Empirical research in the United Kingdom

The background to United Kingdom research

The theoretical discussion in Chapter 2 noted the importance of the difference between the net of tax mortgage interest rate, and the rate of return on equity in a property. Muellbeur & Murphy (1997) argued that the rate of return on a building society share account seems the most appropriate forgone rate of return for the UK. Research in the US presumed that the net of tax mortgage rate was lower than the savings rate. In the UK net of tax mortgage rates tend to be higher than savings rates on basic cash accounts. This is indicated by Figure 3.1, which shows differences in gross rates, which are broadly commensurate with net of tax differences. It has generally been unlikely that a borrower would get a higher rate of return on an investment than the interest cost on borrowing (Spero 1993). However, the net of tax picture becomes increasingly complicated with the introduction of tax-free savings/investment products, and the gradual diminution and final demise of tax relief on mortgage interest rates.4 It is also complicated for earlier periods when, at least up to 1991, interest on savings accounts was taxed at a lower rate (composite tax) than earned income. So the difference in mortgage and savings rates in the UK has been subject to some instability over time.

It is possible that for some UK borrowers there was a fairly consistent expectation that the rate of return on the forgone investment would be higher than the net mortgage interest rate costs. This was possibly so for the UK endowment mortgage choice, a case to be analysed below. Choosing the appropriate opportunity cost of equity in property can be critical to expected mortgage choices, either directly or indirectly through the

Date

Figure 3.1 United Kingdom, Mortgage and Savings Rates (1984 to 2000). Source: Economic Trends.

Date

Figure 3.1 United Kingdom, Mortgage and Savings Rates (1984 to 2000). Source: Economic Trends.

effects upon housing demand. For example, Miles (1992) notes that using an historical mean return of 8% on equities as the opportunity cost of equity can lead to negative user costs of owner occupation. Understanding differences in household behaviour in the mortgage market requires a careful consideration of the appropriate opportunity cost of equity in a property. The mortgage choices facing UK households (that is the endowment) provides a unique opportunity to identify a group of borrowers for whom the opportunity cost of equity can be broadly identified.

Much empirical research in the UK has been time series based, and often primarily interested in the impact of mortgage market rationing on the demand for housing, and the implications of the amelioration or removal of credit rationing. Thus mortgage demand has been represented directly in housing demand equations, usually in the form of a stock measure (aggregate number of mortgages) or a net change in the stock of debt (Hendry 1984; Wilcox 1985; Dicks 1989, 1990; Hall & Urwin 1989; Meen 1990). Leece (1995a) estimated a cross-section analysis of mortgage demand under credit rationing and financial deregulation. More recent research has used pooled cross-section/time series data and has dissagre-gated mortgage demand by mortgage type, that is by the endowment/ repayment mortgage choice (Leece 2000b; Devereaux & Lanot 2003). Pooled cross-section/time series data allows the analysis of both interest cost effects and variations in household behaviour.

One important feature of the UK mortgage market has been the co-existance of two mortgage instruments, each with different taxation implications (see Devereaux & Lanot 2003). These are the standard annuity (repayment) mortgage and the endowment. The endowment is a saving scheme based on an interest only mortgage. Tax relief on the interest only payments means that the endowment would secure the greatest tax benefits over time. Also, under favourable interest rate regimes there can be an investment surplus in excess of the outstanding mortgage balance. The choice of an endowment provides a convenient backdrop to understanding mortgage demand in the UK economy. In particular, Leece (2000b) has used this choice to suggest a separating equilibrium in the mortgage market, a concept to be introduced in Chapter 4. Meanwhile, the UK data provides an excellent opportunity to estimate the Brueckner model of mortgage demand presented in Chapter 2.

