Motivations for Equity Borrowing: A Welfare-switching Effect

Abstract

During the early 2000s, mortgage market innovation together with home price appreciation increased the scope for mortgage equity withdrawal. From a macroeconomic perspective, this proved to be an important transmission mechanism for the wealth (particularly collateral) effects of housing. Microeconomic accounts of equity borrowing are less well developed, since standard models of savings and consumption rarely take housing wealth into account. This paper, however, builds on a small but growing literature assigning a precautionary savings role to consumption from housing wealth. The analysis uses panel data sourced from Britain and Australia to model households’ motivations for equity borrowing. Key among these motivations are pressing, uninsurable, ostensibly short-term, spending needs. In these contexts, it is proposed that equity borrowing assumes a welfare-switching role, substituting privately owned housing wealth for collectively funded safety-nets

1. Introduction

In the past quarter-century, an increasingly globalised economy together with progressively deregulated labour markets, exposed home-buyers to increased financial risk at a time when state safety-nets were thinning out. In parallel, a mortgage finance ‘revolution’ (Green and Wachter, 2010) reduced borrowing costs, changed the character of mortgage contracts and rendered housing wealth more fungible. These developments prompted renewed interest amongst economists in the link between house prices, housing wealth and consumption (Smith et al., 2010). Attention turned, in particular, to the growing salience of the collateral channel for the flow of wealth from housing into the wider economy (Muellbauer and Murphy, 2008). This highlights the significance for 21st-century mortgage markets of equity borrowing: the use of secured loans by existing homeowners to raise funds for discretionary expenditure (Smith, 2012a). In the US, for example, Greenspan and Kennedy (2007) estimate that discretionary extraction of home equity accounted for an astonishing 80 per cent of the rise in home mortgage debt between 1990 and the mid 2000s. Using two panel datasets, the Household Income and Labour Dynamics in Australia Survey and the British Household Panel Survey, we ourselves estimate that home equity-based borrowing by homeowners released £381 billion (in the UK) and A$373 billion (in Australia) between 2001 and 2008. This is more than three times the amount extracted by trading on within owner-occupation (£101 billion and A$69 billion respectively) or by selling up (£81 billion and A$97 billion) over the same period (Ong et al., forthcoming). Arguably, the collateral channel has become the most important of the direct flows of wealth out of housing. As Benjamin and Chinloy (2008, p. 21) point out, it is increasingly clear that “the home owning family’s equity is a piggy bank that can be broken by borrowing”.

In the UK, a series of qualitative studies has suggested that the growing fungibility of housing wealth, through the accumulation of mortgage debt, has a bearing on how owner-occupiers regard their home assets (Searle and Smith, 2010; Smith, 2008). Traditionally, such wealth represented a store of value to be unlocked, if at all, in the later stages, or at the end, of the life course. In an equity borrowing environment, however, home buyers are aware that housing assets can be added into day-to-day decisions around savings, spending and debt. In this context, housing wealth can be actively positioned as a financial buffer, replacing a sense of security that might once have been rooted in the institutions of the welfare state (Searle, 2011; Smith, Searle and Cook, 2007, 2009), further widening the financial divide between owners and renters. Similar findings occur across Europe, even though, on the continent, equity borrowing tends to be less prominent than outright housing assets in households’ financial calculations (Doling and Ronald, 2010). Together, however, these qualitative studies tap into a trend towards self-provisioning in which individuals accept greater responsibility for their own welfare by investing in financial products and property assets. In practice, however, for the majority of home-buying households, the only investment of real significance is their home. So when assets are subsequently tapped into—to meet needs prompted by sudden loss of income, or shocks such as ill health and injury—housing wealth is the obvious safety-net. By encouraging this, governments may be able to fill gaps left by the retreat of welfare states, positioning housing at the centre of a de facto asset-based approach to welfare (Lowe et al., 2011). The downside has become obvious in recent years as research, including our own, exposed the extent to which indebted households were exposed to uninsured or unmanageable investment, as well as credit, risks (Parkinson et al., 2009; Smith et al., 2009). The boom and bust cycles of many European and North American housing markets called in those risks (Searle, 2012; Smith, 2012b), and the US was particularly hard-hit. Here mortgage foreclosures left millions bankrupt or dispossessed, often with minority ethnic groups suffering, proportionately, the most (Wyly, 2010).

The use of equity borrowing to turn housing wealth into a financial safety-net has resonance with a small economics literature on the financial buffering or protection built into the structure of owner-occupied housing (which rolls an investment vehicle together with a bundle of housing services). This has grown out of a larger body of work on the links between savings and consumption. It builds in particular on foundations laid by Skinner (1996) in his discussion of housing wealth as a store of precautionary savings. His argument is that liquid assets (including newly fungible housing assets) are accumulated as a buffer against future income uncertainties or lumpy expenditures. Should a sudden dip in earnings or an acute spending need eventuate, owner-occupied housing may function as a form of self-insurance.

There is relevant supporting evidence from a small number of econometric studies conducted in the UK and US (Carroll et al., 2003; Hurst and Stafford, 2004; Benito, 2007 and 2009). Despite the institutional differences between UK and US mortgage markets, the points of convergence across these studies, which confirm that, where mortgage finance is used to extract equity, housing assets act as a financial buffer, are striking. These findings may be important, especially as accumulating debt is such a risky way of funding pressing needs. They should, therefore, be charted more systematically across jurisdictions whose housing systems are dominated by owner-occupation. This is one aim of the present analysis. It focuses on the UK and Australia which both, by the dawn of the 21st century, had high rates of homeownership and relatively ‘complete’ mortgage markets (see Girouard, 2010). In the early 2000s, Australian and UK homeowners had a wide range of mortgage products to choose from, could enter and exit contracts easily and, as mortgagors, generally had ample opportunity to withdraw funds from, as well as inject equity into, their accounts (Smith and Searle, 2008). The analysis investigates whether equity borrowing (using secured loans to raise cash for a variety of spending needs) functions as a financial buffer which is used by households in financial distress to smooth incomes and consumption in response to economic shocks whose adverse effects are otherwise unmitigated. The modelling exercise weighs what we refer to as a ‘welfare-switching role’ against demands on equity borrowing for debt consolidation and portfolio diversification.

