Mortgages and financial expectations: a household-level analysis.

Brown, Sarah ; Garino, Gaia ; Taylor, Karl 等

1. Introduction and Background

Arguably, the purchase of property is one of the most important investment and consumption decisions an individual or household will make over a lifetime. Furthermore, such purchases are frequently financed by mortgages. There has been a phenomenal rise in mortgage debt over recent years. For example, the growth rate in mortgage debt as a proportion of GDP in the UK between 1992 and 2002 is estimated at 21% (Catte et al. 2004). Similarly, the secured debt to income ratio has increased by 42% between 1995 and 2005 (Council of Mortgage Lenders 2006). Household mortgage debt far outweighs household unsecured debt: In the UK, average household mortgage debt in 2000 was estimated at 48,300 [pounds sterling and at 73,788 [pounds sterling] for new mortgages, as compared to 3281 [pounds sterling] for unsecured debt. Not surprisingly, the extent of household mortgage debt has been of much concern to policy makers. (1) This is especially problematic as financial assets are typically low: average annual savings in the UK in 2000 were estimated at 934 [pounds sterling]. (2) It is apparent, therefore, that savings typically provide insufficient cover for mortgage debt. Hence, the analysis of mortgage debt is important in determining the potential financial stress at the household level. As argued by Hamilton (2003), increases in household borrowing may make households vulnerable to reductions in their income or to changes in the interest rate. Consequently, understanding what factors drive the decision to acquire increasing amounts of mortgage debt and whether or not such indebtedness is sustainable are important issues for policy makers.

We contribute to the literature on household mortgage borrowing by exploring one particular influence on mortgage debt, namely the financial expectations of the individuals within the household. At the macroeconomic level, a number of studies have found that consumer expectations influence household consumption patterns (e.g., Acemoglu and Scott 1994, for the UK; and Carroll, Fuhrer, and Wilcox 1994, for the United States). Surprisingly, empirical analysis into how expectations influence consumption decisions using individual or household-level data has, however, been somewhat scarce. One reason for this may be that skepticism about the use of information derived from subjective survey data may still prevail in the economics literature (Dominitz and Manski 1997). There are, however, a number of recent studies that do exploit subjective information on income expectations, such as the work of Guiso, Jappelli, and Terlizzese (1992, 1996) and Brown et al. (2005).

We explore the relationship between mortgage debt and financial expectations from a theoretical and an empirical perspective. Our theoretical framework predicts a positive association between the expectations of individuals who are optimistic about their future financial situation and the level of mortgage debt. Our empirical analysis based on the British Household Panel Surveys, 1993-2001, supports our theoretical priors.

The British Household Panel Surveys enable us to explore the level of mortgage debt at the household level by tracking a sample of households over the nine-year period ranging from 1993 to 2001. Such an approach allows us to control for changes experienced by households as a result of events such as marriage and childbirth, which may influence the level of mortgage debt. In addition, the time period of our empirical study is particularly interesting from a macroeconomic perspective, since by 1993 the growth in annual UK GDP at constant prices had recovered to around 2.5% (Office for National Statistics) after the depths of recession in 1991, fueled by inflation and high interest rates, at which time growth was negative at -1.4%. Over the period from 1993 to 2001, GDP growth averaged approximately 2.9% per annum, peaking at 4.7% in 1994 and falling to 2% by 2001.

Our use of household-level data is particularly appropriate since, as argued by Leece (1995), the use of aggregate time-series data may mask household responses to changes in the economic environment. Leece (1995) explores mortgage demand at the household level using cross-section data from the British Family Expenditure Survey (FES) and finds that the financial deregulation that occurred during the 1980s affected mortgage demand during this period. Leece (2000) expands his earlier work and finds that mortgage demand is influenced by the type of mortgage undertaken. Cocco and Campbell (2007), who also use the FES, show that rising house prices may stimulate consumption by increasing the household's perceived wealth or by relaxing borrowing constraints. In a U.S. study on household-level data, Crook (2001) identifies the factors that explain U.S. household debt, incorporating unsecured and mortgage debt, over the period from 1990 to 1995 using data from the Survey of Consumer Finances. Income, home ownership, and family size all have a positive impact on the level of household debt. Interestingly, expectations of future changes in interest rates do not influence the level of household debt.

From the theoretical point of view, there exists a large body of literature that analyzes consumption and housing finance choice based mostly on life-cycle models with income risk and borrowing constraints. For example, in Flavin and Yamashita (2002), households maximize a function of the mean and variance of returns to their asset portfolio (inclusive of housing) conditional on the current value of a state variable represented by the ratio of the house value to net worth. In Cocco (2004), agents living for T periods maximize lifetime expected utility over housing size and non-durable consumption, with a mortgage among available financial instruments and labor income risk. Numerical solutions are provided as, generally, closed-form solutions cannot be obtained. Both papers point to an effect of the portfolio constraint imposed by housing demand and indicate that younger and less well-off investors have limited financial wealth to invest in stocks: Their net worth will be used to pay off the mortgage or to buy bonds instead. Expectations are not explicitly modeled, although their action is implicitly embodied in labor income risk.

2. Theoretical Underpinnings

Assumptions

Our stylized life-cycle model is the simplest possible and serves to illustrate our subsequent empirical analysis. We assume two discrete time periods, t = 1 and t = 2, and demonstrate a positive relationship between the level of mortgage debt undertaken by consumers and optimistic financial expectations, represented by a two-point joint distribution of incomes and house prices. At the start of period 1, risk-averse consumers earn certain income, y1, and choose, optimally, a mortgage deposit, D, toward the purchase of one durable and indivisible unit of housing, h, priced [p.sub.1]. To minimize the algebra, and without loss of generality, consumption prices (of the non-durable numeraire) in each period are normalized to 1, the safe interest rate is set to zero, and there is no housing depreciation. The utility function, U([c.sub.t], h), defined over consumption at time t and housing, is twice differentiable and strictly concave. Consumption in period 1 is given by [c.sub.1] = [y.sub.1] - D, yielding utility U([y.sub.1] - D, 1).

There is second-period uncertainty of both consumer incomes and house prices, which are jointly distributed with two possible realizations of each variable, [y.sub.2i] and [p.sub.2j], where i, j = H, L denote the high and low income and house price realizations, respectively; and where [y.sub.2H] > [y.sub.2L] and [p.sub.2H] > [p.sub.2L]. So, in period 2, there are four possible states of nature (HH, HL, LH, LL) that occur with exogenous probabilities [q.sub.HH], [q.sub.HL], [q.sub.LH], and [q.sub.LL], respectively, that sum to 1. In period 1, a competitive risk-neutral lender provides a mortgage of size ([p.sub.1] - D). The mortgage repayment in period 2 is therefore R([p.sub.1] - D), where, as mentioned, D is saved by the borrower, at an interest factor R [greater than or equal to] 1, which includes a risk premium (see Eqn. 2--implying that consumers will always save in D rather than in the safe interest-yielding asset). The distributions of second-period incomes and house prices are common knowledge to both the consumer and the lender. (3)