An estimation of Brueckner's basic results of mortgage demand

There are few studies of mortgage demand for the UK. This section presents the estimates for a basic mortgage demand equation, with expectations based upon the uncertainty case, reflecting the theoretical arguments presented by Brueckner (1994a). The key variables discussed in Chapter 2 were the discrepancy between savings and borrowing rates, the level of housing expenditure, income and initial wealth. The theory also indicated the importance of time preference and attitudes to risk. The latter factors proved to be influential arguments when the net of tax mortgage interest rate was greater than the net of tax savings rate, or when there was uncertainty and the rate of return on savings was higher than the net of tax mortgage rate. In these cases households would not necessarily attempt to maximise their mortgage debt, but would limit the extent of their gearing according to their individual preferences. In this estimation it is argued that the mortgage interest rate was lower than the expected forgone rate of return on savings/investment.5

Empirical counterparts to the key theoretical variables are not always available, and must be proxied by other measures. In the case of this study we have no data on initial wealth. The analysis adopts the usual practice of using personal and household characteristics to represent time preference and attitudes to risk. The empirical specification controls for the endogeneity of housing expenditure by estimating a two stage least squares model. The sample is pooled cross-section and time series data (1991-1994), and is drawn from the British Household Panel Survey (BHPS). There is a sample of 508 borrowers, all of whom are endowment mortgage holders. This latter feature is important because the choice of an interest only endowment mortgage implies an expectation that the rate of return on the endowment fund will exceed the net of tax mortgage interest rate6 (see Leece 1995b).

The choice of time period in this study is important. This is a period of comparative accessibility for the investigator. Subsequently, the mortgage market became more complex with an increasing variety of mortgage choices, and more and more heavy discounting of mortgage debt. Savings and investment opportunities also became increasingly differentiated, with tax free savings products. Tax relief on mortgage interest payments was gradually removed, finally disappearing in April 2000. Switching between providers became a major portion of new mortgage demand. Though these facets of UK financial markets are of interest in themselves, the period 1991 to 1994 is a comparatively less problematic one in which to test a basic mortgage demand model, and to include a variable representing the cost of different mortgage instruments (variable versus fixed rate).

The house price equation, estimating the instrument for housing demand, was identified by using dummy variables for regional location, and lagged values of house price inflation as a dimension of user costs. Interestingly, the measure of user cost was not statistically significant, this may not be too surprising given a period of nominal and real house price declines and the dominance of affordability considerations. Though the theoretical work on housing cycles suggested that lagged house price changes would facilitate overcoming down payment constraints (Stein 1995; Ortalo-Magne & Rady 1998, 1999, 2002), this may also be difficult to detect given the short and particular sample period.

The model does not use the Tobit for censored data that has featured in recent US work. The main reason for this is that endowment balances are typically held to maturity, thus there are likely to be fewer observed zero balances among owner occupiers. In addition, house movers can transfer their endowment. In any case it is not possible to create a consistent sample containing zero debt observations for endowment mortgage holders having paid off their debt, that is we do not know exactly what type of mortgages they used to hold. Also, the theoretical expectation here is that endowment mortgage holders will not be debt minimisers per se. The use of a sample of endowment mortgage holders only can introduce selectivity problems, but statistical testing suggested that this was not the case. This corresponds with Devereaux & Lanot (2003) who find no evidence of this particular selectivity problem in their analysis of mortgage demand by mortgage type.

The results of estimating a basic mortgage demand equation for the UK are presented in Tables 3.1. and 3.2. The estimates are indicated for both the log of the real mortgage balance (Table 3.1) and the log of the loan-to-value ratio (Table 3.2). The focus of analysis is the demand for mortgage debt at the point of house purchase. The instrument representing the real house price at the time of purchase is seen to be statistically significant in both sets of results. Housing expenditure appears to be endogenous to decisions on mortgage size. Interestingly the sign on this variable differs according to whether the loan-to-value ratio (negative sign), or the real mortgage balance (positive sign) is used. This is consistent with the findings of US research (Ling & McGill 1998; Cho et al. 1995).

It is not absolutely clear why there should be a negative sign on the real house price in the loan-to-value ratio equation. Possibly, high levels of housing expenditure raise the risk profile of household portfolios (see Brueckner 1997) and this is compensated for by lower gearing.