The paper proceeds as follows. We begin with a methods section, setting out the data sources, sample design and variable construction. Then the model is specified and estimated. The results are listed and there is a short conclusion.

2. Method

2.1 Data Sources and Sample Design

Closely comparable owner-occupier samples were constructed from two nationally representative panel surveys; the Australian Household Income and Labour Dynamics in Australia (HILDA) Survey and the British Household Panel Survey (BHPS). BHPS and HILDA record a wide range of socioeconomic and demographic information for the same individuals each year (see Lynn, 2006, and Watson, 2009, for details). BHPS commenced in 1991; HILDA in 2001. We concentrate on years common to both surveys (2001–05); a common time-frame is essential for comparison under ceteris paribus conditions and the selected run of years is apposite for its place in a housing cycle that is uniquely synchronised internationally (Renaud and Kim, 2007), coincident with mortgage innovation in both countries and characterised by a period of cheap credit, in which equity borrowing was used to draw from housing wealth to an unprecedented extent. The relevant descriptives are set out in Parkinson et al. (2009). A key objective here is to identify and model cross-nationally the precipitants of equity borrowing. The econometric analysis is based on matched datasets devised to ensure that common variables are constructed in exactly the same way.

The sample used for estimation includes both mortgagors and outright homeowners, recognising that the latter can extract equity by securing a new mortgage against their home. Nevertheless, mortgagors account for 76 per cent of Australian and 92 per cent of UK equity borrowing episodes in this study.¹ BHPS and HILDA are panels based on annual surveys, so the incidence of equity borrowing is measured on an annual accounting basis, based on increases in outstanding mortgage debt among in situ homeowners between adjacent waves covering the period 2001 to 2005. The sample design thus focuses attention on predictors of equity borrowing that are not complicated by residential relocation. Those reducing outstanding mortgage debt across adjacent waves or with unchanged outstanding mortgage debt are equity savers.

Each homeowner has a maximum of four annual episodes of ownership in which equity borrowing can occur.² There are a total of 25 026 annual episodes from 8329 Australian homeowners and 27 731 annual episodes from 7289 British homeowners. Each annual episode is tagged to owners’ personal characteristics as measured in the concurrent wave of HILDA or BHPS.

Equity borrowing is frequent: it occurs in about one in five episodes in both countries and this does not vary greatly from year to year—in the Australian data, for example, the equity borrowing propensities range from a low of 19.1 per cent in 2002 to a high of 23.7 per cent in 2003; they range from 18.9 per cent in 2002 to 20.3 per cent in 2003 in the UK.³ Serial or repeat equity borrowing is also common: one-third of all Australian episodes of equity borrowing are repeat episodes, but their share is even higher at just over one-half of all episodes in the UK. The analysis considers what might motivate or precipitate such behaviour.

2.2 Variable Construction and Sample Descriptives

The panel surveys’ detailed socioeconomic and demographic information is exploited to construct measures of income, employment and life cycle transitions commonly associated with pressing spending needs and adverse financial shocks—circumstances that might prompt homeowners to draw on housing wealth. These are the key variables as far as the welfare-switching model of housing wealth is concerned. In Table 1, variables representing events or transitions that signal challenging financial circumstances are listed first. Additions to or break-up of households, for example, are shocks that are known to cause sharp rises in expenses, or sudden drops in income. To these we add childbirth, unemployment, health conditions and transitions out of the labour force. Next follow family composition, marital status and age variables. If young homeowners release housing equity to meet the acute spending needs prompted by a growing family, the presence of children variables will be important drivers of equity borrowing. The addition of age captures patterns across the life course. Early economic models of consumption and saving predict that retired owners will release housing wealth accumulated over their working lives, with a view to maintaining pre-retirement levels of consumption (Skinner, 1996).

Table 1.

Definition and measurement of variables

Variable	Time-varying or time-invariant	Unit of measurement	Definition
Financial shocks
Childbirth	Time invariant	Person	Dummy variable indicating whether there was an increase in the number of children aged 0–4 years between consecutive years. Coded to one if in any period there was an increase, zero otherwise
New relationship	Time varying	Person	Dummy variable coded to one in the year a new relationship occurs, zero otherwise
End relationship	Time varying	Person	Dummy variable coded to one in the year that either a legal marriage or de facto relationship ends, zero otherwise
Unemployed	Time invariant	Person	Dummy variable coded to one if unemployed during any period, zero otherwise
Transitions out of labour force	Time invariant	Person	Dummy variable coded to one if not in labour force during any period, zero otherwise
Ongoing health condition	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Life cycle and family status
Age	Time varying	Person	Continuous years
Married	Time varying	Person	Dummy variable; yes = 1, zero otherwise
De facto partnership	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Separated	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Divorced	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Single/widow	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Have children aged 0–4 years	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Have children age 5–14 (Aus)/5–15 (UK) years	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Have children aged 15–24 (Aus)/16–18 (UK) years	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Have children aged 25+ (Aus)/Non-dependent children (UK)	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Household employment security and income
Permanent	Time varying	Household	Dummy variable coded to one if one or more members has a permanent contract of employment, zero otherwise. Includes households with a permanent and self employed member or permanent and non-permanent-member
Non-permanent	Time varying	Household	Dummy variable coded to one if one or more members has a non-permanent contract of employment including fixed-term or casual, zero otherwise. Includes households where one member is non-permanent and the other self-employed
Self-employed	Time varying	Household	Dummy variable coded to one if one or more members are self employed with no other waged employment in the household, zero otherwise. Includes households with one member self-employed and the other either unemployed or out of the labour market
Job-seeking	Time varying	Household	Dummy variable coded to one if one or more members are unemployed and looking for work and have no waged or non-waged employment, zero otherwise
Unattached	Time varying	Household	Dummy variable coded to one if one or more household members is out of the labour market, not looking for work or engaged in waged or non waged employment, zero otherwise
Equivalent gross household income/1000	Time varying	Household	Income is equivalised by dividing annual gross income by the square root of total household size
Cluster mean gross household income/1000	Time invariant	Household	The mean equivalised household gross income. Calculated over time-frame 2002–05
Liquidity constraints and portfolio diversification
Lagged housing equity/10 000	Time varying	Household	Value of equity held in the principal residence, lagged one year
Cluster mean of housing equity/10 000	Time invariant	Household	The mean lagged value of equity held in the principal residence, as calculated over the time-frame 2002–05
Own a rental property (Aus)/income from rental property (UK)	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Difficulty raising $2000/£1000 in time of need^a	Time invariant	Person	Dummy variable; yes = 1, zero otherwise if in any period there is difficulty raising $2,000/1,000
Credit card debt/100 (Aus only)	Time invariant	Household	Total household credit card debt from 2002 wealth module.
Business debt/1000 (Aus only)	Time invariant	Household	Total household business debt from 2002 wealth module
Total household assets/10 000 (Aus only)	Time invariant	Household	Total household assets from 2002 wealth module
Non-housing debt (UK only)	Time invariant	Person	Total non-housing debt of individual from 2000 wave of BHPS
Calendar year and area controls
Year 2002	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Year 2003	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Year 2004	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Year 2005	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Live in major city (Aus only)	Time varying	Person	Dummy variable; yes = 1, zero otherwise
Area of UK	Time varying	Person	Categorical variable (individuals are allocated to one area of the UK)
Duration of residence years (Aus only)	Time varying	Person	Continuous number of years lived in property