In order to take up the mortgage in the first period the consumer expects, on average, to be able to repay it in the second period (4) [i.e., E([y.sub.2] + [p.sub.2]) [greater than or equal to] R([p.sub.1] - D), where E([y.sub.2] + [p.sub.2]) = [q.sub.LL]([y.sub.2L] + [p.sub.2L]) + [q.sub.LH]([y.sub.2L] + [p.sub.2H]) + [q.sub.HL]([y.sub.2H] + [p.sub.2L]) + [q.sub.HH]([y.sub.2H] + [p.sub.2H]) and R is defined by the lender's equilibrium condition] (see Eqn. 1 below). However, when both income and house price realizations are low (in state LL), consumers are unable to repay their debt [i.e., [y.sub.2L] + [p.sub.2L] < R([p.sub.1] - D)]. In this case, the lender seizes all of the consumer's resources except an exogenous small amount, e > 0, yielding utility U(e, 0) to the consumer. Conversely, when second-period income is high (in states HH and HL), consumers can repay their debt regardless of the house price realization [i.e., [y.sub.2H] > R([p.sub.1] - D)]. In this case, consumption is either [c.sub.2H] = [y.sub.2H] - R([p.sub.1] - D), yielding utility U([y.sub.2H - R([p.sub.1] - D), 1) if the consumer keeps the house in the second period, or [c.sub.2Hj] = [y.sub.2H] - [p.sub.2j] - R([p.sub.1] - D), yielding utility U([y.sub.2H] + [p.sub.2j] - R([p.sub.1] - D), 0) if the consumer sells the house in the second period for a financial gain [i.e., if [p.sub.2j] > [p.sub.1], j = H, L). Finally, when second-period income is low but the house price is high (in state LH), consumers cannot repay the debt and keep the house since [y.sub.2L] < R([p.sub.1] - D). However, they can repay the debt by selling the house, with consumption [c.sub.2LH] = [y.sub.2L] + [p.sub.2H] - R([p.sub.1] - D), yielding utility U([y.sub.2L] + [p.sub.2H] - R([p.sub.1] - D), 0), so long as [y.sub.2L] + [p.sub.2H] > R([p.sub.1] - D). It is straightforward to verify that these ex post state-dependent conditions are compatible with the ex ante requirement that expected incomes and house prices should be above the debt repayment.

The Model

In equilibrium, the interest factor R is obtained from the lender's zero expected profit condition:

[p.sub.1] - D = [q.sub.LL] ([y.sub.2L] + [p.sub.2L] - e) + (1 - [q.sub.LL]) R([p.sub.1] - D). (1)

From our assumptions it is straightforward to verify that

R = 1/1 - [q.sub.LL] - [q.sub.LL] ([y.sub.2L] + [p.sub.2L] - e)/(1 - [q.sub.LL])([p.sub.1] - D) [greater than or equal to] 1. (2)

By substituting R from Equation 2, we can write the reduced form of the inequality [y.sub.2L] + [p.sub.2L] < R([p.sub.1] - D) (holding in state LL), which gives D < [p.sub.1] - ([y.sub.2L] + [p.sub.2L]) + [q.sub.LL]e, that is, the upper bound to D. Similarly, the inequality [y.sub.2H] > R([p.sub.1] - D) (holding in states HH and HL) becomes D > [p.sub.1] - [q.sub.LL]([y.sub.2L] + [p.sub.2L] - e) - (1 - [q.sub.LL])[y.sub.2H]; the inequality [y.sub.2L] + [p.sub.2H] > R([p.sub.1] - D)(holding in state LH)becomes D > [p.sub.1] - [q.sub.LL]([y.sub.2L] + [p.sub.2L] - e) - (1 - [q.sub.LL])([p.sub.2H] + [y.sub.2L]); and the inequality E([y.sub.2] + [p.sub.2]) > R([p.sub.1] - D) becomes D > [p.sub.1] - [q.sub.LL]([y.sub.2L] + [p.sub.2L] - e) - (1 - [q.sub.LL])E([y.sub.2] + [p.sub.2]). The greater right-hand side to these last three reduced-form inequalities gives the lower bound to D. The domain of the definition of D, the chosen mortgage deposit, is therefore:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (3)

where both the left-hand side and right-hand side are expressed solely in terms of exogenous variables and parameters. In this framework, Equation 3 can be interpreted as the borrowing constraint on the mortgage size ([p.sub.1] - D) and (1 - [q.sub.LL]) as the overall probability of being able to repay the loan; that is, the "optimistic" financial expectation of both parties. It is then straightforward to show that the effect of (1 - [q.sub.LL]) on the total amount of mortgage undertaken, ([p.sub.1] - D), is positive [or, equivalently, that the effect of [q.sub.LL] on ([p.sub.1] - D) is negative]. Again without loss of generality, we consider the case in which consumer preferences are such that high-income consumers always prefer to keep the house in the second period rather than sell it [i.e., U([y.sub.2H] - R([p.sub.1] - D), 1) > U([y.sub.2H] + [p.sub.2j] - R([p.sub.1] - D), 0)].

The consumer chooses the mortgage deposit, D, optimally to maximize expected lifetime utility subject to Equation 3 and the lender's zero expected profit condition:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (4)

where the expressions for consumption are defined above, 8 is a subjective discount factor, and the interest factor is defined by Equation 1 and (1 - [q.sub.LL] - [q.sub.LH] = [q.sub.HH] + [q.sub.HL]).

At an interior solution, which, given Equation 3, is ensured by strict concavity, the first-order condition is

(1 - [q.sub.LL])U' ([c.sub.1], 1) = [delta][q.sub.LH]U' ([c.sub.2LH], 0) + [delta](1 - [q.sub.LL] - [q.sub.LH]) U' ([c.sub.2H], 1). (5)

Comparative statics then give:

dD/[dq.sub.LL] =

[q.LH] [U' ([c.sub.2H], 1) - U' ([c.sub.2LH, 0)]/[[(1 - [q.sub.LL].sup.2] U" ([c.sub.1], 1) + [q.sub.LH] U" ([c.sub.2LH], 0) + (1 - [q.sub.LL] - [q.sub.LH])U" ([c.sub.2H], 1)] > 0, (6)

which is positive, since both the numerator and the denominator are negative by concavity. This implies that d([p.sub.1] - D)/[dq.sub.LL] < 0. Hence, a higher level of (1 - [q.sub.LL]) has a positive effect on mortgage debt. That is, optimistic financial expectations increase mortgage debt.

In sum, the above is a stylized model provided to inform our empirical analysis. In particular, the result encapsulated by Equation 6 is intuitive. Moreover, it is straightforward to verify that it holds also with non-zero safe interest rate and depreciation parameters, with different consumption prices in periods 1 and 2, and with consumer preferences such that the high-income consumer prefers to sell the house in period 2. (5) Our simplifying assumptions are made purely to minimize notation and to illustrate the intuition in the clearest possible way in order to motivate our subsequent empirical analysis.

3. Data and Methodology

In the remainder of the paper, we explore the empirical determinants of the amount of outstanding mortgage debt at the household level in Great Britain, focusing on the role of financial expectations. For the purposes of our empirical study, we exploit information contained in nine waves of the British Household Panel Survey (BHPS), 1993-2001. Prior to 1993, households were not asked to disclose the amount of their mortgage in the BHPS. The BHPS is a random-sample survey, carried out by the Institute for Social and Economic Research, of each adult member from a nationally representative sample of more than 5000 private households (yielding approximately 10,000 individual interviews).

In the 1993-2001 surveys, households were asked "How much is the total amount of your outstanding loans on all the property you (or your household) own, including your current home?" The answers thus provide information on the amount of outstanding mortgage debt. The defining feature of the BHPS, for the purpose of our study, is that it contains information on the total amount of mortgage debt over a relatively long time horizon, 1993-2001, at the household level as well as information relating to the expectations of household members about their future financial situation. Our sample includes households with a head of household aged between 18 and 65 years. We analyze an unbalanced panel of data such that the average number of observations per household is 3.7, with the minimum (maximum) being 1 (9). Our sample comprises 11,478 households, over a maximum of nine years, yielding a total of 42,894 observations.

To explore the relationship between the expectations of household members regarding their future financial situation and the extent of outstanding mortgage debt, we exploit responses to the following question: "Looking ahead, how do you think you will be financially a year from now, will you be: Better off; Worse off; Or about the same?" Answers to this question implicitly incorporate a synthesis of a household member's own financial outlook (e.g., pay and job security) with their expectations about the general economic environment (e.g., future interest rates, tax changes, inflation, and unemployment rates).

