The choice of mortgage instrument in the United Kingdom
Many of the questions explored in the US research are relevant to the UK mortgage market. In the UK the innovation was the fixed rate mortgage (fixed rate = 1) which became popular in the early 1990s (see Figure 8.2). Mirroring US research we can ask several pertinent questions. What is the impact of the FRM-VRM interest rate differential on the choice of a fixed rate mortgage? Do wealth and personal characteristics influence mortgage choice? Are the choice of mortgage instrument and mortgage/housing demand simultaneously determined? The UK also exhibits marked heterogeneity of mortgage products which has implications for consumer decision making and econometric analysis of the choice of mortgage instrument. Thus we can see how far the research findings evident in US work are applicable in the UK.
The main UK research findings
There is considerably less research into the choice of mortgage instrument for the UK, compared to the US. This applies to both cross-section and time series studies. This is surprising given the importance of the mortgage market for the UK economy but less surprising given a paucity of data, at least compared to the US. UK research has focused upon the endowment/repayment mortgage choice (Leece 1995b, 2000b) with more recent work considering the FRM/VRM choice (Leece 2000a, 2001a), and the possible simultaneous determination of mortgage/housing demand with the choice of mortgage instrument (Leece 2001a).
The choice between a repayment mortgage and an endowment mortgage is peculiar to the UK, but given that it is a possible savings/portfolio decision, the results of empirical research may have wider implications for understanding household behaviour. The repayment mortgage is an annuity mortgage with a constant payment consisting of capital and interest. The endowment mortgage is an interest only mortgage with contractual savings in a diversified portfolio of assets, making a single (balloon) payment on maturity. The endowment has the added advantage that maintaining the principal until the debt is paid off maximises any tax relief on the interest payments.
Given an interest only mortgage, then the savings vehicle used to eventually pay off the debt on maturity should reflect the opportunity cost of equity in the property, that is if a repayment mortgage is adopted then the return on this alternative asset is forgone. Thus the choice of an endowment mortgage is indicative of a portfolio decision (see Plaut 1986; Leece 1995b, 2000b). There is an opportunity here for testing the amortisation models presented in Chapter 4 of this book where an interest only mortgage represented zero amortisation (see Plaut 1986).
Using a sample of mortgages taken from the 1986 Family Expenditure Survey, Leece (1995b) estimated a probit model with correction for any selectivity bias arising out of tenure choice. Econometric investigation of the choice between a repayment and an endowment mortgage did not indicate any significant portfolio influences upon the decision (Leece 1995b). The main influences on the choice of the endowment mortgage were income and the nominal mortgage interest rate, suggesting the importance of affordability and cash flow. Thus endowments were popular when they were comparatively cheaper. These arguments are compatible with those models that emphasised borrower impatience and the 'tilt' that were presented in Chapter 7 (e.g. Brueckner 1993).
The sale of endowments in the UK has been at the centre of some controversy, with suggestions that third party originations (mortgage brokers rewarded by commission on sales) have led to widespread 'mis-selling'. The 'mis-selling' relates to the contention that consumers were not made aware of the risk that endowment funds might not grow sufficiently to eventually pay off the mortgage debt. The appropriateness of using a risky investment vehicle for this purpose has also been questioned. A low interest rate environment and depressed stock market returns in recent years have led to widespread endowment shortfalls. The analysis of so called 'mis-selling' is a linguistic and analytical minefield and an issue that has not really been explored in the mortgage analysis literature (for an exception see Leece 2000c), though there is US research on agency problems and third party originations that is relevant to this issue (see LaCour Little & Chun 1999; Alexander et al. 2002). Econometric analysis of the endowment/repayment choice has assumed that borrowers had expectations of a rate of return in excess of the net of tax mortgage rate, even if endowments were 'mis-sold' (Leece 1995b).
There is some UK research into the choice between a fixed and a variable rate mortgage (Leece 2000a, 2001a). The results suggested that UK borrowers took a short-period view, basing their choice upon expected movements in the variable mortgage interest rate. The age of the head of household was the only statistically significant personal characteristic, with older borrowers having a greater likelihood of choosing variable rate debt.8 The use of a limited amount of wealth data produced no statistically significant effects on the choice of mortgage instrument, for either the level of wealth or a measure of liquidity (Leece 2001a). This was also true for a measure of the covariance between interest rates and income though the research did involve the use of a fairly short panel, just five years.
The distinctive aspect of the UK research is the lack of statistical significance on the FRM-VRM differential. Thus an econometric model generally specified along North American lines did not produce the expected statistical results. There was evidence of regressive interest rate expectations with consumers more reluctant to lock into fixed rate debt at high mortgage interest rates.9 Borrowers were also more willing to adopt fixed rate mortgage debt the greater the discrepancy between the mortgage rate and the base rate of interest. The latter variable was a control for the stickiness of mortgage rates, which gives even variable rate debt some element of, albeit uncertain, fixity. Thus this analysis emphasised the importance of the loan rate function as discussed in Chapter 7.