In Australia, this is the response to a direct question; in the UK, we use a proxy based on whether households’ annual savings are less than £1000 a year.

The demographics are followed by a group of employment and income variables. The importance of individual unemployment risk is addressed in Hurst and Stafford (2004); here we augment individual risk by household unemployment measures that acknowledge the importance of threats to household employment. A gross household income measure is also included.

Next are variables proxying alternative motivations for equity borrowing. A property-biased wealth portfolio resulting from windfall gains may motivate equity borrowing as homeowners seek to rebalance their investments. A housing equity variable is therefore added,⁴ as well as an owner status variable. Mortgage finance is a cheap source of credit; equity borrowing could merely reflect debt consolidation choices as households reduce debt repayments by folding other unsecured and expensive borrowing into mortgages secured against their home. In Australia, household credit card balances and business loans can be included, but in the UK models an aggregate non-housing debt measure is used. Both Benito (2009) and Hurst and Stafford (2004) employ liquidity constraint variables on the grounds that homeowners lacking other liquid assets are more likely to fall back on housing wealth for consumption smoothing purposes. In both country models we make use of survey responses to questions concerning perceived difficulties raising £1000 ($2000) in time of need.

Finally, a vector of controls for calendar year, residential location, age and duration of residence completes our model specification. The first of these controls accounts for changes in background economic conditions. The second recognises that, while national house price indices indicate that the sample period (2001–05) was characterised by strong price growth, there were variations across regional markets.

Longitudinal data allow experimentation with alternative measures of event variables. A shock such as marital breakdown could precipitate a ‘spike’ in the probability of equity borrowing in the episode during which it occurs. A time-varying measure that switches an indicator variable equal to one in the episode it occurs (zero otherwise) captures these spikes. Alternatively, shocks can cause long-term scarring effects that permanently elevate the chances of equity borrowing. An indicator variable which takes the value 1 in episodes concurrent and subsequent to the event or transition is then appropriate, as for example in the case of relationship breakdown and divorce variables in Table 1. There is yet another possibility if the event or change of circumstances was anticipated for some time beforehand. A time-invariant measure is then required. We list in Table 1 the versions that proved most successful in terms of statistical significance.

Table 2 describes the cross-country incidence of personal and household characteristics in homeowner episodes between 2002 and 2005. Shocks such as ending or starting couple relationships, birth of children and unemployment occur infrequently in both countries. Break-up and formation of relationships are uncommon in this sample design because they are associated with residential moves. Marrieds account for around three-quarters of all episodes in each country and children are also frequently present. Among middle-aged and younger homeowners, permanent job contracts and ‘job rich’ households are believed to be increasingly important to financially sustainable homeownership, and this is reflected in the 49 per cent (63 per cent) of episodes where Australian (UK) households have one or more adults in permanent employment. With outright owners included in the sample design, retirees are also prominent, hence one-quarter of each country’s episodes belong to households where no adults are attached to the labour force.

Table 2.

Personal and household characteristics and events, 2002–05

	Total Australia episodes		Total UK episodes
	N	Percentage	N	Percentage
Financial shocks
Childbirth	605	2.5	1 374	7
New relationship	176	1	111	1
End relationship	232	1	99	1
Unemployed	278	1	748	4
Out of labour force	8 522	36	6 696	35
Ongoing health condition	6 103	26	2 743	14
Life cycle and family status
Married	17 695	75	13 743	72
De facto partnership	1 557	7	1 394	7
Separated	568	2	188	1
Divorced	1 161	5	1 015	5
Widow	1 519	6	1 744	9
Single never married	1 062	11	1 025	5
Have no children^a	3 277	14	10 719	56
Have children aged 0–4 years	3 190	14	1 758	9
Have children age 5–14 years (Aus)/5–15 (UK)	6 758	29	4 693	25
Have children aged 15–24 years (Aus)/16–18 (UK)	6 055	26	681	4
Have children aged 25+ (Aus)/non-dependent children (UK)	10 173	43	2 451	13
Household employment security
Permanent	11 435	49	12 018	63
Non-permanent	2 535	11	265	1
Self-employed	2 893	12	1 082	6
Job-seeking	193	1	140	1
Unattached	6 506	27	5 604	29
Liquidity constraints and portfolio diversification
Own a rental property (Aus)/income from rental property (UK)	3 391	14	427	2
Difficulty raising $2000/£1000	2 203	9	3 033	16
Controls
Year 2002	6 235	26	4 529	24
Year 2003	5 896	25	4 802	25
Year 2004	5 753	24	4 839	25
Year 2005	5 678	24	4 939	26
Live in major city (Aus)/Area of UK	13 022	55	19 109	100

The large difference reflects the survey questions used to elicit this information. HILDA asks if respondents have children including those aged 25 and over (regardless of their residence). In the BHPS information on children is inferred from the household type and current living arrangements. Where older children are living at home households may identify themselves as lone-parent or couple with non-dependent children. Where children have left home, parents may identify their household type as single or couple.