Response rates for heads of households are shown in Table 1A. From the responses to this question, we create a Financial Expectations Index (FEI), as in Brown et al. (2005), where individuals who answer "Worse off" to the above question are coded as 0, those who answer neither "Worse off or Better off" are coded 1, and individuals who answer "Better off" are coded as 2. Thus, the index ranks individuals according to their financial expectations from having a bleak outlook to being optimistic about their financial future. From Table 1A, it is apparent that over the period heads of households tend to be financially optimistic rather than pessimistic, which may reflect the start of the economic recovery following the recession of the early 1990s.

We also explore whether financial expectations vary over time, since it is possible that any correlation between financial expectations and mortgage debt might simply be capturing a household fixed effect rather than predictions about future income. Table 1B shows the number of times household heads are optimistic, pessimistic, or expect no income change over the time period, including the proportion of households that are always optimistic and always pessimistic. Clearly, there is variation in households' financial expectations over time, with only a small proportion of individuals always in the same category for all years and more than 50% of respondents only reporting two years (out of nine) in which financial expectations do not change.

In order to explore the relationship between mortgage debt and financial expectations, we proceed in two stages. First, we estimate a housing tenure model, since it is apparent that the housing tenure of households will influence the level of mortgage debt. The housing tenure variable is defined as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

The responses to this variable over time are shown in Table 1C. The percentage of households owning a home with a mortgage has fallen over time, while, conversely, the percentage renting privately has risen. We model housing tenure by specifying a multinomial logit (MNL) model in which the unit of observation is the head of household h at time t. In the set of explanatory variables, we include labor market status, a quadratic in age, gender, ethnicity, number of children, household size, household income, educational attainment, household savings, and investments. We also include the financial expectations of the head of household, [FEI.sub.ht], to control for differences in financial optimism across tenure types. We then use the predicted values from the MNL framework to calculate an inverse mills ratio term relating to selection into category 1 (i.e., owner occupiers with a mortgage, [htc.sub.ht] = 1), which is then included in our mortgage debt equation. (6)

Given that our focus is on the relationship between financial expectations and the level of mortgage debt, we select those households in which [htc.sub.ht] = 1, and, in the second stage of our analysis, we explore the determinants of the logarithm of the amount of outstanding mortgage debt. (7) Figure 1 illustrates the distribution of the logarithm of the amount of outstanding mortgage debt across the sample of owner occupiers with a mortgage. We estimate the following reduced-form mortgage equation:

ln ([m.sub.ht]) = [[beta]'.sub.1][X.sub.ht] + [[beta]'.sub.2][FEI.sub.ht] + [v.sub.ht], (7)

where

[v.sub.ht] = [[alpha].sub.h] + [[eta].sub.ht]. (8)

Our notation is defined as follows. The mortgage debt of household h at time t is given by [m.sub.ht] such that t = 1,..., 9, where h = 1,..., [n.sub.h]; [X.sub.ht] denotes a vector of head of household and household characteristics, including the inverse mill ratio term derived from the housing tenure model to control for potential sample selection into category [htC.sub.ht] = 1; [[alpha].sub.h] represents the "household" specific unobservable effect; and [[eta].sub.ht] is a random error term, [[eta].sub.ht] ~ IN(0, [[sigma].sup.2.sub.h]). Our theoretical framework (i.e., Eqn. 6) predicts [[beta].sub.2] > 0. We assume that [[alpha].sub.h] is IN(O, [[sigma.sup.2.sub.[alpha]) and independent of [[eta].sub.ht] and [X.sub.ht]. Hence, the correlation between the error terms of households over time is a constant given by

[rho] = corr([v.sub.tl], [v.sub.tk]) = [[sigma].sup.2.sub.[alpha]]/[[sigma].sup.2.sub.[alpha]] + [[sigma].sup.2.sub.[eta]] l [not equal to] k, (9)

where [rho] represents the proportion of the total unexplained variance in the dependent variable contributed by the panel-level variance component. Thus, the magnitude of [rho] yields information pertaining to the intra-household correlation of mortgage debt over time. As a result of issues pertaining to identification restrictions, we regard our mortgage debt equation as a reduced-form specification potentially capturing demand and supply influences.

[FIGURE 2 OMITTED]

In waves 1993 to 2001 of the BHPS, homeowners are asked: "About how much would you expect to get for your home if you sold it today?" We use the responses to this question to derive a measure of house value, which is used to weight the level of mortgage debt. Figure 2 represents the distribution of mortgage debt as a proportion of the house value. Not surprisingly, the majority of households with a mortgage do not have a mortgage value greater than the value of the house; only 1.5% of mortgagees fall into this latter group. We repeat the analysis represented by Equations 7 and 8 with the weighted level of mortgage debt as the dependent variable in order to explore the robustness of our findings.

In addition to exploring the influence of the head of household's financial expectations, we also investigate the role played by the expectations of other household members. Hence, we repeat the analysis described above and replace [FEI.sub.ht] with the sum of the financial expectations index across all household members, HFEI = [summation] over (i[member of]h)] FEI.

Although the focus of our paper lies in the role of financial expectations, we include a number of additional controls in our econometric analysis for personal and demographic characteristics in the vector [X.sub.ht]. We control for household income, highest educational qualification of the head of household, and the logarithm of total savings and investments to proxy household wealth. Demographic controls include the marital status of the head of household, the number of children (aged less than 18 years), region of residence, and household size. We also control for whether the household has an endowment or repayment mortgage and whether the household has a mortgage protection plan or structural or contents insurance. Reasons for having an additional mortgage are also included, which include an extension to the house, home improvements, car purchase, or another unspecified reason. Table 2 presents summary statistics for the variables used in our empirical analysis.

4. Results

Housing Tenure and Financial Expectations

Before focusing on the determinants of the level of mortgage debt, we will briefly comment on the characteristics that influence housing tenure. Table 3A, B report the determinants of housing tenure, where we consider the influence of individual (i.e., the head of household) and household expectations, respectively. The findings presented in Table 3A indicate that the head of household's expectations are important in influencing housing tenure. For example, a 1-point move up the financial expectations index (i.e., becoming more financially optimistic) yields just under a 1% increase in the probability that the head of household will be an owner occupier with a mortgage rather than own the property outright (the reference category), ceteris paribus. Other factors of interest are that younger heads of household are more likely to rent property, as are students, the unemployed, and those not in the labor market. For example, for heads of household not in the labor market, there is a 21.6% higher probability of renting from the council rather than owning a property outright. Higher income, savings, and investment are all associated with a lower probability of renting relative to owning a property without a mortgage. Married or cohabiting individuals and those with some education (relative to those with no educational qualifications) also have a lower probability of renting. (8)

Mortgage Debt and Financial Expectations

Turning our focus to the determinants of the level of mortgage debt, our aim is to verify whether the empirical evidence corresponds with our theoretical priors. Since we focus solely on mortgagees (i.e., [htc.sub.ht] = 1), we include an inverse mills ratio term in our set of explanatory variables to control for potential sample selection bias. (9) The determinants of the level of mortgage debt and the determinants of the proportion of mortgage debt relative to the estimated house value are shown in Table 4A, based on individual financial expectations, and in Table 4B, based on household financial expectations. Throughout the results, [rho] is large, indicating relatively high intra-household correlation of mortgage debt over time.

Our empirical findings accord with our theoretical priors in that the head of household's financial expectations index is characterized by a positive and significant estimated coefficient, indicating that the more optimistic the head of household is about their financial situation in the following year, the greater is the amount of mortgage debt, as shown in Table 4A, column 1.