Model 3 in Table 8.1 reports the results of estimating an econometric model of the choice of mortgage instrument using UK data. The sample of mortgage holders was drawn from the British Household Panel Survey, covering the years 1991-1994. The estimates confirm the lack of statistical significance of personal characteristics (e.g. age) and the negative sign on the absolute level of the mortgage interest rate, that is regressive expectations. However, such rules and heuristics may be contingent. For example, the post-1995 behaviour of UK borrowers appears different. This is illustrated in Figure 8.3 which plots the proportion of UK fixed rate debt against the FRM-VRM differential for 1996-2000.
The relationship in Figure 8.3 is quite distinct and suggests that households respond negatively to larger FRM-VRM differentials. However, the salient point is the fact that the FRM-VRM premium over a range of values is negative (covering 1997 (4) to 1999 (2)), reflecting the inverse yield curve obtaining at the time. It is perhaps not surprising that the rule based upon regression to the mean might be obscured, or no longer pertinent, when fixed rate mortgages are selling at a discount to variable rate debt. For example, even with a lower expectation of an interest rate rise a negative premium on the FRM would induce increased take up of that type of mortgage contract.
Table 8.1 Comparison of treatments of the misclassification problem (2).
Model 3
Probit with jack-knife
Model 1 Model 2 correction (excluding
Probit with correction for Probit incorporating points with high misclassification misclassification relative influence)
Model 3
Probit with jack-knife
Model 1 Model 2 correction (excluding
Probit with correction for Probit incorporating points with high misclassification misclassification relative influence)
|
Coefficient |
S.E. |
Coefficient |
S.E |
Coefficient |
S.E | |
|
Constant |
21.5024 |
8.3082 |
11.0669 |
2.5117 |
20.3578 |
3.8913 |
|
Age |
-6.4592 |
2.3851 |
-3.5329 |
0.8201 |
-6.4583 |
1.1984 |
|
Variable rate |
-1.5857 |
0.5783 |
-0.7641 |
0.1599 |
-1.3875 |
0.2564 |
|
Gap |
-1.3990 |
0.5797 |
-0.6319 |
0.2679 |
-0.9332 |
0.3065 |
|
Lincome |
0.1015 |
0.2685 |
-0.0151 |
0.1394 |
-0.0441 |
0.1668 |
|
Trend |
-0.3385 |
0.1849 |
-0.1522 |
0.07152 |
-0.3256 |
0.1118 |
|
Corriv |
-0.1849 |
0.3032 |
-0.7373 |
0.1689 |
-0.0899 |
0.2145 |
|
Had |
0.4736 |
0.5068 |
0.2355 |
0.2619 |
0.2828 |
0.3062 |
|
Alpha |
0.1285 |
0.0490 |
- |
- |
- | |
|
Log |
-165.2252 |
-167.1500 |
-96.77313 | |||
|
likelihood |
The table shows the coefficient estimates and standard errors for three different ways of dealing with classification problems in a dependent variable, including estimates that make no attempt to correct for this problem. The results shown as Model 1 are derived from a maximum likelihood technique that estimates the degree of classification error and controls for this. Model2 is estimated on a sample that is not corrected for classification error. Model 3 is estimated using a sample where possibly misclassified observations are excluded.
The table shows the coefficient estimates and standard errors for three different ways of dealing with classification problems in a dependent variable, including estimates that make no attempt to correct for this problem. The results shown as Model 1 are derived from a maximum likelihood technique that estimates the degree of classification error and controls for this. Model2 is estimated on a sample that is not corrected for classification error. Model 3 is estimated using a sample where possibly misclassified observations are excluded.
ro 50
ro 50
Premium on fixed rate mortgages Figure 8.3 United Kingdom, the proportion of fixed rate mortgages and the fixed rate premium.
Premium on fixed rate mortgages Figure 8.3 United Kingdom, the proportion of fixed rate mortgages and the fixed rate premium.
The higher interest rates, indicated by larger bubbles in Figure 8.3, correspond with high take up of FRMs. This is due to the negative or small average premium and the fact that the level of the VRM rate has a strong negative correlation with the size of the differential (—0.66). Thus high FRM take up corresponds with comparatively high rates of interest but negative premiums. This is all rather suggestive of borrowers' concern with the comparative costs of mortgage instruments.10 There is no research into the nature of household decision making in this market when negative premiums obtain and the diagram is suggestive of potentially very interesting research into the contingency of rules and heuristics, and the manner in which interest rate expectations are generated among consumers.
The mortgage choices that have been considered in this section of the book have been particular to the UK, that is the repayment versus endowment, and the choice between generally short-term fixed rate debt and a variable rate mortgage.11 However, they offer important insights into the nature of household decision making. Both types of mortgage choice appeared to be effected more by cash flow and short-term considerations rather than portfolio planning. UK households adopting a cash flow perspective have benefited from and been encouraged by intense competition, and periods of heavy discounting of mortgage debt. There has also been a proliferation of mortgage contracts from which to choose. One of the most interesting aspects of the UK research is the possibility of a misclassified dependent variable, an issue that relates to mortgage contract heterogeneity.
Post a comment