Notes: On deleting episodes where one or more variable measures are missing, the sample used for estimation is 23 562 (Australia) and 19 109 (UK). See Table 1 for detailed definitions.

3. Estimation Strategy and Results

A random effects logit model of equity borrowing has been estimated. Given the existence of a latent continuous dependent variable $Y_{it}^{*}$ for individual i in episode t the model’s specification is;

Y_{it}^{*} = {X'}_{it} β + u_{i} + ε_{it}, i = 1, 2, \dots . N t = 1, 2 \dots \dots . T Y_{it} = 1 if Y_{it}^{*} > 0 Y_{it} = 0 if Y_{it}^{*} \leq 0

(1)

where, $X'_{it}$ is a vector of explanatory variables; $u_{i}$ is the unobserved individual specific effect; $ε_{it} ~ N (0, σ^{2})$ and $u_{i} | X_{i t} ~ (0, ψ^{2})$ .⁵

The exogeneity assumptions are

E (u_{i} | X_{it}) = 0 = E (ε_{it} | X_{it}, u_{i})

Among the explanatory variables housing equity is lagged given likely simultaneity with equity borrowing. We have also added a correction term by including cluster means for the continuous housing equity and income variables to address endogeneity due to omitted homeowner-specific variables (Rabe-Hesketh and Skrondal, 2008, p. 115).

The model estimates are presented in Table 3 (Australia) and Table 4 (UK). Despite institutional differences, findings are sufficiently consistent across the two jurisdictions to integrate discussion of the results.

Table 3.

Modelling estimates: Australia (odds ratios) (N = 23 562; groups = 7943)

Equity borrowing	Model 1	Model 2	Model 3
Financial shocks
Birth of child	1.04 [0.078]	1.03 [0.077]	1.03 [0.077]
Relationship end	1.37 [0.24]	1.37 [0.235]	1.33 [0.231]
New relationship	1.10 [0.20]	1.17 [0.214]	1.13 [0.207]
Unemployment	1.38 [0.113]***	1.39 [0.116]***	1.36 [0.115]***
Not in labour force	0.61 [0.028]***	0.78 [0.040]***	0.80 [0.04]***
Health condition	1.08 [0.053]	1.14 [0.056]**	1.09 [0.055]
Age	0.95 [0.002]***	0.95 [0.003]***	0.95 [0.003]***
Married	1.37 [0.138]***	1.26 [0.127]*	1.37 [0.139]**
De facto partnership	1.41 [0.163]***	1.32 [0.153]**	1.32 [0.153]*
Separated	1.56 [0.243]***	1.59 [0.246]***	1.57 [0.244]***
Divorced	1.30 [0.172]*	1.30 [0.170]*	1.28 [0.168]*
Widow	0.73 [0.134]	0.80 [0.145]	0.82 [0.151]
Children aged 0–4 years	1.16 [0.072]**	1.17 [0.072]**	1.15 [0.072]*
Children age 5–14 years	1.28 [0.058]***	1.27 [0.059]***	1.27 [0.059]***
Children aged 15–24 years	1.43 [0.06]***	1.37 [0.060]***	1.36 [0.060]***
Children aged 25+	0.98 [0.060]	0.98 [0.064]	0.97 [0.063]
Job-seeking		0.50 [0.106]***	0.52 [0.110]***
Unattached		0.43 [0.037]***	0.45 [0.038]***
No permanent employment		0.73 [0.044]***	0.76 [0.045]***
Self-employed		0.87 [0.049]**	0.91 [0.054]
Equivalent gross household/1000		0.99 [0.001]	0.99 [0.001]
Cluster mean equivalent gross household income/1000		1.00 [0.001]	1.00 [0.001]**
Liquidity constraints and portfolio diversification
Lagged housing equity/10 000			1.03 [0.002]***
Cluster mean lagged housing equity/10 000			0.96 [0.002]***
Own a rental property			1.34 [0.070]***
Difficulty raising $2000			1.31 [0.062]***
Credit/$100			1.01 [0.001]***
Business debt/$1000			0.99 [0.000]
Controls added^a	No	No	Yes
Wald chi²	1698.02***	1713.78***	2017.01***
Likelihood ratio test of ρ = 0	117.55***	99.78***	79.26***

Controls have been added for calendar year, city location and years of residence at current address.

Notes: Omitted categories from the polychotomous categories are single/never married, where the household has no children present and one or more adults in the household are employed on permanent contracts. The reference person defined by the omitted categories of dichotomous and polychotomous categories is a healthy, childless, single/never married homeowner who is employed on a permanent contract. The sample of home owner episodes is 23 562 from 7943 individuals. * p <0.05; ** p <0.01; *** p <0.001. Standard errors shown in parenthesis.

Table 4.