Turning to the other explanatory variables, the level of outstanding mortgage debt is positively associated with the head of household's age, albeit at a decreasing rate. Other factors that have a positive and significant influence on the level of mortgage debt are household income, whether the head of household is married or cohabiting, and the educational attainment of the head of household. For example, heads of household with a degree have around a 31% higher level of mortgage debt than those with no education, ceteris paribus. Household size and having structural insurance, on the other hand, are associated with lower levels of mortgage debt. For those mortgagees with an endowment or repayment mortgage, the level of mortgage debt is significantly higher at around 6% and 7%, respectively. Having contents insurance or other types of insurance are also associated with higher levels of mortgage debt.

The second column of Table 4A reports consistent findings with the alternative dependent variable--mortgage debt as a proportion of house value. Clearly, the head of household's financial expectations index is characterized by a significant positive estimated coefficient and, hence, supports the previous findings--although the size of the estimated coefficient is marginally smaller than that for the previous dependent variable. Mortgage debt as a proportion of house value is also positively related to household income, being a male head of household, having an endowment mortgage, and contents insurance. Conversely, older heads of household, household size, the number of children, higher levels of savings and investments, having an additional mortgage for home improvements, or an extension are all negatively associated with the amount of mortgage debt relative to house value.

Table 4B presents the results from estimating Equation 7, including the sum of financial expectations of all individuals within the household. Our findings indicate that the summation of expectations within the household is characterized by a positive and significant estimated coefficient for both the amount of mortgage debt and mortgage debt as a proportion of house value. Thus, our results indicate that, even when controlling for household size, households with higher levels of financial optimism amongst their members are associated with greater levels of mortgage debt.

The coefficients on the regional dummy variables reported in Table 4A, B show that the two areas that have the lowest mortgage debt relative to London, the reference category, are the North East and Wales. Such a finding is not surprising given the relatively low house prices in these two regions. With respect to mortgage debt as a proportion of house value, all regions have a mortgage amount that is closer to the estimated value of the house than that in the London region. Although all monetary figures have been deflated, as compared to the reference category, 2001, mortgage debt relative to house value was significantly lower in the earlier years. (10,11)

Robustness Checks

We explore the robustness of our empirical findings in three ways. First, we replace the financial expectations index with dummy variables denoting financial optimism and pessimism. Second, we control for the truncation of the sample using a Tobit model rather than a sample selection correction. Finally, we distinguish between household expectations and aggregate expectations in order to further explore the issue of household fixed effects.

We replace the financial expectations index with two dummy variables for whether the head of household is financially optimistic or financially pessimistic to see how robust the results are to an alternative definition of our key variable of interest. (12) The results are shown in Table 5, panel A, where the same set of control variables is employed as in Table 4A. Both financial optimism and financial pessimism are statistically significant in influencing the level of mortgage debt and mortgage debt as a proportion of house value. A financially optimistic head of household has 2.1% (5%) higher mortgage debt (mortgage debt as a proportion of house value) than those who predict that their financial situation will stay the same (i.e., the reference category), ceteris paribus. Similarly, financial pessimism works in the opposite direction, with a financially pessimistic head of household having 3.1% (3.4%) lower mortgage debt (mortgage debt as a proportion of house value) than those who predict that their financial situation will stay the same, ceteris paribus. (13) For household financial expectations, we replace the index with four dummy variables capturing (i) whether one individual within the household is optimistic; (ii) whether one individual within the household is pessimistic; (iii) whether two or more individuals within the household are optimistic; and (iv) whether two or more individuals within the household are pessimistic. The results, which are presented in Table 5, panel B, reiterate the finding that financial optimism is associated with a higher level and proportion of mortgage debt.

We explore the robustness of our findings further by dealing with the truncation of the sample in an alternative way by, specifically, a Tobit model in which mortgage debt is truncated at zero:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (10)

The results of estimating Equation 10 are presented in Table 6, panel A, based on individual financial expectations, and in Table 6, panel B, based on household financial expectations. In Table 6, panel A, the financial expectations index is positively related to mortgage debt and mortgage debt as a proportion of house value, indicating that the more financially optimistic individuals are the higher is the level of mortgage debt. This result also holds when we consider the aggregate expectations of individuals within the household (see Table 6, panel B).

Finally, one could argue that the positive correlation found between financial expectations and mortgage debt stems from optimistic aggregate expectations about future income rather than household-specific effects. To separate the aggregate effects from the household specific effects, we explore household financial expectations relative to the average level of aggregate household expectations in each year. (14) Thus, we create an index [FEI.sub.rt], which represents the head of household's expectations relative to aggregate expectations (calculated at the mean for each year), and we also define [HFEI.sub.rt] for sum of the FEI across all household members relative to the aggregate mean expectations. (15) The mean values for [FEI.sub.rt] and [HFEI.sub.rt] are given by 1.0227 and 0.9949, respectively. Table 7 presents the results from weighting the FEI by aggregate expectations for heads of household (panel A) and for all individuals within the household (panel B). The results indicate that higher household financial expectations relative to the yearly average are positively associated with higher mortgage debt.

Mortgage Debt and Income

Finally, we compare the relative magnitude of the mortgage level as a proportion of household income across optimistic and pessimistic heads of households over the period ranging from 1993 to 2001. Figures 3 and 4 show the annual median actual and predicted mortgage level as a proportion of income, respectively, for both financially optimistic and pessimistic heads of households. The percentage growth in GDP year-on-year is also plotted on the right-hand vertical axis in each figure (United Kingdom Office of National Statistics 2007).

The predicted mortgage level is derived by estimating separate mortgage debt equations for financially optimistic and financially pessimistic heads of households. We then use these results to calculate predicted mortgage debt for each group and year. Overall the model accurately predicts the trend in mortgage level as a proportion of income over time, although it does overpredict actual year-on-year values. Clearly, growth in GDP peaked in 1994 at 4.7% and started to fall after 1997. Correspondingly, there is also evidence of the actual and predicted mortgage levels relative to income falling after 1998 for optimistic heads of households. It appears that the proportion of outstanding mortgage debt relative to household income for optimistic individuals lags the business cycle by one year, based on actual and predicted values. This is despite the Bank of England's base interest rate being at its peak in 1998, averaging 6.94%, and falling thereafter to an average of 4.96% in 2001. As such, the trends depicted in Figures 3 and 4 indicate that the level of mortgage debt may not be inversely related to the price of debt. In general, it is apparent that the mortgage debt series of optimistic heads of households lies clearly above that of pessimistic heads of households, providing further evidence indicating that financial optimism is associated with higher levels of mortgage debt at the household level.

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

5. Concluding Remarks

In this paper we have explored an issue that is extremely topical among both economists and policy makers--namely, mortgage debt at the household level. Given that the UK (along with a number of other countries) currently has high, and arguably unsustainable, rates of house price inflation and growing household debt, gaining an insight into what factors influence mortgage debt is a very important issue (Nickell 2002). Our main focus has been on the influence of financial expectations on the level of mortgage debt. To be specific, our theoretical framework predicts an intuitively positive association between optimistic financial expectations and mortgage debt. In order to test our theoretical predictions we have explored the determinants of the level of outstanding mortgage debt using data derived from nine waves of the BHPS, 1993-2001. Our findings indicate that the expectations of household members regarding their future financial situation are an important determinant of mortgage debt.

Received March 2006; accepted March 2007.

References

Acemoglu, D., and A. Scott. 1994. Consumer confidence and rational expectations: Are agents' beliefs consistent with economic theory? Economic Journal 104:1-19,

Brown, S., G. Garino, K. Taylor, and S. Wheatley Price. 2005. Debt and financial expectations: An individual and household level analysis. Economic Inquiry 43:100-20.

Carroll, C., J. Fuhrer, and D. Wilcox. 1994. Does consumer sentiment forecast household spending? If so, why? American Economic Review 84:1397-408.

Catte, Pietro, Nathalie Girouard, Robert Price, and Christophe Andre. 2004. Housing markets, wealth and the business cycle. In OECD economic outlook, IV, n. 75: 514-43.