Table 3.1 Mortgage demand: two stage least squares (dependent variable: log of loan-to-value ratio)

Variable

Coefficient

¿-value

Constant

0.7971

1.375

Log (age of reference person)

-0.0553

-0.993

Log (variable rate)

-0.2451

-2.002

Log (weighted premium on FRM)

-0.0001

-1.896

Log (savings rate - variable mortgage rate)

-0.2017

-1.292

Log (household income)

0.0457

1.368

Log (real house price)

-0.0456

-2.243

Children present under 5 years of age

0.0480

0.980

Male head of household

0.0155

0.526

Marital status of reference person

-0.1098

-3.470

  • R2 = 0.1808)
  • R2 = 0.1808)
Table 3.2 Mortgage demand: two stage least squares (dependent variable: real mortgage balance)

Constant

-3.8081

-6.567

Log (age of reference person)

-0.0553

-0.993

Log (variable rate)

-0.2451

-2.002

Log (weighted premium on FRM)

-0.0001

-1.896

Log (savings rate - variable mortgage interest rate)

-0.2017

-1.292

Log (household income)

0.0457

1.368

Log (real house price)

0.8440

12.133

Children present under 5 years of age

0.0480

0.980

Male head of household

0.0155

0.526

Marital status of reference person

-0.1098

-3.470

  • R2 = 0.6128)
  • R2 = 0.6128)

Alternatively, if housing wealth proxies high levels of non-housing wealth then there might be some substitution of non-housing wealth for mortgage debt (see Hendershott & Lemmon 1975; Jaffee & Rosen 1979; Ioannides 1989); though some researchers have found a positive relationship between non-housing wealth and mortgage debt (Jones 1994; Jones 1995; Cho et al. 1995). However, the elasticity of demand of gearing with respect to housing expenditure, in the UK study, is small with an elasticity of just —0.0456. So any wealth effect may not be large for this sample of mortgage holders.

Given the expectation that endowment mortgage holders have a comparatively high cost of equity then it may appear superfluous to include the discrepancy between the rate of interest on cash savings and the net of tax mortgage interest rate. Though statistically insignificant results cannot be interpreted as establishing any specific hypothesis the result is of interest. The lack of statistical significance of the savings rate variable may suggest that this is not the appropriate opportunity cost of equity, and that the expectation of a persistently high rate of return on investment, compared to the mortgage interest rate, is incorporated in the selectivity of the sample.

Though the lack of statistical significance of income is disappointing, interest rate effects are significant. Higher net of tax mortgage interest rates reduce mortgage demand.7 The elasticity of demand with respect to the net of tax mortgage interest rate is not high in either set of results, being 0.2451 with respect to both dependent variables. Brueckner's theoretical analysis suggests that when the discrepancy between the rate of return on investment and the net of tax mortgage interest rate is high then the elasticity of demand for mortgage debt with respect to the net of tax mortgage interest rate will be low. Thus this result also lends some confirmation to the assumption that the expected forgone return on equity in property, for interest only mortgage holders, was comparatively high. Interestingly, Devereaux & Lanot find an elasticity of 0.20 for the mortgage demand of endowment mortgage holders covering a slightly earlier period of time (1985 to 1989).8

The model also includes a measure of the premium paid by fixed rate mortgage holders, and this is statistically significant with a negative sign. The interpretation here is that households choose between fixed rate and variable rate debt on the basis of comparative costs. The modelling of the simultaneous choice of mortgage instrument and mortgage demand will be discussed in Chapter 8. For now, note that this study found these costs to be exogenously determined. The premium used is that on five-year fixed rate debt. However, a large number of borrowers would pay smaller premiums not identified by the data (that is one-, two-and three-year fixed rates). In this case multiplying the five-year premium by the probability of take up, estimated by a discrete choice model, gives a proxy for actual costs paid. Thus borrowers with a high likelihood of take up are assumed to be willing to pay more.9

In the uncertainty case the borrower's degree of impatience and attitudes to risk can influence mortgage demand and so estimates of the impact of borrower characteristics become important. Only one such characteristic, marital status, is statistically significant. US research has taken this variable to represent impatience and attitudes to risk. Thus the negative sign here can be interpreted as indicating that married households consist of a more patient and more risk averse group of individuals, the usual expectation with this variable. Other factors proxied by marital status, such as the presence of a second earner, are not consistent with the negative sign. Thus this estimate, added to the other results, provides an indication of the efficacy of Brueckner's mortgage demand model in guiding empirical specifications.

Get Out Of Debt Free

Get Out Of Debt Free

Debt is a major issue for a  lot of individuals these days. The issue is, even if they know they want to get out of it, they have a difficult time figuring out how to begin. Now, there isn’t one way to get out of debt, and the most beneficial program ought to be tailored to each person’s individual state of affairs.

Get My Free Ebook


Post a comment