Modelling Estimates: UK (odds ratios) (N = 19 109; groups = 6930)

	Model 1	Model 2	Model 3
Financial shocks
Birth of child	0.91 [0.082]	0.92 [0.083]	0.92 [0.081]
Relationship end	2.26 [0.556]**	2.23 [0.547]**	2.14 [0.522]**
New relationship	0.83 [0.205]	0.83 [0.206]	0.84 [0.204]
Unemployment	1.22 [0.123]*	1.23 [0.126]*	1.24 [0.125]*
Out of labour force	0.36 [0.040]***	0.50 [0.056]***	0.50 [0.056]***
Health condition	1.02 [0.077]	1.08 [0.081]	1.06 [0.080]
Life cycle and family status
Age	0.95 [0.003]***	0.95 [0.003]***	0.96 [0.003]***
Married	1.26 [0.132]*	1.18 [0.124]	1.25 [0.128]*
De facto partnership	1.25 [0.149]	1.19 [0.142]	1.22 [0.142]
Separated	1.25 [0.274]	1.29 [0.281]	1.32 [0.281]
Divorced	1.44 [0.192]**	1.40 [0.187]*	1.42 [0.185]**
Widow	0.50 [0.113]**	0.51 [0.117]**	0.53 [0.118]**
Children aged 0–4 years	0.98 [0.079]	1.02 [0.082]	1.05 [0.083]
Children age 5–15years	1.41 [0.073]***	1.42 [0.075]***	1.47 [0.077]***
Children aged 16–18 years	1.31 [0.126]**	1.29 [0.124]**	1.37 [0.131]**
Non-dependent children	1.20 [0.086]*	1.11 [0.080]	1.13 [0.081]
Job-seeking		0.92 [0.243]	0.92 [0.244]
Unattached		0.50 [0.061]***	0.52 [0.063]***
No permanent employment		1.13 [0.200]	1.22 [0.215]
Self-employed		0.71 [0.066]***	0.77 [0.073]**
Equivalent gross household income/1000		1.01 [0.003]**	1.01 [0.003]**
Cluster mean equivalent gross household income/1000		0.99 [0.004]	1.00 [0.004]
Liquidity constraints and portfolio diversification
Lagged housing equity/10 000			1.01 [0.003]**
Cluster mean lagged housing equity/10 000			0.97 [0.005]***
Income from rental property			1.95 [0.258]***
Difficulty raising £1000			1.08 [0.062]
Non-housing debt/1000			1.02 [0.005]***
Controls added	No	No	Yes
Wald chi²	1419.84***	1410.35***	1525.83***
Likelihood ratio test of ρ = 0	41.40***	36.47***	18.60 ***

Notes: See notes to Table 3.

3.1 Model 1

Model 1 estimates effects on the likelihood of equity borrowing of a variety of biographical shifts as well as life cycle and family status variables. The biographical shift variables capture equity borrowing effects prompted by discrete events or transitions occurring during the study time-frame. The status variables capture more enduring or permanent impacts on equity borrowing. Three effects merit attention.

First, there is a clear indication in both countries’ estimates that equity borrowing declines steeply with age (see also, Benito, 2007). Odds ratio estimates are identical in both countries. In Australia, for example, the predicted probability of equity borrowing at the mean values of other variables declines from 44 per cent at 25 years through 29 per cent at 45 years to 5 per cent at 65 years. Curiously, those who have had less time to accumulate housing equity (younger people) are more likely to engage in equity borrowing. This could reflect a variety of factors. Equity borrowing may be the only option for younger people, who hold most of their wealth in housing, yet have pressing spending needs to meet. It is, on the other hand, surprising that older people (who have generally spent a lifetime accumulating home equity) are not themselves more inclined to use equity borrowing to meet spending needs or support declining incomes. It is possible, of course, that older homeowners preferentially use other channels of equity borrowing such as trading down or selling up. Alternatively the reluctance of older owners might be explained by mental accounting processes that are specific to older cohorts. The idea here is that households maintain ‘mental accounts’ that enable them to control spending impulses by accumulating savings in non-fungible assets that are not easily spent, or which are only used in emergencies when households are in financial distress. This might explain Skinner’s (1996) observation that windfall gains in housing wealth are generally not spent when households are young and are only spent by retired households in financial stress. The evidence here, however, is that there could also be a cohort effect: today’s older people may hold traditional views about bequesting housing wealth, whereas future cohorts could be more inclined to spend down their housing wealth, not least because of the familiarity they gain in middle age with mortgage instruments that facilitate equity borrowing.

Secondly, we turn to household characteristics. Again, the models are alike in that the direction of effects is uniform. In Australia, all marital statuses, apart from widowhood, are strongly and significantly associated with an elevated likelihood of equity borrowing compared with being single/never married. In the UK, these effects are more muted and not always statistically significant, although here, widowhood significantly depresses the likelihood of equity borrowing.

Couples’ elevated likelihoods of equity borrowing may reflect consumption economies of scale and the repayment risk-sharing enabled by two incomes. Yet the UK findings also suggest that the end of a relationship prompts a spike in equity borrowing (statistical significance is not quite achieved for Australia). Model 1 also suggests a more permanent scarring effect associated with transitions out of relationships. Being separated (in Australia) or divorced (in both countries) increases the odds of equity borrowing by between a third and a half, compared with single/never married. A key difference between these marital categories, and widowhood (which depresses the odds of equity borrowing), is that death of a spouse is often insured and implies a legacy effect on the remaining partner’s wealth. Additionally, of course, widowhood typically occurs late in life and is commonly experienced as a permanent transition whose financial impact (loss of economies of scale and risk-sharing) is enduring rather than bridgeable, whereas relationship breakdown may hold the prospect of reconciliation or repartnering.

Becoming unemployed significantly lifts the odds of equity borrowing, but transitioning out of the labour force in any wave of the panel depresses them. Bridging dips in income due to spells of unemployment that are frequently short-term is a financially feasible response. Yet transitions out of the labour force are more likely to be permanent and insurable through retirement pensions. The onset of a health condition is unexpectedly insignificant. Health conditions generate additional spending needs and may require time off work, precipitating a sudden drop in income. Yet these circumstances fail to occasion equity borrowing. This may reflect the extent to which social security provisions are favourable to health conditions, but there are other possible explanations that we discuss later when interpreting model 3 findings.

Thirdly, in both jurisdictions, households with dependent children of school age have higher odds of equity borrowing than the childless. It is notable that, with the presence of children controlled for, the event of childbirth is insignificant. In Australia, the effect of dependent children increases across the age categories with the odds of borrowing raised by over two-fifths (43 per cent) when young adults are present, a finding that might reflect parental plans to assist their children’s entry into the housing market. In the UK, the effect is strongest among the 5–15 years age group (at 40 per cent). The indication overall is that acute spending needs occasioned by the presence of school- and college-aged children gives rise to equity borrowing among younger households at a stage of the life cycle where they typically have few other liquid assets.