Cocco, J. 2004. Portfolio choice in the presence of housing. Review of Financial Studies 18:535-67.

Cocco, J., and J. Y. Campbell. 2007. How do house prices affect consumption? Journal of Monetary Economics 54:591 621.

Council of Mortgage Lenders. 2006. "A trillion pounds success story." Housing Finance, April. Accessed July 2007. Available http://www.cml.org.uk/cml/publications/research.

Crook, J. 2001. The demand for household debt in the USA: Evidence from the 1995 survey of consumer finance. Applied Financial Economics 11:83-91.

Dominitz, J., and C. Manski. 1997. Using expectations data to study subjective income expectations. Journal of the American Statistical Association 92:855-67.

Flavin, M., and T. Yamashita. 2002. Owner-occupied housing and the composition of household portfolio. American Economic Review 92:345-62.

Gale, D., and M. Hellwig. 1985. Incentive compatible debt contracts: The one period problem. Review of Economic Studies 52:647-63.

Greenspan, A. 2004. Understanding household debt obligations. Remarks, Credit Union National Association Governmental Affairs Conference, Washington, DC, February 23. Available http://www.federalreserve.gov/ boardDocs/speeches/2004/20040223/default.htm.

Guiso, L., T. Jappelli, and D. Terlizzese. 1992. Earnings uncertainty and precautionary saving. Journal of Monetary Economics 30:307-37.

Guiso, L., T. Jappelli, and D. Terlizzese. 1996. Income risk, borrowing constraints and portfolio choice. American Economic Review 86:158-72.

Hamilton, R. 2003. Trends in households' aggregate secured debt. Bank of England Quarterly Bulletin Q1:271-80.

Jost, P. 1996. On the role of commitment in a principal-agent relationship with an informed principal. Journal of Economic Theory 68:510-30.

Krasa, S., and A. P. Villamil. 2000. Optimal contracts when enforcement is a decision variable. Econometrica 68:119-34.

Leece, D. 1995. Rationing, mortgage demand and the impact of financial deregulation. Oxford Bulletin of Economics and Statistics 57:43-66.

Leece, D. 2000. Choice of mortgage instrument, liquidity constraints and the demand for housing debt in the UK. Applied Economies 32:1121-32.

Mookherjee, D., and I. Png. 1989. Optimal auditing, insurance and redistribution. Quarterly Journal of Economics 104:397-415.

Nickell, S. 2002. House prices, household debt and monetary policy. Household Debt and Monetary Policy Speech Glasgow. Accessed July 2007. Available http://www.bankofengland.co.uk/speeches/speech187.pdf.

Papademos, L. 2004. The economic outlook and the ECB's monetary policy: Some key issues. Speech, Nomura Annual Euro Conference, Tokyo, November 11.

Townsend, R. 1979. Optimal contracts and competitive markets with costly state verification. Journal of Economic Theory 21:265-93.

United Kingdom Office of National Statistics. "GDP Growth." Accessed July 2007. Available http://www.statistics. gov.uk/cci/nugget.asp?id = 192.

(1) For example, the accumulation of debt has been noted by the European Central Bank (ECB), which has reported that falling interest rates have allowed households to borrow more and accumulate more debt. Consequently, household debt in the euro area has increased significantly in recent years. In 2004, it was estimated at 54% of GDP. See http:// www.ecb.int/press/key/date/2004/html/sp041111.en.html for the speech by Lucas Papademos, Vice-President of the ECB, delivered at the Nomura Annual Euro Conference, "A Challenging Future for Europe," Tokyo, November 11, 2004. In the United States, remarks on the amount of household debt relative to assets were made by the Chairman of the Federal Reserve Board Alan Greenspan, "Understanding Household Debt Obligations," at the Credit Union National Association, Governmental Affairs Conference, Washington, D.C., February 23, 2004 (Greenspan 2004).

(2) The figures for average household mortgage debt, unsecured debt, and savings are calculated from the British Household Panel Survey, 2000.

(3) Exploring cases in which the lender is not informed about the realization of the consumer's resources is beyond the scope of this paper, as it requires the incorporation of the appropriate incentive constraints into a "mortgage contract" between consumer and lender. There is a large and established theoretical literature on loan contracts with costly state verification; see, for example, Townsend (1979), Gale and Hellwig (1985), Mookherjee and Png (1989), Jost (1996), and Krasa and Villamil (2000).

(4) We are grateful to an anonymous referee for highlighting this point.

(5) Algebraic proofs of these results are available from the authors on request.

(6) Our over-identifying instruments are the following labor market status dummy variables: employed; self employed; unemployed; not in the labor market; and being a full-time student. These are intuitively appealing instruments, given that obtaining a mortgage is conditional on labor market status. Since the choice of over-identifying instruments is always a contentious issue, we have explored changes to the set of instruments as well as validity tests for our choice of instruments and different approaches to allowing for sample selection. We discuss these issues further in section 4.

(7) Zero reported mortgage debt is included as zero in our dependent variable, as there is no reported mortgage debt between zero and unity.

(8) Both individual and household financial expectations are positively associated with renting a property from the council. We have explored this further and find that a possible explanation relates to the fact that council renting is concentrated among younger individuals who tend to be more financially optimistic. If the financial expectations index is interacted with age, the marginal effect on the interaction term is negative and significant. Mortgage Debt and Financial Expectation

(9) The inverse mills ratio term has a positive estimated coefficient, indicating that its exclusion would bias our results downward. In general, the sample selection equation is well specified, with the chi-squared statistic being significant at the 1% level. We have also explored the robustness of our findings by omitting the inverse mills ratio term. Our findings with respect to the relationship between financial expectations and the level of mortgage debt are largely unchanged. To test for the validity of the instruments we test the joint significance of the labor market status variables in the sample selection equation. We find that these variables are jointly significant in all models,

supporting the use of these instruments. Secondly, we regress the residual from the mortgage equation on the over-identifying instruments. Our findings indicate an insignificant relationship between the residuals and the labor market status variables in all models, thereby further endorsing the validity of this set of over-identifying instruments.

(10) To explore the robustness of our findings, we replicated the analysis of Table 4A, B, replacing regional dummy variables with regional average house prices in each year and replacing the year dummy variables with the Bank of England base interest rate. The significant estimated coefficient of financial expectations remains across each specification, with the magnitude of the impact being largely unaffected in comparison to those reported in Table 4A, B. For example, for individual financial expectations, the estimated coefficients were 0.0242 and 0.0249 for LMORT and PMORT, respectively. For household financial expectations, the corresponding estimated coefficients were 0.0138 and 0.0094. The full results are omitted for brevity, but are available on request.

(11) We have also investigated how well the financial expectations index predicts future income. The summary statistics presented in Table 1B indicate that household expectations vary over time. When we regress future household income on lagged income, the financial expectations index, and the demographic variables used in Table 4A, the coefficient on the financial expectations index is positive and statistically significant, although it is outweighed by the coefficient on lagged household income. This result is confirmed regardless of housing tenure and for mortgagees only; the respective coefficients (t-statistics) are 0.0333 (3.64) and 0.0166 (2.45). We also regress household income growth on the financial expectations index. Once again, expectations about future income have predictive power, with estimated coefficients (t-statistics) of 0.0286 (2.79) and 0.0253 (2.02). Such findings indicate that financial expectations are not capturing a household fixed effect and that the index does help predict future income and income growth at the household level.

(12) We are grateful to an anonymous referee for suggesting this robustness check.

(13) The effect of financial optimism remains if the omitted category is financial pessimism.

(14) We are grateful to an anonymous referee for suggesting this approach.

(15) A ratio equal to unity implies that the financial expectations of the household are equal to the yearly average. If the ratio is greater than unity, financial expectations of the household are higher than the mean; conversely, a ratio less than unity implies that household financial expectations are lower than the mean.

