0.084** (0.040) j0.082 (0.13)

House

0.16 0.16* (0.044)

OLS

0.16 j0.17* (0.049)

GARCH

Entire period 622 obs.

j0.0096 (0.064) 0.29 (0.35)

Adj R2 Refin (j1) Refin (j2) Unemp

Var.

0.25 (2.20)

j0.12 (0.12)

j0.14 (0.13)

j0.77* (0.13)

0.044 (087)

j0.15 (0.13)

j1.07*(0.47)

0.19 (0.69)

0.065 (0.12)

j0.049 (0.062) j0.22 (0.25)

0.18 j0.25*(0.074)

GARCH

1.23 (2.26)

j0.14 (0.11)

j0.63* (0.19) j0.12 (0.13)

j0.14 (0.16)

j0.11 (0.13)

0.43 (1.46)

j0.044 (0.78) j1.30* (.48)

j4.54 (7.33)

j0.019 (0.082) j0.018 (0.093) j0.38* (0.17) j0.80* (0.16) j0.18 (0.20)

0.49 (0.34)

j0.18 (0.15)

j0.10 (0.090) 0.024 (0.21)

0.13 j0.23 (0.17)

GARCH

1.60 (7.50)

j0.85* (0.19) j0.17 (.22)

j0.35 (0.23)

0.21** (0.11) j0.16 (0.13)

0.72 (0.56)

0.74 (1.71)

j0.077 (0.11) 0.00083 (0.23) j0.21 (.16)

0.33 j0.14 (.12)

OLS

1.76 (7.50)

j0.17 (0.13) j0.35 (0.23) j0.87* (0.19) 0.11 (0.22)

0.21**(0.11)

0.71 (0.56)

j0.075 (0.11) 0.0048 (0.23) j0.020 (0.16) 0.80 (1.71)

0.12 (0.12)

Ridge

June 26, 2000–December 31, 2001 79 obs.

0.059 (0.081)

0.19 j0.19* (0.059) j0.088 (0.058) j0.17 (0.18)

OLS

January 31, 1995–June 25, 2000 301 obs.

Table 2 Regression results showing explanatory factors of the change in refinancing index of consumer mortgage loans: 1990–2001 weekly results

84 J Real Estate Finan Econ (2006) 33: 75–86


85

period results show a negative one-period lag in the dependent variable. This suggests a reduction in the trend and possibly a faster response to changes in the independent variables consistent with Bennett et al. (2001). The refinancing rate is negatively related to changes in stock returns indicating a strong substitution effect between investing in the house or the stock market consistent with the Bcash out’’ argument of Giliberto and Thibodeau, 1989; and Brady et al. (2000). There is a significant negative relationship between changes in refinancing activity and changes in current mortgage rates only. Mortgagors are moving more quickly with the decision of refinancing or are more sensitive to the idea of refinancing (Bennett et al., 2001). Individuals may be reacting faster because they think rates will not drop much lower. Or, mortgagors may be more sophisticated regarding the process of refinancing. At this time too, mortgage institutions are developing more efficient ways of processing refinancing activity thus reducing administration time. The results for the third regime—June 26, 2000 to December 31, 2001 are different depending on whether the GARCH or OLS methodology is used. In the GARCH results, the change in refinancing rate is negatively related only to the immediate and one-week lag of the change of mortgage rate. This suggests a faster response to the possibility of refinancing (Bennett et al., 2001) albeit slower than that of the preceding regime. The OLS and Ridge regression results show a negative relationship at a one-period lag. There is no lag in the dependent variable. The presence of multicollinearity in the OLS results suggests that perhaps the model is not effectively measuring changes in refinancing activity and needs revision. This could be the result of a transitional phase in mortgage refinancing.

Conclusions The results show that the economic variables with the strongest ties to changes in mortgage refinancing activity are changes in mortgage rates (current and lagged) and a lagged dependent variable. The house that is financed by the mortgage represents a portion of the individual’s investment portfolio. The regression results change over time suggesting that the original model may need to be modified to reflect the presence of new explanatory factors of mortgage refinancing activity. Refinancing activity is showing a faster response to changes in mortgage rates. The significant lag in the dependent variable and mortgage rate coefficients are declining. This is consistent with the Bennett et al. (2001) conclusion that consumers are becoming more sophisticated and mortgage markets more efficient due to technological and organizational changes. The results also suggest that rates may not be expected to drop further.