3.2 Model 2

Model 2 introduces a series of employment status variables together with income. First, note that the statistical significance of the unemployment event variable is unchanged. Yet the focus here is on a series of mutually exclusive and exhaustive measures of households’ employment status. We use two-category variables for two types of household in which no adult is employed. The first of these is labelled ‘job-seeking’ to signal that at least one adult is unemployed (and therefore job-seeking); others are either job-seeking or out of the labour force. The second is labelled ‘unattached’ to signal that all adults have withdrawn from the labour force. Those in households with earnings are assigned to one of three categories signalling differing degrees of employment security. There are: households with at least one adult employed on a permanent contract; households with no adult employed on a permanent contract (because employment is casual/fixed-term contract, or there is a mix of self-employed and casual/fixed-term employment); and, households primarily reliant on self-employment with no other wage income. The omitted category is households with at least one adult employed on a permanent contract.

The salient finding is that, for Australia, all included categories of household employment status show a propensity for equity borrowing that is well below that of the omitted income-secure group and these differences are statistically significant. The odds ratio estimates indicate that, as households’ attachment to labour markets and secure forms of employment strengthens, Australian homeowners are more likely to resort to equity borrowing to meet their spending needs. Yet in the UK the findings are slightly different. Here, complete detachment from the labour market and self-employment also depress the odds of equity borrowing⁶. However, the estimates for job-seeking and non-permanent categories are insignificant. That aside, the propensity to borrow increases with income in the UK, although not in Australia: for every additional £1000 of income, the estimated odds of equity borrowing increase by 1 per cent. The findings are consistent with the idea that households attached, however remotely to the labour market, can and do use equity borrowing to meet pressing spending needs associated with biographical disruption. The final fitted model 3 pulls this picture together by introducing a set of variables designed to capture a range of other explanations for equity borrowing.

3.3 Model 3

Debt consolidation has been an important driver of 21st century mortgage markets. Mortgages in both countries did (and still do) allow substantial sums to be borrowed relatively cheaply. Such sums may be secured against property, but in a deregulated environment, can be spent on anything. So owner-occupiers who have accumulated outstanding debts that are not secured on property have a strong incentive to roll these into their mortgages. Measures of non-housing debt (in the UK), and of business and credit card debt (in Australia), test for this.

Housing wealth is the centre-piece of most homeowners’ wealth portfolios. Their willingness to engage in equity borrowing may be bolstered where house price inflation has boosted the value of outstanding home equity. Yet households can have little choice if safety-nets fail and there are no other resources available. The variable ‘difficulty of raising A$2000/£1000’ checks for the availability of other resources; the lagged housing equity variable will detect price-inflation effects.

For wealthier homeowners, on the other hand, who have no urgent need to spend from their housing wealth, soaring home equity may precipitate portfolio imbalance and prompt portfolio diversification. Equity borrowing could be used for this (Schwartz et al., 2010). However, it is a costly and awkward way to achieve that end. It is unlikely that returns on many other investments would exceed the cost of equity borrowing sufficiently to justify this route. There is, however, one exception, and this is where equity borrowing is used as leverage for other property investments. In this case, equity borrowing may be cost-effective (it is generally cheaper to borrow against a primary residence than to incur the transactions costs and higher interest rates associated with loans secured on second properties). It may also bring substantial returns in a buoyant market.

Model 3 tests these propositions and adds controls. There is some support for them all, although their addition does not adversely affect the statistical significance of variables representing the welfare-switching hypothesis (with one exception—health conditions in Australia). The personal debt variables in both countries’ models point to debt consolidation motives. Yet these same findings may also signal debt-cycling; non-secured debts can be transferred into mortgages, releasing credit cards for further debt accumulation. Moreover, debt secured against homes can increase credit ratings (Bridges et al., 2006). So what appears to be debt consolidation could turn out to be ‘debt addition’ that signals consumption-smoothing motives.

Concerning liquidity constraints the evidence in Australia is strong: those who have difficulty raising at least $2000 in emergencies are almost a third more likely to engage in equity borrowing episodes than those who do not. The UK findings point in the same direction, but are not statistically significant.

House price appreciation also has a role: a rise in housing equity of $10 000 increases the odds of borrowing by 3 per cent in Australia. A smaller, positive, but again statistically significant effect is observed for the UK. There is some support for the idea that a mix of needs and confidence in the property market underpin the propensity for equity borrowing. The confidence theme is also apparent in the large positive and statistically significant effect on equity borrowing likelihoods of rental property ownership. This lends some weight to the argument that a particular style of portfolio diversification motivation can prompt equity borrowing, although it seems that Australian owners are not more inclined to equity borrow against owner-occupied homes to finance investment property despite negative gearing tax advantages.

The final fitted model indicates that the significance for equity borrowing of household employment, life cycle and marital and family status variables is unchanged by allowing for portfolio and debt consolidation motivations. The borrowing ‘spikes’ occasioned by unemployment (in Australia) and relationship dissolution (in the UK) remain. The odds of equity borrowing continue to be depressed for those households weakly attached to the labour force. Yet as the degree of detachment narrows so the chances of equity borrowing improve, underlining the possibility that equity borrowing is a strategy that people employ when they can reasonably hope to bridge the income gap.

The financial demands of school-age children remain strong across the models. In the final fitted model, having children aged 5–15 in the UK raises the odds of equity borrowing by almost 50 per cent compared with not having children. In Australia, the effects of children in all three dependent age groups are strong, consistent and significant, and the older the children, the more marked the effect.