Sarah Brown, Department of Economics, University of Sheffield, 9 Mappin Street, Sheffield S1 4DT, United Kingdom.

Gaia Garino, Department of Economics, University of Leicester, University Road, Leicester, Leicestershire LE1 7RH, United Kingdom; E-mail [email protected]; corresponding author.

Karl Taylor, Department of Economics, University of Sheffield, 9 Mappin Street, Sheffield S1 4DT, United Kingdom.

We are grateful to the Data Archive at the University of Essex for supplying data from the British Household Panel Surveys 1993 to 2000. We would like to thank two anonymous referees for constructive comments. We are also grateful to Professor Gianni De Fraja and Professor Stephen Pudney for helpful comments and advice as well as to the seminar participants at the University of Birmingham. The normal disclaimer applies.

Table 1A. Financial Expectations Over Time

 1993 1994 1995 1996 1997

Better off (%) 26.53 28.14 29.70 31.38 31.04
Worse off (%) 22.36 19.36 16.17 14.23 13.21
Or about the same? (%) 51.11 52.50 54.13 54.39 55.75

 1998 1999 2000 2001

Better off (%) 32.65 31.70 32.49 29.41
Worse off (%) 11.77 12.37 11.83 10.57
Or about the same? (%) 55.58 55.93 55.68 60.02

Table 1B. Persistence of Financial Expectations Over Time

 About the Same
 Worse Off (%) (%) Better Off (%)

1 years/9 years 62.24 (2645) 35.69 (9121) 46.27 (6054)
2 years/9 years 21.53 (915) 21.51 (5499) 23.60 (3088)
3 years/9 years 8.85 (376) 14.82 (3788) 13.21 (1729)
4 years/9 years 4.05 (172) 10.11 (2583) 7.67 (1003)
5 years/9 years 1.91 (81) 7.25 (1852) 4.62 (605)
6 years/9 years 0.89 (38) 4.89 (1250) 2.63 (344)
7 years/9 years 0.42 (18) 3.27 (837) 1.31 (171)
8 years/9 years 0.09 (4) 1.74 (444) 0.51 (67)
9 years/9 years 0.02 (1) 0.72 (185) 0.18 (24)

Italic numbers in parentheses are the number of times
household heads hold a particular type of expectation.

Note: Italic numbers in parentheses are the number of
times household heads hold a particular type of expectation.

Table 1C. Housing Tenure Over Time

 1993 1994 1995 1996 1997

Owned outright (%) 23.03 22.04 19.84 20.47 18.26
Owned mortgage (%) 47.45 48.97 50.94 49.62 46.63
Rented council (%) 10.81 11.05 11.91 12.33 12.57
Rented private (%) 18.41 17.94 17.32 17.58 22.54

 1998 1999 2000 2001

Owned outright (%) 19.67 23.01 22.84 24.01
Owned mortgage (%) 46.73 43.75 44.79 44.40
Rented council (%) 11.89 10.80 10.79 10.98
Rented private (%) 21.70 22.43 21.58 20.61

Table 2. Summary Statistics

 Full Sample
 1993-2001

 Mean SD Max Min

LMORT 4.714 4.977 14.876 -0.423
PMORT 0.207 0.456 46.154 0
FEI 1.206 0.601 2 0
HFEI 1.505 0.879 4 0
Age 42.186 12.186 65 18
[Age.sup.2] 1928.171 1050.790 4225 324
Male 0.714 0.452 1 0
White 0.919 0.273 1 0
Married 0.545 0.498 1 0
Cohabiting 0.119 0.325 1 0
No. of children 0.721 1.049 8 0
Household size 1.972 0.835 9 1
Employed 0.682 0.466 1 0
Self-employed 0.112 0.316 1 0
Unemployed 0.052 0.222 1 0
Not in labor market 0.028 0.164 1 0
Student 0.056 0.231 1 0
L(savings+investments) 1.431 2.030 8.722 -0.550
L(household income) 6.896 0.959 10.709 -3.017
Degree 0.154 0.361 1 0
Further education 0.230 0.421 1 0
A Level 0.117 0.321 1 0
GCSE (grades [greater
 than or equal to] C) 0.176 0.381 1 0
GCSE (grades < C) 0.036 0.187 1 0

Other education 0.045 0.208 1 0
Mortgage type
 Endowment mortgage -- -- -- --
 Repayment mortgage -- -- -- --
Type of insurance
 Mortgage protection
 plan -- -- -- --
 Structural insurance -- -- -- --
 Contents insurance -- -- -- --
 Other insurance -- -- -- --
Reason for extra
 mortgage payments
 Building extension -- -- -- --
 Home improvements -- -- -- --
 Car purchase -- -- -- --
 Other reason -- -- -- --
 42,894

 Mortgagees Only
 1993-2001

 Mean SD Max Min

LMORT 9.889 0.884 14.876 -0.423
PMORT 0.574 0.586 46.154 0
FEI 1.244 0.608 2 0
HFEI 1.592 0.909 4 0
Age 40.608 9.993 65 18
[Age.sup.2] 1748.890 844.259 4225 324
Male 0.825 0.380 1 0
White 0.928 0.259 1 0
Married 0.660 0.474 1 0
Cohabiting 0.129 0.335 1 0
No. of children 0.797 1.025 7 0
Household size 2.079 0.768 8 1
Employed 0.800 0.400 1 0
Self-employed 0.129 0.335 1 0
Unemployed 0.023 0.150 1 0
Not in labor market 0.010 0.097 1 0
Student 0.015 0.123 1 0
L(savings+investments) 1.824 2.149 8.722 -0.550
L(household income) 7.229 0.722 10.709 -2.852
Degree 0.196 0.397 1 0
Further education 0.280 0.449 1 0
A Level 0.130 0.337 1 0
GCSE (grades [greater
 than or equal to] C) 0.176 0.381 1 0
GCSE (grades < C) 0.037 0.189 1 0

Other education 0.034 0.180 1 0
Mortgage type
 Endowment mortgage 0.424 0.494 1 0
 Repayment mortgage 0.189 0.391 1 0
Type of insurance
 Mortgage protection
 plan 0.313 0.463 1 0
 Structural insurance 0.402 0.490 1 0
 Contents insurance 0.208 0.406 1 0
 Other insurance 0.034 0.180 1 0
Reason for extra
 mortgage payments
 Building extension 0.068 0.252 1 0
 Home improvements 0.156 0.362 1 0
 Car purchase 0.016 0.124 1 0
 Other reason 0.062 0.242 1 0
 19,941

For brevity, we have omitted summary statistics for year and
region. SD = standard deviation; Max = maximum; Min = minimum.