References Abrahams SW (1997, June) The new view in mortgage prepaymenst: insight from analysis at the loan-by-loan level. Fixed Income J pp 8–21. Archer WR, Ling DC, McGill GA (1997a, June) The effect of income and collateral constraints on residential mortgage terminations. Reg Sci Urban Econ 26:235–261. Archer WR, Ling DC, McGill GA (1997b) Demographic versus option-driven mortgage terminations. J Hous Econ 6:137–163

86


Bennett P, Keane F, Mosser PC (1999, March) Mortgage refinancing and the concentration of mortgage coupons. Current Issues in Economics and Finance 5:1–6 Bennett P, Peach R, Peristiani S (2001, November) Structural changes in the mortgage market and the propensity to refinance. J Money, Credit Bank 33:955–975 Bollerslev T, Chou RY, Kroner KF (1992) ARCH modeling in finance: a review of the theory and empirical evidence. J Econom 52:5–59 Brady PJ, Canner GB, Maki DM (2000, July) The effects of recent mortgage refinancing. Fed Reserve Bull pp441–450 Caplin A, Freeman C, Tracy J (1997) Collateral damage: refinancing constraints and regional recessions. J Money, Credit Bank 29:496–516 (November, part 1) Chan S (2001) Spatial lock-in: do falling house prices constrain residential mobility? J Urban Econ 49:567–586 Chen AH, Ling DC (1989) Optimal mortgage refinancing with stochastic interest rates. Areua Journal 17:278–300 (Fall) Flannery MJ, James CM (1984, September) The effect of interest rate changes on the common stock returns of financial institutions. J Finance 39:1141–1153 Follain JR, Ondrich J (1997) Ruthless prepayment? evidence from multifamily mortgages. J Urban Econ 41:78–101 Followill RA, Johnson LL (1989, April) Taxes and mortgage refinancing. Appraisal J pp197–206 Giliberto MS, Thibodeau TG (1989) Modeling conventional residential mortgage refinancings. J Real Estate Finance Econ 2:285–299 Green RK, LaCour-Little M (1999) Some truths about ostriches: who doesn"t prepay their mortgages and why they don"t. J Hous Econ 8, 233–248. G-Yohannes A (1988) Mortgage refinancing. J Consum Aff 22:85–96 (Summer) Harding JP (2000) Mortgage valuation with optimal intertemporal refinancing strategies. J Hous Econ 9:233–266 Hendershott PH, Van Order R (1988) Pricing mortgages: an interpretation of the models and results. J Financ Serv Res 1:19–56 (September) Kane EJ, Unal H (1988, June) Change in market assessments of deposit institution riskiness. J Financ Serv Res 1:207–229 Kau JB, Keenan DC (1995) An overview of option theoretic pricing of mortgages. J Hous Res 6(2):217–244 McConnell JJ, Singh M (1994, July) Rational prepayments and the valuation of collateralized mortgage obligations. J Finance 49:891–921 Peristiani S, Bennet P, Monsen G, Peach R, Raiff J (1997, July) Credit, equity, and mortgage refinancings. Federal Reserve Bank of New York Economic Policy Review 3:83–99 Schwartz ES, Torous WN (1989, June) Valuing commercial mortgages: an empirical investigation of the contingent-claims approach to pricing risky debt. J Finance 44:375–392 Stanton R (1995) Rational prepayment and the valuation of mortgage-backed securities. Rev Financ Stud 8:677–708 (Autumn) Stein JC (1995, May) Prices and trading volume in the housing market: a model with down-payment effects. Q J Econ 110:379–406 Stone CA, Zissu AM (1990) Choosing a discount point/contract rate combination. J Real Estate Finance Econ 3:283–293 Tai LS, Przasnyski ZH (1999) Evaluating refinancing strategies for commercial real estate. Real Estate Finance 16(2):59–69 (Summer) Thorbecke W, Alami T (1994) The effect of changes in the federal funds rate on stock prices in the 1970s. J Econ Bus 46:13–20 Todd S (2001) The effects of securitization on consumer mortgage costs. Real Estate Econ 29(1):29– 54 White H (1980, May) A heteroskedasticity consistent covariance matrix and a direct test for heteroskedasticity. Econometrica 48:817–838 Yang TLT, Maris BA (1993) Mortgage refinancing with asymmetric information. J Am Real Estate Urban Econ Assoc 21(4):481–511 Zuckerman G (2002, August 8) Refinancing rocks bond market. Wall Street J, p. C1