The curiosity in the findings pertains to limiting long-term health conditions. While raising the odds of equity borrowing in all models, it is generally statistically insignificant. This is a finding that bears closer scrutiny. One complicating factor is that a very wide range of health states are included in these measures. Some will be limiting in the long term; others resolved relatively quickly. Some are progressive; others quite curable. If these things matter for equity borrowing (if some are insured whilst others are not; if some give rise to a bridgeable income gap while others signal permanent loss of labour market attachment), their effects may easily cancel one another out. The different models may capture this to an extent. In Australia, for example, the addition of income and employment controls in model 2 causes health conditions to become statistically significant: where household labour market participation permits it, health conditions do occasion equity borrowing. This positive effect persists in model 3, but loses significance, perhaps because those with on-going health conditions are also likely to be liquidity constrained and carrying high levels of non-housing debt.

4. Conclusion

This paper models the effects of variables which previous studies suggest may motivate equity borrowing. To measures of income and housing equity, we add variables capturing life cycle events that stress household budgets. The model specification allows for a broad welfare role for housing wealth, in which homeowners engage in equity borrowing to meet spending needs prompted by a range of events and transitions. We thus conceptualise equity borrowing not as an unthinking hedonistic response to rising home prices, but rather as a product of the way households now use mortgage debt to roll housing wealth into budgeting strategies to manage a wide range of spending needs.

Importantly, we include variables that allow for debt consolidation, portfolio diversification and liquidity constraint motivations. Even controlling for these, and notwithstanding use of data from two national panel surveys, sourced from countries with historically different institutional arrangements and housing policies, the findings are strikingly similar. This suggests that the analysis is tapping into themes common in societies with high rates of homeownership, complete mortgage markets and budgetary restraint in the public sector. There are four key findings.

First, equity borrowing has a role in maintaining family welfare in the presence of children. This is particularly evident in relation to children of school age. Equity borrowing thus seems more geared to supporting the needs of growing families than to income smoothing around childbirth or levering non-dependent children into homeownership.

Secondly, there are strong life-cycle effects that are independent of the presence of children, but not protective of older age. In fact, the propensity for equity borrowing declines steeply with age. This may reflect the fact that younger homeowners have an undiversified wealth portfolio, which is centred on housing. For them, equity borrowing is the least disruptive and most cost-effective way to access this resource, especially in a rising market. It may also capture a cohort effect: today’s older homeowners have little familiarity with mortgage market innovations; they retain a strong bequest motive in relation to their housing wealth. Future generations of homeowners who are experienced in the use of innovative mortgage instruments, could be more willing to dip into their housing equity in older age. So equity borrowing, perhaps combined with a new generation of equity release schemes, could enhance the role of housing wealth in the funding of retirement. This, of course, assumes that today’s younger people do not arrive at older age already mired in debt and overexposed to investment and credit risks.

Thirdly, relationship breakdowns produce a spike in equity borrowing, but in both countries there is also evidence of divorce having a more enduring effect. The spike effect could reflect the costs of buying out a partner’s share of the family home. Yet more permanent scarring may result from a loss of risk sharing and of economies of scale in consumption. This could be sufficient to motivate equity borrowing over a longer time-horizon.

Fourthly, labour market attachment is a prerequisite of equity borrowing. In both countries, households completely detached from the labour force have odds of equity borrowing around one-half those of households with at least one person in permanent employment. This is an indication of prudence that deters equity borrowing where there are no future earnings to service the additional debt. At the same time, the closer households are to the permanent labour market, the more likely they are to engage in equity borrowing, and this suggests that some see their present circumstances as a bridge towards more secure employment and incomes.

Intriguingly, some events which we might expect, on economic grounds, to prompt equity borrowing, do not. This includes the onset of limiting long-term health conditions and death of a spouse. These, however, are events which—to some degree—are routinely ‘insurable’ or insured by the state. This provides a safety-net sufficient to mitigate the need for equity borrowing. These events are also circumstances which might prompt use of alternative channels of housing equity withdrawal that do not involve taking on additional debt, but rather are a means of extracting equity outright, such as trading down.

We conclude that equity borrowing has a welfare-switching role. It enables households to regard their homes as a store of precautionary savings (as anticipated by Skinner, 1996); it enables borrowers to use secured loans as a financial buffer (as noted by Hurst and Stafford, 2004 and Benito, 2009). However, a detailed analysis of equity borrowing in two jurisdictions, suggests further that where homeownership is the norm and mortgage markets are complete, such borrowing occupies a particular financial niche. It bridges those income gaps that are opened up by life events and transitions whose financial effects are neither insurable through private means nor (any longer) underpinned by state safety-nets. Equity borrowing has, de facto, established a role for housing wealth as an asset base for welfare. It may thus be conceptualised as integral to a process of ‘welfare-switching’, away from reliance on collective resources, towards systems of self-provisioning.

Whether or not this is sustainable is of course open to question. We have already shown that, over time, the debt to income ratios of equity borrowers and those with conventional mortgage repayment profiles diverge (Parkinson et al., 2009). We also speculate that this might lead equity borrowers towards a pattern of trading down or selling up (Ong et al., forthcoming). This is consistent with findings elsewhere that the ‘refinance ratchet’ is risky for lenders and borrowers alike (Khandani et al., 2009; Mian and Sufi, 2011; Searle, 2012; Smith, 2012b). Home equity-based borrowing fuelled by home price appreciation is increasingly implicated in the risk of default when housing markets go into recession. Equity borrowing may, then, be a rather risky platform from which to advocate individualised asset-based welfare over more inclusive styles of insurance.

Footnotes

Acknowledgements

The paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey and the British Household Panel Survey (BHPS). The HILDA project was initiated and is funded by the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and is managed by the Melbourne Institute of Applied and Economic and Social Research (MIAESR). The BHPS is made available through the Economic and Social Research Council (ESRC) Data Archive. The data were originally collected by the ESRC Research Centre on Micro-social Change at the University of Essex. The findings and views reported in this paper are those of the authors and should not be attributed to FaHCSIA, MIAESR or the University of Essex.

Funding

The research for this paper was funded under the international collaboration scheme of the Australian Research Council (LX 0775767) and the UK’s Economic and Social Research Council (RES-000-22-1985).

Notes

References

Benito

(2007) Housing equity as a buffer: evidence from UK households. Working Paper No. 324, Bank of England, London.