Table 3A. Housing Tenure and Head of
Household's Financial Expectations

 Owner Occupier
 (Mortgage)

 M.E. TSTAT

FEI 0.0096 (2.96)
Age 0.0477 (22.02)
[Age.sup.2] -0.0007 (25.53)
Male 0.1170 (14.91)
White 0.0694 (6.72)
Married 0.1593 (17.68)
Cohabiting 0.0844 (7.85)
No. of children -0.0199 (5.97)
Household size -0.0386 (8.86)
Employee 0.0567 (3.51)
Self-employed 0.0595 (3.12)
Unemployed -0.1308 (5.93)
Not in labor market -0.1237 (4.24)
Student -0.1844 (8.37)
L(savings+investments) 0.0128 (8.68)
L(household income) 0.1351 (25.51)
Degree 0.1489 (13.45)
Further education 0.1535 (18.09)
A Level 0.1388 (13.45)
GCSE (grades [greater
 than or equal to] C) 0.1106 (12.03)
GCSE (grades < C) 0.0983 (5.82)
Other education 0.0871 (6.02)
Observations 42,894
[chi square] (120) 25,689.65 p = [0.000]
Pseudo [R.sup.2] 0.2378

 Rent Rent
 (Council) (Private)

 M.E. TSTAT M.E. TSTAT

FEI 0.0121 (4.58) -0.0019 (0.55)
Age -0.0239 (21.92) -0.0219 (16.55)
[Age.sup.2] 0.0002 (17.55) 0.0003 (16.53)
Male 0.0103 (2.80) -0.0936 (15.85)
White -0.0072 (1.23) 0.0329 (5.24)
Married -0.1029 (18.71) -0.0787 (12.66)
Cohabiting -0.0376 (10.19) -0.0252 (3.91)
No. of children -0.0153 (7.28) 0.0498 (23.34)
Household size 0.0114 (4.80) 0.0174 (5.73)
Employee 0.0350 (3.73) 0.0471 (5.46)
Self-employed 0.0683 (4.13) -0.0828 (10.08)
Unemployed 0.0748 (4.05) 0.1752 (9.21)
Not in labor market 0.2187 (7.24) -0.0174 (1.15)
Student 0.0713 (3.81) 0.1575 (8.67)
L(savings+investments) -0.0018 (2.09) -0.0203 (16.85)
L(household income) -0.0352 (16.82) -0.0810 (28.30)
Degree 0.0117 (2.06) -0.1864 (53.97)
Further education -0.0150 (3.21) -0.1364 (35.14)
A Level -0.0293 (6.03) -0.1262 (33.29)
GCSE (grades [greater
 than or equal to] C) -0.0300 (6.58) -0.0898 (22.35)
GCSE (grades < C) -0.0313 (4.75) -0.0629 (9.28)
Other education -0.0255 (3.41) -0.0625 (10.42)
Observations
[chi square] (120)
Pseudo [R.sup.2]

(i) Other controls: year and region dummies in each panel;
(ii) M.E. denotes marginal effect and TSTAT denotes the
t-statistic; (iii) The base category is owned outright.

Table 3B. Housing Tenure and Household Financial Expectations

 Owner Occupier
 (Mortgage)

 M.E. TSTAT

HFEI 0.0037 (1.13)
Age 0.0476 (21.98)
[Age.sup.2] -0.0007 (25.51)
Male 0.1167 (14.87)
White 0.0696 (6.73)
Married 0.1591 (17.62)
Cohabiting 0.0846 (7.86)
No. of children -0.0199 (6.00)
Household size -0.0395 (8.96)
Employee 0.0569 (3.52)
Self-employed 0.0599 (3.14)
Unemployed -0.1302 (5.90)
Not in labor market -0.1230 (4.22)
Student -0.1844 (8.38)
L(savings+investments) 0.0128 (8.67)
L(household income) 0.1353 (25.54)
Degree 0.1491 (15.05)
Further education 0.1538 (18.13)
A Level 0.1389 (13.46)
GCSE (grades [greater
 than or equal to] C) 0.1108 (12.06)
GCSE (grades < C) 0.0985 (5.83)
Other education 0.0872 (6.02)
Observations 42,894
[chi square] (120) 25,680.22 p = [0.000]
Pseudo [R.sup.2] 0.2377

 Rent (Council) Rent (Private)

 M.E. TSTAT M.E. TSTAT

HFEI 0.0076 (3.96) 0.0011 (0.43)
Age -0.0240 (22.02) -0.0219 (16.34)
[Age.sup.2] 0.0002 (17.67) 0.0003 (16.38)
Male 0.0102 (2.78) -0.0937 (15.74)
White -0.0072 (1.24) 0.0327 (5.21)
Married -0.1046 (18.89) -0.0787 (12.58)
Cohabiting -0.0381 (10.34) -0.0256 (3.98)
No. of children -0.0154 (7.32) 0.0498 (22.93)
Household size 0.0099 (4.12) 0.0172 (5.58)
Employee 0.0350 (3.74) 0.0470 (5.45)
Self-employed 0.0689 (4.16) -0.0831 (10.06)
Unemployed 0.0759 (4.09) 0.1743 (9.17)
Not in labor market 0.2160 (7.19) -0.0161 (1.06)
Student 0.0711 (3.81) 0.1578 (8.67)
L(savings+investments) -0.0018 (2.11) -0.0203 (16.66)
L(household income) -0.0352 (16.80) -0.0811 (27.40)
Degree 0.0122 (2.14) -0.1865 (47.85)
Further education -0.0147 (3.14) -0.1366 (33.52
A Level -0.0292 (6.02) -0.1262 (31.71)
GCSE (grades [greater
 than or equal to] C) -0.0298 (6.53) -0.0900 (21.92)
GCSE (grades < C) -0.0311 (4.71) -0.0630 (9.28)
Other education -0.0254 (3.39) -0.0626 (10.39)
Observations
[chi square] (120)
Pseudo [R.sup.2]

(i) Other controls: year and region dummies; (ii) M.E. denotes
marginal effect and TSTAT denotes the t-statistic; (iii) The
base category is owned outright.

Table 4A. Mortgage Debt and Individual Financial
Expectations (Sample: [htc.sub.ht] = 1)

 LMORT PMORT

FEI 0.0245 (3.85)# 0.0239# (4.40)
Age 0.0832 (12.91)# -0.0082# (2.14)
[Age.sup.2] -0.0014 (17.08)# -0.0001# (2.46)
Male 0.1014 (3.84)# 0.0366# (3.01)
White -0.0226 (0.58)# 0.0189# (1.18)
Married 0.0654 (3.25)# -0.0059# (0.46)
Cohabiting 0.0754 (3.64)# 0.0620# (4.39)
No. of children 0.0022 (0.32)# -0.0153# (3.71)
Household size -0.0176 (2.08)# -0.0343# (5.96)
L(savings+investments) -0.0021 (0.99)# -0.0069# (4.19)
L(household income) 0.1017 (10.27)# 0.0471# (6.75)
Degree 0.3102 (11.21)# 0.0046# (0.33)
Further education 0.1379 (5.92)# 0.0104# (0.81)
A Level 0.1091 (3.94)# -0.0106# (0.71)
GCSE (grades [greater
 than or equal to] C) 0.0767 (2.94)# -0.0062# (0.45)
GCSE (grades < C) 0.1318 (2.90)# 0.0150# (0.65)
Other education 0.0790 (1.67)# 0.0112# (0.48)
Endowment mortgage 0.0613 (4.96)# 0.0424# (4.92)
Repayment mortgage 0.0751 (5.02)# -0.0095# (0.93)
Mortgage protection plan 0.0167 (1.85)# 0.0117# (1.57)
Structural insurance -0.0361 (3.33)# -0.0167# (1.92)
Contents insurance 0.0563 (4.19)# 0.0409# (3.88)
Other insurance 0.0030 (0.15)# 0.0056# (0.32)
Building extension -0.0311 (1.44)# -0.0276# (1.84)
Home improvements 0.0101 (-0.66)# -0.0334# (3.17)
Car purchase 0.0019 (0.05)# 0.0086# (0.30)
Other reason for extra
 mortgage 0.1822 (8.57)# 0.0614# (4.04)
South East 0.1789 (5.73)# 0.0614# (4.10)
South West 0.1105 (2.65)# 0.0647# (3.47)
East Anglia -0.0222 (0.41)# 0.0428# (1.75)
East Midlands -0.1594 (3.80)# 0.0504# (2.66)
West Midlands -0.1448 (3.42)# 0.0424# (2.25)
North West -0.1726 (4.37)# 0.0544# (3.12)
York and Humberside -0.1800 (4.20)# 0.0761# (4.10)
North East -0.3777 (7.26)# 0.0640# (3.00)
Wales -0.2037 (5.85)# 0.0674# (4.04)
Scotland -0.1202 (3.73)# 0.1249# (8.24)
1993 -0.1390 (8.50)# 0.0603# (4.50)
1994 -0.1134 (7.26)# 0.0377# (2.88)
1995 -0.1035 (6.85)# 0.0865# (6.66)
1996 -0.1103 (7.55)# 0.0499# (3.89)
1997 -0.0806 (5.73)# 0.0749# (6.04)
1998 -0.0743 (5.47)# 0.0627# (5.09)
1999 -0.0834 (6.75)# 0.0277# (2.47)
2000 -0.0216 (1.85)# 0.0119# (1.08)
Inverse mills ratio term 0.1718 (4.70)# 0.0969# (4.08)
[rho] 0.7272 0.1383
[chi square] (47) 2626.81 p = [0.000] 2601.34 p = [0.000]
Observations 19,941

Italic numbers in parentheses are t-statistics.