Benito

(2009) Who withdraws housing equity and why?, Economica, 76, pp. 51–70.

Benjamin

J. D.

Chinloy

(2008) Home equity, household savings and consumption, Journal of Real Estate Finance and Economics, 37, pp. 21–32.

BHPS (British Household Panel Survey) (1991–2000) British Household Panel Survey Waves 11–16. Institute for Social & Economic Research (www.iser.essex.ac.uk/bhps).

Bridges

Disney

Henley

(2006) Housing wealth and the accumulation of financial debt: evidence from UK households, in: Bertola

Disney

Grant

(Eds) The Economics of Consumer Credit, pp. 135–180. Cambridge, MA: MIT Press.

Carroll

C. D.

Dynan

K. E.

Krane

S. D.

(2003) Unemployment risk and precautionary wealth: evidence from households’ balance sheets, The Review of Economics and Statistics, 85(3), pp. 586–604.

Doling

Ronald

(2010) Home ownership and asset-based welfare, Journal of Housing and the Built Environment, 25, pp. 165–173.

Girouard

(2010) Housing and mortgage markets: an OECD perspective, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 38–57. Chichester: Wiley-Blackwell.

Green

Wachter

(2010) The housing finance revolution, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 414–446. Chichester: Wiley-Blackwell.

10.

Greenspan

Kennedy

(2007) Sources and uses of equity extracted from homes. Finance and Economics Discussion Series 2007-20, Divisions of Research & Statistics and Monetary Affairs, Federal Reserve Board, Washington, DC.

11.

Hurst

Stafford

(2004) Home is where the equity is: mortgage refinancing and household consumption, Journal of Money, Credit and Banking, 36(6), pp. 985–1014.

12.

Khandani

A. E.

Merton

R. C.

(2009) Systematic risk and the refinancing ratchet effect. Working Paper No. WP 15362, NBER, Cambridge, MA.

13.

Lowe

Searle

B. A.

Smith

S. J.

(2011) From housing wealth to mortgage debt: the emergence of Britain’s asset-shaped welfare state, Social Policy and Society, 11(1), pp. 105–116.

14.

Lynn

(Ed.) (2006) Quality Profile: British Household Panel Survey, Vol 2. Colchester: University of Essex Press.

15.

Mian

Sufi

(2011) House prices, home equity-based borrowing, and the US household leverage crisis, American Economic Review, 101(5), pp. 2132–2156.

16.

Muellbauer

Murphy

(2008) Housing markets and the economy: the assessment. Oxford Review of Economic Policy, 24, pp. 1–33.

17.

Ong

Parkinson

Searle

B. A.

Smith

S. J.

Wood

G. A.

(forthcoming) Channels from housing wealth to consumption, Housing Studies.

18.

Parkinson

Searle

B. A.

Smith

S. J.

Stoakes

Wood

(2009) Mortgage equity withdrawal in Australia and Britain: towards a wealth-fare state?, European Journal of Housing Policy, 9(4), pp. 363–387.

19.

Rabe-Hesketh

Skrondal

(2008) Multilevel and Longitudinal Modeling Using Stata, 2nd edn. College Station, TX: Stata Press.

20.

Renaud

Kim

(2007) The global housing price boom and its aftermath, Housing Finance International, 22(2), pp. 3–15.

21.

Schwartz

Hampton

Lewis

Norman

(2010) A survey of housing equity withdrawal and injection in Australia, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 147–175. Chichester: Wiley-Blackwell.

22.

Searle

B. A.

(2011) Housing wealth and recession, Journal of Financial Economic Policy, 33(1), pp. 33–48.

23.

Searle

B. A.

(2012) Recession repossession and family welfare, Child and Family Law Quarterly, 24(1), pp. 1–23.

24.

Searle

B. A.

Smith

S. J.

(2010) Housing wealth as insurance: insights from the UK, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 339–360. Chichester: Wiley-Blackwell.

25.

Skinner

J. S.

(1996) Is housing wealth a side show?, in: Wise

(Ed.) Advances in the Economics of Aging, pp. 241–268. Chicago, IL: University of Chicago Press.

26.

Smith

S. J.

(2008) Owner occupation: living with a hybrid of money and materials, Environment and Planning A, 40, pp. 520–535.

27.

Smith

S. J.

(2012a) Crisis and innovation in the housing economy: a tale of three markets, in: Haliossis

(Ed.) Financial Innovation and Economic Crisis: An Introduction, pp. 71–102. Cambridge, MA: MIT Press.

28.

Smith

S. J.

(2012b) Care-full markets: miracle or mirage?, Tanner Lectures on Human Values, 31, pp. 245–276.

29.

Smith

S. J.

Searle

B. A.

(2008) Dematerialising money: the ebb and flow of wealth between housing and other things, Housing Studies, 23(1) pp. 21–43.

30.

Smith

S. J.

Searle

B. A.

(Eds) (2010) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations. Chichester: Wiley-Blackwell.

31.

Smith

S. J.

Searle

B. A.

Cook

(2007) Banking on housing: spending the home. End of Award Report, RES-154-25-0012, ESRC.

32.

Smith

S. J.

Searle

B. A.

Cook

(2009) Rethinking the risks of owner occupation, Journal of Social Policy, 38, pp. 83–102.

33.

Smith

S. J.

Searle

B. A.

Powells

G. D.

(2010) Introduction, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 1–28. Chichester: Wiley-Blackwell.

34.

Watson

(Ed.) (2009) HILDA user manual: release 7. Melbourne Institute of Applied Economic and Social Research, University of Melbourne.

35.

Wood

Kemp

P. A.

(2003) The taxation of Australian landlords: would the British tax treatment of rental investments increase tax burdens if introduced in Australia?, Urban Studies, 40(4), pp. 747–765.

36.

Wyly

E. K.

(2010) The subprime state of race, in: Smith

S. J.

Searle

B. A.

(Eds) The Blackwell Companion to the Economics of Housing: The Housing Wealth of Nations, pp. 381–413. Chichester: Wiley-Blackwell.