Note: Italic numbers in parentheses are t-statistics indicated with #.

Table 4B. Mortgage Debt and Household Financial
Expectations (Sample: [htc.sub.ht] = 1)

HFEI 0.0067 (2.61)# 0.0098# (2.70)
Age 0.0810 (12.60)# -0.0092# (2.42)
[Age.sup.2] -0.0014 (16.78)# -0.0001# (2.22)
Male 0.0989 (3.73)# 0.0356# (2.92)
White -0.0254 (0.66)# 0.0180# (1.13)
Married 0.0633 (3.14)# -0.0078# (0.61)
Cohabiting 0.0742 (3.59)# 0.0611# (4.32)
No. of children 0.0027 (0.39)# -0.0154# (3.72)
Household size -0.0177 (2.08)# -0.0358# (6.14)
L(savings+investments) -0.0021 (1.02)# -0.0069# (4.20)
L(household income) 0.0973 (9.88)# 0.0456# (6.54)
Degree 0.3118 (11.25)# 0.0057# (0.41)
Further education 0.1379 (5.91)# 0.0110# (0.86)
A Level 0.1090 (3.93)# -0.0105# (0.70)
GCSE (grades [greater
 than or equal to] C) 0.0769 (2.94)# -0.0061# (0.44)
GCSE (grades < C) 0.1307 (2.87)# 0.0149# (0.64)
Other education 0.0797 (1.68)# 0.0112# (0.48)
Endowment mortgage 0.0610 (4.93)# 0.0416# (4.82)
Repayment mortgage 0.0763 (5.11)# -0.0095# (0.92)
Mortgage protection plan 0.0171 (1.89)# 0.0118# (1.59)
Structural insurance -0.0362 (3.34)# -0.0168# (1.93)
Contents insurance 0.0565 (4.21)# 0.0411# (3.90)
Other insurance 0.0031 (0.16)# 0.0055# (0.31)
Building extension 0.0308 (1.43)# -0.0280# (1.87)
Home improvements -0.0101 (0.66)# -0.0335# (3.17)
Car purchase 0.0024 (0.06)# 0.0091# (0.32)
Other reason for extra
 mortgage 0.1823 (8.58)# 0.0619# (4.07)
South East 0.1768 (5.65)# 0.0611# (4.07)
South West 0.1113 (2.67)# 0.0650# (3.48)
East Anglia -0.0227 (0.42)# 0.0431# (1.76)
East Midlands -0.1587 (3.78)# 0.0506# (2.66)
West Midlands -0.1452 (3.42)# 0.0426# (2.26)
North West -0.1742 (4.40)# 0.0540# (3.09)
York and Humberside -0.1810 (4.22)# 0.0756# (4.07)
North East -0.3798 (7.28)# 0.0633# (2.96)
Wales -0.2017 (5.78)# 0.0678# (4.06)
Scotland -0.1206 (3.73)# 0.1252# (8.25)
1993 -0.1413 (8.64)# 0.0586# (4.37)
1994 -0.1150 (7.37)# 0.0365# (2.78)
1995 -0.1054 (6.98)# 0.0853# (6.58)
1996 -0.1113 (7.62)# 0.0494# (3.85)
1997 -0.0818 (5.83)# 0.0745# (6.01)
1998 -0.0749 (5.52)# 0.0625# (5.07)
1999 -0.0834 (6.75)# 0.0278# (2.48)
2000 -0.0217 (1.86)# 0.0120# (1.10)
Inverse mills ratio term 0.1554 (4.23)# 0.0902# (3.81)
[rho] 0.7291 0.1397
[chi square] (47) 2600.87 p = [0.000] 2581.41 p = [0.000]
Observations 19,941

Italic numbers in parentheses are t-statistics.

Note: Italic numbers in parentheses are t-statistics indicated with #.

Table 5. Mortgage Debt and Financial
Expectations (Sample: [htc.sub.ht] = 1)

Panel A: Individual LMORT PMORT

Whether optimistic 0.0212 (2.47) 0.0503 (5.67)
Whether pessimistic -0.0314 (-2.50) -0.0342 (2.45)
Inverse mills ratio term 0.1150 (3.35) 0.0439 (2.83)
[rho] 0.7376 0.1301
[chi square] (48) 2180.26 p = [0.000] 2181.05 p = [0.000]
Observations 19,941

Panel B: Household
Whether 1 person optimistic 0.0221 (2.63) 0.0430 (4.83)
Whether [greater than or
 equal to] 2 people
 optimistic 0.0105 (0.56) 0.0815 (3.87)
Whether 1 person
 pessimistic 0.0120 (1.24) -0.0277 (-2.55)
Whether [greater than or
 equal to] 2 people
 pessimistic -0.0145 (-0.91) -0.0662 (-3.71)
Inverse mills ratio term 0.1205 (3.51) 0.0453 (1.89)
[rho] 0.7375 0.1299
[chi square] (50) 2179.99 p = [0.000] 2184.83 p = [0.000]
Observations 19,941

(i) Other controls as in Table 4;
(ii) italic numbers in parentheses are t-statistics.

Table 6. Mortgage Debt and Financial
Expectations--Tobit Model (Sample: All)

Panel A: Individual LMORT PMORT

FEI 0.1384 (3.24)# 0.0186 (3.20)#
[rho] 0.5093 (7.81)# 0.3766 (6.77)#
[chi square] (47) 22,315.46 p = [0.000] 12,075.80 p = [0.000]
Observations 42,894
Uncensored
 (i.e., mortgagees) 19,941
Left censored 22,953

Panel B: Household
HFEI 0.0368 (2.28)# 0.0031 (2.89)#
[rho] 0.5094 (3.88)# 0.3769 (6.81)#
[chi square] (47) 22,317.75 p = [0.000] 12,062.59 p = [0.000]
Observations 42,894
Uncensored
 (i.e., mortgagees) 19,941
Left censored 22,953

Controls as in Table 4, excluding the inverse mills ratio
term. Italic numbers in parentheses are t-statistics.

Note: Italic numbers in parentheses are t-statistics
indicated with #.

Table 7. Mortgage Debt and Relative Financial
Expectations (Sample: [htC.sub.ht] = 1)

Panel A: Individual LMORT PMORT

[FEI.sub.rt] 0.0296 (3.82)# 0.0310 (3.81)#
Inverse mills ratio term 0.1627 (4.51)# 0.0721 (2.71)#
[rho] 0.7274 0.1385
[chi square] (47) 2623.36 p = [0.000] 2268.87 p = [0.000]
Observations 42,894

Panel B: Household
[HFEI.sub.rt] 0.0122 (2.83)# 0.0162 (2.24)#
Inverse mills ratio term 0.1763 (4.82)# 0.0893 (3.34)#
[rho] 0.7291 0.1763
[chi square] (47) 2610.09 p = [0.000] 2263.32 p = [0.000]
Observations 42,894

Controls as in Table 4. Italic numbers in parentheses are t-statistics.

Note: Italic numbers in parentheses are t-statistics indicated with #.